sample_number int64 8 203k | abstract stringlengths 21 17.4k | article stringlengths 0 18.9k | compressed_article stringlengths 0 10.9k | compression_time float64 0 0.78 | original_tokens int64 0 3.79k | compressed_tokens int64 0 2.29k | token_saving stringclasses 2 values |
|---|---|---|---|---|---|---|---|
20,456 | bettini _ et al . _ [ nature nanotech * 1 * , 182 ( 2006 ) ] reported the first experimental realization of linear atomic chains ( lacs ) composed of different atoms ( au and ag ) .
different contents of au and ag were observed in the chains from what found in the bulk alloys , which rises the question of what is the wire composition if in equilibrium with a bulk alloy . in this work
we address the thermodynamic driving force for species fractionation in lacs under tension , and we present density - functional theory results for ag - au chain alloys . a pronounced stabilization of wires with an alternating ag - au sequence
is observed , which could be behind the experimentally observed au enrichment in lacs from alloys of high ag content . | metal nanowires have attracted great interest with the increasing miniaturization of electronic and mechanical devices @xcite , given their very interesting properties related to quantum transport @xcite . from the experimental point to view , two techniques have been mostly used in nanowires research : mechanically controllable break junctions ( mcbj ) @xcite and _ in situ _ high - resolution transmission electron microscopy ( hrtem ) @xcite .
while mcbj is used to obtain electronic properties of nanowires , hrtem gives structural information , providing atomic resolution with real - time image acquisition .
a combination of both techniques is frequently used in order to establish a correlation between atomic arrangement and quantum conductance @xcite . since 1997
these techniques have revealed the existence of some quite unexpected structures , such as , extremely thin metal nanotubes with a square cross - section @xcite , helical multishell gold nanowires @xcite , single - walled metal nanotubes @xcite , and the ultimate nanowires , linear atomic chains ( lacs ) @xcite .
the stability of these structures has been inferred using statistics over large number of observed structures ( in hrtem ) and on conductance histograms for a large number of contact - breaking events ( in mcjb ) .
these experimental analyses have shown the remarkable existence of stable ( or metastable ) nanowires for alkali @xcite and noble metals.@xcite structural motifs for the atomic - wire compositions considered in this work .
dark ( light ) balls indicate ag ( au ) atoms .
see text for discussions . ]
there have been many theoretical studies of these structures , mostly using classical atomistic models , as well as first principles ones @xcite .
of particular importance is the work of tosatti _ et al . _
@xcite , that investigated the stability of these nanostructures considering the metal wires connected to atomic reservoirs at both ends , determining the local minima ( different metastable structures ) as a function of the tension of the wire . using this approach
the authors were able to predict the stability of magic sizes for au nanowires .
although metal nanowires have been the object of intense theoretical and experimental investigations , most of these studies were carried out for pure metals .
the first lac experimental realization using metallic alloys was reported only in 2006 @xcite .
they reported the observation of lacs of different lengths composed of ag and au atoms from ag - au nanoalloys of different compositions .
they observed that when the nanowires are rich in ag @xcite , in the final states of the stretching leading to lac formation , the observed behavior was similar to the ones for au rich nanowires @xcite . from classical molecular dynamics @xcite it was concluded that this unusual behavior is consequence of the au enrichment at the region where the lacs were formed .
the concentration of these nanoalloys changes significantly in time from the nominal initial values .
they basically behave as if they had a much high au content than in their nominal composition .
this behavior was addressed by fao @xcite by considering the enthalpy as defining the stability of these atomic chains under tension , in line with what pioneered by tosatti @xcite . a grand canonical perspective , however , which is the adequate framework to study the thermodynamical trends in these systems , has not been done until this date and will be the focus of this work .
the main objective of this paper is to determine the relative concentrations of two species ( ag and au in this case ) in lac when in equilibrium with bulk alloys of given composition . for the purpose
the relevant free energy is defined and calculated from first principles , analogously to what was done in ref .
@xcite for single - species wires and tubes , which is now generalized to a grand - canonical setting .
the observed capability of the wires to change composition with respect to the bulk reservoir demands such grand canonical treatment , beyond what was presented in ref .
@xcite . in this way , with our approach we should be able to predict for specific bulk alloy concentrations what would be the corresponding likely concentrations for lacs formed by stretching a nanowire .
it is important to remember , however , that this method takes an equilibrium viewpoint , which assumes atom exchange between reservoirs and wire , and should only be taken as a determination of equilibrium trends .
the kinetics of the formation process will very probably originate deviations from this limiting behavior .
it is however important to know what the thermodynamical trends are , which is what we address here .
density - functional theory ( dft ) calculations were carried out for free - standing infinite nanowires by preparing arrays of sufficiently separated lacs in periodic boundary conditions .
tetragonal supercells are used with the wires aligned along the @xmath0-axis , with four atoms of au or ag and binary alloys of them .
the considered geometries are presented in fig .
[ studiedgeometries ] .
the lac geometries are relaxed for fixed supercell size , keeping the @xmath1 and @xmath2 sizes such that the minimum distance among wires is 15.0 , while the @xmath0 direction was varied in increments of 0.05 , from 6.00 until the wire snapped . we have considered lac lengths between 7 and 11 . chemical potential difference @xmath3 calculated for a bulk agau alloy .
squares indicate @xmath4 values at bulk concentrations 0 , 7/32 , 13/32 , 20/32 , 26/32 and 1 ; crosses show the interpolated estimates for concentrations 0 , 0.2 , 0.4 , 0.6 , 0.8 and 1.,title="fig : " ] [ chemical ] the dft calculations were performed with the siesta method @xcite , using the perdew - burke - ernzerhof ( pbe ) generalized gradient approximation ( gga ) for exchange and correlation @xcite .
norm - conserving scalar - relativistic pseudopotentials @xcite replaced nuclei and core electrons , considering the normal core - valence separation for ag and au . double-@xmath5
polarized bases were used for valence electrons @xcite .
integrals in real space were performed on a mesh of 250 ry cutoff @xcite .
the integration over the first brillouin zone was carried out using the special @xmath6-points method @xcite with @xmath6-meshes of @xmath7 and @xmath8 for wire and bulk structures , respectively . in order to calculate the nanowire stability for an alloy we have generalized the string tension analysis to different atomic species .
the free energy @xmath9 we will seek to minimize is @xmath10 where @xmath11 and @xmath12 represent the number of au and ag atoms in the lac ( for the case of agau alloys ) , and @xmath13 and @xmath14 , are the corresponding chemical potentials .
it is the zero - temperature grand - canonical free energy ( per unit cell ) that gives the stability for a wire of unit - cell length @xmath15 ( length of cell along @xmath0 ) .
the analysis can also be done as a function of @xmath16 , the force pulling the wire , in which case the relevant free energy would be @xcite @xmath17 here we will restrict ourselves to @xmath15-dependence and thus @xmath9 .
we will consider the low temperature limit , disregarding entropy effects .
the generalization of the present scheme to finite temperature is costly but conceptually straightforward , either by including the entropy of harmonic vibrations @xcite ( for a fully quantum albeit harmonic treatment ) , or thermodynamic integration for the free energy @xcite , using molecular dynamics for classical nuclei , or with path - integral monte - carlo or molecular dynamics for quantum nuclei .
if we define @xmath18 we can write @xmath19 where @xmath20 the @xmath21 term controls the energetics related to their having more or less atoms in the chain , regardless of species , while the @xmath22 term controls the energetics related to swapping atoms of different species in a wire with a given number of atoms . we will confine ourselves here mainly to the latter behavior considering wires with a constant @xmath23 , and will consider the free energy @xmath24 the chemical potentials are defined by the bulk alloy out of which the wire forms , which constitute the effective reservoirs . in particular , the relative chemical potential can be calculated as the statistical average energy change when swapping any au atom in the alloy into a ag atom , @xmath25 this has been obtained by making random substitutions of au by ag in disordered bulk alloy samples of different relative concentrations , relaxing the structures before and after the swap , and computing the energy difference .
face - centered cubic supercell structures with 32 atoms were used for the purpose .
the original concentrations of ag in au were 0 , 7/32 , 13/32 , 20/32 , 26/32 , and 1 .
the results are shown in fig .
[ chemical ] . both for @xmath13 and @xmath14 the zero is defined for their respective pure bulk crystals.@xcite
the curve shows a quite linear behaviour except at very low ag concentrations .
free energy @xmath26 versus length @xmath15 for atomic chains of different compositions , for different bulk compositions .
each group of curves correspond to different bulk alloy composition , as specified with the percentages on the right , and contains six curves , for the wire configurations i to vi ( see fig .
[ studiedgeometries ] ) , as specified in the legend .
the curves for different bulk compositions have been shifted upwards by 85.55 ev for convenience . ]
fig . [ omegaauag ] shows the free energy @xmath26 as a function of cell length @xmath15 , for the six atomic - wire configurations shown in fig .
[ studiedgeometries ] , and for the different bulk composition of the lead alloys specified on the figure .
starting from the left , the regime up to @xmath27 corresponds to zigzag wires in which three consecutive atoms define quite an equilateral triangle , i.e. , essentially a band with two rows of atoms and edge - sharing triangles .
the regime at larger distances corresponds to the zigzag structures shown in fig .
[ studiedgeometries ] , agreeing with what known for au wires @xcite , which gradually stretch with growing @xmath15 , until they snap at @xmath15-values slightly above what displayed .
the transitions between regimes , which involve bond " breaking , are discontinuous and hysteretic , and thus will be quite dependent on quantum and thermal fluctuations.@xcite the experimentally observed lacs correspond to the intermediate regime .
[ omegaauag ] shows that for most alloy concentrations the structure iv of fig .
[ studiedgeometries ] , with ag - au alternation is most stable in the lac regime , with higher stabilization for higher ag content in the bulk alloy .
this implies that ag - rich bulk alloys tend to produce wires with [ ag ] tending towards 50% , i.e. the wires will be au enriched .
these results agree with the mixed and au enriched wires seen in the experimental atomically resolved micrographs.@xcite reducing the ag concentration in the alloy there is a transition to a situation in which the wire could change equilibrium concentration with stretching , as seen for 20% . for @xmath28
, the wire would tend to pure au , thus depleting ag ( although the configuration ii with 25% ag is close in energy , and indeed lower for mechanically unstable shorter @xmath15 ) . upon stretching ,
though , the agau alternating wire dominates again , thus replenishing ag . in summary , we have defined and calculated from first - principles the grand - canonical zero - temperature free energy governing the fractionation of two species between an alloyed atomic chain and the bulk alloy .
we have applied this to ag / au alloy chains and explained the tendency for au enrichment in lacs observed experimentally for ag rich alloys .
pasa acknowledges the funding and support of the erasmus mundus program and the hospitality of the department of earth sciences of the university of cambridge .
pasa and dsg acknowledge financial support of the brazilian agencies fapesp and cnpq .
they also wish to thank prof .
daniel ugarte for many helpful discussions .
ea acknowledges the hospitality of the institute of physics at the university of campinas .
the calculations were done using the camgrid high - throughput facility of the university of cambridge .
this is for convenience with the numbers .
the bare @xmath4 coming from @xmath29 gives large numbers referring to the separate cores and electrons .
here we use @xmath30 , the two latter terms being the dft energy per atom of bulk au and bulk ag , respectively . the same is done in the definition of @xmath31 . | metal nanowires attracted interest with miniaturization electronic mechanical devices interesting properties related quantum transport two techniques used in nanowires research : mechanically controllable break junctions mcbj high - resolution transmission electron microscopy hrtem mcbj used electronic properties hrtem gives structural information atomic resolution real - time image acquisition combination of techniques used correlation between atomic arrangement quantum conductance since 1997 techniques revealed unexpected structures thin metal nanotubes with square cross - section helical multishell gold nanowires single - walled metal nanotubes linear atomic chains stability structures inferred using statistics over observed structures conductance histograms for contact - breaking events mcjb experimental analyses shown existence of stable nanowires for alkali @xcite noble metals structural motifs for atomic - wire compositions dark balls indicate atoms many theoretical studies of structures using classical atomistic models first principles ones importance work of tosatti et al investigated stability of nanostructures considering metal wires connected to atomic reservoirs determining local minima metastable structures tension wireapproach authors stability of magic sizes for au nanowires metal nanowires object of intense experimental investigations most studies for pure metals first lac experimental realization using metallic alloys reported in 2006 of lacs of different lengths ag and au atoms from ag - au nanoalloys different compositions when nanowires rich in ag final states stretching to lac formation behavior similar to au rich nanowires concluded unusual behavior consequence of au enrichment region lacs formed concentration nanoalloys changes significantly from values behave as if high au content behavior addressed by fao considering enthalpy as defining stability of atomic chains under tension grand canonical perspective framework study thermodynamical trends not done focus of work main objective to determine relative concentrations of two species ( ag and au in lac when in equilibrium with bulk alloys composition relevant free energy defined and calculated from first principles analogously to single - species wires tubes generalized to grand - canonical setting capability of wires to change composition bulk demands grand canonical treatment approach predict specific bulk alloy concentrations for lacs formed stretching nanowire method equilibrium viewpoint assumes atom exchange between reservoirs wire of equilibrium trends kinetics formation originate deviations important to know thermodynamical trends density - functional theory calculations for free - standing infinite nanowires separated lacs in periodic boundary conditions tetragonal supercells used wires aligned along @xmath0-axis four atoms of au or ag binary alloys geometries presented in fig . lac geometries relaxed for fixed supercell size @xmath1 @xmath2 sizes minimum distance among wires 15.0 @xmath0 direction varied 0.05 considered lac lengths between 7 and 11 chemical potential difference @xmath3 calculated for bulk agau alloy squares indicate @xmath4 values at bulk concentrations 0 7 13 20 26/32 1 crosses show interpolated estimates for concentrations 0 0.2 0.4 0.6 0.8 1. calculations with siesta method generalized gradient approximation for exchange correlationconserving scalar - relativistic pseudopotentials @xcite replaced nuclei core electrons normal core - valence separation for ag au double-@xmath5 polarized bases for valence electrons integrals performed on 250 ry cutoff @xcite integration over first brillouin zone @xmath6-points method-meshes of @xmath7 @xmath8 for wire bulk structures nanowire stability for alloy generalized string tension analysis to atomic species free energy @xmath9 is @xmath10 @xmath11 represent au ag atoms in chemical potentials zero - temperature grand - canonical free energy cell stability for wire unit - cell length @xmath15 analysis function of @xmath16 force pulling wire relevant free energy @xcite to @xmath15-dependence @xmath9 low temperature limit disregarding entropy effects generalization to finite temperature costly straightforward entropy of harmonic vibrations @xcite thermodynamic integration for free energy molecular dynamics for classical nuclei path - integral monte - carlo dynamics for quantum nucleidefine @xmath18 write @xmath19 @xmath20 @xmath21 controls energetics more less atoms chain species @xmath22 controls swapping atoms different species wire confine latter behavior considering wires constant @xmath23 consider free energy @xmath24 chemical potentials defined by bulk alloy wire effective reservoirs relative chemical potential calculated average energy change swapping au atom into ag atom @xmath25 obtained random substitutions au by ag disordered bulk alloy samples different relaxing structures before after swap computing energy difference face - centered cubic supercell structures 32 atoms used original concentrations ag au 0 7/32 13/32 20/32 26/32 1 results shown fig . chemical @xmath13 @xmath14 zero defined pure bulk crystals curve linear behaviour low ag concentrations free energy @xmath26 versus length @xmath15 atomic chains different compositions bulk compositions group curves correspond different bulk alloy composition percentages six curves wire configurations i to vi curves bulk compositions shifted upwards by 85.55 ev convenience fig omegaauag shows free energy @xmath26 function cell length @xmath15 six atomic - wire configurations fig . different bulk composition lead alloys regime up to @xmath27 corresponds zigzag wires three consecutive atoms define equilateral triangle two edge - triangles regime larger distances zigzag structures fig . au wires @xcite stretch with growing @xmath15 snap at-values transitions between regimes discontinuous hysteretic dependent on quantum thermal fluctuations observed lacs correspond intermediate regime omegaauag shows alloy concentrations structure iv of fig with ag - au alternation stable in lac regime higher stabilization for higher ag content bulk alloy ag - rich bulk alloys produce wires with ag ] towards 50% au enriched results agree with mixed au enriched wires experimental atomically resolved micrographs reducing ag concentration wire change equilibrium concentration with stretching 20% for @xmath28 wire pure au ag configuration ii with 25% ag close in energy lower for mechanically unstable shorter @xmath15stretching agau alternating wire dominates replenishing ag defined calculated grand - canonical zero - temperature free energy fractionation two species between alloyed atomic chain bulk alloy applied ag / au alloy chains explained tendency au enrichment ag rich alloys pasa acknowledges funding erasmus mundus program hospitality department earth sciences university cambridge acknowledge financial support brazilian agencies fapesp cnpq prof . daniel ugarte discussions acknowledges hospitality institute physics university of campinas calculations camgrid high - throughput facility university cambridge convenience @xmath4 @xmath29 large numbers separate cores electrons use @xmath30 terms dft energy per atom bulk au bulk ag same definition @xmath31 | 0.43554 | 2,742 | 1,338 | , Saving $0.1 in GPT-4. |
20,469 | we report here on calculation of probabilities of intercombination transitions @xmath0 @xmath1 - @xmath2 @xmath3 , @xmath4 @xmath3 for be - like ions along the isoelectronic sequence for large range of @xmath5 .
our results obtained with the @xmath6 expansion method agree well with experimental data including the recent ones . | intercombination , or spin - forbidden , transitions are due to deviations from the pure ls - coupling .
since their the very existence , at least for small nuclear charges @xmath5 , is a consequence of the subtle relativistic spin - orbit interaction , the agreement between theoretical and experimental results is an important measure of our understanding of atomic structure and interactions .
in addition , intercombination lines are very important for plasma diagnostics , for example , in astrophysics and tokamak studies [ 8 ]
. since spin - forbidden transitions for not too large @xmath5 are much weaker than spin - allowed ones , the reliable determination of their transition probabilities requires serious experimental efforts .
recently two important experiments on measurement of the intercombination transitions probabilities in be - like ions were carried out .
et al _ [ 19 ] have improved their ion trap experiment on the transition @xmath7 - @xmath8 @xmath9 in c iii and obtained the value of @xmath10 @xmath11 which differs considerably from the preliminary result of the same group @xmath12 @xmath11 [ 26 ] . in another experiment granzow _
et al _ [ 14 ] have measured the probability of the @xmath7 - @xmath13 transition in na viii si xi ions .
similar measurements for smaller nuclear charges @xmath5 from n iv to ne vii [ 9,15 ] were done more than ten years ago , so it is possible now to compare theoretical predictions along larger interval of @xmath5 .
these two latest experimental works initiated a wave of theoretical calculations [ 11,13,3,28,12 ] where most sophisticated and elaborated up to now methods were used , e.g. multiconfiguration hartee - fock ( mchf ) , multiconfiguration dirac - fock ( mcdf ) , configuration interaction ( ci ) methods , multiconfiguration relativistic random - phase approximation ( mcrrpa ) . in this paper
we report the results of calculations of @xmath7 - @xmath8 and @xmath7 - @xmath13 transition probabilities along [ be ] isoelectronic sequence by the @xmath5 expansion method .
this approach is known to be very accurate in determination of energy levels of few - electron multicharged ions but as we will see in what follows gives also good results for intercombination probabilities . in these calculations we use the mz code [ 25 ] based on the perturbation theory on @xmath6 parameter .
the main principles of this approach and the code structure are well described in [ 25 ] so we refer the reader to this book for details .
both intercombination transitions studied in this paper arise from mixing of triplet terms with the singlet ones .
this mixing not only makes the spin - forbidden transition to be possible but also influences the level energies .
therefore the correspondence between the experimental and theoretical energies is an additional important check for accuracy of the calculations .
earlier the mz method was successfully applied for calculation of the level energies for configurations @xmath14 @xmath15 of na viii - s xiii ions [ 24,1 ] .
these references contain detailed examination of energy calculations by mz so here we immediately pass to the discussion of obtained results . here
we calculated energies for configurations @xmath16 and @xmath17 for @xmath18 . in most cases
the difference between calculated in this work and experimental [ 17 ] energies is of order of a few units of @xmath19 or less which is typical for calculations with the mz code . for intercombination transitions two energy differences , namely @xmath20 and @xmath21 , play a crucial role since these quantities explicitly enter the formula for transition probability in the first order of perturbation theory .
in fact , our results for these energies have the highest accuracy except of the latest variational calculations [ 29 ] .
note that experimental energies for levels @xmath2 @xmath3 @xmath22 have a stable non - depending on @xmath5 shift of @xmath23 @xmath24 comparing to our results .
since all these energies were determined from the same experiment [ 4 ] , it is quite possible that there was a systematic error in those measurements .
until now only four experiments on measurement of @xmath0 @xmath1 - @xmath2 @xmath3 transition probability were carried out , that is for c iii [ 19 ] , fe xxiii [ 5 ] , kr xxxiii [ 6 ] and xe li [ 21 ] . the best known in this list is the widely used in solar plasma diagnostics intercombination line 1909 in c iii which has been discussed in many theoretical papers ( see references in ref . [ 27 ] ) .
as we have noted above , the recently reported experimental results on c iii [ 19 ] do not overlap with the previous measurements of the same group for this transition probability [ 26 ] . the new value @xmath25 @xmath11 agrees well with our result @xmath26 @xmath27 and mcrrpa value of 118 @xmath11 [ 3 ] but contradicts the latest ci [ 11 ] , mchf [ 13 ] and mcdf [ 28 ] calculations which give @xmath28 , @xmath29 , and @xmath30 @xmath11 respectively . as one can see from fig .
1 , the agreement between mz probabilities and available data for fe , kr and xe is also good .
in addition to above mentioned ci , mcdf , mchf , and mcrrpa calculations , we show on this figure the extensive mcdf calculations for be - like ions up to xe li [ 2 ] , model potential results [ 20 ] and mc calculations with superstructure code [ 22 ] as well .
although different theoretical results give practically the same both @xmath5-dependence and absolute values of probabilities for large @xmath31 , the situation for small nuclear charges is not very clear yet and new accurate measurements for @xmath32 would be of great importance . on fig .
2 the theoretical and experimental results for transition probability @xmath33 @xmath34 @xmath35 for @xmath36 are shown .
at present there are old experimental data for @xmath37 [ 9 ] , @xmath38 [ 15 ] and latest data for @xmath39 [ 14 ] .
the theoretical results on fig .
2 include mcdf [ 12 ] , mchf [ 7 ] , ci [ 16 ] and hf - with - relativistic - corrections [ 10 ] calculations . as was shown by fritzsche and grant
[ 12 ] , for small @xmath5 the mcdf method is rather slow convergent with increase of the number of configurations included .
this is probably to be a reason for large discrepancy between that paper and results of kim _ et al _ [ 18 ] .
it is seen from fig .
2 that different theoretical approaches show again similar @xmath5-dependence , except for smallest @xmath5 , where more accurate measurements would be desirable . the new experimental results by granzow _
et al _ [ 14 ] seem to give @xmath5-dependence rather different from theoretical one , although in sufficiently narrow region . at least for na viii
the value of @xmath40 is considerably larger than most recent theoretical results .
this feature needs further investigation , especially since the authors claim that this value appears to be an overestimate .
we are grateful to drs . s.fritzsche and i.p.grant for making their results available prior to publication and to drs .
p.l.smith and e.trbert for interesting discussions .
yu.v.r . acknowledges the financial support from israeli ministries of absorption and science and hospitality of dr .
i.yu.tolstikhina during his stay in nagoya . | intercombination spin - forbidden transitions due to deviations from ls - coupling existence small nuclear charges @xmath5 consequence of relativistic spin - orbit interaction agreement between theoretical experimental results atomic structure interactions intercombination lines important for plasma diagnostics astrophysics tokamak studies spin - forbidden transitions for @xmath5 weaker than spin - allowed determination transition probabilities requires experimental efforts two experiments on intercombination transitions in be - like ions carried et al improved ion trap experiment transition @xmath7 - @xmath8 obtained value of @xmath10 @xmath11 differs from preliminary result @xmath12 @xmath11 granzow et al measured probability @xmath7 - @xmath13 transition in na viii si xi ions similar measurements for smaller nuclear charges @xmath5 to vii done ten years ago possible to compare theoretical predictions along larger interval @xmath5 works initiated wave of theoretical calculations methods used multiconfiguration hartee - fock relativistic random - phase approximationpaper report results @xmath7 - @xmath8 transition probabilities along be ] isoelectronic sequence by @xmath5 expansion method approach accurate energy levels few - electron multicharged ions good results intercombination probabilities use mz code [ 25 ] perturbation theory @xmath6 parameter principles code structure described in [ 25 ] intercombination transitions arise from mixing triplet terms with singlet ones mixing spin - forbidden transition influences level energies correspondence between experimental theoretical energies accuracy mz method applied for level energies for configurations @xmath14 @xmath15 of na viii - s xiii ions [ 24,1 ] references detailed examination energy calculations mz discussion results calculated energies for configurations @xmath16 @xmath17 @xmath18 difference between calculated experimental energies few units of @xmath19 or less typical mz code intercombination transitions energy differences @xmath20 @xmath21 crucial enter formula transition probability first perturbation theory results highest accuracy except latest variational calculations [ 29 ] experimental energies for levels @xmath2 stable non - depending on @xmath5 shift of @xmath23 @xmath24 from same experiment possible systematic error in measurements four experiments on @xmath0 - @xmath3 transition probability carried for c iii fe xxiii kr xxxiii xe li 21 best known is solar plasma diagnostics intercombination line 1909 in c iii discussed in theoretical papers in 27 results on c iii 19 overlap with previous measurements same group for transition probability new value @xmath25 @xmath11 agrees with result @xmath26 @xmath27 and mcrrpa value of 118 @xmath11 contradicts latest ci mchf mcdf 28 calculations @xmath28 @xmath29 @xmath30 @xmath11 agreement between mz probabilities data for fe , kr xe good extensive mcdf calculations for be - like ions up to xe li model potential results mc calculations with superstructure code theoretical results same @xmath5-dependence and absolute values of probabilities for large @xmath31 situation for small nuclear charges not clear new accurate measurements for @xmath32 importancefig . 2 theoretical experimental results transition probability @xmath33 shown old data for @xmath37 9 @xmath38 15 latest data @xmath39 14 theoretical results include mcdf mchf ci 16 hf - with - relativistic - corrections 10 calculations fritzsche grant small @xmath5 mcdf method slow convergent with increase configurations large discrepancy between paper results kim _ et al _ 18 ] different theoretical approaches show similar @xmath5-dependence except smallest @xmath5 more accurate measurements desirable new experimental results granzow et al 14 give @xmath5-dependence different theoretical narrow region na viii value @xmath40 larger than recent theoretical results needs investigation authors claim overestimate grateful to drs . s.fritzsche.grant results drs . p.l.smith e.trbert for discussions acknowledges financial support israeli ministries of absorption science hospitality dr . i.yu.tolstikhina stay nagoya | 0.265816 | 1,668 | 835 | , Saving $0.0 in GPT-4. |
20,472 | recent monte carlo simulation studies of a lennard - jones fluid confined to a mesoscopic slit - pore have reported evidence for `` critical depletion '' in the pore local number density near the liquid - vapour critical point . in this note
we demonstrate that the observed depletion effect is in fact a simulation artifact arising from small systematic errors associated with the use of long range corrections for the potential truncation .
owing to the large near - critical compressibility , these errors lead to significant changes in the pore local number density .
we suggest ways of avoiding similar problems in future studies of confined fluids . | in recent papers@xcite , one of us has reported grand canonical monte carlo simulation studies of a lennard - jones fluid confined between two structureless attractive walls arranged in a slit - pore geometry .
the behaviour of the number density profile across the pore , @xmath0 , was studied for various values of the thermodynamic parameters , namely the chemical potential , @xmath1 , and temperature , @xmath2 . at certain values of @xmath1 and @xmath2 ( apparently close to those of the bulk liquid - vapour critical point )
it was observed that the average local density in the middle of the pore fell markedly below the value obtained in a fully periodic simulation performed at the _ same _ @xmath1 and @xmath2 .
these findings were used to argue in favour of the existence of a generic `` critical depletion '' phenomenon , namely the proposed tendency of a critical fluid to be expelled by a confining medium , even when the confining walls strongly attract the fluid particles @xcite . such a scenario is supported by experimental findings for sf@xmath3 adsorbed in mesoporous materials @xcite , for which a dramatic reduction in adsorption was observed as the bulk critical temperature was approached from above along the critical isochore . in this note
we point out that the apparent critical depletion reported in references @xcite is actually a simulation artifact arising from systematic errors associated with the corrections applied to the configurational energy to compensate for the truncation of the interparticle potential .
using new simulations , we show that if one chooses a sufficiently large truncation distance or alternatively avoids the use of truncation corrections altogether , then the depletion effect disappears .
the simulation arrangement and procedure employed in this work are the same as those described in refs .
@xcite , and accordingly we merely summarise the principal features .
grand canonical monte carlo simulations @xcite were performed for a lennard - jones fluid , having an interparticle potential of the form : @xmath4\;. \label{eq : lj}\ ] ] where @xmath5 and @xmath6 are respectively the lennard - jones well depth and scale parameters .
two distinct geometries were studied : 1 . a fully periodic system .
2 . a slit - pore geometry , in which the fluid is confined between two parallel structureless walls , having periodic boundary conditions in the directions parallel to the walls . in the latter case ,
the walls were taken to exert a potential on the fluid particles of the form : @xmath7\;,\ ] ] where @xmath8 is a parameter that tunes the strength of the wall - fluid interactions relative to those of the fluid interparticle interactions . as in the previous studies of this system @xcite ,
the reduced temperature was set to the value @xmath9 , believed to be close to the bulk critical temperature .
the chemical potential @xmath1 of the periodic system was then tuned until the equilibrium density reached the value @xmath10 , believed to be close to the bulk critical density .
the resulting value of @xmath1 was then fed into a simulation of the slit - pore system at the same temperature and with the choice @xmath11 . in both the periodic and slit - pore arrangements
, the lennard - jones interparticle potential was truncated at some radius and a compensating correction applied to the configurational energy . for the periodic system ,
this correction was calculated in the standard fashion by assuming a spherical cutoff surface of radius @xmath12 centred on each particle , combined with a uniform density approximation ( uda ) for @xmath13 .
this yields for the energy correction per particle : @xmath14\;,\ ] ] where @xmath15 is the average number density of the system . in the case of the slit - pore system ,
the truncation correction used was that given by schoen et al .
@xcite , which assumes a cylindrical cutoff surface ( on whose principal axis each particle lies ) extending across the whole width @xmath16 of the pore , i.e. such that the cylinder ends coincide with the pore walls .
this yields @xmath17 where @xmath18 is the radius of the cutoff cylinder .
we have studied the effect of the cylindrical cutoff radius @xmath18 on the density profile @xmath0 of the slit - pore system . for each choice of @xmath18 , the @xmath1 value employed in the simulation was that yielding an average density @xmath10 in a periodic system of the same dimensions and with @xmath19 . for small cylindrical cutoffs ( @xmath20 ) ,
figure [ fig : cutdep ] shows that the local density in the pore middle is depleted with respect to the density of the periodic system ( dashed line ) at the same @xmath21 .
this is the result reported in refs .
however , new results for a larger choice of the cylindrical cutoff ( also included in figure [ fig : cutdep ] ) show that this depletion reduces as @xmath18 is increased , and in fact vanishes for @xmath22 .
this dependence of the depletion on the choice of @xmath18 was missed in the previous studies @xcite .
we have also performed simulations in which we dispense with the use of cutoff corrections altogether and simply simulate a system of particles interacting via a truncated lennard - jones potential .
the results ( figure [ fig : notrunc ] ) , exhibit no sign of a density depletion in the pore middle with respect to the periodic system .
the dependence of the density depletion on the choice of the cylindrical cutoff @xmath18 ( figure [ fig : cutdep ] ) points to a breakdown of the uniform density approximation invoked in the derivation of the truncation correction for the internal energy , eq .
[ eq : cylcor ] .
this approximation assumes that the number density outwith the cutoff surface is uniform , having the average system density @xmath23/v .
however , figures [ fig : cutdep ] and [ fig : notrunc ] show that for a slit system , @xmath0 exhibits considerable structure across the pore , especially close to the walls .
accordingly , one must expect some measure of systematic error to be associated with eq .
[ eq : cylcor ] .
tests show that for the choice of cutoff @xmath20 employed in references @xcite , this error is very small , so that in most circumstances eq .
[ eq : cylcor ] represents a good approximation .
it seems , however , that in the critical region , even a very small error in the truncation correction can lead to large effects on the local pore number density .
the reason for this is the large near - critical compressibility , reflected in the fact that near @xmath24 , isotherms of @xmath25 become very flat ( see eg .
figure @xmath26 of @xcite ) . since the error in the truncation correction acts rather like a shift in the bulk ( chemical potential ) field with respect to the periodic system
, large alterations to the local pore density may result .
this is in accord with the observation @xcite that the depletion is large close to the critical point , but diminishes as one moves to higher temperatures along the critical isochore . 1 .
employ a very large value for the truncation range and test for any systematic dependence of results on its value .
this is clearly a very computationally intensive solution .
2 . employ a truncated potential and dispense with corrections altogether .
such a system is clearly well defined but causes complications if one wishes to model real substances .
3 . employ a cut and shifted potential which tends smoothly to zero at the cutoff . in summary
, we have demonstrated that the apparent critical depletion reported in @xcite was actually an artifact arising from systematic errors in the energy correction for the tail truncation in the slit - pore geometry .
although numerically small , these errors can ( in the critical region ) strongly influence the fluid local number density of the confined system compared to a periodic system at the same temperature and chemical potential .
thus our findings underline the care that must be taken when implementing any sort of truncation corrections for near critical fluid models .
further simulation studies of the near - critical properties of a confined fluid are in progress and a detailed account of these and their implications for theory and experiment on critical depletion will be presented in a later paper .
nbw is grateful to r. evans and a. maciolek for stimulating his interest in critical depletion , and for most useful correspondence .
he also thanks the royal society ( grant number 19076 ) , the royal society of edinburgh and the epsrc ( grant no .
gr / l91412 ) for financial support .
ms is grateful to g.h .
findenegg for discussions and acknowledges financial support from the sonderforschungsbereich 448 `` mesoskopisch strukurierte verbundsysteme '' | recent papers@xcite one reported canonical monte carlo simulation studies of lennard - jones fluid confined between two structureless walls in slit - pore geometry behaviour number density profile across pore @xmath0 studied for thermodynamic parameters chemical potential temperature at certain values and close to bulk liquid - vapour critical point average local density in middle pore fell below value periodic simulation at same @xmath1 and @xmath2 findings argue critical depletion '' phenomenon tendency critical fluid to expelled by confining medium even walls attract fluid particles scenario supported by experimental findings for sf@xmath3 adsorbed in mesoporous materials dramatic reduction in adsorption observed as bulk critical temperature approached from above apparent critical depletion simulation artifact from systematic errors corrections to configurational energy for truncation interparticle potential new simulations if large truncation distance or avoids truncation corrections depletion effect disappears simulation arrangement procedure same as in refs. @xcite summarise principal featuresmonte carlo simulations @xcite performed for lennard - jones fluid interparticle potential : @xmath4\; @xmath5 @xmath6 lennard - jones well depth scale parameters two geometries studied fully periodic system slit - pore geometry fluid confined between walls periodic boundary conditions walls walls exert potential on fluid particles @xmath7\; @xmath8 wall - fluid interactions fluid interparticle interactions reduced temperature set to @xmath9 close to bulk critical temperature chemical potential @xmath1 tuned until equilibrium density reached @xmath10 close to bulk critical density resulting value @xmath1 fed into simulation slit - pore system same temperature choice @xmath11 periodic slit - pore lennard - jones interparticle potential truncated at radius compensating correction applied to configurational energy periodic system correction calculated spherical cutoff surface of radius @xmath12 on each particle uniform density approximation for @xmath13 energy correction per particle : @xmath14\; @xmath15 average number density systemslit - pore system truncation correction by schoen et al . assumes cylindrical cutoff surface across width of pore cylinder ends coincide with pore walls yields @xmath17 where @xmath18 is radius of cutoff cylinder studied effect of cylindrical cutoff radius @xmath18 on density profile slit - pore system each choice of @xmath18 @xmath1 value average density @xmath10 in periodic system same dimensions with @xmath19 small cylindrical cutoffs ( @xmath20 ) local density in pore middle depleted density periodic system at same @xmath21 reported in new results for larger choice cutoff show depletion reduces as @xmath18 increased vanishes for @xmath22 dependence on choice @xmath18 missed in previous studies performed simulations dispense with cutoff corrections system of particles interacting via truncated lennard - jones potential results exhibit no density depletion in pore middle periodic systemdependence of density depletion on cylindrical cutoff @xmath18 figure points to breakdown of uniform density approximation in derivation of truncation correction for internal energy cylcor approximation assumes number density outwith cutoff surface uniform average system density @xmath23/v figures and show for slit system @xmath0 exhibits structure across pore especially close to walls expect systematic error with eq cylcor ] tests show for cutoff @xmath20 error is small most represents good approximation in critical region small error in truncation correction can lead to large effects on local pore density large near - critical compressibility in near @xmath24 isotherms of @xmath25 become flat figure error in truncation correction acts like shift in bulk chemical potential field large alterations to local pore density may result accord with observation depletion large close to critical point diminishes as higher temperatures along critical isochore employ large value for truncation range test for systematic dependence of results computationally intensive solution employ truncated potential dispense with correctionssystem defined causes complications model real substances employ cut shifted potential zero at cutoff demonstrated apparent critical depletion in @xcite artifact from errors in energy correction tail truncation slit - pore geometry small errors influence fluid local density confined system periodic system same temperature findings underline care truncation corrections for near critical fluid models further simulation studies near - critical properties confined fluid progress detailed account implications for theory critical depletion presented later paper nbw grateful to r. evans a. maciolek interest critical depletion useful correspondence thanks royal society royal society of edinburgh epsrc for financial support grateful to g.h . findenegg acknowledges financial support from sonderforschungsbereich 448 `` mesoskopisch strukurierte verbundsysteme | 0.313428 | 1,890 | 911 | , Saving $0.1 in GPT-4. |
20,473 | the short interval between successive outbursts of rs oph strongly suggests that it has a high mass white dwarf accreting at a high rate .
this , in turn , suggests the possibility of prominent x - ray emission from rs oph in quiescence
. however , archival quiescent x - ray observations of rs oph show it to be a modest soft x - ray source but not a strong 210 kev x - ray source . in this aspect ,
rs oph differs markedly from t crb .
we speculate on the possible mechanisms that could significantly suppress the 210 kev x - ray emission in rs oph . . | the 2006 outburst of rs oph is almost certainly due to a thermonuclear runaway of hydrogen rich fuel that has accumulated since the previous outburst in 1985 . according to the theories of nova outbursts , both the white dwarf mass and
the mass accretion rate must be high for two consecutive nova outbursts to be separated by such a short interval . according to the calculations of @xcite , a 1.25 m@xmath0 white dwarf accreting at @xmath110@xmath2 m@xmath0yr@xmath3
would accumulate a sufficient amount of fuel in 20 years .
this corresponds to a boundary layer luminosity of @xmath11.4 @xmath4 ergss@xmath3 , assuming that half the potential energy is radiated away in the accretion disk and the white dwarf is rotating slowly .
if rs oph harbors a 1.4 m@xmath0 white dwarf , it requires less fuel to achieve a thermonuclear runaway .
even then , an accretion rate of @xmath110@xmath5 m@xmath0yr@xmath3 is required , implying a boundary layer luminosity of @xmath13 @xmath6 ergss@xmath3 .
the boundary layer emission is likely to be in the hard or soft x - rays @xcite , where the contribution from the red giant mass donor is expected to be minimal .
we have therefore searched the heasarc archive for past , pointed x - ray observations of rs oph , excluding those that were performed during the 1985 and 2006 outbursts , and found 7 observations using 4 satellites .
these quiescent x - ray observations are summarized in table 1 . in calculating the luminosity ,
the distance to rs oph was assumed to be 1.44 kpc , and no correction was made for absorption .
cccccc date & satellite & sequence & energy range & count rate & luminosity + & & & ( kev ) & ( cts@xmath3 ) & ( ergss@xmath3 ) + 1979 - 09 - 25 & _ einstein _ &
i03190 & 0.23.5 & @xmath7 & @xmath8 + 1981 - 04 - 03 & _ einstein _ & i10703 & 0.23.5 & @xmath9 & @xmath10 + 1991 - 03 - 02 & _ rosat _ & rp300038a00 & 0.22.4 & 0.0035 & 9.4@xmath11 + 1992 - 03 - 04 & _ rosat _ & rp300038a01 & 0.22.4 & 0.0116 & 3.1@xmath12 + 1997 - 02 - 04 & _ rxte _ & 20027 - 01 - 01 - 00 & 210 & 0.16 & 2.4@xmath13 + 1997 - 02 - 06 & _ rxte _ & 20027 - 01 - 01 - 01 & 210 & 0.10 & 1.7@xmath13 + 1997 - 03 - 13 & _ asca _ & 35003000 & 0.72.0 & 1.3@xmath14 & 8.6@xmath11 + & & & 210 & 1.2@xmath14 & @xmath15 + rs oph was observed twice with _
ipc well before the 1985 outburst , and was undetected both times .
the exposure times were 1.7 ksec in 1979 sep and 5.5 ksec in 1981 apr . in table 1
, we give approximate upper limits , based on an estimate that 10 source counts would have resulted in a detection .
the luminosity limits are then inferred assuming the plasma model parameters that fit the _ rosat _ data described below . the _ rosat _ pspc detections have already been published by @xcite and by @xcite .
the former reported on the first observation , and the latter was a compilation of a large number of _ rosat _ observations of novae and recurrent novae . in figure 1 ( left ) , we show the pspc spectrum from the second _ rosat _ observation , together with a mekal model fit , from which we obtain a plasma temperature of kt@xmath10.6 kev and an absorbing column of n@xmath169.6 @xmath17 @xmath18 .
note , however , that the fit is neither unique nor well constrained due to the low statistical quality of the data .
the estimated n@xmath19 value suggests that the intrinsic ( unabsorbed ) luminosity is about 10 times higher than the observed ( absorbed ) value . while these _ rosat _ observations do establish rs oph as a soft x - ray source , what about the 210 kev band
the two _ rxte _ observations should be considered non - detections , even though the observed count rates are higher than the predictions of the background model .
the count rate excesses are statistically significant , and are above the estimated systematic errors of the background model that reproduces the non x - ray background and the average cosmic x - ray background
. however , give the location of rs oph and the collimator response of _ rxte _ pca , contamination due to galactic x - ray background and faint , discrete sources in the field of view at the observed level is quite possible .
imaging x - ray telescopes are necessary to securely detect rs oph in quiescence .
the _ asca _ observation with an exposure time of @xmath134 ksec is the most sensitive observation to date of 210 kev emission from rs oph in quiescence .
of the two types of detectors on - board _ asca _ , sis is more sensitive below 2 kev , while gis is more sensitive for harder emission .
the values tabulated in table1 are from gis in two energy bands .
while rs oph is securely detected below 2 kev ( @xmath20 ctss@xmath3 ) , it is not so in the 210 kev band ( @xmath21 ctss@xmath3 ) .
the analysis of sis data produces similar results . due to the marginal statistical quality of the data ,
spectral fitting can not provide tight constraints on the spectral shape .
however , the same kt@xmath10.6 kev plasma model that was used for the _ rosat _ spectrum provides a good description of the data below 2 kev , resulting in a luminosity compatible with the first _
this model is shown with the sis data in figure 2 ( right ) .
the possible excess of data over this model above 2 kev may indicate that the emitting plasma is hotter than 0.6 kev overall , or that there is an additional contribution from a hotter plasma .
this is very different from the highly absorbed , bright x - ray component seen in the quiescent recurrent nova , t crb @xcite , which is otherwise very similar to rs oph .
t crb has a very hard , highly absorbed x - ray emission with little flux below 2 kev , with an observed 210 kev luminosity of 12 @xmath22 ergss@xmath3 .
a 210 kev gis count rate of 1.2@xmath14 ctss@xmath3 corresponds to an absorbed luminosity of 25@xmath12 ergss@xmath3 for a kt=30 kev bremsstrahlung spectrum absorbed with n@xmath19 of 10@xmath2310@xmath24 @xmath18 .
the unabsorbed luminosity may be up to a factor of 10 higher for the extreme absorption case .
compared with the expected accretion luminosity from the boundary layer , both the observed soft x - ray luminosity and the undetected hard x - ray luminosity of rs oph are extremely low .
thus , we have a case of `` missing boundary layer '' problem in rs oph that is perhaps more extreme than any seen in non - magnetic cataclysmic variables ( cvs ; @xcite ) .
in fact , we may not have detected any boundary layer emission from rs oph at all : the properties of the detected soft emission is quite similar to those seen in several other symbiotic systems @xcite , for which a colliding winds interpretation has been put forth .
it is unlikely that absorption alone can extinguish the hard x - ray flux sufficiently .
the above analysis shows that even an extreme n@xmath19 of 10@xmath24 @xmath18 is insufficient in itself to reduce a 210 kev x - ray component at the t crb level to below _
asca _ detectability level .
moreover , such a high column density is significantly above those seen in the hard x - ray component of t crb , the x - rays from rs oph in outburst ( up to 5@xmath25 @xmath18 for the early shock emission ; @xcite ) , or in optical data for rs oph is quiescence ( @xmath110@xmath26 @xmath18 ; @xcite ) .
if the boundary layer is completely optically thick , hard x - rays will not result .
note that the resulting soft x - ray component is likely to be hot in this case .
for example , a 1.4 m@xmath0 white dwarf accreting at @xmath110@xmath5 m@xmath0yr@xmath3 over 10% of its surface area should emit a kt@xmath150 ev blackbody - like soft component , which can only be hidden by a high , though realistic , intrinsic column . in any case , the luck of a hard component is puzzling since , empirically speaking , the boundary layer always appear to have a surface layer optically thin enough to produce hard x - rays @xcite . if the white dwarf in rs oph is spinning rapidly , near the break - up frequency , while that in t crb is not , this would explain the low boundary layer luminosity in the former . while uv spectroscopy of non - magnetic cvs show that they harbor slowly rotating white dwarfs @xcite , accretion of a few tenths of a solar mass should be enough to spin up the cv primaries to near break - up .
cv primaries are slow rotators probably because of the angular momentum removed during nova outbursts , which eject more mass than accreted @xcite and have an extended common envelope phase @xcite .
neither is thought to be the case with rs oph .
the high white dwarf mass inferred for rs oph almost certainly means that it has accreted a significant amount of mass , and therefore angular momentum . if indeed the nova mechanism is ineffective in removing angular momentum from the white dwarf in rs oph , it is quite possibly a fast rotator . however attractive
this possibility may be , it remains a pure conjecture .
independent studies of this using , e.g. , the morphology of the ejecta @xcite , as well as future quiescent x - ray observations , are strongly encouraged .
anupama , g.c . & mikoajewska , j. 1999 , a&a , 344 , 177 bode , m.f . ,
obrien , t.j . , osborne , j.p .
, page , k.l . , senziani , f. , skinner , g.k . ,
starrfield , s. , ness , j .- u . ,
drake , j.j . , schwarz , g. , beardmore , a.p .
, darnley , m.j . ,
eyres , s.p.s . , evans , a. , gehrels , n. , goad , m.r . ,
jean , p. , krautter , j. , novara , g. 2006 , apj , 652 , 629 king , a.r . ,
wynn , g.a .
& regev , o. 1991 , mnras , 251 , 30p livio , m. & pringle , j.e .
1998 , apj , 505 , 339 luna , g.j.m . ,
sokoloski , j.l . & mukai , k. , 2007 , in preparation .
mrset , u. , wolff , b. & jordan , s. 1997 , a&a , 319 , 201 obrien , t.j . , bode , m.f .
, porcas , r.w . ,
muxlw , t.w.b . ,
eyres , s.p.s . , beswick , r.j . ,
garrington , s.t . ,
davis , r.j .
& evans , a. 2006 , nature , 442 , 279 orio , m. 1993 , a&a , 274 , l41 .
orio , m. , covington , j. & gelman , h. 2001 , a&a , 373 , 542 patterson , j. & raymond , j.c .
1985 , apj , 292 , 535 patterson , j. & raymond , j.c . 1985 ,
apj , 292 , 550 sion , e.m .
1999 , pasp , 111 , 532 sokoloski , j.l . , luna , g.j.m . ,
mukai , k. & kenyon , s.j .
2006 , nature , 442 , 276 van teeseling , a. , beuermann , k. & verbunt , f. 1996 , a&a , 315 , 467 yaron , o. , prialnik , d. , shara , m.m . & kovetz , a. 2005 , apj , 623 , 398 | 2006 outburst rs oph due to thermonuclear runaway hydrogen fuel accumulated since 1985 white dwarf mass mass accretion rate high for two nova outbursts calculations @xcite 1.25 m@xmath0 white dwarf accreting at @xmath110@xmath3 fuel in 20 years corresponds boundary layer luminosity @xmath11.4 @xmath4 ergss@xmath3 half potential energy radiated white dwarf slowly rs oph harbors 1.4 m@xmath0 white dwarf requires less fuel thermonuclear runaway accretion rate @xmath110@xmath5 m@xmath0yr@xmath3 required boundary layer luminosity @xmath13 @xmath6 ergss@xmath3 boundary layer emission in hard or soft x - rays contribution from red giant mass donor minimal searched heasarc archive for x - ray observations rs oph found 7 observations 4 satellites summarized in table 1 distance to rs oph assumed 1.44 kpc no correction for absorptiondate satellite sequence energy range count rate luminosity kev cts@xmath3 ergss@xmath3 1979 - 09 - 25 einstein i03190 0.23.5 @xmath7 1981 - 04 - 03 i10703 0.23.5 @xmath9 1991 - 03 - 02 rosat rp300038a00 0.22.4 0.0035 9.4@xmath11 1992 - 03 - 04 rp300038a01 0.22.4 0.0116 3.1@xmath12 1997 - 02 - 04 20027 - 01 - 01 - 00 210 0.16 2.4@xmath13 1997 - 02 - 06 210 0.10 1.7@xmath13 1997 - 03 - 13 35003000 0.72.0 1.3@xmath14 8.6@xmath11 210 1.2@xmath14 @xmath15 oph observed twice ipc before 1985 outburst undetected exposure times 1.7 ksec 1979 sep 5.5 ksec 1981 apr table 1 approximate upper limits 10 source counts detection luminosity limits inferred plasma model parameters rosat data pspc detections published @xcite first observation compilation rosat observationsfigure 1 left pspc spectrum second _ rosat observation mekal model fit plasma temperature kt@xmath10.6 kev absorbing column n@xmath169.6 @xmath17 @xmath18 fit neither unique constrained low statistical quality estimated n@xmath19 value intrinsic unabsorbed luminosity 10 times higher than observed _ rosat observations establish rs oph soft x - ray source 210 kev band two _ rxte _ observations non - detections observed count rates higher than predictions background model count rate excesses statistically significant above estimated systematic errors background model location rs oph collimator response contamination due to galactic x - ray background discrete sources level possible imaging x - ray telescopes necessary to detect rs oph in quiescence _ asca _ observation exposure time @xmath134 ksec most sensitive 210 kev emission rs oph quiescence detectors sis sensitive below 2 kev gis sensitive for harder emission values table1 from gis two energy bands rs oph detected below 2 kev not in 210 kev band @xmath21@xmath3 ) analysis sis data similar resultsmarginal statistical quality spectral fitting constraints shape kt@xmath10.6 kev plasma model for rosat spectrum data below 2 kev luminosity compatible with first sis data figure 2 excess data above 2 kev may emitting plasma hotter than 0.6 kev additional contribution from hotter plasma different from highly absorbed bright x - ray in recurrent nova t crb @xcite similar to rs oph hard highly absorbed x - ray emission little flux below 2 kev observed 210 kev luminosity 12 @xmath22 ergss@xmath3 210 kev gis count rate 1.2@xmath14 absorbed luminosity 25@xmath12 ergss@xmath3 kt=30 kev bremsstrahlung spectrum absorbed n@xmath19 10@xmath2310 @xmath18 unabsorbed luminosity 10 higher for extreme absorption case observed soft x - ray luminosity hard x - ray luminosity of rs oph low missing boundary layer problem in rs oph extreme than non - magnetic cataclysmic variables boundary layer emission rs oph properties soft emission similar to symbiotic systems @xcite colliding winds interpretationunlikely absorption extinguish hard x - ray flux analysis shows extreme n@xmath19 of 10@xmath24 @xmath18 insufficient reduce 210 kev x - ray component t crb level below asca detectability level high column density above hard x - ray component t crb rs oph 5@xmath25 @xmath18 @xmath110@xmath26 @xmath18 if boundary layer optically thick hard x - rays result soft x - ray component likely hot 1.4 m@xmath0 white dwarf accreting at @xmath110@xmath5@xmath3 over 10% surface area emit kt@xmath150 ev blackbody - like soft component hidden by high intrinsic column luck hard component puzzling boundary layer optically thin produce hard x - rays white dwarf in rs oph spinning rapidly near break - up frequency t crb not low boundary layer luminosity uv spectroscopy non - magnetic cvs slowly rotating white dwarfs accretion few tenths of solar mass spin up cv primaries near break - upcv primaries slow rotators angular momentum removed during nova outbursts eject more mass than accreted @xcite extended common envelope phase neither with rs oph high white dwarf mass for rs oph means accreted significant mass angular momentum if nova mechanism ineffective angular momentum white dwarf in rs oph possibly fast rotator pure conjecture independent studies morphology ejecta @xcite future quiescent x - ray observations encouraged anupama , g.c . mikoajewska , j. 1999 , a&a , 344 , 177 bode obrien osborne j page senziani skinner starrfield ness drake schwarz beardmore darnley eyres evans gehrels goad jean krautter novara. 2006 , apj , 652 , 629 king , a.r wynn g.a regev 1991 , mnras , 251 , 30p livio m. pringle , j.e.1998 apj 505 , 339 luna g.j.m sokoloski j.l mukai 2007 preparation mrset wolff b jordan 1997 a&a 319 , 201 obrien bode porcas muxlw eyres s beswick.j garrington davis evans 2006 nature 442 , 279 orio 1993 a&a 274 , l41 orio covington j gelman h 2001 a&a 373 , 542 patterson j. raymond j.c 1985 apj 292 , 535 apj 292 , 550 sion 1999 pasp 111 , 532 sokoloski j.l luna g.j.m mukai kenyon s.j 2006 nature 442 , 276 teeseling beuermann verbunt f. 1996 a&a 315 , 467 yaron prialnik shara m.m kovetz a. 2005 apj 623 , 398 | 0.432475 | 3,069 | 1,649 | , Saving $0.1 in GPT-4. |
20,480 | we introduce some techniques to investigate dynamical mass generation .
the gross - neveu model @xcite ( gn ) is used as a toy model , because the gn mass gap is exactly known @xcite , making it possible to check reliability of the various methods .
very accurate results are obtained . also application to su(@xmath0 )
yang - mills ( ym ) is discussed . | recently , there was growing evidence that the ym - vacuum favours a condensate of mass dimension 2 @xcite .
a reasonable candidate is @xmath1 where @xmath2 labels an arbitrary gauge transformation .
( [ 1 ] ) is a gauge invariant operator and so has some physical meaning .
it reduces to the local composite operator ( lco ) @xmath3 in the landau gauge .
the landau gauge is in a way the most natural one to perform calculations . with another gauge fixing , ( [ 1 ] )
is no longer local .
more precisely , thinking of abelian projection and imposing the maximally abelian gauge ( mag ) , we should consider the gauge invariant operator @xmath4 where the index @xmath5 runs only over the off - diagonal gluons . in mag , this simplifies to the lco @xmath6 .
+ + we mentioned mag , because interesting attempts have been undertaken by kondo e.a . @xcite and recently freire @xcite to construct low energy effective theories starting from the full ym lagrangian .
their efforts rest mainly on the principle of abelian dominance @xcite , which can be understood by means of massive off - diagonal gluons and the appelquist - carazzone decoupling theorem @xcite .
mag requires a 4-ghost interaction to preserve renormalizibility @xcite , and this non - trivial ghost interaction was used in @xcite to produce an effective charged gluon mass .
our analysis learned that , in contrast to their results , the ghost condensate alone resulted in a tachyonic mass . we guess that a combination of the off - diagonal ghost and gluon condensate might provide us with a real mass . + + for the sake of simplicity , we present the different approaches using the gn - model .
available results on ym in the landau gauge will be quoted too .
we start from the manifestly u(@xmath0)-invariant gn lagrangian in @xmath7 dimensional euclidean spacetime with a source @xmath8 coupled to the lco @xmath9 @xmath10 this is an asymptotically free theory with a chiral @xmath11 symmetry if @xmath12 , preventing a perturbative non - zero value for @xmath13 and the fermion mass . when @xmath14 , new logarithmic infinities @xmath15 ( multiplicative mass renormalization ) and @xmath16 ( vacuum energy divergences ) appear . in order to remove the latter ,
a new coupling @xmath17 must be incorporated into ( [ 3 ] ) , and we get @xmath18 where @xmath19 contains all counterterm information . @xmath20 defining the bare quantities as @xmath21 the finite , generating energy functional @xmath22 satisfies a homogeneous renormalization group equation ( rge ) @xmath23 with @xmath24 the above reasoning to treat a lco seems to have 2 problems : first , @xmath17 is at this stage still arbitrary , so we have a problem of uniqueness . putting @xmath25 is a bad choice , it leads to a non - homogeneous rge for @xmath26 and a non - linear rge for the effective action @xmath27 .
secondly , the @xmath28 term spoils a simple energy interpretation . these are reflections of the criticism uttered by banks and raby on the use of lco @xcite .
both problems can be solved by choosing @xmath17 such a function of @xmath29 , so that if @xmath29 runs according to @xmath30 , @xmath17 will run properly according to @xmath31 @xcite . indeed , because of ( [ 6 ] ) and ( [ 7 ] ) , we have @xmath32 where @xmath33a solution of ( [ 8 ] ) is @xmath34 where @xmath35 is a particular solution of @xmath36the integration constant has been put to zero , in order to avoid an independent coupling constant and to have multiplicatively renormalizable vacuum divergences @xmath37
. we will solve ( [ 11 ] ) by a laurent expansion @xmath38 notice that @xmath39-loop results require @xmath40-loop knowledge of @xmath30,@xmath41 and @xmath42 .
the generating functional @xmath26 now fulfills @xmath43we conclude that the lco @xmath44 has a finite and multiplicatively renormalizable vev @xmath45 .
the effective action @xmath46 , obeys the following rge @xmath47 introducing unity via the hubbard - stratonovich transformation @xmath48\exp-\frac{1}{2z_{\zeta}\zeta}\int d^{2-\varepsilon}x\left[\frac{\sigma}{g}+\mu^{\frac{\varepsilon}{2}}z_{2}\overline{\psi}\psi-\mu^{\frac{-\varepsilon}{2}}z_{\zeta}\zeta j \right]^2\ ] ] we finally arrive at @xmath49\exp-\int d^{2-\varepsilon}x\left[{\cal l}(\sigma,\overline{\psi},\psi)-\mu^{\frac{-\varepsilon}{2}}\frac{\sigma}{g}j\right]\ ] ] where @xmath50+\frac{\sigma^{2}}{2g^{2}z_{\zeta}\zeta}+\mu^{\frac{\varepsilon}{2}}g\sigma\overline{\psi}\psi\frac{z_{2}}{g^{2}z_{\zeta}\zeta}\ ] ] @xmath8 is now a real source , in the sense that it appears linearly for @xmath51 so that we have a straightforward energy interpretation and @xmath52 .
eq.([19 ] ) is a new effective , renormalized lagrangian for gn , equivalent to the original ( [ 3 ] ) but encapsulating non - perturbative information .
perturbing around @xmath53 , we recover the original perturbation series with its infrared renormalon problems .
if we calculate the effective potential for @xmath54 , we could perturb around a non - perturbative vacuum free of renormalons . + + we wo nt rederive @xmath55 ,
the results can be found in the original papers @xcite . out of ( [ 19 ] )
, we immediately see that @xmath56 gives birth to a fermion mass . after improving the renormalization prescriptions , very accurate results for the effective fermion mass
were obtained ( see table i ) . for comparison
, we also displayed the @xmath57 and @xmath58 results .
[ tabel1 ] .deviation in terms of percentage for the mass gap with lco method [ cols="^,^,^,^",options="header " , ] the source inversion method was also tested on the chiral gross - neveu model , again with good results @xcite .
we have dealt with 2 different approaches concerning dynamical mass generation .
each method seems to give ( very good ) results in case of gn .
the almost exact gn results does nt mean we can generalize immediately to the ym case ( in the landau gauge ) .
other sources of non - perturbative effects besides infrared renormalons , such as instantons , will contribute to the dynamical mass .
+ + in the future , also mag ym deserves our attention , first trying to clear the widely accepted abelian dominance in a somewhat analytical way .
secondly , there might exist a connection between those condensate formations and the curci - ferrari lagrangian @xcite .
+ + as a general conclusion , we state it is possible to get non - perturbative information on e.g. a dynamical gluon mass in a sector of the ym - vacuum accessible to perturbation theory .
t. suzuki , i. yotsuyanagi , _ phys.rev . _ * d42 * ( 1990 ) 4257 + h. hioki , s.kitahara , s. kiura , y. matsubara , o. miyamura , s. ohno , t. suzuki , _
phys.lett . _
* b272 * ( 1991 ) 326 , _
erratum ibid._*b281 * ( 1992 ) 416 | evidence ym - vacuum favours condensate mass dimension 2 @xcite reasonable candidate @xmath1 @xmath2 labels arbitrary gauge transformation ( 1 ] ) gauge invariant operator physical meaning reduces to local composite operator @xmath3 in landau gauge natural calculations gauge fixing ( [ 1 ] ) no longer local abelian projection maximally abelian gauge mag ) consider gauge invariant operator @xmath4 index @xmath5 runs over off - diagonal gluons simplifies to lco @xmath6 attempts kondo @xcite freire @xcite construct low energy theories full ym lagrangian efforts principle abelian dominance @xcite massive off - diagonal gluons appelquist - carazzone decoupling theorem mag requires 4-ghost interaction renormalizibility ghost interaction used effective charged gluon mass analysis ghost condensate resulted tachyonic mass combination off - diagonal ghost gluon condensate might real mass present different approaches using gn - model results on ym in landau gauge quotedstart u@xmath0)-invariant gn lagrangian @xmath7 euclidean spacetime source @xmath8 lco @xmath9 asymptotically free theory chiral @xmath11 symmetry @xmath12 preventing non - zero value @xmath13 fermion mass @xmath14 logarithmic infinities @xmath15 multiplicative mass renormalization vacuum energy divergences appear remove new coupling @xmath17 3 ] @xmath18 @xmath19 contains counterterm information @xmath20 quantities @xmath21 finite generating energy functional @xmath22 satisfies homogeneous renormalization group equation @xmath23 @xmath24 reasoning lco 2 problems @xmath17 arbitrary problem uniqueness @xmath25 bad choice leads non - homogeneous rge @xmath26 non - linear rge @xmath27 @xmath28 term spoils energy interpretation criticism banks raby use lco @xcite solved @xmath17 function @xmath29 @xmath29 runs @xmath17 @xmath31[ 6 7 ] ) @xmath32 solution ( 8 ] ) @xmath34 @xmath35 integration constant zero avoid independent coupling constant renormalizable vacuum divergences @xmath37 solve ( [ 11 ] ) laurent expansion @xmath38 @xmath39 results require @xmath40-loop knowledge @xmath30 generating functional @xmath26 fulfills lco @xmath44 finite renormalizable vev @xmath45effective action @xmath46 obeys rge @xmath47 unity hubbard - stratonovich transformation @xmath48\exp-\frac{2z\zeta}{2-\varepsilon}x\left\frac\sigma\frac\varepsilon}\overline\psi}-\zeta} @xmath49\exp{2-\varepsilon}x\left\sigma\overline\psi-\mu{\frac{\varepsilon}{2}}\frac{\sigma}{g}j @xmath50+\frac{\sigma}z\zeta}}\mu{\frac{\varepsilon}{2}}g\sigma\overline{\psi}\frac{z_{2}}\zeta}\zeta @xmath8 real source appears linearly @xmath51 straightforward energy interpretation @xmath52 new renormalized lagrangian equivalent original non - perturbative information perturbing @xmath53 recover original perturbation series infrared renormalon problems calculate effective potential @xmath54 perturb non - perturbative vacuum free renormalons rederive @xmath55 results original papers @xcite19 ) @xmath56 birth fermion mass improving renormalization prescriptions accurate results effective fermion mass obtained see table i comparison displayed @xmath57 @xmath58 results tabel1.deviation percentage mass gap with lco method source inversion method tested on chiral gross - neveu model good results 2 approaches dynamical mass generation each method good results in gn exact gn results generalize to ym case other non - perturbative effects infrared instantons contribute to dynamical mass future mag ym deserves attention clear abelian dominance connection between condensate formations curci - ferrari lagrangian @xcite conclusion possible get non - perturbative information on dynamical gluon mass in sector ym - vacuum accessible to perturbation theory t. suzuki . yotsuyanagi phys.rev . d42 * ( 1990 ) 4257 hioki.kitahara kiura matsubara miyamura ohno suzuki phys.lett . b272 * ( 1991 ) 326 ibid.*b281 * ( 1992 ) 416 | 0.30527 | 1,839 | 987 | , Saving $0.1 in GPT-4. |
20,482 | antares is the first undersea neutrino detector ever built and presently the neutrino telescope with the largest effective area operating in the northern hemisphere .
a three - dimensional array of photomultiplier tubes detects the cherenkov light induced by the muons produced in the interaction of high energy neutrinos with the matter surrounding the detector .
the detection of astronomical neutrino sources is one of the main goals of antares .
the search for point - like neutrino sources with the antares telescope is described and the preliminary results obtained with data collected from 2007 to 2010 are shown .
no cosmic neutrino source has been observed and neutrino flux upper limits have been calculated for the most promising source candidates . | cosmic rays were firstly detected nearly one century ago and widely studied since then but their origin is still unknown .
high energy gamma rays have been detected from astronomical sources .
in particular astronomical objects with relativistic outflows , such blazars and micro - quasars , have been firmly established as sources of high energy gamma rays @xcite .
if these sources accelerate hadronic cosmic rays too they should produce also neutrinos which could travel unaffected to the earth .
the observation of point - like sources of cosmic neutrinos would provide a powerful insight into cosmic particle acceleration mechanisms .
the extremely low interaction cross - section of neutrinos with matter and the low neutrino fluxes expected from astronomical sources @xcite require large instrumented volumes to detect a statistically significant neutrino signal in a reasonable amount of time .
charged current interactions of high energy muon - neutrinos produce muons which can travel hundreds of meters before decaying .
muon neutrinos can therefore be detected even if they interact far away from the detector and their direction can be reconstructed with sub - degree accuracy because muon direction closely follows that of the parent neutrino .
the tracks of relativistic muons propagating in optically transparent dielectric media , like water or ice , can be reconstructed detecting the induced cherenkov light with a three dimensional array of photo - multipliers .
the large background from down - going muons due to cosmic ray interactions in the atmosphere can be reduced by placing the detector deeply underwater / under - ice and selecting only up - going muons as neutrino candidates .
up - going neutrinos produced in the interactions of cosmic rays in the atmosphere constitute an irreducible background .
antares is the first undersea neutrino detector ever built and presently the neutrino telescope with the largest effective area operating in the northern hemisphere @xcite .
located in the mediterranean sea 40 kilometers offshore the france coast near toulon , the antares experiment is predominantly sensitive to neutrinos from the southern hemisphere in the tev to pev energy range .
the cherenkov light is detected by an array of photomultiplier tubes , each housed in a pressure resistant glass sphere @xcite .
the photo - multipliers are grouped in triplets along strings which are anchored to the sea bed at a depth of 2475 meters .
antares final configuration ( may 2008 ) consists of twelve detection strings spaced approximately 60 meters .
each detector string contains up to 25 triplets , separated by 14.5 meters . the arrival time and charge of the photomultiplier signals
are digitized @xcite and transmitted to shore .
muon tracks are reconstructed by measuring the photon arrival times and the photomultiplier positions @xcite .
the preliminary results presented here have been obtained analyzing data collected in 2007 - 2010 period with detector configurations containing between five and twelve detection strings .
the overall live time is 813 days .
muon tracks have been reconstructed using a maximum likelihood fit of the observed photon arrival times .
tracks with bad reconstruction quality , quantified by a parameter based on the value of the likelihood function , have been rejected . the angular uncertainty obtained from the fit
has been also required to be smaller than one degree .
selecting up - going tracks only 3058 candidate neutrinos have been selected out of @xmath0 million reconstructed tracks .
see for a detailed description of the analysis of data collected from 2007 to 2008 .
the search for point - like neutrino sources has been carried on looking for an excess of events over the atmospheric neutrino background anywhere in the field of view ( all - sky search ) and by testing the presence of a signal at the locations of 51 candidate sources ( candidate list search ) .
candidate sources have been selected from catalogs of high energy gamma ray sources .
the list includes also the most significant spot in the all - sky search performed by the icecube collaboration with the 22 string detector configuration .
see table [ candidatesourcestab ] for the list of candidate sources .
no significant excess of events has been found at the location of any candidate source .
the most promising source candidate is an unidentified gamma - ray source , hess j1023 - 575 , while the second one is a well - known blazar , 3c 279 .
neutrino flux upper limits , calculated assuming a @xmath1 spectrum , are shown in table [ candidatesourcestab ] .
such limits are the most stringent to date for many candidate sources at @xmath2 .
no significant cluster of neutrino candidates has been found in the all - sky search either , see figure [ skymap+limits ] .
the position of the most signal - like cluster is @xmath3 where five events are within one degree .
such amount of events is compatible with the background only hypothesis .
the two - point auto - correlation function of neutrino candidates has been calculated looking for an excess at any level of angular separation .
no significant excess of events has been found at any angular scale .
no significant excess of events has been found either with the full sky search or the candidate list search .
limits on the high energy neutrino flux have been calculated for a number of selected source candidates .
the limits obtained for many of the candidate sources at declination below @xmath4 are the most stringent to date .
the author gratefully acknowledges the financial support of the spanish ministerio de ciencia e innovacin ( micinn ) , grants fpa2009 - 13983-c02 - 01 , aci2009 - 1020 and consolider multidarkcsd2009 - 00064 and of the generalitat valenciana , prometeo/2009/026 .
j.a.hinton and w.hofmann , _ ara&a _ * 47 * , 523 ( 2009 ) e.waxman and j.bahcal , _ phys .
d _ * 59 * , 023002 ( 1998 ) antares collab .
( ageron , m. , _ et al_. ) , _ nucl . instr .
_ , * a656 * , 11 ( 2011 ) antares collab .
( amram , p. , _ et al_. ) _ nucl .
_ , * a484 * , 369 ( 2002 ) antares collab .
( aguilar , j. a. , _ et al_. ) , _ nucl .
_ , * a622 * , 59 ( 2010 ) antares collab .
( aguilar , j. a. , _ et al_. ) , _ astropart .
_ , * 34 * , 652 ( 2011 ) antares collab .
( aguilar , j. a. , _ et al_. ) , _ astropart .
_ , * 34 * , 539 ( 2011 ) antares collab .
( adrian - martinez , s. , _ et al_. ) , to appear in _ the astrophysical journal letters _ , arxiv:1108.0292v1 | cosmic rays detected one century ago studied origin unknown high energy gamma rays detected from astronomical sources objects with relativistic outflows blazars micro - quasars established as sources of high energy gamma rays sources accelerate rays produce neutrinos travel to earth observation of point - like sources neutrinos insight into particle acceleration mechanisms low interaction cross - section of neutrinos with matter low neutrino fluxes require large volumes to detect neutrino signal charged current interactions of high energy muon - neutrinos produce muons travel hundreds of meters before decaying muon neutrinos detected direction reconstructed with tracks of relativistic muons in transparent dielectric media water reconstructed light with photo - multipliers large background from down - going muons reduced by detector underwater selecting up - going muons neutrino candidates up - going neutrinos constitute irreducible background antares first undersea neutrino detector largest effective area in northern hemisphere in mediterranean sea 40 kilometers offshore france coast near toulon sensitive to neutrinos from southern hemisphere tev to pev energy rangecherenkov light detected by photomultiplier tubes each in pressure resistant glass sphere multipliers grouped in triplets anchored to sea bed at depth 2475 meters antares final configuration may 2008 twelve detection strings spaced 60 meters each string contains 25 triplets separated by 14.5 meters arrival time charge signals digitized transmitted to shore muon tracks reconstructed by photon arrival times photomultiplier positions preliminary results obtained analyzing data 2007 - 2010 five overall live time 813 days tracks reconstructed using maximum likelihood fit of photon arrival times tracks with bad reconstruction quality rejected angular uncertainty required to smaller than one degree 3058 candidate neutrinos selected out of million reconstructed tracks 2007 to 2008 search for point - like neutrino sources for excess of events over atmospheric neutrino background signal at 51 candidate sources sources selected from high energy gamma ray sources list includes most significant spot in all - sky search by icecube collaboration with 22 string detector configuration table candidatesourcestab for list sources no significant excess of events found at any candidate sourcepromising source candidate unidentified gamma - ray source hess j1023 - 575 second blazar , 3c 279 neutrino flux upper limits @xmath1 spectrum shown in table candidatesourcestab limits stringent for sources at @xmath2 no significant cluster neutrino candidates found in all - sky search figure signal - like cluster @xmath3 five events within one degree compatible with background only hypothesis two - point auto - correlation function neutrino calculated excess angular separation no significant excess angular scale full sky search candidate list search limits high energy neutrino flux calculated for selected source candidates limits sources below @xmath4 most stringent author acknowledges financial support spanish ministerio de ciencia innovacin grants fpa2009 - 13983-c02 - 01 - 1020 generalitat valenciana prometeo/2009/026 j.a.hinton w.hofmann * 47 * , 523 ( 2009 ) e.waxman j.bahcal phys. 59 * 023002 ( 1998 ) antares collab ageron, m. al a656 11 ( 2011 ) amram pa484 369 ( 2002 ) antares collab aguilar j. a. nucl. a622 * 59 ( 2010 ) antares collab aguilar astropart * 34 * , 652 ( 2011 ) antares collab aguilar astropart * 34 * 539 ( 2011 ) antares collab adrian - martinez al astrophysical journal letters arxiv:1108.0292v1 | 0.261446 | 1,504 | 760 | , Saving $0.0 in GPT-4. |
20,486 | we will comment on the perihelion / periastron advance of celestial bodies due to the cosmological constant @xmath0 .
it is well known that the cosmological constant @xmath0 causes the perihelion / periastron shift ; however , there seems to still exist a discrepancy among the various derived precession formulae .
we will point out that the expression @xmath1 is the general formula for any orbital eccentricity @xmath2 and the expression @xmath3 comes from the nearly circular ( @xmath4 ) approximation .
example.eps gsave newpath 20 20 moveto 20 220 lineto 220 220 lineto 220 20 lineto closepath 2 setlinewidth gsave .4 setgray fill grestore stroke grestore | the 2011 nobel prize in physics was awarded for the discovery of the accelerating expansion of the universe @xcite . according to the present literature , theory that is most suitable for explaining
this phenomenon is that of the cosmological constant @xmath0 , generally referred to as dark energy .
although the cosmological constant or dark energy can be considered to account for the current expansion of the universe in the simple terms , the details are still far from clear . meanwhile ,
several attempts have also been made for explaining this accelerating expansion without dark energy @xcite . to verify the existence of the cosmological constant / dark energy and/or the potentiality of alternative gravitational theories , it is most natural to investigate their properties in the context of classical tests of general relativity , i.e. , the perihelion / periastron advance of celestial bodies and the deflection of light . to this end
, we must elucidate the difference in such effects among the cosmological constant and the alternative gravitational theories .
therefore , it is important to derive a rigorous formula for such effects resulting from the cosmological constant @xmath0 .
recently , the effect of the cosmological constant on the bending of light rays was intensively investigated and discussed by many authors ; the detailed reports can be found in @xcite and the references therein . the influence of the cosmological constant on the orbital motion is a historical problem ( see section 45 of @xcite ) . in principle , it is clear that the cosmological constant contributes to the perihelion / periastrion advance , although this contribution is too small to detect because of @xmath5 .
however , there is a discrepancy in the obtained precession formulae , in terms of the difference in the eccentricity dependence @xmath6 @xcite and @xmath7 @xcite .
the contribution of cosmological constant on the orbital dynamics is also investigated in other papers @xcite , and constraints on the cosmological constant from the sun s motion through the milky way is recently discussed @xcite .
in fact , this issue has been resolved in @xcite using the general framework of radial / central force perturbation .
subsequently , it was shown that the physical interpretation of the formula derived in @xcite is the precession of hamilton s vector @xcite .
a similar problem as that in @xcite was treated in terms of the gaussian planetary equation @xcite , and multiple scales method @xcite . in @xcite
, it was shown that the correct eccentricity dependence is @xmath7 as previously shown in @xcite . however , in spite of these facts , some confusion still seems to exist ( refer to e.g. , @xcite ) .
therefore , the purpose of this short note is to arrange the discussions on the perihelion / periastron advance due to the cosmological constant @xmath0 .
in order to examine the additional perihelion / periastron advance due to @xmath0 , let us assume the spacetime to be spherically symmetric ; without loss of generality , we write the metric as @xmath8c^2 dt^2 + \left [ 1 - \frac{r_g}{r } + w(r ) \right]^{-1}dr^2 + r^2 d\omega^2,\end{aligned}\ ] ] where @xmath9 , @xmath10 and @xmath11 is an additional term that is a function of @xmath12 and generally depends on the gravitational theories .
then , the geodesic equation of the test particle with mass @xmath13 becomes @xmath14.\end{aligned}\ ] ] here , @xmath15 is the proper time , and @xmath16 and @xmath17 are two constants of motion : the total energy and angular momentum , respectively . according to the standard approach ,
we define the newtonian potential @xmath18 as @xmath19 = -\frac{gm}{r } + \frac{{\cal l}^2}{2 r^2 } + \frac{1}{2}w(r)c^2,\end{aligned}\ ] ] where we set @xmath20 and let @xmath21 ; then , we have , @xmath22 if we rewrite @xmath23 and @xmath24 ( @xmath25 is coordinate time ) , eq . ( [ eq : eom1 ] )
becomes completely equivalent to eq .
( 15 ) of @xcite .
thus , henceforth , we follow the approach employed in @xcite . the perihelion / periastron shift @xmath26 due to the perturbation potential @xmath27 with respect to newtonian potential @xmath28
is expressed as ( see eq .
( 30 ) in @xcite ) @xmath29 in which @xmath30 and @xmath31 , @xmath32 being the semi - major axis of the orbit .
it should be noted that in @xcite , @xmath27 , or equivalently @xmath33 , is defined including the mass of the orbiting particle @xmath13 ( refer to eq .
( 15 ) in @xcite ) . however , in our case , @xmath27 or @xmath33 does not contain @xmath13 ; hence , hereafter , @xmath13 does not appear in the precession formula .
we are interested in the power - law perturbing potentials of the type @xmath34 . noting the relation , @xmath35 we have the precession formula , @xmath36 for discussing the effect of cosmological constant @xmath0 , the schwarzschild de sitter or kottler metric @xcite , @xmath37 is widely used and @xmath38 term causes famous general relativistic perihelion / periastron shift . in the case of perihelion / periastron advance due to @xmath0
, we consider @xmath39 and then @xmath40 . it should be noted in this case , @xmath41 ; thus , @xmath42 and @xmath43 . the integral part in eq . ( [ eq : peri - formula2 ] ) is @xmath44 therefore , from eqs .
( [ eq : peri - formula2 ] ) and ( [ eq : peri - formula3 ] ) , the perihelion / periastron advance due to @xmath0 is given by the following equation ( again , note @xmath45 and @xmath42 ) @xmath46 eq .
( [ eq : adv1 ] ) is in agreement with the results of @xcite .
let us summarize the outline of above derivation .
we start from the sds / kottler metric eq .
( [ eq : sdsk ] ) and its `` newtonian approximation ( weak field approximation ) ''
. then the perturbation potential @xmath47 or perturbation force @xmath48 is determined .
the result , eq .
( [ eq : adv1 ] ) , is obtained exactly by `` the direct integration of given perturbation potential or force '' e.g. eq .
( [ eq : peri - formula3 ] ) , with the help of the knowledge on gaussian hypergeometric function @xmath49 , see eq . ( 37 ) in @xcite .
we mention that eq .
( [ eq : peri - formula3 ] ) can be shortly calculated by the variable transformation , @xmath50 which produces the integral @xmath51 it is worthy to note when integrating eq .
( [ eq : peri - formula3 ] ) , the integral part is not expanded with respect to the orbital eccentricity @xmath2 .
the perihelion / periastron advance due to @xmath0 is considered by means of the standard perturbation method . here
let us discuss this approach according to @xcite . adopting sds / kottler metric eq .
( [ eq : sdsk ] ) , we begin with the second - order geodesic equation for time - like world - line , @xmath52 here @xmath53 is the @xmath54 component of orbital angular momentum , component of angular momentum is expressed by @xmath55 instead of @xmath53 . ] and right - hand side of ( [ eq : geo2 ] ) can be considered as the perturbation to the keplerian motion . the keplerian motion is described by @xmath56 and its solution is @xmath57 since it is well - known that the first term in right - hand side of eq .
( [ eq : geo2 ] ) causes famous general relativistic precession formula , @xmath58 , and eq .
( [ eq : geo2 ] ) is linear differential equation , then we concentrate on the perturbation due to the cosmological constant @xmath0 ; @xmath59 inserting eq . ( [ eq : adv3 ] ) into right - hand side of eq . ( [ eq : geo3 ] ) , we have @xmath60 here , it is worthy to mention when obtaining eq .
( [ eq : adv1 ] ) , the part @xmath61 is exactly integrated with changing the variable @xmath62 with @xmath54 ( see eq .
( [ eq : peri - formula3 ] ) ) .
however , in the standard literature such as @xcite , this part is expanded with respect to the orbital eccentricity @xmath2 , @xmath63 and only @xmath64 term is retained in @xcite .
comparing with the derivation process of general relativistic precession formula , the precession formula due to @xmath0 is derived as ( see eq .
( 14.25 ) in @xcite ) @xmath65 it should be emphasized here although the expansion itself is `` truncated at the first order in @xmath2 '' , the orbital angular momentum @xmath53 is given by @xmath66 _ then the quadratic terms in orbital eccentricity @xmath2 is remained in the final result due to @xmath53 but these terms are not cared in the lowest order perturbation method considered_. inserting eq .
( [ eq : rindler2 ] ) into eq .
( [ eq : rindler1 ] ) , it is found @xmath67 which is also same results of @xcite . if we expand the integral part in eq .
( [ eq : peri - formula3 ] ) up to the second order in @xmath2 and integrate , then we have @xmath68 and combining with eq .
( [ eq : peri - formula2 ] ) , eq .
( [ eq : adv2 ] ) is recovered . the eccentricity dependence of eqs .
( [ eq : adv1 ] ) and ( [ eq : adv2 ] ) is plotted in fig .
it is clear that eqs .
( [ eq : adv1 ] ) and ( [ eq : adv2 ] ) behave quite differently as the orbital eccentricity becomes large . ) and ( [ eq : adv2 ] ) . in this plot
, we set @xmath69 , @xmath70 , @xmath71 .
in this short note , we commented on the perihelion / periastron advance of celestial bodies due to the cosmological constant @xmath0 .
as we stated before , our prescription is based on @xcite , in which the problem of perihelion / periastron shift due to the cosmological constant @xmath0 has practically been resolved .
however , we hope that this note contributes to clear up the confusion about the perihelion / periastron advance due to the cosmological constant @xmath0 , and is helpful in discussing the probability of alternative theories of gravitation since @xmath0-like terms arise also in various theoretical contexts such as @xmath72 @xcite and @xmath73 gravity @xcite . just before closing this letter
, we would mention the following issue ; for high orbital eccentricity @xmath2 and large central mass @xmath74 , the general relativistic value of perihelion / periastron advance becomes large , while _ eqs .
( [ eq : adv1 ] ) and ( [ eq : adv2 ] ) tend to zero _ ( see again fig .
[ fig1 ] ) .
this property in eqs .
( [ eq : adv1 ] ) and ( [ eq : adv2 ] ) seems to be physically counter - intuitive . as a possibility
, this counter - intuitive feature may have roots in the fact that essentially , schwarzschild de sitter / kottler metric does not become asymptotically flat spacetime
. however the derivation of eqs .
( [ eq : adv1 ] ) and ( [ eq : adv2 ] ) may assume implicitly that the background metric is minkowskian .
therefore , the problem on perihelion / periastron advance due to cosmological constant @xmath0 should be investigated and discussed further in a careful manner in the context of real expanding universe ( see , for example , @xcite , and similar topic was recently discussed by @xcite ) .
we would like to acknowledge two anonymous referees for reading our manuscript carefully and for giving fruitful comments and suggestions , which significantly improved the quality of the manuscript .
we also thank m. kasai for the fruitful discussions . | 2011 nobel prize awarded for discovery accelerating expansion universe @xcite theory suitable is cosmological constant @xmath0 dark energy cosmological constant for current expansion details clear attempts for explaining without dark energy @xcite to verify cosmological constant dark energy alternative gravitational theories investigate properties in classical tests general relativity perihelion / periastron advance celestial bodies deflection of light elucidate difference in effects among cosmological constant and alternative gravitational theories important to derive rigorous formula for effects from cosmological constant @xmath0 effect cosmological constant on bending of light rays investigated detailed reports in @xcite references influence cosmological constant on orbital motion historical problem ( see section 45 of @xcite ) cosmological constant contributes to perihelion / periastrion advance contribution too small to detect because of @xmath5 discrepancy in precession formulae difference in eccentricity dependence @xmath6 @xcite and @xmath7 @xcite contribution of cosmological constant on orbital dynamics investigated in papers constraints on constant from sun s motion through milky way discussed issue resolved in @xcite using framework radial / central force perturbation physical interpretation of formula in @xcite is precession of hamilton s vector similar problem treated in gaussian planetary equation multiple scales method correct eccentricity dependence is @xmath7 confusion purpose short note to arrange discussions on perihelion / periastron advance due to cosmological constant @xmath0 examine additional perihelion advance assume spacetime spherically symmetric write metric as @xmath8c^2 dt^2 + 1 -^2 + r^2 d\omega^2 @xmath9 @xmath10 @xmath11 is additional term function of @xmath12 depends on gravitational theories geodesic equation of test particle with mass @xmath13 becomes @xmath14. @xmath15 is proper time @xmath16 and @xmath17 are constants of motion : total energy and angular momentumstandard approach define newtonian potential @xmath18 @xmath19 = -\frac{gm}{r } + l}^2}{2 r^2 } +{1}{2}w(r)c^2 set @xmath20 @xmath21 @xmath22 rewrite @xmath23 @xmath24 @xmath25 coordinate time equivalent to 15 ) @xcite follow approach @xcite perihelion / periastron shift @xmath26 due perturbation potential @xmath27 newtonian potential @xmath28 expressed eq 30 ) @xmath29 @xmath30 @xmath31 @xmath32 semi - major axis orbit @xcite @xmath27 @xmath33 defined mass orbiting particle @xmath13 eq . 15 ) @xmath27 @xmath33 contain @xmath13 appear precession formula interested power - law perturbing potentials @xmath34 relation @xmath35 precession formula @xmath36 effect cosmological constant @xmath0 schwarzschild kottler metric @xcite @xmath37 @xmath38 causes relativistic perihelion / periastron shiftperihelion periastron advance due to @xmath0 consider @xmath39 @xmath40 noted @xmath41 @xmath42 @xmath43 integral part in eq . - formula2 is @xmath44 from perihelion periastron advance due to @xmath0 given by equation note @xmath45 @xmath42 ) @xmath46 eq . adv1 agreement with results @xcite summarize derivation start from sds / kottler metric eq . sdsk newtonian approximation perturbation potential @xmath47 force @xmath48 determined result eq . ( : adv1 ] ) obtained by direct integration of perturbation potential force '' eq . peri - formula3 knowledge gaussian hypergeometric function @xmath49 see eq . ( 37 ) @xcite eq . formula3 calculated by variable transformation @xmath50 produces integral @xmath51 integrating eq . - formula3 integral part not expanded orbital eccentricity @xmath2 perihelion periastron advance due to @xmath0 considered standard perturbation method discuss approach according to @xcite adopting sds / kottler metric eq . begin second - order geodesic equation time - world - line @xmath52 @xmath53 @xmath54 component orbital angular momentum expressed @xmath55 right - hand side eq : geo2 perturbation keplerian motion keplerian motion described @xmath56 solution @xmath57 first term right - hand side eq geo2 causes general relativistic precession formula @xmath58 eq geo2 linear differential equation concentrate perturbation cosmological constant @xmath0 @xmath59 inserting eq . eq : adv3 ] ) right - hand side eq @xmath60 obtaining eq . eq : adv1 part @xmath61 integrated changing variable @xmath62 @xmath54 peri - formula3 standard literature @xcite part expanded orbital eccentricity @xmath2 @xmath63 @xmath64 term retained @xcite derivation process general relativistic precession formula precession formula due @xmath0 derived eq14.25 @xcite @xmath65 emphasized expansion truncated first order @xmath2 orbital angular momentum @xmath53 given by @xmath66 quadratic terms orbital eccentricity @xmath2 remained final result due to @xmath53 terms not cared lowest order perturbation method inserting eq . rindler2 ] into eq eq rindler1 found @xmath67 same results @xcite expand integral part eq . peri - formula3 second order @xmath2 integrate @xmath68 combining with eq peri - formula2 eq . eq : adv2 ] ) recovered eccentricity dependence of eqs adv1 adv2 plotted in fig . eqs adv1 adv2 behave differently orbital eccentricity becomes large plot set @xmath69 @xmath70 @xmath71 commented perihelion / periastron advance celestial bodies due to cosmological constant @xmath0 prescription based on @xcite problem perihelion / periastron shift due to cosmological constant @xmath0 resolvedhope note confusion about perihelion / periastron advance due to cosmological constant @xmath0 helpful discussing alternative theories gravitation @xmath0-like terms arise in theoretical contexts @xmath72 @xcite @xmath73 gravity @xcite before closing mention issue high orbital eccentricity @xmath2 large central mass @xmath74 relativistic value of perihelion / periastron advance becomes large eqs . eq adv1 adv2 ] tend to zero see fig . fig1 ] property in physically counter - intuitive may roots in schwarzschild de sitter / kottler metric become asymptotically flat spacetime derivation of eqs may assume background metric is minkowskian problem on perihelion / periastron advance due to cosmological constant @xmath0 should be investigated discussed context of expanding universe see @xcite similar topic acknowledge two anonymous referees for reading manuscript giving comments suggestions improved quality manuscript thank m. kasai for discussions | 0.430736 | 2,857 | 1,516 | , Saving $0.1 in GPT-4. |
20,490 | the character of asymmetry between the spectra of @xmath0 and @xmath1 , obtained recently in the e781 experiment ( fnal ) is discussed on the basis of the quark gluon string model(qgsm ) . as it was shown in the description of the asymmetry between the spectra of leading and nonleading charmed mesons measured in @xmath2 interactions at @xmath3= 340 gev / c in the wa89 experiment in previous studies ,
the asymmetries between @xmath4 and @xmath5 meson spectra and between @xmath6 and @xmath7 meson spectra can be fitted by qgsm curves obtained with the same parameter of string fragmentation , @xmath8=10 , as well as the asymmetry between the d - meson spectra in @xmath9 collisions at the e791 experiment .
the forms of @xmath0/@xmath10 asymmetry dependenses measured in @xmath11 and @xmath12 collisions at @xmath3= 600 gev / c in the e781 experiment are different .
it is shown in the framework of qgsm that they depend on whether the diquarks of beam and target particles took part in charmed baryon formation or not .
the qgsm results are compared with the calculations carried out by the other authors . | the difference in @xmath13 spectra ( x=@xmath14 ) of leading and nonleading particles has been discussed recently and several theoretical models explained successfully the asymmetry as an effect of an interplay between the quark contents of the projectile and of the produced hadron .
the charmed particles of leading type ( lp ) containing at least one ordinary quark of the same type as the beam particle have higher average @xmath13 value than the particles ( nlp ) of the same sort having no one quark in common with the projectile . the asymmetry a(x ) is usually defined as : @xmath15 two experiments have measured recently the asymmetries between the spectra of @xmath4 and @xmath5 , @xmath6 and @xmath7 mesons as well as between @xmath0 and @xmath10 in @xmath11 interactions : wa89 @xcite ( cern ) at @xmath3= 340 gev / c and e781 @xcite(fnal ) at @xmath3= 600 gev / c .
it seems to be interesting to consider these results from the point of view of quark gluon string model ( qgsm ) and to compare with @xmath16 asymmetry dependencies obtained in the @xmath9 experiments @xcitein order to understand the influence of quark composition of beam particle on the features of heavy flavored particle production .
the nonperturbative approach accepted in our qgsm model @xcite exploits the properties of fragmentation functions in order to insert the asymmetry .
the process of multiparticle production can be illustrated in qgsm with the cut n - pomeron diagram .
the inclusive production cross section of hadrons h is written as a sum over n - pomeron cylinder diagrams : @xmath17 where function @xmath18 is a particle distribution in the configuration of n cut cylinders and @xmath19 is the probability of this process .
the parameter of the supercritical pomeron used here is @xmath20 .
the distribution functions are discribed in previous papers @xcite .
it should be mentioned here that a structure function of i - th quark wich has a fraction of energy @xmath21 in the interacting hadron,@xmath22 , and @xmath23 , a fragmentation function of this quark into the considered type of produced hadrons h , are constracted according to the regge asymptotic rules proposed in @xcite . in the case of hyperon beam
they depend on the parameter of the regge trajectory of @xmath24-mesons ( @xmath25 ) because of s - quark contained in @xmath26 .
the following favoured fragmentation function into @xmath27-mesons was written , for instance , for the strange valence quark fragmentation : @xmath28 where @xmath29=2@xmath30(0)@xmath31 .
an additional factor @xmath32 provides the parametrization of the probability of heavy quark production in the interval z=0 to z @xmath33 1 .
the function for the nonleading fragmentation of d - quark chain into @xmath34 has an additional @xmath35 according to the same rules : @xmath36 where @xmath37(0 ) .
the dd - diquark fragmentation includes the constant @xmath38 which could be interpreted as `` leading '' parameter , but this value is fixed due to baryon number sum rule and should be approximately equal to @xmath38 taken for @xmath0 spectra in our previous calculations @xcite : @xmath39 where the term @xmath40 means the probability for initial diquark to have z close to 0 .
some fractions of sea quark pairs in hyperon , @xmath41 and @xmath25 , are to be taken into account as far as they suppress the leading / nonleading asymmetry .
the structure functions of ordinary quark pairs in the quark sea of hyperon can be written by the same way as the valence quark distributions . as soon as we accounted @xmath41 and @xmath25 fraction in the quark sea of hyperon some fraction of charmed sea quark
are to be considered as well .
this small heavy quark admixture plays an important role due to its strong impact on the difference between leading and nonleading charmed meson spectra .
the charmed sea quark structure function is similar to the distribution of strange sea quarks @xcite : @xmath42 where @xmath43 is the weight of charm admixture in the quark sea of hyperon .
in fact it is not necessarily to be equal to the charmed quark fraction in quark sea of pion @xcite .
this is only one parameter we can vary for @xmath44 interaction after the best fit of pion experimental data which had been done before .
the value of @xmath43 can be estimated in the description of the wa89 data on @xmath45- and @xmath46-meson asymmetries .
fragmentation functions into @xmath46 mesons are the following : @xmath47
the main parameter of qgsm scheme which is responsible for leading / nonleading charm asymmetry is @xmath8 , defined above .
the fraction of charmed sea quarks , @xmath49 , is the second parameter in this calculations which makes the asymmetry lower because of the equal amounts of @xmath5 and @xmath4 mesons produced by each sea charmed quark pair .
two sets of this couple of parameters were chosen in the description of @xmath9 reaction data : @xmath8=4 , @xmath50=0 and @xmath8= 10 , @xmath51 .
we concider here these two values of @xmath8 taking the @xmath52 as more or less free parameter .
the two curves provided the fits of e791 and wa92 pion beam experiment data @xciteare presented in fig.1 with two sets of parameters discussed above as well as the results of two other approaches @xcite .
+ it should be mentioned that the smaller fraction of charmed sea quarks was taken into account ( @xmath53 for to describe both @xmath54 asymmetries in @xmath11 reaction instead of @xmath55 in the case of pion beam at the e791 experiment .
the asymmetry between the spectra of @xmath0 and @xmath10 measured in @xmath11 collisions can be easy obtained practically in the same calculations described here .
what is important , the leading @xmath0 baryon is formated from single d - quark of projectile particle .
no diquark from @xmath26 hyperon takes part in leading charm baryon production .
it allows us to take the results of our calculation for d - meson production at @xmath3= 600 gev / c and to compare with the @xmath0/@xmath10 asymmetry measured in in @xmath11 collisions in the e781 experiment @xcite.the parameter @xmath8 means in this case a parametrization parameter for the density of uc - diquark in the string fragmentation , it can be different than @xmath8 taken for d - meson production .
the energy of interaction has to be also changed .
the parameter of the fraction of sea charm quarks , @xmath49 , must be of the same value as for d - meson calculations because it does nt know what leading particle is produced .
the asymmetry between the spectra of @xmath0 and @xmath10 measured in @xmath11 collisions at @xmath3= 600 gev / c is shown in fig.2 .
+ the complete calculations carried out with the fragmentation function written for @xmath0 and @xmath10 production with proton beam give the good description of data with the value of diquark fragmentation parameter @xmath56 ( see fig.3 ) .
we can conclude here only that the data of the wa89 experiment on leading asymmetry in charmed meson production at @xmath11 interactions can be described within the framework of quark - gluon string model with the same asymmetry parameter @xmath57 as e791 data for @xmath9 reaction .
good statistics on @xmath58 and @xmath59 production is needed to estimate the comparable influence of light , strange and charmed sea quark fractions for different beams . @xmath58 and
@xmath59 asymmetries measured with @xmath26 beam are discribed with the different weight of charmed quark sea of interacting hyperon ( @xmath49=0.01 ) than it was done for @xmath60 beam interaction ( @xmath49=0.05 ) .
@xmath61 asymmetry is higher than @xmath58 asymmetry because strange quark pairs suppressing the asymmetry at @xmath45 production have lower weight in quark sea of hyperon than ordinary @xmath41 pairs which cause the suppression of @xmath58 meson asymmetry .
data of the e781 experiment on charmed baryon production asymmetry measured in @xmath11 collisions can be also described within the framework of quark - gluon string model with the asymmetry parameter @xmath62 for uc - diquark density in d - quark string fragmentation .
the asymmetry between the spectra of @xmath0 and @xmath10 does nt grow so rapidly with @xmath16 as it was predicted in other approaches .
the rapid growing of asymmetry in the case of proton beam can be explained by the important role of ud - diquark fragmentation into @xmath0 . 99 wa89 collaboration , m.i.adamovich _ et.al._ , europian phys .
selex collaboration , m.iori _ et.al .
_ these proceedings , hep - ex/0009049(2000 ) .
e791 collaboration , aitala _
et.al._ , phys.lett.*b411*(1997)230 .
wa92 collaboration , m.i.adamovich _
et.al._ , nucl.phys.*b495*(1997)3 .
kaidalov a.b .
, phys.lett.*116b*(1982)459 .
piskounova o.i .
, nucl.phys.(proc.suppl.)*b50*(1996)179 , phys.of at.nucl .
* 60*(1997)439 .
piskounova o.i.,preprint fian-140,1987 .
kaidalov a.b.,sov.j.nucl.phys .
* 45*(1987)1450 .
kaidalov a.b . and
piskounova o.i . , phys.of at.nucl . *
kaidalov a.b . and
piskounova o.i . , sov.j.nucl.phys . * 43*(1986)994 ; piskounova o.i.,phys.of at.nucl . *
likhoded a.k . and
slabospitsky s.r.,phys.of at .
arakelyan g.h .
, phys.atom.nucl.61(1998)1570 .
arakelyan g.h .
, hep - ph/9906544 ( 1999 ) . | difference in @xmath13 spectra@xmath14 leading nonleading particles discussed models explained asymmetry interplay between quark contents projectile produced hadron particles leading type quark have higher @xmath13 value particles nlp no quark projectile asymmetry a(x ) defined as @xmath15 experiments measured asymmetries between spectra @xmath4 @xmath5 @xmath6 mesons @xmath0 @xmath10 @xmath11 interactions @xcite at @xmath3= 340 gev / c @xcite @xmath3= 600 gev / c consider results quark gluon string model compare with @xmath16 asymmetry dependencies @xmath9 experiments understand influence quark composition heavy flavored particle production nonperturbative approach qgsm model exploits fragmentation functions asymmetry multiparticle production illustrated in qgsm with n - pomeron diagram inclusive production cross section of hadrons h sum over n - pomeron cylinder diagrams @xmath17 function @xmath18 particle distribution n cylinders @xmath19 probability process parameter supercritical pomeron @xmath20 distribution functions discribed in previous papersstructure function i - th quark fraction energy @xmath21 interacting hadron fragmentation function into hadrons constracted regge asymptotic rules @xcite hyperon beam depend on regge trajectory @xmath24-mesons @xmath25 s - quark in @xmath26 fragmentation function into @xmath27-mesons valence quark fragmentation @xmath28 @xmath29=2@xmath30(0)@xmath31 additional factor @xmath32 probability heavy quark production interval z=0 to z @xmath33 1 function nonleading fragmentation d - quark chain into @xmath34 additional @xmath35 @xmath36 @xmath37(0 ) dd - diquark fragmentation includes constant @xmath38 leading parameter fixed baryon number sum rule equal to @xmath38 @xmath0 spectra @xmath39 @xmath40 probability initial diquark z close to 0 fractions of sea quark pairs in hyperon @xmath41 @xmath25 suppress leading / nonleading asymmetry structure functions of ordinary quark pairs in hyperon written valence quark distributions @xmath41 @xmath25 fraction quark sea charmed sea quark consideredsmall heavy quark admixture leading nonleading charmed meson spectra charmed sea quark structure similar to distribution strange sea quarks weight charm admixture in quark sea hyperon not equal to charmed quark fraction in quark sea pion one parameter for @xmath44 interaction after best fit pion experimental data value @xmath43 estimated in wa89 data @xmath45- @xmath46-meson asymmetries fragmentation functions @xmath46 mesons @xmath47 main parameter leading nonleading charm asymmetry @xmath8 fraction charmed sea quarks @xmath49 second parameter asymmetry lower equal amounts @xmath5 @xmath4 mesons quark pair two sets parameters chosen @xmath9 reaction data @xmath8=4 @xmath50=0= 10 concider two values @xmath8 @xmath52 as free parameter two curves fits of e791 wa92 pion beam experiment data presented in fig.1 with two sets parameters results two other approaches smaller fraction of charmed sea quarks @xmath53 @xmath54 asymmetries in @xmath11 reaction instead @xmath55 pion e791 experimentasymmetry between spectra @xmath0 collisions obtained calculations @xmath0 baryon from single d - quark projectile particle no diquark from @xmath26 hyperon charm baryon production results d - meson production at @xmath3= 600 gev / c compare with @xmath0/@xmath10 asymmetry collisions experiment @xmath8 density uc - diquark string fragmentation different than d - meson production energy interaction changed parameter fraction sea charm quarks @xmath49 same value d - meson calculations leading particle asymmetry between spectra @xmath0 collisions at= 600 gev / c shown in fig.2 complete calculations fragmentation for @xmath0 proton beam data diquark fragmentation parameter @xmath56 fig.3 data wa89 experiment asymmetry charmed meson production at @xmath11 interactions described quark - gluon string model with same asymmetry parameter @xmath57 as data @xmath9 reaction statistics on @xmath58 @xmath59 production to estimate influence light strange charmed sea quark fractions for different beams@xmath58 @xmath59 asymmetries @xmath26 beam different weight charmed quark sea interacting hyperon @xmath49=0.01 ) @xmath60 beam interaction.05 ) @xmath61 asymmetry higher strange quark pairs suppressing asymmetry @xmath45 production lower weight quark sea hyperon @xmath41 suppression @xmath58 meson asymmetry e781 experiment charmed baryon production asymmetry @xmath11 collisions described quark - gluon string model asymmetry parameter @xmath62 uc - diquark density d - quark string fragmentation asymmetry between spectra @xmath0 @xmath10 grow rapidly @xmath16 rapid growing asymmetry proton beam explained role ud - diquark fragmentation @xmath0 wa89 collaboration.adamovich hep - ex/0009049(2000 e791.lett.*b411*(1997)230.adamovich nucl.phys.*b495*(1997)3 phys.lett.*116b*(1982)459 piskounova nucl.phys.*b50*(1996)179(1997)439fian-140,1987 kaidalov a.b.nucl.phys 45*(1987)1450 piskounova o at.nucl piskounova.nucl.phys 43*(1986)994 piskounova at.nucl likhoded a.k slabospitsky s.r. arakelyan g.h.atom.nucl.61(1998)1570 arakelyan g hep ph/9906544 1999 | 0.349404 | 2,411 | 1,298 | , Saving $0.1 in GPT-4. |
20,492 | the advent of international wideband communication by optical fibre has produced a revolution in communications and the use of the internet .
many african countries are now connected to undersea fibre linking them to other african countries and to other continents .
previously international communication was by microwave links through geostationary satellites .
these are becoming redundant in some countries as optical fibre takes over , as this provides 1000 times the bandwidth of the satellite links . in the 1970 s and 1980 s
some two dozen large ( 30 m diameter class ) antennas were built in various african countries to provide the satellite links .
twenty six are currently known in 19 countries . as these antennas
become redundant , the possibility exists to convert them for radio astronomy at a cost of roughly one tenth that of a new antenna of similar size .
hartrao , ska africa and the south african department of science and technology ( dst ) have started exploring this possibility with some of the african countries . | new undersea fibre optic cables provide communication bandwidths a thousand times greater than radio links via communications satellites .
large antennas at satellite earth stations are becoming redundant as a result .
this trend has now reached africa with the advent of fibre optic cables down the east and west coasts of the continent .
more than two dozen large satellite earth station antennas built in the 1970 s and 80 s exist in africa .
twenty six are currently known in 19 countries .
antennas that are redundant could be refitted for radio astronomy at comparatively small cost .
africa currently has only one antenna equipped to participate in global radio astronomy using very long baseline interferometry ( vlbi ) for high resolution imaging , this being the 26 m ex - nasa antenna at hartebeesthoek in south africa .
the window of opportunity to create an african network of vlbi - capable radio telescopes from redundant large satellite antennas exists now .
this african vlbi network ( avn or afvn ) would initially work with existing vlbi arrays such as the european vlbi network ( evn ) , but could operate independently if sufficient antennas become available .
such a network would bring new science opportunities to participating countries on a short time scale , enable participation in ska pathfinder development and science , and would help create the environment for bringing the square kilometre array ( ska ) to africa .
it would provide a ready made vlbi extension for the meerkat radio telescope currently being developed as an ska precursor .
from the 1960 s , communication via satellites orbiting the earth was introduced to carry voice , data and tv signals , to supplement undersea cables . the radio bands allocated for this were 5.925 @xmath0 6.425 ghz for uplink and 3.700 @xmath0 4.200 ghz for downlink .
these are within the frequency range known as `` c - band '' .
intelsat defines earth stations for this band as standard a , b , f or h depending on their technical characteristics @xcite .
initially a standard a antenna had to be at least 30 m in diameter , and the antennas built in africa from 1970 to 1985 are this size . from 1985 new technology
enabled satellite transmitter power to increase and the required size was greatly reduced .
ska africa partner countries identified with large antennas are south africa ( 3 antennas ) , ghana , kenya , madagascar and zambia .
other african countries in which large antennas have been located are algeria ( 2 ) , benin , cameroon ( 2 ) , congo peoples republic , egypt ( 2 ) , ethiopia , malawi , morocco , niger , nigeria ( 3 ) , senegal , tunisia , uganda and zimbabwe .
these two groups are shown in fig .
[ fig : antenna_map ] .
in addition , gabon probably has one and one is reported in the congo democratic republic . the antenna in mozambique
was dismantled and probably also the one in togo .
south africa s ska partners in africa are botswana , ghana , kenya , madagascar , mauritius , mozambique , namibia and zambia .
locations of identified large satellite earth station antennas in africa ] operational converted antennas are the 30 m ceduna antenna in australia @xcite and the 32 m yamaguchi antenna in japan @xcite .
the warkworth 30 m antenna in new zealand was handed over for conversion on 19 november 2010 @xcite . in peru
the sicaya 32 m antenna is being converted with assistance from the yamaguchi team and saw first light in march 2011 @xcite .
three antennas at the goonhilly downs station in the uk decommissioned in 2008 are proposed for conversion for use with e - merlin and the evn @xcite . on 10
may 2011 it was announced that the 32 m elfordstown antenna outside cork in ireland is to be converted @xcite .
widely separated radio telescopes operating together create a virtual telescope equal in size to the project distance , or baseline , between the telescopes .
the angular resolution of a telescope depends on its size , so the larger the separation , the better the resolution . practically , to create good images a number of telescopes are needed , separated by small , medium and large distances .
the hartrao 26 m radio telescope is valuable in providing long baselines to radio telescope arrays in on other continents , e.g. europe ( european vlbi network - evn ) and australia ( australia telescope long baseline array - at - lba ) , and thus high angular resolution imaging .
however , the large distance between the south african telescope and the others makes for less than ideal imaging .
it would be helpful to have antennas filling the gap between south africa and europe and south africa and australia ( for example ) and the image quality would be substantially improved .
they would be a powerful addition to the radio telescope arrays on other continents . to create an interferometer able to operate independently to produce images ,
a minimum of four antennas are needed to provide both phase and amplitude closure .
thus a minimal african vlbi network able to operate independently of the radio telescopes on other continents could be formed if four suitable antennas were brought into use on the continent and on neighbouring islands .
science capability would improve substantially if more large antennas could join the network . with four antennas ,
50% of the phase information and 33% of the amplitude information can be recovered ; with ten antennas these rise to 80% and 78% respectively @xcite .
astronomical objects suitable for study by vlbi are those that are radio - bright and of small angular size
. objects of very large physical size meet this requirement if they are quite distant , and radio galaxies and quasars are examples .
radio - bright supernovae in external galaxies are good examples of objects whose evolution can be studied with vlbi .
masers in star - forming regions in the milky way are examples of nearby bright , compact sources .
methanol masers at 6.668 ghz and 12.178 ghz are of particular interest currently . repeated mapping would help in investigating the causes of their variability , especially those found to be showing regular variations .
measurement of their annual parallaxes by repeated vlbi observations enables their distances to be determined , and thus the locations of the spiral arms in the milky way , where these occur , to be measured accurately .
this is only being done by northern vlbi arrays currently , leaving the fourth quadrant of the milky way , in the far southern skies , so far unmapped by this method .
distances to pulsars can be obtained by astrometric vlbi .
transient sources and gamma - ray bursts are all potential targets . the on - going study of southern hemisphere radio galaxies ( tanami project ) would be enhanced with extra telescopes able to see the southern skies .
microquasars in the milky way are targets , as are radio - loud interacting binary stars such as circinus x-1
. spacecraft in interplanetary space and at other planets that are equipped with 8.4 ghz frequency - stable transmitters are being used for precise spacecraft position determination and for study of the interplanetary medium .
this technique is known as the planetary radio interferometry and doppler experiment ( pride ) .
the hartrao 26 m telescope , and others , are regularly observing the venus express ( vex ) spacecraft in orbit around venus for this purpose . the russian phobos - grunt mission to mars moon phobos
is due for launch late in 2011 and will use this technique . converted satellite antennas with 8.4 ghz receivers in africa would be good for this owing to their collecting area and location near the equator .
antennas equipped with dual 2.3 + 8.4 ghz receivers would be able to participate in geodetic and astrometric vlbi with the current generation of radio telescopes carrying out this research .
this technique provides a very precise absolute location for the telescope , which can be transferred to co - located relative position measuring systems such as the global navigation satellite systems gps and glonass , and the laser ranging and doris systems used to obtain precise measurements of satellite orbits .
the only vlbi array that runs essentially continuously is the us very long baseline array ( vlba ) .
the evn now runs much more frequently than the traditional three four - week sessions per year of the past , with monthly e - vlbi ( real - time data streaming to the correlator ) and target of opportunity ( too ) vlbi .
however there would still be substantial non - vlbi time available for operation of each telescope as a stand - alone instrument .
the time available for single - dish astronomy would be valuable for student training purposes and for selected research projects .
techniques that could be developed would be radiometry - measuring the brightness of broad - band emission sources , spectroscopy - measuring emission and absorption line strengths , and pulsar timing - measuring the arrival times of pulses from neutron stars . establishing a capability for unattended queue - scheduled single - dish observing ( as on the hartrao 26 m telescope ) would greatly assist in permitting time series to be built up of the behaviour of scientifically interesting variable sources that can be observed with these techniques .
intelsat documentation and internet mapping services using high resolution photography of the earth s surface enabled the current existence of the large antennas to be confirmed .
postage stamps commemorating their inauguration were another valuable source of information .
government level contact has been made with some of the countries by dst to further investigate the possibility of conversion .
ghana was one of those approached , and was quick to seize the opportunity .
this opportunity is a direct result of the african ska project . the ghanaian government and vodafone , the owners of the kuntunse satellite earth station outside accra , have agreed to convert the 32 m antenna at the station for radio astronomy .
this antenna has been out of service for about ten years ; an adjacent 16 m antenna is carrying the remaining satellite traffic . a hartrao team and prof .
michael jones from oxford university visited kuntunse in march 2011 for an initial feasibility investigation , which indicated that it was in a suitable condition to proceed ( fig .
[ fig : kuntunse ] ) .
a hartrao / ska africa team visited in may to assist in starting the rehabilitation .
this antenna is located roughly half way between south africa and europe .
[ fig : uv_kuntunse ] shows how it fills the gap in uv coverage for vlbi of the evn with hartebeesthoek .
it is 6 north of the equator and so it can see the entire plane of the milky way and almost the entire sky .
vlbi uv diagram for a source at + 20declination .
the red tracks show how the kuntunse antenna improves the uv coverage by filling the gap between europe and south africa . ]
vlbi uv diagram for a source at + 20declination .
the red tracks show how the kuntunse antenna improves the uv coverage by filling the gap between europe and south africa . ]
the drive systems of out of service antennas will generally need some rectification , and this is in process with the kuntunse antenna .
more challenging is the development of radio astronomy receiver systems .
the initial aim is to build receivers for operation in radio astronomy bands within the original intelsat band range , possibly re - using the existing microwave feed horn .
the most commonly used frequency band for vlbi in the evn is 4.8 5.1 ghz ; it falls between the current transmit and receive bands of the antenna and should not be difficult to implement with the existing feed .
the 6.7 ghz methanol maser line lies a little above the feed design upper frequency and tests will be needed to see if it works at this frequency .
wide ( octave ) band feeds and multi - band feeds are now commonplace . a feed covering 4.5 to 9 ghz ,
either continuously or only in the relevant radio astronomy bands , would include the two bands already discussed and the 8.4 ghz vlbi band and would permit a wide range of science to be done .
various options exist for developing capability outside these bands .
the antenna design is the so - called `` beam waveguide '' in which large diameter pipes with 45 mirrors pass the signal to a focus in the room below the antenna , where the receiver is located .
most of the african antennas are of this design .
it comes with some advantages and disadvantages compared to the cassegrain design widely used on radio telescopes of this size , but the technical issues are solvable .
the potential for converting obsolete large antennas for radio astronomy has been recognised world - wide .
the opportunity exists for african countries to re - use these expensive assets that are becoming redundant to promote science development on the continent , at relatively low cost .
lastnum intelsat , _ intelsat earth station standards ( iess ) document iess-207 ( rev .
4 ) _ , 2005 , ( luxembourg : intelsat ) , http://www.intelsat.com/wp-content/uploads/2012/12/iess-207e.pdf mcculloch p. m. , ellingsen s. p. , jauncey d. l. , carter s. j. b. , cimo g. , lovell j. e. j. , and dodson r. g. , 2005 , _ astron j _ , * 129 * , 2034 mansfield i. , 2010 , _ telecom new zealand converts satellite dish into radio telescope _ , http://www.cellular-news.com/story/46534.php finn e. , 2011 , _ peru : interstellar contact made in huancayo observatory _ , http://archive.peruthisweek.com/news-14300-environmentnature-peru-interstellar-contact-made-huancayo-observatory heywood i. , kloeckner h .-
, beswick r. , garrington s. t. , hatchell j. , hoare m. g. , jarvis m. j. , jones i. , muxlow t. w. b. , and rawlings s. , 2011 , _ astronomy with megastructures : joint science with e - elt and ska _ ed .
i. hook , d. rigolpoulou , s. rawlings and a. karastergiou , ( preprint arxiv:1103.1214 ) | undersea fibre optic cables provide bandwidths greater than radio links large antennas at satellite stations redundant trend reached africa fibre two dozen large satellite antennas built in 1970 s and 80 s exist in africa twenty six known in 19 countries antennas redundant could be refitted for radio astronomy at small cost africa has one antenna global radio astronomy 26 m ex - nasa antenna at hartebeesthoek in south africa opportunity to create african network of vlbi - capable radio telescopes from redundant satellite antennas exists african vlbi network work with arrays could operate independently if antennas available network science opportunities ska pathfinder development environment for square kilometre array ska ) to africa vlbi extension for meerkat radio telescope from 1960 s communication via satellites introduced voice data tv signals undersea cables radio bands allocated 5.925 6.425 ghz for uplink 3.700 4.200 ghz for downlink c - band intelsat defines stations for this band as standard a b f or hstandard antenna 30 m diameter antennas built in africa from 1970 to 1985 this size 1985 technology enabled satellite transmitter power size reduced ska africa partner countries large antennas are south africa ghana kenya madagascar zambia other african countries large antennas are algeria benin cameroon congo egypt ethiopia malawi morocco niger nigeria senegal tunisia uganda zimbabwe groups shown in fig . antenna_map gabon has one congo democratic republic antenna mozambique dismantled togo south africa s ska partners are botswana ghana kenya madagascar mauritius mozambique namibia zambia large satellite station antennas operational converted antennas are 30 m ceduna antenna australia 32 m yamaguchi antenna in japan warkworth 30 m antenna new zealand handed for conversion 19 november 2010 peru sicaya 32 m antenna converted march 2011 three antennas at goonhilly downs station uk proposed for conversion 32 m elfordstown antenna outside cork ireland be convertedseparated radio telescopes together create virtual telescope equal to project distance between angular resolution depends on size larger separation better resolution good images telescopes needed separated by small medium large distances hartrao 26 m radio telescope valuable long baselines to arrays other continents europe australia high angular resolution imaging large distance between south african telescope others less than ideal imaging helpful to antennas filling gap between south africa europe australia image quality powerful addition to radio telescope arrays on other continents interferometer minimum of four antennas needed phase and amplitude closure minimal african vlbi network could be formed if four antennas on continent neighbouring islands science capability improve if more large antennas network with four antennas 50% phase 33% amplitude information recovered with ten antennas to 80% and 78% astronomical objects suitable for study by vlbi are radio - bright small angular size objects large physical size meet requirement if distant radio galaxies quasars examples radio - bright supernovae in external galaxies good examples objects evolution studied with vlbimasers in star - forming regions milky way are bright compact sources methanol masers at 6.668 12.178 ghz interest repeated mapping causes variability regular variations annual parallaxes by vlbi enables distances locations spiral arms milky way done by northern vlbi arrays fourth quadrant southern unmapped distances to pulsars obtained by astrometric vlbi transient sources gamma - ray bursts potential targets study of southern hemisphere radio galaxies project enhanced with extra telescopes microquasars in milky way targets radio - loud interacting binary stars circinus x-1 spacecraft interplanetary with 8.4 ghz frequency - stable transmitters used for position determination study interplanetary medium planetary radio interferometry and doppler experiment ( hartrao 26 m telescope venus express spacecraft russian phobos - grunt mission to mars moon phobos 2011 this technique converted satellite antennas with 8.4 ghz receivers in africa good collecting area location near equator antennas dual 2.3 + 8.4 ghz receivers participate in geodetic astrometric vlbi with current generation radio telescopestechnique provides precise location for telescope transferred to co - located position measuring systems global gps glonass laser ranging doris systems only vlbi array runs continuously is us very long baseline array ) evn runs more frequently than traditional three four - week sessions with monthly e - vlbi real - time data streaming correlator target of opportunity vlbi substantial non - vlbi time available for operation telescope time for single - dish astronomy valuable for student training research projects techniques developed radiometry emission spectroscopy emission absorption line strengths pulsar timing arrival times pulses from neutron stars capability for unattended queue - scheduled single - dish observing on hartrao 26 m telescope time series of behaviour of variable sources intelsat documentation internet mapping services high resolution photography existence of large antennas postage stamps commemorating inauguration valuable source information government level contact with countries to investigate conversion ghana approached opportunity result of african ska project ghanaian government vodafone owners kuntunse satellite earth station agreed to convert 32 m antenna for radio astronomyantenna out of service ten years adjacent 16 m antenna carrying remaining satellite traffic hartrao team prof . michael jones oxford university visited kuntunse march 2011 initial feasibility investigation suitable condition proceed fig kuntunse hartrao / ska africa team visited may rehabilitation antenna half way between south africa europe fig uv_kuntunse fills gap in uv coverage for vlbi evn with hartebeesthoek 6 north of equator entire plane milky way sky vlbi uv diagram source 20declination tracks kuntunse antenna improves uv coverage gap drive systems of out of service antennas need rectification process with kuntunse antenna challenging development of radio astronomy receiver systems aim to build receivers for radio astronomy bands within original intelsat band range possibly re - using existing microwave feed horn commonly used frequency band for vlbi evn 4.8 5.1 ghz between current transmit receive bands antenna not difficult implement with existing feed 6.7 ghz methanol maser line above feed design upper frequency tests needed wide band multi - band feeds commonplacefeed covering 4.5 to 9 ghz continuously or relevant radio astronomy bands two bands discussed 8.4 ghz vlbi band wide range science options for developing capability outside bands antenna design beam waveguide large diameter pipes with 45 mirrors pass signal to focus room below antenna receiver african antennas this design advantages disadvantages compared to cassegrain design technical issues potential converting obsolete large antennas for radio astronomy recognised world - wide opportunity for african countries to re - use expensive assets redundant promote science development low cost intelsat earth station standards iess document iess-207 rev . 4 ) 2005 luxembourg intelsat.intelsat/uploads/2012/12/iess-207e.pdf mcculloch ellingsen jauncey carter cimo lovell dodson 2005 astron j 129 2034 mansfield 2010 telecom new zealand converts satellite dish into radio telescope.cellular-news.com/story/46534 finn 2011 peru interstellar contact huancayo observatory.peruthisweek/news-14300 heywood kloecknerbeswick r garrington hatchell j hoare jarvis jones muxlow rawlings s 2011 astronomy megastructures joint science e - elt ska i. hook d. rigolpoulou rawlings a. karastergiou preprint arxiv:1103.1214 | 0.501656 | 3,104 | 1,531 | , Saving $0.1 in GPT-4. |
20,502 | over the last years , lattice calculations in pure yang - mills gauge theory seem to have come more or less to a consensus .
the ghost propagator is not enhanced and the gluon propagator is positivity violating , infrared suppressed and non - vanishing at zero momentum . from an analytical point of view , several groups are agreeing with these results . among them ,
the refined gribov - zwanziger ( rgz ) framework also accommodates for these results .
the question which rises next is , if our models hold the right form for the propagators , how to extract information on the real physical observables , i.e. the glueballs ?
how do the operators which represent glueballs look like ?
we review the current status of this matter within the rgz framework .
address = ghent university , department of physics and astronomy + krijgslaan 281-s9 , b-9000 gent , belgium address = ghent university , department of physics and astronomy + krijgslaan 281-s9 , b-9000 gent , belgium address = departamento de fsica , universidade de coimbra , + p-3004 - 516 coimbra , portugal address= departamento de fsica terica , instituto de fsica , uerj - universidade do estado do rio de janeiro + rua so francisco xavier 524 , 20550 - 013 maracan , rio de janeiro , brasil | as is well known , the yang - mills action needs to be gauge fixed in order to define the path integral .
one way to do this is trough the gauge fixing proposed by faddeev and popov .
however , as is shown by gribov , this gauge fixing suffers from gribov copies @xcite .
if we take e.g. the landau gauge , @xmath0 , it is very easy to show that there exist gauge equivalent fields @xmath1 also fulfilling the landau gauge @xmath2 when having zero modes of the faddeev - popov operator @xmath3 therefore , gribov proposed to restrict the region of integration to @xmath4 @xmath5 the region which is free of zero modes of the faddeev - popov operator @xmath6 .
we should mention however , that there are still gribov copies inside the gribov region @xcite . in 1989 , zwanziger implemented this restriction to the gribov region to all orders : the gribov - zwanziger action was born @xcite .
let us here immediately present the localized version of this action , namely @xmath7 with @xmath8 the usual faddeev - popov action , @xmath9 and @xmath10 and @xmath11 given by @xmath12 the fields @xmath13 are a pair of complex conjugate bosonic fields , while @xmath14 are anticommuting fields .
@xmath15 is not free , but fixed by a horizon condition , @xmath16 .
looking at this horizon condition , we see that it is equivalent with a dimension two condensate .
therefore , we could investigate other dimension two condensates , namely @xmath17 and @xmath18 . including these condensates
gives rise to the refined gz ( rgz ) action @xcite .
we have shown that this can be done in a renormalizable way .
we stress that the masses related to the condensates respectively @xmath19 and @xmath20 are dynamically generated .
let us now look at the ghost and the gluon propagator .
firstly , it is sure that perturbation theory fails in the infrared region as perturbatively , the gluon propagator blows up , while all lattice calculations show that there is infrared suppression .
secondly , the gribov - zwanziger predicts a gluon propagator which is infrared suppressed , and therefore clearly goes beyond perturbation theory .
unfortunately , at zero momentum , this gluon propagator vanishes , which contradicts the current numerical simulations that this propagator is non - vanishing at zero momentum @xcite .
in addition , the ghost propagator shows infrared enhancement , which is also no longer observed in the infrared .
therefore , something is still missing in the gz framework .
thirdly , let us investigate the refined gz action . in this case , the tree level gluon propagator behaves like @xmath21 this propagator is clearly infrared suppressed , also violating positivity and is no longer zero at zero momentum .
also the ghost propagator is no longer enhanced as an effect of the condensates .
this agrees again qualitatively with the latest lattice results .
+ in @xcite , we have investigated our specific form of the gluon propagator on the lattice .
two main conclusions could be drawn : 1 ) if @xmath22 , the fits are of low quality , therefore , the condensate @xmath23 is indispensable ; 2 ) nice fits for the form were found when including the condensate @xmath24 .
let us recall some estimates : @xmath25gev@xmath26 , @xmath27 gev@xmath26 and @xmath28 gev@xmath26 . from the estimate of @xmath20 , we find that @xmath29 which is in the same ballpark as other , rather independent , approaches @xcite . + in conclusion
, we could state that the rgz framework provides a possible explanation of the behavior of the gluon and the ghost propagator .
in fact , one can show that more condensates can be present , and alternative rgz frameworks exist @xcite .
+ let us also stress that in 2d , the ghost propagator is still enhanced , while the gluon propagator is zero at zero momentum , @xcite .
we have also provided evidence that in 2d , the gz formalism still holds the right results as refinement is impossible in 2d @xcite .
now that the propagators match with the lattice , we can wonder whether the rgz also holds information about the particles of ( quenched ) qcd ?
the idea to proceed is the following : we want to find an operator @xmath30 , so that the correlator @xmath31 can be put into a spectral representation : @xmath32 we then introduce @xmath33 so that when going to minkowski space , i.e. @xmath34 , we find a discontinuity along the positive real axis . in addition
we want @xmath35 to be positive for @xmath36 in order to give a particle interpretation to the correlator is proportional to the cross section , and thus has to be positive . ] and obviously @xmath37 has to be renormalizable .
+ we have worked out two different angles to attack this problem . in the first approach , we have introduced @xmath38-particles @xcite @xmath39 to make the quadratic part of the action diagonal . with the @xmath38-field strengths , @xmath40 and @xmath41 , we have proposed the following operator @xmath42 the corresponding correlator @xmath43 can be put in a spectral representation : @xmath44 whereby @xmath45 .
unfortunately , the operator @xmath46 is not renormalizable .
+ let us therefore also discuss the second angle . in @xcite , we have investigated the renormalization of the typical scalar glueball operator @xmath47 with the gz action .
let us start by noting that the faddeev - popov action @xmath48 is invariant under the brst transformation @xmath49 : @xmath50 , @xmath51 this brst invariance lies at the origin of the slavnov - taylor identity , which allows us to prove the renormalizability .
moreover , the brst charge allows us to define the sub - space of the physical states and to establish the unitarity of the @xmath52 matrix .
unfortunately , the gz action is no longer invariant under the brst transformation @xmath49 .
the new fields transform as @xmath53 and thus the breaking is proportional to @xmath15 , @xmath54 . despite this breaking ,
the gz action is still renormalizable , due to a rich set of ward identities .
moreover , only two renormalization constants are needed , which is the same as in the yang - mills theory .
let us also mention that by introducing extra fields , a symmetry can be restored again @xcite . with the breaking in mind
, we can renormalize @xmath47 within the gz framework as done in @xcite .
this renormalization is however far from trivial due to the breaking of the brst symmetry .
we find that @xmath47 mixes with the following operator : @xmath55 moreover , we have constructed a renormalization group invariant given by @xmath56 therefore , we would propose the following correlator @xmath57(x ) \left [ \frac { \beta } { g^2 } f^2 -2 \gamma_c
\mathcal e \right](y ) } \nonumber\\ & \not= \left ( \frac { \beta } { g^2}\right)^2 \braket { f^2(x ) f^2(y ) } \;.\end{aligned}\ ] ] notice that the breaking of the brst symmetry is the reason that this correlator contains extra terms beside @xmath58 . in ordinary yang - mills theory , a similar correlator can be found , but it reduces the @xmath59 due to the presence of the brst symmetry . unfortunately , with this correlator , we are unable to find a good spectral representation .
this was already found in @xcite : @xmath60 whereby the unphysical part , @xmath61 , displays cuts along the imaginary axes beginning at the unphysical values @xmath62 and the physical part , @xmath63 , has a cut beginning at the physical threshold @xmath64 .
+ + for further clarification , let us show the relation between @xmath65 and @xmath66 . at lowest order
, we have that @xmath67 and thus it are in fact the two last terms which cause the bad spectral representation in @xmath59 , and the question remains how we can find a renormalizable operator with a good spectral representation .
+ we can already mention that so far , we have only investigated spectral representations within the gz framework .
the hope is that within the rgz framework , we can still find an operator with the required properties as some aspects change of the analysis .
also , the study of glueballs looks very promising within the ( r)gz framework , see , for example , the recent results in @xcite .
this question of finding descent operators is not only relevant within the gribov - zwanziger context , but for all people involved in infrared propagators qcd . how to find good spectral operators starting from unphysical operators ?
d. dudal and n. vandersickel are supported by the research foundation - flanders ( fwo ) .
the work of o. oliveira is supported by fct under project cern / fp/83644/2008 .
silvio p. sorella acknowledges support from cnpq - brazil , faperj , sr2-uerj , capes and claf .
d. dudal , j. a. gracey , s. p. sorella , n. vandersickel and h. verschelde , phys . rev .
d * 78 * ( 2008 ) 065047 .
a. cucchieri and t. mendes , pos * lat2007 * ( 2007 ) 297 [ arxiv:0710.0412 [ hep - lat ] ] . i. l. bogolubsky , e. m. ilgenfritz , m. muller - preussker and a. sternbeck , pos * lat2007 * , 290 ( 2007 ) [ arxiv:0710.1968 [ hep - lat ] ] .
d. dudal , o. oliveira and n. vandersickel , phys .
d * 81 * ( 2010 ) 074505 [ arxiv:1002.2374 [ hep - lat ] ] .
j. a. gracey , arxiv:1009.3889 [ hep - th ] .
a. cucchieri and t. mendes , phys .
* 100 * , 241601 ( 2008 ) .
d. dudal , s. p. sorella , n. vandersickel and h. verschelde , phys .
b * 680 * , 377 ( 2009 ) .
l. baulieu , d. dudal , m. s. guimaraes , m. q. huber , s. p. sorella , n. vandersickel and d. zwanziger , phys .
d * 82 * ( 2010 ) 025021 .
d. dudal , s. p. sorella , n. vandersickel and h. verschelde , jhep * 0908 * ( 2009 ) 110 .
d. dudal and n. vandersickel , arxiv:1010.3927 [ hep - th ] .
d. dudal , m. s. guimaraes and s. p. sorella , arxiv:1010.3638 [ hep - th ] . | yang - mills action needs gauge fixed path integral gauge fixing proposed by faddeev popov gauge fixing suffers from gribov copies @xcite landau gauge @xmath0 gauge equivalent fields @xmath1 fulfilling landau gauge @xmath2 zero modes of faddeev - popov operator @xmath3 gribov restrict region integration to @xmath4 @xmath5 free of zero modes @xmath6 gribov copies inside gribov region @xcite 1989 zwanziger implemented restriction to gribov region gribov - zwanziger action born @xcite localized version action @xmath7 with @xmath8 faddeev - popov action @xmath9 @xmath10 @xmath11 @xmath12 fields @xmath13 complex conjugate bosonic fields @xmath14 anticommuting fields @xmath15 not free fixed by horizon condition @xmath16 equivalent with dimension two condensate investigate other dimension two condensates @xmath17 @xmath18 condensates refined gz action @xcite masses related to condensates @xmath19 and @xmath20 dynamically generated look at ghost gluon propagatorperturbation theory fails in infrared region gluon propagator blows up lattice calculations show infrared suppression gribov - zwanziger predicts gluon propagator infrared suppressed beyond perturbation theory at zero momentum gluon propagator vanishes contradicts numerical simulations non - vanishing at zero @xcite ghost propagator shows infrared enhancement no longer observed infrared something missing in gz framework investigate refined gz action tree level gluon propagator behaves like @xmath21 infrared suppressed violating positivity no longer zero at momentum ghost propagator no longer enhanced condensates agrees with latest lattice results investigated specific form of gluon propagator on lattice conclusions if @xmath22 fits low quality condensate @xmath23 indispensable nice fits for form found including condensate @xmath24 recall estimates : @xmath25gev @xmath27 @xmath28 estimate @xmath20 @xmath29 same ballpark as other approaches @xcite rgz framework provides possible explanation of behavior gluon ghost propagator more condensates present alternative rgz frameworks exist @xcitestress in 2d ghost propagator enhanced gluon propagator zero momentum @xcite provided evidence 2d gz formalism holds right results refinement impossible in 2d @xcite propagators match with lattice wonder rgz holds information about particles quenched qcd ? idea proceed find operator @xmath30 correlator @xmath31 spectral representation @xmath32 introduce @xmath33 minkowski space @xmath34 discontinuity along positive real axis want @xmath35 positive for @xmath36 particle interpretation correlator proportional to cross section @xmath37 renormalizable worked two angles attack first approach introduced @xmath38-particles @xcite @xmath39 quadratic part action diagonal @xmath38-field strengths @xmath40 @xmath41 proposed operator @xmath42 correlator @xmath43 spectral representation : @xmath44 @xmath45 operator @xmath46 not renormalizable discuss second angle @xcite investigated renormalization of typical scalar glueball operator @xmath47 with gz action faddeev - popov action @xmath48 invariant under brst transformation @xmath49 @xmath50 @xmath51 brst invariance origin of slavnov - taylor identity renormalizability brst charge sub - space physical states unitarity of @xmath52 matrix gz action no longer invariant under brst transformation @xmath49 new fields transform as @xmath53 breaking proportional to @xmath15 , @xmath54 despite gz action still renormalizable due to ward identities two renormalization constants needed same as yang - mills theory introducing extra fields symmetry restored @xcite renormalize @xmath47 within gz framework @xcite renormalization trivial due to breaking brst symmetry @xmath47 mixes with operator @xmath55 constructed renormalization group invariant by @xmath56 propose correlator @xmath57(x ) \beta } { g^2 } f^2 -2\gamma_c(y ) f^2(x ) f^2(y ) breaking of brst symmetry correlator contains extra terms beside @xmath58yang - mills theory similar correlator reduces @xmath59 brst symmetry correlator find good spectral representation found in @xcite @xmath60 unphysical part @xmath61 displays cuts imaginary axes at unphysical values @xmath62 physical @xmath63 cut at physical threshold @xmath64 relation between @xmath65 @xmath66 @xmath67 two last terms cause bad spectral representation in @xmath59 find renormalizable operator with good spectral representation investigated spectral representations gz framework hope rgz framework find operator required properties study of glueballs promising r)gz framework recent results @xcite finding descent operators relevant gribov - zwanziger context infrared propagators find good spectral operators from unphysical operators ? d. dudal n. vandersickel supported by research foundation - flanders work o. oliveira supported by fct project cern / fp/83644/2008 silvio p. sorella support from cnpq - brazil faperj sr2-uerj capes claf d. dudal j. a. gracey s p. sorella n. vandersickel h.verschelde phys 78 2008 ) 065047 cucchieri t mendes lat2007 2007 297 arxiv:0710.0412 hep - lat bogolubsky ilgenfritz muller preussker lat2007 290 2007 arxiv:0710.1968 hep - lat d. dudal oliveira n vandersickel 81 2010 ) 074505 arxiv:1002.2374 hep - lat a gracey arxiv:1009.3889 hep - th cucchieri t mendes 100 241601 2008 ) d. dudal sorella n. vandersickel verschelde phys 680 377 2009 ). baulieu d. dudal guimaraes huber sorella vandersickel zwanziger phys 82 2010 ) 025021. dudal sorella vandersickel verschelde 0908 2009 ) 110 dudal n vandersickel arxiv:1010.3927 hep - th dudal guimaraes sorella arxiv:1010.3638 hep - th | 0.351699 | 2,644 | 1,394 | , Saving $0.1 in GPT-4. |
20,506 | we report on the result of some investigations concerning the radiative decays of decuplet baryons and the non - leptonic weak decays of the octet baryons in the context of topological chiral soliton models .
our results are compared with those of alternative baryon models . for the radiative
decays we find that the predictions are similar to those of quark models . in the case of
the non - leptonic weak decays , we find that although the predicted @xmath0-wave amplitudes are in rather good agreement with the observed values , the model is not able to reproduce the empirical @xmath1-wave amplitudes .
thus , in contrast to previous expectations , the skyrme model does not seem to provide a solution to the long - standing @xmath0-wave/@xmath1-wave puzzle. | during the last fifteen years it has become clear that the applicability of the skyrme s topological soliton model for baryon structure@xcite goes beyond all the original expectations .
indeed , this model provides a very useful framework to investigate not only the structure of non - strange baryons and their interactions@xcite , but of strange hyperons as well . in this type of models , baryons emerge as soliton configurations of the pseudoscalar mesons .
the corresponding non - linear chiral action describes the main features of qcd at low energies .
when the model is extended to the strange sector , and in order to account for the non - negligible strange quark mass , appropriate chiral symmetry breaking terms have to be included .
one can treat the resulting effective action by starting from a flavor symmetric formulation wherein non vanishing kaon fields arise from a rigid rotation of the classical pion field .
the associated collective coordinates , which parametrize these large amplitude fluctuations off the soliton , are canonically quantized to generate states which possess the quantum numbers of physical hyperons .
it turns out that the resulting collective hamiltonian can be diagonalized exactly even in the presence of flavor symmetry breaking @xcite .
this approach leads to a good description of the hyperon spectra , magnetic moments , etc ( for a review see ref.@xcite ) .
consequently , it provides an alternative and interesting scheme to investigate other various hyperon properties . in the present contribution we will concentrate on the application of the model to the study of the electromagnetic decays of the decuplet baryons and of the non - leptonic weak decays of the octet hyperons .
the study of the decuplet baryon radiative decays is of great interest at the present time due to the upcoming experiments at jlab@xcite .
in fact , although recently the reaction @xmath2 has carefully been analyzed at mami , the decay parameters are still unknown for those @xmath3 to @xmath4 transitions which involve strange baryons .
the experiments at jlab are expected to provide some data on these radiative decays soon and thus give more insight in the pattern of flavor symmetry breaking .
theoretical studies of these decays have been performed in a number of quark - based models like i.e. the non relativistic quark model @xcite , a quenched lattice calculation @xcite , etc . here , we report the corresponding results in the skyrme model . on the other hand ,
the non - leptonic weak decays of hyperons represent a long - standing problem in hadron physics .
for example , although quark models with qcd enhancement factors have been quite successful in predicting @xmath0-wave decay amplitudes ( see ref.@xcite and references therein ) , they have serious difficulties in reproducing the empirical @xmath1-wave amplitudes .
indeed , this is a problem ( so - called `` @xmath0-wave/@xmath1-wave puzzle '' ) which seems to be common to other approaches as e.g. heavy baryon chiral perturbation theory@xcite . within the context of the topological soliton models a possible solution to this problem
has been suggested rather long ago@xcite .
it was shown that in the skyrme model , in addition to the standard pole terms , the @xmath1-wave amplitudes receive extra contributions from contact terms .
unfortunately , at the time this suggestion was made the model was hampered by several serious problems , like e.g. very poor predictions for the hyperon spectrum@xcite , far too small results for the @xmath0-wave non - leptonic decay amplitudes@xcite , etc .
consequently , it was hard to draw definite conclusions about the real relevance of such contact terms . as mentioned above the introduction of a more refined version of the model
has led to the solution of such problems .
in fact , it has been recently shown@xcite that the correct @xmath0-wave absolute values can be naturally obtained within such scheme .
thus , we are now in position to verify whether chiral soliton models can provide a unified and consistent description of the hyperon non - leptonic decays .
the strategy is to use an effective chiral lagrangian to describe the weak interactions . assuming that the corresponding low energy constants are fixed to describe the known @xmath5 meson decay amplitudes , the hyperon weak matrix elements
are then evaluated using the topological soliton model wave functions .
the rest of this contribution is organized as follows . in sec .
2 we give a brief description of the collective coordinate approach to the @xmath6 skyrme model . in sec .
3 we present our predictions concerning the radiative decays of the decuplet baryons within this approach . in sec .
4 we describe how to evaluate the non - leptonic hyperon decays and give our results for the @xmath0-wave and @xmath1-wave decay amplitudes .
finally in sec .
5 some conclusions are given .
as mentioned in the introduction , in the skyrme model baryons appear as topological excitations of a chiral effective action which depends only on meson fields . in the present work ,
the following effective action is used @xmath7 where @xmath8 is the skyrme action @xmath9 + { 1\over{32 \epsilon^2 } } \mbox{tr}\left [ [ u^\dagger \partial_\mu u , u^\dagger \partial_\nu u]^2\right ] \right\ } \ , .\ ] ] here , @xmath10 is the dimensionless skyrme parameter .
furthermore the chiral field @xmath11 is the non linear realization of the pseudoscalar octet .
@xmath12 is the wess - zumino action @xmath13 \ , \end{aligned}\ ] ] where @xmath14 and @xmath15 is the number of colors .
finally , @xmath16 represents the symmetry breaking terms @xmath17 + \sqrt{3 } { f_\pi^2 m_\pi^2 - f_k^2 m_k^2 \over{6 } } \mbox{tr } \left [ \lambda_8 \left ( u + u^\dagger \right ) \right ] \right .
\nonumber \\ & & \qquad \left . + { f_k^2 - f_\pi^2\over{12 } } \mbox{tr } \left [ ( 1- \sqrt{3 } \lambda_8 ) \left ( u ( \partial_\mu u)^\dagger \partial^\mu u + u^\dagger \partial_\mu u ( \partial^\mu u)^\dagger \right ) \right ] \right\ } \ , . \label{sb}\end{aligned}\ ] ] here , @xmath18 is the kaon decay constant while @xmath19 and @xmath20 are the pion and kaon masses , respectively . in all our numerical calculations below we
will take @xmath21 and the rest of the parameters appearing in @xmath22 to their empirical values . in the soliton picture
we are using the strong interaction properties of the low lying @xmath23 and @xmath24 baryons are computed following the standard @xmath6 collective coordinate approach to the skyrme model .
we introduce the ansatz @xmath25 for the chiral field . here
, @xmath26 is the chiral angle which parameterizes the soliton .
the collective rotation matrix @xmath27 is @xmath6 valued . substituting the configuration eq.([ansatz ] ) into @xmath22 yields ( upon canonical quantization of @xmath28 ) the collective hamiltonian @xmath29-\frac{3}{8\theta_k } + \phi \ \left ( 1 - d_{88 } \right ) \ . \label{slowham}\ ] ] in eq ( [ slowham ] ) , @xmath30 and @xmath31 are the moments of inertia in the pionic and kaonic directions , respectively , and @xmath32 the symmetry breaking strength ( see ref.@xcite for details ) .
@xmath33 denotes the spin operator while @xmath34 $ ] refers to the quadratic casimir operator of @xmath6 .
the eigenfunctions and eigenvalues of the collective hamiltonian are identified as the baryon wavefunctions @xmath35 and masses @xmath36 . due to the symmetry breaking terms in @xmath16
this hamiltonian is obviously not @xmath6 symmetric .
as shown by yabu and ando @xcite it can , however , be diagonalized exactly .
this diagonalization essentially amounts to admixtures of states from higher dimensional @xmath6 representations into the octet ( @xmath4 ) and decuplet ( @xmath3 ) states .
this procedure has proven to be quite successful in describing the hyperon spectrum and static properties @xcite .
to study the hyperon radiative decays one has to introduce the electromagnetic field @xmath37 in the effective action described in the previous section .
this is accomplished by replacing the usual derivatives @xmath38 appearing in @xmath8 and @xmath16 by covariant derivatives .
these covariant derivatives are defined as @xmath39 \ , \ ] ] where @xmath40 is the electric charge matrix @xmath41 . in the case of wess - zumino action @xmath12 ,
the gauging procedure is slightly more complicated since extra terms have to be added in order to ensure gauge invariance .
details of this procedure as well as the explicit form of the gauged wess - zumino action can be found in e.g. ref.@xcite .
the resulting gauged effective action can be cast in the following form @xmath42 where @xmath43 is the hadron e.m .
current and @xmath44 is a tensor which is responsible , for example , of the seagull contributions to the hadron polarizabilities . for the calculation of the radiative
decays only the second term in eq.([gauged ] ) is relevant . for the e.m .
decays of the decuplet hyperons we are interested in both @xmath45 and @xmath46 transitions are allowed .
the corresponding partial widths can be expressed in terms of @xmath43 as @xmath47 where @xmath48 is the photon momentum and @xmath49 are the spherical bessel functions . in our model ,
the general form of @xmath50 can be directly obtained from the gauged effective action .
using the ansatz eq.([ansatz ] ) one can then get the explicit expression of the integrands in eqs.([ge2],[gm1 ] ) .
these expressions can be found in ref.@xcite .
it should be noticed that in deriving eq.([ge2 ] ) some approximations according to the siegert s theorem have been made .
the impact of such approximations on the e2@xmath51 transition has been recently studied in the two flavor reduction of the model@xcite .
it turns out that the kinematical corrections coming from the low - momentum expansion of the bessel functions are of the order of 5 % .
in addition to this , pion fluctuations off the rotating soliton have to be included to consistently satisfy the continuity equation @xmath52 at subleading order in @xmath53 .
these induced fields account for shortcomings in the collective quantization and give corrections of @xmath54% for the @xmath55 case .
unfortunately such an inclusion of induced fields seems to be unfeasible in the three flavor model with symmetry breaking included . in any case , since the decuplet decays are almost completely m1dominated , these uncertainties do not affect the total decay widths in any significant way .
results for the total decay widths are given in table 1 .
there we list the corresponding widths normalized to the @xmath55 value . as in the case of magnetic moments ,
the calculated @xmath55 transition amplitude turns to be roughly @xmath56 smaller than the empirical value .
however , it has been recently shown that in the @xmath57 model the inclusion of next - to - leading order quantum corrections solves not only the nucleon mass problem , but also that of the magnetic moments@xcite .
similar improvements are expected in the present model . from table 1
we observe that the predictions of the present model are very similar to those of quark - based models .
in particular , in both types of models the @xmath58spin selection rules that predict small @xmath59 and @xmath60 are rather well satisfied .
._total decay widths normalized to that of the @xmath61 transition in various models .
the data for the quark model ( qm ) and lattice calculation ( latt ) are taken from refs.[6,7 ] . _ [ cols="<,^,^,^",options="header " , ] | last fifteen years applicability skyrme s topological soliton model for baryon structure@xcite beyond original expectations model provides useful framework structure non - strange baryons interactions strange hyperons baryons emerge as soliton configurations pseudoscalar mesons non - linear chiral action describes features qcd at low energies model extended to strange sector strange quark mass chiral symmetry breaking terms included from flavor symmetric formulation non vanishing kaon fields arise from rigid rotation of classical pion field collective coordinates parametrize large amplitude fluctuations off soliton quantized generate states quantum numbers of physical hyperons collective hamiltonian diagonalized flavor symmetry breaking approach leads to description hyperon spectra magnetic moments provides alternative scheme to investigate hyperon properties on application model electromagnetic decays of decuplet baryons non - leptonic weak decays of octet hyperons study decuplet baryon radiative decays interest due to upcoming experiments at jlab@xcite reaction @xmath2 analyzed at mami decay parameters unknown for @xmath3 to @xmath4 transitions strange baryonsexperiments at jlab expected to provide data on radiative decays give insight in flavor symmetry breaking theoretical studies of decays performed in quark - based models like non relativistic quark model quenched lattice calculation report results in skyrme model non - leptonic weak decays of hyperons represent problem in hadron physics quark models with qcd enhancement factors successful in predicting @xmath0-wave decay amplitudes difficulties reproducing empirical @xmath1-wave amplitudes problem @xmath0-wave/@xmath1-wave puzzle common to other approaches heavy baryon chiral perturbation theory topological soliton models possible solution suggested long ago in skyrme model @xmath1-wave amplitudes receive extra contributions from contact terms model hampered by problems poor predictions for hyperon spectrum small results for @xmath0-wave non - leptonic decay amplitudes hard to draw definite conclusions about relevance of contact terms introduction of refined version of model led to solution problems correct @xmath0-wave absolute values can be naturally obtainedverify chiral soliton models unified hyperon non - leptonic decays strategy use effective chiral lagrangian describe weak interactions low energy constants fixed @xmath5 meson decay amplitudes hyperon weak matrix elements evaluated using topological soliton model wave functions contribution organized sec. 2 description collective coordinate approach to @xmath6 skyrme model sec 3 predictions radiative decays of decuplet baryons sec 4 evaluate non - leptonic hyperon decays results for @xmath0-wave and @xmath1-wave decay amplitudes sec. 5 conclusions skyrme model baryons appear as topological excitations of chiral effective action depends on meson fields effective action used @xmath7 @xmath8 is skyrme action @xmath9 @xmath10 is dimensionless skyrme parameter chiral field @xmath11 is non linear realization of pseudoscalar octet @xmath12 is wess - zumino action @xmath13 @xmath14 and @xmath15 is number of colors@xmath16 symmetry breaking terms @xmath17 + \sqrt{3 } f_\pi^2 m^2 - f_k^2 m_k^2 \over{6 \mbox{tr\lambda_8 u + u^\dagger \right\nonumber f_k^2 - f_\pi^2\over{12 } \mbox{tr \sqrt{3 }\lambda_8\partial_\mu u)^\dagger{aligned @xmath18 kaon decay constant @xmath19 @xmath20 pion kaon masses numerical calculations take @xmath21 parameters @xmath22 to empirical values soliton picture strong interaction properties low lying @xmath23 @xmath24 baryons computed following standard @xmath6 collective coordinate approach skyrme model ansatz @xmath25 for chiral field @xmath26 chiral angle parameterizes soliton collective rotation matrix @xmath27 is @xmath6 valued configuration eq.[ansatz @xmath22 yields quantization @xmath28 collective hamiltonian @xmath29-\frac{3}{8\theta_k } + \phi\left ( 1 - d_{88 }\right{slowham} @xmath30 @xmath31 moments of inertia pionic kaonic directions @xmath32 symmetry breaking strength ref.@xcite @xmath33 spin operator @xmath34 quadratic casimir operator @xmath6 eigenfunctions eigenvalues hamiltonian baryon wavefunctions @xmath35 masses @xmath36 symmetry breaking terms @xmath16 hamiltonian not @xmath6 symmetric can diagonalized diagonalization admixtures states higher dimensional @xmath6 representations octet @xmath4 decuplet @xmath3 ) states procedure successful describing hyperon spectrum static properties study hyperon radiative decays introduce electromagnetic field @xmath37 action replacing derivatives @xmath38 @xmath8 @xmath16 by covariant derivatives defined @xmath39 @xmath40 is electric charge matrix @xmath41 wess - zumino action @xmath12 gauging procedure complicated extra terms added ensure gauge invarianceprocedure gauged wess - zumino action in ref.@xcite gauged effective action form @xmath42 @xmath43 hadron e.m current tensor seagull contributions hadron polarizabilities radiative decays second term in eq.([gauged ] ) relevant decays decuplet hyperons @xmath45 @xmath46 transitions allowed partial widths expressed @xmath43 @xmath47 @xmath48 photon momentum @xmath49 spherical bessel functions general form of @xmath50 obtained from gauged effective action ansatz eq.([ansatz ] ) explicit expression integrands in eqs.([ge2],[gm1 ] ) expressions in ref.@xcite deriving eq.([ge2 ] ) approximations siegert s theorem made impact on e2@xmath51 transition studied kinematical corrections from low - momentum expansion bessel functions 5 % pion fluctuations off rotating soliton included satisfy continuity equation @xmath52 subleading in @xmath53 induced fields account shortcomings collective quantization give corrections of @xmath54% for @xmath55 caseinclusion of induced fields unfeasible in three flavor model with symmetry breaking decuplet decays m1dominated uncertainties affect total decay widths results for total decay widths in table 1 widths normalized to @xmath55 value calculated @xmath55 transition amplitude roughly @xmath56 smaller than empirical value in @xmath57 model inclusion of next - - order quantum corrections solves nucleon mass problem magnetic moments similar improvements expected in present model predictions model similar to quark - based models @xmath58spin selection rules predict small @xmath59 and @xmath60 well satisfied_total decay widths normalized to @xmath61 transition in various models data for quark model and lattice calculation ( taken from refs.[6,7 ] | 0.427062 | 2,805 | 1,480 | , Saving $0.1 in GPT-4. |
20,512 | in the adiabatic and weak - modulation quantum pump , net electron flow is driven from one reservoir to the other by absorbing or emitting an energy quantum @xmath0 from or to the reservoirs . in our approach , high - order dependence of the scattering matrix on the time
is considered .
non - sinusoidal behavior of strong pumping is revealed .
the relation between the pumped current and the ac driving amplitude varies from power of @xmath1 , @xmath2 to @xmath3 when stronger modulation is exerted .
open experimental observation can be interpreted by multi - energy - quantum - related processes . | generally speaking , the transport of matter from low potential to high potential excited by absorbing energy from the environment can be described as a pump process .
the driving mechanics of classic pumps is straightforward and well understood@xcite .
the concept of a quantum pump is initiated several decades ago@xcite with its mechanism involving coherent tunneling and quantum interference .
research on quantum pumping has attracted heated interest since its experimental realization in an open quantum dot@xcite .
the current and noise properties in various quantum pump structures and devices were investigated such as the magnetic - barrier - modulated two dimensional electron gas@xcite , mesoscopic one - dimensional wire@xcite , quantum - dot structures@xcite , mesoscopic rings with aharonov - casher and aharonov - bohm effect@xcite , magnetic tunnel junctions@xcite , chains of tunnel - coupled metallic islands@xcite , the nanoscale helical wire@xcite , the tomonaga - luttinger liquid@xcite , and garphene - based devices@xcite .
theory also predicts that charge can be pumped by oscillating one parameter in particular quantum configurations@xcite .
a recent experiment@xcite based on two parallel quantized charge pumps offers a way forward to the potential application of quantum pumping in quantum information processing , the generation of single photons in pairs and bunches , neural networking , and the development of a quantum standard for electrical current .
correspondingly , theoretical techniques have been put forward for the treatment of the quantum pumps@xcite .
one of the most prominent is the scattering approach proposed by brouwer who presented a formula that relates the pumped current to the parametric derivatives of the scattering matrix of the system . driven by adiabatic and weak modulation ( the ac driving amplitude is small compared to the static potential ) , the pumped current was found to vary in a sinusoidal manner as a function of the phase difference between the two oscillating potentials .
it increases linearly with the frequency in line with experimental finding .
although the quantum pump has been extensively discussed in literature , little attention was paid to experimentally observed deviation from the weak - pumping theory with only the first - order parametric derivative of the scattering matrix considered .
we improved the scattering approach by expanding the scattering matrix to higher orders of the time and modulation amplitude , which enables us to go further in investigation of the problem .
we start with the scattering matrix approach detailed by moskalets _ et al_. @xcite to describe the response of a mesoscopic phase - coherent sample to two slowly oscillating ( with a frequency @xmath4 ) external real parameters @xmath5 ( gate potential , magnetic flux , etc . ) , @xmath6 @xmath7 and @xmath8 measure the static magnitude and ac driving amplitude of the two parameters , respectively .
the phase difference between the two drivers is defined as @xmath9 .
the mesoscopic conductor is connected to two reservoirs at zero bias . the scattering matrix @xmath10 being a function of parameters @xmath5 depends on time .
it is assumed that the external parameter changes sufficiently slowly to validate an instant scattering " description . to investigate the deviation from the small amplitude @xmath11 limit ,
we expand the scattering matrix @xmath12 into taylor series of @xmath13 to second order at @xmath7 with the terms linear and quadratic of @xmath11 present in the expansion , @xmath14 with @xmath15 it can be seen from the equations that higher orders of the fourier spectra enter into the scattering matrix . as a result , both the nearest and next nearest sidebands are taken into account , which implies that a scattered electron can absorb or emit an energy quantum of @xmath0 or @xmath16 before it leaves the scattering region . in principle ,
third or higher orders in the taylor series can be obtained accordingly . however , the higher - order parametric derivatives of the scatter matrix diminish dramatically and approximate zero .
numerical calculation demonstrates that even in relatively large amplitude modulation , their contribution is negligible .
the pumped current depends on the values of the scattering matrix within the energy interval of the order of @xmath17 near the fermi energy . in the low - temperature limit
( @xmath18 ) , an energy interval of @xmath19 is opened during the scattering process .
the mesoscopic scatterer is coupled to two reservoirs with the same temperatures @xmath20 and electrochemical potentials @xmath21 .
electrons with the energy @xmath22 entering the scatterer are described by the fermi distribution function @xmath23 , which approximates a step function at a low temperature . due to the interaction with an oscillating scatterer ,
an electron can absorb or emit energy quanta that changes the distribution function .
a single transverse channel in one of the leads is considered . applying the hypothesis of an instant scattering
, the scattering matrix connecting the incoming and outgoing states can be written as @xmath24 here @xmath25 is an element of the scattering matrix @xmath10 ; the time - dependent operator is @xmath26 , and the energy - dependent operator @xmath27 annihilates particles with total energy e incident from the @xmath28 lead into the scatter and obey the following anticommutation relations @xmath29
= \delta _ { \alpha \beta } \delta \left ( { e - e ' } \right).\ ] ] note that above expressions correspond to single- ( transverse ) channel leads and spinless electrons . for the case of many - channel leads
each lead index ( @xmath30 , @xmath31 , etc . ) includes a transverse channel index and any repeating lead index implies implicitly a summation over all the transverse channels in the lead .
similarly an electron spin can be taken into account . using eqs .
( 2 ) and ( 4 ) and after a fourier transformation we obtain @xmath32 . \\
\end{array}\ ] ] the distribution function for electrons leaving the scatterer through the lead @xmath28 is @xmath33 , where @xmath34 means quantum - mechanical averaging . substituting eq .
( 6 ) we find @xmath35 . \\
\end{array}\ ] ] the distribution function for outgoing carriers is a nonequilibrium distribution function generated by the nonstationary scatterer .
the fourier amplitudes of the scattering matrix @xmath36 ( @xmath37 ) is the probability for an electron entering the scatterer through the lead @xmath31 and leaving the scatterer through the lead @xmath28 to emit ( to absorb ) an energy quantum @xmath0 and @xmath38 ( @xmath39 ) is that of the energy quantum @xmath16 process .
@xmath40 is the probability for the same scattering without the change of an energy with the second - order term @xmath41 much smaller than the zero - order term @xmath42 in weak - modulation limit ( @xmath43 ) and can be omitted therein . using the distribution functions @xmath23 for incoming electrons and @xmath44 for outgoing electrons ,
the pumped current measured at lead @xmath28 reads @xmath45 substituting eqs .
( 7 ) and ( 3 ) we get @xmath46 } . \\ \end{array}\ ] ] quantum pumping properties beyond former theory based on first - order parametric derivative of the scattering matrix are demonstrated in eq .
( 9 ) . by taking higher orders of the fourier spectrum of the scattering matrix into consideration , double @xmath0 energy quantum ( or a @xmath16 energy quantum ) emission ( absorption ) processes coact with single @xmath0 quantum processes .
in the weak - modulation limit , the second term in the right - hand side of eq .
( 9 ) is small , which implies that double @xmath0 quantum processes are weak and therefore not observable .
as the ac driving amplitude is enlarged , this term increases markedly and contribution from double @xmath0 quantum processes takes effect . as a result ,
the dependence of the pumped current on the phase difference between two driving oscillations deviates from sinusoidal and changes from @xmath47 to @xmath48 , which is observed in experiment@xcite .
moreover , the relation between the pumped current and the ac driving amplitude @xmath11 is reshaped .
it is also seen that the linear dependence of the pumped current on the oscillation frequency holds for multi - quanta - related processes . in the next section ,
numerical results of the pumped current in a two - oscillating - potential - barrier modulated nanowire are presented and comparison with experiment is given .
we consider a nanowire modulated by two gate potential barriers with equal width @xmath49 separated by a @xmath50 width well ( see fig .
the electrochemical potential of the two reservoirs @xmath21 is set to be @xmath51 mev according to the resonant level within the double - barrier structure . the two oscillating parameters in eq .
( 1 ) correspond to the two ac driven potential gates @xmath52 with all the other notations correspond accordingly .
we set the static magnitude of the two gate potentials @xmath53 mev and the ac driving amplitude of the modulations equal @xmath54 . in fig .
2 , the dependence of the pumped current on the phase difference between the two ac oscillations is presented . in weak - modulation regime (
namely @xmath55 is small ) , sinusoidal behavior dominates . here , three relatively large @xmath55 is selected to reveal the deviation from the sinusoidal dependence .
( the magnitude of the pumped current mounts up in power - law relation as a function of @xmath55 as shown in fig .
3 . the sinusoidal curve for small @xmath55 would be flat and invisible in the same coordinate range . )
it can be seen from the figure that the @xmath56-@xmath57 relation varies from sinusoidal ( @xmath47 ) to double - sinusoidal ( @xmath58 ) as the ac oscillation amplitude is increased .
the interpretation follows from eq .
the single @xmath0 quantum emission ( absortion ) processes feature a sinusoidal behavior while the @xmath16 quantum emission ( absortion ) processes feature a double - sinusoidal behavior when the fourier index is doubled .
as @xmath55 is increased , double @xmath0 quantum processes gradually parallel and outweigh the single @xmath0 quantum ones .
it is also demonstrated that when the single @xmath0 quantum processes have the effect of @xmath47 dependence , the double @xmath0 quantum processes induce a @xmath59 contribution with a sign flip , which can be understood from the sign change of the derivative of the scattering matrix .
the effect of three- and higher @xmath60 quantum processes is small even for large @xmath61 comparable to @xmath62 .
the experimental observations@xcite as a deviation from the weak - modulation limit are revealed by our theory .
experiment@xcite also discovered that for weak pumping the dependence of the pumped current on the pumping strength obeys a power of 2 relation , as expected from the simple loop - area argument@xcite ; for strong pumping , power of 1 and 1/2 relation is observed beyond former theory .
we presented in fig .
3 the numerical results based on our theory of the @xmath56-@xmath63 relation at a fixed @xmath64 . to demonstrate its power - law dependence , natural logarithm of the variables is applied . from eq .
( 9 ) , it can be seen that for large ac driving amplitude @xmath63 , contribution of double @xmath0 quantum processes ( formulated in the second term on the right hand side of the equation ) causes the @xmath56-@xmath63 relation to deviate from its weak - modulation limit , the latter of which is @xmath65 . for different phase difference between the two ac drivers , the deviation is different . at @xmath66
the pumped current is invariably zero regardless of the order of approximation determined by time - reversal symmetry . at @xmath67 ,
@xmath48 is exact zero , and no difference is incurred by introducing higher order effect .
if we shift the value of @xmath68 to @xmath69 , the abating effect of the double @xmath0 quantum processes has the order of @xmath70 with the small second - order parametric derivative of the scattering matrix smoothing that effect a bit .
consequently , a power of @xmath71 relation is obtained and visualized by our curve fit , which is analogous to experimental findings . for different values of @xmath68 ,
sharper abating and augmental effect occurs with analogous mechanisms .
it is possible that the experiment@xcite was done at the phase difference close to @xmath72 while trying to approach maximal pumped current in the adiabatic and weak - pumping limit .
based on the ac scattering approach , we go further to expand the time - dependent scattering matrix to higher orders of the modulation amplitude and the time .
it is demonstrated in our theory that @xmath73 quantum emission ( absorption ) processes coact with those of single @xmath0 quantum when we go beyond the small - frequency and weak - modulation limit .
nonsinusoidal dependence on the phase difference between two oscillating modulations is incurred by higher order fourier components .
the pumped current versus modulation amplitude relation has a power law of @xmath71 passage with the increase of the oscillating amplitude .
numerical results for a two - ac - gate modulated nanowire interpret experimental findings at large ac driving amplitudes .
the author would like to express sincere appreciation to professor wenji deng , professor jamal berakdar , professor liliana arrachea , and dr .
brian m. walsh for valuable enlightenment to the topic from discussions with them .
along with fits to @xmath74 ( red solid circle ) below 35 mev , @xmath75 ( green upward triangle ) below 41 mev , and @xmath76 above 41 mev ( blue downward triangle ) . to demonstrate its power - law dependence ,
natural logarithm of the variables is applied .
the phase difference between the two ac driver @xmath77 .
inset is the zoom - in of the circled region . ] | transport of matter from low to high absorbing energy pump process driving mechanics of classic pumps concept quantum pump initiated decades ago coherent tunneling quantum interference research on quantum pumping interest since experimental realization open quantum dot current noise properties in quantum pump structures devices investigated magnetic - barrier - modulated dimensional electron gas mesoscopic one - dimensional wire quantum - dot structures mesoscopic rings with aharonov - bohm effect magnetic tunnel junctions tunnel - coupled metallic islands nanoscale helical wire tomonaga - luttinger liquid garphene - based devices theory predicts charge pumped by oscillating one parameter in quantum configurations recent experiment two parallel quantized charge pumps potential application quantum pumping in quantum information processing generation of single photons neural networking quantum standard for electrical current theoretical techniques for treatment quantum pumps prominent scattering approach by brouwer pumped current to derivatives scattering matrix driven by adiabatic weak modulation pumped current phase difference between oscillating potentials increases linearly with frequencyquantum pump discussed little attention to deviation from weak - pumping theory first - order parametric derivative scattering matrix improved scattering approach expanding to higher orders time modulation amplitude further investigation start with scattering matrix approach by moskalets et al response mesoscopic phase - coherent sample to two oscillating external parameters @xmath6 @xmath8 measure static magnitude driving amplitude parameters phase difference between defined as @xmath9 mesoscopic conductor connected to two reservoirs at zero bias scattering matrix @xmath10 function of parameters depends on time external parameter changes slowly instant scattering investigate deviation from small amplitude @xmath11 limit scattering matrix @xmath12 into taylor series of @xmath13 to second order at @xmath7 terms linear quadratic higher orders of fourier spectra enter into scattering matrix nearest sidebands scattered electron can absorb or emit energy quantum of @xmath0 or @xmath16 before scattering region third or higher orders in taylor series can obtainedhigher - order derivatives scatter matrix diminish approximate zero large amplitude modulation contribution negligible pumped current depends on values scattering matrix energy interval @xmath17 near fermi energy low - temperature limit @xmath18 energy interval @xmath19 opened during scattering mesoscopic scatterer coupled to two reservoirs same temperatures @xmath20 electrochemical potentials @xmath21 electrons energy @xmath22 entering scatterer described by fermi distribution function @xmath23 approximates step function at low temperature oscillating scatterer electron absorb emit energy distribution function single transverse channel in considered scattering scattering matrix connecting incoming outgoing states written as @xmath24 @xmath25 element scattering matrix @xmath10 time - dependent operator @xmath26 energy - dependent operator @xmath27 annihilates particles total energy from @xmath28 scatter anticommutation relations @xmath29 = \delta\alpha\beta expressions correspond to single- transverse channel leads spinless electrons many - channel leads each lead index includes transverse channel index implies summation transverse channels electron spin eqs2 ) 4 ) fourier transformation obtain @xmath32 distribution function electrons leaving scatterer @xmath28 @xmath33 @xmath34 quantum - mechanical averaging substituting eq 6 ) find @xmath35 distribution outgoing carriers nonequilibrium nonstationary scatterer fourier amplitudes scattering matrix @xmath36 probability electron entering scatterer @xmath31 leaving @xmath28 emit energy quantum @xmath0 @xmath38 energy quantum @xmath16 process @xmath40 probability same scattering without change energy second - order term @xmath41 smaller than zero - order term @xmath42 weak - modulation limit omitted distribution functions @xmath23 incoming @xmath44 outgoing electrons pumped current lead @xmath28 reads @xmath45 7 ) 3 ) get @xmath46 quantum pumping properties scattering matrix demonstrated in eq 9 higher orders fourier spectrum scattering matrix double @xmath0 emission absorption processes coact with single @xmath0 quantum processes weak - modulation limit second term right - hand side eq9 ) small implies double @xmath0 quantum processes weak not observable ac driving amplitude enlarged term increases contribution @xmath0 processes dependence pumped current on phase difference oscillations deviates changes from @xmath47 to @xmath48 observed experiment@xcite relation pumped current ac driving amplitude @xmath11 reshaped linear dependence on oscillation frequency holds multi - quanta - processes numerical results pumped current two - oscillating potential - barrier modulated nanowire comparison experiment nanowire modulated two potential barriers equal width @xmath49 separated @xmath50 width electrochemical potential two reservoirs @xmath21 set @xmath51 mev two oscillating parameters correspond ac driven potential gates @xmath52 static magnitude gate potentials @xmath53 mev ac driving amplitude equal @xmath54 fig . 2 dependence pumped current on phase difference ac oscillations presented weak - modulation regime @xmath55 small sinusoidal behavior dominates large @xmath55 selected deviation from sinusoidal dependence magnitude pumped current mounts power - law relation function @xmath55sinusoidal curve for small @xmath55 flat invisible same coordinate range @xmath56-@xmath57 relation varies from sinusoidal to double - sinusoidal as oscillation amplitude increased interpretation follows from eq single @xmath0 processes sinusoidal @xmath16 double - sinusoidal when fourier index doubled @xmath55 increased double @xmath0 processes outweigh single @xmath0 single @xmath0 processes effect @xmath47 dependence double induce @xmath59 contribution with sign flip effect of three- higher @xmath60 quantum processes small for large @xmath61 comparable to @xmath62 experimental observations deviation from weak - modulation limit revealed by theory weak pumping dependence pumped current on pumping strength obeys power of 2 relation strong pumping power of 1 and 1/2 relation observed presented in fig . 3 numerical results of @xmath56-@xmath63 relation at fixed @xmath64 power - law dependence natural logarithm of variables appliedlarge ac driving amplitude @xmath63 double @xmath0 quantum processes causes @xmath56-@xmath63 relation deviate from weak - modulation limit @xmath65 different phase difference ac drivers deviation different @xmath66 pumped current zero - @xmath67 @xmath48 exact zero no difference introducing higher order effect shift value @xmath68 to @xmath69 abating effect double @xmath0 processes order @xmath70 small second - order parametric derivative scattering matrix effect power of @xmath71 relation visualized curve fit analogous experimental findings different values @xmath68 sharper abating augmental effect occurs analogous mechanisms experiment@xcite done phase difference close @xmath72 maximal pumped current weak - pumping limit time - dependent scattering matrix higher orders modulation amplitude time @xmath73 quantum emission processes coact with single @xmath0 quantum beyond small - frequency weak - modulation limit nonsinusoidal dependence phase difference oscillating modulations higher order fourier components pumped current versus modulation amplitude relation power law @xmath71 passage increase oscillating amplituderesults two - ac - gate modulated nanowire interpret experimental findings large ac driving amplitudes author appreciation professor wenji deng jamal berakdar liliana arrachea brian walsh enlightenment fits @xmath74 below 35 mev @xmath75 below 41 mev @xmath76 above 41 mev power - law dependence natural logarithm variables applied phase difference two ac driver @xmath77 zoom - in circled region | 0.496874 | 3,040 | 1,533 | , Saving $0.1 in GPT-4. |
20,529 | the observation of two charged bottomonium - like structures @xmath0 and @xmath1 has stimulated extensive studies of the properties of @xmath0 and @xmath1 . in this talk , we briefly introduce the research status of @xmath0 and @xmath1 combined with our theoretical progress . | in the past 8 years , experiments have made big progress on the observations of charmonium - like or bottomonium - like state @xmath2 , @xmath3 , @xmath4 , which have stimulated theorists interest in revealing their underlying structures .
thus , studying charmonium - like and bottomonium - like states is an active and important research field in hadron physics at present , which can further deepen our understanding of the properties of @xmath2 , @xmath3 , @xmath4 and improve our knowledge about non - perturbative qcd .
very recently , the belle collaboration @xcite reported the first observation of two charged bottomonium - like structure , which makes the family of bottomonium - like abundant . by analyzing the @xmath5 ( @xmath6 ) and @xmath7 ( @xmath8 ) invariant mass spectra of @xmath9 hidden - bottom dipion decays , belle observed that there exist two enhancement structures around 10610 mev and 10650 mev , which are named as @xmath0 and @xmath1 , where @xmath0 and @xmath1 are close to the thresholds of @xmath10 and @xmath11 respectively . due to the peculiarities of @xmath0 and @xmath1 ,
theorists have paid more attentions to the observed novel phenomena by different approaches @xcite . in the following
, we will introduce the theoretical progress on the study of @xmath0 and @xmath1 .
before the observation of two @xmath13 structures @xcite , the belle collaboration measured the @xmath14@xmath15@xmath16 , @xmath17 processes near the peak of the @xmath12 resonance at @xmath18 gev @xcite , which indicates that there exist the anomalous @xmath16 and @xmath17 productions , i.e , the branching ratios of @xmath19 and @xmath20 are larger than the dipion - transition rates between the lower members of the @xmath21 family by two orders of magnitude @xcite . for solving this puzzling phenomena ,
rescattering mechanism was proposed in ref .
since @xmath12 is above the threshold of @xmath22 meson pair , the coupled channel effect becomes important , which makes the intermediated hadronic loops constructed by @xmath23 mesons play crucial role to understanding the anomalous @xmath16 and @xmath17 production @xcite . as an alternative explanation , tetraquark state @xmath24[\bar{b}\bar{q}]$ ]
was introduced in ref .
later , authors of ref .
@xcite studied the belle data by analyzing the dipion invariant mass spectrum and the @xmath25 distribution of @xmath26 decays into @xmath16 and @xmath17 , and claimed that the tetraquark interpretation can well describe the anomalous rates observed by belle @xcite .
later , in ref .
@xcite we indicated that the @xmath25 distribution of @xmath27 given by ref .
@xcite is not consistent with the belle data , and proposed an alternative approach to try to explain belle observation of @xmath12 decays into @xmath16 and @xmath17 , where the interference between the direct dipion transition and the final state interaction corresponding to the first and the second diagrams of fig .
[ fig:1 ] , respectively . under this scenario
, we can explain the anomalous rates of @xmath16 and @xmath17 production in @xmath12 decays , especially the inverse rates @xmath28 . besides , the dipion invariant spectrum and the @xmath25 distribution of @xmath29 can be reproduced .
however , the @xmath25 distribution of @xmath30 can not be described by the scenario in ref .
@xcite .
this fact mentioned above shows that a new puzzle appears in the @xmath30 decay , which can not be understood by the tetraquark state picture @xcite or rescattering mechanism @xcite .
thus , we need to consider new mechanism involved in the @xmath30 decay .
since two @xmath13 structures are from the hidden - bottom dipion decays of @xmath12 @xcite , we realized that there exists the extra intermediate @xmath0 and @xmath1 contributions to the @xmath12 decays just depicted by the last two diagrams in fig .
[ fig:1 ] .
thus , in ref .
@xcite we included all mechanisms listed in fig . [ fig:1 ] to redo the analysis of the @xmath20 decay . the results presented in fig .
[ fig:2 ] indicate that including the intermediate @xmath0 and @xmath1 contribution can produce the @xmath25 distribution of @xmath20 consistent with the experimental data @xcite .
this observation provides an indirect evidence to the existence of two charged @xmath13 structures , and gives a possible approach to solve the puzzles existing in the @xmath12 hidden - bottom dipion decays .
however , we must find the source to generate the @xmath0 and @xmath1 structures . in the following
, we introduce the exotic state explanations to @xmath0 and @xmath1 .
since the charged @xmath0 and @xmath1 are close to the @xmath10 and @xmath11 thresholds , respectively , @xmath0 and @xmath1 can be as good candidate of exotic state . before the observation of these two charged bottomonium - like states , the analysis in refs .
@xcite indicates that there probably exists a loosely bound s - wave @xmath10 molecular state , where the one - boson - exchange ( obe ) model is applied to the dynamical calculation . for further understanding the structure of @xmath0 and @xmath1 ,
different theoretical groups have performed the study of @xmath0 and @xmath1 considering different exotic state assignments to @xmath0 and @xmath1 , which mainly include molecular states composed of @xmath31 mesons @xcite , and hidden - bottom tetraquark states @xcite . by the obe model
, we systematically calculated the interaction between @xmath32 and @xmath33 mesons , where @xmath34 , @xmath35 , @xmath36 , @xmath37 meson exchanges are introduced when deducing the effective potential of @xmath38 molecular systems .
if @xmath0 and @xmath1 are @xmath10 and @xmath11 molecular states respectively , we can construct their flavor wave functions @xcite , @xmath39 and @xmath40 $ ] for @xmath0 , @xmath41 and @xmath42 for @xmath1 . under the breit approximation
, the effective potential of the @xmath10 and @xmath11 systems can be related to the scattering amplitude @xmath43 where @xmath44 and @xmath45 denote the masses of the initial and final states respectively .
thus , we obtain the potential in the coordinate space @xmath46 by performing the fourier transformation @xmath47 where the monopole form factor ( ff ) @xmath48 is introduced , which reflects the structure effect of the vertex of the heavy mesons interacting with the light mesons . @xmath49 denotes the exchange meson mass .
we consider both s - wave and d - wave interactions between @xmath32 and @xmath33 mesons , which make the @xmath50 and @xmath51 to be further expressed as @xmath52\big\rangle=\left ( \begin{array}{c } \big|b^*\bar{b}^*(^1s_0)\big\rangle\\ \big|b^*\bar{b}^*(^5d_0)\big\rangle , \end{array}\right ) \nonumber\\ \big|z_{b^*\bar{b}^*}^{(\alpha)}[\mathrm{1}]\big\rangle&=&\left ( \begin{array}{c } \big|b^*\bar{b}^*(^3s_1)\big\rangle\\ \big|b^*\bar{b}^*(^3d_1)\big\rangle\\ \big|b^*\bar{b}^*(^5d_1)\big\rangle\end{array}\right ) , \
, \big|z_{b^*\bar{b}^*}^{(\alpha)}[\mathrm{2}]\big\rangle=\left ( \begin{array}{c } \big|b^*\bar{b}^*(^5s_2)\big\rangle\\ \big|b^*\bar{b}^*(^1d_2)\big\rangle\\ \big|b^*\bar{b}^*(^3d_2)\big\rangle\\ \big|b^*\bar{b}^*(^5d_2)\big\rangle\end{array}\right ) \nonumber\\ \label{11}\end{aligned}\ ] ] with @xmath53 , where we use the notation @xmath54 to denote the total spin @xmath55 , angular momentum @xmath56 , total angular momentum @xmath57 of the @xmath10 or @xmath11 system .
indices @xmath55 and @xmath58 indicate that the couplings between @xmath59 and @xmath60 occur via the @xmath55-wave and @xmath58-wave interactions , respectively . in general , the total effective potentials of the @xmath10 and @xmath11 systems are @xmath61}&=&\big\langle { z_{b^*\bar{b}^*}^{(\alpha)}}[\mathrm{j}]\big)\big| \sum_{e=\mathrm{\pi,\eta,\sigma,\rho,\omega}}\mathcal{v}_e^{b^*\bar{b}^*}(r)\big|{z_{b^*\bar{b}^*}^{(\alpha)}}[\mathrm{j}]\big\rangle,\end{aligned}\ ] ] which are @xmath62 and @xmath63 matrices , respectively ( see ref .
@xcite for more details ) .
[ tab:1 ] llllllll state & @xmath64 ( gev ) & @xmath65 ( mev ) & @xmath66 ( fm)&state & @xmath64 ( gev ) & @xmath65 ( mev ) & @xmath66 ( fm ) + & 2.1&-0.22&3.05 & & 2.2&-0.81&1.38 + @xmath67&2.3&-1.64&1.31 & @xmath1&2.4&-3.31&0.95 + & 2.5&-4.74&0.84 & & 2.8&-14.94&0.52 + the numerical result listed in table [ tab:1 ] indicates that @xmath68 and @xmath69 can be explained as the @xmath70 and @xmath51 molecular states @xcite .
besides explaining @xmath0 and @xmath1 under the exotic state assignments , we also explored whether there exist other mechanisms resulting in @xmath0 and @xmath1 enhancement structures , which is an interesting research topic .
since the observation of the @xmath0 and @xmath1 structures is from the hidden - bottom dipion decays of @xmath12 , studying the decay mechanisms existing in the @xmath12 decays can provide some hints to understand @xmath0 and @xmath1 .
just presented in sect .
[ sec:2 ] , there are at least three different decay mechanisms just shown in fig .
[ fig:1 ] . apart from the first two mechanisms corresponding to the first two diagrams of fig .
[ fig:1 ] , we noticed that the last two diagrams in fig .
[ fig:1 ] can be naturally replacements as the diagrams in fig .
[ fig:3 ] , where the emitted single pion directly from @xmath12 makes the intermediate @xmath38 meson pair with low momenta , which interact with each other to transit into final states by exchanging @xmath32 mesons .
this decay mechanism is named as the initial single pion emission ( ispe ) mechanism in ref .
@xcite . by the ispe mechanism @xcite
, we studied the line shapes of @xmath71 ( @xmath6 ) and @xmath72 ( @xmath8 ) .
we found sharp structures around 10610 mev and 10650 mev in the obtained theoretical line shapes of @xmath71 and @xmath72 distributions , which could naturally correspond to the @xmath0 and @xmath1 structures newly observed by belle @xcite . in fig .
[ fig:4 ] , we shown the numerical result in the scenario of ispe . if the ispe mechanism is a universal mechanism existing the decay of heavy flavor quarkonia , we extend this mechanism to study the hidden - charm dipion decays of higher charmonia @xcite . we predicted the charged charmonium - like structures around the @xmath73 and @xmath74 thresholds . since these novel phenomena are accessible by the besiii , belle , babar experiments , and belle - ii or superb , we suggest future experiments to carry out the search for these charged charmonium - like structures .
@xmath75@xmath76 + + @xmath77@xmath78 +
recently two charged bottomonium - like structures @xmath0 and @xmath1 were announced by the belle collaboration @xcite , which inspired theorists interest in understanding their structure and underlying mechanism behind these novel phenomena . at present , different explanations to @xmath0 and @xmath1 were proposed due to the peculiarities of @xmath0 and @xmath1 .
we briefly introduce the research progress on the study of @xmath0 and @xmath1 .
especially , we emphasize our theoretical work . in ref .
@xcite , we indicated that two @xmath13 structures are important to understand the anomalous @xmath25 distribution of @xmath20 .
@xmath10 and @xmath11 molecular state assignments provide the possible explanation to the structure of @xmath0 and @xmath1 .
in addition , we found the ispe mechanism existing @xmath12 hidden - bottom dipion decays @xcite , which results in the enhancement structures around the @xmath10 and @xmath11 thresholds .
this observation also provides a possibility to understand @xmath0 and @xmath1 .
what is more important it that we also predicted charged charmonium - like structures close to the @xmath73 and @xmath74 thresholds .
the further experimental search for these structures will be helpful to test the ispe mechanism proposed in ref .
@xcite . in conclusion
, we still need to pay more theoretical and experimental efforts to reveal the underlying mechanism relevant to the charged bottomonium - like structures @xmath0 and @xmath1 .
x.l . would like to thank the organizers of apfb2011 conference , especially prof .
hyun - chul kim for his warm invitation . x.l . also thank lanzhou university for financial support ( from the 985 project ) .
we enjoy the collaboration with prof .
shi - lin zhu , prof .
xue - qian li , dr . jun he , zhi - gang luo and zhi - feng sun .
this project is partly supported by the national natural science foundation of china under grants no .
11175073 , no 11005129 , no . 11035006 , no .
11047606 ; the ministry of education of china ( fanedd under grant no . 200924 , dpfihe under grant no .
20090211120029 , ncet under grant no .
ncet-10 - 0442 , the fundamental research funds for the central universities ) ; and the west doctoral project of chinese academy of sciences .
ali a. , hambrock c. , aslam m.j . : a tetraquark interpretation of the belle data on the anomalous @xmath85 and @xmath86 production near the @xmath12 resonance .
phys . rev .
lett . * 104 * , 162001 ( 2010 ) [ erratum - ibid . *
107 * , 049903 ( 2011 ) ] | past 8 years experiments progress observations charmonium - state @xmath2 stimulated theorists interest revealing underlying structures studying charmonium bottomonium - states active important research field hadron physics deepen understanding properties @xmath2 improve knowledge non - perturbative qcd recently belle collaboration reported first observation two charged bottomonium - like structure family bottomonium abundant analyzing @xmath5 invariant mass spectra @xmath9 hidden - bottom dipion decays observed two enhancement structures around 10610 mev 10650 mev @xmath0 and @xmath1 close to thresholds @xmath10 @xmath11 peculiarities @xmath0 theorists paid attentions observed novel phenomena different approaches introduce theoretical progress study @xmath0 @xmath1 before observation two @xmath13 structures belle collaboration measured @xmath14 processes near peak @xmath12 resonance at @xmath18 gev indicates anomalous @xmath16 @xmath17 productionsbranching ratios of @xmath19 @xmath20 larger than dipion - transition rates @xmath21 family two orders rescattering mechanism proposed @xmath12 above threshold @xmath22 meson pair coupled channel effect important intermediated hadronic loops by @xmath23 mesons anomalous @xmath16 @xmath17 production alternative explanation tetraquark state @xmath24 introduced authors studied belle data dipion invariant mass spectrum @xmath25 distribution of @xmath26 decays into claimed tetraquark interpretation describe anomalous rates indicated @xmath25 distribution of @xmath27 not consistent with belle data proposed alternative approach explain belle observation @xmath12 decays into interference between direct dipion transition final state interaction first second diagrams explain anomalous rates of @xmath16 @xmath17 production in @xmath12 decays especially inverse rates @xmath28 dipion invariant spectrum @xmath25 distribution of @xmath29 reproduced @xmath25 distribution of @xmath30 not described by scenario ref @xcitefact shows new puzzle in @xmath30 decay understood by tetraquark state picture or mechanism need consider new mechanism decay two @xmath13 structures from hidden - bottom dipion decays of @xmath12 extra intermediate @xmath0 and @xmath1 contributions to @xmath12 decays diagrams in included all mechanisms in analysis @xmath20 decay results:2 indicate intermediate @xmath0 and @xmath1 contribution @xmath25 distribution of @xmath20 consistent with experimental data indirect evidence two charged @xmath13 structures possible approach to solve puzzles in @xmath12 decays must find source to generate @xmath0 and @xmath1 structures introduce exotic state explanations to @xmath0 and @xmath1 charged @xmath0 close to @xmath10 and @xmath11 thresholds good candidate of exotic state before states analysis in indicates loosely bound s - wave @xmath10 molecular state one - boson - exchange model applied to dynamical calculationunderstanding structure @xmath0 @xmath1 theoretical groups exotic state assignments molecular states @xmath31 mesons hidden bottom tetraquark states model calculated interaction @xmath32 @xmath33 mesons @xmath34 meson exchanges introduced effective potential @xmath38 molecular systems @xmath0 @xmath1 @xmath10 @xmath11 molecular states construct flavor wave functions @xcite @xmath39 @xmath40 @xmath0 @xmath41 @xmath42 @xmath1 effective potential @xmath10 @xmath11 systems related scattering amplitude @xmath43 @xmath44 @xmath45 denote masses initial final states obtain potential coordinate space @xmath46 fourier transformation @xmath47 monopole form factor @xmath48 introduced reflects structure effect heavy mesons interacting light mesons @xmath49 denotes exchange meson massconsider s d - wave interactions @xmath32 @xmath33 mesons @xmath50 @xmath51 expressed @xmath52\big\rangle\left|b^*\bar{b}^*(^1s_0)^*\bar{b}^*(^5d_0){array{b^*\bar{b}^*}^{\alpha)\rangle\left|b^*\bar{b}^*(^3s_1)^*\bar{b}^*(^3d_1)^*\bar{b}^*(^5d_1){array_{b^*\bar{b}^*}^{\alpha)}{2}\rangle\left|b^*\bar{b}^*(^5s_2)^*\bar{b}^*(^1d_2)^*\bar{b}^*(^3d_2)^*\bar{b}^*(^5d_2){array{aligned @xmath53 notation @xmath54 total spin @xmath55 angular moment@xmath56 total angular momentum @xmath57 @xmath10 @xmath11 system indices @xmath55 @xmath58 indicate couplings @xmath59 @xmath60 @xmath55 @xmath58-wave interactions total effective potentials @xmath10 @xmath11 systems @xmath61}\big\langle^*\bar\alpha\mathrm\mathrm{\pi\eta\sigma\rho\omega}}\mathcal{v^*\bar\rangle{aligned @xmath62 @xmath63 matrices see @xcite details tab:1 @xmath64 ( gev @xmath65 mev ) @xmath66 ( fm 2.1&-0.22&3.05 2.2&-0.81&1.38 @xmath67&2.3&-1.64&1.31 @xmath1&2.4&-3.31&0.95 2.5&-4.74&0.84 2.8&-14.94&0.52 numerical result table indicates @xmath68 @xmath69 explained @xmath70 @xmath51 molecular statesexplaining @xmath0 @xmath1 under exotic state assignments explored other mechanisms enhancement structures interesting research topic observation @xmath0 @xmath1 structures from hidden - bottom dipion decays @xmath12 studying decay mechanisms in @xmath12 decays understand @xmath0 @xmath1 three decay mechanisms shown in fig fig:1 first diagrams last two diagrams replacements diagrams:3 emitted single pion from @xmath12 makes intermediate @xmath38 meson pair low momenta transit final states exchanging @xmath32 mesons decay mechanism initial single pion emission ( ispe ) mechanism in @xcite studied line shapes of @xmath71 @xmath72 found sharp structures around 10610 mev 10650 mev in theoretical line shapes distributions correspond to @xmath0 @xmath1 structures observed by @xcite fig :4 numerical result in scenario of ispe if ispe mechanism universal mechanism decay of heavy flavor quarkonia extend mechanism to study hidden - charm dipion decays of higher charmonia @xcitepredicted charged charmonium - structures around @xmath73 thresholds novel phenomena accessible by besiii belle babar experiments belle - superb suggest future experiments for structures @xmath75 + two charged bottomonium - like structures @xmath0 @xmath1 announced by belle collaboration inspired theorists understanding structure underlying mechanism different explanations to @xmath0 @xmath1 proposed peculiarities research progress emphasize theoretical work two @xmath13 structures important understand anomalous @xmath25 distribution of @xmath20 @xmath10 @xmath11 molecular state assignments provide possible explanation structure @xmath0 found ispe mechanism @xmath12 hidden - bottom dipion decays results in enhancement structures around @xmath10 thresholds possibility understand @xmath0 @xmath1 predicted charged charmonium - like structures close to @xmath73 @xmath74 thresholds further experimental search for structures test ispe mechanism need more theoretical experimental efforts to reveal underlying mechanism charged bottomonium - like structures @xmath0 @xmath1 thank organizers apfb2011 conference prof .hyun - chul kim invitation lanzhou university financial support 985 project collaboration prof shi - lin zhu xue - qian li jun he zhi - gang luo zhi - feng sun project supported national natural science foundation china grants 11175073 11005129 11035006 11047606 ministry education china grant 200924 20090211120029 - 0442 fundamental research funds central universities west doctoral project chinese academy sciences ali a. hambrock c. aslam m.j tetraquark interpretation data anomalous @xmath85 @xmath86 production @xmath12 resonance . rev . lett 104 162001 2010 107 049903 2011 | 0.553372 | 3,557 | 1,850 | , Saving $0.1 in GPT-4. |
20,532 | we present our preliminary bima co(1 - 0 ) images of ii zw 96 which show huge molecular gas concentrations outside the merging disks .
the dominant extra - disk co concentrations in ii zw 96 correspond to the two star - forming `` knots '' hidden by dust , whereas the others are associated with the two spiral disks .
the unique huge co concentrations , @xmath0 kpc away from the merging nuclei , account for @xmath1 of the total co emission , yet only at a moderate gas surface density as compared with that of ultraluminous infrared galaxies ( uligs ) .
we tentatively conclude that the huge extra - disk gas concentrations are most likely the overlap region of the gas disks , not necessarily coinciding with the stellar disks overlap region .
the gas disks overlap region could be formed prior to the merging of the stellar disks and become the overlap regions found between the disks like that in arp 244 ( the `` antennae '' galaxies ) .
these overlap regions would eventually nurture the _ extreme starbursts _ and evolve into the extended gas disk surrounding the double nuclei in final merging in most uligs .
# 1_#1 _ # 1_#1 _ = # 1 1.25 in .125 in .25 in | starburst galaxy mergers involve highly dissipative collapse of tremendous amounts of molecular gas from their location in two separate galaxies , into the inner few 100 pc of the merged system .
the high gas densities in the core of the merger prompt powerful star formation ; the dust associated with the gas absorbs the uv / optical light and re - radiates in the far - ir , resulting in an ultraluminous infrared galaxy ( ulig ) . in most uligs , merging is sufficiently advanced that the gas has mostly collapsed into the central a few 100 pc .
notable examples are the nearest uligs like arp 220 ( downes & solomon 1998 ( ds98 ) ; sakamoto et al .
there are , however , a few very highly luminous galaxies with extranuclear star forming regions ( just below the @xmath2 ulig cutoff ) where the gas has apparently not yet collapsed into the nuclei ( e.g. , arp 299 aalto et al . 1997 ; casoli et al . 1999 ; vv 114 yun et al . 1994 ) .
these systems are of crucial importantance as they give us a glimpse into the ( apparently brief ) intermediate stage between well - separated gas - rich pairs and the merged late - stage uligs .
ii zw 96 ( @xmath3 , nuclear separation @xmath4 kpc ) , shows four distinct star - forming regions , as revealed by optical and near - ir observations ( goldader et al .
1997 ( g97 ) ) .
two of those are apparently associated with galaxy nuclei .
the other two sources lie well away from the two nuclei , extremely weak in optical , but prominent in near - ir , implying heavy dust obscuration in these extra - disk star forming regions .
in order to explore the extra - disk star forming regions heavily obscured by dust , we have observed ii zw 96 at co(1 - 0 ) using the 10-element bima array at hat creek , california .
the data presented here are mainly compact c and d array observations plus some limited long baseline b array observations .
the combined data give a synthesized beam of @xmath5 ( @xmath6 kpc ) .
additional b array observations at @xmath7 resolution are being conducted currently to further improve the resolution .
a total h@xmath8 mass of @xmath9 ( standard conversion ) was estimated from a single - dish co survey ( gao & solomon 1999 ) .
our bima observations have recovered most of the single - dish flux and revealed that ii zw 96 is a morphologically complex and extraordinary system with most of co emission located outside the two merging stellar disks ( fig . 1 ) .
the dominant co concentrations correspond to the two extra - disk star - forming regions hidden by heavy dust , whereas the others are associated with the two galaxy disks and the region between them .
the huge co concentrations far away from the galaxy disks account for more than @xmath10 of the total co emission with a h@xmath8 mass of @xmath11 .
huge concentrations of molecular gas in the overlap regions have been observed already in merger systems like arp 299 and the `` antennae '' galaxies arp 244 .
the enormous amount of molecular gas located far away from merging disks makes ii zw 96 quite unique .
moreover , the gas density of @xmath12 ( as seen by a beam resolution of @xmath13 kpc ) at the peak of the concentrations , although at least @xmath14 higher than either nuclear regions of the merging galaxies , is orders of magnitude less than that of uligs and even arp 299 .
the peak gas surface density would still be several times less than that of uligs even if the entire co emission covered by current larger synthesized beam is from the central @xmath15 kpc region .
in addition , the extra - disk gas concentrations appear to be peaked with a small offset from the star forming knots identified in near - ir ( fig . 1 ) and are mostly at higher velocities ( fig . 2 ) .
only a small amount of molecular gas has been observed in se galaxy which appears to have little association with the extra - disk gas concentrations ( fig . 2 ) .
nw galaxy has a rather extended gas distribution , gas rotation motion and connection of the disk gas with the extra - disk co concentration , whereas much less gas is observed in the se disk .
it is difficult to speculate that most of the gas in the extra - disk concentrations could have come from the nw galaxy .
could the se galaxy have been originally very gas - rich with extended gas disk and left most of the gas behind after colliding with the nw galaxy ?
the star forming knots seen in near - ir in the extra - disk gas concentrations show properties of the hii regions as that of the se nuclear region indicating powerful young starbursts ( g97 ) .
these same spectral features are also found in arp 299 and vv 114 . in all these systems ,
the apparent ages of the starbursts are @xmath16830 myr , much younger than the allowed range of uligs .
in fact , goldader et al . (
g97 ) suggested that these three systems might be examples of a short - lived phase through which many
pass on their way to becoming more advanced , centrally concentrated ulig mergers .
because the hi disks are more extended than the stellar disks , they may merge first prior to the merging of the stellar and molecular gas disks when a pair of gas - rich spirals approach each other .
indeed , we may have seen such early merging and formation of the hi disks overlap region in ngc 6670 where huge concentration of hi has been observed between the two galaxies whose hi disks have been highly disrupted , whereas little distortion is observed in the stellar disks ( wang et al .
one plausible idea is that the hi gas disk overlap regions formed much earlier in early mergers ( ngc 6670 ) may be the progenitors of the apparent overlap regions in intermediate mergers ( arp 244 ) , which may evolve into the extended gas disks around the much more concentrated double nuclear gas disks in uligs ( arp 220 , ds98 ) . ii zw 96 fits in between early mergers like ngc 6670 and intermediate mergers like arp 244 . as merging progresses and
the gas overlap region further concentrates , these huge co concentrations will eventually merge with the double nuclear gas disks .
they could thus , be the progenitors of the so called _ extreme starburst _ regions , which dominate the overall starburst power output of the entire system , yet be confined within sub - kpc scale ( ds98 ) .
they may also appear to be the multiple `` nuclei '' claimed in some uligs ( cf .
borne et al .
although there are now several galaxy mergers where huge gas concentrations in the overlap regions the `` interaction zones '' have been observed , ii zw 96 appears to be the only one with most of the gas concentrations ( presumably the overlap regions ) located outside the stellar merging disks . is this unique ?
are there more such systems to be found , especially in the distant universe where galaxy collisions are expected to be much more frequent ?
perhaps , one should not be totally surprised when alma has mapped some huge molecular gas concentrations with little or no optical correspondences , or the peaks of the molecular gas concentrations offset from the optical peaks by several kpc , in merging / star - forming galaxies at high redshift .
aalto , s. , radford , s.j.e . ,
scoville , n.z . , & sargent , a.i .
1997 apj , 475 , l107 borne , k.d . ,
bushouse , h. , lucas , r.a . , & colina , l. 2000 apj , in press casoli , f. , willaime , m .- c . , viallefond , f. , & gerin , m. 1999 a&a , 346 , 663 downes , d. , & solomon , p.m. 1998 apj , 507 , 615 ( ds98 ) gao , y. , & solomon , p.m. 1999 apj , 512 , l99 goldader , j.d .
1997 aj , 113 , 1569 ( g97 ) sakamoto , k. et al .
1999 apj , 514 , 68 wang , w .- h .
2000 apj , in revision yun , m.s . ,
scoville , n.z . , & knop , r.a .
1994 apj , 430 , l109 | starburst galaxy mergers involve collapse of molecular gas from galaxies into inner 100 pc merged system high gas densities in prompt star formation dust absorbs uv light re - radiates in in ultraluminous infrared galaxy ( ulig ) in most uligs merging advanced gas collapsed into central 100 pc examples are arp 220 sakamoto few highly luminous galaxies with extranuclear star forming regions below @xmath2 ulig cutoff where gas not collapsed into nuclei ( arp 299 casoli 1999 vv 114 yun et 1994 these systems give glimpse into intermediate stage between separated gas - rich pairs and merged late - stage uligs ii zw 96 ( @xmath3 nuclear separation @xmath4 kpc shows four star - forming regions optical observations two associated with galaxy nuclei other two sources lie away from nuclei prominent in heavy dust obscuration in extra - disk star forming regions to by observed ii zw 96 at co(1 - 0 ) using 10-element bima array at hat creek , californiadata compact c d array observations limited long baseline b array observations combined data synthesized beam of @xmath5 ( @xmath6 kpc ) additional b array observations at @xmath7 resolution conducted improve resolution total h@xmath8 mass of @xmath9 estimated from single - dish co survey 1999 bima observations recovered single - dish flux ii zw 96 morphologically complex system co emission outside merging stellar disks dominant co concentrations to extra - disk star - forming regions others with galaxy disks co concentrations far from galaxy disks for more than @xmath10 total co emission with h@xmath8 mass of @xmath11 concentrations molecular gas in overlap regions observed in systems like arp 299 arp 244 molecular gas from disks makes ii zw 96 unique gas density of @xmath12 @xmath13 kpc at peak less than uligs arp 299 peak gas surface density times less than uligs co emission from central @xmath15 kpc region extra - disk gas concentrations small offset from star forming knots in near - irat higher velocities fig . 2 ) small molecular gas in se galaxy little association with extra - disk gas concentrations nw galaxy extended gas distribution rotation motion connection disk gas with extra - disk concentration less gas in se disk difficult to speculate gas in extra - disk concentrations from nw galaxy se galaxy originally gas - rich extended gas disk left gas behind after colliding with nw galaxy ? star forming knots in near - ir extra - disk gas concentrations show properties hii regions se nuclear region powerful young starbursts spectral features in arp 299 vv 114 apparent ages starbursts younger than allowed range uligs goldader systems short - lived phase advanced centrally concentrated ulig mergers hi disks more extended stellar may merge first prior to merging stellar molecular gas disks gas - rich spirals early merging hi disks overlap region in ngc 6670 huge concentration of hi between two galaxies hi disks disrupted little distortion in stellar disksplausible idea hi gas disk overlap regions formed in early mergers ngc 6670 ) progenitors overlap regions intermediate mergers ( arp 244 ) evolve into extended gas disks around concentrated double nuclear gas disks in uligs ( arp 220 , ds98 ) ii zw 96 between early ngc 6670 and intermediate mergers arp 244 merging gas overlap concentrates co concentrations merge with double nuclear gas disks could be progenitors of extreme starburst regions dominate starburst power output confined within sub - kpc scale ( ds98 ) may multiple `` nuclei '' in some uligs several galaxy mergers huge gas concentrations in overlap regions observed ii zw 96 only one with gas concentrations outside stellar merging disks unique ? more such systems distant universe galaxy collisions frequent ? alma mapped huge molecular gas concentrations with little optical correspondences offset several kpc in merging star - forming galaxies at high redshift aalto radford scoville sargent 1997 apj 475 l107 bushouse lucas colina2000 apj press casoli willaime viallefond gerin 1999 a&a 346 , 663 downes solomon p.m 1998 apj , 615 ds98 ) gao solomon.m 1999 apj 512 l99 goldader 1997 113, 1569 g97 ) sakamoto 1999 apj 514 , 68 wang 2000 apj revision yun scoville n knop 1994 apj 430 , l109 | 0.312041 | 1,853 | 914 | , Saving $0.1 in GPT-4. |
20,535 | we consider perturbations of the complex quadratic map @xmath0 and corresponding changes in their quasi - mandelbrot sets . depending on particular perturbation , visual forms of quasi - mandelbrot set changes either sharply ( when the perturbation reaches some critical value ) or continuously . in the latter case
we have a smooth transition from the classical form of the set to some forms , constructed from mostly linear structures , as it is typical for two - dimensional real number dynamics .
two examples of continuous evolution of the quasi - mandelbrot set are described .
complex dynamics is a well - known source of beautiful images which bring this branch of pure mathematics into attention of wide groups of non - mathematicians , and provide interesting connections between math and art ( see , for example @xcite and references therein ) . on the other hand ,
the situation with real two - dimensional dynamics is more complicated .
there are many examples of artistic interest ( @xcite ) , however , they may be considered as an `` island '' in a wide `` sea '' of other systems without any esthetic value .
it is also remarkable that visual patterns given by complex and real dynamics are quite different .
mostly based on round structures , typical julia sets of complex maps are sometimes compared with `` baroque architecture '' , while analogs of julia sets and basins of attraction in the case of real maps usually consist of quasi - parallel bands , long tongues , and other linear structures .
however , from the mathematical viewpoint one - dimensional complex- and two - dimensional real maps are not completely different entities : each 1-dim complex map can be considered as 2-dim real map by , for example , separating real and imaginary parts .
for example , the simplest nonlinear complex map ( and the most famous one ) @xmath1 is equivalent to a real map @xmath2 @xmath3 where @xmath4 and @xmath5 are the real and imaginary parts of complex parameter @xmath6 . in general complex analytic maps
form a particular subset of two - dimensional real maps which satisfy the cauchy - riemann conditions .
the goal of our study is to understand what happens in a zone located in some sense `` near '' this subset .
we consider one - parameter family of maps having a complex analytic structure for a particular value of the parameter , and then study properties of the map when the parameter is close , but is not equal to this particular value .
pure mathematical aspects of this problem have been studied in a number of papers @xcite , though a general picture is still far from being well understood . in the present paper
we address more informal question of visual appearance of sets created by such maps , and try to see a transition from pure complex visual forms to those typical to a real dynamics .
our starting point is the quadratic map ( 1 ) . instead of fixing @xmath6 to consider a sequence of corresponding julia sets we have chosen the mandelbrot set as the pattern to perturb .
there are two reason for this .
one is technical simplicity of this set to calculate , the second is the wide - spread popularity of the mandelbrot set and its role in bringing mathematical achievements to a more general public .
there two definition of the mandelbrot set which are equivalent for complex analytic maps .
first , in is the set of @xmath6 for which the julia set of the map ( 1 ) is connected .
second , it is the set of @xmath6 for which the orbit of the zero - point of the map ( 1 ) does not escape to infinity .
we take the second definition to calculate the ( quasi)-mandelbrot set for real maps , using the name `` connectedness locus '' for the set created according to the first definition .
we leave studies of the connectedness loci as well as julia sets for real perturbations of complex maps for a future work .
to start our investigations we should fix the perturbations imposed on the quadratic map .
some maps do not show continuous transitions to visual patterns of real dynamics , as it can be seen by close look on the map @xmath7 @xmath8 studied in @xcite . for @xmath9 this system is equivalent to the map ( 1 ) with the following redefinition of variables : @xmath10 , @xmath11 , @xmath12 , @xmath13 . in order to understand what happens for other values of the parameter @xmath14
we remind a reader some simple facts from general theory of two - dimensional real algebras ( see , for example , @xcite ) . to define a particular algebra of this type
it is enough to fix the square of imaginary unit : @xmath15 , where @xmath16 and @xmath17 are real numbers .
there are three qualitatively different possibilities : * if @xmath18 the resulting algebra is isomorphic to the algebra of complex numbers with the imaginary unit @xmath19 , and the isomorphism has the form @xmath20 . *
if @xmath21 the resulting algebra is isomorphic to the algebra of dual numbers with the condition @xmath22 .
* if @xmath23 the resulting algebra is isomorphic to the algebra of perplex numbers with the condition @xmath24 .
sometimes the perplex numbers are called as double numbers , split - complex numbers or hyperbolic numbers ( in the latter case two other number systems are correspondingly elliptic and parabolic numbers ) .
the map ( 4 - 5 ) is specially constructed so that it coincides with the initial quadratic map ( 1 ) for perturbed two - dimensional algebra . if @xmath25 this algebra is isomorphic to the algebra of complex numbers , and the mandelbrot set for the map ( 4 - 5 ) is nothing else but the `` classical '' mandelbrot set transformed according to corresponding linear transformation . for @xmath26 an abrupt transition occurs , and we obtain and infinite band an analog of the mandelbrot set for dual numbers . for smaller @xmath6 the result is a rectangular representing this set for the algebra of perplex numbers .
the abrupt nature of transition for @xmath26 makes the map ( 4 - 5 ) useless for our present study .
the map ( 4 - 5 ) represents a rather peculiar example , and the other known map , studied by peckhman in @xcite gives us a success .
this map conserves the meaning of @xmath27 as a complex variable with real part @xmath28 and imaginary part @xmath29 , however , it violates the cauchy - riemann conditions by adding a linear term depending on @xmath30-conjugate : @xmath31 or , in real variables @xmath32 @xmath33 in figs . 1 6 we present several pictures of a part of quasi - mandelbrot set boundary in the lower - right quadrant for @xmath16 from @xmath34 to @xmath35 with the step equal to @xmath34 .
we can see clearly how the deviation from the famous structure of the mandelbrot set , touching the set at it outer range for @xmath36 , then goes deep inside and transform it into the structure typical for real maps .
at @xmath37 almost all structure remaining from complex origin of the map under investigation is washed out .
qualitatively similar patterns appear for negative @xmath16 in the same range of @xmath38 , so we do not present corresponding pictures here .
figure 1 : a part of the `` original '' mandelbrot set boundary with no perturbation .
figure 2 : a fragment of quasi - mandelbrot set boundary with @xmath36 .
the distortion have touched an outer part of a bulb .
figure 3 : the same boundary for @xmath39 .
linear structures have been formed in the outer part of the bulb .
figure 4 : the same boundary for @xmath40 .
the bulb have been almost completely destroyed .
figure 5 : the same boundary for @xmath41 .
only remnants of the non - perturbed structures are visible .
figure 6 : the same boundary for @xmath37 .
almost all complex analytic structure have been washed out .
finally , we consider a family of maps which , to our knowledge , is not studied previously .
it is possibly the simplest way to disturb the map ( 2 - 3 ) , though from the viewpoint of complex dynamics it looks crazy .
this map has the form @xmath42 @xmath43 and represents a complex analytic map for @xmath44 .
the first equation of the real decomposition of the complex quadratic map ( 2 ) remains unchanged .
we start with an auxiliary map which has the same equation ( 9 ) for the @xmath28-component and represent the quadratic map for a disturbed algebra with @xmath45 .
it has the following equation for @xmath29 @xmath46 the properties of the map ( 9,11 ) is similar to those for ( 4,5 ) , and the algebra becomes isomorphic to perplex numbers for @xmath47 showing a sharp change of its quasi - mandelbrot set for @xmath48 . on the other side , positive @xmath17
keep the considered algebra within the class of complex numbers .
the map ( 9,10 ) is not equivalent to ( 9,11 ) , though , as the fate of a particular trajectory is typically determined by its behavior at the range of variables @xmath28 and @xmath29 of the order of unity , we may expect that it should be a remarkable change of dynamics for @xmath49 being somewhere in the interval @xmath50 .
our numerical experiment confirms these considerations .
we find qualitative changes in the shape of the quasi - mandelbrot set for @xmath49 less than unity . in figs .
7 10 some of `` snapshots '' of the boundary are presented . at the final picture created for @xmath51
we can see a typical pattern of a real number dynamics .
it is also remarkable that unlike previous case of the map ( 6 ) , the real - like features of the map(9,10 ) do not propagate from the outer part of the set , but steadily develop in the whole range of the boundary structure , which becomes more and more elongated with decreasing @xmath49 . on the other hand , for @xmath52
the change is rather slow and not so interesting from the viewpoint of visual impression .
even for @xmath53 some structure remaining from the original mandelbrot set still survives , though linear structures are well developed for this value of @xmath49 .
figure 7 : a part of quasi - mandelbrot set boundary for the map ( 9 ) ( 10 ) with @xmath54 .
the whole structure is slightly disturbed .
figure 8 : the same boundary for @xmath55 .
bulbs have been deformed , but not destroyed .
figure 9 : the same boundary for @xmath56 .
bulbs have been deformed into quasi - linear structures .
figure 10 : the same boundary for @xmath51 .
bulbs became elements of linear structures .
in the present paper we have studied visual appearance of quasi - mandelbrot sets for perturbed quadratic complex map .
two different cases can be distinguished . in a particular case when a perturbed map is equivalent to the initial one for a perturbed two - dimensional real algebra , the form of the quasi - mandelbrot set change sharply reflecting change in the isomorphic class of the algebra .
for general perturbations , however , we expect a smooth transition from complex to real visual patterns .
two studied in the present paper maps show this transition clearly . our present work can be pushed further in several ways
. continuous nature of transition can be used in constructing `` movies '' showing how one fractal structure is replaced by another , which may make the beauty of fractals dynamical .
another possible application of the transition range of parameters found in the present paper is creation of generalized julia sets , which can provide interesting changes in their visual patterns .
other possible directions include non - polynomial perturbations of complex analytical maps which have already been studied in mathematical literature @xcite . | author is grateful to julien clinton sprott for discussions . 99 h.o.peitgen and p.h.richeter , _ the beauty of fractals : images of complex dynamical systems _ , springer - verlag , new york ( 1986 ) .
j.c.sprott and c.a.pickover , comput . and graphics * 18 * , 309 ( 1995 ) .
m. michelitsch and o. rssler , comput . and graphics * 16 * , 435 ( 1992 ) .
b. bielefeld , s.sutherland , f. tangerman and j. veerman , exp . math . * 2 * , 281 ( 1993 ) . o.isaeva and s.kuznetsov `` on possibility of realization of the mandelbrot set in coupled continuous system '' [ arxiv : nlin/0509013 ] .
i. yaglom , _ complex numbers in geometry _ , academic press , new york ( 1968 ) .
b.peckham , int .
j. bifurcation and chaos * 8 * , 73 ( 1998 ) .
b.peckham , int .
j. bifurcation and chaos * 10 * , 391 ( 2000 ) . | grateful julien clinton sprott discussions h.o.peitgen p.h.richeter beauty fractals complex dynamical systems springer - verlag new york 1986 ) j.c.sprott c.a.pickover comput 309 1995 ) m michelitsch o rssler comput graphics 435 1992 ) bielefeld s.sutherland f tangerman j. veerman math 281 1993 ) o.isaeva s.kuznetsov realization mandelbrot set coupled continuous system nlin/0509013 yaglom complex numbers geometry academic press new york 1968 ) b.peckham bifurcation chaos 73 ( 1998 ) 391 ( 2000 ) | 0.070138 | 279 | 152 | , Saving $0.0 in GPT-4. |
20,539 | the capabilities of the cms experiment to study the low-@xmath0 parton structure and qcd evolution in the proton and the nucleus at lhc energies are presented through four different measurements , to be carried out in pb - pb at @xmath1 = 5.5 tev : ( i ) the charged hadron rapidity density @xmath2 and ( ii ) the ultraperipheral ( photo)production of @xmath3 ; and in p - p at @xmath4 = 14 tev : ( iii ) inclusive forward jets and ( iv ) mueller - navelet dijets ( separated by @xmath5 8) . | at high energies , the cross - sections of all _ hadronic objects _ ( protons , nuclei , or even photons `` fluctuating '' into @xmath6 vector states ) are dominated by scatterings involving gluons .
gluons clearly outnumber quarks in the small momentum fraction ( low-@xmath0 ) range of the parton distribution functions ( pdfs ) as a consequence of the qcd parton splitting probabilities described by the dglap @xcite and bfkl @xcite evolution equations .
the fast growth of the gluon densities @xmath7 for decreasing @xmath0 conspicuously observed in dis @xmath8 at hera @xcite , can not however continue indefinitely since this would violate unitarity even for scatterings with @xmath9 . for small enough @xmath0 values
, gluons must start to recombine in a process known as gluon saturation @xcite .
this phenomenon occurs when the size occupied by the partons becomes similar to the size of the hadron @xmath10 , or in terms of the _ saturation momentum _ @xmath11
when : @xmath12 .
@xmath11 grows with the number @xmath13 of nucleons in the `` target '' , the collision energy @xmath4 , and the rapidity of the gluon @xmath14 , according to : @xmath15 .
the @xmath13 dependence implies that , at equal energies , saturation effects will be enhanced by factors as large as @xmath16 6 in a heavy nucleus ( @xmath13 = 208 for pb ) compared to protons .
theoretically , the regime of low-@xmath0 qcd can be effectively described in the `` color glass condensate '' ( cgc ) framework , where all gluon fusions and multiple scatterings are `` resummed '' into classical high - density gluon wavefunctions @xcite .
the corresponding evolution is given in this case by the bk / jimwlk @xcite _ non - linear _ equations .
+ experimentally , the most direct way to access the low-@xmath0 pdfs in hadronic collisions is by measuring perturbative probes ( heavy-@xmath17 , jets , high-@xmath18 hadrons , prompt @xmath19 , ... ) at large @xmath4 and _ forward _ rapidities @xcite . for a @xmath20 parton scattering ,
the _ minimum _ @xmath0 probed in a process with a particle of momentum @xmath18 produced at pseudo - rapidity @xmath21 , is @xmath22 where @xmath23 .
thus , @xmath24 decreases by a factor of @xmath2510 every 2 units of rapidity .
the experimental capabilities of the cms experiment are extremely well adapted for the study of low-@xmath0 phenomena with proton and ion beams .
the acceptance of the cms / totem system is the largest ever available in a collider , and the detector is designed to measure different particles with excellent momentum resolution @xcite : jets @xmath26 , @xmath19 and @xmath27 ( @xmath28 3 ) , muons ( @xmath28 2.5 ) , hadrons ( @xmath28 6.6 ) , plus neutrals in the zero - degree calorimeters ( zdcs , @xmath29 8.3 ) .
we present a selection of four observables measurable in cms which are sensitive to parton saturation effects in the proton and nucleus wave - functions at lhc energies .
other relevant measurements ( e.g. forward drell - yan in p - p at 14 tev ) are discussed in @xcite .
in high - energy heavy - ion collisions , the _ hadron _ rapidity density @xmath31 is directly related to the number of initially released _ partons _ at a given @xmath21 .
cgc approaches which effectively take into account a reduced initial parton flux in the nuclear pdfs , reproduce successfully the absolute hadron yields ( as well as their centrality and @xmath1 dependences ) at sps rhic energies @xcite . at lhc ,
the expected pbpb multiplicities are @xmath32 2000 ( fig .
[ fig : dndeta ] , left ) .
cms simulation studies from hit counting in the innermost si pixel layer ( @xmath28 2.5 ) indicate that the occupancy remains less than 2% and that , on an event - by - event basis , the reconstructed @xmath2 is within @xmath252% of the true primary multiplicity ( fig .
[ fig : dndeta ] , right ) @xcite .
-0.4 cm in central pbpb at the lhc @xcite .
right : range of particle rapidities covered by cms ( si tracker , hf , castor ) and totem ( t1 , t2 trackers ) .
the hijing pbpb distribution of primary simulated tracks within @xmath28 2.5 ( black dots ) is compared to the reconstructed hits in the first layer of the si tracker ( red crosses ) @xcite.,title="fig:",width=283 ] in central pbpb at the lhc @xcite .
right : range of particle rapidities covered by cms ( si tracker , hf , castor ) and totem ( t1 , t2 trackers ) .
the hijing pbpb distribution of primary simulated tracks within @xmath28 2.5 ( black dots ) is compared to the reconstructed hits in the first layer of the si tracker ( red crosses ) @xcite.,title="fig:",width=328,height=249 ] -0.4 cm ultraperipheral collisions ( upcs ) of heavy ions generate strong electromagnetic fields ( equivalent to a flux of quasi - real photons ) which can be used to study @xmath7 via @xmath36 photoproduction @xcite .
lead beams at 2.75 tev have lorentz factors @xmath19 = 2930 leading to maximum photon energies @xmath37 100 gev ( for a nuclear radius @xmath38 6.5 fm ) and c.m .
energies @xmath39 160 gev and @xmath40 1 tev .
the @xmath0 values probed in @xmath41pb processes at @xmath42 = 2.5 can be as low as @xmath43 .
full simulation+reconstruction @xcite of input distributions from the starlight mc @xcite show that cms can measure @xmath44 , @xmath45 within @xmath28 2.5 , in upcs tagged with neutrons detected in the zdcs .
[ fig : upc_ups_cms ] shows the reconstructed @xmath46 around the @xmath3 mass for 0.5 nb@xmath47 integrated pbpb luminosity . with a total yield of @xmath25400 @xmath3 ,
detailed @xmath18,@xmath21 studies can be carried out , to constrain the low-@xmath0 gluon density in the pb nucleus .
( left ) , @xmath45 ( right ) invariant mass distributions from @xmath48pb@xmath49pb@xmath50 ( @xmath51 , signal ) and @xmath52 ( background ) in upc pbpb at @xmath1 = 5.5 tev in cms.,title="fig:",width=291,height=245 ] ( left ) , @xmath45 ( right ) invariant mass distributions from @xmath48pb@xmath49pb@xmath50 ( @xmath51 , signal ) and @xmath52 ( background ) in upc pbpb at @xmath1 = 5.5 tev in cms.,title="fig:",width=291,height=245 ]
jet measurements at tevatron have provided valuable information on the proton pdfs . at 14 tev ,
the production of jets with @xmath55 20100 gev in the cms forward calorimeters ( hf and castor ) probes the pdfs down to @xmath56 @xcite .
figure [ fig : pp_fwd_jets]-left shows the single inclusive jet spectrum in both hfs ( 3@xmath575 ) expected for a short first run with just 1 pb@xmath47 integrated luminosity .
the spectrum has been obtained from a preliminary study using pythia 6.403 with jet reconstruction at the _ particle - level _ ( i.e. _ no _ detector effects are included apart from the hf tower @xmath58 granularity ) @xcite .
although at such low @xmath59 s systematic uncertainties can be as large as @xmath2530% , the available statistics for this study is very high .
= 14 tev ( 1 pb@xmath47 ) obtained from pythia 6.403 at the particle - level ( _ no _ full detector response , underlying - event or hadronization corrections ) .
left : single inclusive jets in hf ( @xmath60 ) ( compared to a nlo calculation with scale @xmath61 = @xmath59 @xcite ) .
right : dijets separated by @xmath62 89 with the mller - navelet kinematics cuts described in the text , as a function of @xmath63.,title="fig:",width=283,height=257 ] = 14 tev ( 1 pb@xmath47 ) obtained from pythia 6.403 at the particle - level ( _ no _ full detector response , underlying - event or hadronization corrections ) . left : single inclusive jets in hf ( @xmath60 ) ( compared to a nlo calculation with scale @xmath61 = @xmath59 @xcite ) .
right : dijets separated by @xmath62 89 with the mller - navelet kinematics cuts described in the text , as a function of @xmath63.,title="fig:",width=283,height=257 ] inclusive dijet production at large pseudorapidity intervals mller - navelet ( mn ) jets has been considered an excellent testing ground for bfkl @xcite and non - linear qcd @xcite evolutions .
the large rapidity separation between partons enhances the available longitudinal momentum phase space for bfkl radiation .
gluon saturation effects are expected to reduce the ( pure bfkl ) mn cross section by a factor of @xmath252 for jets separated by @xmath67 9 @xcite . in order to estimate the expected statistics for a short run without pile - up ( 1 pb@xmath47 ) , we have selected the pythia events which pass the mn kinematics cuts : @xmath68 gev , @xmath69 , and @xmath70 = 6 10 @xcite .
figure [ fig : pp_fwd_jets]-right shows the results for @xmath70 = 89 .
the expected dijet yields for this @xmath21 separation indicate that these studies are clearly statistically feasible at the lhc . +
* acknowledgments . * supported by 6th eu framework programme meif - ct-2005 - 025073 . +
altarelli g and parisi g 1977 298 ; dokshitzer yu l 1977 _ sov . phys .
jetp _ * 46 * 641 kuraev e a _ et al . _
1977 _ zh . eksp .
. fiz _ * 72 * 3 ; balitsky i i _ et al . _
1978 _ sov .
j. nucl .
_ * 28 * 822 see e.g. devenish r 2002 _ proceeds pic 2002 , stanford , ca , 20 - 22 jun 2002 _ , _ preprint _ hep - ex/0208043 gribov l _
1983 _ phys .
rept . _ * 100 * 1 ; mueller a h and qiu j w 1986 427 gelis f 2007 these proceeds , _ preprint _ hep - ph/0701225 and refs .
therein balitsky i 1996 99 ; kovchegov yu 2000 d * 61 * 074018 ; jalilian - marian j _
_ 1997 415 ; iancu e _ et al .
_ 2001 583 denterria d 2006 a to appear , _ preprint _ hep - ex/0610061 betts r [ cms collaboration ] 2007 , these proceedings cms / totem _ prospects for diffractive and forward physics at the lhc _
, cern - lhcc-2006 - 039/g-124 kharzeev d _
_ 2005 609 armesto
n _ et al .
_ 2005 022002 cms _
physics tdr : high density qcd with heavy - ions _ ,
cern - lhcc-2007 - 009 armesto n and pajares c 2000 _ int . j. mod .
_ * a15 * 2019 baltz a _ et al .
_ 2007 _ ultraperipheral collisions at the lhc _ , in preparation klein s r and nystrand j 1999 c 60 014903 ; baltz a _ et al .
_ 2002 89 012301 jager b _
_ 2004 d * 70 * 034010 mller a h and navelet h 1987 727 ; vera a s and schwennsen f 2007 hep - ph/0702158 marquet c and royon c 2006 131 | high energies cross - sections of hadronic objects protons nuclei photons dominated by scatterings involving gluons gluons outnumber quarks in small momentum fraction low-@xmath0 ) range of parton distribution functions qcd parton splitting probabilities by dglap bfkl @xcite evolution equations fast growth of gluon densities @xmath0 continue indefinitely unitarity for scatterings @xmath9 for small @xmath0 values gluons recombine in gluon saturation occurs when size partons similar to size hadron @xmath10 saturation momentum grows with number of nucleons target collision energy rapidity of gluon @xmath14 @xmath13 dependence implies equal energies saturation effects enhanced by factors large as @xmath16 6 in heavy nucleus ( = 208 compared to protons regime of low-@xmath0 qcd described in color glass condensate framework gluon fusions multiple scatterings resummed into high - density gluon wavefunctionsevolution by bk / jimwlk @xcite non - linear equations low-@xmath0 pdfs in hadronic collisions measuring perturbative probes heavy-@xmath17 jets high-@xmath18 hadrons @xmath19 at large @xmath4 forward rapidities for @xmath20 parton scattering minimum @xmath0 probed momentum at pseudo - rapidity @xmath22 @xmath23 @xmath24 decreases by @xmath2510 every 2 units rapidity capabilities cms experiment for low-@xmath0 phenomena with proton ion beams cms / totem system largest in collider detector measure particles with excellent momentum resolution jets muons hadrons neutrals in zero - degree calorimeters 8.3 of four observables in cms sensitive to parton saturation effects in proton nucleus wave - functions at lhc energies other measurements forward drell - yan at 14 tev discussed in in high - energy heavy - ion collisions hadron rapidity density related to initially released partons at @xmath21cgc approaches reduced initial parton flux in nuclear pdfs reproduce absolute hadron yields centrality @xmath1 dependences at sps rhic energies @xcite at lhc expected pbpb multiplicities are @xmath32 2000 fig dndeta left cms simulation studies from in innermost si pixel layer ( @xmath28 2.5 ) indicate occupancy less than 2% reconstructed @xmath2 within @xmath252% of true primary multiplicity dndeta -0.4 cm in central pbpb at lhc range of particle rapidities covered by hijing pbpb distribution of primary simulated tracks @xmath28 2.5 compared to reconstructed hits in first layer of si tracker central pbpb at lhc particle rapidities hijing pbpb distribution of primary simulated tracks within @xmath28 2.5 compared to reconstructed hits in first layer of si tracker -0.4 cm ultraperipheral collisions of heavy ions generate strong electromagnetic fields quasi - real photons study @xmath7 via @xmath36 photoproductionlead beams 2.75 tev lorentz factors @xmath19 = 2930 maximum photon energies @xmath37 100 gev nuclear radius @xmath38 6.5 fm c.m energies @xmath39 160 gev @xmath40 1 tev @xmath0 values @xmath41pb processes @xmath42 = 2.5 low @xmath43 simulation+reconstruction @xcite input distributions starlight mc cms measure @xmath44 @xmath45 @xmath28 2.5 neutrons zdcs_cms reconstructed @xmath46 @xmath3 mass 0.5 nb@xmath47 pbpb luminosity total yield @xmath25400 @xmath3 @xmath18,@xmath21 studies constrain low-@xmath0 gluon density pb nucleus @xmath45 invariant mass distributions upc pbpb @xmath1 = 5.5 tev cms invariant mass distributions = 5.5 tev jet measurements tevatron information proton pdfs 14 tev production jets @xmath55 20100 gev cms forward calorimeters pdfs to @xmath56 @xcitefigure pp_fwd_jets shows single inclusive jet spectrum hfs 3@xmath575 short first run 1 pb@xmath47 integrated luminosity spectrum obtained preliminary study pythia 6.403 jet reconstruction particle - level no detector effects tower @xmath58 @xmath59 s uncertainties @xmath2530% statistics high 14 tev ( 1 pb@xmath47 ) obtained pythia 6.403 particle - level no full detector response underlying - event hadronization corrections left single inclusive jets hf @xmath60 ) compared @xcite dijets separated @xmath62 89 mller - navelet kinematics cuts 14 tev 1 pb@xmath47 ) pythia 6.403 particle - level no full detector response underlying - event hadronization corrections single inclusive jets hf @xmath60 dijets separated @xmath62 89 mller - navelet kinematics cuts inclusive dijet production large pseudorapidity intervals mller - navelet jets excellent testing ground for bfkl @xcite non - linear qcd @xcite evolutionslarge rapidity separation between partons enhances longitudinal momentum space bfkl radiation gluon saturation effects reduce bfkl mn cross section factor @xmath252 for jets separated by @xmath67 @xcite estimate statistics short run without pile - up selected pythia events mn kinematics cuts @xmath68 @xmath70 = 6 10 @xcite figure shows results for @xmath70 = 89 expected dijet yields @xmath21 separation indicate studies statistically feasible at lhc supported by 6th eu framework programme meif - ct-2005 - 025073 altarelli g parisi g 1977 298 dokshitzer yu 1977 46 * 641 kuraev 1977 72 balitsky 1978 28 * 822 devenish 2002 proceeds 2002 stanford 20 - 22 jun hep - ex/0208043 gribov l 1983 100 mueller qiu j 1986 427 2007 ph/0701225 balitsky 1996 99 kovchegov 2000 61 * 074018 jalilian - marian j 1997 415 iancu e2001 583 denterria 2006 preprint hep - ex/0610061 betts collaboration 2007 proceedings diffractive forward physics lhc cern lhcc-2006 - 039/g-124 kharzeev 2005 609 armesto n 022002 high density qcd heavy ions cern lhcc-2007 - 009 armesto pajares 2000 2019 baltz 2007 ultraperipheral collisions lhc klein s nystrand 1999 60 014903 baltz 2002 89 012301 jager 2004 70 034010 mller navelet 1987 727 vera schwennsen 2007 hep - ph/0702158 marquet royon 2006 131 | 0.423504 | 2,959 | 1,517 | , Saving $0.1 in GPT-4. |
20,544 | we discuss the parametric instabilities of large - amplitude parallel propagating alfvn waves using the 2-d pic simulation code .
first , we confirmed the results in the past study [ sakai _ et al .
_ , 2005 ] that the electrons are heated due to the modified two stream instability and that the ions are heated by the parallel propagating ion acoustic waves . however ,
although the past study argued that such parallel propagating longitudinal waves are excited by transverse modulation of parent alfvn wave , we consider these waves are more likely to be generated by the usual , parallel decay instability .
further , we performed other simulation runs with different polarization of the parent alfvn waves or the different ion thermal velocity .
numerical results suggest that the electron heating by the modified two stream instability due to the large amplitude alfvn waves is unimportant with most parameter sets . | large amplitude , low - frequency alfvn waves constitute one of the most essential elements of magnetic fluctuations in the solar corona and solar wind@xcite . due to small collisionless dissipation rates
, the waves can propagate long distance and efficiently convey such macroscopic quantities as momentum , energy , and helicity .
since loading of such quantities is completed when the waves damp away , it is important to examine how the waves can dissipate . among various possible dissipation processes of the alfvn waves ,
parametric instabilities have been believed to be important@xcite .
further , the parametric instabilities of these alfvn waves can cause the developed turbulence , which are believed to produce _ in situ _ the localized structures ( shocks and discontinuities ) in the solar wind@xcite .
however , most past studies using kinetic simulation code mainly discussed the parametric instabilities among parallel propagating waves in one - dimensional system , although the real solar wind is three - dimensional . actually , arguments including obliquely propagating waves have been developed by several authors using the fluid systems@xcite .
these studies clarified that the growth rates involving the oblique waves are usually much smaller than those among parallel propagating waves . on the other hand , sakai _
et al._@xcite recently performed two - dimensional particle - in - cell ( 2-d pic ) simulations .
they reported that the left - hand ( lh- ) polarized alfvn waves are dissipated through the transverse modulational instability when ion and electron beta is very low ( @xmath0 ) .
this is in contrast to the past studies using the fluid systems which indicated that parametric instabilities including obliquely propagating daughter waves ( transverse modulation or filamentation instability ) can be dominant only in the high @xmath1 ( the squared normalized sound velocity ) plasma .
one of the purposes of the present study is to clarify the reason of this difference on the type of parametric instabilities between the fluid and the pic simulations .
sakai _ et al._@xcite also reported the electron heating through the modified two stream instability ( mtsi)@xcite which is caused by the difference in ion and electron bulk velocities ( cross - field currents ) . in a presence of finite amplitude alfvn waves ,
the cross - field currents can initially be given by the walen relation .
however , sakai _
et al._@xcite only discussed lh- mode in their paper .
another purpose of this study is to confirm the importance of the mtsi driven by alfvn waves with both right - hand ( rh- ) and left - hand ( lh- ) polarizations , and also by a mixture of alfvn waves ( in which the walen condition is not satisfied ) .
the paper is organized as follows . in section 2
, we present our simulation model . in section 3
, we present our simulation results .
a summary and conclusions of this paper are given in section 4 .
we have carried out the 2-d pic simulation@xcite with the system size @xmath2 and @xmath3 with 100 super - particles ( both ions and electrons ) per cell .
we use periodic boundary conditions in both @xmath4 and @xmath5 directions , with the uniform ambient magnetic field ( @xmath6 ) in the @xmath4-direction , and the homogeneous plasma ( the zeroth order longitudinal bulk velocity @xmath7 ( @xmath8 ) and the number density @xmath9 ) .
the ion to electron mass ratio is @xmath10 , and the ratio of the electron plasma to the cyclotron frequency is @xmath11 ( @xmath12 and @xmath13 are plasma and the cyclotron angular frequencies of electrons ( @xmath14 ) and ions ( @xmath15 ) , which include the ) .
alfvn velocity @xmath16 , thus , when the beta ratio is @xmath17 for both ions and electrons , electron thermal velocity is @xmath18 , and ion thermal velocity is @xmath19 .
the debye length @xmath20 ( grid size ) and the electron skin depth @xmath21 . as initial conditions , finite amplitude , monochromatic alfvn waves are given @xmath22 where @xmath23 ( complex transverse magnetic field ) and @xmath24 ( complex transverse bulk velocity of each species ) , @xmath25 , @xmath26 and @xmath27 satisfy the dispersion relation of the parallel propagating electromagnetic waves in the two fluid ( electron - ion ) system@xcite , @xmath28 and @xmath29 is given by the walen relation@xcite @xmath30 we adopt the notation that the positive ( negative ) @xmath26 corresponds to the rh- ( lh- ) polarized waves . in our pic simulation the charge neutrality can not be assumed and displacement current can not be neglected due to the values assigned to @xmath31 and @xmath32 . while these parameters used in the present study are different from those in the past studies using the hybrid simulation code@xcite , our results can basically be explained within the framework of the hall - mhd theory as we will see later .
.[tab : table2]parameters used in simulation runs . [ cols="^,^,^,^,^,^",options="header " , ] we performed four simulation runs with different parameters as tabulated in table [ tab : table2 ] .
the system size is @xmath33 and @xmath34 in run 1 , 3 and 4 , and @xmath35 and @xmath34 in run2 .
we have confirmed that the results remain essentially the same when we extended the transverse system size to @xmath36 .
run 1 is almost the same as the run in sakai _
et al._@xcite(lh- parent wave satisfying the walen relation ) . in run 2
the rh- parent waves are given , and in run 3 , magnetic and the velocity perturbations are given independently ( and thus the walen relation is not satisfied ) in order to model the fluctuations at the chromosphere . in run 4 , we introduce higher ion temperature in order to examine the roles of kinetic modulational instability@xcite .
initial alfvn waves are all forward propagating in runs 1 , 2 , and 4 , while forward and backward propagating waves are mixed in run 3 . to discuss the propagating direction of alfvn waves , the two - dimensional ( @xmath37 ) spacial fourier transforming analysis were performed in this study . since we focus on the low - frequency waves in this paper , the vertical and horizontal axes ( @xmath38 and @xmath39 ) of the figures indicating the fourier analysis ( fig .
[ 3.2.3 ] , [ 3.2.5 ] , [ 3.2.7 ] ) are normalized by the ion skin depth ( @xmath40 ) .
time histories of magnetic field energy ( thick line ) , ion ( dashed line ) and electron ( thin line ) total energies in run 1 . ]
first , we show the results of run 1 , in which physical parameters correspond to those in sakai _
et al._@xcite . fig .
[ 3.2.1 ] shows the time history of magnetic field energy , ion and electron total energies , respectively . as discussed in sakai _
et al._@xcite , the electron energy is enhanced by the excitation of mtsi ( the arrow in fig . [ 3.2.1 ] ) .
[ 3.2.2](a ) clearly shows the obliquely propagating waves with parallel electric field , corresponding to fig.3(b ) in sakai _
et al._@xcite .
we have also confirmed that the propagation angle of the waves is about @xmath41 , which agrees with the simulation results in sakai _
et al._@xcite and the linear analysis in wu _ et al._@xcite ( we note that while wu _
et al._@xcite discuss the case @xmath42 , the growth rate of the mtsi is not sensitive to @xmath32@xcite).the initial perpendicular velocity ( @xmath43 ) is obtained from eq.([eq32002 ] ) @xmath44 where @xmath45 , @xmath46 .
@xmath43 in run 1 is @xmath47 .
we remark that the kinetic effects of electrons are important for the mtsi observed here , since @xmath48@xcite .
snapshots of longitudinal electric field @xmath49 at ( a)@xmath50 and ( b)@xmath51 in run 1 . ] after these obliquely propagating waves excited by the mtsi are damped by electron landau damping , low - frequency parallel propagating waves become dominant ( fig .
[ 3.2.2](b ) ) .
dissipation of these parallel propagating ion acoustic waves heat the ions as discussed in sakai _
et al._@xcite .
although sakai _
et al._@xcite argued that such parallel propagating longitudinal waves are excited by transverse modulation of parent alfvn wave , we consider these waves are more likely to be generated by the usual , parallel decay instability .
[ 3.2.3 ] shows the power spectrum of the complex magnetic field @xmath52 ( @xmath53 ) in the @xmath54 wave number space at @xmath55 .
the anti - parallel propagating daughter waves ( @xmath56 ) are dominantly excited around @xmath57 .
presence of such anti - parallel propagating waves provides a strong evidence of the occurrence of the decay instability , while the daughter alfvn waves excited by transverse modulation have finite propagation angles .
in addition , these daughter waves are consistent with the linear analysis of hall - mhd system for parallel propagating waves@xcite .
thus , we conclude that under this set of parameters , the one - dimensional decay instability dominantly dissipates the parent alfvn wave , as the hall - mhd theory predicts@xcite . while numerical results of the present study and sakai _ et al._@xcite are quite similar , their interpretation on the parametric instability of parent alfvn wave seems misleading .
next , we discuss the rh- mode case ( run 2 ) . fig .
[ 3.2.4 ] shows the time history of energies in the same format as fig .
[ 3.2.1 ] .
in contrast to fig . [ 3.2.1 ] ( run 1 ) , electron heating due to the excitation of mtsi is not evident . from eq .
( [ eq32004 ] ) , @xmath43 in run 2 is @xmath58 , which is larger than @xmath59 .
we here find that even when @xmath60 , which is a rough criterion of the mtsi@xcite , the cross field current is not large enough to excite the mtsi under the particular set of parameters used . on the other hand
, the decay instability takes place in a way similar to run 1 ( not shown ) . the power spectrum of the complex magnetic field @xmath52 ( @xmath53 ) in @xmath54 wave number space at @xmath55 in run 1 .
the vertical and horizontal axes ( @xmath38 and @xmath39 ) are normalized by the ion skin depth . ]
time histories of magnetic field energy ( thick line ) , ion ( dashed line ) and electron ( thin line ) total energies in run 2 . ] in run 3 , we discuss the case in which the walen relation is not satisfied for the initial alfvn wave . these initial conditions are intended to simulate the plasma and the magnetic field fluctuations in the solar wind and solar surface , where both rh- and lh- polarized alfvn waves are considered to be mixed@xcite . in such a case , amplitude of transverse magnetic field and transverse bulk velocities vary in time@xcite . since such linear oscillations modify the magnitude of the cross field current , they let the system stay in the stable and unstable regimes in turn , so that the mtsi growth is effectively reduced . the power spectrum of the complex magnetic field @xmath52 ( @xmath53 ) in @xmath54 wave number space at @xmath61 in run 3 .
the vertical and horizontal axes ( @xmath38 and @xmath39 ) are normalized by the ion skin depth . ] finally , we discuss the results of run 4 , in which parameters are the same as those in run 1 except for @xmath62 .
in contrast to run 1 , mtsi is suppressed by the ion kinetic effects ( fig .
[ 3.2.6 ] ) .
further , as fig . [ 3.2.7 ] shows , the decay instability is not the dominant instability in this case , because of the occurrence of kinetic
modulational instability@xcite . in fig .
[ 3.2.7 ] , both the dominant daughter waves ( @xmath63 ) and the higher harmonic waves are observed .
time histories of magnetic field energy ( thick line ) , ion ( dashed line ) and electron ( thin line ) total energies in run 4 . ] the power spectrum of the complex magnetic field @xmath52 ( @xmath53 ) in @xmath54 wave number space at @xmath64 in run 4 .
the vertical and horizontal axes ( @xmath38 and @xmath39 ) are normalized by the ion skin depth . ]
in this paper , we discussed the parametric instabilities of alfvn waves using a 2-d pic simulation code .
first , we re - ran the simulation of sakai _
et al._@xcite using essentially the same set of parameters .
while our numerical results are basically the same , we conclude that the generation of longitudinal waves are due to the usual , parallel decay instability , rather than the transverse modulational instability proposed by sakai _ et al._@xcite , since the one - dimensional decay instability is dominant in a low beta plasma as the hall - mhd theory predicted@xcite .
further , we performed several other simulation runs using different parent alfvn wave polarizations and different ion thermal velocities . the condition to excite the mtsi
is easily violated by such modifications of the parameters .
thus , the electron heating by the mtsi of the alfvn waves seems unimportant in the upper chromoshpere , since the alfvn waves excited by the convection in the solar surface are likely to be the superposition of alfvn waves with different polarizations and propagating directions , but not the monochromatic , lh- polarized wave .
further , a similar process on obliquely propagating waves have also been suggested by markovskii and hollweg @xcite according to the proton heating . the more detailed analysis on this issue is desired . while the dominant parametric instabilities are quasi - one dimensional in a low beta case , the transverse modulational instability can be more important in different circumstances .
for instance , the alfvn wave excited by the ion beam instabilities can be dominantly dissipated through the transverse modulational instability , which is nonlinearly driven by the obliquely propagating waves excited by the same beam instabilities@xcite .
further , even if the parametric instabilities are one dimensional , two or three dimensional density structures are excited as a consequence of its nonlinear evolution@xcite .
when the inhomoginity of the background plasmas exist , heating of electrons by phase mixed alfvn waves take place@xcite .
we remark that even in the uniform background plasmas , the electron heating regardless of the mtsi is actually observed in pic simulation ( _ e.g. _ , sakai _
et al._@xcite and fig .
[ 3.2.4 ] ) . the electron heating is possibly caused by numerical noise originated from the super - particles in the pic simulation . to avoid this and confirm the recent results , the electromagnetic vlasov simulations are in plan
this paper has been supported by jsps research fellowships for young scientists in japan .
99 tu c y and marsch e 1995 _ space sci . rev .
_ , * 73*(1 - 2 ) , 1 | large amplitude low - frequency alfvn waves essential magnetic fluctuations in solar corona solar wind to small collisionless dissipation rates waves propagate long distance convey macroscopic momentum energy helicity loading quantities completed when waves damp important to examine how waves dissipate dissipation parametric instabilities important parametric instabilities cause developed turbulence produce localized structures shocks discontinuities in solar wind past studies discussed parametric instabilities among parallel propagating waves in one - dimensional system real solar wind three - dimensional arguments including obliquely propagating waves developed by authors using fluid systems studies growth rates oblique waves smaller than parallel propagating waves sakai _ et al._ performed two - particle - in - cell simulations reported left - hand polarized alfvn waves dissipated through transverse modulational instability when ion electron beta low ( contrast to past studies parametric instabilities including obliquely propagating waves dominant only in high @xmath1 plasma present study to clarify difference on parametric instabilities between fluid pic simulationssakai et al. reported electron heating through modified two stream instability caused by difference in ion and electron bulk velocities cross - field currents finite amplitude alfvn waves cross - field currents given by walen relation discussed lh- mode purpose study confirm importance of mtsi driven by alfvn waves with right and left - polarizations mixture of alfvn waves walen condition not paper organized section 2 simulation model section 3 simulation results summary conclusions in section 4 carried out 2-d pic simulation@xcite with system size @xmath2 and @xmath3 with 100 super - particles ( ions and electrons per cell periodic boundary conditions in @xmath4 directions uniform ambient magnetic field homogeneous plasma zeroth order longitudinal bulk velocity number density ion to electron mass ratio is ratio electron plasma to cyclotron frequency is are plasma cyclotron angular frequencies of electrons ions velocity beta ratio is @xmath17 for ions and electrons electron thermal velocity is @xmath18 ion thermal velocity @xmath19debye length @xmath20 grid size electron skin depth @xmath21 initial conditions finite amplitude monochromatic alfvn waves given @xmath22 @xmath23 complex transverse magnetic field @xmath24 transverse bulk velocity @xmath25 @xmath26 @xmath27 satisfy dispersion relation parallel electromagnetic waves fluid electron - ion system@xcite @xmath28 @xmath29 given walen relation positive negative @xmath26 corresponds to rh- ( lh- ) polarized waves charge neutrality assumed displacement current neglected values @xmath31 @xmath32 parameters present study different from past studies hybrid simulation code results explained hall - mhd theory table2]parameters simulation runs performed four simulation runs different parameters table system size @xmath33 @xmath34 in run 1 3 4 @xmath35 @xmath34 run2 results same extended transverse system size to @xmath36 run 1 same as@xcite(lh- parent wave satisfying walen relation run 2 rh- parent waves given run 3 magnetic velocity perturbations independently walen relation not satisfied model fluctuations at chromosphererun 4 higher ion temperature examine kinetic modulational instability@xcite initial alfvn waves forward propagating in runs 1 2 4 forward backward mixed run 3 discuss propagating direction alfvn waves two - dimensional @xmath37 ) spacial fourier transforming analysis performed focus low - frequency waves vertical horizontal axes @xmath38 @xmath39 ) figures 3.2.3 3.2.7 normalized by ion skin depth @xmath40 ) time histories magnetic field energy ion electron total energies run 1 results run 1 physical parameters correspond sakai _ et al._@xcite fig 3.2.1 shows time history magnetic field energy ion electron total energies electron energy enhanced by excitation mtsi [ 3.2.2](a ) shows obliquely propagating waves parallel electric field corresponding fig.3(b ) sakai propagation angle waves about @xmath41 agrees with simulation results sakai _ et al._@xcite linear analysis wu _ et al._@xcite growth rate mtsi not sensitive to @xmath32 initial perpendicular velocity ( @xmath43 ) obtained from eq.[eq32002 @xmath44 @xmath45 @xmath43 run 1 @xmath47 kinetic effects electrons important for mtsi @xmath48@xcite longitudinal electric field @xmath49@xmath50@xmath51 run 1 propagating waves mtsi damped electron landau damping low - frequency parallel waves dominant fig 3.2.2](b ) dissipation waves heat ions sakai _ et al._@xcite parallel waves excited transverse modulation parent alfvn wave likely generated by parallel decay instability 3.2.3 ] power spectrum complex magnetic field @xmath52 @xmath53 ) @xmath54 wave @xmath55 anti - parallel propagating daughter waves @xmath56 ) excited around @xmath57 evidence decay instability daughter alfvn waves excited transverse modulation finite propagation angles daughter waves consistent with linear analysis hall - mhd system parallel waves one - dimensional decay instability dissipates parent alfvn wave hall - mhd theory predicts numerical results study sakai _ et al._@xcite similar interpretation parametric instability parent alfvn wave misleadingdiscuss rh- mode case run 2 ) fig 3.2.4 ] shows time history energies as 3.2.1 contrast to 3.2.1 electron heating due to excitation mtsi not evident @xmath43 run 2 is @xmath58 larger than @xmath59 @xmath60 cross field current not large enough to excite mtsi decay instability similar to run 1 power spectrum complex magnetic field @xmath52 ( @xmath53 ) @xmath54 wave space at @xmath55 run 1 vertical horizontal axes @xmath38 @xmath39 normalized by ion skin depth time histories of magnetic field energy ion electron total energies in run 2 run 3 discuss walen relation not satisfied for initial alfvn wave conditions simulate plasma magnetic field fluctuations solar wind surface rh- lh- polarized alfvn waves mixed transverse magnetic field transverse bulk velocities vary in time linear oscillations modify cross field current system stay in stable unstable regimes mtsi growth reduced power spectrum magnetic field @xmath52 ( @xmath53 ) @xmath54 wave space at @xmath61 in run 3vertical horizontal axes @xmath38 @xmath39 normalized by ion skin depth discuss results run 4 parameters same as 1 except @xmath62 mtsi suppressed by ion kinetic effects fig . 3.2.6 ] fig . 3.2.7 decay instability not dominant kinetic modulational instability 3.2.7 dominant daughter waves @xmath63 higher harmonic waves observed histories of magnetic field energy ion electron total energies in run 4 power spectrum of complex magnetic field @xmath52 @xmath53 @xmath54 wave number space at @xmath64 in run 4 vertical horizontal axes @xmath38 normalized by ion skin depth discussed parametric instabilities of alfvn waves using 2-d pic simulation code re - ran simulation of sakai _ et al._@xcite same parameters generation of longitudinal waves due to parallel decay instability transverse modulational instability one - dimensional decay instability dominant in low beta plasma performed simulation runs using different alfvn wave polarizations ion thermal velocities condition to excite mtsi violated by modifications parameterselectron heating by mtsi alfvn waves unimportant in upper chromoshpere excited by convection solar surface likely superposition of different polarizations directions not monochromatic lh- polarized wave similar process on obliquely propagating waves suggested by markovskii hollweg proton heating detailed analysis desired dominant parametric instabilities quasi - one dimensional in low beta transverse modulational instability important in different circumstances alfvn wave excited by ion beam instabilities dissipated through transverse modulational instability driven by obliquely propagating waves excited same beam instabilities instabilities one dimensional two three dimensional density structures excited nonlinear evolution inhomoginity background plasmas heating of electrons by phase mixed alfvn waves in uniform background plasmas electron heating regardless of mtsi observed in pic simulation possibly caused by numerical noise from super - particles confirm electromagnetic vlasov simulations in plan paper supported by jsps research fellowships for young scientists in japan 1995 space sci . rev . | 0.504958 | 3,523 | 1,776 | , Saving $0.1 in GPT-4. |
20,549 | we consider a simple model and make use of field transformations in lagrangian field theories to illustrate the impossibility of measuring off - shell effects in nucleon - nucleon bremsstrahlung and related processes . | the issue of how to describe the interaction of a particle which is not on its mass shell has a long history .
for example , in their derivation of the low - energy theorem of compton scattering in 1954 , gell - mann and goldberger @xcite took into account that the electromagnetic vertex of an off - shell nucleon is more complicated than that of a free nucleon . in case of the electromagnetic interaction , some restrictions on the general form of the off - shell vertex result from the ward - takahashi identity @xcite .
it is a natural and legitimate question to ask whether it is possible to extract the off - shell behavior of particular interaction vertices from empirical information similarly as one , say , extracts the electromagnetic form factors of the nucleon from elastic electron scattering . in this context
one might think of the electromagnetic interaction of a bound nucleon @xcite .
another example would be an investigation of the off - shell nucleon - nucleon amplitude entering the nucleon - nucleon - bremsstrahlung process @xcite .
it has only been recently that the relevance of field transformations in the framework of lagrangian field theories @xcite has been emphasized in addressing this question . for the case of pions , compton scattering @xcite and pion - pion bremsstrahlung @xcite
have been considered using chiral perturbation theory .
it was shown that off - shell effects both depend on the model used and on the choice of representation for the fields . from
that we concluded that off - shell effects are not only model dependent but also representation dependent , making a unique extraction of off - shell effects impossible . in practice , the freedom of choosing appropriate field variables has been used in refs .
@xcite to obtain the most general effective lagrangian describing low - energy ( virtual ) compton scattering . here
, we will extend our previous discussion @xcite to the investigation of a spin one - half system @xcite which should remove any uncertainty that the results of the previous works somehow depended on the simplicity of a spin - zero process .
as our toy model we take @xmath0 where @xmath1 is the covariant derivative of the proton field , @xmath2 and @xmath3 are the proton charge and anomalous magnetic moment respectively , @xmath4 is the photon field , and @xmath5 is the electromagnetic field strength tensor . the fields @xmath6 and @xmath7 refer to protons and neutrons , respectively . for calculational simplicity , we neglect the electromagnetic coupling to the neutron magnetic moment . in the framework of eq .
( [ eq : l0 ] ) it is straightforward to calculate the born diagrams of proton - neutron bremsstrahlung : @xmath8 u_p(p_1 ) , \label{eq : nngborn}\end{aligned}\ ] ] which corresponds to the usual choice for nucleon - nucleon bremsstrahlung for the electromagnetic parts .
obviously , for pedagogical reasons , eq . ( [ eq : nngborn ] ) has a much simplified interaction for the strong part .
next , we perform a change of variables leading to different off - shell behavior in the nucleon - nucleon amplitude as well as the photon - nucleon vertex , and study the effect on the total @xmath9 bremsstrahlung amplitude . for that purpose
we consider the transformation @xmath10 where @xmath11 and @xmath12 are real constants , determining the overall strength of the transformation .
equation ( [ eq : wftrans ] ) results in a class of equivalent lagrangians @xmath13 where the explicit expressions of @xmath14 and @xmath15 can be found in ref .
consider , for example , @xmath16 and @xmath17 , in which case the @xmath11 term will generate an off - shell dependence in the proton legs of the @xmath9 amplitude @xmath18 \overline{u}_n(p_4)u_n(p_2 ) , \label{eq : l1strongvertex}\ ] ] in conjunction with a @xmath19 contact term @xmath20 one can show that in the amplitude the off - shell dependence from eq .
( [ eq : l1strongvertex ] ) precisely cancels the contact term of eq .
( [ eq : l1acontact ] ) . in the present context , such a cancellation for the charge , but not the magnetic , parts
is simultaneously enforced by gauge invariance .
a similar observation was also made in ref .
@xcite for real compton scattering off a charged pion . however , the implication of field transformations is even more general , because it also has consequences for terms which are not fixed by gauge invariance , such as the magnetic terms above .
the general case @xmath21 is discussed in ref.@xcite .
the above result is a simple illustration of the equivalence theorem of lagrangian field theory @xcite , according to which all lagrangians of eq .
( [ dl ] ) result in identical @xmath22-matrix elements .
one can also make use of this observation in order to show that what appears as an off - shell effect in an @xmath22-matrix element for one lagrangian may originate in a contact term from an equivalent lagrangian @xcite .
we conclude that off - shell effects can not in any unambiguous way be extracted from an @xmath22-matrix element .
9 m. gell - mann and m.l .
goldberger , phys .
96 ( 1954 ) 1433 .
ward , phys .
78 ( 1950 ) 182 ; y. takahashi , nuovo cim . 6
( 1957 ) 371 .
naus and j.h .
koch , phys .
c 36 ( 1987 ) 2459 ; h.w.l .
naus , s.j .
pollock , j.h .
koch , and u. oelfke , nucl .
a509 ( 1990 ) 717 ; p.c .
tiemeijer and j.a .
tjon , phys .
c 42 ( 1990 ) 599 ; x. song , j.p .
chen , and j.s .
mccarthy , z. phys .
a 341 ( 1992 ) 275 .
see e.g. v.r .
brown , p.l .
anthony , and j. franklin , phys . rev .
c 44 ( 1991 ) 1296 ; v. herrmann and k. nakayama , phys . rev .
c 46 ( 1992 ) 2199 ; a. katsogiannis and k. amos , phys .
c 47 ( 1993 ) 1376 ; m. jetter and h.v .
von geramb , phys .
c 49 ( 1994 ) 1832 ; g.h .
martinus , o. scholten , and j.a .
tjon , phys .
b 402 ( 1997 ) 7 ; r.l . workman and h.w .
fearing , phys .
c 34 ( 1986 ) 780 .
chisholm , nucl .
( 1961 ) 469 ; s. kamefuchi , l. oraifeartaigh , and a. salam , nucl .
( 1961 ) 529 ; s. coleman , j. wess , and b. zumino , phys .
177 ( 1969 ) 2239 ; s. scherer and h.w .
fearing , phys .
d 52 ( 1995 ) 6445 .
s. scherer and h.w . fearing , phys .
c 51 ( 1995 ) 359 ; s. scherer , in proceedings of the conference on perspectives in nuclear physics at intermediate energies , trieste , 1995 ( nucl - th/9506028 ) .
fearing , phys .
81 ( 1998 ) 758 ; few - body syst .
, in press , nucl - th/0006040 .
lvov , int
phys . a 8 ( 1993 ) 5267 ; s. scherer , a.yu .
korchin , and j.h .
koch , phys .
c 54 ( 1996 ) 904 ; h.w .
fearing and s. scherer , few - body syst .
23 ( 1998 ) 111 .
h.w . fearing and s. scherer , accepted for publication in phys .
c ( 2000 ) ( nucl - th/9909076 ) .
kaloshin , phys .
62 ( 1999 ) 1899 [ yad .
62 ( 1999 ) 2049 ] . | issue interaction particle not on mass shell long history derivation low - energy theorem compton scattering in 1954 gell - mann and goldberger electromagnetic vertex of off - shell nucleon more complicated than free nucleon restrictions on form off - shell vertex result from ward - takahashi identity question extract off - shell behavior of interaction vertices from empirical information electromagnetic form factors nucleon from elastic electron scattering think electromagnetic interaction of bound nucleon example investigation off - shell nucleon - amplitude entering nucleon - nucleon - bremsstrahlung process recently relevance of field transformations in lagrangian field theories emphasized case pions compton scattering pion - pion bremsstrahlung considered using chiral perturbation theory off - shell effects depend on model and choice of representation concluded model dependent representation dependent unique extraction impossible freedom of choosing field variables used in to obtain effective lagrangian describing low - energy compton scattering extend previous discussion to investigation of spin one - half system remove uncertainty results previous works depended on simplicity of spin - zero processtoy model @xmath0 @xmath1 covariant derivative proton field proton charge anomalous magnetic moment photon field electromagnetic field strength tensor fields refer protons neutrons neglect electromagnetic coupling neutron magnetic moment framework eq . calculate born diagrams proton - neutron bremsstrahlung @xmath8 u_p(p_1 ){eq : nngborn}{aligned usual choice nucleon - nucleon bremsstrahlung electromagnetic parts eq . nngborn simplified interaction strong part perform change variables off - shell behavior nucleon - nucleon amplitude photon - nucleon vertex study effect total @xmath9 bremsstrahlung amplitude consider transformation @xmath10 @xmath11 @xmath12 real constants determining overall strength transformation equation eq : wftrans ] results class equivalent lagrangians @xmath13 explicit expressions @xmath14 @xmath15 in ref .@xmath16 @xmath17 @xmath11 term off - shell dependence in proton legs @xmath9 amplitude @xmath18\overline_n(p(p_2{eq : l1strongvertex @xmath19 contact term @xmath20 off - shell dependence from eq cancels contact term eq cancellation for charge not magnetic parts enforced by gauge invariance similar observation in @xcite compton scattering off charged pion implication field transformations general consequences for terms not fixed by gauge invariance magnetic terms general case @xmath21 discussed in ref.@xcite equivalence theorem of lagrangian field theory lagrangians of eq result in identical @xmath22-matrix elements off - shell effect in @xmath22-matrix element lagrangian originate in contact term from equivalent lagrangian @xcite off - shell effects not extracted from @xmath22-matrix element gell - mann goldberger, phys . 96 ( 1954 ) 1433 ward , phys . 78 ( 1950 ) 182 ; takahashi nuovo cim . 6 ( 1957 ) 371 naus jkoch phys 36 1987 ) 2459 naus pollock koch oelfke 1990 ) 717 tiemeijer j tjon 42 1990 ) 599 song chen j mccarthy 341 ( 1992 ) 275 v brown anthony franklin 44 1991 ) 1296 herrmann nakayama 46 ( 1992 ) 2199 katsogiannis k amos phys 47 ( 1993 ) 1376 m jetter geramb phys 49 1994 ) 1832 martinus scholten j tjon phys 402 ( 1997 ) 7 r workman h.w fearing phys 34 ( 1986 ) 780 chisholm nucl 1961 ) 469 kamefuchi oraifeartaigh salam nucl 1961 ) 529 coleman j wess zumino phys . 177 ( 1969 ) 2239 scherer h.w fearing phys 52 ( 1995 ) 6445 scherer fearing phys 51 1995 ) 359scherer conference nuclear physics intermediate energies trieste 1995 nucl th/9506028 fearing phys . 81 1998 ) 758 few - body syst. nucl th/0006040 lvov int phys . 8 1993 ) 5267 scherer korchin koch . 54 1996 ) 904 fearing scherer few - body syst . 23 1998 ) 111 accepted publication 2000 nucl th/9909076 kaloshin phys . 62 1999 ) 1899 2049 | 0.309507 | 1,892 | 967 | , Saving $0.1 in GPT-4. |
20,552 | in gravitational microlensing , interference visibility is established when the spectral resolution of the observation is less than the reciprocal of the time delay difference .
using high resolution spectroscopy fringes will be visible at large path differences , even when the source emission is incoherent .
this is possible because coherence can be established after interference has already occurred .
results of a laboratory test are presented which demonstrate this retroactive establishment of coherence . | interference can occur when there are two paths for light propagation from a source to an observer . in gravitational microlensing ( @xcite , @xcite , @xcite ) wavetrains from a point on the source star travel on two paths , and the wave amplitudes
are combined before the light is detected .
the lensing geometry has the potential to produce interference , and applications of gravitational lensing interference have been considered for wavelength regimes from radio waves to gamma rays ( @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite ) .
as many of these authors note , interference visibility will be diminished if the difference of the length of the two paths exceeds the coherence length of the light .
the light sources in gravitational microlensing are incoherent . in many microlensing examples
the light sources are stars and stellar photospheres emit primarily blackbody radiation .
the temporal coherence length of this light is @xmath0 , where @xmath1 is the speed of light , @xmath2 is the coherence time and @xmath3 is the half - power bandwidth of the source emission spectrum . because of the large bandwidth of the blackbody spectrum , the coherence length @xmath4 is about equal to the peak wavelength .
blackbody emission is often labeled `` incoherent '' .
blackbody emission can be modeled as a continuous stream of short wavepackets .
figure 1 shows wavepackets of length @xmath4 traveling on two paths from source to observer around a lensing mass . because the wavepackets do not overlap at arrival
, the short coherence length of the light would seem to indicate that gravitational lensing interference is undetectable except in cases where the path length difference amounts to just a few wavelengths .
however , this is not correct , as the experiment described below demonstrates .
coherence can be established retroactively .
_ after _ incoherent light has passed through the interference apparatus ( in the lab or in the universe ) , it can be filtered to a narrow bandwidth and that filtered light will have a new , longer coherence length .
this allows gravitational microlensing interference to be observed for large path length differences . in this letter
i present results of a laboratory experiment , carried out using a michelson interferometer , which demonstrates retroactive coherence .
an analysis of partial coherence for two path geometries is presented in @xcite .
i adapt their analysis to the present purpose . to describe the degree of interference seen in the patterns formed by his interferometers michelson defined a quantity called the interference visibility @xmath5 . here the maximum and minimum of the intensity @xmath6 and @xmath7
are taken as the path length difference @xmath8 varies .
@xmath9 ranges from 0 to 1 , and generally has a maximum at zero path difference . for two - path gravitational microlensing , or for light that passes through a michelson interferometer ,
the detected intensity is @xmath10 where @xmath11 and @xmath12 are the intensities on the two paths , @xmath13 is a complex function describing the time variation of the arriving field strength on the shorter path , and @xmath14 is the same function shifted in time by the path delay difference .
@xmath15 is normalized , i.e. @xmath16 , and assumed to be stationary .
@xmath17 is the duration of the observation , the time over which the intensity is averaged , and is taken to be much longer than @xmath2 and @xmath18 .
the total intensity is then @xmath19 the last term describes the interference is also called the real part of the complex degree of coherence @xmath20 . for an optical frequency band centered at @xmath21 the magnitude of the degree of coherence
is related to the interference visibility by @xmath22 .
the autocorrelation of the wave amplitude @xmath23 is the fourier transform of the power spectral density @xmath24 of the light and the coherence length is the width of the function @xmath25 .
this means that decreasing the bandwidth of the observation , i.e. forcing @xmath24 to zero outside a narrow range of frequencies , increases the coherence length and makes interference visible at large path length differences .
the experiment described below shows that this is true even when the bandwidth restriction occurs after the light has passed through the interference apparatus .
i set up a laboratory analog of gravitational lensing interference using a michelson interferometer , as shown in figure 2 . in the experiment light from a laboratory thermal source arrives at a partially reflecting beam splitter which divides the light among two paths .
these paths are the two orthogonal arms of the interferometer . at the ends of the arms mirrors return the light to the beam splitter with a path length difference that can be adjusted .
the beam splitter recombines the light before it passes through the exit aperture .
filtering the light restores interference visibility . in gravitational lensing
the path length varies with time because of relative motion among the source , lensing masses , and the observer . in the michelson interferometer ,
the path lengths are stable but the interferometer can be adjusted to give various path lengths through various parts of the instrument .
i tilted one mirror slightly so zero path length difference occurs only at a line dividing the pattern vertically .
the pattern is also divided horizontally .
the top half of the exit aperture of the interferometer is unblocked and here the full bandwidth of the thermal source contributes to the interference pattern .
the bottom half of the exit aperture is covered by a narrow - band ( @xmath26 ) filter . in figure 3
the results of the experiment are displayed . for white light
the visibility falls almost to zero within a few fringes . for filtered light
the visibility remains near one for the entire width of the pattern .
considering the frequency domain it is easy to see how coherence is restored through bandwidth restriction : the white light interference pattern is composed of many narrow - band patterns overlaid .
the spacing of the fringes is proportional to wavelength .
for all wavelengths , the patterns are aligned in the center of the image , at zero path difference .
the central few fringes have high visibility but , away from the center of the image , the pattern develops colored stripes , as red and blue fringe patterns separate in phase . at the edge of the pattern
the visibility has fallen nearly to zero because of cancellation , among the various wavelengths , of constructive and destructive interference . by selecting a narrow range of frequencies cancellation
is avoided and the interference becomes visible across the entire field .
this is shown in the right panel of figure 3 . in the time domain
a puzzle remains : how can short coherence length wave packets , which do nt overlap as they exit the interferometer , create visible interference at large path length differences ?
the answer is that the filter stores multiply delayed copies of the wave train incident upon it .
interference occurs among these copies .
light storage in a filter is illustrated in figure 4 .
in the experiment described here a single narrow filter is used to restrict the passband of the observation .
most of the optical power from the source is blocked by this filter . in a gravitational microlensing observation a spectrophotometer could be used instead of a filter .
this would divide the spectrum into many narrow passbands , each acting as a narrow filter .
this allows all the optical power to be used in the observation . for a spectrophotometer
the coherence time @xmath2 is the reciprocal of the frequency resolution .
although temporal coherence can be increased by using high spectral resolution , limited _ lateral _ spatial coherence remains a factor reducing lensing interference visibility for most geometries studied so far .
as described in the references above , the path length difference varies from point to point across the disk of a source star .
the total variation is generally many wavelengths , and this substantially reduces interference visibility .
occasionally , microlensing events include caustic crossings . at the first crossing two new , high magnification paths appear .
these paths have the same length at first so the size of the laterally coherent region ( @xcite , @xcite ) can cover a much larger part of the source disk than in the more common schwarzschild case .
a michelson interferometer is useful , not just as a laboratory analog of gravitational lensing interference , but also as a tool for observation of the effect .
microlensed light , collected with a telescope , can be passed through a michelson interferometer before detection .
then , if the interferometer path length difference matches the lensing path length difference , the two arriving wavepackets seen in figure 1 can be realigned , and bandwidth reduction of visibility can be avoided . for slight mismatches of the two path length difference
the required bandwidth for full visibility is @xmath27 .
even though a complex representation of the wave amplitude is used in this discussion of interference , and even though the uncertainty principle @xmath28 applies to the coherence time , the effects described here are classical rather than quantum mechanical .
no use is made , in this analysis , of photon quanta .
the use of high resolution spectroscopy establishes coherence retroactively , allowing gravitational microlensing interference to be visible for path length differences many times larger than the wavelength of the light .
alcock , c. , et.al 2000 , , 542 , 281 born and wolfe deguchi , s. and watson , w. d. 1986 , , 307 , 30 derue , f. , et.al . 1999 , , 351 , 87 gould , a. 1992 , , 386 , 5 gould , a. and gaudi , b.s .
1997 , , 486 , 687 jaroszynski , m. and paczynski , b. 1995 , , 455 , 443 mandzhos , a. v. 1981 , soviet astron .
lett . , 7 , 213 peterson , j. b. and falk , t. 1991 , , 374 , 5 schneider , p. and schmid - burgk , j. 1985 , , 148 , 369 stanek , k. z. , paczynski , b. and goodman , j. 1993 , 413 , 7 ulmer , a. and goodman , j. 1995 , , 442 , 67 wozniak , p. r. , et.al .
2001 , acta astron . | interference two paths for light propagation from source to observer in gravitational microlensing wavetrains from travel two paths wave amplitudes combined before light detected lensing geometry interference applications considered for from radio waves to gamma rays interference visibility diminished if difference length two paths exceeds coherence length light light sources in microlensing incoherent stars and stellar photospheres emit blackbody radiation temporal coherence length light is @xmath0 @xmath1 speed of light @xmath2 coherence time @xmath3 half - power bandwidth of source emission spectrum large bandwidth blackbody spectrum coherence length @xmath4 equal to peak wavelength blackbody emission labeled incoherent modeled as continuous stream of short wavepackets figure 1 shows wavepackets of length @xmath4 traveling on two paths from source to observer around lensing mass wavepackets overlap at arrival short coherence length light gravitational lensing interference undetectable except path length difference few wavelengths not correct coherence can established retroactivelyincoherent light through interference apparatus filtered to narrow bandwidth longer coherence length allows gravitational microlensing interference for large path length differences present results laboratory experiment michelson interferometer retroactive coherence analysis of partial coherence for two path geometries in @xcite interference michelson defined interference visibility @xmath5 maximum minimum intensity @xmath6 @xmath7 taken as path length difference @xmath8 varies @xmath9 ranges 0 to 1 maximum at zero path difference for two - path gravitational microlensing detected intensity is @xmath10 @xmath11 @xmath12 intensities on two paths @xmath13 complex function time variation arriving field strength on shorter path @xmath14 function shifted time by path delay difference @xmath15 normalized assumed stationary @xmath17 duration of observation intensity averaged longer than @xmath2 @xmath18 total intensity @xmath19 last term interference part of complex degree coherence @xmath20 optical frequency band at @xmath21 magnitude degree coherence related to interference visibility by @xmath22autocorrelation wave amplitude @xmath23 is transform power spectral density @xmath24 light coherence length width function @xmath25 decreasing bandwidth forcing @xmath24 to zero frequencies increases coherence length makes interference visible at large path length differences experiment shows bandwidth restriction after light through interference apparatus set laboratory analog gravitational lensing interference michelson interferometer figure 2 light from thermal source arrives at partially reflecting beam splitter divides light two paths paths orthogonal arms interferometer mirrors return light with path length difference adjusted splitter recombines light before exit aperture filtering light restores interference visibility gravitational lensing path length varies with time motion source lensing masses observer michelson interferometer path lengths stable adjusted various tilted one mirror zero path length difference at line dividing pattern vertically divided horizontally top half exit aperture unblocked full bandwidth thermal source contributes to interference pattern bottom half covered by narrow - band @xmath26 ) filter figure 3 results for white light visibility falls almost to zero fringesfiltered light visibility remains near one entire width pattern frequency domain coherence restored through bandwidth restriction white light interference pattern many narrow - band patterns overlaid spacing fringes proportional to wavelength all patterns aligned in center zero path difference central fringes have high visibility pattern develops colored stripes blue patterns edge visibility nearly to zero cancellation constructive destructive interference selecting narrow range frequencies cancellation avoided interference visible across entire field shown in right panel figure 3 puzzle short coherence length wave packets create visible interference at large path length differences ? filter stores delayed copies wave train interference occurs among copies light storage in filter illustrated in figure 4 experiment single narrow filter passband observation optical power blocked by filter gravitational microlensing observation spectrophotometer could be used instead of filter spectrum into narrow passbands each as narrow filter allows all optical power spectrophotometer coherence time @xmath2 reciprocal of frequency resolution temporal coherence increased high spectral resolution limited _ lateral _ spatial coherence lensing interference visibility for most geometriespath length difference varies across disk source star total variation many wavelengths reduces interference visibility microlensing events include caustic crossings first crossing two new high magnification paths appear paths same length first laterally coherent region larger part source disk than schwarzschild case michelson interferometer useful gravitational lensing interference tool for observation microlensed light passed through interferometer before detection if interferometer path length difference matches lensing path length difference two arriving wavepackets realigned bandwidth reduction avoided for slight required bandwidth for full visibility is @xmath27 complex representation wave amplitude uncertainty principle @xmath28 applies effects classical quantum mechanical no use of photon quanta high resolution spectroscopy establishes coherence retroactively gravitational microlensing interference visible for path length differences larger than wavelength alcock ,. et.al 2000 , , 542 , 281 born wolfe deguchi watson 1986 , , 307 , 30 derue 1999 , , 351 , 87 gould 1992 , , 386 , 5gaudi 1997 486 , 687 jaroszynski paczynski b. 1995 455 , 443 mandzhos 1981 soviet astron 7 , 213 peterson falk 1991 374 , 5 schneider schmid - burgk 1985 148 , 369 stanek paczynski goodman 1993 413 , 7 ulmer goodman 1995 , 67 wozniak 2001 acta astron | 0.349746 | 2,229 | 1,107 | , Saving $0.1 in GPT-4. |
20,557 | the @xmath0 process was studied in the snd experiment at vepp-2 m @xmath1 collider in the energy region 0.600.97 gev . from the analysis of the energy dependence of measured cross section
the branching ratios @xmath2 and @xmath3 were obtained . +
* pacs : 13.65.+i,14.40.cs * | in 1998 and 2000 the experiments with spherical neutral detector ( snd)@xcite at vepp-2 m @xmath1 collider were carried out in the energy range @xmath4 mev where cross section of @xmath1 annihilation into hadrons is determined by the @xmath5 and @xmath6 meson decays .
the integrated luminosity of 9 pb@xmath7 collected in the experiment corresponds to @xmath8 and @xmath9 produced @xmath5 and @xmath6 mesons , respectively .
one of the goals of the experiment was the investigation of the rare process @xmath10 our preliminary study @xcite of the process ( [ ppg ] ) was based on 1/3 of collected statistics .
its results were the first measurement of @xmath11 and the measurement of @xmath12 confirming the only previous measurement of this decay by gams : @xmath13 @xcite .
the theoretical study of the @xmath14 decays was begun by p.singer in ref .
@xcite where the transitions via @xmath15 ( fig .
[ diad]a ) and @xmath16 intermediate states were suggested . the vector meson dominance ( vmd )
calculation with these intermediate states leads to branching ratios @xmath17 and @xmath18 for @xmath19 and @xmath20 , respectively @xcite .
.the branching ratios for different contributions to @xmath19 decay .
the chiral loop or @xmath21 contribution is calculated in the framework of chiral perturbation theory ( @xmath22 ) @xcite , unitarized chiral perturbation theory ( @xmath23 ) @xcite , linear sigma model ( @xmath24 ) @xcite and @xmath25 pole model @xcite . [ cols="<,^,^,^,^,^,^,^",options="header " , ] the spreads in parameter values correspond to different models describing @xmath26 cross section above 1 gev .
all models reproduce the experimental data well .
therefore parameter midrange was taken as a final result .
its spread was regarded as the model error .
the branching ratios obtained this way with statistical and systematic errors are listed in the last row of table [ t4 ] .
the systematic error includes the model error , uncertainties in the detection efficiency and integrated luminosity .
the variation of the background level within its systematic error practically does not change the @xmath27 central value and results in following additional uncertainties of the @xmath5 meson branching ratios : 7% for @xmath28 and 12% for @xmath29 .
since these uncertainties affect statistical significance of the results they were added to statistical errors .
the energy dependence of the cross section in the model with @xmath30 mev and @xmath31 mev is shown in fig .
[ f6 ] together with the curve corresponding to @xmath32 .
the @xmath33 value for the latter model is equal to 0.5% .
from table [ t1 ] it is seen that selected events are mainly concentrated in two energy regions : 180 events near @xmath6 peak and 92 events in the range 920970 mev above the @xmath26 reaction threshold .
the angular and energy distributions in the latter region agree with @xmath15 mechanism .
meson in the @xmath34 rest frame .
point with error bars are experimental data , solid line is a simulation with @xmath6 decay into pure @xmath16 state .
dashed line is a simulation with @xmath6 decaying into a mixture of @xmath16 and @xmath35 states .
, scaledwidth=98.0% ] meson in the @xmath34 rest frame .
point with error bars are experimental data , solid line is a simulation with @xmath6 decay into pure @xmath16 state .
dashed line is a simulation with @xmath6 decaying into a mixture of @xmath16 and @xmath35 states .
, scaledwidth=98.0% ] recoil photon spectrum for events from 760800 mev energy range is shown in fig.[f3 ] . although this energy region is dominated by @xmath6 peak the contributions of both @xmath6 and @xmath5
decays must be taken into account to obtain the theoretical spectrum .
the spectrum calculated in model 3 from table [ t4 ] ( fig.[f3 ] ) is in a good agreement with the experimental one . in this model
it is supposed that @xmath20 decay proceeds through @xmath16 intermediate state .
another way is to describe the @xmath6 decay by a sum of contributions of @xmath36 and @xmath37 mechanisms . to do this
we fix @xmath38 at @xmath39 and fit the @xmath37 decay contribution to a value yielding observed @xmath20 branching ratio . in case of constructive interference
this leads to @xmath40 and the photon spectrum close to that expected for @xmath16 mechanism . as was shown in the previous section , assumption of destructive interference results in @xmath27 inconsistent with the pdg table value .
the second theoretical spectrum in fig.[f3 ] corresponds to the model 2 from table [ t3 ] with destructive interference of @xmath41 and @xmath35 amplitudes in @xmath5 decay . for this model
the consistency between theoretical and experimental spectra calculated using kolmogorov test @xcite is about 1% , which was one of the reasons to discard this model .
additional information about mechanism of @xmath6 decay can be obtained from the analysis of angular distributions .
one of such distribution is shown in fig.[f4 ] .
the same figure displays the theoretical distributions obtained under assumptions that @xmath6 decay proceeds through either pure @xmath16 intermediate state or a mixture of @xmath16 and @xmath35 .
one can see that our limited statistics does not allow to distinguish these two models .
the branching ratios measured in this work , @xmath42 @xmath43 are in a good agreement with our preliminary results @xcite and gams measurement @xmath44 @xcite , but have higher accuracy .
the probability of @xmath19 decay significantly exceeds vmd model prediction @xmath45 .
this excess can be explained by the contribution of the decay via scalar state @xmath21 .
two mechanisms , @xmath21 and @xmath46 , can be separated using difference in energy dependence of their amplitudes .
our result on @xmath47 decay @xmath48 differs from zero by 2.4 standard deviations and is consistent with the predictions of chiral models @xcite .
the magnitude of @xmath49 is sensitive to @xmath25 parameters .
as can be seen from table [ t4 ] , the models with @xmath50 mev give the most consistent description of the experimental data .
the value of the branching ratio @xmath51 , obtained by averaging of our measurement with the gams result exceeds theoretical predictions , @xmath52 @xcite and @xmath53 @xcite , by 1.3 standard deviations .
it is necessary to make some remarks about these predictions .
the result of ref .
@xcite is based on table value of @xmath54 kev @xcite .
it must be corrected taking into account newer measurement @xmath55 kev @xcite , which is close to the value for charged @xmath5 , @xmath56 kev @xcite .
this decreases the predicted @xmath27 and worsens agreement with the experiment . in the refs .
@xcite and @xcite the values of @xmath57 equal to 15 gev@xmath7 and 15.9 gev@xmath7 were used to calculate @xmath58 . on the other hand
, the use of these @xmath57 values for vmd calculation of @xmath59 and @xmath60 leads to too large values conflicting with experimental data .
for example , the @xmath57 obtained from @xmath61 width assuming intermediate @xmath62 state is equal to @xmath63 gev@xmath7 @xcite .
therefore , our opinion is that @xmath64 is the maximum branching ratio acceptable within vmd model and the additional theoretical study is required to explain the large value of @xmath27 .
this work is supported by `` russian fund for basic researches '' grants no .
01 - 02 - 16934-a and 00 - 15 - 96802 and stp `` integration '' fund , grant no . a0100 .
achasov et al . , nucl .
meth . a 449 ( 2000 ) 125 .
achasov et al . , jetp letters , 71 ( 2000 ) 355 .
d. alde et al . , phys .
b 340 ( 1994 ) 122 .
p. singer , phys .
( 1962 ) 2789 .
a. bramon , a. grau , g. pancheri , phys .
b 283 ( 1992 ) 416 . a. bramon , a. grau , g. pancheri , phys .
b 289 ( 1992 ) 97 .
a. bramon et al . , phys .
b 517 ( 2001 ) 345 . e. marco et al . , phys . lett .
b 470 ( 1999 ) 20 .
j. e. palomar , s. hirenzaki , e. oset , e - print hep - ph/0111308 .
a. gokalp , o. yilmaz , phys .
b 508 ( 2001 ) 25 .
d. guetta , p. singer , phys .
d 63 ( 2001 ) 017502 .
a. gokalp , o. yilmaz , phys .
b 494 ( 2000 ) 69 . a.v .
bozhenok , v.n .
ivanchenko , z.k .
silagadze , nucl .
meth . a 379 ( 1996 ) 507 .
achasov et al . , nucl . phys .
b 569 ( 2000 ) 158 . e.a .
kuraev , v.s .
fadin , sov .
41 ( 1985 ) 466 .
clegg , a. donanachie , z. phys .
c 62 ( 1992 ) 455 .
achasov et al . , phys .
b 486 ( 2000 ) 29 . k.w .
edwards et al .
, phys . rev .
d 61 ( 2000 ) 072003 .
tsai , phys .
d 4 ( 1971 ) 2821 .
r. barate et al . , z. phys .
c 76 ( 1997 ) 15 .
s. anderson et al . , phys . rev .
d 61 ( 2000 ) 112002 .
aitala et al . , phys .
86 ( 2001 ) 770 .
asner et al . , phys .
d 61 ( 2000 ) 012002 .
hbook reference manual , cern program library y250 , 1998 , p.90 .
review of particle physics .
j. c 15 ( 2000 ) 1 . m.n .
achasov et al . , e - print hep - ex/0109035 | 1998 2000 experiments with spherical neutral detector@xcite at vepp-2 m @xmath1 collider energy range @xmath4 mev cross section @xmath1 annihilation into hadrons determined by @xmath5 @xmath6 meson decays integrated luminosity 9 pb@xmath7 corresponds to @xmath8 @xmath9 produced @xmath5 mesons rare process @xmath10 preliminary study based on 1/3 statistics results first measurement @xmath11 @xmath12 previous measurement decay by @xmath13 theoretical study @xmath14 decays begun by p.singer transitions via @xmath15 @xmath16 intermediate states suggested vector meson dominance calculation leads to branching ratios @xmath17 @xmath18 for @xmath19 @xmath20 branching ratios contributions to @xmath19 decay chiral loop @xmath21 contribution calculated chiral perturbation theory linear sigma model pole model spreads in parameter values correspond different models describing @xmath26 cross section above 1 gev all models reproduce experimental data parameter midrange final resultspread model error branching ratios with statistical systematic errors listed in last row table [ t4 systematic error includes model error uncertainties detection efficiency integrated luminosity variation background level change @xmath27 central value results in additional uncertainties @xmath5 meson branching ratios : 7% for @xmath28 12% @xmath29 uncertainties affect statistical significance added to statistical errors energy dependence of cross section with @xmath30 @xmath31 mev shown in fig . [ f6 ] curve @xmath32 @xmath33 value equal to 0.5% table t1 selected events concentrated in two energy regions : 180 events near @xmath6 peak 92 events range 920970 mev above @xmath26 reaction threshold angular energy distributions region agree with @xmath15 mechanism @xmath34 rest frame simulation @xmath6 decay into @xmath16 state mixture @xmath16 @xmath35 states=98.0% @xmath6 decay state mixture states recoil photon spectrum for events from 760800 mev energy range shown in fig.[f3 ]energy region dominated by @xmath6 peak contributions @xmath6 decays theoretical spectrum spectrum calculated in model 3 fig.[f3 with experimental one @xmath20 decay proceeds through @xmath16 intermediate state @xmath6 decay by sum of contributions @xmath36 @xmath37 mechanisms fix @xmath38 at @xmath39 fit @xmath37 decay contribution to @xmath20 branching ratio constructive interference leads to @xmath40 photon spectrum close to @xmath16 mechanism destructive interference results in @xmath27 inconsistent with pdg table value second theoretical spectrum in fig.[f3 ] corresponds to model 2 with destructive interference of @xmath41 @xmath35 amplitudes in @xmath5 decay consistency between theoretical experimental spectra about 1% model information @xmath6 decay from angular distributions in fig.[f4 ] theoretical distributions @xmath6 decay proceeds through @xmath16 intermediate state or mixture of @xmath16 @xmath35 limited statistics distinguish models branching ratios measured @xmath42 @xmath43 with preliminary results @xmath44 have higher accuracyprobability @xmath19 decay exceeds model prediction @xmath45 excess explained by contribution decay via scalar state @xmath21 mechanisms @xmath21 @xmath46 separated energy dependence amplitudes result @xmath47 decay @xmath48 differs zero 2.4 standard deviations consistent with predictions chiral models @xcite magnitude @xmath49 sensitive to @xmath25 parameters models with @xmath50 mev consistent experimental data branching ratio @xmath51 exceeds predictions @xmath52 @xmath53 1.3 standard deviations result ref @xcite based on table value @xmath54 kev @xcite corrected newer measurement @xmath55 kev @xcite close to value @xmath5 @xmath56 kev decreases predicted @xmath27 worsens agreement experiment values @xmath57 equal to 15 15.9 gev@xmath7 used calculate @xmath58 use values for calculation @xmath59 @xmath60 leads large values experimental data @xmath57 from @xmath61 @xmath62 state equal to @xmath63 gev@xmath7opinion @xmath64 maximum branching ratio vmd model additional study required value @xmath27 work supported russian fund basic researches grants 01 - 02 - 16934-a 00 - 15 - 96802 stp integration fund grant a0100 achasov al nucl 449 ( 2000 ) 125 jetp letters 71 ( 2000 ) 355 alde phys . b 340 ( 1994 ) 122 singer . 1962 ) 2789 bramon grau pancheri 283 ( 1992 ) 416 289 ( 1992 ) 97 phys 517 ( 2001 ) 345 marco phys . 470 ( 1999 ) 20 palomar hirenzaki oset - ph/0111308 gokalp yilmaz phys . b 508 ( 2001 ) 25 guetta singer . 63 ( 2001 ) 017502 gokalp phys 494 ( 2000 ) 69 bozhenok ivanchenko silagadze nucl 379 ( 1996 ) 507 achasov phys . b 569 ( 2000 ) 158 kuraev fadin41 1985 ) 466 clegg donanachie phys 62 1992 ) 455 achasov phys 486 2000 ) 29 edwards phys 61 2000 ) 072003 tsai 4 ( 1971 ) 2821 barate phys 76 ( 1997 ) 15 anderson phys 61 2000 ) 112002 aitala . 86 ( 2001 ) 770 asner phys 61 2000 ) 012002 hbook reference manual cern program library y250 1998 p.90 review particle physics 15 ( 2000 ) 1 achasov print ex/0109035 | 0.347624 | 2,565 | 1,309 | , Saving $0.1 in GPT-4. |
20,559 | as compared to electron storage rings , one advantage of recirculating linacs is that the beam properties at target are no longer dominated by the equilibrium between quantum radiative diffusion and radiation damping because new beam is continually injected into the accelerator .
this allows the energy spread from a cebaf - type machine to be relatively small ; the measured energy spread from cebaf at 4 gev is less than 100 parts per million accumulated over times of order several days . in this paper ,
the various subsystems contributing to the energy spread of a cebaf - type accelerator are reviewed , as well as the machine diagnostics and controls that are used in cebaf to ensure that a small energy spread is provided during routine running .
examples of relevant developments are ( 1 ) stable short bunches emerging from the injector , ( 2 ) precision timing and phasing of the linacs with respect to the centroid of the beam bunches on all passes , ( 3 ) implementing 2 khz sampling rate feedback systems for final energy stabilization , and ( 4 ) continuous beam energy spread monitoring with optical transition radiation devices .
we present measurement results showing that small energy spreads are achieved over extended periods . | in this paper we summarize the present status on energy spread measurement and control in the jefferson lab nuclear physics accelerator called cebaf .
a schematic diagram of cebaf appears in fig .
[ cebafacc ] and a summary of relevant beam parameters is given in table [ cebafpar ] , where all sizes are _ rms _ quantities .
cw beam , originating in the injector , is recirculated up to five times through each linac .
the beam may be directed into up to three experimental halls simultaneously , the beam current in the halls being at the third subharmonic of the accelerator operating frequency of 1497 mhz . because of the low charge - per - bunch at even the highest operating current , collective effects are not an important source of energy spread in cebaf .
[ cebafacc ] locates some of the feedback system hardware , discussed in section 6 below .
.cebaf accelerator parameters [ cols="<,^,^",options="header " , ] [ momod ]
as mentioned in the introduction , sources of bunch energy centroid fluctuations are corrected by a fast digital feedback system @xcite .
the system corrects beam position and energy near the targets of the nuclear physics experiments utilizing energy measurements obtained from the bend magnets which deliver the beam to the various halls , see fig .
[ cebafacc ] .
the system is capable of suppressing beam motion in the frequency band from 0 to 80 hz and also performs narrow band suppression at the first twelve power line harmonics .
the system operates with a 2.1 khz sampling rate and utilizes two vme board computers to compute the corrections .
energy corrections are fed back as analogue signals to the gradient set points in the rf controls of a few cavities in the linac called vernier cavities . for the standard optics in cebaf , the horizontal dispersion is maximum in the middle of the bend magnets delivering the beam to the halls .
its value is approximately 4 m , meaning position fluctuations at 10 ppm correspond to 40 @xmath0 m of beam motion .
the feedback system suppresses the fluctuations to around 20 @xmath0 m , limited by bpm noise @xcite .
the beam noise to be corrected is primarily at frequencies of 60 hz and its first few harmonics .
because the energy information is so closely tied to the magnetic fields in the beam delivery lines to the halls , a question arises about the stability of the magnetic fields themselves at the 10@xmath1 level . the total magnetic field at several points within the magnets
have been verified to be stable to 10 ppm , and the power supplies deliver current having similarly small fluctuations .
recently , we have installed a magnetic flux loop monitor through the dipole strings to the halls .
this monitor will provide better quantitative information than we currently possess on residual fluctuations in the magnets , and will be able to address the issue of magnet stability directly .
the accelerator is equipped with slow wire scanners using 22 @xmath0 m diameter tungsten wires .
beam profiles for currents in the 2 to 5 @xmath0a range can be accurately measured once a minute with such scanners .
more recently , we have developed a profile monitor that can measure even the most intense beams using forward optical transition radiation ( otr ) @xcite .
a very thin ( 1/4 @xmath0 m ) carbon foil inserted into the beam path is not invasive to physics experiments for most cebaf energies and currents .
presently , otr monitors are installed in each of the experimental hall a and c beam transport lines at the high dispersion points of the beam optics .
these monitors provide the experiments and the accelerator with 5 hz measurement rates for each instrument by using a common image processing hardware . a dedicated software , developed under the epics @xcite control system , multiplexes up to four video input channels connected to a single maxvideo image processing board @xcite .
the global processing speed is 10 hz , 5 hz for each of the two otr monitors .
the hall a otr measures beams in the 1 to 180 @xmath0a operational range ; in hall c , the dynamic range extends down to 0.1 @xmath0a .
the resolution of these monitors is limited by the ccd camera to about 2 pixels .
this amounts to approximately 70 @xmath0 m of _ rms _ beam size .
[ li2000f3 ] shows that the energy spread is relatively stable and below @xmath2 for a wide range of beam currents .
the horizontal size , measured at the 4 m dispersion point , is mostly due to the energy spread . neglecting the betatron beam size , 40 @xmath0 m , and the camera resolution , 70 @xmath0 m , overestimates the actual energy spread by less than 25% .
continuous small energy spread became an operational requirement at cebaf in dec .
1999 , for a hypernuclear experiment housed in hall a , and continued until may 2000 with a similar experiment in hall c. both experiments ran simultaneously during one month last march , with 2-pass beam for hall a and a 4-pass beam for hall c. delivering two beams with tight energy spread and energy stability requirements instead of one proved demanding .
the energy requirements for each experiment were similar : @xmath3 , with energy stability better than @xmath4 .
in addition , hall a needed the transverse beam sizes at the target to be less than 200 @xmath0 m but greater than 100 @xmath0 m and a beam position stable within 250 @xmath0 m .
the energy spread requirements have been routinely achieved for the hall under feedback control . because the feedback system can correct the energy fluctuations only in a single hall as presently configured , there were uncertainties that the spread in the other hall would remain small .
[ li2000f4 ] shows energy spreads and relative energies in the hall c beam recorded over a 2-week period , with hall c transport line providing the energy corrections to the energy feedback system .
small energy spreads were delivered throughout the period to hall c , however drifts led to energy spread increases in the hall a beam , as seen in fig .
[ li2000f4i ] . throughout the experiments in either hall , the energy spread and stability of both beams were continuously recorded .
the otr monitors have been critical in this task .
they were initially too cumbersome to be easily used by all operator crews .
the implementation of scripts that periodically check and adjust the camera illumination , that initialize the image processing board according to the beam , and that set a data valid " flag quickly improved the instrument availability to 95% @xcite .
after these improvements , the machine crews were able to correct quickly unacceptable energy spreads , usually without interrupting beam delivery .
we are planning to improve the energy spread monitoring for two reasons : 1 ) at lower energies ( @xmath51.2 gev ) , the beam current had to be lowered to under 50 @xmath0a to have acceptable radiation levels on sensitive beam - line equipment .
2 ) experiments scheduled in 2002 require monitoring an energy spread of @xmath6 .
as an alternate to otr monitoring , we are planning to use synchrotron light beam monitoring , which is less invasive to the experimenters .
however , the resolution of such a device is limited to about 100 @xmath0 m in the visible using the bending magnets of the hall transport lines .
we are starting a development effort in order to reach about 30 @xmath0 m resolution utilizing the uv synchrotron emission .
we have demonstrated the ability of a cebaf - type accelerator to produce beams with small energy spreads over long periods of time .
we ensure that the energy spread remains small by : ( 1 ) ensuring the bunch length out of the injector is small , ( 2 ) ensuring that the beam remains close to the crest phase on each separate pass ( soon continuously and automatically ! ) , and ( 3 ) providing continuous fast correction of 60 cycle harmonic noise on the beam .
we have developed beam diagnostic devices to continuously monitor and record beam conditions with 5 hz update rates using digitization of multiple video monitors .
r. abbott , _ et .
al . _ , design , commissioning , and operation of the upgraded cebaf injector " , proc . of the 1994 linac conf . , 777 ( 1994 ) g. a. krafft , correcting @xmath7 and @xmath8 to obtain very short bunches at cebaf " , proc . of the microbunches workshop ,
aip conference proceedings 367 , 46 ( 1996 ) c. g. yao , a new scheme for measuring the length of very short bunches at cebaf " , proc . of the 1990 beam instrumentation workshop , aip
conference proceedings 229 , 254 ( 1990 ) g. a. krafft , status of the continuous electron beam accelerator facility " , proc . of the 1994 linac conf . ,
9 ( 1994 ) r. kazimi , c. k. sinclair , and g. a. krafft , setting and measuring the longitudinal optics in cebaf injector " , these proceedings d. x. wang , g. a. krafft , and c. k. sinclair , measurement of femtosecond electron bunches using a rf zero - phasing method " , phys .
e , * 57 * , 2283 ( 1998 ) a detailed calculation is performed in an internal note , g. a. krafft , j. j. bisognano , and r. miller , revised energy spread estimate " , cebaf - tn-0050 ( 1987 ) .
a similar calculation is performed in l. merminga and g. a. krafft , energy spread from rf amplitude and phase errors " , proc . of the 1996 european part .
conf . , 756 ( 1996 ) g. a. krafft , _ et .
_ , energy vernier system for cebaf " , proc . of the 1993 part .
, 2364 ( 1993 ) summarizes the results in the internal note @xcite above .
s. n. simrock , _ et .
_ , operation of the rf controls in the cebaf injector " , proc . of the 2nd european part .
, 824 ( 1990 ) y. chao , _ et .
_ , commissioning and operation experience with the cebaf recirculation arc beam transport system " , proc .
of the 1993 part .
conf . , 587 ( 1993 ) g. a. krafft , _ et .
_ , measuring and adjusting the path length at cebaf " , proc . of the 1995 part .
, 2429 ( 1995 ) d. hardy , _ et .
_ , automated measurement of path length and @xmath7 " proc . of the 1997 part .
conf . , 2265 ( 1997 ) v. a. lebedev , j. musson , and m. g. tiefenback ,
high- precision beam - based rf phase stabilization at jefferson lab " , proc . of the 1999 part .
conf . , 1183 ( 1999 ) m. g. tiefenback and k. brown , beam - based phase monitoring and gradient calibration of jefferson laboratory rf systems " , proc . of the 1997 part .
conf . , 2271 ( 1997 ) r. dickson and v. a. lebedev , fast digital feedback system for energy and beam position stabilization " , proc .
of the 1999 part .
conf . , 646 ( 1999 ) p. piot , _ et .
_ , high current cw beam profile monitors using transition radiation at cebaf " , proc . of the 1996 beam instrumentation workshop ,
aip conference proceedings 390 , 298 - 305 ( 1997 ) j .- c .
denard , _ et .
_ , high power beam profile monitor with optical transition radiation " , proc . of the 1997 part .
conf . , 2198 ( 1997 )
vancouver , bc .
l. dalesio , _ et .
conf . on accelerators and large experimental physics control systems ( icaleps ) ,
278 ( 1992 ) d. hardy , _ et . al .
_ , multivideo source image processing for beam profile monitoring system " , proc . of the 2000 beam instrumentation workshop ,
aip proceedings to be published . | paper status energy spread measurement control jefferson lab nuclear physics accelerator cebaf schematic diagram in fig . summary beam parameters table [ cebafpar sizes rms quantities cw beam originating injector recirculated five times through each linac beam directed into three experimental halls current third subharmonic accelerator operating frequency 1497 mhz low charge - per - bunch current collective effects not important source energy spread cebaf cebafacc locates feedback system hardware discussed section 6 accelerator parameters bunch energy centroid fluctuations corrected by fast digital feedback system @xcite corrects beam position energy near targets nuclear experiments energy measurements bend magnets halls fig . system beam motion 0 to 80 hz performs narrow band suppression at first twelve power line harmonics operates 2.1 khz sampling rate utilizes two vme board computers compute corrections energy corrections fed back as analogue signals to gradient set points rf controls linac vernier cavities horizontal dispersion maximum in middle of bend magnets delivering beam hallsvalue approximately 4 m position fluctuations at 10 ppm correspond to 40 @xmath0 m beam motion feedback system suppresses fluctuations to 20 @xmath0 m limited by bpm noise @xcite beam noise corrected at 60 hz first harmonics energy information tied to magnetic fields beam delivery lines question about stability fields at 10@xmath1 level total magnetic field at magnets stable to 10 ppm power supplies deliver current small fluctuations installed magnetic flux loop monitor through dipole strings information on residual fluctuations magnet stability accelerator equipped with slow wire scanners 22 @xmath0 m diameter tungsten wires beam profiles for currents 2 to 5 @xmath0a range measured once a minute developed profile monitor intense beams using forward optical transition radiation thin ( 1/4 @xmath0 m ) carbon foil beam not invasive to experiments otr monitors installed in experimental hall a c beam transport lines high dispersion points monitors provide 5 hz measurement rates common image processing hardware software multiplexes four video input channels single maxvideo image processing boardprocessing speed 10 hz 5 hz two otr monitors hall a measures beams 1 to 180 @xmath0a range hall c range 0.1 @xmath0a resolution limited ccd camera 2 pixels approximately 70 @xmath0 m beam size energy spread stable below @xmath2 beam currents horizontal size 4 m dispersion point due to energy spread betatron beam size 40 @xmath0 m camera resolution 70 @xmath0 m overestimates energy spread less than 25% continuous small energy spread requirement cebaf 1999 hypernuclear experiment hall a continued until may 2000 experiment hall c experiments ran simultaneously 2-pass beam a 4-pass beam c two beams tight energy spread stability demanding energy requirements similar @xmath3 stability better than @xmath4 hall a needed transverse beam sizes less than 200 @xmath0 m greater than 100 @xmath0 m beam position stable within 250 @xmath0 m energy spread requirements achieved feedback control feedback system energy fluctuations single hall uncertainties spread other hall small energy spreads energies hall c beam 2-week period hall c transport line energy corrections feedback systemenergy spreads delivered to hall c drifts led to energy spread increases in hall a beam seen in fig . [ li2000f4i ] experiments energy spread stability of both beams continuously recorded otr monitors critical initially cumbersome implementation of scripts camera illumination initialize image processing board according to beam set data valid " flag improved instrument availability to 95% @xcite machine crews unacceptable energy spreads without interrupting beam delivery planning to improve energy spread monitoring for at lower energies ( @xmath51.2 gev beam current to under 50 @xmath0a acceptable radiation levels on equipment experiments in 2002 require monitoring energy spread of @xmath6 alternate to planning to use synchrotron light beam monitoring less invasive resolution limited to 100 @xmath0 m starting development effort to reach 30 @xmath0 m resolution uv synchrotron emission demonstrated cebaf - type accelerator to produce beams with small energy spreads over long ensure energy spread small by bunch length injector small beam close to crest phase each providing continuous correction of 60 cycle harmonic noisedeveloped beam diagnostic devices monitor record beam conditions 5 hz update rates digitization video monitors r abbott design commissioning operation upgraded cebaf injector 1994 linac conf . 777 1994 g a. krafft correcting @xmath7 @xmath8 short bunches cebaf microbunches workshop conference proceedings 367 46 1996 yao new scheme measuring length short bunches cebaf 1990 beam instrumentation workshop 229 254 1990 krafft continuous electron beam accelerator facility 1994 linac conf . 9 1994 ) kazimi sinclair krafft setting measuring longitudinal optics cebaf injector wang krafft sinclair measurement femtosecond electron bunches rf zero - phasing method 2283 1998 detailed calculation internal note a. krafft bisognano r. miller revised energy spread estimate cebaf - tn-0050 1987 ) similar calculation l. merminga g a. krafft energy spread rf amplitude phase errors 1996 european conf . 756 ( 1996 ) krafft energy vernier system cebaf 1993 part2364 ( 1993 ) summarizes results internal note simrock operation rf controls cebaf injector 2nd european part 824 ( 1990 ). chao commissioning operation experience cebaf recirculation beam transport system 1993 conf 587 ( 1993 ) g. a. krafft measuring adjusting path length cebaf 1995 2429 ( 1995 ) d. hardy automated measurement path length 1997 conf 2265 ( 1997 ). lebedev musson tiefenback high- precision beam - based rf phase stabilization jefferson lab 1999 1183 ( 1999 ) tiefenback. brown beam - based phase monitoring gradient calibration jefferson laboratory rf systems 1997 2271 ( 1997 ) r. dickson v. lebedev digital feedback system energy beam position stabilization 1999 conf 646 ( 1999 ) p. piot high current beam profile monitors transition radiation cebaf 1996 beam instrumentation workshop conference proceedings 298 - 305 ( 1997 ) denard high power beam profile monitor optical transition radiation 1997 conf2198 1997 ) vancouver l. dalesio conf accelerators large experimental physics control systems 278 1992 ) d. hardy multivideo image processing beam profile monitoring system 2000 beam instrumentation workshop proceedings published | 0.44143 | 2,798 | 1,387 | , Saving $0.1 in GPT-4. |
20,560 | the three - nucleon bound state problem is studied employing nucleon - nucleon potentials derived from a basic quark - quark interaction . we analyze the effects of the nonlocalities generated by the quark model .
the calculated triton binding energies indicate that quark - model nonlocalities can yield additional binding in the order of few hundred kev . during the last decade the development of quark - model based interactions for the hadronic force
has led to nucleon - nucleon ( @xmath0 ) potentials that provide a fairly reliable description of the on - shell data . as a consequence of the internal structure of the nucleon ,
such interaction models are characterized by the presence of nonlocalities .
these nonlocalities are reflected in the off - shell properties and emerge from the underlying dynamics in a natural way .
the relevance and/or necessity of considering the nonlocal parts of nucleon - nucleon potentials in realistic interactions is still under debate .
indeed , over the past few years several studies have appeared in the literature which stress the potential importance of nonlocal effects for the quantitative understanding of few - body observables and , specifically , for the triton binding energy @xcite .
however , the majority of these investigations @xcite explore only nonlocalities arising from the meson - exchange picture of the @xmath0 interaction .
the effects of nonlocalities resulting from the quark substructure of the nucleon have only been addressed once so far @xcite and more systematic studies are lacking altogether . in this work
we study the triton binding energy by means of a local and a nonlocal potential , derived from the same constituent quark - model
. we will pay special attention to the nonlocal effects originating from the quark model .
the nonlocal potential is derived by means of the resonating group method ( rgm ) and the local one is obtained through the born - oppenheimer approximation .
these interactions are employed in faddeev calculations of the three - nucleon binding energy . | the underlying idea of the quark model we use is that the constituent quark mass is a consequence of the spontaneous chiral symmetry breaking ( scsb ) .
then , between the chiral symmetry breaking scale ( @xmath1 1 gev ) and the confinement scale , ( @xmath2 0.2 gev ) qcd may be simulated in terms of an effective theory of constituent quarks interacting through the goldstone modes associated with the scsb .
perturbative features of qcd are incorporated through the one - gluon exchange potential .
a more extensive description of the quark - model hamiltonian can be found in the literature @xcite . based on the quark model hamiltonian , two different procedures have been used in the literature to obtain baryonic interactions .
the first one is the rgm .
it allows to treat the inter - cluster dynamics in an exact way once the hilbert space has been fixed .
the potential derived from this method contains all the nonlocalities associated with quark antisymmetry . for the present study we will make use of the @xmath0 potential derived through a lippmann - schwinger formulation of the rgm equations in momentum space @xcite .
it reproduces the @xmath0 phase shifts up to 250 mev lab energy with quite a good accuracy .
the second method is based on the born - oppenheimer approximation .
it provides a clear - cut prescription for removing the nonlocalities while preserving the general properties of the interaction for lower partial waves , i.e. , those coming form quark antisymmetry .
this local interaction has been widely applied to a great variety of physical problems , obtaining reasonable results . in general , the phase shifts are reproduced with a comparable accuracy to the rgm results @xcite . in both cases , for a correct description of the @xmath3 phase shift it is necessary to take into account the coupling to the @xmath4 channel @xcite , providing the required additional attraction . in order to achieve almost phase shift equivalence between the local and nonlocal interaction models , which is mandatory
if one wants to reliably judge the influence of the nonlocalities , we have done a fine tuning of the potential parameters .
the results obtained for the two - body system with the local and nonlocal potentials are presented in table [ table2 ] and figs .
[ fig1 ] and [ fig2 ] .
the @xmath3 and the @xmath5 phase shifts and the low - energy scattering parameters as well as the deuteron binding energy are practically the same for both potential models , and also in very good agreement with experimental data .
the three - body system is solved performing a five channel fadeev calculation including the @xmath3 and @xmath5 @xmath0 partial waves as input .
note that since in our model there is a coupling to the @xmath6 system , a fully consistent calculation would require the inclusion of two more three - body channels . however , their contribution to the @xmath7 binding energy is known to be rather small @xcite and therefore we neglect them for simplicity reasons . to solve the three - body faddeev equations in momentum space
we first perform a separable finite - rank expansion of the @xmath8 sector utilizing the est method @xcite . in ref .
@xcite it was shown that with a separable expansion of sufficiently high rank , reliable and accurate results on the three - body level can be achieved . in the present case it turned out that separable representations of rank 6 - 8 for @xmath9 and rank 6 for @xmath5 , are sufficient to get converged results .
the results of the triton bound state calculation are summarized in table [ table4 ] .
let us first emphasize that the predicted binding energies for both models are comparable to those obtained from conventional potentials of the @xmath0 interaction such as the paris , bonn , or nijmegen models @xcite . comparing the values for our local and nonlocal models
, one observes that there is about 150 kev more binding for the nonlocal potential .
is this the enhancement we can expect from the nonlocalities due to the quark substructure of the nucleon ? in order to answer this question we need to go back again to the @xmath0 results and scrutinize the on - shell properties carefully . for the @xmath3 partial wave
the differences in the low - energy scattering parameters and in the phase shift are indeed very small , see table [ table2 ] and fig .
[ fig1 ] respectively .
unfortunately , for the @xmath5 partial wave the situation is much more complicated . while the deuteron binding energy and also the @xmath10 and @xmath11 phase shifts are in excellent agreement , ( see fig .
[ fig2 ] ) this can not be said about the mixing parameter @xmath12 . in this case , it is difficult to estimate reliably the effect from the obvious deviation from phase equivalence on the triton binding energy . however , one can clearly separate the effects from the two involved partial waves . for this purpose
, we carried out additional @xmath7 calculations where we combined the @xmath3 of the local model with the @xmath5 of the nonlocal model and vice versa .
corresponding results are compiled in table [ table5 ] where we show triton binding energies ( in mev ) for different combinations of the local and nonlocal models .
they strongly suggest that the nonlocalities present in the @xmath3 alone are already responsible for the enhancement of around 150 kev in the triton binding energy . the shift in the binding energy is independent of whether we use the local or nonlocal version model for the @xmath5 partial wave . on the other hand , the nonlocalities present in the @xmath5 partial wave seem to even decrease the binding energy .
however , we suspect that here the effect of the nonlocalities is obscured by the fact that the two models are not strictly phase equivalent . in summary , we have calculated the three - nucleon bound state problem utilizing @xmath0 potentials derived from a basic quark - quark interaction .
one of these potentials was generated by means of the resonating group method so that nonlocalities resulting from the internal structure of the nucleon were preserved .
the other potential is based on the born - oppenheimer approximation and is strictly local .
these potentials are made nearly phase equivalent by fine tuning of some of the model parameters .
the corresponding calculations of the triton bound state indicate that the nonlocalities resulting from the quark substructure of the nucleon yield additional attraction and , specifically , can lead to an increase of the binding energy by up to 200 kev .
thus , the effect of those nonlocalities on the three - nucleon binding energy is certainly appreciable .
in particular , it is of the same magnitude as the one resulting from nonlocalities that occur in the meson - exchange picture of the @xmath0 interaction . | quark model constituent quark mass consequence spontaneous chiral symmetry breaking between chiral symmetry breaking scale @xmath1 1 gev ) and confinement scale @xmath2 0.2 gev ) qcd simulated constituent quarks interacting through goldstone modes scsb perturbative features of qcd incorporated through one - gluon exchange potential extensive description quark - model hamiltonian in literature @xcite two procedures used baryonic interactions first rgm inter - cluster dynamics once hilbert space fixed potential contains nonlocalities quark antisymmetry present study @xmath0 potential through lippmann - schwinger formulation rgm momentum space @xcite reproduces @xmath0 phase shifts up to 250 mev lab energy good accuracy second method born - oppenheimer approximation removing nonlocalities preserving properties interaction for lower partial waves quark antisymmetry local interaction applied to physical problems reasonable results phase shifts reproduced comparable accuracy to rgm results @xcite for correct description @xmath3 phase shift coupling to @xmath4 channel @xcite additional attractionachieve phase shift equivalence between local nonlocal interaction models mandatory influence potential parameters results for two - body system local nonlocal potentials presented in table table2 figs @xmath3 @xmath5 phase shifts low - energy scattering parameters deuteron binding energy same for both potential models agreement with experimental data three - body system solved five channel fadeev calculation including @xmath3 @xmath5 @xmath0 partial waves input model coupling to @xmath6 system calculation two more three - body channels contribution to @xmath7 binding energy small neglect them for simplicity solve three - body faddeev equations momentum space separable finite - rank expansion of @xmath8 sector est method @xcite separable expansion high rank reliable accurate results on three - body level achieved separable representations of rank 6 - 8 for @xmath9 rank 6 for @xmath5 sufficient converged results results triton bound state calculation summarized in table table4 ] predicted binding energies for both models comparable to conventional potentials of @xmath0 interactioncomparing values local nonlocal models 150 kev more binding for nonlocal potential enhancement from nonlocalities due to quark substructure nucleon ? need to @xmath0 results scrutinize on - shell properties @xmath3 partial wave differences in low - energy scattering parameters phase shift small see table table2 fig fig1 ] @xmath5 partial wave situation complicated deuteron binding energy @xmath10 @xmath11 phase shifts not mixing parameter @xmath12 difficult to estimate effect from deviation from phase equivalence on triton binding energy separate effects from two partial waves carried additional @xmath7 calculations combined @xmath3 local model with @xmath5 nonlocal model results compiled in table table5 ] show triton binding energies mev for different combinations local nonlocal models suggest nonlocalities in @xmath3 responsible for enhancement 150 kev in triton binding energy shift binding energy independent of local or nonlocal version model for @xmath5 partial wave nonlocalities in @xmath5 partial wave decrease binding energy suspect effect nonlocalities obscured by two models not strictly phase equivalentcalculated three - nucleon state problem utilizing @xmath0 potentials from quark - quark interaction one generated resonating group method nonlocalities from internal structure nucleon preserved other potential based on born - oppenheimer approximation strictly local potentials made nearly phase equivalent by fine tuning model parameters calculations indicate nonlocalities from quark substructure nucleon yield additional attraction increase binding energy up to 200 kev effect on three - nucleon binding energy appreciable same magnitude as nonlocalities in meson - exchange picture interaction | 0.263909 | 1,476 | 758 | , Saving $0.0 in GPT-4. |
20,575 | the nonminimal coupling of the kinetic term to einstein s tensor helps the implementation of inflationary models due to the gravitationally enhanced friction .
we calculate the parametrized post - newtonian ( ppn ) parameters for the scalar
tensor theory of gravity with nonminimally derivative coupling .
we find that under experimental constraint from the orbits of millisecond pulsars in our galaxy , the theory deviates from einstein s general relativity in the order of @xmath0 , and the effect of the nonminimal coupling is negligible if we take the scalar field as dynamical dark energy . with the assumed conditions that the background scalar field is spatially homogeneous and evolves only on cosmological timescales and the contribution to stress
energy in the solar system from the background scalar field is subdominant , the scalar field is required to be massless . | the simplest generalization of einstein s general relativity is the scalar tensor theory of gravity such as brans dicke theory @xcite . in brans dicke theory , in addition to the massless spin two graviton , a massless scalar field is also responsible for the exchange of gravitational interaction .
the scalar field called dilaton appears in the low energy effective bosonic string theory , and the scalar degree of freedom arises naturally upon compactification of higher dimensions .
the most general scalar tensor theory of gravity which gives at most the second order equation of motion in four
dimensional spacetime is horndeski theory @xcite with the following lagrangian , @xmath1 where @xmath2,\\ \begin{split } l_5=&g_5(\phi , x)g_{\mu\nu}\nabla^\mu\nabla^\nu\phi-\frac{1}{6}g_{5,x}\left[(\box\phi)^3\right.\\ & \left.-3(\box\phi)(\nabla_\mu\nabla_\nu\phi)(\nabla^\mu\nabla^\nu\phi ) + 2(\nabla^\mu\nabla_\alpha\phi)(\nabla^\alpha\nabla_\beta\phi)(\nabla^\beta\nabla_\mu\phi)\right ] , \end{split}\end{gathered}\ ] ] @xmath3 , @xmath4 , the functions @xmath5 , @xmath6 , @xmath7 and @xmath8 are arbitrary functions of @xmath9 and @xmath10 , and @xmath11 .
the theory not only includes the nonminimal coupling @xmath12 , but also includes the nonminimal coupling between the kinetic term @xmath10 and @xmath13 .
for the special nonminimal coupling @xmath14 @xcite , the theory is conformal invariant . in general , there is no conformal invariance .
however , if we only consider the nonminimal interactions which are quadratic in @xmath9 and linear in @xmath13 , then we only need to consider the nonminimally derivative terms @xmath15 , @xmath16 , @xmath17 , @xmath18 , @xmath19 and @xmath20 @xcite .
because of the divergencies of @xmath21 , @xmath22 and @xmath23 , the independent interactions are @xmath15 , @xmath16 and @xmath17 .
the inflationary solution was discussed in @xcite for the nonminimal coupling @xmath24r$ ] , but the equation of motion for the scalar field is fourth order . if the kinetic term is coupled to einstein tensor by the coupling @xmath25 , then the field equations contain no more than second derivatives @xcite , because it is the special case of horndeski theory of @xmath26 with @xmath27 due to the divergence of @xmath28 .
even for massless scalar field without a potential , the nonminimally derivative coupling @xmath25 has de - sitter solution @xcite .
if the canonical kinetic term @xmath10 is absent , the nonminimally derivative coupling term behaves like a dark matter @xcite .
the introduction of the nonminimally derivative coupling to einstein tensor causes the scalar field to evolve more slowly because of the gravitationally enhanced friction , so the new higgs inflation was successfully implemented without either violating the unitarity bound or fine - tuning the coupling constant @xmath29 @xcite .
furthermore , the challenge to slow - roll inflation with single field due to the lyth bound was also alleviated @xcite .
the cosmological consequences of the theory with nonminimally derivative coupling were discussed extensively @xcite .
we can also distinguish scalar
tensor theories of gravity from einstein s general relativity by calculating the parametrized post - newtonian ( ppn ) parameters and compare them with the solar system experiments .
for the original brans
dicke theory , the ppn parameter @xmath30 .
if the scalar field has a potential @xmath31 , then the ppn parameter @xmath32 is @xcite @xmath33 where the mass is @xmath34 the ppn analysis for the nonminimally derivative coupling @xmath35 with massless scalar field was performed in ref . .
in this paper , we follow ref . to calculate the ppn parameters for the scalar tensor theory with the nonminimally derivative coupling @xmath36 . in sec .
2 we review the general equation of motion for the theory , and the ppn analysis was carried out in sec . 3
, we conclude the paper in sec .
the action for the scalar tensor theory of gravity with nonminimally derivative coupling to einstein tensor is @xmath37\nabla^\mu \phi \nabla^\nu \phi - v(\phi)\right\}+s_m,\ ] ] where the nonminimally coupling constant @xmath38 has the dimension of length squared , @xmath31 is the potential of the scalar field , and @xmath39 is the action for the matter . varying the action with respect to the metric @xmath40 , we get the gravitational field equation @xmath41,\ ] ] with @xmath42 @xmath43 and @xmath44 @xmath45 . \end{split}\ ] ] then , we get @xmath46 @xmath47 where @xmath48 . in terms of @xmath49
, we can express einstein s field equation as @xmath50 similarly , we derive the equation of motion for the scalar field as @xmath51\nabla_{\mu}\nabla_{\nu}\phi - v'(\phi)=0.\ ] ]
in this section , we will calculate the metric in the solar system to the first
order beyond the newtonian approximation . the scalar field perturbation is written as @xmath52 .
the metric is expanded around the flat background @xmath53 to the first
order , @xmath54 , and the perturbation @xmath55 is at least first order in @xmath56 , where @xmath57 is the characteristic velocity of bodies in the solar system .
the order of the post - newtonian expansion is determined by @xmath58 , where @xmath59 and @xmath60 .
our objective is to compute the metric @xmath55 to the order of @xmath61 , or @xmath62 .
following ref . , we also make the following two assumptions : ( c1 ) we assume that the background scalar field @xmath63 is spatially homogeneous and evolves only on cosmological timescales which are much longer than the characteristic timescales of the solar system or laboratory , so the spatial derivatives and second and higher time derivatives of @xmath63 are neglected , and we only consider @xmath64 in the second order of @xmath65 ; ( c2 ) we assume the contribution to stress
energy in the solar system from the background scalar field is subdominant , so we can replace ricci tensor in @xmath66 by @xmath67 . to the zeroth
order , from eq .
we get @xmath68 where @xmath69 . to the first order , eq .
( [ gra : field : equation : s ] ) becomes @xmath70 , \end{split}\ ] ] where @xmath71 and @xmath72 . to the zeroth
order , eq . ( [ sca : field : equation ] ) gives @xmath73 the linear perturbed expansion of eq .
( [ sca : field : equation ] ) is @xmath74 so we get @xmath75 since the order of @xmath76 is at least @xmath77 , so @xmath78 to solve eq .
, we need to calculate the perturbed ricci tensor . in order to simplify the calculation
, we use the gauge conditions @xmath79 with the above gauge choice , the perturbed ricci tensor is @xmath80 @xmath81 @xmath82 @xmath83 where @xmath84 , @xmath85 and @xmath86 . throughout this paper ,
we only label second and higher order terms like @xmath87 .
we know the stress energy tensor of a perfect fluid is @xmath88 so we get @xmath49 as @xmath89 now we are ready to compute the newtonian limit of the @xmath65 and @xmath90 . to do this
we should collect all @xmath91 terms in eq .
( [ pert : gra ] ) . from equation ( [ delta : sca :
equation : know ] ) we know @xmath92 , so the @xmath93 component of the right hand side of equation ( [ pert : gra ] ) is @xmath94 and the @xmath95 components of the right hand side of equation ( [ pert : gra ] ) is @xmath96 , to the order of @xmath91 , we get @xmath97 therefore , the solution is @xmath98 to the order of @xmath77 , the @xmath99 components of equation ( [ pert : gra ] ) becomes @xmath100 substituting the solution @xmath90 in eq . to the above equation , we get the solution @xmath101 where @xmath102 so far the solutions are the same as those in general relativity .
then we need to get the second order solution of @xmath65 . from equations and
, we can get the @xmath62 terms of the sources @xmath103 by introducing the post - newtonian potential @xmath104 @xcite , @xmath105 then equation ( [ r00:2 ] ) for the ricci tensor becomes @xmath106 from equation ( [ per : s ] ) we know that @xmath107 using the post - newtonian potentials @xcite @xmath108 we get @xmath109 combining eqs . , and , we get @xmath110 the solution is @xmath111 using the effective newtonian constant and the dimensionless variable @xmath10 , @xmath112 finally we get @xmath113 @xmath114 @xmath115 or @xmath116 @xmath117 @xmath118 comparing our results with the definitions of the ppn parameters @xcite , @xmath119 @xmath120 @xmath121 then we have @xmath122 note that in einstein s general relativity @xmath123 , @xmath124 and all other ppn parameters are zero .
although the results are similar to that in ref . , the nonminimal coupling @xmath125 considered there introduces extra degree of freedom , the coupling @xmath126 considered in this paper introduces no extra degree of freedom and the equation of motion is still the second order differential equation . by using the experimental constraints on the ppn parameters shown in table [ tab : t1 ] @xcite ,
we get the current limit on @xmath10 .
the current stringent limit on @xmath10 comes from the constraint on the ppn parameter @xmath127 at the 95% confidence level@xcite , so @xmath128 at the 95% confidence level .
horndeski theory is the most general scalar
tensor theory of gravity which introduces no extra degree of freedom because the equation of motion of the theory contains at most the second derivatives .
we consider the special case of horndeski theory in which the kinetic term of the scalar field is nonminimally coupled to einstein tensor @xmath129 , and we calculated the ppn parameters in this theory .
the results we obtained are similar to those in ref . .
however , due to the nonminimal coupling of the kinetic term of the scalar field to ricci tensor @xmath130 , the theory considered in ref . introduces extra degree of freedom . comparing with the results in einstein s general relativity , the ppn parameters have small corrections from @xmath131 which is only important on the cosmological timescale . under the assumed conditions ( c1 ) and ( c2 ) , the consistent solution requires that both @xmath132 and @xmath133 are zero , i.e. , the scalar field should be massless . by using the constraint on the ppn parameter
@xmath134 derived from the statistics of the period derivatives of 21 millisecond pulsars @xcite , we find the strongest limit @xmath135 , so the theory with nonminimally derivative coupling deviates from einstein s general relativity in the order of @xmath0 .
the first time derivative of the orbital period of a selected sample of binary pulsars was also applied to test @xmath136 theory of gravity @xcite .
the ratio of the nonminimal kinetic term to the canonical one is @xmath137 , so the nonminimal term becomes important only if @xmath138 and the nonminimal term is negligible if we choose @xmath139 from inflationary models @xcite .
note that the contribution of the nonminimal coupling to the critical energy density of the universe is @xmath10 , so the effect of the nonminimal coupling is negligible if the scalar field is taken as dynamical dark energy .
if we use the observational constraint @xmath140 and @xmath141 @xcite , then we get the contribution of the canonical kinetic energy is @xmath142 , so roughly we get @xmath143 and the high friction limit assumed in inflationary models @xcite is guaranteed unless the equation of state parameter @xmath144 is extremely close to be -1 .
this research was supported in part by the natural science foundation of china under grant no . 11475065 , and the program for new century excellent talents in university under grant no .
ncet-12 - 0205 . | generalization einstein s relativity is scalar tensor theory gravity brans dicke theory @xcite brans dicke theory massless spin graviton massless scalar field responsible for exchange gravitational interaction scalar field dilaton in low energy bosonic string theory scalar degree freedom arises compactification higher dimensions general scalar tensor theory second order equation motion four spacetime is horndeski theory @xcite lagrangian @xmath1 @xmath2 l_5=&g_5(\phi x_\nu\nabla\phi-\frac{1}{6}g_{5,x}\left\phi\right @xmath3 @xmath4 functions @xmath5 @xmath6 @xmath7 @xmath8 are arbitrary functions of @xmath9 @xmath10 @xmath11 theory includes nonminimal coupling @xmath12 kinetic term @xmath10 @xmath13 special nonminimal coupling @xmath14 @xcite theory conformal invariant no conformal invarianceconsider nonminimal interactions quadratic @xmath9 linear derivative terms @xmath15 divergencies independent interactions @xmath15 inflationary solution discussed nonminimal coupling @xmath24r equation of motion scalar field fourth order kinetic term coupled to einstein tensor coupling @xmath25 field equations contain no second derivatives special case horndeski theory @xmath26 @xmath27 divergence @xmath28 massless scalar field potential nonminimally derivative coupling @xmath25 has de - sitter solution canonical kinetic term @xmath10 absent nonminimally derivative coupling behaves like dark matter introduction nonminimally derivative coupling einstein tensor causes scalar field slowly gravitationally enhanced friction new higgs inflation implemented without violating unitarity bound fine - tuning coupling constant @xmath29 challenge slow - roll inflation single field lyth bound alleviated cosmological consequences theory nonminimally derivative coupling discussed distinguish scalar tensor theories gravity from einstein s general relativity parametrized post - newtonian parameters solar system experimentsoriginal brans dicke theory ppn parameter @xmath30 scalar field potential @xmath31 ppn parameter @xmath32 @xcite @xmath33 mass @xmath34 ppn analysis nonminimally derivative coupling @xmath35 massless scalar field ref . paper follow ref calculate ppn parameters scalar tensor theory nonminimally derivative coupling @xmath36 sec 2 review equation motion ppn analysis sec 3 conclude paper action scalar tensor theory gravity nonminimally derivative coupling einstein tensor @xmath37 nonminimally coupling constant @xmath38 dimension length squared @xmath31 potential scalar field @xmath39 action matter varying action metric @xmath40 gravitational field equation @xmath41 @xmath42 @xmath43 @xmath44 @xmath45 @xmath46 @xmath47 @xmath48 express einstein s field equation @xmath50 derive equation motion scalar field @xmath51 - v'()=0 calculate metric solar system first order beyond newtonian approximation scalar field perturbation written @xmath52metric expanded around flat background @xmath53 first order @xmath54 perturbation @xmath55 first order in @xmath56 @xmath57 characteristic velocity of bodies solar system order post - newtonian expansion determined by @xmath58 @xmath59 @xmath60 objective compute metric @xmath55 to order @xmath61 or @xmath62 two assumptions background scalar field @xmath63 spatially homogeneous evolves on cosmological timescales longer than characteristic timescales solar system spatial derivatives second higher time derivatives of @xmath63 neglected consider @xmath64 second order of @xmath65 contribution to stress energy solar system from background scalar field subdominant replace ricci tensor in @xmath66 by @xmath67 zeroth order @xmath68 @xmath69 first order becomes @xmath70 @xmath71 and @xmath72 zeroth order gives @xmath73 linear perturbed expansion @xmath74 get @xmath75 order of @xmath76 is @xmath77 to solve eq calculate perturbed ricci tensorsimplify calculation use gauge conditions @xmath79 perturbed ricci tensor @xmath80 @xmath81 @xmath82 @xmath83 @xmath84 @xmath85 @xmath86 label second higher order terms like @xmath87 stress energy tensor perfect fluid is @xmath88 get @xmath49 as @xmath89 ready compute newtonian limit @xmath65 @xmath90 collect @xmath91 terms in eq . : gra equation delta know @xmath92 @xmath93 component is @xmath94 @xmath95 components @xmath96 order @xmath91 get @xmath97 solution is @xmath98 order @xmath77 @xmath99 components becomes @xmath100 substituting solution @xmath90 get solution @xmath101 @xmath102 solutions same as general relativity need get second order solution of @xmath65 get @xmath62 terms of sources @xmath103 introducing post - newtonian potential @xmath104 @xcite @xmath105 equation ( r00:2 ricci tensor becomes @xmath106 equation know @xmath107 post - newtonian potentials @xcite @xmath108 get @xmath109 combining eqs get @xmath110 solution @xmath111 effective newtonian constant dimensionless variable @xmath10 @xmath112 @xmath113 @xmath117 @xmath118 results ppn parameters @xcite @xmath119 @xmath121 @xmath122 einstein s general relativity @xmath123 @xmath124 ppn parameters zero results similar ref . nonminimal coupling @xmath125 introduces extra freedom coupling @xmath126 no freedom equation motion second order differential equation experimental constraints on ppn parameters table t1 current limit on @xmath10 limit @xmath10 constraint ppn parameter @xmath127 95% confidence level @xmath128 95% confidence level horndeski theory general scalar tensor theory gravity introduces no extra freedom equation of motion contains second derivatives special case kinetic term scalar field nonminimally coupled to einstein tensor @xmath129 calculated ppn parameters results similar ref . nonminimal coupling kinetic term scalar field ricci tensor @xmath130 theory introduces extra freedom results einstein s general relativity ppn parameters small corrections from @xmath131 important on cosmological timescaleassumed conditions c1 c2 solution requires @xmath132 @xmath133 zero scalar field massless constraint ppn parameter @xmath134 derivatives 21 millisecond pulsars strongest limit @xmath135 theory derivative coupling deviates from einstein s general relativity @xmath0 first time derivative orbital period binary pulsars @xmath136 theory gravity ratio nonminimal kinetic term to canonical @xmath137 important if @xmath138 negligible @xmath139 from inflationary models @xcite contribution nonminimal coupling to critical energy density universe @xmath10 negligible if scalar field dynamical dark energy observational constraint @xmath140 @xmath141 @xcite contribution canonical kinetic energy @xmath142 @xmath143 high friction limit in inflationary models @xcite guaranteed unless equation state parameter @xmath144 close -1 research supported by natural science foundation china grant 11475065 program new century excellent talents university grant ncet-12 - 0205 | 0.428693 | 2,967 | 1,546 | , Saving $0.1 in GPT-4. |
20,590 | we present a pre - survey study of using pan - starrs high sampling rate video mode guide star images to search for tnos .
guide stars are primarily used by pan - starrs to compensate for image motion and hence improve the point spread function . with suitable selection of the guide stars within the pan - starrs 7 @xmath0 field of view
, the lightcurves of these guide stars can also be used to search for occultations by tnos .
the best target stars for this purpose are stars with high signal - to - noise ratio ( snr ) and small angular size . in order to do this , we compiled a catalog using the snr calculated from stars with @xmath1 mag in the tycho2 catalog then cross matched these stars with the 2mass catalog and estimated their angular sizes from @xmath2 color .
we also outlined a new detection method based on matched filter that is optimized to search for diffraction patterns in the lightcurves due to occultation by sub - kilometer tnos .
a detection threshold is set to compromise between real detections and false positives .
depending on the theoretical size distribution model used , we expect to find up to _ a hundred events _ during the three - year life time of the pan - starrs-1 project .
the high sampling ( 30 hz ) of the project facilitates detections of small objects ( as small as 400 m ) , which are numerous according to power law size distribution , and thus allows us to verify various models and further constrain our understanding of the structure in the outer reach of the solar system .
we have tested the detection algorithm and the pipeline on a set of _ engineering _ data ( taken at 10 hz in stead of 30 hz ) .
no events were found within the engineering data , which is consistent with the small size of the data set and the theoretical models .
meanwhile , with a total of @xmath3 star - hours video mode data ( @xmath4 ) , we are able to set an upper limit of @xmath5 deg@xmath6 at 95% confidence limit . | since the first kuiper belt object ( not including pluto ) was discovered in 1992 @xcite , more than 1000 trans - neptunian objects ( tnos ) have been found .
a picture is emerging of the principal characteristics of the tnos that is rich in information but leaves many unanswered questions .
for example , a number of dynamical groups have been identified : classical , resonant , scattered disk @xcite , and more recently the extended scattered disk @xcite . what is the number and size distribution of the smaller objects of these dynamical groups ? is there an extension of the kuiper belt beyond 50 au comprising bodies too small to have been detected in direct surveys ?
meanwhile , could a possible close stellar encounter in the early history of the solar system @xcite be responsible for the mass deficit and the depletion of larger objects ( d @xmath7 150 km ) beyond 50 au ?
the size distribution of these small tnos provides important clues to the dynamical evolution of the early solar system .
kilometer - size tnos with 4% albedo are expected to have @xmath8 , which is still way below the detection limit of the largest ground - based telescopes .
yet the vast majority of the tno population is beyond the limit of direct observation for example , the keck pencil beam survey @xcite has a detection limit of @xmath9 .
a number of authors @xcite have suggested the possibility of detecting tnos by stellar occultation .
an occultation manifests itself as the shadow created by a tno occulting a background star sweeping across the earth , causing flux reduction in the lightcurve .
hence , as opposed to direct observation , stellar occultation by tnos provides a unique way of detecting kilometer to sub - kilometer size objects in the foreseeable future .
recently a few groups have attempted to search for occultations by tnos @xcite , but no real detection has been confirmed yet .
projects committed to search for tnos by occultation either do not have adequate time resolution , observe too few stars with good snr or do not have enough observing time .
these projects do a blind search for occultations , and these events are rare due to the reasons mentioned above .
therefore , the key to a successful detection is to have high sampling ( @xmath10 20 hz ) , high signal to noise ratio of background stars ( snr @xmath10 80 ) , small stellar angular sizes ( @xmath11 0.1 mas ) and , most importantly , many star - hours of observations ( @xmath7 a few hundred thousands ) .
the panoramic survey telescope and rapid response system ( pan - starrs ) is a project consisting of 4 telescopes that can cover over 6000 @xmath0 per night or scan the whole visible sky from hawaii ( @xmath12 ) in a week to a detecting limit @xmath13 magnitude .
pan - starrs prototype telescope ( ps1 ) will monitor up to 60 guide stars in a very high sampling rate video mode .
we have compiled a list which allows us to select guide stars from anywhere in the field ; guide stars could be in any of the 64@xmath1464 otccds ( see @xmath15[sec : ps1 ] ) and read at @xmath16 hz .
each field will have multiple choices of guide stars that were ranked based on the predicted event rates , which depend on their snrs and angular sizes . by properly selecting guide stars
, the pan - starrs video mode images would be ideal for searching tnos . in the next section ,
we briefly describe the pan - starrs system . in @xmath15[sec : diff ] and @xmath15[sec : det ] we discuss the diffraction profiles and present our detection algorithm . in @xmath15 [
sec : angular ] we describe the prediction of stellar angular size using @xmath2 color . in @xmath15[sec : rate ] we estimate the number of expected events based on different number density estimations , sampling rates and stellar angular sizes .
the methodology and compilation of the guide star list is in @xmath15[sec : catalog ] . in @xmath15[sec : engdata ] , we show and describe the quality of the engineering data obtained in fall 2008 .
conclusions are in the final section .
the full version of pan - starrs will contain four identical telescopes and camera systems , which is also known as pan - starrs-4 @xcite .
ps1 is a single telescope and camera system built to test all requirements that are needed in the full version project .
ps1 telescope has an aperture of 1.8 m in diameter with 3-element wide field corrector that delivers a 7 deg@xmath17 flat field in the focal plane .
its 8 m focal length gives a plate scale of 38.5 @xmath18m / arcsecond rendering a pixel resolution of 0.258 arcsecond / pixel .
the telescope can monitor up to 240 guide stars at @xmath16 hz ( in reality , probably about 60 guide stars at 30hz with high snr ) .
high frequency image motion can be removed by on - chip fast guiding , where localized image motion can be calculated from a collection of nearby guide stars . hence different image motions across the field of view can be corrected .
this can only be achieved with the innovative design of an orthogonal transfer charge coupled device ( otccd ) .
the otccd allows on - chip charge shifting , in both rows and columns , which can compensate for the image motion in the focal plane @xcite .
the ps1 giga - pixel camera ( gpc1 ) contains an 8x8 array , minus the 4 corners , of otccds . for reasons of economy of manufacture , facilitating fast readout and other practical concerns , the otccds each consist of an 8x8 orthogonal transfer array ( ota ) of independently - addressable cells of size 590x598 pixels .
hence gpc1 comprises approximately 1.4 billion pixels .
typically , each ota will have a single cell devoted to a guide star reading at high frequency . of the 590x598 pixels of this cell
, only a small fraction around the guide star will be read , thus allowing a fast read out with relatively low read noise .
.filter bandpass , @xmath19 ( the ab magnitude that produces @xmath20/sec / pixel ) and average sky brightness @xmath18 in magnitudes per arcsecond at haleakala for ps1 filters .
[ cols="^,^,^,^",options="header " , ] c
we have made a pre - survey study of using lightcurves from guide star video mode images to search for occultation by tnos near 43 au .
simulations were made to calculate the null hypothesis distribution ( ) and detection efficiency for various angular sizes and snrs . under 30 hz sampling
, ps1 can detect objects as small as @xmath21 m at 43 au ( ) .
using @xmath2 indices , we predicted the angular sizes of all matched stars between tycho2 and 2mass catalogs with @xmath22 mag and @xmath23 mag above the ps1 southern declination limit @xmath24 .
the distribution of stellar angular sizes peak around 0.02 mas in ps1 @xmath12 sky ( ) . on average , there are about 180 stars with @xmath25 mag for ps1 7 deg@xmath17 field of view .
given that we can choose only 60 guide stars for each ps1 target field , we ranked our guide star candidates by their number of expected events based on their angular sizes , snrs and model by p&s05 .
as mentioned before , multiple choices of guide star will ensure us the freedom of selecting guide stars away from gaps in the ccd array or dead pixels that would develop over time .
based on the differential size distribution from different models and the threshold to have one false positive in the ps1 three - year lifetime , we estimated the number of expected events could be somewhere from 1 to @xmath26 ( ) .
the engineering data allowed us to investigate the quality of the lightcurve and develop the detection pipeline for the upcoming real data .
we have established that the detection technique performs as well with the filtered engineering data .
we also realize that the true snr is limited by systematics .
some of the systematics will be removed as we move to real data stream , however some of the systematics will remain . for example , scintillation will limit the snr to about 200 .
we have recalculated the event rates with the worst case scenario and found that the event rates were compromised by a factor of four .
even with this pessimistic estimation the event rate that ps1 will find can allow us to place a constraint on the size distribution and hence the evolution of the tnos .
using the available engineering data and detection efficiency at 0.4 and 0.5 km , we were able to derive an effective solid angle @xmath27 and @xmath28 deg@xmath17 and set the 95% confidence upper limit on surface number density at @xmath29 and @xmath30 deg@xmath6 ( ) . in future work , for the upcoming real data , we will set the threshold based on maximum true positive to false positive ratio , which allows more candidate events for further investigations .
meanwhile , the video mode lightcurves can also be used to search for objects in sedna like orbits from 100 to 1000 au .
we will work on a new detection algorithm that is capable of searching for objects in this region .
work at ncu was supported by the grant nsc 96 - 2112-m-008 - 024-my3 .
work at the cfa was supported in part by the nsf under grant ast-0501681 and by nasa under grant nng04g113 g .
work at asiaa was supported in part by the thematic research program as-88-tp - a02 .
and we also like to thanks iic for hosting j - h wang for his staying . | first kuiper belt object discovered 1992 @xcite 1000 trans - neptunian objects tnos found picture emerging characteristics tnos rich unanswered questions dynamical groups identified classical resonant scattered disk @xcite extended scattered disk @xcite number size distribution of smaller objects ? extension kuiper belt beyond 50 au bodies small surveys ? possible close stellar encounter early responsible for mass deficit depletion of larger objects ( d @xmath7 150 km ) beyond 50 au ? size distribution small tnos clues to dynamical evolution early solar system kilometer - size tnos with 4% albedo expected have @xmath8 below detection limit largest ground telescopes majority tno population beyond limit direct observation keck pencil beam survey @xcite detection limit @xmath9 authors suggested detecting tnos by stellar occultation occultation shadow tno occulting background star flux reduction lightcurve stellar occultation detecting kilometer to sub - kilometer size objects groups attempted search for occultations by tnos @xcite no real detection confirmed projects search tnos occultation adequate time resolution observe too few stars enough observing timeprojects search for occultations events rare to reasons key to successful detection high sampling @xmath10 20 hz high signal to noise ratio 80 ) small stellar angular sizes @xmath11 0.1 mas many star - hours observations @xmath7 thousands panoramic survey telescope rapid response system pan - starrs ) project 4 telescopes cover over 6000 @xmath0 per night scan sky from hawaii week to @xmath13 magnitude pan - starrs prototype telescope 60 guide stars high sampling rate video mode compiled list select guide stars from field 64@xmath1464 otccds @xmath16 hz each field multiple choices of guide stars ranked based on predicted event rates snrs angular sizes selecting stars pan - starrs video images ideal for searching tnos describe pan - starrs system @xmath15 diff det discuss diffraction profiles detection algorithm angular prediction of stellar angular size using @xmath2 color rate expected events based number density estimations sampling rates stellar angular sizes methodology compilation guide star list in @xmath15[sec : catalog ]@xmath15 engdata quality engineering data fall 2008 conclusions final section full version pan - starrs four telescopes camera systems known pan - starrs-4 @xcite ps1 single telescope camera system requirements full version project ps1 telescope aperture 1.8 m 3-element wide field corrector 7 deg@xmath17 flat field focal plane 8 m focal length plate scale 38.5 @xmath18m / arcsecond pixel resolution 0.258 arcsecond / pixel telescope 240 guide stars @xmath16 hz 60 stars at 30hz high snr high frequency image motion removed on - chip fast guiding localized image motion calculated from nearby guide stars different image motions corrected design orthogonal transfer charge coupled device otccd ) allows on - chip charge shifting image motion focal plane ps1 giga - pixel camera 8x8 array otccds economy manufacture fast readout otccds 8x8 orthogonal array independently - addressable cells size 590x598 pixels gpc1 comprises 1.4 billion pixels each ota single cell guide star reading high frequency590x598 pixels small fraction around guide star read fast read low noise bandpass @xmath19 magnitude @xmath20/sec pixel average sky brightness @xmath18 magnitudes per arcsecond haleakala ps1 filters pre - survey study lightcurves guide star video occultation tnos near 43 au simulations null hypothesis distribution detection efficiency angular sizes snrs 30 hz sampling ps1 objects small @xmath21 m at 43 au @xmath2 indices predicted angular sizes matched stars between tycho2 2mass catalogs @xmath22 @xmath23 mag above ps1 southern declination limit @xmath24 stellar angular sizes peak 0.02 mas ps1 @xmath12 sky average 180 stars @xmath25 mag ps1 7 deg@xmath17 field of view 60 guide stars each ps1 target field ranked guide star candidates expected events angular sizes snrs model multiple choices guide star freedom selecting stars gaps ccd array dead pixels differential size distribution false positive ps1 three - year lifetime estimated expected events 1 to @xmath26 engineering data quality lightcurve develop detection pipeline upcoming real dataestablished detection technique performs with filtered engineering data true snr limited by systematics some systematics removed real data stream remain scintillation snr to 200 recalculated event rates worst case scenario compromised by factor four pessimistic estimation event rate ps1 on size distribution evolution tnos using engineering data detection efficiency at 0.4 0.5 km effective solid angle @xmath27 @xmath28 deg@xmath17 set 95% confidence upper limit on surface number density at @xmath29 @xmath30 deg@xmath6 future work set threshold based on maximum true positive to false positive ratio more candidate events video mode lightcurves search for objects in sedna orbits from 100 to 1000 au work on new detection algorithm region work at ncu supported by grant nsc 96 - 2112-m-008 - 024-my3 cfa nsf grant nasa grant nng04g113 asiaa supported thematic research program as-88-tp - a02 iic for hosting j - h wang | 0.315397 | 2,175 | 1,123 | , Saving $0.1 in GPT-4. |
20,604 | the presence of a domain wall trapped by a sub - micron notch is probed in two ways : through electronic transport measurements and by magnetic force microscopy ( mfm ) .
we observe complex magnetic features which are consistent with numerical simulations predicting the existence of multiple magnetic configurations stabilized by the notch structure .
domain walls ( dw ) play a critical role in much of the current research involving spin - dependent electronic devices@xcite . in the case of a highly constrained dw ,
extremely large magnetoresistance ratios have been reported@xcite .
logic gates based on domain wall motion are being studied@xcite .
recently , the movement of trapped dws via the spin - transfer torque has been experimentally demonstrated in thin films@xcite .
although the dw characteristics are key parameters in the aforementioned experiments , only a few theoretical@xcite and experimental studies@xcite have been devoted to trapped dws . in this letter we present an experimental study of trapped dws . using a device geometry previously proposed by ono , et al.@xcite and subsequently employed by grollier , et al.@xcite , we demonstrate that multiple stable locations exist for the trapped dw in such devices . | exchange biased spin - valve films consisting of ta(3)/ nife(1.5)/ irmn(5)/ cofe(2.5)/ cu(3)/ cofe(4)/ nife(15)/ ta(2.5 ) ( thicknesses in nm ) were sputtered onto a si substrate . to fabricate the devices
, we used a combination of e - beam lithography and reactive ion etching to produce a thin carbon hardmask .
this mask protected the devices during the ar ion milling used to etch the spin valve film .
a second e - beam step was used to pattern thick ta / au leads allowing 4-probe device resistance measurements to be made .
the left side of fig . [ fig1](a ) is a schematic plane view of the device .
there are two important features : a 50 @xmath0 m square nucleation pad and a long ( 100 @xmath0 m ) , narrow ( 0.5 @xmath0 m ) propagation line .
the four gold leads ( i+ , v+ , i- and v- ) for measuring the propagation line resistance are also shown .
a rectangular notch is positioned one - third of the way along the propagation line ( see fig.[fig1](a ) right ) .
different notch geometries have been fabricated , varying the notch width ( w ) from 500 nm to 50 nm .
typical curves of the propagation line resistance as a function of the applied field ( h@xmath1 ) are shown in fig.[fig1](b ) .
the resistance measurements were performed at room temperature using a standard 4-probe technique .
the gmr signal is recorded during two different field loops .
the first , corresponding to the ( @xmath2 ) curve , is a minor loop during which the magnetization of the cofe / nife bilayer ( called free layer in the following ) is completely reversed .
the minimum and maximum states of resistance correspond respectively to the parallel and antiparallel alignment of the spin - valve magnetizations .
intermediate plateaus at values corresponding to either 1/3 or 2/3 of the total resistance variation are present during the switching of the free layer magnetization .
these plateaus indicate the existence of a magnetic configuration in which a dw present in the free layer is trapped at the notch@xcite . in order to check the stability of this configuration at remanence ,
the propagation line resistance has been recorded during a second field loop : the ( @xmath3o@xmath3 ) curve in fig . [ fig1](b ) . after preparing the sample in a magnetic state where a dw is trapped in the notch ,
h@xmath1 is reversed .
note that the resistance remains at an intermediate level for -20 oe@xmath4 h@xmath5 + 3 oe .
so in this field range the dw stays trapped by the notch . to further investigate this specific magnetic configuration ,
mfm measurements have been performed .
using a commercial digital instruments scanning probe microscope equipped with a standard mesp tip , both topographic and mfm measurements were performed at room temperature .
no external magnetic field was applied during the mfm scans , and the tip - sample separation was 18 nm . the topographic images for four different notch shapes ( w= 500 , 200 , 100 and 50 nm ) are denoted topo . " in fig .
the depth of the notch is 150 nm for all widths except for samples with w=50 nm where proximity effects in the negative e - beam resist made the notch depth 110 nm .
with the electrical characterization technique previously described , we have been able to prepare the free layer of our samples in two different magnetic states at remanence named a and b in the following ( see the resistance levels denoted a and b in fig . [ fig1](b ) ) . the state a is characterized by the presence of a dw trapped by the notch .
conversely , no dw is present in the state b. the mfm scans obtained in the states a and b , for samples with w=500 , 200 , 100 and 50 nm , correspond respectively to the pictures labelled mfma and mfmb in fig .
independent of the notch geometry , dark shapes ( resembling a bat ) centered on the notch are present on the mfma pictures , while two black and white areas surrounding the notch are observable in the mfmb pictures .
the contrast of mfmb images have been reproduced well using the micromagnetic model oommf@xcite ( not shown here ) . using the same model , the micromagnetic configuration and resulting
mfm image at remanence has also been computed in the case of a trapped dw ( state b ) for different reverse fields values .
figure [ fig3 ] shows the results of these simulations .
the absolute value of the reverse field is increased in 20 oe increments , starting from 60 oe for simulation ( i ) and ending at 120 oe for simulation ( iv ) .
note that the effect of temperature is not taken into account in these simulations , leading to an overestimation of the propagation field values . depending on the reverse field value , different wall shapes and wall positions can be attained at remanence .
so the simulations predict the existence of multiple possible trapped dw positions at remanence . but none of these simulations produces mfm images similar to the observed bat " shape which our magnetoresistance measurements unambiguously associate with the presence of a trapped dw .
note , however , that the observed bat " shape may be obtained via a superposition of the different simulated states and of their mirrors .
this strongly suggests that the dw is imaged in different positions during the same mfm scan .
hence , the observed bat " shape likely originates from dw motion induced by the tip .
such motion is not surprising .
the effect of mfm tips on micromagnetic configuration was widely studied when the technique was introduced ; specifically , dw motion effects induced by mfm tips were observed by mamin , et al.@xcite . in order to confirm this hypothesis
we have performed mfm scans in which we vary the scanning history .
figure [ fig4 ] shows the results obtained when a dw is present in the free layer for devices with w= 500 , 200 , 100 nm .
the smaller notch case ( w= 50 nm ) will be discussed later .
three different scanning procedures have been used ; they are named ( a ) , ( b ) and ( c ) in the following .
each of them includes two successive scans and the pictures are always recorded during the second scan .
the procedure ( a ) consists in a scan starting from the bottom " of the device and ending at the top " of the element immediately followed by a scan performed in the opposite direction ( starting from the top " ) . during procedure ( b )
the two successive scans are performed in the same direction from the bottom " to the top " part of the device . for procedure ( c )
the two successive scans are performed from the top " to the bottom " . the global trajectory followed by the tip during each procedure
is noted by arrows in fig .
the mfm pictures recorded using procedures ( a ) , ( b ) and ( c ) are labeled respectively ( a ) , ( b ) and ( c ) in fig .
[ fig4 ] . if procedure ( a ) leads to the observation of the full bat " shape as expected .
note that ( a ) is the default procedure currently used in scanned probe microscopy .
] , procedures ( b ) and ( c ) generate a discontinuity during the magnetic signal recording . only the top ( bottom ) wing of the bat " is present in images when procedure ( b ) ( procedure ( c ) ) is performed .
these three behaviors can be easily understood by taking the dw motion induced by the mfm tip into account . during the first scan of procedure ( a )
, the tip leaves the dw in a position bordering the top of the notch , and then during the second scan , the tip , moving from top - to - bottom , displaces the dw through the notch .
so the dw is imaged in all of the trapping positions induced by the notch , and this leads to the complete bat " observation . at the end of the first scan of procedure ( b )
, the tip leaves the dw in a position bordering the top of the notch . during the second scan ,
the mfm tip is unable to image the dw in the bottom position , simply because the domain wall is situated in the top position .
so the bottom wing of the bat " is missing . when the tip approaches close enough to the domain wall , the wall is attracted in an irreversible way , and this leads to the discontinuity in the magnetic signal observed in ( b ) .
then the dw experiences the trapping positions situated on the tip trajectory .
this leads to the imaging of the top part of the bat " .
if the two scanning directions of procedure ( b ) are reversed , we expect that the top wing of the bat " will be missing , as observed in the images recorded using procedure ( c ) .
the same experiments were performed on the smallest notch devices ( w= 50 nm ) , but the results obtained were not reproducible from one device to another .
we believe that this is due to weaker pinning in the shallower notch depth realized at this dimension . in conclusion ,
stripe shaped spin - valves including a sub - micron notch have been fabricated in order to investigate trapped dws . combining micromagnetic simulations , transport , and mfm measurements
, we have demonstrated that multiple stable states can be created by the notch .
furthermore , the shape and the stability of the wall appears to be dependent on the trapping position .
these results should be taken into account when dealing with spin - dependent electronic devices involving trapped dws because the magneto - electronic properties can change drastically from one stable position to another . | biased spin - valve films ta(3) nife(1.5) irmn(5) cofe(2.5) cu(3) cofe(4) nife(15) ta(2.5 ) thicknesses nm sputtered si substrate used e - beam lithography reactive ion etching thin carbon hardmask mask protected devices ar ion milling second e - beam thick ta / au leads 4-probe resistance measurements schematic view device features 50 @xmath0 m square nucleation pad 100 m narrow ( 0.5 @xmath0 m ) propagation line four gold leads i+ v+ i- v- measuring propagation line resistance shown rectangular notch one - third along propagation line notch geometries notch width 500 nm to 50 nm curves propagation line resistance field shown in fig.(b ) resistance measurements performed at room temperature standard 4-probe technique gmr signal recorded during two field loops minor loop magnetization cofe / nife bilayer reversed minimum maximum states resistance correspond to parallel antiparallel alignment spin - valve magnetizationsintermediate plateaus at 1/3 or 2/3 total resistance variation during switching free layer magnetization plateaus indicate magnetic configuration dw trapped at notch@xcite stability propagation line resistance recorded during second field loop : ( @xmath3o@xmath3 ) curve in fig .(b after preparing sample in magnetic state dw trapped in notch h@xmath1 reversed resistance remains intermediate level for -20 oe@xmath4 h@xmath5 + 3 oe dw stays trapped by notch mfm measurements performed using commercial digital instruments scanning probe microscope standard mesp tip topographic mfm measurements performed at room temperature no external magnetic field tip - sample separation 18 nm topographic images for four notch shapes ( w= 500 200 100 50 nm ) denoted topo . depth notch is 150 nm for except w=50 nm notch depth 110 nm electrical characterization technique free layer samples in two magnetic states at remanence a and b state a dw trapped by notch no dw in state b.mfm scans in states a b samples w=500 200 100 50 nm correspond to pictures mfma mfmb in fig . dark shapes resembling bat on notch present on mfma pictures two black and white areas surrounding notch in mfmb pictures contrast mfmb reproduced using micromagnetic model oommf@xcite micromagnetic configuration mfm image at remanence computed trapped dw ( state b ) for different reverse fields values figure fig3 shows results absolute value reverse field increased in 20 oe increments from 60 at 120 oe iv effect of temperature not overestimation of propagation field values reverse field value different wall shapes positions at remanence simulations predict multiple trapped dw positions at remanence none produces mfm images similar to bat " shape with trapped dw observed bat " shape may obtained via superposition of simulated states mirrors suggests dw imaged in different positions during mfm scan bat " shape likely originates from dw motion induced by tip effect of mfm tips on micromagnetic configuration studied dw motion effects observed by mamin , et al.@xcite confirm hypothesis performed mfm scans scanning history figure shows results when dw present in free layer for devices with w= 500 200 100 nm smaller notch case ( w= 50 nm ) discussed later three scanning procedures used ( a ) b ) c ) each includes two scans pictures recorded during second scan procedure ( a ) scan from bottom ending at top followed by scan opposite direction ( b ) scans same direction from bottom to procedure ( c ) scans from top to bottom global trajectory by tip noted by arrows in fig mfm pictures recorded using procedures a b c ) labeled ( ) in fig if procedure ( a ) leads to full bat " shape default procedure in scanned probe microscopy procedures ( b ) c ) generate discontinuity during magnetic signal recording only top bottom wing of bat " present in images when procedure ( b ) c performed three behaviors understood by dw motion induced by mfm tipfirst scan ( a ) tip leaves dw bordering top notch second scan tip to displaces dw through notch dw imaged in trapping positions notch leads to complete bat " observation end first scan ( b ) tip leaves dw bordering top notch second scan mfm tip image dw bottom position domain wall in top position bottom wing bat " missing tip approaches close to domain wall wall attracted leads to discontinuity in magnetic signal ( b ) dw experiences trapping positions on tip trajectory leads to imaging top part of bat " if two scanning directions reversed top wing bat " missing in images procedure ( c ) experiments performed on smallest notch devices ( w= 50 nm ) results not reproducible device due to weaker pinning in shallower notch depth stripe shaped spin - valves including sub - micron notch fabricated to investigate trapped dws micromagnetic simulations transport mfm measurements demonstrated multiple stable states created by notch shape stability of wall dependent on trapping positionresults dealing spin - dependent electronic devices trapped dws magneto - electronic properties change stable position | 0.348178 | 2,094 | 1,031 | , Saving $0.1 in GPT-4. |
20,611 | almost all properties of a photodissociation region ( pdr ) depend on its metallicity .
the heating and cooling efficiencies that determine the temperature of the gas and dust , the dust composition , as well as the elemental abundances that influence the chemical structure of the pdr are just three examples that demonstrate the importance of metallicity effects in pdrs .
pdrs are often associated with sites of star formation . if we want to understand the star formation history of our own galaxy and of distant low - metallicity objects we need to understanding how metallicity acts on pdr physics and chemistry . | it is common to define pdrs as regions where stellar far ultraviolett ( fuv : 6 ev@xmath0 13.6 ev ) radiation dominates the physical and chemical properties of the local interstellar medium ( ism ) .
fuv radiation is predominantely produced in massive stars , which do not live long enough to exit their parental cloud of gas and dust within their lifetime .
pdrs are the products of the strong mutual feedback between young massive stars and their parental clouds and hence closely related to the process of star formation ( sf ) .
metallicity ( @xmath1 ) is thought to affect the fragmentation properties of gravitationally unstable gas , especially in the early universe ( bromm _ et al .
_ @xcite ; schneider _ et al . _
this should also affect the initial mass function ( imf ) .
however , jappsen _ et al .
_ @xcite have shown , that the cooling of hot , very low-@xmath1 gas does not seem not to depend on metallicity , since it is dominated by h@xmath2 cooling .
hence , the cooling depends on the amount of h@xmath2 available for cooling , which might still be a function of metallicity due to the fact that molecular hydrogen is formed on metallic surfaces , i.e. dust .
figure [ influence ] lists the most prominent metallicity effects .
an altered metallicity affects the physics ( primarily energy balance and radiative transfer ) and the chemistry of pdrs .
the most important , direct effects are altered elemental abundances leading to a different chemical structure and different total column densities of important species like c@xmath3 , c , and co as well as an altered dust - to - gas ratio @xmath4 .
a lower @xmath4 leads to a weaker attenuation and fuv radiation penetrates deeper into the cloud , destroying fragile molecule up to larger depths . additionally , gas heating by fuv photons can act deeper in the cloud leading to a different temperature structure compared to solar metallicity pdrs . on the other hand ,
the probability for a photon to escape the cloud is increased , improving the cooling efficiency at any particular depth while the smaller elemental abundance reduces the cooling rate .
observations of low metallicity environments also confirm that the dust composition changes at lower @xmath1 .
the pah ( polycyclic aromatic hydrocarbons ) fraction drops significantly ( draine _ et al . _
@xcite , madden _ et al . _
pahs dominantly contribute to the total photoelectric heating ( pe ) efficiency , thus , at lower values of @xmath1 the pe efficiency is significantly diminished .
_ @xcite also showed , that the spectral shape of the fuv radiation is different in low @xmath1 dwarf galaxies .
reduced metallicities lead , for a given stellar type , to hotter main sequence temperatures and to harder , i.e. more energetic , spectral seds .
this may affect the dust composition via evaporation and photodissociation processes and leads to a different pe efficiency .
since all these effects are non - linear and mutually coupled it is very difficult to give an analytical solution and one has to rely on numerical computations .
sometimes the results are completely counter - intuitive like displayed in figure [ invertedt ] .
the heating and cooling efficiencies at the cloud surface depend on parameters like density @xmath5 , metallicity and fuv intensity @xmath6 . with increasing fuv intensities and densities above @xmath7 @xmath8 , the dominant heating contribution shifts from h@xmath2-deexcitation to photoelectric ( pe ) heating ( c.f .
_ @xcite ) . since @xmath9 is much less effective for @xmath10 compared to the cooling terms , which scale linearly with @xmath1 , the gas temperature drops if @xmath6 increases , i.e. if we move closer to the fuv source !
vice versa this means , a hypothetical cloud gets hotter the further away it is placed from a strong fuv source . even though this might be an academic example due to its particular assumptions , it demonstrates the difficulty in predicting metallicity effects .
an inversion occurs , i.e. the surface temperature drops for an intenser fuv illumination . ]
the model results presented here are computed with the kosma-@xmath11 pdr code .
( stoerzer _ et al . _
@xcite , rllig _ et al . _
this code features spherical geometry and an isotropical fuv irradiation .
it calculates the chemical and physical cloud structure as function of radius as well as the total cloud emission .
the metallicity is a free model parameter and changes the elemental abundances and the gas - to - dust ratio .
the chemical network in kosma-@xmath11 is fully modular and individual species can be easily added to or removed from the network .
recently , the code has been expanded to calculate the dust temperatures and continuum emission for any given dust distribution ( rllig & szczerba @xcite ) .
the model setup can be easily expanded to clump ensembles as presented by cubick _
( this issue ) to calculate the pdr emission of a large , complex superposition of differently sized clouds .
_ @xcite gives an overview over pdr modelling as well as a comparison between a number of different model codes .
the kosma-@xmath11 results are also available online ( http://www.astro.uni-koeln.de/workgroups/theo_astronomy/pdr/ ) .
* left : * photodissociation rates of h@xmath2 ( black ) and co ( red ) vs. depth [ cm ] . *
right : * chemical abundances vs. depth .
the lower metallicity leads to a lower abundance at the surface of the cloud .
additionally , the weaker shielding shifts the typical h - h@xmath2 and c@xmath3-c - co transitions deeper into the cloud . ] . * left : * photodissociation rates of h@xmath2 ( black ) and co ( red ) vs. depth [ cm ] . *
right : * chemical abundances vs. depth .
the lower metallicity leads to a lower abundance at the surface of the cloud . additionally , the weaker shielding shifts the typical h - h@xmath2 and c@xmath3-c - co transitions deeper into the cloud . ]
a reduced metallicity leads to a decreased shielding of fuv radiation . in the left panel in figure [ shielding ] we compare the h@xmath2 and co photodissociation rates for @xmath12 and @xmath13 .
the right panel shows the chemical abundances vs. depth .
the lower metallicity leads to a lower abundance of all metallic species at the surface of the cloud . additionally , the weaker shielding shifts the typical h - h@xmath2 and c@xmath3-c - co transitions deeper into the cloud leading to an almost negligible amount of co in the @xmath12 case while the c@xmath3 column depth is much less affected .
and co(1 - 0 ) surface brightnesses of an ensemble of model pdrs vs. metallicity .
the red lines show the ratio [ cii]/co(1 - 0 ) .
* right:*[cii ] survey results : comparing normal metallicity regions with low - metallicity galaxies ( from madden @xcite).,title="fig : " ] and co(1 - 0 ) surface brightnesses of an ensemble of model pdrs vs. metallicity .
the red lines show the ratio [ cii]/co(1 - 0 ) .
* right:*[cii ] survey results : comparing normal metallicity regions with low - metallicity galaxies ( from madden @xcite).,title="fig : " ] the reduced column densities of the more fragile species at lower metallicities are visible as lower surface brightnesses of the pdr .
co , being one of the most important chemical species , suffers strongly from the weak shielding due to low @xmath1 .
figure [ ciivsco ] compares the ensemble averaged clump emission ( see below ) of [ cii ] , and co(1 - 0 ) vs. metallicitiy ( on the 12+log(o / h ) scale ) . for lower metallicity values ,
the co surface brightness decreases significantly .
this is a combination of reduced column densities plus a much smaller projected area of the co core of the cloud due a transition from c to co at much higher depths .
this projection effect is particularly strong in the low density case @xmath14 @xmath8 , visible as kink in the co curve at 12+log(o / h)=8.4 .
the right panel in figure [ ciivsco ] , taken from madden @xcite , shows the [ cii ] intensity vs. co(1 - 0 ) comparing normal metallicity regions with low - metallicity galaxies .
lines of constant @xmath15}/i_\mathrm{co}$ ] run diagonally across the plot .
the figure clearly shows the metallicity dependence of the [ cii]/co ratio .
a similar behavior has been found by nakagawa _
@xcite in low metallicity clouds in the large magellanic clouds ( lmc ) .
the ratio @xmath15}/i_\mathrm{co}$ ] is a good measure of pdr emission relative to the emission of the cold molecular material and is a good indicator of star formation .
active star formation in the milky way and in nearby star burst galaxies shows a ratio @xmath15}/i_\mathrm{co}=6300 $ ] while dwarf galaxies show ratios between 6000 and 70000 .
this makes dwarf galaxies one of the prime targets to study low metallicity pdrs . due to their small angular size , single dish far - infrared ( fir )
observations of dwarf galaxies are unable to resolve individual molecular clouds .
hence we apply a clump ensemble approach ( cubick _ et al . _
@xcite ) to model the beam averaged pdr emission of these galaxies .
figure [ ngc6822 ] compares the model results for the pdr emission of ngc 6822v , an hii region complex in the dwarf irregular galaxy of the local group at a distance of 470 kpc and a @xmath16 ( 12+log(o / h)=8.37 ) , with fir observations . the blue crosses in figure [ ngc6822 ] show the results for a single component model , i.e. the galaxy is assumed to consists of numerous identical model clouds .
the red crosses denote a clump - mass distribution @xmath17 and mass - size relation @xmath18 . taking into account the observed mid-@xmath19 co and fine - structure transitions
it is not possible to find a good model fit for a single component model . for a clump ensemble
we derive a mean gas density @xmath20 @xmath8 , @xmath21 , @xmath22 . 99 bromm , v. ; ferrara , a. ; coppi , p. s. ; larson , r. b. , 2001 , mnras , 328 , 969 cubick , m. , stutzki , j. , ossenkopf , v. , kramer , c. , rllig , m. 2008 , a&a , submitted draine , b. t. et al .
, 2007 , apj , 663 , 866 israel , f .
p. ; baas , f. ; rudy , r.j .
; skillman , e.d . ; woodward , c.e.,2003 , a&a 397,87 jappsen , a .-
k . ; klessen , r.s . ; glover , s.c.o . ; mac low , m.m . , 2007 , arxiv:0709.3530 nakagawa , t. ; mochizuki , k. ; okuda , h. ; yui , y. y. ; doi , y. ; shibai , h. ; nishimura , t. ; low , f. j. , 1999 , proceedings of star formation 1999 , nagoya , japan madden , s.c . , 2000 , new astronomy reviews , volume 44 , issue 4 - 6 , p. 249
- 256 madden , s. c. ; galliano , f. ; jones , a. p. ; sauvage , m. , 2006 , a&a , 446 , 877 rllig , m. ; ossenkopf , v. ; jeyakumar , s. ; stutzki , j. ; sternberg , a. , 2006 , a&a , 45 , 917 rllig , m. et al . , 2007 , a&a , 467 , 187 rllig , m. & szczerba , r. , 2008 , in preparation schneider , r. ; ferrara , a. ; natarajan , p. ; omukai , k. , 2002 , apj , 571 , 30 stoerzer , h. ; stutzki , j. ; sternberg , a. , 1996 , a&a , 310 , 592 sternberg , a. ; dalgarno , a. , 1995 , apj , 338 , 197 | common define pdrs as regions where stellar far ultraviolett ( fuv ) radiation dominates physical chemical properties local interstellar medium fuv radiation produced in massive stars exit parental cloud pdrs products of feedback between young stars parental clouds related to star formation metallicity ( @xmath1 fragmentation properties of gravitationally unstable gas especially in early universe initial mass function cooling of hot low-@xmath1 gas depend on metallicity dominated by h@xmath2 cooling cooling depends on h@xmath2 available for cooling might function of metallicity molecular hydrogen formed on metallic surfaces figure lists prominent metallicity effects altered metallicity affects physics energy balance radiative transfer chemistry of pdrs important effects are altered elemental abundances different chemical structure different total column densities of species altered dust - to - gas ratio @xmath4 lower @xmath4 leads to weaker attenuation fuv radiation penetrates deeper into cloud destroying fragile molecule gas heating by fuv photons can act deeper in cloud different temperature structure compared to solar metallicity pdrsprobability photon to escape cloud increased improving cooling efficiency smaller elemental abundance reduces cooling rate observations low metallicity environments confirm dust composition changes at lower @xmath1 pah aromatic hydrocarbons fraction drops pahs contribute to total photoelectric heating efficiency at lower values @xmath1 pe efficiency diminished spectral shape fuv radiation different in low @xmath1 dwarf galaxies reduced metallicities lead to hotter temperatures harder more energetic spectral seds affect dust composition evaporation photodissociation different pe efficiency effects non - linear mutually coupled difficult to solution on numerical computations results counter - intuitive figure heating cooling efficiencies at cloud surface depend on density metallicity fuv intensity @xmath6 increasing fuv intensities densities above @xmath7 @xmath8 dominant heating contribution shifts from h@xmath2-deexcitation to photoelectric pe heating @xmath9 less effective for @xmath10 gas temperature drops if @xmath6 increases closer to fuv source hypothetical cloud gets hotter further away from strong fuv sourceacademic example demonstrates difficulty predicting metallicity effects inversion occurs surface temperature drops for intenser fuv illumination model results computed with kosma-@xmath11 pdr code code features spherical geometry isotropical fuv irradiation calculates chemical physical cloud structure radius total cloud emission metallicity free model parameter changes elemental abundances gas - to - dust ratio chemical network in modular species added or removed code expanded to calculate dust temperatures continuum emission for dust distribution model setup can expanded to clump ensembles calculate pdr emission of large complex superposition of differently sized clouds @xcite gives overview over pdr modelling comparison between model codes kosma-@xmath11 results available online http://www.astro.uni-koeln.de/workgroups/theo_astronomy/pdr/ photodissociation rates of h@xmath2 co vs. depth cm chemical abundances vs depth lower metallicity leads to lower abundance at surface cloud weaker shielding shifts typical h - h@xmath2 c@xmath3-c - co transitions deeper into cloud photodissociation rates of h@xmath2 vs. depthchemical abundances vs. depth lower metallicity leads lower abundance surface cloud weaker shielding shifts h - h@xmath2 c@xmath3-c - co transitions deeper cloud reduced metallicity decreased shielding fuv radiation left h@xmath2 co photodissociation rates for @xmath12 @xmath13 right chemical abundances vs. depth lower metallicity lower abundance metallic species surface cloud weaker shielding shifts h - h@xmath2 c@xmath3-c - co transitions deeper cloud negligible amount co in @xmath12 c@xmath3 column depth less affected co(1 - 0 ) surface brightnesses model pdrs vs. metallicity red lines show ratio cii/co(1 - 0 survey results normal metallicity regions low - metallicity galaxies co(1 - surface brightnesses red lines show ratio/co(1 - 0 reduced column densities fragile species at lower metallicities visible as lower surface brightnesses pdr co important chemical suffers weak shielding due low @xmath1 figure compares ensemble averaged clump emission of cii ] co(1 - 0 ) vsmetallicitiy 12+log(o / h ) scale lower metallicity values co surface brightness decreases combination reduced column densities smaller projected area co core cloud transition c to co higher depths projection effect strong low density case visible kink co curve 12+log(o / h)=8.4 right panel figure shows cii ] intensity vs. co(1 - 0 ) comparing normal metallicity regions low - metallicity galaxies lines constant run diagonally across plot figure shows metallicity dependence cii]/co ratio similar behavior found low metallicity clouds large magellanic clouds ratio measure pdr emission cold molecular material indicator star formation active star formation milky way nearby star burst galaxies ratio=6300 dwarf galaxies show ratios between 6000 and 70000 dwarf galaxies prime targets study low metallicity pdrs small angular size far - infrared observations dwarf galaxies resolve individual molecular clouds clump ensemble approach model averaged pdr emission galaxiesfigure ngc6822 compares model results pdr emission ngc 6822v hii region complex dwarf irregular galaxy local group distance 470 kpc @xmath16 ( 12+log(o / h)=8.37 ) fir observations blue crosses show results single component model galaxy numerous identical model clouds red crosses denote clump - mass distribution @xmath17 mass - size relation @xmath18 mid-@xmath19 fine - structure transitions possible find good model fit single component model clump ensemble mean gas density @xmath20 @xmath8 @xmath21 @xmath22 bromm v ferrara coppi larson 2001 mnras , 328 , 969 cubick stutzki ossenkopf kramer rllig 2008 draine b t. 2007 apj , 663 , 866 israel baas rudy skillman woodward,2003 a&a 397,87 jappsen klessen glover mac low 2007 arxiv:0709.3530 nakagawa mochizuki okuda yui doi shibai nishimura low f1999 star formation nagoya japan madden s 2000 new astronomy reviews volume 44 issue 4 - 6 p. 249 - 256 madden galliano f jones sauvage 2006 a&a 446 , 877 rllig ossenkopf jeyakumar stutzki j a 2006 a&a 45 , 917 rllig 2007 a&a 467 , 187 rllig szczerba 2008 preparation schneider r ferrara natarajan p omukai 2002 apj 571 , 30 stoerzer stutzki sternberg 1996 a&a 310 , 592 sternberg dalgarno 1995 338 , 197 | 0.432348 | 2,878 | 1,438 | , Saving $0.1 in GPT-4. |
20,621 | the present study reports the formation of self - assembled droplet pattern on the pdms polymer coated over grooved side of dvd under saturated vapours of alcohols .
comparison of the results with breath figures formed over unconstrained side of dvd is made .
four different environments namely methanol , ethanol , 2-propanol and n - butanol are used for the analysis .
it is observed that the pattern formation occurs with methanol and ethanol vapours and not with 2-propanol and n - butanol .
the difference is pattern formation with different alcohols is attributed to the variation in chain length and the presence of hydrophobic groups in alcohols , as given by traube s rule .
the distortion of patterns over constrained surface is attributed to the depinning of contact lines . | the study of dynamics of a droplet s contact line on structured surfaces is of importance for its applications in self - cleaning @xcite , anti - corrosion @xcite , bio - sensing @xcite and so on .
the contact line dynamics is greatly influenced by the presence of surface heterogeneities @xcite .
these heterogeneities acts as droplet pinning sites .
although the dynamics of the contact line for hydrophobic / super - hydrophobic surfaces were extensively studied , the contact line dynamics during the phase transition process like evaporation and condensation were relatively less studied @xcite .
+ breath figure is one such process involving the phase change process of evaporation and condensation to form self - assembled droplet patterns @xcite .
usually breath figures are formed with super saturated vapor of water .
ding et al .
, was the first to report the formation of self - assembled droplet pattern using non - aqueous vapors of ethanol and methanol @xcite .
they proved that the breath figure method is not only suitable for water vapor but is also compatible with organic solvents like ethanol and methanol .
but the influence of constraints on the underlying surface on the formation of breath figures using non - aqueous vapor is so far not reported to the best of our knowledge .
+ alcohols are the simplest form of amphiphilic molecules having hydrophobic and hydrophilic groups @xcite .
these hydrophobic and hydrophilic group of the alcohols play an important role in deciding the contact area of the droplets .
when a droplet of liquid is impinged on the surface ( preferentially solid / soft solid surface ) , the hydrophobic group of an added alcohol preferentially goes to the surface @xcite .
the hydrophobicity of the alcohols increases with increase in the chain length of alcohols and at the interface , surface tension decreases with increase in number of @xmath0 group according to traube s rule @xcite .
+ the present paper deals with the experimental verification of the effect of hydrophobic interactions in the formation of self - assembled droplet formation and traube s rule of surface tension at interfaces .
the effect of contact line dynamics of these alcohols on the constrained surface is also investigated .
when a liquid droplet is placed on a solid surface having heterogeneities , the contact line can stay pinned until the contact angle reaches a critical angle @xcite . the contact angle for smooth surface under thermodynamic equilibrium conditions
is defined by young s equation , which is given as @xcite , @xmath1 where @xmath2 , @xmath3 and @xmath4 are the solid - vapour , solid - liquid and liquid - vapor interfacial tensions respectively and @xmath5 is the young s contact angle which is the intrinsic contact angle of the surface at the three phase contact line . for a heterogeneous rough surface
, the contact angle can vary over a finite range @xcite . on a smooth surface ,
the contact line will not advance ( recede ) until the contact angle reaches a critical value of the advancing ( receding ) contact angle .
the difference between the advancing and receding contact angle is called contact angle hysteresis , which characterizes the strength of pinning of contact line at the asperities of the rough surface @xcite . + after pinning
the droplets start to evaporate .
there are two different modes of evaporation of a liquid droplet on a polymer surface , namely , the constant contact radius ( ccr ) mode and the constant contact angle ( cca ) .
the ccr mode is characterized by the pinned contact line and decreased contact angle whereas the cca mode is characterized by the receding contact line and constant contact angle @xcite .
+ the behavior of droplet evaporation is also affected by the surface morphology and chemical composition .
orejon et al.@xcite in their investigation of the dynamics of the three - phase contact line of evaporating droplets on different surfaces with varying substrate hydrophobicity proposed that the depinning of contact line on rough surfaces occurs when the unbalanced young s force is large enough to overcome an intrinsic energy barrier arising from surface roughness .
ccr - cca transition , a consequence of the pinning - depinning phenomenon was observed during the evaporation on the surfaces with relatively large stripe widths @xcite .
commercially available digital versatile disk was used as substrate in the present study . in the present study , poly - dimethyl siloxane ( pdms ) from sigma - aldrich and chloroform ( purity 99.9 % ) purchased from merck ( emplura )
were used as polymer and solvent respectively .
methanol , ethanol , n - butanol and 2-propanol purchased from merck ( emplura ) were used as environment .
+ the aluminium layer in the dvd were peeled off using forceps and the side facing the aluminium layer was used as constrained surface consisting of parallel grooves .
the other side was used as the smooth surface . + for the preparation of patterned films using the supersaturated vapor of methanol , 5 ml of methanol is added to the petri - dish and kept inside the glass chamber . the polymer solution was prepared and casted on the surface of the disk .
the casted disk was then placed inside the chamber containing methanol vapors .
after complete evaporation of solvent , the films were characterized using confocal laser scanning microscope ( clsm ) .
same procedure is followed to prepare patterned films with other vapors namely ethanol , 2-propanol and n - butanol .
+ in the beginning of the formation of self - assembled droplet pattern , the non - aqueous vapor starts to condense due to evaporation of the solvent .
the droplets condense on the polymer solution surface . spreading and dissolving of droplets in the polymer solution
should nt take place if pores are to be formed .
the spreading behavior of the droplet on the surface is defined using spreading co - efficient as , @xmath6 where @xmath7 and @xmath8 are the surface tension of the solution and condensed liquid respectively .
@xmath9 is the interfacial tension of the liquid and solution .
@xmath10 indicates spreading behaviour is more and @xmath11 indicates non - spreading behaviour of the liquid .
once the solvent and liquid completely evaporates , the polymer film is left with array of pores .
the patterned films on smooth and constrained surfaces using methanol , ethanol , 2-propanol and n - butanol vapours were characterized for surface morphology using clsm and are shown in figure [ 1 ] . with methanol and ethanol vapours ,
the films were patterned with pores .
the smooth surface showed a hexagonal array of self - assembled droplet pattern as expected due to marangoni convection and formation of a lubricating thin polymer layer as a result of instant precipitation of the polymer@xcite .
the instant precipitation prevents the droplets from coalescing with each other thus forming hexagonal pore patterns .
the clsm images of grooved surface showed a different pattern of pores .
this distortion in pattern formation is ascribed to the depinning mechanism of contact lines .
this depinning occurs due to the presence of constraints on the underlying surface @xcite .
+ for a droplet on a rough surface having number of grooves parallel to each other , the contact line can align in two ways .
it can lie either parallel to the grooves and get pinned or lies at an angle to the grooves . in the latter case
, the contact line is displaced continuously without any pinning @xcite .
this prevents the self - assembly of droplets into a hexagonal pattern . as a result ,
a distorted hexagonal ring of pores with cluster of pores inside is observed .
+ the clsm images of surfaces in 2-propanol and n - butanol environment does nt show any patterns and is shown in figure [ 1 ] .
the 3d images of the polymer film formed over the substrate is shown in figure [ 2 ] . for 2-propanol and n - butanol
, there is an increase in @xmath0 group .
this increase in @xmath0 group reduces the surface tension which in turn affects the contact area of the droplet .
the increase in droplet contact area favours the coalescence of droplets . on complete evaporation of the solvent , a layer of polymer film without any significant patterns
is observed .
this result also proves the traube s rule relating surface tension of organic compounds to the number of hydrocarbon @xmath0 groups present in the molecule . according to traube s rule , for every extra @xmath0 group , the reduction in surface tension triples @xcite .
the formation of self - assembled droplet pattern on the pdms polymer coated over grooved side of dvd under saturated vapors of alcohols is reported .
it is observed that with the increase in alkyl chain of alcohols , the formation of droplets pattern decreases .
hexagonal pattern of self - assembled droplets on smooth surface was observed for surfaces patterned under methanol and ethanol vapors .
a distorted hexagonal ring with cluster of pores inside was observed for grooved surface under same vapor conditions .
the distortion in pattern formation is attributed to the depinning effect of contact line on rough surface . with 2-proppanol and n - butanol ,
no patterns of droplets were observed .
this confirms the traub s rule of decrease in surface tension with increase in @xmath0 group . | study of dynamics droplet s contact line on surfaces for applications in self - cleaning anti - corrosion bio - sensing contact line dynamics influenced by surface heterogeneities heterogeneities droplet pinning sites dynamics contact line for hydrophobic / super - hydrophobic surfaces studied dynamics during phase transition process like evaporation condensation less studied breath figure process self - assembled droplet patterns formed with super saturated vapor water ding et al first to formation self - assembled droplet pattern using non - aqueous vapors of ethanol methanol breath figure method suitable for water vapor compatible with organic solvents influence of constraints on surface on formation breath figures using non - aqueous vapor not reported alcohols are amphiphilic molecules hydrophobic and hydrophilic groups contact area of droplets when droplet liquid impinged on surface hydrophobic group of added alcohol goes to surface hydrophobicity increases with chain length of alcohols surface tension decreases with increase in number of grouppresent paper hydrophobic interactions in self - assembled droplet formation traube s rule of surface tension at interfaces effect contact line dynamics alcohols on constrained surface investigated liquid droplet on solid surface contact line pinned until reaches critical angle contact angle for smooth surface under thermodynamic equilibrium conditions defined by young s equation are solid - vapour liquid - vapor interfacial tensions young s contact angle intrinsic contact angle surface at three phase contact line heterogeneous rough surface contact angle vary over finite range on smooth surface contact line advance until contact angle reaches critical value difference called contact angle hysteresis characterizes strength of pinning contact line at asperities rough surface after pinning droplets start evaporate two modes of evaporation liquid on polymer surface constant contact radius constant contact angle ccr mode characterized by pinned contact line decreased contact angle cca mode receding contact line constant contact angle droplet evaporation affected by surface morphology chemical composition orejon et@xcite dynamics three - phase contact line evaporating droplets varying substrate hydrophobicity proposed depinning contact line on rough surfaces occurs when unbalanced young s force energy barrier surface roughness ccr - cca transition consequence pinning - depinning observed during evaporation large stripe widths digital versatile disk substrate study poly - dimethyl siloxane from sigma - aldrich chloroform purity 99.9 % from merck polymer solvent methanol ethanol n - butanol 2-propanol merck environment aluminium layer peeled off forceps side constrained surface other side smooth surface preparation patterned films supersaturated vapor methanol 5 ml methanol added petri - dish glass chamber polymer solution prepared casted on surface disk disk placed inside chamber containing methanol vapors after evaporation solvent films characterized using laser scanning microscope same procedure films with other vapors ethanol 2-propanol n - butanol formation self - assembled droplet pattern non - aqueous vapor evaporation solvent droplets condense on polymer solution surfacespreading dissolving droplets in polymer solution if pores formed spreading behavior droplet defined using spreading co - efficient as @xmath6 are surface tension solution liquid interfacial tension @xmath10 indicates spreading @xmath11 non - spreading once solvent liquid evaporates polymer film left with pores patterned films on smooth constrained surfaces using methanol ethanol 2-propanol n - butanol vapours characterized clsm shown in figure [ 1 ] with methanol ethanol vapours films patterned with pores smooth surface showed hexagonal self - assembled droplet pattern due to marangoni convection formation lubricating thin polymer layer instant precipitation precipitation prevents droplets coalescing forming hexagonal pore patterns clsm images of grooved surface showed different pattern pores distortion pattern formation to depinning mechanism of contact lines due to constraints on underlying surface droplet on rough surface contact line can parallel grooves or angle contact line displaced without prevents self - assembly droplets into hexagonal patterndistorted hexagonal ring of pores inside observed clsm images in 2-propanol n - butanol show patterns shown in figure [ 1 ] 3d images polymer film substrate in figure [ 2 ] 2-propanol n - butanol increase in @xmath0 group reduces surface tension affects contact area droplet favours coalescence evaporation solvent polymer film without patterns observed proves traube s rule surface tension to hydrocarbon @xmath0 groups extra group reduction surface tension triples formation self - assembled droplet pattern on polymer grooved side under saturated vapors alcohols reported increase alkyl chain formation pattern decreases hexagonal pattern on smooth surface observed for under methanol ethanol vapors distorted hexagonal ring cluster pores observed for grooved surface under vapor conditions distortion to depinning effect of contact line surface 2-proppanol n - butanol no patterns droplets observed confirms traub s rule decrease surface tension with increase @xmath0 group | 0.323547 | 2,027 | 977 | , Saving $0.1 in GPT-4. |
20,631 | we describe the dynamical properties which may be inferred from hst / stis spectroscopic observations of luminous compact blue galaxies ( lcbgs ) between @xmath0 . while the sample is homogeneous in blue rest - frame color , small size and line - width , and high surface - brightness , their detailed morphology is eclectic . here
we determine the amplitude of rotation versus random , or disturbed motions of the ionized gas .
this information affirms the accuracy of dynamical mass and m / l estimates from keck integrated line - widths , and hence also the predictions of the photometric fading of these unusual galaxies .
the resolved kinematics indicates this small subset of lcbgs are dynamically hot , and unlikely to be embedded in disk systems .
lcbgs , stis spectroscopy , internal kinematics , dynamics , evolution | the evolution of lcbgs is a matter of debate .
these galaxies are unusual in their blue colors , small sizes and line - widths , yet large luminosities .
we have suggested that at least a subset of these sources are the progenitors of des such as ngc 205 ( koo et al .
1995 , guzman et al .
1998 ) , while others counter these are bulges in formation ( hammer et al .
2001 ; barton & van zee 2001 ) .
surveys at intermediate redshift are not uniformly defined , and each contains heterogeneous samples objects span a range in size , color , luminosity , surface - brightness , and image concentration .
the broad `` lcbg '' class contributes as much as 45% of the comoving sfr between @xmath1 ( guzman et al .
1997 ) ; the proposed de progenitors are a fraction of this class . here
we focus on an extreme lcbg sub - class that are among the smallest , bluest and highest surface - brightness ( koo et al . 1995 ) : m@xmath2@xmath3@xmath4 ( h@xmath5=70 km / s / mpc , @xmath6=1 , @xmath7=0.7 ) , rest - frame b - v@xmath30.25 , half - light radii of @xmath8@xmath32 kpc , mean surface - brightness within @xmath8 of @xmath319 mag / arcsec@xmath9 ( rest - frame b band ) , and integrated line - widths of @xmath10@xmath365 km / s .
many of these are good candidates for de progenitors .
if so , their internal kinematics should reveal they are dynamically hot , while deep imaging should show they lack outer disks .
we have derived ionized - gas position velocity and position line - width diagrams from stis long - slit measurements along what appears to be the photometric major axes of 6 lcbgs between @xmath11 , and one other source at @xmath12 which is 2 - 3 mag lower luminosity than the others .
one example is given in figure 1 .
with 0.2 arcsec slits , stis delivers instrumental resolutions ( @xmath13 ) of 13 - 19 km s@xmath14 .
line - emission is not always centered on the continuum ( hoyos et al .
2004 ) ; the continuum centroid is adopted as the kinematic center .
we find keck hires integrated line - widths ( koo et al 1995 ) agree in the mean with the resolved velocity dispersions from stis spectroscopy : integrated dispersions are not due to large - scale , bulk , motion .
this secures our previous dynamical estimates of m / l and their use as constraints on photometric fading ( e.g. , guzman et al . 1998 ) .
only the low-@xmath15 , low-@xmath16 system shows clear rotation and substantially different integrated versus resolved line - widths .
-0.45 in -0.05 in -0.9 in versus ellipticity or inclination for lcbgs ( dotted stars ) , ngc 7673 ( circle / lower limit ) and early - types galaxies from simien & prugniel ( circles , dotted diamonds ) .
lines represent trajectories as a function of inclination for disk - galaxies of different luminosity assuming they lie on the tully - fisher relation and @xmath17 km s@xmath14 ( andersen et al .
shaded regions are adopted from kormendy & kennicutt ( 2004 ) .
, title="fig : " ] -1.6 in -0.1 in the lack of rotation coupled with their ellipticity squarely places these sources in the `` spheroidal '' region of the v/@xmath13 - ellipticity plane , illustrated in figure 2 . for fair comparison to local samples
, ellipticities are measured at the half - light radius near rest - frame v - band from hst images ; rotation velocities are set to half the difference between minimum and maximum velocities ; and @xmath13 is the observed central velocity dispersion .
while there is a range of observed v/@xmath13 , lcbgs lie well below the region inhabited by disk systems , with values comparable to local des and other spheroidals , particularly if recent observations of des with larger rotational components are considered ( pedraz et al .
2002 , van zee et al .
2004 ) : * lcbgs are dynamically hot*.
do we know that we are nt just sampling a bulge , or a face - on nuclear star - burst ?
ngc 7673 has been suggested by homeier et al .
( 2002 ) as a nearby example .
indeed this source has the right color , surface - brightness , size , luminosity and integrated line - width , and has a faint , extended outer disk .
however , homeier & gallagher s ( 1999 ) h@xmath18 velocity map shows clear evidence for rotation in the inner , star - burst region .
our re - analysis confirms this : we would see similar structure _ if it existed _ in our stis spectra .
independently , deep cfht imaging reveals no strong evidence for extended , normal disks around the types of ( and specific ) sources presented here ( barton et al . 2005 ) . in summary ,
the preponderance of evidence is against bulge formation in disk systems and in favor of a de - like descendant scenario _ for all of the specific sample presented here _ with m@xmath2@xmath19@xmath20 ( 6 out of 7 ) .
however , the disturbed morphology and kinematics makes clean interpretations difficult . what is the gas really telling us about dynamics ?
are these systems in dynamical equilibrium ? while their morphology and resolved kinematics would argue otherwise , the agreement between integrated velocity dispersions and resolved profiles indicates the systems can not be too far out of equilibrium .
stellar velocity and dispersion profiles would provide a much clearer dynamical picture .
finally we comment on issues raised at the conference about environment : is ngc 205 a good example of a faded , lcbg descendant ?
if so , where are the m31-like neighbors ? are there field des ? if dsphs are the low - mass cousins of des , then the presence of isolated dsphs such as tucana , cetus , and the recently discovered apples 1 ( pasquali et al .
2004 ) should give us pause about accepting assertions that des do not exist outside of rich environments or far from massive galaxies .
the space density of lcbgs presented here is 1.25@xmath210.15@xmath22 mpc@xmath23 for m@xmath24 between @xmath25 . even allowing a factor of @xmath26 higher relic density ( given the time interval in this redshift slice and assuming
the lcbg phase is a few@xmath2710@xmath28 yr ) to find even one descendant requires an all - sky local survey volume reaching out to @xmath310 mpc . at this distance
the half - light radius of ngc 205 is @xmath312 arcsec .
are our local surveys this complete ? | evolution lcbgs debate galaxies unusual blue colors small sizes line - widths large luminosities suggested sources progenitors des ngc 205 koo 1995 1998 others counter bulges formation hammer 2001 barton 2001 surveys intermediate redshift uniformly defined heterogeneous samples size color luminosity surface - brightness image concentration lcbg class contributes 45% comoving sfr between @xmath1 guzman 1997 proposed progenitors fraction focus on extreme lcbg sub - class smallest bluest highest surface - brightness koo 1995@xmath2@xmath3@xmath4=70 km / s @xmath7=0.7 rest - frame b - v@xmath30.25 half - light radii @xmath8@xmath32 mean surface - brightness @xmath319 mag line - of @xmath10@xmath365 km / s good candidates for de progenitors internal kinematics dynamically hot deep imaging lack outer disks derived ionized - gas position velocity position line - width diagrams from measurements photometric axes 6 lcbgs between @xmath11 one source at @xmath12 2 - 3 mag lower luminosityexample figure 1 0.2 arcsec slits stis delivers resolutions 13 - 19 km s@xmath14 line - emission not centered continuum 2004 continuum centroid kinematic center keck integrated line - koo 1995 agree resolved velocity dispersions stis spectroscopy dispersions not due to large - scale bulk motion secures dynamical estimates m / l constraints photometric fading low-@xmath15-@xmath16 system clear rotation different integrated resolved line - widths -0.45 -0.05 -0.9 in versus ellipticity for lcbgs ngc 7673 early - galaxies lines represent trajectories inclination disk - galaxies different luminosity tully - fisher relation @xmath17 km s@xmath14 shaded regions from kormendy & kennicutt 2004 -1.6 -0.1 lack rotation ellipticity places sources spheroidal region v/@xmath13 - ellipticity plane figure 2 ellipticities measured half - light radius near rest - frame v - band hst images rotation velocities half difference minimum maximum velocities @xmath13 central velocity dispersionrange observed v/@xmath13 lcbgs below region disk systems values comparable to local des spheroidals recent observations rotational components pedraz 2002 van zee 2004 lcbgs dynamically hot sampling bulge nuclear star - burst ngc 7673 suggested by homeier et al . 2002 nearby example source right color surface - brightness size luminosity integrated line - width faint extended outer disk homeier & gallagher s 1999 ) h@xmath18 velocity map shows evidence rotation in inner star - burst region re - analysis confirms similar structure in stis spectra deep cfht imaging no evidence for extended normal disks around sources barton 2005 ) preponderance evidence against bulge formation disk systems de - like descendant scenario m@xmath2@xmath19@xmath20 ( 6 out of 7 ) disturbed morphology kinematics interpretations difficult dynamics systems in dynamical equilibrium ? agreement integrated velocity dispersions resolved profiles indicates systems not out of equilibrium stellar velocity dispersion profiles clearer dynamical picturecomment issues conference environment ngc 205 faded lcbg descendant ? where m31-like neighbors ? field des ? dsphs low - mass cousins des presence isolated dsphs tucana cetus apples 1 pasquali et 2004 pause assertions des outside rich environments far massive galaxies space density lcbgs 1.25@xmath210.15@xmath22 mpc@xmath23@xmath24 between @xmath25 @xmath26 higher relic density lcbg phase@xmath2710@xmath28 yr descendant requires all - sky local survey @xmath310 mpc half - light radius ngc 205 @xmath312 arcsec local surveys complete ? | 0.262551 | 1,570 | 832 | , Saving $0.0 in GPT-4. |
20,640 | in this work we consider the problem of finding optimal regularization parameters for general - form tikhonov regularization using training data .
we formulate the general - form tikhonov solution as a spectral filtered solution using the generalized singular value decomposition of the matrix of the forward model and a given regularization matrix .
then , we find the optimal regularization parameter by minimizing the average of the errors between the filtered solutions and the true data .
we extend the approach to the multi - parameter tikhonov problem for the case where all the matrices involved are simultaneously diagonalizable . for problems where this is not the case
, we describe an approach to compute optimal or near - optimal regularization parameters by using operator approximations for the original problem .
several tests are performed for 1d and 2d examples using different norms on the errors , showing the effectiveness of this approach . spectral filtering , regularization , multi - parameter tikhonov , optimal filters , learning approach | we consider a linear ill - posed problem that can be modeled as @xmath0 where @xmath1 is the desired solution , @xmath2 models the forward process , @xmath3 is additive random noise , and @xmath4 is observed data .
the goal is to obtain an approximate solution of @xmath5 given @xmath6 and @xmath7 .
prior information regarding the probability distribution of @xmath8 may be incorporated , if known .
the matrix @xmath6 is ill - conditioned and its singular values decay to zero without significant spectral gap . due to the ill - posed nature of the problem , the exact ( inverse ) solution of will be contaminated by noise and may not be a good approximation of @xmath5 .
thus , we seek an approximation to the solution by solving a nearby problem that is well - posed , a process called _ regularization_. a well - known regularization method is tikhonov regularization @xcite , @xmath9 where @xmath10 is a regularization parameter and @xmath11 is a regularization matrix .
the matrix @xmath11 must have the same number of columns as @xmath6 , and we assume that the null spaces of @xmath6 and @xmath11 intersect trivially such that the matrix @xmath12 $ ] has full row rank and the solution @xmath13 is unique . if @xmath10 and @xmath11 are chosen appropriately , the solution to should approximate the desired solution @xmath5 .
typical choices of @xmath11 include the identity matrix and the discrete first or second derivative operators .
if @xmath14 , we say that is in _
standard form_. otherwise , we say it is in _
general form_. for the latter case , we can transform the problem to standard form .
that is , we can use the substitution @xmath15 if @xmath11 is invertible , or otherwise , we can define @xmath16 , where @xmath17 is the @xmath6-weighted generalized inverse of @xmath11 defined by @xmath18 ; see ( * ? ? ?
* section 2.3 ) for further details .
next we provide a motivating example to illustrate the potential benefits of considering the general - form tikhonov regularization . using the deriv2 test problem from @xcite ,
we provide in figure [ fig : differentl ] reconstructions for various choices of @xmath19 where in each case , the regularization parameter @xmath10 was selected to produce the smallest mean squared error . on the reconstructed solution . here , the true solution is provided , along with reconstructions corresponding to @xmath20 ( standard tikhonov ) and @xmath21 , where @xmath22 represents a discrete approximation of the first derivative . ] this example illustrates that @xmath14 is not suited for all problems and that better reconstructions may be obtained with @xmath23 .
we remark that this example is used for illustrative purposes only , so specific details are not included here .
complete numerical investigations will be presented in section [ sec : numerics ] .
the difficulty in selecting an appropriate regularization matrix has recently led researchers to investigate an extension of the general - form tikhonov regularization that contains several regularization terms instead of only one .
this extension is known as the multi - parameter tikhonov problem , @xmath24 where multiple regularization parameters @xmath25 $ ] and multiple regularization matrices @xmath26 are incorporated .
several studies on the multi - parameter tikhonov problem include @xcite .
a key ingredient for both the one - parameter and multi - parameter tikhonov problems is the choice of the regularization parameter(s ) .
standard techniques to find the regularization parameter for tikhonov regularization have been studied extensively in the literature and include the discrepancy principle , the l - curve criterion , and the generalized cross - validation ( gcv ) method @xcite .
some of these techniques have been extended to the multi - parameter tikhonov problem .
for example , a higher dimension l - curve is considered in @xcite , the discrepancy principle is discussed in @xcite , and an extension of gcv is described in @xcite . in @xcite , the regiska s parameter choice rule is generalized to the multi - parameter tikhonov case . in this paper , we propose a learning approach for computing _ optimal _ regularization parameters for both tikhonov problems and .
previous work on learning approaches in the context of regularization methods for solving inverse problems can be found in @xcite . however
, none of these works specifically address the general - form tikhonov and multi - parameter tikhonov problems .
we follow the work of chung et .
@xcite where training data is used to find optimal regularization parameters for a variety of spectral regularization methods , including standard form tikhonov regularization .
we first extend these approaches to the general - form tikhonov problem .
for general - form tikhonov , the generalized singular value decomposition ( gsvd ) @xcite of @xmath27 can be used to write the solution to as a spectral filtered solution .
we take advantage of this fact to develop efficient methods for computing an optimal regularization parameter , and compare its performance to that of filtered svd solutions on several numerical examples .
another main contribution of our work is to develop an efficient learning approach for the multi - parameter tikhonov problem .
the extension to the multi - parameter case is more complex , not only in terms of finding multiple @xmath10 s , but also because in general , there is no natural extension of the gsvd for @xmath6 and multiple regularization matrices @xmath28 . for problems where matrices @xmath6 and @xmath28 are _ simultaneously diagonalizable _
( sd ) , such as those arising in image processing , we propose an efficient learning approach for computing optimal regularization parameters .
however , for problems that do not exhibit this nice property , we follow a framework for regularization introduced in @xcite , where matrix approximations are used to compute regularization parameters , but the original problem is solved .
it is worth mentioning that in @xcite and @xcite , multi - parameter learning approaches for denoising problems ( @xmath29 ) are proposed . in
@xcite a parameter learning approach for multiple @xmath30-norm regularization terms is presented .
the learning problem is formulated as a bilevel optimization problem and solved using semismooth newton methods .
although the same techniques could be extended to our problem , we believe that such an approach applied to the problem where @xmath31 would be more computationally expensive than the one we propose here .
the paper is organized as follows .
background on the general - form tikhonov regularization problem is provided in section [ sec : background ] .
we provide the solution as a spectral filtered solution and describe some standard approaches for selecting regularization parameters .
in section [ sec : optfilt ] , we extend the optimal filter framework to both the general - form tikhonov regularization problem and the multi - parameter tikhonov regularization problem and describe an empirical bayes risk framework for computing optimal regularization parameters . in particular , numerical methods for computing optimal regularization parameters for the multi - parameter problem are discussed .
numerical results are presented in section [ sec : numerics ] , and conclusions and future work can be found in section [ sec : conclusions ] .
a closed form solution to the general - form tikhonov regularization problem can be obtained using the gsvd of the matrix pair @xmath32 .
given matrices @xmath33 , and @xmath34 , the gsvd can be obtained by first considering the _ reduced _ qr - factorization of the stacked matrix , @xmath35 = \left[\begin{array}{c } { { \bf q}}_{{{\bf a } } } \\ { { \bf q}}_{{{\bf l } } } \end{array}\right ] { { \bf r}}.\ ] ] we have @xmath36 and @xmath37 .
let @xmath38 and @xmath39 be the cs decomposition @xcite of \{@xmath40 } , where @xmath41 , @xmath42 , and @xmath43 are orthogonal matrices ; @xmath44 where @xmath45 is diagonal and @xmath46 is diagonal ( not necessarily square ) , satisfying @xmath47 .
the existence of the cs decomposition can be found in ( * ? ? ?
* section 22.1 ) . then the gsvd of @xmath32 is defined as follows @xmath48 where for convenience
, we have ordered the entries in @xmath49 and @xmath50 such that @xmath51 @xmath52 using the gsvd , the general - form tikhonov solution can be written as @xmath53 where @xmath54 is a diagonal matrix with _ filter factors _ ,
@xmath55 on the diagonal .
notice that for @xmath56 , @xmath57 can be written as @xmath58 , where @xmath59 are the generalized singular values .
it is easy to see that for standard - form tikhonov regularization where @xmath60 and @xmath49 contain the right singular vectors and singular values of @xmath6 ( with reordering ) and @xmath61 for all @xmath62 , the filter factors are given by @xmath63 .
therefore , by allowing @xmath23 , we not only affect the definition of the filter factors , but also alter the basis vectors @xmath64 used to represent the regularized solution @xmath13 .
for ill - posed inverse problems , it is well known that the problem satisfies the discrete picard condition @xcite .
that is , the values @xmath65 decay on average faster than the values @xmath66 , until an index is reached where the noise components dominate the solution .
after that point , the coefficients @xmath65 stabilize around the noise level and @xmath66 s continue decreasing , resulting in amplification of errors in the reconstruction .
a sample picard plot is provided in figure [ fig : picard ] .
since the signal is contained primarily in the subspace spanned by @xmath64 for small @xmath62 and including @xmath64 for larger @xmath62 results in errors , a good value of the regularization parameter @xmath10 should filter out the terms in for larger values of @xmath62 .
in other words , the filter factors determined by @xmath10 should go to zero to counteract the amplification of the error .
figure [ fig : picard ] shows filter factors corresponding to the parameter @xmath10 that produces the smallest mean squared errors . [ cols="^,^ " , ]
in this paper , we described a learning approach to compute optimal regularization parameters for general - form and multi - parameter tikhonov regularization .
we formulated the problem as an empirical bayes risk minimization problem and developed numerical algorithms for computing the optimal or near - optimal regularization parameters .
various error measures were considered .
numerical results illustrate that gsvd filtered solutions give better performance than optimal svd filtered solution , with less training data .
in addition , optimal parameters for multi - parameter tikhonov can provide results comparable to gcv , with less computational cost for a large set of images , and these parameters can be used to reconstruct different types of images .
f. s. viloche bazn , l. s. borges , and j. b. francisco . on a generalization of regiska s parameter choice rule and its numerical realization in large - scale multi - parameter tikhonov regularization .
, 219(4):21002113 , 2012 .
j. liu , t. liu , l. de rochefort , j. ledoux , i. khalidov , w. chen , j. tsiouris , c. wisnieff , p. spincemaille , m. prince , and y. wang .
morphology enabled dipole inversion for quantitative susceptibility mapping using structural consistency between the magnitude image and the susceptibility map .
, 59(3):25602568 , 2012 . | consider linear ill - posed problem modeled as @xmath0 @xmath1 desired solution forward process additive random noise observed data goal obtain approximate solution of @xmath5 @xmath6 @xmath7 information probability distribution @xmath8 incorporated matrix @xmath6 ill - conditioned values decay to zero without spectral gap exact solution contaminated by noise not good approximation of @xmath5 seek approximation solving nearby problem well - posed process regularization_ regularization method tikhonov regularization @xmath9 @xmath10 regularization parameter @xmath11 regularization matrix matrix @xmath11 same columns as @xmath6 null spaces of intersect matrix @xmath12 full row rank solution @xmath13 unique if @xmath10 @xmath11 chosen appropriately solution approximate desired solution @xmath5 typical choices of @xmath11 include identity matrix discrete first or second derivative operators @xmath14 _ standard form_ _ general form_ transform problem to standard formuse substitution @xmath15 if @xmath11 invertible define @xmath16 @xmath17 @xmath6-weighted inverse of @xmath11 defined by @xmath18 see section 2.3 ) for details motivating example benefits general - form tikhonov regularization deriv2 test problem from @xcite figure reconstructions for choices of @xmath19 regularization parameter @xmath10 selected smallest mean squared error true solution provided reconstructions to @xmath20 and @xmath21 @xmath22 represents discrete approximation of first derivative example illustrates @xmath14 not suited for all problems better reconstructions with @xmath23 example for illustrative purposes specific details not included complete numerical investigations in section numerics difficulty selecting appropriate regularization matrix led researchers extension tikhonov regularization several regularization terms one extension multi - parameter tikhonov problem @xmath24 multiple regularization parameters @xmath25 regularization matrices @xmath26 incorporated studies on multi include @xcite key ingredient choice of regularization parameter(s )techniques to find regularization parameter for tikhonov studied include discrepancy principle l - curve criterion generalized cross - validation gcv method techniques extended to multi - parameter tikhonov problem higher dimension l - curve discrepancy principle extension of gcv regiska s parameter choice rule generalized to multi - parameter tikhonov case propose learning approach for computing optimal regularization parameters for tikhonov problems previous work on in @xcite address general - form multi - parameter tikhonov problems follow work of chung et @xcite training data optimal regularization parameters for spectral regularization methods including tikhonov extend approaches to general - form tikhonov problem generalized singular value decomposition ( gsvd ) of @xmath27 used to write solution spectral filtered solution develop efficient methods for computing optimal regularization parameter compare performance to filtered svd solutions on examples contribution efficient learning approach for multi - parameter tikhonov problem extension to multi - parameter case complex no natural extension of gsvd for @xmath6 and multiple regularization matricesproblems matrices @xmath6 @xmath28 simultaneously diagonalizable propose efficient learning approach computing optimal regularization parameters problems follow framework regularization @xcite matrix approximations compute regularization parameters original problem solved @xcite multi - parameter learning approaches for denoising problems @xmath29 ) proposed @xcite parameter learning approach for multiple @xmath30-norm regularization terms presented learning problem bilevel optimization problem solved using semismooth newton methods techniques could extended to problem approach @xmath31 more computationally expensive paper organized background on general - form tikhonov regularization problem section solution spectral filtered solution standard approaches for selecting regularization parameters section extend optimal filter framework to general - form multi - parameter problem describe empirical bayes risk framework for computing optimal regularization parameters numerical methods for computing optimal regularization parameters multi - parameter problem discussed numerical results in section [ : numerics conclusions future work section closed form solution general regularization problem using gsvd of matrix pair @xmath32 matrices @xmath33 @xmath34 gsvd obtained considering reduced qr - factorization stacked matrix @xmath35 =\left\begin{array\bf\bf a\bf l\end{array\bf r}} @xmath36 @xmath37 @xmath38 @xmath39 cs decomposition of@xmath40 @xmath41 @xmath43 orthogonal matrices @xmath44 @xmath45 diagonal @xmath46 diagonal not square satisfying @xmath47 cs decomposition in section 22.1 gsvd of @xmath32 defined @xmath48 ordered entries @xmath49 @xmath50 @xmath51 @xmath52 gsvd general - form tikhonov solution written as @xmath53 @xmath54 diagonal matrix filter factors @xmath55 diagonal @xmath56 @xmath57 written as @xmath58 @xmath59 generalized singular values standard - form tikhonov regularization @xmath60 @xmath49 contain right singular vectors singular values of @xmath6 @xmath61 @xmath62 filter factors given by @xmath63 allowing @xmath23 definition filter factors alter basis vectors @xmath64 regularized solution @xmath13ill - posed inverse problems problem satisfies discrete picard condition @xcite values @xmath65 decay faster than @xmath66 until noise components dominate coefficients @xmath65 stabilize around noise level @xmath66 amplification errors reconstruction sample picard plot figure signal in subspace @xmath64 small @xmath62 larger results errors regularization parameter @xmath10 filter terms larger values @xmath62 filter factors @xmath10 to zero counteract amplification error figure shows filter factors parameter @xmath10 smallest squared errors paper described learning approach compute optimal regularization parameters for general - form multi - parameter tikhonov regularization formulated problem empirical bayes risk minimization problem developed numerical algorithms computing optimal parameters error measures considered results gsvd filtered solutions better performance than optimal svd filtered less training data optimal parameters multi - parameter tikhonov provide results comparable to gcv less computational cost for large set images reconstruct different types images francisco generalization of regiska s parameter choice rule numerical realization in large - scale multi - parameter tikhonov regularization219(4):21002113 2012 liu de rochefort ledoux khalidov chen tsiouris wisnieff spincemaille prince y wang dipole inversion quantitative susceptibility mapping structural consistency magnitude image susceptibility map 59(3):25602568 2012 | 0.427029 | 2,627 | 1,330 | , Saving $0.1 in GPT-4. |
20,644 | gamma ray bursts ( grbs ) are known to have short - time variability and power - law behavior with the index @xmath0 in the power spectrum density . reanalyzing the expanded data , we have found a ) the power - law comes from the global profile of the burst and not from the self - similar shots nor rapid fluctuations in the luminosity profile .
b ) the power indices vary from burst to burst and the value @xmath0 is given simply as the mean value of the distribution ; there is no systematic correlation among grbs to yield the power law . | the origin and the fundamental mechanism of the gamma - ray bursts ( grbs ) have not been revealed despite the fact that more than thirty years have passed since the first discovery of them . in the numerous luminosity profiles of grbs
, we notice that most grbs have very rapid milli - second time variability in their intensity @xcite though there is vast variety in luminosity profile itself for each grb .
we believe that this characteristic variability must have important information to reveal the grb mechanism .
therefore , in this letter , we would like to concentrate on the analysis of this variability . in studying time sequence of objects ,
the fourier transform technique is useful and often yields indispensable information on the scaling properties and characteristic time scale .
for example , @xcite used this method for 214 light curves of long grbs ( @xmath1 ) and reported that the averaged power spectrum density ( psd ) of grbs shows the power - law with the index @xmath2 over two decades in frequency range .
this value is very closed to @xmath3 which suggests the kolmogorov spectrum of velocity fluctuations in turbulent medium . on the other hand
, the power - law in psd does not necessarily specify the whole mechanism as we know in various examples .
actually the power - law in psd can be derived by many reasons ; some specific burst profile in luminosity , levy - type random noise in the background , and the superposition of similar shots and so on .
for example , the power - law in psd of x - ray emission from the active galactic nuclei ( agn ) is considered to be originated from the superposition of many similar shots @xcite .
@xcite tried to reproduce the power - law index , -5/3 , and the individual psd distribution of grbs by the superposition of many decaying pulse shots . here in this paper
, we would like to determine the origin of the power - law in psd of grbs by analyzing the detail of psd for each burst data .
first , we demonstrate that the individual psd of grbs does not exactly have the power - law index @xmath0 but has wide variation which is well approximated by the gaussian distribution .
second , we argue that power - law index of psd is determined by the profiles of an individual shot in the light curve and not by the superposition of many similar shots nor by levy - type random noise .
then we discuss the general nature of psd of superposed shots and apply this method to the actual data of grb light curves .
finally we observe how the averaged psd shows clear power - law index @xmath0 .
in our analysis , we use the data detected by the burst and transient source experiment ( batse ) on compton gamma ray observatory ( cgro ) with 64ms resolution .
we use the light curves in the energy band @xmath4 kev , excluding unclear high and low energy bands , and subtract the background by the linear fitting .
then for excluding the noise contamination , we select the data with peak count rate larger than 250 counts per 64ms bin ; finally 297 data set remain . in order to see the generality of the power index @xmath3
, we first calculate the power spectra , which is the absolute square of the fourier transform of the time sequence , for individual light curves .
we assume the following fitting function for the power spectrum @xmath5 , expecting the coexistence of the power - law component and the thermal white - noise component .
@xmath6 figure [ in.dist ] shows the distribution of power - law index @xmath7 thus obtained .
the distribution is well fitted by the gaussian and the mean value is @xmath8 , which is close to the value @xmath0 reported in the reference @xcite .
we emphasize here that the individual data has the power - law psd though the power index is widely distributed with the variance 0.65 .
now let us consider the meaning of the power - law in psd .
as is well known , the power - law itself in the psd can be realized in various origin .
the specifications of the origin is significant for the analysis of the central engine of grbs .
@xcite shows at least the following three ways to realize the power - law in psd . 1
. specific global profile of a burst .
2 . levy - type random process .
3 . superposition of self - similar shots .
we examine the above possibilities in the following in this order . examining the first possibility
, we realize that each grb data generally has multiple shots and rapid fluctuations as well as noise whose origin is not seem to be grb .
we try to identify these components in each grb data step by step .
first we discuss the light curve with a single shot .
this simple type of light curve shows a clear power - law behavior in its psd .
we consider the following two model cases ; a ) the decay type , and b ) the grow - and - decay type .
\a ) the decay type the decay type shot has the flux @xmath9 at time @xmath10 as @xmath11 where the positive parameters @xmath12 , @xmath13 and @xmath14 respectively represent `` intensity '' , `` sharpness '' and `` inverse of the duration '' of the shot .
@xmath15 is the step function .
power spectrum @xmath16 for this shot is written as @xmath17 ^ 2 \left [ f_{0}^2 + f^2 \right ] ^{p-1},\ ] ] where @xmath18 is the gamma function .
\b ) the grow - and - decay type the grow - and - decay type shot generally has asymmetric profile in the burst ; we separate the grow ( left side ) and decay ( right side ) of the shot .
each side of the shot is written as @xmath19 power spectrum of this type is written as @xmath20 , \label{eq.gn_p}\end{aligned}\ ] ] where @xmath21 .
since the grb data generally has multiple shots , the above single shot profiles must be superposed before we use .
for simplicity for fitting procedure , we assume that all shots in the individual light curve has the same profile except the overall amplitude ; as a fitting parameter , but it does not change the index of psd at all . ]
i.e. the @xmath22-th shot in the light curve @xmath23 is written as @xmath24 , where @xmath25 is the relative amplitude and @xmath26 is the location of the @xmath22-th shot with @xmath27 being the fixed function for each grb data .
then we can easily calculate the psd for multiple shots since the fourier transforms of the first and the k - th shots are simply related with each other as @xmath28 .
the light curve @xmath29 which has @xmath30 shots @xmath31 has the psd @xmath32,\end{aligned}\ ] ] where @xmath5 is psd for @xmath27 .
it is apparent that the last expression consists of the `` constant '' term and the `` oscillating '' term ; the former term determines the global profile of psd and the latter fluctuations around it .
we can actually see this structure in psd of grb data .
we first identify the maximum shot in the grb data and locally fit this shot by the single - shot form argued in the above .
then we subtract this first fit from the original data yielding one - shot subtracted data .
second , we identify the maximum shot in this subtracted data and locally fit this shot by the single - shot form with the same parameter @xmath13
. then we subtract this second fit from the one - shot subtracted data yielding two - shots subtracted data . after repeating this process several times
, the reduced data becomes almost flat .
we applied this method for many individual grb data .
one typical example for grb910602(#257 ) is shown in fig.[fit.lc ] . here
we use the decay type shot for fitting identifying two shots in the data .
the full psd of this light curve is plotted in fig.[fit.pw ] with the solid line .
the psd of the first fit is plotted with the chain line which already shows smooth power - law .
the psd of the superposition of the first and the second fits is plotted with the broken line which shows the oscillation around the chain line .
it is almost clear that the psd of the superposition of first few fits are sufficient to faithfully reproduce the main part of the original psd .
other grb data sets show similar behavior as this one .
thus we have confirmed the first possibility : specific global profile of a burst determines the power - law behavior in psd . in order to exclude the second possibility , the levy - type random process
, we have examined the peak distribution analysis .
we calculate the distribution function of all peak - intensity in the data of artificially produced levy - type light curve and that in the actual data of grbs .
the former shows a clear power law however the latter does not .
we can now claim that the power - law in psd of grbs is not originated from the levy - type random process . in order to exclude the third possibility i.e. superposition of self - similar shots
, we calculate the psd of the few - shots subtracted data . for the previous data
set , we calculate the psd of the two - shots subtracted data in fig.[fit.pw ] with the dotted line .
this residual component is 10 - 100 times smaller than the main component and is almost flat . if grbs have self - similar structure in their light curves , the few - shots subtracted data should also show power law ( but its range must be narrower than that of original data ) .
thus we can now claim that the power - law in psd of grbs is not originated from the superposition of self - similar shots we have also applied the same analysis to the data set of the soft gamma - ray repeaters ( sgr ) .
we use the data of sgr1900@xmath3314 .
the first one - shot fit has been enough to reproduce the power - law behavior in its original psd is @xmath34 and power index of original psd is @xmath35 . ] and the one - shot subtracted data shows a flat psd except several very sharp peaks which are considered to be originated from the central pulsar of sgr .
we have averaged all the 297 psd of grb data and obtain the figure [ fit.mu ] . before taking the average
, we have normalized each grb luminosity data so that the maximum of the count rate be unity .
this is because the original light curves have two to three orders of difference in maximum count rates and therefore the naively averaged psd is determined by few grb data .
since the magnitude of the count rate itself is not physical ( we do nt know the distance ! ) , such discrimination is not a proper manipulation .
we have also tried the total - count - rate ( fluence ) normalization as well as maximum - count - rate normalization ; yielding no significant difference between them .
thus averaged psd shows much clear power - law ( fig.[fit.mu ] ) than the individual psd of grb data .
@xmath36 from fig.[fit.mu ] we again observe that the individual psd shows power - law and the power index simply fluctuates variously .
after the superposition of many psd data , there appears smooth power - law behavior with the power index of the central value .
it is important to observe in fig .
[ fit.mu ] that there is no systematic correlation in each psd to yield the global smooth power - law behavior .
especially there is no systematic distribution of time scales nor correlation of turning points which , if any , would have yielded clear envelope when many psd are superposed . the above reasoning for the appearance of the smooth power - law in averaged psd is also supported by the fact that the power index @xmath0 which is close enough to the mean value @xmath8 of the index distribution within 14 percent of the variance .
if the clear power - law were realized by the envelope of many systematic distribution of burst time scales , the averaged power index would be significantly smaller than the mean power index .
analyzing 297 power spectrum density ( psd ) of grbs , we obtain the following results in this paper .
a ) individual grb data shows power law behavior in the psd . the power index ( @xmath7 ) distribution is well approximated by the gaussian form with the mean @xmath8 and the variance @xmath37 .
b ) power law behavior in psd for individual grb data is determined by the shot profile and not by the levy - type noise nor the superposition of many self - similar shots .
c ) power law index in averaged psd is simply determined by the mean value of the index ( @xmath7 ) distribution .
we found no correlation mechanism for producing clear power law in averaged psd . from the result b ) , we doubt the turbulence @xcite as the origin of the power law in psd of grbs .
this is because the self - similar cascade of eddy , which would naturally yield many self - similar shots in the emission , yields the kolmogorov power law spectrum of velocity fluctuations in turbulent medium .
the result b ) provides sharp contrast with the case of power law in the x - ray emission from agn .
many authors have demonstrated that some self - similar cascade model , which yields many self - similar shots , successfully describes the emission profile and the power law in psd of agn .
thus we naturally expect that the emission mechanism of agn and that of grb are quite different with each other despite the similar power law behavior in their psd . as we have seen in our analysis
, some special form of shots , which yield the power law in psd , is very characteristic and would be a good criterion for restricting various models of grb generation mechanisms .
the special shot profile is more severe checking point of the validity of the model than simply the power - law behavior of psd @xcite .
an urgent interest then would be the question whether the popular `` internal shock model '' can explain these special shot profiles .
there are still important questions to be answered in the near future .
1 ) the shots in the light curves of grbs have distinctly different two types ; the decay type and the grow - and - decay type .
this property is different from that of blazars , even if their similarity in emission process is often suggested .
2 ) what is the distribution of the sharpness parameter @xmath13 and the ratio of power - law component and noise @xmath38 in equation ( [ fit.eq ] ) ? is it also gaussian or any other form from which we can extract the information of the distance or classification of grb ? | origin mechanism of gamma - ray bursts grbs not revealed despite thirty years since first discovery luminosity profiles grbs rapid milli - second variability in intensity vast variety in luminosity profile for each grb characteristic variability grb mechanism concentrate on analysis variability fourier transform technique yields information on scaling properties characteristic time scale @xcite used method for 214 light curves of long grbs reported averaged power spectrum density ( psd ) of grbs shows power - law with index @xmath2 over two decades frequency value closed to @xmath3 suggests kolmogorov spectrum of velocity fluctuations in turbulent medium power - law in psd specify whole mechanism derived by many reasons specific burst profile luminosity random noise superposition of similar shots power - law in psd x - ray emission from active galactic nuclei originated from superposition of many similar shots tried to reproduce power - law psd distribution by shots determine origin of power - law in psd of grbs by analyzing detail psd for each burst datademonstrate individual psd grbs power - law index @xmath0 wide variation approximated by gaussian distribution argue power - law index psd determined by profiles individual shot light curve not superposition shots random noise discuss nature psd superposed shots apply method to data grb light curves observe averaged psd shows clear power - law index @xmath0 use data burst transient source experiment compton gamma ray observatory 64ms resolution use light curves energy band @xmath4 kev subtract background by linear fitting select data with peak count rate larger than 250 counts per 64ms bin 297 data set remain generality power index @xmath3 calculate power spectra square fourier transform time sequence for individual light curves assume fitting function for power spectrum @xmath5 coexistence power - law thermal white - noise figure shows distribution of power - law index @xmath7 distribution fitted by gaussian mean value @xmath8 close to value @xmath0 reference @xcite individual data has power - law psd power index widely distributed variance 0.65 consider meaning power - law in psd realized various origin specifications origin significant for analysis central engine grbs @xcite shows three ways to realize power - law in psd 1 specific global profile of burst levy - type random process 3 superposition of self - similar shots examine possibilities order first each grb data has multiple shots rapid fluctuations noise origin not grb identify components in grb data step by step discuss light curve with single shot curve shows clear power - law behavior in psd consider two model cases decay type grow - and - decay type decay type shot has flux @xmath9 @xmath10 @xmath11 positive parameters @xmath12 @xmath13 @xmath14 represent intensity sharpness inverse of duration shot @xmath15 is step function power spectrum @xmath16 written as @xmath17 @xmath18 is gamma function grow - and - decay type shot has asymmetric profile in burst separate grow decay side shot each side written as @xmath19 power spectrum written as @xmath20 grb data has multiple shots single shot profiles must be superposed before use simplicity fitting procedure assume all shots light curve same profile except amplitude change index psd @xmath22-th shot curve @xmath23 written as @xmath24 @xmath25 relative amplitude @xmath26 location-th shot @xmath27 fixed function each grb data calculate psd for multiple shots fourier transforms first k - th shots related as @xmath28 light curve @xmath29 @xmath30 shots @xmath31 has psd @xmath32 @xmath5 psd for @xmath27 last expression consists constant oscillating term former determines global profile psd latter fluctuations structure in psd grb data identify maximum shot fit single - shot form subtract first fit data one - shot subtracted data identify maximum shot subtracted data fit shot single - shot form same parameter @xmath13 subtract second fit two - shots subtracted data repeating process several times reduced data becomes almost flat applied method for many grb data example grb910602(#257 ) fig.[fit.lc ] decay type shot fitting identifying two shots full psd light curve plotted in fig.[fit.pw ] solid linepsd first fit plotted chain line shows smooth power - law superposition first second fits plotted broken line oscillation around chain line psd fits original psd other grb data sets show similar behavior confirmed first possibility global profile burst determines power - law behavior exclude second possibility levy - type random process examined peak distribution analysis calculate distribution peak - intensity artificially produced levy - type light curve actual data grbs former shows clear power law latter not claim power - law psd not from levy - type random process exclude third possibility superposition self - similar shots calculate psd few - shots subtracted data calculate psd two - shots subtracted data fig dotted line residual component 10 - 100 times smaller main component almost flat if grbs self - similar structure light curves few - shots subtracted data show power law range narrower original data claim power - law psd grbs not from superposition self - similar shots applied analysis soft gamma - ray repeaters sgr data sgr1900@xmath3314 first one - shot fit power - law behavior original psd @xmath34 power index @xmath35one - shot data shows flat psd except sharp peaks from central pulsar sgr averaged 297 psd grb data figure normalized each grb luminosity data maximum count rate original light curves have two to three orders difference in maximum count rates averaged psd determined by few grb data magnitude count rate not physical distance discrimination not proper manipulation tried total - count - rate normalization maximum - count - rate normalization no significant difference averaged psd shows clear power - law than individual psd of grb data individual psd shows power - law power index fluctuates variously after superposition of many psd data smooth power - law behavior with power index central value no systematic correlation in each psd global smooth power - law behavior no systematic distribution of time scales nor correlation of turning points smooth power - law in averaged psd by power index @xmath0 close to mean value @xmath8 distribution within 14 percent of variance if clear power - law realized by distribution time scales averaged power index smaller than mean power index analyzing 297 power spectrum density ( psd ) of grbs resultsgrb data shows power law behavior in psd power index @xmath7 ) distribution approximated by gaussian form mean @xmath8 variance @xmath37 power law behavior psd determined by shot profile not levy - type noise superposition of self - similar shots power law index in averaged psd determined by mean value index @xmath7 ) distribution no correlation clear power law psd doubt turbulence @xcite origin of power law in psd grbs self - similar cascade of eddy yields kolmogorov power law spectrum velocity fluctuations in turbulent medium with power law x - ray emission from agn self - similar cascade model describes emission profile power law psd agn expect emission mechanism of agn grb different despite similar power law behavior special form shots yield power law in psd characteristic criterion for restricting models grb generation mechanisms special shot profile more severe checking validity model than power - law behavior of psd @xcite urgent internal shock model '' explain special shot profiles important questions to be answered shots in light curves of grbs types decay grow - and - decay typeproperty different blazars similarity emission process suggested 2 distribution sharpness parameter @xmath13 ratio power - law component noise @xmath38 equation fit.eq ] ) ? gaussian other form extract information distance classification grb ? | 0.503766 | 3,203 | 1,573 | , Saving $0.1 in GPT-4. |
20,647 | a detailed description is provided of a numerical model of the microcavity optical parametric oscillator ( opo ) .
the model is solved on a two dimensional grid , in the time domain , using an alternating - direction implicit method .
results are presented for a system pumped with a gaussian spatial profile beam .
it is shown that at high pump powers the opo signal forms a ring , with low emission at the centre where the pumping is strongest .
+ to be published in physica status solidi ( c ) | the polariton physics in a strongly - coupled semiconductor microcavity includes interesting non - linear optical effects .
when resonantly pumped with an off - axis laser , optical parametric oscillator ( opo ) behaviour can be obtained , with the spontaneous appearence of signal and idler emission@xcite .
since the excitonic non - linearity is @xmath0 , the phase matching conditon requires @xmath1 and @xmath2 , which means that the pump , signal and idler can all be close to resonance with the polariton dispersion .
this behaviour has been investigated theoretically by a number of authors , using analytic@xcite and numerical@xcite models . however , all these models differ from the experimental setup by assuming that the pump is a plane wave , with uniform excitation of the system , and uniform polariton fields .
while such treatments provide useful insights into the opo , a fuller understanding of the observed behaviour requires a theory which takes into account the finite excitation spot used in experiments .
the purpose of this paper is to describe a numerical treatment of a two dimensional microcavity polariton system pumped with a gaussian beam , solved in the time domain on a discrete spatial grid .
the model for the opo is based on the non - linear optics treatment of ref.@xcite .
it consists of coupled two - dimensional cavity and exciton fields , @xmath3 and @xmath4 , with a local @xmath0 non - linearity in the exciton to represent the effects of short - range interaction .
both fields have dispersions , described by effective masses @xmath5 and @xmath6 , and are damped with widths @xmath7 and @xmath8
. the coupling is characterised by @xmath9 , the rabi splitting , and @xmath10 , the detuning at normal incidence .
the cavity field is driven by a spatially dependent external pump field @xmath11 .
the time evolution of the fields is thus determined by @xmath12 to be exact , the form for the @xmath0 nonlinearity should be @xmath13 . however , for a numerical treatment
, it is better to chose @xmath14 , as this is independent of the origin of energy , which can thus be shifted to the exciton frequency .
the modified form correctly gives the scattering effects of interest here , but does not describe frequency multiplying effects , which would require extremely small time - steps in the numerics . for the numerical treatment ,
this model is discretized on a two dimensional grid , with the differential operators are replaced by hopping terms .
the solution is obtained in the time domain , as an initial value problem , using the crank - nicholson method@xcite .
this leads to a semi - implicit time - step formula of the form @xmath15 where the vector @xmath16 describes the cavity photon and exciton amplitudes on the site @xmath17 at time - step @xmath18 .
the precise form of the matrix operator @xmath19 will be specified below .
the problem with solving eq.([eq : crank ] ) in two dimensions is that @xmath19 is a large matrix ( size @xmath20 , where n is the length of the edge of the grid ) , whose structure does not lend itself to efficient numerical solution : there are non - zero terms far from the diagonals .
it is better to use an operator - splitting treatment , the alternating direction implicit method@xcite , writing @xmath21 , where @xmath22 and @xmath23 respectively contain only @xmath24 and @xmath25 hopping terms .
then , each time - step is divided into two sub - steps : first , @xmath26 is solved for every @xmath27 , then @xmath28 for every @xmath29 .
each of these one dimensional strip problems is independent , so it is necessary to solve only @xmath30 @xmath31 sets of equations .
moreover the matrices @xmath22 and @xmath23 contain non - zero terms only on the diagonal and first three sub - diagonals , which permits a very efficient solution by factorisation@xcite . the part of @xmath22 involving coupling to site @xmath29 is @xmath32 where each @xmath19 is a @xmath33 matrix : @xmath34 the hopping terms are @xmath35 and @xmath36 , with @xmath37 the @xmath24-direction grid spacing .
the value of @xmath4 used in the non - linear term is the old value at that site , ie from the right hand side of eq.([eq : xup ] ) or eq.([eq : yup ] ) .
@xmath23 has a similar form , except @xmath38 replaces @xmath37 , though in practice an isotropic grid is used .
the factors of @xmath39 multiplying @xmath40 are needed to avoid over - counting , because they occur in both @xmath22 and @xmath23 .
eq.([eq : xup ] ) , and similarly eq.([eq : yup ] ) , is most efficently solved in two stages : first an intermediate variable @xmath41 is calculated using @xmath42 then it is straightforward to show that @xmath43 . to incorporate periodic boundary conditions , eq.([eq : strip ] ) has to be modified by adding hopping terms in the off - diagonal corners of the matrix , coupling the sites at either end of a strip : @xmath44 adding these terms breaks the useful diagonal form of the matrix , so a direct solution of eq.([eq : periodic ] ) would be inefficient .
however , as there are just four extra non - zero entries , the problem can be handled at little numerical cost using the woodbury formula@xcite to incorporate the extra terms .
@xmath45 is written as @xmath46 where the @xmath47 and @xmath48 are chosen to generate the end hopping terms .
an efficent choice is @xmath49 where @xmath50 and @xmath51 are the first two elements along the main diagonal of @xmath22 .
@xmath52 is a slightly modified version of @xmath22 : the @xmath53 terms produce , as well as the required hopping terms , four additional elements along the main diagonal which have to be subtracted from @xmath22 to obtain the correct @xmath45 .
evaluating the woodbury formula requires additional solutions of eq.([eq : trick ] ) with @xmath54 and @xmath55 on the right hand side , but this is very inexpensive once the factorisation of @xmath52 is obtained . hence adding periodic boundary conditions makes little difference to the numerical costs .
the results described in this section are for a microcavity with physical parameters typical of a good gaas structure : the rabi splitting , @xmath56 mev , and the line - widths of the exciton and cavity mode are @xmath57 mev .
the structure is on - resonance at normal incidence ( @xmath58 ) , and pumped at 16@xmath59 with a gaussian beam whose fwhm is 17.05@xmath60 m .
the cavity mode effective mass is taken to be @xmath61 , and the exciton mass @xmath62 .
the spatial grid consists of @xmath63 points with a separation of 0.4@xmath60 m , and the time step is 12 fs .
the calculation is started with the fields set to zero , and the solution is then propagated until a steady state is reached , typically after about @xmath64 time steps or @xmath65 ps .
the pump , signal and idler are then separated by storing the data for a series of time steps , fourier transforming to the frequency domain , and applying band - pass filters .
this process can be carried out in a continuous way using a rolling time window , which allows the observation of dynamical effects occuring on time - scales slow compared to the inverse of the mode separation .
the calculated images show the spatial dependence of the steady - state cavity field intensities in the signal and idler modes at various pump powers . it should be noted that they represent the internal cavity field , which can not be directly observed experimentally .
the emitted light will undergo some additional spatial filtering due to the small @xmath66-width of the polariton branches . with various powers .
( a ) and ( e ) are the signal and idler intensities for a peak power of 0.2 , just above threshold .
the following rows represent pump powers of 0.3 , 0.4 and 0.5 , with the signal states on the left , the idlers on the right .
the lengths of the edges of the squares are 51.2@xmath60 m , and the fwhm of the pump beam is 17.05@xmath60 m . ]
fig.1 shows the calculated signal and idler images for the steady state at various pump powers above threshold .
close to threshold , ( a ) and ( e ) , only a small spot near to the centre of the image is bright .
this is easily understood : for a gaussian beam profile , the threshold will be reached first where the pump is strongest . at higher pump powers , the emission spot expands , but , instead of the centre becoming brighter , the spot turns into a ring with low intensity in the middle .
it can be explained qualitatively using an analytic plane - wave model where the contributions of the signal and idler to the polariton blue - shifts are taken into account@xcite .
the shifts cause the opo to switch off above a certain power , which depends on the incidence angle and detuning of the pump beam .
this happens first at the centre of the spot , leading to the dark middle . in a gaussian beam
, there will always be a region where the pump power is the range to allow the opo , so an expanding emission ring should be expected as the power is increased above threshold .
of course , the strong spatial dispersion of the polariton means that a such a simple picture , relating the opo intensity to the local value of the pump , is not completely accurate , but it seems to give a reasonable guide .
the dispersion effects are responsible for the fact that the ring is not perfectly symmetrical , and breaks up significantly at higher pump powers
. however , a bilateral symmetry is always mainatined , defined by the incidence plane , which cuts the figure along the @xmath24-axis .
finally , at higher powers than those shown in the fig.1 , the numerics predict that a steady state may not be reached .
instead , the signal and idler undergo a periodic oscillation , maintaining the general ring shape , but with the maximum intensity moving from side to side .
it has been shown that a full understanding of the microcavity opo requires a two - dimensional treatment of the spatial structure of the pump beam and the polariton fields .
a numerical model describing the spatial and temporal evolution of the system has been derived , and used to predict that the emission spot should form a ring at high pump powers .
these results demonstrate that experimental measurements without spatial resolution provide only a limited picture of the opo behaviour .
n. a. gippius , s. g. tikhodeev , v. d. kulakovskii , d. n. krizhanovskii and a. a. tartakovskii , cond - mat/0312214 ( 2003 ) ; v. d. kulakovskii , d. n. krizhanovskii , a. a. tartakovskii , n. a. gippius and s. g. tikhodeev , uspekhi fiz .
nauk * 173 * , 995 ( 2003 ) . | polariton physics in coupled semiconductor microcavity includes non - linear optical effects pumped with off - axis laser optical parametric oscillator behaviour spontaneous signal idler emission@xcite excitonic non - linearity @xmath0 phase matching conditon requires @xmath1 @xmath2 pump signal idler resonance with polariton dispersion investigated theoretically numerical models pump plane wave uniform excitation uniform polariton fields understanding requires theory finite excitation spot paper numerical treatment of two dimensional microcavity polariton system pumped with gaussian beam time discrete spatial grid model opo based on non - linear optics treatment.@xcite coupled two - dimensional cavity exciton fields @xmath3 @xmath4 local @xmath0 non - linearity in exciton range interaction fields have dispersions effective masses @xmath5 @xmath6 damped with widths @xmath7 @xmath8 coupling characterised by @xmath9 splitting @xmath10 detuning at normal incidence cavity field driven by spatially dependent external pump field @xmath11time evolution fields determined by @xmath12 @xmath0 nonlinearity @xmath13 chose @xmath14 independent of origin energy shifted to exciton frequency modified form gives scattering effects describe frequency multiplying effects small time - steps model discretized on two dimensional grid differential operators replaced by hopping terms solution time domain initial value problem using crank - nicholson method@xcite leads to semi - implicit time - step formula @xmath15 vector @xmath16 describes cavity photon exciton amplitudes site @xmath17 at time - step @xmath18 precise form matrix operator @xmath19 specified below problem solving eq two dimensions @xmath19 large matrix n length edge grid non - zero terms far from diagonals use operator - splitting treatment alternating direction implicit method@xcite writing @xmath21 @xmath22 @xmath23 contain only @xmath24 @xmath25 hopping terms each time - step divided into two sub - steps @xmath26 solved for @xmath27 @xmath28 @xmath29 independent solve only @xmath30 @xmath31 sets equationsmatrices @xmath22 contain non - zero terms diagonal three sub - diagonals efficient solution by factorisation@xcite @xmath22 coupling site @xmath29 is @xmath32 @xmath19 is @xmath33 matrix @xmath34 hopping terms are @xmath35 @xmath37 @xmath24-direction grid spacing value @xmath4 non - linear term is old value site eq. @xmath23 similar form @xmath38 replaces @xmath37 isotropic grid used factors of @xmath39 multiplying @xmath40 needed avoid over - counting occur in @xmath22 @xmath23 eq.([eq : xup yup solved in two stages intermediate variable @xmath41 calculated using @xmath42 @xmath43 periodic boundary conditions eq.( modified adding hopping terms off - diagonal corners matrix coupling sites strip @xmath44 adding breaks diagonal form matrix direct solution inefficient four extra non - zero entries problem handled little cost using woodbury formula@xcite extra terms @xmath45 written as @xmath46 @xmath47 @xmath48 generate end hopping termsefficent choice @xmath49 @xmath50 first two elements main diagonal @xmath22 @xmath52 modified version @xmath53 terms produce four additional elements subtracted from obtain correct @xmath45 woodbury formula requires additional solutions @xmath54 @xmath55 right side inexpensive factorisation @xmath52 obtained adding periodic conditions little difference numerical costs results microcavity parameters gaas structure rabi splitting @xmath56 mev line - widths exciton cavity mode @xmath57 mev structure on - resonance normal incidence @xmath58 ) pumped at 16@xmath59 gaussian beam fwhm 17.05@xmath60 m cavity mode effective mass @xmath61 exciton mass @xmath62 spatial grid @xmath63 points separation 0.4@xmath60 m time step 12 fs calculation started fields set zero solution propagated until steady state reached after @xmath64 time steps or @xmath65 ps pump signal idler separated storing data time steps transforming frequency domain applying band - pass filters process rolling time window observation dynamical effects time - scales slow mode separationcalculated images show spatial dependence steady - state cavity field intensities signal idler modes at pump powers represent internal cavity field not observed experimentally emitted light spatial filtering small polariton branches ( a ) ( e ) are signal idler intensities for peak power 0.2 above threshold following rows represent pump powers 0.3 0.4 0.5 signal states left idlers right lengths edges squares 51.2@xmath60 m fwhm pump beam 17.05@xmath60 m fig.1 shows calculated signal idler images steady state pump powers above threshold close to threshold small spot centre bright gaussian beam profile threshold reached first where pump strongest higher pump powers emission spot expands spot turns into ring low intensity middle explained using analytic plane - wave model contributions signal idler polariton blue - shifts shifts cause opo switch off above power depends on incidence angle detuning pump beam happens first centre spot leading to dark middle gaussian beam region pump power opo expanding emission ring expected as power increased above thresholdstrong spatial dispersion polariton means simple picture relating opo intensity to local value pump not accurate reasonable guide dispersion effects responsible for ring not symmetrical breaks up at higher pump powers bilateral symmetry mainatined defined by incidence plane cuts figure along @xmath24-axis at higher powers numerics predict steady state reached signal idler undergo periodic oscillation maintaining ring shape maximum intensity moving side to side full understanding of microcavity opo requires two - dimensional treatment spatial structure pump beam polariton fields numerical model spatial temporal evolution system derived used predict emission spot form ring at high pump powers results demonstrate experimental measurements without spatial resolution provide limited picture of opo behaviour n a. gippius s. tikhodeev kulakovskii krizhanovskii tartakovskii cond - mat/0312214 ( 2003 ) fiz . nauk * 173 , 995 ( 2003 ) | 0.35519 | 2,521 | 1,286 | , Saving $0.1 in GPT-4. |
20,651 | mass estimates , based on scaling relationships , are presented of central black holes in luminous quasars at a range of redshifts ( @xmath0 , @xmath1 ) .
the data show that very massive ( @xmath2 ) black holes appear already at @xmath3 , indicating that they form very early or very fast .
# 1_#1 _ # 1_#1 _ = # 1 1.25 in .125 in .25 in | recent advances that allow virial black - hole masses in nearby quiescent and active galaxies to be measured have opened a new era : with mass measurements we can address issues of the physics of black - hole evolution and its connection with galaxy evolution . some obvious first questions to ask are what is the typical mass of black holes in agns at high redshift , and is this different from those of lower-@xmath4 agns ? also , does the typical mass of luminous quasars change beyond the redshift ( @xmath5 ) at which the co - moving space density of quasars drops dramatically ?
a study addressing these issues ( vestergaard 2004 , hereafter paper i ) is summarized below .
as summarized by vestergaard at this meeting , obtaining direct virial black - hole mass ( @xmath6 ) measurements of large samples of distant quasars is very time - consuming and is , hence , impractical .
instead , @xmath6 _ estimates _ can be more easily obtained with scaling relationships and measurements of broad - line width and continuum luminosity in a single - epoch uv or optical spectrum ( wandel , peterson , & malkan 1999 ; vestergaard 2002 ; mclure & jarvis 2002 ) . the mass estimates presented here were obtained with such scaling relationships ( paper i ) .
but is it reasonable to assume that scaling relations , even those based on high - ionization lines , are valid for all agns and quasars ? while the efficacy of the scaling relations is discussed in more detail in paper i , a few key points are worth emphasizing here . from the demonstrated virial relationships ( peterson & wandel 1999 , 2000 ; onken & peterson 2002 ) for the broad emission lines , including the high - ionization lines ,
we know that for a given object the distance to the line - emitting gas and its velocity dispersion each scale between the emission lines : for example , @xmath7(h@xmath8 ) @xmath9(h@xmath8 ) @xmath10(civ ) @xmath9(civ ) , where @xmath11 is the light travel time delay between the variations in the ionizing continuum and responding line emission . therefore , any broad emission - line width ( for which the virial relationship holds ) can be used as a measure of the velocity dispersion , @xmath12(line ) , at the distance of this line - emitting gas , @xmath13(line ) @xmath14(line ) .
in addition , since all four agns that can be tested exhibit this virial relationship , it is fair to assume it is universal in spite of the small number of agns on which this conclusion is based . moreover ,
for the reasonable assumption that the broad - line region is photoionized one can show that if the radius
luminosity relationship did not apply to the luminous and high - redshift quasars , then their spectra would look very different than they actually do : composite quasar spectra grouped by luminosity and redshift are surprisingly similar at least up to redshifts of @xmath154.5 ( dietrich 2002 ) .
also , the application of the radius luminosity relation to the most luminous ( high - redshift ) quasars is only an extrapolation of at most 1.5 dex beyond the luminosity range of 4 dex over which the relationship is defined ( kaspi 2000 ) .
thus , there are no immediate concerns that the radius luminosity relationship
figure 1 shows the distribution of estimated @xmath6 values for several samples of quasars : 87 objects from the bright quasar survey at @xmath16 , a sample of @xmath17 quasars , and recently published samples of @xmath18 quasars , including sdss edr data ( anderson 2001 ; fan 2001 ; constantin 2002 ) . a cosmology with @xmath19 = 75 @xmath20 ,
q@xmath21 = 0.5 , and @xmath22 = 0 is used .
the reader is referred to paper i for further details . for these samples
figure 2a shows the distributions of the bolometric luminosities , @xmath23 , determined by applying bolometric correction factors to measured monochromatic continuum luminosities .
the lower envelopes in both the @xmath23 and @xmath6 values are due to the flux limits in the sample selection or in the original surveys .
figure 2b displays the redshift distribution of the eddington luminosity ratio of the samples .
however , the uncertainty in the @xmath24 values is significant ( @xmath250.7dex ) due to the need to apply an average bolometric correction to obtain @xmath23 and due to the accuracy of the @xmath6 estimates ( 1@xmath26error is@xmath150.5dex relative to reverberation masses ) . the observed range in @xmath24 values is the result of selection flux limits ( lower envelope ) and the combination of a minimum line - width adopted ( @xmath152000km / s ) to identify quasars and the observed upper @xmath23 envelope
nonetheless , these crude @xmath24 estimates show that massive , high-@xmath4 quasars have eddington ratios of order 0.1@xmath271.0 , as commonly expected , as opposed to super - eddington or * distant , luminous quasars have very massive black holes , even at the highest redshifts . in other words ,
black holes with very large masses , @xmath6 @xmath28 , exist at large redshifts beyond the space density drop at @xmath5 and even as early as @xmath29 6 .
* there appears to be a real physical ceiling of @xmath6 @xmath30 and @xmath23 @xmath31 ^ -1 ergss@xmath32 .
this is partly owing to the steep cutoff in the upper end of the luminosity ( and mass ) function .
in addition , this may signify a maximum sustainable mass and luminosity . *
the early appearance of supermassive black holes suggests that they form early or fast .
in fact , they may even form faster than the stars in the host galaxy as suggested by studies of high-@xmath4 host galaxies ( see below ) . the notion that active massive black holes at high-@xmath4 may reside in relatively young galaxies
is based on largely circumstantial evidence that host galaxies of agns at @xmath333 do not typically appear fully formed or have old stellar populations ( see paper i for a more detailed discussion ) : ( a ) the presence of large masses of cold molecular gas in high-@xmath4 hosts ( significant fuel exists for future star formation ) which for some high-@xmath4 sources are clearly separated from the agn nucleus or is distinctly extended ( ohta 1996 ; papadopoulos 2000 ; carilli 2002 , 2003 ) , ( b ) inferred high star - formation rates from the far - ir luminosities in 2001 ) , ( c ) the presence of large - scale young stellar populations confirmed in at least one well - studied @xmath18 radio galaxy with high far - ir luminosity , 4c41.17 ( dey 1997 ) , ( d ) the strikingly different morphology of host galaxies of high-@xmath4 radio - quiet quasars and radio galaxies compared to their lower-@xmath4 cousins : radio galaxies exhibit multiple components embedded in large - scale diffuse emission indicative of non - relaxed systems in the process of forming ( van breugel 1998 ) while radio - quiet quasars appear to have smaller scale - lengths and a fraction of their final stellar mass ( ridgway 2001 ; papovich 2001 ) , and ( e ) the presence of massive , active black holes ( @xmath34 ; marked ` lbg - agn ' in figures 1 and 2 ) in lyman - break galaxies at @xmath35 ; these are typically small , young starforming galaxies ( steidel 2002 ) .
rix ( 2001 ) and omont ( 2001 ) provide independent arguments in favor of a faster build - up of black - hole mass anderson , s. f. , et al .
2001 , , 122 , 503 carilli , c. l. , 2001 , , 555 , 625 carilli , c. l. , et al .
2002 , , 123 , 1838 carilli , c. l. , 2003 , science , 300 , 773 constantin , a. , 2002 , , 565 , 50 dey , a. , 1997 , , 490 , 698 dietrich , m. , et al .
2002 , , 581 , 912 fan , x. , et al .
2001 , , 121 , 31 kaspi , s. , et al .
2000 , , 533 , 631 mclure , r. , & jarvis , m. 2002 , , 337 , 109 ohta , k. , 1996 , nature , 382 , 426 omont , a. , 2001 , , 374 , 371 onken , c. , & peterson , b. m. 2002 , , 572 , 746 papadopoulos , p. p. , 2000
, , 528 , 626 papovich , c. , dickinson , m. , & ferguson , h. c. 2001 , , 559 , 620 peterson , b. m. , & wandel , a. 1999 , , 521 , l95 peterson , b. m. , & wandel , a. 2000 , , 540 , l13 ridgway , s. e. , 2001 , , 550 , 122 rix , h .- w . , 2001 , in gravitational lensing : recent progress and future goals , ed .
t. brainerd & c. kochanek ( san francisco : asp ) , 169 steidel , c. c. , 2002 , , 576 , 653 van breugel , w. j. m. , 1998 , , 502 , 614 vestergaard , m. 2002 , , 571 , 733 vestergaard , m. 2004 , , 600 , in press ( paper i ) wandel , a. , peterson , b. m. , & malkan , m. a. 1999 , , 526 , 579 | advances black - hole masses in galaxies opened new era measurements address physics black - hole evolution connection galaxy evolution questions typical mass of black holes in agns at high redshift different from lower-@xmath4 agns ? typical mass of luminous quasars change beyond redshift ( @xmath5 ) co - moving space density drops ? study vestergaard 2004 paper i summarized direct virial black - hole mass ( @xmath6 ) measurements distant quasars time - consuming impractical @xmath6 estimates obtained with scaling relationships measurements broad - line width continuum luminosity in single - epoch uv optical spectrum ( wandel peterson malkan 1999 vestergaard 2002 mclure jarvis 2002 ) mass estimates obtained with scaling relationships assume scaling relations valid for all agns quasars ? efficacy paper key points emphasizingdemonstrated virial relationships ( peterson wandel 1999 2000 onken peterson 2002 ) for broad emission lines high - ionization lines given object distance to line - emitting gas velocity dispersion scale between emission lines @xmath7 @xmath11 light travel time delay between ionizing continuum line emission broad emission - line width virial relationship holds measure of velocity dispersion @xmath12 at distance line - emitting gas @xmath13 all four agns exhibit virial relationship assume universal small number of agns assumption broad - line region photoionized if radius luminosity relationship to luminous high - redshift quasars spectra different composite quasar spectra luminosity redshift similar to redshifts of @xmath154.5 ( dietrich 2002 ) application radius luminosity relation to most luminous high - redshift quasars extrapolation of 1.5 dex beyond luminosity range 4 dex ( kaspi 2000 )no immediate concerns radius luminosity relationship figure 1 shows distribution estimated @xmath6 values samples quasars 87 objects survey @xmath16 @xmath17 @xmath18 sdss edr data anderson 2001 fan 2001 constantin 2002 cosmology @xmath19 = 75 @xmath20@xmath21 = 0.5 @xmath22 = 0 used paper i details figure 2a shows distributions bolometric luminosities @xmath23 correction factors monochromatic luminosities lower envelopes @xmath23 @xmath6 values due flux limits selection surveys figure 2b displays redshift distribution eddington luminosity ratio uncertainty @xmath24 values significant @xmath250.7dex ) due average bolometric correction accuracy @xmath6 estimates 1@xmath26error@xmath150.5dex reverberation range @xmath24 values selection flux limits minimum line - @xmath152000km / s upper @xmath23 envelope @xmath24 estimates show massive high-@xmath4 quasars have eddington ratios 0.1@xmath271.0 super - eddington distant luminous quasars massive black holes highest redshiftsblack holes large masses @xmath6 @xmath28 exist at redshifts beyond space density drop at @xmath5 early @xmath29 6 physical ceiling of @xmath6 @xmath23 @xmath31 -1 ergss@xmath32 to steep cutoff upper end luminosity mass function maximum sustainable mass luminosity early appearance of supermassive black holes form early fast may form faster than stars host galaxy studies high-@xmath4 host galaxies active massive black holes at high-@xmath4 in young galaxies based on circumstantial evidence host galaxies at @xmath333 fully formed old stellar populations presence of large masses cold molecular gas in high-@xmath4 hosts fuel for future star formation separated from agn nucleus extended high star - formation rates from far - ir luminosities in 2001 presence of large - scale young stellar populations in @xmath18 radio galaxy with high far - ir luminositydey 1997 d different morphology host galaxies high-@xmath4 radio - quiet quasars radio galaxies lower-@xmath4 cousins radio galaxies multiple components large - scale diffuse emission non - relaxed systems forming breugel 1998 radio - quiet quasars smaller scale - lengths fraction final stellar mass ridgway 2001 papovich 2001 massive active black holes @xmath34 lbg - agn figures 1 2 lyman - break galaxies @xmath35 small young starforming galaxies steidel 2002 rix 2001 omont ( 2001 ) arguments faster build - up black - hole mass anderson 2001 122 , 503 carilli 555 , 625 2002 123 , 1838 science 300 , 773 constantin 2002 565 , 50 dey 1997 490 , 698 dietrich 2002 581 , 912 fan 2001 121 , 31 kaspi 2000 533 , 631 mclure jarvis 2002 337 , 109 ohta nature 382 , 426 omont 2001 374 , 371 onken peterson2002 572 , 746 papadopoulos 2000 528 , 626 papovich dickinson ferguson 2001 559 , 620 peterson wandel 1999 521 , l95 peterson 2000 540 , l13 ridgway 2001 550 , 122 rix gravitational lensing recent progress future goals brainerd. kochanek san francisco asp 169 steidel c.. 2002 576 , 653 van breugel 1998 502 , 614 vestergaard 2002 571 , 733 vestergaard 2004 600 press wandel peterson b m. malkan m. 1999 526 , 579 | 0.347622 | 2,343 | 1,246 | , Saving $0.1 in GPT-4. |
20,655 | the pax project at gsi darmstadt plans to polarize an antiproton beam by repeated interaction with a hydrogen target in a storage ring . many of the beam particles are required to remain within the ring after interaction with the target ,
so small scattering angles are important .
hence we concentrate on low momentum transfer ( small @xmath0 ) , a region where electromagnetic effects dominate the hadronic effects .
a colliding beam of polarized electrons with energy sufficient to provide scattering of antiprotons beyond ring acceptance may polarize an antiproton beam by spin filtering .
expressions for spin observables are provided and are used to estimate the rate of buildup of polarization of an antiproton beam .
address = school of mathematics , trinity college dublin , ireland address = school of mathematics , trinity college dublin , ireland | the buildup of the polarization of an antiproton beam by spin filtering off a polarized hydrogen target has been described by differential equations @xcite involving the spin observables for antiproton - electron and antiproton - proton elastic scattering .
relativistic expressions for spin observables due to single photon exchange in elastic spin 1/2 - spin 1/2 particle collisions have been presented recently @xcite .
there has been much recent theoretical debate as to whether electrons in a hydrogen gas target are effective in transferring polarization to the antiproton beam @xcite .
it is claimed that electrons should scatter antiprotons out of the beam to contribute to spin filtering .
the maximum antiproton - electron scattering angle for various antiproton beam energies is lower than the ring acceptance angles under consideration by pax @xcite and we suggest that a beam of polarized electrons with energy sufficient to provide scattering of antiprotons beyond ring acceptance may polarize an antiproton beam by spin filtering .
for electromagnetic interactions to first order the double spin asymmetries equal the polarization transfer observables ( @xmath1 ) and all the single and triple spin asymmetries are zero ( @xmath2 ) where @xmath3 .
spin filtering requires evaluation of the angular integration of the product of the observables @xmath4 and @xmath5 with @xmath6 .
azimuthal averaging indicates that the observables with single @xmath7 ( i.e. @xmath8 , @xmath9 , @xmath10 and @xmath11 ) do not contribute to spin filtering .
the quantities @xmath12 , @xmath13 , @xmath14 , @xmath15 and @xmath6 which we now present , play the important role .
the spin averaged differential cross - sections have been presented in @xcite . in the following @xmath16
is the electron mass , @xmath17 is the ( anti)proton mass , @xmath18 is the magnetic moment of the proton , @xmath19 is the fine structure constant , @xmath20 and @xmath0 are mandelstam variables and @xmath21^{1/2}\!.$ ]
to leading order in small @xmath0 the relevant spin observables for single photon exchange antiproton - proton scattering are as follows @xcite .
the @xmath22 sign refers to the first term in the expansion in @xmath0 .
@xmath23 \frac{\left(1 - d_\mathrm{xx}\right ) + \left(1 - d_\mathrm{yy}\right)}{2}\ \frac{d\,\sigma}{d\,\omega } & \approx & \displaystyle { \frac{-\ \alpha^2 \left(k^2 + m^{\,2}\right)}{k^2\ m^{\,2}\ s\ t}\left[m^{\,2 } - 2\,k^2 \left(\mu_\mathrm{p } -1\right)\right]^{\,2 } } \nonumber \\[1ex ] k_\mathrm{zz}\ \frac{d\,\sigma}{d\,\omega } & \approx & \displaystyle{\frac{-\,2\ \alpha^2 \
\mu_\mathrm{p}^2}{s\ t}\ \left(2\,k^2 + m^{\,2}\right)}\\[1ex ] \left(1 - d_\mathrm{zz}\right)\ \frac{d\,\sigma}{d\,\omega } & \approx & \displaystyle { \frac{-\ , 2\,\alpha^2 \left(k^2 + m^{\,2}\right)}{k^2\ m^{\,2}\ s\ t}\left[m^{\,2 } - 2\,k^2 \left(\mu_\mathrm{p } -1\right)\right]^{\,2}}\,.\nonumber\end{aligned}\ ] ] the leading @xmath0 terms in the relevant observables for antiproton - electron scattering are @xcite : @xmath24 \frac{\left(1 - d_\mathrm{xx}\right ) + \left(1 - d_\mathrm{yy}\right)}{2}\ \frac{d\,\sigma}{d\,\omega } & \approx & \displaystyle{\frac{-m^2\,\alpha^2\,\left(\,s - m^2 + m^{\,2}\,\right)^2}{4\ k^2\ s^2\ t } } \nonumber \\[1ex ] k_\mathrm{zz}\ \frac{d\,\sigma}{d\,\omega } & \approx & \displaystyle{\frac{-\ , \alpha^2\ \mu_\mathrm{p}}{s\ t}\left(\,s - m^2 - m^{\,2}\,\right)}\\[1ex ] \left(1 - d_\mathrm{zz}\right)\ \frac{d\,\sigma}{d\,\omega } & \approx & \displaystyle{\frac{-m^2\,\alpha^2\,\left(\,s + m^2 - m^{\,2}\,\right)^2}{2\ k^2\ s^2\ t}}\,.\nonumber\end{aligned}\ ] ]
the maximum scattering angle for @xmath25 antiprotons scattering off electrons in an atomic target is @xmath26 as shown in figure 1 ( a ) .
this is below the acceptance angles under consideration at pax @xcite so that all scattering off atomic electrons will be within the ring ; therefore no spin filtering is possible and electrons will not contribute to the polarization buildup of the antiproton beam .
an opposing electron beam of sufficient energy would increase the scattering angles of the antiprotons beyond acceptance as seen in figure 1 ( b ) , hence allowing spin filtering . for @xmath25 antiprotons scattering off electrons in an atomic target .
( b ) with a colliding electron beam the maximum scattering angle increases as the electron beam momentum increases in a direction opposite to that of the antiproton beam .
, title="fig : " ] for @xmath25 antiprotons scattering off electrons in an atomic target .
( b ) with a colliding electron beam the maximum scattering angle increases as the electron beam momentum increases in a direction opposite to that of the antiproton beam .
, title="fig : " ]
when circulating at frequency @xmath27 through a polarized target of areal density @xmath28 and polarization @xmath29 oriented normal to the ring plane , ( or longitudinally with rotators ) @xmath30 j \end { array } \right ] \ , = \ , - \ , n \ , \nu \left [ \begin{array}{ccc } % i_\mathrm{\ , out } & & p_\mathrm{e } \ , a_\mathrm{\ , out } \\[2ex ] p_\mathrm{e } \
, a_\mathrm{\ , all } - p_\mathrm{e } \
, k_\mathrm{in } & & i_\mathrm{\ , all } - d_\mathrm{\ , in } % \end { array } \right ] \
, \left [ \begin{array}{c } n \\[2ex ] j \end { array } \right]\end{aligned}\ ] ] describes the rate of change of the number of beam particles @xmath31 and their total spin @xmath32 @xcite .
these coupled differential equations involve angular integration of the spin observables presented earlier , as seen in the following table .
the minimum angle @xmath33 relates to the average transverse electron separation , and @xmath34 is the acceptance angle .
the time ( @xmath35 ) dependence of the polarization of the beam is given by solving the system @xmath36 where the discriminant of the quadratic equation for the eigenvalues is @xmath37 and where @xmath38 is a loss of polarization quantity .
the approximate rate of change of polarization for sufficiently short times , and the limit of the polarization for large times are respectively : @xmath39 [ cols="^,^ " , ]
we have shown that , in principle , a sufficiently high density polarized electron beam could be used to polarize an antiproton beam by spin filtering in a storage ring .
this can be done by a pure electromagnetic process .
the polarization buildup rate is proportional to the density of the electrons in the beam .
we are now investigating if the rate of polarization buildup and maximum polarization achievable with current technologies are sufficient to provide a useful figure of merit when attempting to polarize antiprotons .
dob would like to thank the embark initiative
( ircset ) , for a postgraduate research scholarship .
nhb is grateful to enterprise ireland for the award of a grant under the international collaboration program to facilitate a visit to infn at the university of torino , where discussions with m. anselmino and m. boglione are acknowledged .
we are also grateful to e. leader , n. n. nikolaev and f. rathmann for helpful comments .
n. n. nikolaev and f. f. pavlov , arxiv : hep - ph/0601184 .
a. i. milstein and v. m. strakhovenko , _ phys .
e _ * 72 * 066503 ( 2005 ) .
d. s. obrien and n. h. buttimore , arxiv : hep - ph/0609233 . f. rathmann _ et al .
_ , _ phys .
lett . _ * 94 * 014801 ( 2005 )
. v. barone _ et al . _
[ pax collaboration ] , arxiv : hep - ex/0505054 .
d. s. obrien and n. h. buttimore , _ czech .
j. phys . _
* 56 * c265 ( 2006 ) [ arxiv : hep - ph/0605099 ] . | polarization antiproton beam by spin filtering off polarized hydrogen target described by differential equations spin observables antiproton - electron proton elastic scattering relativistic expressions for spin observables single photon exchange in particle collisions presented recently debate electrons in hydrogen gas target transferring polarization to antiproton beam electrons scatter antiprotons spin filtering maximum antiproton - electron scattering angle for beam energies lower than ring acceptance angles polarized electrons with energy scattering beyond ring acceptance may polarize antiproton beam by spin filtering electromagnetic interactions double spin asymmetries equal polarization transfer observables single triple spin asymmetries zero spin filtering requires evaluation angular integration observables @xmath4 @xmath5 with @xmath6 azimuthal averaging indicates observables with single @xmath7 contribute to spin filtering quantities @xmath12 play role spin averaged differential cross - sections presented in @xcite@xmath16 electron mass anti)proton mass magnetic moment proton fine structure constant @xmath20 @xmath0 mandelstam variables @xmath21^{1/2}\.$ @xmath0 relevant spin observables single photon exchange antiproton - proton scattering @xmath22 sign first term expansion @xmath0\frac{\left(1 -{xx{yy{2}{d\sigma\omega\frac{-\alpha^2 \left(k^2 + m^{\,2}{k^2 m^{\,2}[m^{,2 - 2\,k^2(\mathrm{p -1^{\,2\mathrm{zz\frac{d\sigma\omega\frac{-\,2\\alpha^2\mathrm{p}^2\left(2\,k^2 + m^{\,2}\right)[1ex\left(1 - d\mathrm{zz}\right){d\sigma\omega\frac{- 2\alpha^2 \left(k^2 + m^{\,2}{k^2\ m^{\,2}\left[m^{\,2 } - 2\,k^2 (\mathrm{p -1^{\,2}}{aligned leading @xmath0 terms antiproton - electron scattering @xcite @xmath24\left(1 -\mathrm{xx} +{yy}\right}{2}\frac{d\sigma\omega\frac{-m^2\\alpha^2 - m^2 + m^{,2^2}{4\ k^2\ s^2 t\mathrm{zz\frac{d\sigma}{d\omega\frac{-\alpha^2\mathrm\left( - m^2 - m^{\,2}[1ex - d\mathrm{zz}\right{d\sigma}{d\omega{-m^2\\alpha^2 + m^2 - m^{,2})^2}{2\ k^2\ s^2\ t}} maximum scattering angle for @xmath25 antiprotons atomic target is @xmath26 figure 1 ( a below acceptance angles @xcite scattering electrons within ring no spin filtering electrons not contribute to polarization buildup antiproton beamopposing electron beam energy scattering angles antiprotons beyond acceptance in figure 1 ( b ) allowing spin filtering @xmath25 antiprotons scattering off electrons in atomic target colliding electron beam maximum scattering angle increases as electron beam momentum increases direction opposite antiproton beam title="fig @xmath25 antiprotons scattering off electrons in atomic target colliding electron beam maximum scattering angle increases as electron beam momentum increases direction opposite antiproton beam circulating at frequency @xmath27 through polarized target areal density @xmath28 polarization @xmath29 oriented normal to ring plane or longitudinally with rotators ) @xmath30 j \end { array } \right ] \left\begin{array}{ccc }_\mathrm{ out\mathrm\mathrm{in\end { array }\right\left{array[2ex array{aligned describes rate of change of number of beam particles @xmath31 total spin @xmath32 @xcite coupled differential equations involve angular integration spin observables table minimum angle @xmath33 average transverse electron separation @xmath34 acceptance angle time ( @xmath35 dependence of polarization beam solving system @xmath36 discriminant equation is @xmath37 @xmath38 is loss of polarization quantity approximate rate change polarization for short times limit large times are : @xmath39 high density polarized electron beam polarize antiproton beam by spin filtering in storage ring electromagnetic process polarization buildup rate proportional to density electrons beam investigating if rate polarization buildup maximum polarization with current technologies antiprotons embark initiative for postgraduate research scholarship grateful to enterprise ireland for grant visit to infn university of torino discussions m. anselmino boglione grateful to e. leader n. n. nikolaev f. rathmann for comments pavlov arxiv hep - ph/0601184 a. i. milstein v m. strakhovenko 72 * 066503 ( 2005 ) d. s. obrien n. hbuttimore arxiv hep - ph/0609233 rathmann phys . lett 94 * 014801 2005 ) barone pax collaboration arxiv hep - ex/0505054 obrien buttimore . 56 * c265 ( 2006 ) arxiv hep - ph/0605099 | 0.415432 | 2,190 | 1,120 | , Saving $0.1 in GPT-4. |
20,667 | although sr@xmath0ru@xmath1o@xmath2 has not been reported to exhibit superconductivity so far , ac susceptibility measurements revealed multiple superconducting transitions occurring in the sr@xmath0ru@xmath1o@xmath2 region cut from sr@xmath0ru@xmath1o@xmath2-sr@xmath1ruo@xmath3 eutectic crystals .
based on various experimental results , some of us proposed the scenario in which sr@xmath1ruo@xmath3 thin slabs with a few layers of the ruo@xmath1 plane are embedded in the sr@xmath0ru@xmath1o@xmath2 region as stacking faults and multiple superconducting transitions arise from the distribution of the slab thickness . to examine this scenario , we measured the resistivity along the @xmath4 plane ( @xmath5 ) using a sr@xmath0ru@xmath1o@xmath2-region sample cut from the eutectic crystal , as well as along the @xmath6 axis ( @xmath7 ) using the same crystal . as a result , we detected resistance drops associated with superconductivity only in @xmath5 , but not in @xmath7 .
these results support the sr@xmath1ruo@xmath3 thin - slab scenario .
in addition , we measured the resistivity of a single crystal of pure sr@xmath0ru@xmath1o@xmath2 with very high quality and found that pure sr@xmath0ru@xmath1o@xmath2 does not exhibit superconductivity down to 15 mk . | the ruddlesden - popper series of layered perovskites sr@xmath8ru@xmath9o@xmath10 is a fascinating research subject because the @xmath11=1 member sr@xmath1ruo@xmath3 is now believed to be a spin - triplet superconductor @xcite .
the @xmath11=2 member of this series sr@xmath0ru@xmath1o@xmath2 is known to exhibit an enhanced pauli paramagnetism with a metamagnetic transition @xcite .
although crystals of sr@xmath0ru@xmath1o@xmath2 are highly refined , the superconductivity in sr@xmath0ru@xmath1o@xmath2 has not been discovered so far .
recently , sr@xmath0ru@xmath1o@xmath2-sr@xmath1ruo@xmath3 eutectic crystals were successfully grown @xcite and surprisingly , multiple superconducting transitions were observed in the ac susceptibility measurements using a sr@xmath0ru@xmath1o@xmath2-region sample cut from eutectic crystals @xcite , as presented in fig .
[ ec](a ) .
however , it has been revealed that this superconductivity is not a bulk property of sr@xmath0ru@xmath1o@xmath2 because these superconducting transitions are easily suppressed by small ac magnetic fields and no anomaly was observed in the specific heat @xcite .
the most plausible scenario of this superconductivity is that sr@xmath1ruo@xmath3 thin slabs with a few layers of ruo@xmath1 planes are embedded in the sr@xmath0ru@xmath1o@xmath2 region and the multiple superconducting transitions arise from the distribution of the slab thickness . in order to obtain additional evidence for the origin of the superconductivity observed in the sr@xmath0ru@xmath1o@xmath2 region cut from eutectic crystals , we measured the resistivity along the @xmath4 plane ( @xmath5 ) of a sr@xmath0ru@xmath1o@xmath2-region sample cut from the eutectic crystal , which we designate below as a eutectic sr@xmath0ru@xmath1o@xmath2 sample , as well as along the @xmath6 axis ( @xmath7 ) of the same sample .
in addition , we measured @xmath7 of a pure ( i.e. non - eutectic ) sr@xmath0ru@xmath1o@xmath2 sample with very high quality in order to clarify whether or not pure sr@xmath0ru@xmath1o@xmath2 exhibits superconductivity at low temperatures .
resistivity was measured using a conventional four - probe method with an ac current .
single crystals of the sr@xmath0ru@xmath1o@xmath2-sr@xmath1ruo@xmath3 eutectic system and those of single - phase sr@xmath0ru@xmath1o@xmath2 were grown by a floating - zone method .
the size of the eutectic sr@xmath0ru@xmath1o@xmath2 sample was approximately 1.5 @xmath12 0.7 mm@xmath13 in the @xmath4 plane and 0.3 mm along the @xmath6 axis .
the results of ac susceptibility and specific heat measurements using this eutectic sr@xmath0ru@xmath1o@xmath2 sample were reported in ref .
the dimensions of the pure sr@xmath0ru@xmath1o@xmath2 sample are 0.44 @xmath12 0.28 mm@xmath13 in the @xmath4 plane and 1.14 mm along the @xmath6 axis .
the @xmath5 measurements on the eutectic sr@xmath0ru@xmath1o@xmath2 sample were performed down to 0.3 k with a @xmath14he cryostat ( oxford instruments , model heliox vl ) .
after the @xmath5 measurements , we removed the electrical leads and attached another set of wires again on the same sample and performed @xmath15 measurements down to 0.1 k with an adiabatic demagnetization refrigerator ( cambridge magnetic refrigeration , mfridge50 ) .
the @xmath15 measurements using the pure sr@xmath0ru@xmath1o@xmath2 sample were performed with a @xmath14he-@xmath16he dilution refrigerator ( cryoconcept , model dr - jt - s-100 - 10 ) down to 15 mk . in this study
, we used a cylinder of permalloy ( hamamatsu photonics k.k .
, e989 - 28 ) in order to reduce remanent fields such as the earth field .
figure [ ec](b ) shows temperature dependence of @xmath5 and @xmath15 of the eutectic sr@xmath0ru@xmath1o@xmath2 sample .
the values of @xmath5 and @xmath15 are approximately 1 @xmath17 cm and 300 @xmath17 cm at 1.5 k , respectively .
the in - plane resistivity is nearly the same as those of pure sr@xmath0ru@xmath1o@xmath2 crystals with very high quality @xcite .
this low value of the in - plane resistivity indicates that sr@xmath0ru@xmath1o@xmath2 in the eutectic system crystallized with high quality . also , macroscopic sr@xmath1ruo@xmath3 domains in the eutectic system are high quality because its @xmath18 is nearly 1.5 k @xcite , which is one of the best @xmath18 of sr@xmath1ruo@xmath3 reported .
it is interesting that both sr@xmath1ruo@xmath3 and sr@xmath0ru@xmath1o@xmath2 spontaneously crystallize with high quality in this eutectic system .
( a ) temperature dependence of the real part of the ac susceptibility for a eutectic sr@xmath0ru@xmath1o@xmath2 sample with @xmath19=0.58 @xmath20 t and @xmath21=3011 hz(sample 1 in ref .
( b ) temperature dependences of the resistivity along the @xmath6 axis and that along the @xmath4 plane for the eutectic sr@xmath0ru@xmath1o@xmath2 sample measured at @xmath21=89.1 hz ( @xmath22=0.5 ma - rms for @xmath5 and @xmath22=0.1 ma - rms for @xmath7 ) .
( c ) a schematic image of two monolayers of ruo@xmath1 planes ( sr@xmath1ruo@xmath3 ) intercalated in bilayers ( sr@xmath0ru@xmath1o@xmath2 ) .
oxygens are located at the corner of the octahedra.,width=556 ] in @xmath5 measurements , two clear resistance drops were observed at 1.05 and 1.32 k. these transition temperatures well coincide with those observed in the ac susceptibility [ fig .
[ ec](a ) ] .
however , in @xmath15 measurements , no obvious transition was observed .
these results are consistent with the sr@xmath1ruo@xmath3 thin - slab scenario because they imply that superconducting inclusions embedded in the eutectic sr@xmath0ru@xmath1o@xmath2 are too thin along the @xmath6 axis to shortcircuit the current path along the interlayer direction .
this behavior is in sharp contrast with @xmath15 for the sr@xmath1ruo@xmath3-ru eutectic system , in which the emergence of the 3-k " superconductivity in the interface of ru lamellae results in a large drop in @xmath15 @xcite .
although we observed two clear transitions in the @xmath5 measurement using the eutectic sr@xmath0ru@xmath1o@xmath2 sample , @xmath5 does not become zero down to low temperatures .
this non - zero resistivity implies that sr@xmath1ruo@xmath3 inclusions in this eutectic sr@xmath0ru@xmath1o@xmath2 sample do not completely form a path between the voltage contacts .
the sr@xmath1ruo@xmath3 inclusions are probably well separated in this sample . in some cases , eutectic sr@xmath0ru@xmath1o@xmath2 samples exhibit zero resistivity ( e. g. , ref .
@xcite ) , probably because sr@xmath1ruo@xmath3 inclusions in such samples link a path between the voltage contacts .
now , on the basis of various experiments , we believe that the origin of the superconductivity observed in the eutectic sr@xmath0ru@xmath1o@xmath2 sample is the presence of several monolayers of ruo@xmath1 planes intercalated in sr@xmath0ru@xmath1o@xmath2 as stacking faults . for example , two monolayers of ruo@xmath1 planes intercalated in sr@xmath0ru@xmath1o@xmath2 is schematically drawn in fig .
[ ec](c ) .
in fact , such stacked monolayers of ruo@xmath1 planes have been observed with a transmission electron microscope @xcite .
we are also interested in the possibility of superconductivity in pure sr@xmath0ru@xmath1o@xmath2 . in order to examine the superconductivity in pure sr@xmath0ru@xmath1o@xmath2
, we consider it important ( i ) to use single crystals of single - phase sr@xmath0ru@xmath1o@xmath2 with very high quality , ( ii ) to cool down the sample to sufficiently low temperature , and ( iii ) to perform measurements in zero field by excluding the geomagnetic field .
therefore , we measured @xmath15 down to 15 mk using a single crystal of pure sr@xmath0ru@xmath1o@xmath2 with the in - plane residual resistivity of 0.4 @xmath17 cm , which is one of the highest - quality sr@xmath0ru@xmath1o@xmath2 grown so far . in addition , by placing the sample in a cylinder of permalloy , we reduced the residual field to be lower than 0.1 @xmath20 t .
figure [ sp ] shows temperature dependence of @xmath15 of the pure sr@xmath0ru@xmath1o@xmath2 sample .
the @xmath15 monotonically decreases with decreasing temperature and no anomaly indicating a superconducting transition was observed down to 15 mk . from this measurement , we conclude that pure sr@xmath0ru@xmath1o@xmath2 does not become superconducting down to 15 mk .
temperature dependence of the resistivity along the @xmath6 axis of the pure sr@xmath0ru@xmath1o@xmath2 sample with very high quality measured at @xmath22=0.01 ma - rms with @xmath21=7 hz without the geomagnetic field .
the inset shows the low - temperature region below 0.3 k.,width=336 ]
we measured the resistivity along the @xmath6 axis as well as along the @xmath4 plane of a eutectic sr@xmath0ru@xmath1o@xmath2 sample .
the resistance drops due to the multiple superconducting transitions were observed only for @xmath5 , but not for @xmath15 .
this result indicates that superconductors with thin thickness along the @xmath6 axis are embedded in the eutectic sr@xmath0ru@xmath1o@xmath2 .
now , we are convinced that the origin of the superconductivity observed in the eutectic sr@xmath0ru@xmath1o@xmath2 sample is the presence of sr@xmath1ruo@xmath3 inclusions embedded in sr@xmath0ru@xmath1o@xmath2 as stacking faults . in order to search for superconductivity in pure sr@xmath0ru@xmath1o@xmath2
, we also measured the resistivity along the @xmath6 axis using a single crystal of best - quality pure sr@xmath0ru@xmath1o@xmath2 .
however , no resistance anomaly associated with the superconducting transition was observed down to 15 mk .
therefore , we conclude that pure sr@xmath0ru@xmath1o@xmath2 does not become superconducting down to 15 mk .
this work has been supported by the grant - in - aid for the global coe program `` the next generation of physics , spun from universality and emergence '' from the ministry of education , culture , sports , science , and technology ( mext ) of japan .
it is also supported by grants - in - aid for scientific research from mext and from the japan society for the promotion of science ( jsps ) .
one of the authors ( s. k. ) is financially supported as a jsps research fellow .
1 url # 1#1urlprefix[2][]#2 maeno y , hashimoto h , yoshida k , nishizaki s , fujita t , bednorz j g and lichtenberg f 1994 _ nature ( london ) _ * 372 * 532 | ruddlesden - popper series layered perovskites sr@xmath8ru research @xmath11=1 member sr@xmath1ruo@xmath3 spin - triplet superconductor @xcite @xmath11=2 member sr@xmath0ru@xmath1o@xmath2 enhanced pauli paramagnetism metamagnetic transition @xcite crystals@xmath0ru@xmath1o refined superconductivity@xmath0ru discovered sr@xmath0ru eutectic crystals grown @xcite multiple superconducting transitions observed ac susceptibility measurements@xmath0ru sample fig . superconductivity not bulk property sr@xmath0ru@xmath1o@xmath2 transitions suppressed by small magnetic fields no anomaly heat @xcite plausible scenario sr@xmath1ruo@xmath3 thin slabs layers ruo@xmath1 planes embedded sr@xmath0ru@xmath1o@xmath2 region multiple superconducting transitions arise from distribution slab thicknessevidence superconductivity sr@xmath0ru@xmath1o@xmath2 region from eutectic crystals measured resistivity @xmath4 plane sr@xmath0ru@xmath1o@xmath2-region sample eutectic crystal @xmath6 axis measured @xmath7 pure non - eutectic sr@xmath0ru@xmath1o@xmath2 sample high quality superconductivity at low temperatures resistivity measured conventional four - probe method ac current crystals@xmath0ru eutectic system single - phase sr grown by floating - zone method size eutectic sr@xmath0ru sample approximately 1.5 @xmath12 0.7 mm@xmath13 @xmath4 plane 0.3 mm @xmath6 axis results ac susceptibility heat measurements eutectic@xmath0ru sample reported in ref dimensions pure sr@xmath0ru@xmath1o@xmath2 sample 0.44 @xmath12 0.28 mm@xmath13 @xmath4 plane 1.14 mm @xmath6 axis@xmath5 measurements eutectic sr@xmath0ru sample 0.3 k @xmath14he cryostat oxford instruments heliox vl after removed electrical leads attached wires performed @xmath15 measurements 0.1 k adiabatic demagnetization refrigerator cambridge magnetic refrigeration @xmath15 measurements pure sr@xmath0ru@xmath1o@xmath2 sample @xmath14he-@xmath16he dilution refrigerator cryoconcept - - - 15 mk used cylinder permalloy hamamatsu photonics reduce remanent fields figure shows temperature dependence @xmath5 @xmath15 eutectic@xmath0ru sample values @xmath5 approximately 1 @xmath17 cm and 300 @xmath17 cm at 1.5 k in - plane resistivity same as pure sr@xmath0ru@xmath2 crystals high quality @xcite low resistivity indicates@xmath0ru crystallized high quality macroscopic sr@xmath1ruo@xmath3 domains system high quality @xmath18 nearly 1.5 k @xcite bestsr@xmath1ruo@xmath3@xmath2 crystallize eutectic system temperature dependence ac susceptibility eutectic@xmath0ru@xmath1o@xmath2 sample @xmath19=0.58 @xmath20 @xmath21=3011 hz temperature dependences resistivity @xmath6 axis @xmath4 plane eutectic@xmath0ru sample @xmath21=89.1 hz @xmath22=0.5 ma - rms @xmath5=0.1 ma - rms @xmath7 schematic image two monolayers@xmath1 planes intercalated bilayers@xmath0ru@xmath1o@xmath2 oxygens corner octahedra @xmath5 measurements two resistance drops at 1.05 1.32 k transition temperatures coincide ac susceptibility @xmath15 measurements no transition results consistent sr@xmath1ruo@xmath3 thin - slab scenario superconducting inclusions eutectic sr@xmath0ru@xmath1o@xmath2 too thin @xmath6 axis shortcircuit current pathbehavior with @xmath15 sr@xmath1ruo@xmath3-ru eutectic system 3-k " superconductivity ru lamellae results large drop in @xmath15 @xcite observed two transitions in @xmath5 measurement sample @xmath5 become zero to low temperatures non - zero resistivity implies sr@xmath1ruo@xmath3 inclusions form path between voltage contacts sr@xmath1ruo@xmath3 inclusions separated some eutectic samples exhibit zero resistivity sr@xmath1ruo@xmath3 inclusions link path between voltage contacts origin of superconductivity in sr@xmath0ru@xmath1o@xmath2 sample several monolayers of ruo@xmath1 planes in@xmath0ru@xmath2 as stacking faults two monolayers planes in@xmath2 drawn in fig . [ ec](c ) stacked monolayers observed with transmission electron microscope @xcite interested in superconductivity in pure sr@xmath0ru@xmath1o@xmath2examine superconductivity in pure sr@xmath0ru@xmath1o important use single crystals high quality cool sample to low temperature perform measurements in zero field excluding geomagnetic field measured @xmath15 down to 15 mk using single crystal in - plane residual resistivity 0.4 @xmath17 cm highest - quality sample in cylinder of permalloy reduced residual field lower than 0.1 @xmath20 t figure shows temperature dependence of @xmath15 decreases with temperature no anomaly superconducting transition down to 15 mk pure sr@xmath0ru become superconducting down to 15 mk temperature dependence of resistivity along @xmath6 axis measured at @xmath22=0.01 ma - rms @xmath21=7 hz without geomagnetic field low - temperature region below 0.3 k measured resistivity along @xmath6 axis @xmath4 plane of sample resistance drops due to multiple superconducting transitions observed only for @xmath5 not for @xmath15result indicates superconductors thin thickness @xmath6 axis embedded in eutectic sr@xmath0ru@xmath1o@xmath2 origin superconductivity is sr@xmath1ruo@xmath3 inclusions stacking faults superconductivity measured resistivity @xmath6 axis single crystal best - quality pure no resistance anomaly superconducting transition observed down to 15 mk conclude pure sr@xmath0ru@xmath1o@xmath2 become superconducting down to 15 mk work supported by grant - in - aid global coe program next generation physics ministry education japan supported by grants - in - aid scientific research from mext japan society promotion science one s. k. ) financially supported jsps research fellow maeno y hashimoto h yoshida k nishizaki s fujita t bednorz j g lichtenberg f 1994 _ nature ( london ) * 372 * 532 | 0.434194 | 3,092 | 1,602 | , Saving $0.1 in GPT-4. |
20,674 | observational evidence for strong magnetic fields throughout the envelopes of evolved stars is increasing .
many of the instruments coming on line in the near - future will be able to make further contributions to this field . specifically , maser polarization observations and dust / line polarization in the sub - mm regime has the potential to finally provide a definite picture of the magnetic field strength and configuration from the asymptotic giant branch ( agb ) all the way to the planetary nebula phase . while current observations are limited in sample size ,
strong magnetic fields appear ubiquitous at all stages of ( post-)agb evolution .
recent observations also strongly support a field structure that is maintained from close to the star to several thousands of au distance .
while its origin is still unclear , the magnetic field is thus a strong candidate for shaping the stellar outflows on the path to the planetary nebula phase and might even play a role in determining the stellar mass - loss . + * keywords.*planetary nebulae stars : agb and post - agb magnetic fields | strongly asymmetric planetary nebulae ( pne ) have been shown to be common .
the research into their shaping processes has become a fundamental part of our attempts to further the understanding of the return of processed material into the ism by low- and intermediate - mass stars at the end of their evolution . whereas the standard interacting winds scenario
@xcite can explain a number of the pn properties , an important discovery has been that the collimated outflows of the pre - pne ( p - pne ) , where such outflows are common , have a momentum that exceeds that which can be supplied by radiation pressure alone @xcite .
the source of this momentum excess has been heavily debated during the past several years , with the most commonly invoked cause being magnetic fields , binary or disk interaction or a combination of these ( e.g. * ? ? ?
due to a number of similarities with the jets and outflows produces by young stellar objects , the study of p - pne outflows provides further research opportunities into a potentially universal mechanism of jet launching . here
i will review the observational evidence for strong magnetic fields in pne as well as around their agb and post - agb progenitors .
i will give an overview of the methods that can be used to study magnetic fields , especially in light of the plethora of new instruments that will be available shortly .
finally , i will discuss a number of questions related this topic that we can expect to be answered with the new instruments in the next few years .
with the exception of observations where the magnetic field strength is estimated assuming forms of energy equilibrium , such as synchrotron observations , the magnetic field strength and structure is typically determined from polarization observations .
circular polarization , generated through zeeman splitting , can be used to measure the magnetic field strength .
it measures the total field strength when the splitting is large and the line - of - sight component of the field when the splitting is small .
the predominant source of magnetic field strength information during the late stages of stellar evolution comes from maser circular polarization observations , and particularly the common sio , h@xmath0o and oh masers .
these can show circular polarization fractions ranging from @xmath1% ( h@xmath0o ) up to @xmath2% ( oh ) and are , because of their compactness and strength , excellent sources to be observed with high angular resolution .
unfortunately , the analysis of maser polarization is not straightforward ( for a review , see * ? ? ?
* ) , and it has taken a long time before maser observations were acknowledged to provide accurate magnetic field measurements .
more recently , the first attempts have been made to detect the zeeman splitting of non - maser molecular lines in circumstellar envelopes , such as cn @xcite .
as many of these occur at shorter wavelength in the ( sub-)mm regime , the advent of the atacama large ( sub-)millimeter array will further enhance these types of studies .
linear polarization , probing the structure of the plane - of - the - sky component of the magnetic field , can be observed both in the dust ( through aligned grains ) and molecular lines ( through radiation anisotropy - the goldreich - kylafis effect ) .
typical percentages of linear polarization range from up to a few percent ( e.g. dust , co , h@xmath0o masers ) to several tens of percent ( oh and sio masers ) . again
the interpretation of maser polarization depends on a number of intrinsic maser properties , but in specific instances maser linear polarization can even be used to determine the full 3-dimensional field morphology .
in addition to the geometry , the linear polarization of most notably dust , can also be used to obtain a value for the strength of the plane - of - the - sky component of the magnetic field .
this is done using the chandrasekhar - fermi method , which refers to the relation between the turbulence induced scatter of polarization vectors and the magnetic field strength .
most agb magnetic field measurements come from maser polarization observations ( sio , h@xmath0o and oh ) . these have revealed a strong magnetic field throughout the circumstellar envelope . in figure
[ bvsr ] , i have indicated the magnetic field strength in the regions of the envelope traced by the maser measurements throughout agb envelopes .
while a clear trend with increasing distance from the star is seen , the lack of accurate information on the location of the maser with respect to the central stars makes it difficult to constrain this relation beyond stating that it seems to vary between @xmath3 ( solar - type ) and @xmath4 ( toroidal ) .
future observations of co polarization might be able to provide further constraints .
as the masers used for these studies are mostly found in oxygen - rich agb stars , it has to be considered that the sample is biased .
however , recent cn zeeman splitting observations @xcite seem to indicate that similar strength fields are found around carbon - rich stars . beyond determining the magnetic field strength ,
the large scale structure of the magnetic field is more difficult to determine , predominantly because the maser observations often probe only limited line - of - sights .
even though specifically oh observations seem to indicate a systematic field structure , it has often been suggested that there might not be a large scale component to the field that would be necessary to shape the outflow @xcite .
so far the only shape constraints throughout the envelope have been determined for the field around the supergiant star vx sgr ( fig.[vxsgr ] ) , where maser observations spanning 3 orders of magnitude in distance are all consistent with a large scale , possibly dipole shaped , magnetic field .
similar to the agb stars , masers are the major source of magnetic field information of post - agb and p - pne , with the majority of observations focused on oh masers .
these have revealed magnetic field strengths similar to those of agb stars ( few mg ) and a clear large scale magnetic field structure ( e.g. * ? ? ?
the most promising results have come after the detection of the so - called water - fountain sources .
these sources exhibit fast and highly collimated h@xmath0o maser jets that often extend beyond even the regular oh maser shell . with the dynamical age of the jet of order
100 years , they potentially are the progenitors of the bipolar ( p-)pne .
although the masers are often too weak for a detection of the magnetic field , observations of the arch - type of the water - fountains , w43a , have revealed a strong toroidal magnetic field that is collimating the jet ( fig .
[ w43a ] and * ? ? ?
* ) . during the pn phase , masers are rare and weak and until now only the pn k3 - 35 has had a few mg magnetic field measured in its oh masers @xcite .
fortunately , there are a few other methods of measuring pn magnetic fields . the field orientation in the dust of the nebula
can be determined using dust continuum polarization observations and current observations seem to indicate toroidal fields , with the dust alignment likely occurring close to the dust formation zone @xcite .
faraday rotation studies are potentially also able to study the magnetic field in the interaction region between the interstellar medium and the stellar outflow .
in contrast to agb stars , the central stars of pne also show atomic lines that can be used to directly probe the magnetic fields on the surface of these stars .
while measurements are still rare , observations of for example the central star of ngc 1360 , indicate a field of order several kg @xcite .
despite the strong observational evidence for evolved star magnetic fields , the origin of these fields is still unclear . in single stars , differential rotation between the agb star core and the envelope
could potentially result in sufficiently strong magnetic field @xcite .
however , as the energy loss due to a rotating magnetic field drag drains the rotation needed to maintain the field within several tens of years , an additional source of energy is needed ( e.g. * ? ? ? * ) .
if agb stars would be able to have a sun - like convective dynamo , magnetically dominated explosions could indeed result from single stars .
alternatively , the energy could be provided by the interaction with a circumstellar disk , although the origin of the disk is then another puzzle .
another explanation for maintaining a magnetic field is the interaction between a binary companion or potentially a heavy planet , with common - envelope evolution providing paths to both magnetically as well as thermally driven outflows @xcite .
a companion could be the cause of the precession seen in a number of water - fountain and ( p-)pne jets .
however , to date , the majority of the stars with measured magnetic fields do not show any other indication of binarity .
.energy densities in agb envelopes [ cols="<,<,^,^,^,^",options="header " , ] [ energy ]
until a more complete sample of magnetically active agb stars , post - agb stars and ( p-)pne is known , it is hard to observationally determine the effect of the magnetic field on these late stages of evolution . starting with the agb phase ,
a number of theoretical works have described the potential of magnetic fields in ( at least partly ) driving the stellar mass - loss through alfvn waves(e.g . * ? ? ?
* ) , or through the creation of cool spots on the surface above with dust can form easier @xcite .
as current models of dust and radiation driven winds are still unable to explain especially the mass - loss of oxygen - rich stars , magnetic fields might provide the missing component of this problem , with tentative evidence already pointing to a relation between the magnetic field strength and mass - loss rate .
other theoretical works have focused on the magnetic shaping of the stellar winds ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
but to properly determine the possible effect of the magnetic fields , it is illustrative to study the approximate ratios of the magnetic , thermal and kinematic energies contained in the stellar wind . in table.[energy ] i list these energies along with the alfvn velocities and typical temperature , velocity and temperature parameters in the envelope of agb stars .
while many values are quite uncertain , as the masers that are used to probe them can exist in a fairly large range of conditions , it seems that the magnetic energy dominates out to @xmath5 au in the circumstellar envelope .
this would correspond to the so - called launch region of magneto - hydrodynamic ( mhd ) outflows , which typically extend to no more than @xmath6 , with @xmath7 the inner - most radial scale of launch engine ( e.g. * ? ? ?
a rough constraint on @xmath7 thus seems to be @xmath8 au , close to the surface of the star .
while progress in studying the magnetic fields of evolved stars has been significant , a number of crucial questions remain to be answered .
several of these can be addressed with the new and upgraded telescopes in the near future .
for example , the upgraded evla and emerlin will uniquely be able to determine the location of the masers in the envelope with respect to the central star , giving us , together with polarization observations , crucial information on the shape and structure of the magnetic field throughout the envelopes .
alma will be able to add further probes of magnetic fields with for example high frequency masers and co polarization observations , significantly expanding our sample of stars with magnetic field measurements . with the alma sensitivity
, polarization will be easily detectable even in short observations and thus , even if not the primary goal , polarization calibration should be done .
the new low - frequency arrays can potentially be used to determine magnetic fields in the interface between the ism and pne envelopes through faraday rotation observations .
* under what conditions does the magnetic field dominate over e.g. binary interaction when shaping outflows ?
* are magnetic fields as widespread in evolved stars as they seem ? * what is the origin of the agb magnetic field - can we find the binaries / heavy planets that might be needed ? * is there a relation between agb mass - loss and magnetic field strength ? wv acknowledges the support by the deutsche forschungsgemeinschaft ( dfg ) through the emmy noether research grant vl 61/3 - 1 , and the work by the various researchers that have been crucial in the development of the area of evolved star magnetic field research ( including those that i neglected to reference in this review ) . | asymmetric planetary nebulae ( pne ) common research into shaping processes fundamental of understanding return of processed material into ism by low- intermediate - mass stars at end evolution standard interacting winds scenario explain pn properties important discovery collimated outflows of pre - pne ( p - pne ) have momentum exceeds by radiation pressure source of momentum excess debated commonly cause being magnetic fields , binary or disk interaction or combination similarities with jets outflows young stellar objects study of p - pne outflows provides research opportunities into universal mechanism of jet launching review observational evidence for strong magnetic fields in pne agb and post - agb progenitors overview of methods study magnetic fields new instruments available discuss questions related with new instruments exception magnetic field strength estimated energy equilibrium synchrotron observations magnetic field strength structure determined from polarization observations circular polarization through zeeman splitting to measure magnetic field strength measures total field strength when splitting large and line - of - sight component when smallpredominant source magnetic field strength information late stellar evolution from maser circular polarization observations particularly common sio h@xmath0o oh masers show polarization fractions from @xmath1% up to @xmath2% ( oh compactness strength excellent with high angular resolution analysis of maser polarization not straightforward long time before maser observations provide accurate magnetic field measurements recently first attempts to detect zeeman splitting of non - maser molecular lines in circumstellar envelopes at shorter wavelength sub advent atacama large ( sub-)millimeter array enhance studies linear polarization structure plane - of sky magnetic field observed in dust molecular lines percentages of linear polarization range from few percent to several tens of percent ( oh sio masers interpretation of maser polarization depends on intrinsic maser properties linear polarization can used to determine full 3-dimensional field morphology linear polarization of dust can used to value for strength of plane - of - the - sky component magnetic fielddone using chandrasekhar - fermi method refers relation between turbulence scatter polarization vectors magnetic field strength agb magnetic field measurements from maser polarization observations sio@xmath0o oh revealed strong magnetic field circumstellar envelope in figure indicated magnetic field strength in regions traced by maser measurements clear trend with increasing distance from star lack of accurate information on location maser difficult to constrain relation between @xmath3 and @xmath4 toroidal future observations co polarization might provide constraints masers in oxygen - rich agb stars sample is biased recent cn zeeman splitting observations indicate similar strength fields around carbon - rich stars field strength large scale structure difficult to determine maser observations probe limited line - of - sights oh observations indicate systematic field structure large scale component to shape outflow @xcite shape constraints for field around supergiant star vx sgr maser observations spanning 3 orders of magnitude consistent with large scale dipole shaped magnetic field masers major source of magnetic field information post - agb p - pne majority of observations focused on oh masersrevealed magnetic field strengths similar agb stars few mg clear large scale magnetic field structure promising results after detection water - fountain sources sources exhibit fast highly collimated maser jets beyond regular oh maser shell dynamical age jet 100 years progenitors of bipolar p-)pne masers weak for detection magnetic field observations water - fountains revealed strong toroidal magnetic field collimating jet fig during pn phase masers rare weak only pn k3 - 35 few mg magnetic field measured in masers other methods measuring pn magnetic fields field orientation in dust nebula determined using dust continuum polarization observations observations indicate toroidal fields dust alignment close to dust formation zone faraday rotation studies study magnetic field in interaction between interstellar medium stellar outflow central stars of pne show atomic lines probe magnetic fields measurements rare observations central star ngc 1360 indicate field of order several kg strong evidence for evolved star magnetic fields origin unclear in single stars differential rotation between agb star core envelope could result in strong magnetic fieldenergy loss due to rotating magnetic field drains rotation field additional source energy needed if agb stars have sun - like convective dynamo magnetically dominated explosions could result from single stars alternatively energy could provided by interaction with circumstellar disk origin puzzle explanation for magnetic field is interaction between binary companion or heavy planet common - envelope evolution paths to magnetically thermally driven outflows companion could cause of precession in water - fountain ( p-)pne jets majority stars with magnetic fields show indication of binarity.energy densities in agb envelopes until complete sample of magnetically active agb stars known hard to determine effect magnetic field on late stages evolution theoretical works described potential of magnetic fields driving stellar mass - loss through alfvn waves creation of cool spots on surface dust models of dust radiation winds explain mass - loss of oxygen - rich stars magnetic fields might provide missing component tentative evidence to relation between magnetic field strength and mass - loss rate theoretical works focused on magnetic shaping of stellar winds* ? to determine effect magnetic fields study ratios magnetic thermal kinematic energies in stellar wind table.[energy ] i list energies alfvn velocities typical temperature parameters in envelope of agb stars values uncertain masers large range conditions magnetic energy dominates to @xmath5 au in circumstellar envelope launch region of magneto - hydrodynamic mhd outflows extend to no more than @xmath6 @xmath7 inner - most radial scale of launch engine constraint on @xmath7 @xmath8 au close to surface star progress studying magnetic fields of evolved stars significant crucial questions remain to be answered addressed with new upgraded telescopes upgraded evla emerlin determine location of masers in envelope central star crucial information on shape structure magnetic field alma add further probes magnetic fields with high frequency masers co polarization observations expanding sample of stars with magnetic field measurements alma sensitivity polarization detectable in short observations polarization calibration should be done new low - frequency arrays determine magnetic fields in interface between ism and pne envelopes through faraday rotation observationsconditions magnetic field dominate over binary interaction shaping outflows ? magnetic fields widespread in evolved stars ? origin of agb magnetic field binaries heavy planets needed ? relation between agb mass - loss magnetic field strength ? acknowledges support deutsche forschungsgemeinschaft emmy noether research grant 61/3 - 1 work researchers crucial evolved star magnetic field research neglected | 0.460571 | 2,636 | 1,255 | , Saving $0.1 in GPT-4. |
20,680 | distorted wave impulse approximation calculations of @xmath0-hypernuclear production rates in stopped @xmath1 reactions on several @xmath2-shell targets used recently in experiments by the finuda collaboration are reported .
chirally motivated @xmath3 in - medium transition amplitudes are employed and the sensitivity of the calculated rates to the initial @xmath1-atomic wave functions and final pion distorted waves is studied .
the calculated rates are compared with measured rates , wherever available , confirming earlier observations that ( i ) the calculated rates are generally lower than the measured rates , and ( ii ) the deeper the @xmath1-nuclear potential , the worse is the discrepancy .
the @xmath4 dependence of the calculated @xmath5 production rates is discussed for the first time , providing a useful tool to resolve the issue of depth of the @xmath1-nuclear potential near threshold . | @xmath0-hypernuclear production in @xmath6 reactions , in which the final state is uniquely identified by measuring the outgoing pion momentum , was reported for the first time in stopped @xmath1 experiments at cern in 1973 @xcite and more recently in experiments at kek @xcite , bnl @xcite and at da@xmath7ne , frascati , by the finuda collaboration @xcite . on the theoretical side , several distorted wave impulse approximation ( dwia ) calculations of @xmath6 hypernuclear production rates
have been reported @xcite , but none of them led to satisfactory agreement with the measured rates . in general , these calculated capture rates fall below the experimentally reported rates , with the exception of the old cern data for @xmath8 @xcite .
the present paper primarily covers the production of @xmath9 , @xmath10 , @xmath11 , @xmath12 and @xmath13 , for all of which preliminary data have recently been reported @xcite .
we focus on the @xmath4 dependence of the calculated rates , hitherto not explored systematically , to look for further tests of the role played by initial- and final - state interactions . in conjunction with previous calculations
, we use the dwia .
several @xmath1 nuclear optical potentials are used to generate the required initial - state @xmath1-atom wave functions , and several pion nuclear optical potentials are used to generate final - state pion distorted waves ( dws ) .
the underlying @xmath14 reaction amplitude is studied in free space , as well as in the nuclear medium , within the chiral lagrangian framework @xcite to generate in - medium branching ratios ( brs ) for stopped @xmath1 reactions .
past works @xcite used brs extrapolated from emulsion experiments @xcite .
we compare results obtained in both approaches .
the present paper is organized as follows .
the capture at rest dwia formalism is outlined in sec .
the choice of the microscopic chiral model for @xmath15 reactions at rest , together with the brs derived in this model and the input wave functions to the dwia calculations , are specified in sec .
results of @xmath0 hypernuclear production rate calculations for stopped @xmath1 reactions are presented and discussed in sec .
iv , with a brief conclusion given in sec . v.
we study the reaction @xmath16 in which a @xmath17 meson is captured on a target nucleus denoted as @xmath4 in its ground state @xmath18 , from an atomic orbit @xmath19 into a final state @xmath20 of a @xmath0 hypernucleus @xmath21 plus an outgoing pion .
the superscript @xmath22 denotes the pion charge ( @xmath23 for @xmath24 and @xmath25 respectively ) .
we follow the capture at rest calculation formalism detailed in ref . @xcite . in the dwia , the nuclear reaction eq .
( [ reakcecela ] ) is induced by the one - baryon transition @xmath26 on a nucleon @xmath27 to a hyperon @xmath28 , with the in - medium @xmath29 matrix assumed here to be of @xmath30-wave type : @xmath31 the charge indices are omitted for simplicity and the dw transition form factor is given by @xmath32 where @xmath33 stands for the nuclear to hypernuclear transition matrix element . the @xmath17-atomic wave function @xmath34 is obtained by solving the klein - gordon equation with a @xmath17-nuclear strong interaction optical potential @xmath35 added to the coulomb potential @xmath36 generated by the nuclear charge distribution plus vacuum polarization .
the dependence on the magnetic quantum number @xmath37 was suppressed on the left - hand side ( l.h.s . ) of eq .
( [ eq : dw ] ) .
the pion dw @xmath38 in the final state is given in terms of a partial - wave expansion : @xmath39 the radial wave function @xmath40 , which reduces to the spherical bessel function @xmath41 for a free pion , solves the klein - gordon equation with the pion - nuclear optical potential , plus the appropriate electromagnetic potential for a charged pion .
the nuclear capture rate per stopped @xmath17 in the reaction eq .
( [ reakcecela ] ) is given by @xmath42 where the fractions @xmath43 are the elementary brs for mononucleonic @xmath17 absorption at rest in the nuclear medium .
the brackets @xmath44 mean that the absolute square of the dw transition form factor is to be averaged on the initial states and summed over the final ones . the kinematical factor in front of @xmath45 in eq .
( [ eq : r / k ] ) appears because of transformation of the two - body scattering amplitude , which describes the elementary reaction eq .
( [ reakce1nukl ] ) , into the many - body center - of - mass frame .
the momentum @xmath46 of the outgoing pion is determined by energy conservation , and @xmath47 stands for the reduced energy in the final state , @xmath48 where @xmath49 , @xmath50 are appropriately averaged in - medium quantities .
finally , @xmath51 denotes the effective nuclear density available to the @xmath1 capture process , @xmath52 where the nucleon density @xmath53 and the @xmath1-atomic radial wave function are normalized according to @xmath54 where @xmath55 denotes the number of neutrons or protons for @xmath56 , respectively .
the last factor on the right - hand side ( r.h.s . ) of eq .
( [ eq : r / k ] ) , @xmath57 is loosely termed the capture rate per hyperon @xmath28 because its derivation assumes that the capture reaction eq .
( [ reakce1nukl ] ) is the only one available .
it can be decomposed into contributions , which correspond to transitions with multipolarity @xmath58 from a given @xmath59 nuclear shell to a given @xmath60 hypernuclear shell , in the following form @xcite : @xmath61 here , @xmath62 is the neutron ( proton ) occupation number of the target nuclear shell for @xmath63 , the clebsch - gordan coefficient squared accounts for the conservation of angular momentum and parity , and the entities @xmath64 are the appropriate averages of the absolute squares of the dwia amplitudes @xmath65 where @xmath66 are the radial parts of the one - baryon wave functions .
( [ eq : r / y])-([eq : il ] ) assume that the dwia capture rate calculation does not depend on the total angular momenta @xmath67 for the orbits in question .
this was justified by the numerical calculation performed in ref .
@xcite , where more general formulae for the dependence on @xmath68 can be found as well , and is checked later in the present paper .
in this section , we specify the entities that are needed to perform numerical calculations of nuclear capture rates .
first , we outline the model used for the one - baryon capture process eq .
( [ reakce1nukl ] ) .
subsequently , we specify the baryon and meson nuclear potentials chosen to generate initial- and final - state wave functions for use in the evaluation of the dwia amplitudes eq .
( [ eq : il ] ) that serve as input to nuclear capture rate calculations . as a function of the nuclear density for capture on neutrons ( solid curve ) and on protons ( dashed curve ) . ] for the calculation of brs @xmath43 of the one - baryon capture process eq .
( [ reakce1nukl ] ) , we adopted the effective potential model based on chiral symmetry , as described in detail in refs .
the required brs are obtained in coupled - channel calculations that include ten meson - baryon channels coupled to the @xmath70 system @xcite .
we also take the effects of pauli blocking and kaon self - energy in the nuclear medium into account @xcite .
the dependence of the calculated brs on the nuclear density is demonstrated in fig .
[ figbrratios ] as a function of the fractional nuclear density @xmath71 , where @xmath72 is nuclear - matter density .
although the central nuclear density varies along the periodic table roughly in the range @xmath73 , the brs shown in the figure do not change much over this range of densities .
therefore , the precise dependence on @xmath74 may be neglected , and we assume that the @xmath69 capture reaction takes place at a proton ( or neutron ) density @xmath75 . for further applications ,
we denote the brs obtained at nuclear density @xmath75 by br1 , and the brs obtained in vacuum by br2 .
the use of a microscopic model for the @xmath69 brs is one of the novelties of the present paper .
past works used brs derived indirectly by extrapolating from measurements performed in carbon and freon emulsions @xcite , a procedure that is prone to systematic errors .
we use these emulsion brs ( labeled here as br3 ) to compare with brs obtained from our microscopic chiral model .
this is shown in table [ tabbrratios ] , where the maximum difference between the various brs ( br1 , br2 , and br3 ) is as large as @xmath76 .
however , this variation is still small compared to other effects discussed below .
the ratio of 1/2 for brs on a proton to brs on a neutron follows from charge independence , which is implemented by conserving isospin in our model for @xmath69 . .@xmath77 branching ratios ( in units of @xmath78 ) .
[ cols="<,^,<,^,^,<,^,^ , < " , ] calculated capture rates for the @xmath80 @xmath81 and @xmath82 transitions [ see eq .
( [ eq : r / y ] ) ] are presented in table [ tabcaprates1p ] for @xmath8 and @xmath83 targets . for completeness
, we also included the @xmath84 @xmath85 transition discussed in sec . iv b1 and added experimental rates from kek @xcite and bnl @xcite . here , the reason for choosing kek over cern @xcite and finuda @xcite is that for @xmath11 production , the kek @xcite rates are the closest to the isospin factor 2 expected in getting these rates from the @xmath86 rates which were measured only at bnl @xcite .
capture rate in @xmath11 varies from @xmath87 @xcite to @xmath88 @xcite .
] of course , if further ( @xmath89 ) experiments are done on @xmath8 , which would lead to different results from those of ref .
@xcite , one s preference might change accordingly .
capture rate to @xmath11 over that of @xmath86 is largely caused by the ratio of the @xmath90 brs which in the limit of good isospin is 2 . a slight departure from this ratio
is caused by charge - dependent effects in our calculation , notably from the outgoing pion dws .
] similar to the summed @xmath5 rate discussed in sec .
iv b1 , the summed @xmath79 rate for the deep potential [ @xmath91 is smaller by a factor @xmath92 than for the relatively shallow potential [ @xmath93 . excluding the old cern data @xcite , the calculated @xmath79 capture rates are generally smaller than the experimentally reported rates , with the exception that the [ @xmath93 rates for @xmath86 are larger than the reported rate @xcite .
since the @xmath79 spectral strength is partly mixed into the ( @xmath94 ) quasifree continuum , and its extraction from the measured spectra is considerably more ambiguous than the extraction of the summed @xmath5 production rate , we do not proceed further to confront theory with experiment for the summed @xmath79 production rate .
we performed dwia ( @xmath94 ) calculations for @xmath2-shell targets , using several @xmath1 and pion wave functions to test the sensitivity of the calculated hypernuclear capture rates to the choice of these wave functions .
the calculated capure rates are generally smaller than the measured ones ; the deeper the @xmath1 potential , the smaller is the capture rate . since the absolute normalization of capture at rest experimental rates is a delicate matter , we suggest to focus on the @xmath4 dependence of the measured rates , expecting it to be largely free of the absolute normalization of the data .
the calculated capture rates for a given @xmath1 optical potential decrease as a function of @xmath4 , with the fractional difference between the rates calculated for the two extreme @xmath1 optical potentials , the shallow [ @xmath93 and the deep [ @xmath91 , increasing steadily with @xmath4 .
we find other dependencies of the calculated capture rates to be secondary to the dependence on the @xmath1-atomic wave function in the range studied here .
we argue that a dedicated experimental study of @xmath5 capture rates in @xmath2-shell targets , such as reported recently by the finuda collaboration in a preliminary form @xcite , could yield useful information on the depth of the threshold @xmath1 optical potential by comparing the measured @xmath4 dependence with the @xmath4 dependence of the calculated capture rates listed in table [ tabcaprates1s ] .
this work was supported by the gauk grant no .
91509 and the gacr grant no .
202/09/1441 , as well as by the eu initiative fp7 , hadronphysics2 , under project no . 227431 . | @xmath0-hypernuclear production in @xmath6 reactions final state identified measuring outgoing pion momentum reported in @xmath1 experiments cern 1973 recently kek bnl da@xmath7ne frascati finuda collaboration distorted wave impulse approximation calculations of @xmath6 hypernuclear production rates reported none satisfactory agreement measured rates calculated capture rates fall below experimentally reported rates exception old cern data @xmath8 present paper covers production of @xmath9 preliminary data recently reported focus on @xmath4 of calculated rates tests initial- final - state interactions use dwia @xmath1 nuclear optical potentials generate initial - state @xmath1-atom wave functions pion nuclear optical potentials final - state pion distorted waves @xmath14 reaction amplitude studied in free space nuclear medium chiral lagrangian framework generate in - medium branching ratios for stopped @xmath1 reactions past works used brs from emulsion experiments compare results both approaches present paper organized capture at rest dwia formalism outlined in sec .microscopic chiral model for @xmath15 reactions rest brs input wave functions dwia calculations specified sec . results @xmath0 hypernuclear production rate calculations stopped @xmath1 reactions sec iv conclusion sec v study reaction @xmath16 @xmath17 meson captured target nucleus @xmath4 ground state atomic orbit final state hypernucleus outgoing pion superscript @xmath22 denotes pion charge capture at rest calculation formalism ref @xcite nuclear reaction induced one - baryon transition @xmath26 nucleon to hyperon @xmath28 in - medium @xmath29 matrix @xmath30-wave type charge indices omitted dw transition form factor @xmath32 nuclear hypernuclear transition matrix element @xmath17-atomic wave function @xmath34 klein - gordon equation @xmath17-nuclear strong interaction optical potential @xmath35 added coulomb potential @xmath36 nuclear charge distribution vacuum polarization dependence on magnetic quantum number @xmath37 suppressed left - hand side eq .pion dw @xmath38 final state partial - wave expansion @xmath39 radial wave function @xmath40 reduces spherical bessel function @xmath41 free pion solves klein - gordon equation pion - nuclear optical potential electromagnetic potential charged pion nuclear capture rate per stopped @xmath17 reaction eq given @xmath42 @xmath43 elementary brs mononucleonic absorption rest nuclear medium brackets @xmath44 absolute square dw transition form factor averaged initial states summed over final ones kinematical factor @xmath45 transformation two - body scattering amplitude reaction many - body center - of - mass frame momentum @xmath46 outgoing pion determined energy conservation @xmath47 reduced energy final state @xmath48 @xmath49 @xmath50 averaged in - medium quantities @xmath51 effective nuclear density @xmath1 capture process @xmath52 nucleon density @xmath53 @xmath1-atomic radial wave function normalized @xmath54 @xmath55 number neutrons protons @xmath56 last factor @xmath57 capture rate per hyperon @xmath28 capture reactionreakce1nukl ] only available decomposed into contributions transitions multipolarity @xmath58 from @xmath59 nuclear to @xmath60 hypernuclear shell @xcite : @xmath61 @xmath62 neutron ( ) occupation number target nuclear shell @xmath63 clebsch - gordan coefficient conservation angular momentum parity entities @xmath64 averages of absolute squares dwia amplitudes @xmath65 @xmath66 radial parts of one - baryon wave functions y- il dwia capture rate calculation depend on total angular momenta @xmath67 orbits justified by numerical calculation in ref. @xcite general formulae for dependence on @xmath68 checked later in paper specify entities needed numerical calculations nuclear capture rates outline model for one - baryon capture process specify baryon meson nuclear potentials initial- final - state wave functions dwia amplitudes input to nuclear capture rate calculations function of nuclear density for capture on neutrons protons calculation of brs @xmath43 one - baryon capture process adopted effective potential model based on chiral symmetry describedrequired brs obtained in coupled - channel calculations ten meson - baryon channels @xmath70 system @xcite effects pauli blocking kaon self - energy nuclear medium dependence calculated brs on nuclear density demonstrated in fig . function of fractional nuclear density @xmath71 @xmath72 nuclear - matter density central nuclear density varies @xmath73 brs change over dependence on @xmath74 neglected assume @xmath69 capture reaction at proton or neutron density @xmath75 denote brs obtained at nuclear density @xmath75 by br1 vacuum by br2 microscopic model for @xmath69 brs present paper past works used brs from measurements carbon freon emulsions use emulsion brs compare with brs microscopic model shown in table maximum difference between brs br1 large as @xmath76 variation small compared to other effects ratio of 1/2 for brs on proton to neutron from charge independence conserving isospin in model for @xmath69@xmath77 branching ratios in units of @xmath78 )cols=" calculated capture rates for @xmath80 transitions presented in table for @xmath8 @xmath83 targets included @xmath84 @xmath85 transition added experimental rates from kek @xcite bnl @xcite choosing kek over cern @xcite finuda @xcite @xmath11 production kek @xcite rates closest to isospin factor 2 @xmath86 rates at bnl @xcite capture rate in @xmath11 varies from @xmath87 @xcite to @xmath88 @xcite if further @xmath89 experiments on @xmath8 preference might change capture rate to @xmath11 over @xmath86 caused by ratio of @xmath90 brs limit isospin is 2 slight departure caused by charge - dependent effects summed @xmath5 rate summed @xmath79 rate for deep potential @xmath91 smaller by factor @xmath92 than shallow potential @xmath93 excluding old cern data calculated @xmath79 capture rates smaller than reported rates exception @xmath93 rates for @xmath86 larger than reported rate @xcite@xmath79 spectral strength mixed into @xmath94 ) quasifree continuum extraction from ambiguous than summed @xmath5 production rate theory experiment for summed @xmath79 production rate performed dwia @xmath94 calculations for @xmath2-shell targets using @xmath1 pion wave functions calculated hypernuclear capture rates to wave functions calculated capure rates smaller than measured deeper @xmath1 potential smaller capture rate absolute normalization of capture rates delicate focus on @xmath4 dependence of measured rates free of absolute normalization calculated capture rates for @xmath1 optical potential decrease as function of @xmath4 fractional difference between rates for potentials increasing with @xmath4 other dependencies of calculated capture rates secondary to dependence on @xmath1-atomic wave function argue dedicated experimental study of @xmath5 capture rates in @xmath2-shell targets could yield information on depth threshold @xmath1 optical potential comparing measured @xmath4 dependence with calculated capture rates work supported by gauk grant 91509 gacr grant 202/09/1441 eu initiative fp7 , hadronphysics2 project no. | 0.494583 | 3,037 | 1,565 | , Saving $0.1 in GPT-4. |
20,688 | we present _ ab initio _ calculations of total energies of mn atoms in various interstitial positions .
the calculations are performed by the full - potential linearized plane - wave method .
the minimum energy is found for tetrahedral t(as@xmath0 ) position , but the energy of the t(ga@xmath0 ) site differs by only a few mev . the t(ga@xmath0 ) position becomes preferable in the p - type materials . in samples with one substitutional and one interstitial mn the mn atoms tend to form close pair with antiparallel magnetic moments .
+ we also use the spin - splitting of the valence band to estimate the exchange coupling @xmath1 for various positions of mn .
the exchange parameter is the same for the substitutional and for the t(as@xmath0 ) position and it is somewhat smaller in the case of the t(ga@xmath0 ) position . | diluted magnetic semiconductors , such as ( ga , mn)as , are important materials combining ferromagnetic behavior with a sensitivity to doping characteristic for semiconductors .
this leads to their interesting physical properties and makes possible their applications e.g. in spin - electronics @xcite .
although ( ga , mn)as has been extensively studied in the last years , some aspects of the incorporation of mn into the crystal lattice still remain unclear .
it was assumed that in the well defined samples mn simply substitutes for the host cation .
only recently it was pointed out @xcite that the presence of interstitial mn may explain some peculiar properties of ( ga , mn)as such as the low doping efficiency of mn acceptors @xcite .
the main reason for considering the mn atoms on the interstitial positions was that they act as double donors and partly compensate the mn acceptors in the substitutional positions .
at the same time , channelling rutherford backscattering experiments proved that a large fraction of mn atoms indeed occupies the interstitial positions ( mn@xmath2 ) in the as - grown samples @xcite .
a correlation between the removal of mn interstitials and increase of the conductivity , the curie temperature , and saturation magnetization has been found .
in addition to the self - compensation effect , the interstitial mn atom also reduce the number of local moments that participate in the ferromagnetic state .
this was explained by pairing of mn@xmath3 with the mn atoms in the substitutional positions ( mn@xmath4 ) due to their coulomb attraction @xcite . at the bonding distance , the antiferromagnetic superexchange within the pair
is assumed to outweigh the hole mediated ferromagnetic exchange . as a result ,
the moments of the paired mn atoms have opposite directions and the pair as a whole has no magnetic moment .
the _ ab initio _ studies of the interstitial mn showed the differences in the electronic structure of the mn atom in the interstitial and substitutional positions @xcite and also a possible reaction path for the incorporation of mn into the gaas lattice @xcite .
it was also found that the increase of the lattice constant of ga@xmath5mn@xmath6as with increasing content of mn is partly due to the presence of mn in the interstitial positions @xcite .
recently , blinowski and kacman @xcite investigated the coupling of the local moment on the interstitial mn to the spin of the carriers in the valence band . using a simplified tight - binding model they showed that the local moment on mn atom in the tetrahedral position with four ga neighbors , t(ga@xmath0 ) , is effectively decoupled from the spin of the holes in the valence band .
they claimed that the coupling is weak because , in addition to the reduced number of the holes , also the coupling constant @xmath1 is small for the interstitial mn in the t(ga@xmath0 ) position . as a result ,
the local moment is not subject to the ferromagnetic coupling with the moments around it .
this opens the way for the af exchange to be important in the mn@xmath2-mn@xmath7 pair , as anticipated in ref .
@xcite .
there are , however , several open questions concerning the magnetic interactions of mn in the interstitial positions .
first of all , the spin - polarized band structures @xcite did not show any indication of different values of @xmath1 for substitutional and interstitial mn , at least for the t(as@xmath0 ) positions . in addition , the mn@xmath2-mn@xmath7 pair as a whole is expected @xcite to have only a small magnetic moment , but the exchange coupling with mn@xmath7 being uncompensated by the contribution of mn@xmath2 should strongly polarize the holes .
that is why we performed a more detailed study of the interstitial mn and its spin interactions .
we use the density - functional , full - potential linearized augmented plane wave calculations ( fplapw @xcite ) to obtain the electronic structure of ( ga , mn)as with mn atoms in various crystallographic positions . the calculated total energies are used to compare different positions of the interstitial mn and to estimate the strength of the mn@xmath7-mn@xmath2 pair interactions .
the splitting of the valence band for the majority- and minority - spin electrons is used to compare the corresponding values of the exchange parameters @xmath1 .
we compare the total energies for three interstitial positions of mn in gaas .
there are two inequivalent tetrahedral positions in the zinc - blende structure of gaas , t(as@xmath0 ) and t(ga@xmath0 ) .
they are surrounded by four as and ga atoms , respectively .
the ( unrelaxed ) distances of these nearest neighbors are equal to the length @xmath8 of the ga - as bond , i.e. to the mn - as distance for the substitutional mn@xmath7 .
the local arrangement around the interstitial mn is , in contrast to mn@xmath7 , characterized by another six close neighbors at the distance @xmath9 . in the hexagonal position
, the interstitial mn has three ga and three as atoms at the same distance @xmath10 , and no other close neighbors .
we represent gaas with mn@xmath2 by hexagonal ga@xmath11as@xmath11mn supercells .
the c - axis coincides with the body diagonal of the conventional cubic unit cell .
the symmetry of the cell does not change if we shift the mn interstitial along the c - axis from t(as@xmath0 ) to hexagonal and t(ga@xmath0 ) positions .
this makes possible to perform all calculations under the same conditions .
.total energy , @xmath12 , and total spin , @xmath13 , of the unit cell of ga@xmath11as@xmath11mn with various interstitial positions of mn .
ga@xmath14zn@xmath15as@xmath11mn samples are used to simulate p - type materials . [ cols="^,^,^,^",options="header " , ] this effect is formally described by the kondo exchange interaction between the local spins @xmath16 at sites @xmath17 occupied by mn and the spin density @xmath18 due to the itinerant holes @xcite , @xmath19 the exchange parameter @xmath1 characterizes the strength of the coupling . within the mean - field theory , the kondo exchange interaction results in the splitting of the valence band edge @xmath20 .
the splitting @xmath21 is proportional to the size @xmath22 of the local spins and to the concentration @xmath23 of magnetic ions , @xmath24 assuming spin @xmath25 for the holes , @xmath26 is the lattice constant .
we use eq .
( [ mfdev ] ) to determine the exchange parameter @xmath1 from our spin - polarized band structures .
the results of the calculations for the unrelaxed geometries presented here and in ref .
@xcite are summarized in table iii .
the resulting values of @xmath1 are overestimated from the same reasons as discussed above in sect .
being divided by a factor of two ( the reduction factor for the energy nominator ) , they approach the realistic values @xcite . in this work , however , we concentrate on the comparison of @xmath1 for different geometries .
the overestimate of @xmath1 due to the reduced band gap is expected to be similar in all cases and not very important in this respect . in fig .
3 , we plot @xmath21 against @xmath23 to visualize the linear increase of the band splitting with the concentration of mn .
the most noticeable result is that all points in fig .
3 , obtained for substitutional mn as well for mn in the interstitial t(as@xmath0 ) and t(ga@xmath0 ) positions lie around the same line .
this means that the value of @xmath1 does not depend much on the position of the mn atom in the lattice .
this result is not very surprising in the case of the substitutional mn and t(as@xmath0 ) position . in either case ,
mn is surrounded by four as atoms at the same distance @xmath8 and the same degree of the hybridization of the valence band states with mn d - states can be expected . the results for the t(ga@xmath0 )
position are not simple to interpret .
the value of @xmath1 for both ga@xmath11mnas@xmath11 and ga@xmath27mnas@xmath27 is smaller than but comparable with the result for the substitutional mn@xmath7 site . assuming for simplicity that only the hybridization of the mn d - states with p - states of the nearest neighbors is relevant for the spin splitting @xmath21 , the values of @xmath1 simply reflect the orbital composition of the valence band .
it is well known that the top of the valence band in gaas , as well as in other iii - v semiconductors , is composed of both anion and cation p - states , in the proportion approximately @xmath28 , @xmath29 @xcite .
this means that the same proportion ( roughly 3:1 ) should be expected also for the exchange parameters @xmath1 parameters corresponding to mn atoms in substitutional and t(ga@xmath0 ) positions .
the strong deviation of the actual density - functional results from this model expectation may indicate that the hybridization of the mn d - states with more distant neighbors has a remarkable influence on the magnetic interactions .
the fact that the magnetic behavior of the interstitial mn is rather insensitive to its position in the crystal has two important implications concerning the mn@xmath7-mn@xmath2 pairs .
first of all , the pair as a whole does not interact much with the spin of the holes because the effects due to mn@xmath4 and mn@xmath2 compensate one another . in this respect ,
our _ ab initio _ results overcome the troubles of the simplified tight - binding models mentioned in the introduction . in addition , the effective annihilation of mn@xmath7 due to pairing with mn@xmath2 is not restricted to the closest pairs with mn@xmath2 in the t(ga@xmath15mn ) position , but it works as long as the mn atoms are close one to another and the exchange coupling in the pair remains antiferromagnetic .
we used the fplapw method to obtain total energies of supercells simulating various geometric and magnetic configurations of mn atoms in ( ga , mn)as . in absence of other defects ,
the ground state of the mn interstitials is the tetrahedral t(as@xmath0 ) position .
the energy of the t(ga@xmath0 ) position , however , is almost the same .
the situation changes in the p - type material where the t(ga@xmath0 ) position has a lower energy .
the hexagonal interstitial position has much higher energy and represents a barrier for diffusion of mn from one to another interstitial site .
the barrier , and consequently also the mobility of the interstitial mn , depends on the doping .
the exchange coupling @xmath1 of mn interstitials with the holes in the valence band is , for both t(as@xmath0 ) and t(ga@xmath0 ) , close to the value of @xmath1 obtained for the substitutional mn .
this is not consistent with the simplest tight - binding picture of mn d - states that hybridize only with the nearest neighbors . in this way ,
our result indicate that the hybridization with more distant neighbors may be also important for the magnetic interactions .
the mn interstitials are attracted to the substitutional mn and form stable and magnetically inactive pairs .
the density - functional estimate for both binding energy of the pair and for the energy of the antiferromagnetic coupling is of order of 0.3 ev .
this fits well with the present day notion of the interstitial mn in ( ga , mn)as .
@xcite . in contrary to the general opinion , however , we found that the efficient pairing is not restricted to mn@xmath2 in the t(ga@xmath15mn ) position and we showed the importance of the close next - nearest neighbors for the properties of the interstitial mn .
this work has been done within the project avoz1 - 010 - 914 of the ascr .
the financial support provided by the grant agency of the ascr ( grant no .
a1010214 ) and by rtn project no .
hprn - ct-2000 - 00143 the ec is acknowledged . | diluted magnetic semiconductors ( ga , mn)as combining ferromagnetic behavior sensitivity to doping leads to interesting physical properties applications in spin - electronics ( ga , mn)as extensively studied incorporation mn into crystal lattice unclear assumed mn substitutes for host cation recently presence interstitial mn may explain properties low doping efficiency of mn acceptors mn atoms on interstitial positions act as double donors compensate mn acceptors in substitutional positions rutherford backscattering experiments proved large fraction of mn atoms occupies interstitial positions ( ) in samples correlation between removal of mn interstitials conductivity curie temperature saturation magnetization found interstitial mn local moments in ferromagnetic state explained by pairing of mn@xmath3 with mn atoms in substitutional positions ( mn@xmath4 ) due to coulomb attraction bonding antiferromagnetic superexchange pair outweigh mediated ferromagnetic exchange moments of paired mn atoms have opposite directions pair has no magnetic moment studies interstitial mn showed differences in electronic structure mn atom in interstitial substitutional positions @xcite possible reaction path for incorporation mn into gaas lattice increase lattice constant ga@xmath5mn@xmath6as with increasing content mn due to presence mn in interstitial positions blinowski kacman investigated coupling local moment interstitial mn to spin carriers valence band showed local moment mn atom in tetrahedral position with four ga neighbors decoupled from spin holes valence band coupling weak reduced number holes coupling constant @xmath1 small for interstitial mn in t(ga@xmath0 ) position local moment not subject to ferromagnetic coupling moments around opens af exchange important in mn@xmath2-mn@xmath7 pair open questions concerning magnetic interactions mn in interstitial positions spin - polarized band structures show different values @xmath1 for substitutional interstitial mn positions mn@xmath2-mn@xmath7 pair expected have small magnetic moment exchange coupling with mn@xmath7 uncompensated by contribution mn@xmath2 should polarize holesperformed detailed study interstitial mn spin interactions density - functional full - potential linearized augmented plane wave calculations electronic structure ga mn)as mn atoms crystallographic positions calculated energies compare positions interstitial mn estimate strength mn@xmath7-mn@xmath2 pair interactions splitting valence band majority minority - spin electrons exchange parameters total energies three interstitial positions mn in gaas two inequivalent tetrahedral positions zinc - structure t(as@xmath0 t(ga@xmath0 ) surrounded by four as ga atoms distances neighbors equal to length ga - as bond mn - as distance substitutional mn@xmath7 local arrangement interstitial mn six close neighbors distance hexagonal position three ga three as atoms same distance no other close neighbors represent gaas with mn@xmath2 by hexagonal ga@xmath11as@xmath11mn supercells c - axis coincides with body diagonal conventional cubic unit cell symmetry change mn interstitial from t(as to hexagonal t(ga@xmath0 positions all calculations under same conditionsenergy @xmath12 spin @xmath13 of unit cell ga@xmath11as@xmath11mn with interstitial positions mn ga@xmath14zn@xmath15as@xmath11mn samples simulate p - type materials effect described by kondo exchange interaction between local spins @xmath16 at sites occupied mn spin density @xmath18 due itinerant holes exchange parameter @xmath1 characterizes strength coupling kondo exchange interaction results in splitting valence band edge @xmath20 splitting @xmath21 proportional to size local spins concentration of magnetic ions spin @xmath25 for holes @xmath26 lattice constant use eq . mfdev determine exchange parameter @xmath1 from spin - polarized band structures results calculations for unrelaxed geometries summarized in table iii values @xmath1 overestimated divided by factor two approach realistic values on comparison @xmath1 for geometries overestimate @xmath1 due to reduced band gap similar not important fig . 3 plot @xmath21 against @xmath23 linear increase band splitting with concentration mn points in fig obtained substitutional mn in interstitial t(as positions same line value of @xmath1 depend on position mn atom in lattice not surprising substitutional mn t position mn surrounded by four atoms same distance same degree hybridization valence band states with mn d - states expected results for t(ga@xmath0 ) position not simple value of @xmath1 for ga@xmath11mnas and ga@xmath27mnas smaller comparable with substitutional mn@xmath7 site hybridization mn d - states with p - states nearest neighbors relevant for spin splitting values @xmath1 reflect orbital composition valence band top valence band composed of anion and cation p - states approximately same proportion roughly 3:1 ) expected for exchange parameters parameters mn atoms in substitutional and t(ga@xmath0 ) positions strong deviation - functional results from model expectation may indicate hybridization of mn d - states with distant neighbors magnetic interactions magnetic behavior of interstitial mn insensitive to position in crystal two implications mn@xmath7-mn@xmath2 pairspair interact with spin holes effects mn@xmath4 and mn@xmath2 compensate our results overcome tight - binding models annihilation of mn@xmath7 due to pairing with mn@xmath2 not restricted to closest pairs in t(ga@xmath15mn ) position works mn atoms close exchange coupling remains antiferromagnetic used fplapw method total energies of supercells simulating configurations mn atoms ground state of mn interstitials is tetrahedral t(as@xmath0 ) position energy of t(ga@xmath0 ) position almost same changes in p - type material t(ga@xmath0 ) position lower energy hexagonal interstitial position higher energy barrier for diffusion mn barrier mobility interstitial mn depends on doping exchange coupling @xmath1 of mn interstitials with holes in valence band for close to value of @xmath1 for substitutional mn not consistent with tight - binding picture mn d - states with nearest neighbors hybridization with distant neighbors important for magnetic interactions mn interstitials attracted to substitutional mn form stable magnetically inactive pairsdensity - functional estimate binding energy pair antiferromagnetic coupling 0.3 ev fits with notion interstitial mn @xcite contrary efficient pairing not restricted to mn@xmath2 t(ga@xmath15mn ) position importance close neighbors properties interstitial mn work done project avoz1 - 010 - 914 ascr financial support by grant agency grant no. a1010214 ) rtn project no . hprn - ct-2000 - 00143 ec acknowledged | 0.431432 | 2,750 | 1,384 | , Saving $0.1 in GPT-4. |
20,692 | one of the types of signals for which the ligo interferometric gravitational wave detectors will search is a stochastic background of gravitational radiation .
we review the technique of searching for a background using the optimally - filtered cross - correlation statistic , and describe the state of plans to perform such cross - correlations between the two ligo interferometers as well as between ligo and other gravitational - wave detectors , in particular the preparation of software to perform such data analysis . | in this section we review the data analysis technique to be used to detect a stochastic background of gravitational radiation .
more details can be found in @xcite .
first , we limit attention to backgrounds which are cosmological in origin and thus can be assumed to be isotropic , unpolarized , gaussian , and stationary .
subject to these assumptions , the stochastic gravitational - wave ( gw ) background is completely described by its power spectrum .
it is conventional to express this spectrum in terms of the gw contribution to the cosmological parameter @xmath0 : @xmath1 note that @xmath2 has been constructed to be dimensionless , and represents the contribution to the overall @xmath3 per _ logarithmic _ frequency interval . in particular , it is _ not _ equivalent to @xmath4 .
note also that since the critical density @xmath5 , which is used in the normalization of @xmath2 , is proportional to the square of the hubble constant @xmath6 @xcite , it is convenient to work with @xmath7 , which is independent of the observationally determined value of @xmath8 .
since a stochastic signal is by definition random , it is impractical to look for one in the output of a single gravitational wave detector @xcite .
however , the effects of such a signal can be detected by cross - correlating the outputs of two independent detectors .
if the output @xmath9 ( @xmath10 ) of the first ( second ) detector consists of a term @xmath11 ( @xmath12 ) due to stochastic gravitational waves and a term @xmath13 ( @xmath14 ) due to instrument noise , and the noise in each instrument is assumed to be uncorrelated both with the gw signal and the noise in the other instrument , the only surviving term in a time averaged correlation @xmath15 is the term @xmath16 due to the gravitational wave background . in practice , one defines a _ cross - correlation statistic _
@xmath17 which is weighted by a filter @xmath18 . in the presence of a stochastic gravitational wave signal , both the mean @xmath19 and variance @xmath20 of the cross - correlation statistic will grow linearly with time , so the signal - to - noise ratio @xmath21 will grow as the square root of the observation time . given a gw background spectrum and a pair of detectors , the signal - to - noise ratio is maximized by using the _ optimal filter _
@xmath22 the denominator , containing the power spectral densities @xmath23 of the noise in the two detectors , serves to suppress the contributions to the cross - correlation statistic from frequencies where one or both detectors are `` noisy '' and most correlations are therefore likely to be accidental .
the numerator represents the average unweighted cross - correlation between the outputs of two detectors , and depends on both the spectrum @xmath2 of the expected gravitational wave background and the locations and orientations of the two detectors .
the latter is described by the _ overlap reduction function _
@xmath24 @xcite .
the overlap reduction function is equal to unity for the case of a pair of interferometers ( ifos ) at the same location with their arms aligned , and is suppressed as the detectors are rotated out of alignment or separated from one another .
the frequency dependence comes about for the following reason : if the wavelength of a wave is comparable to or smaller than the separation between two detectors , the detectors will see different phases of the wave at the same time , and this phase difference will depend on the direction of propagation of the wave . since the stochastic gw background is assumed to be isotropic ,
averaging over different propagation directions suppresses the sensitivity of a pair of detectors to high - frequency waves .
for example , a wave whose wavelength is twice the distance between the two detectors will drive them @xmath25 out of phase if it travels along the line separating them , but _ in _ phase if its direction of propagation is perpendicular to this line .
figure [ fig : overlap ] shows the overlap reduction functions for several detectors of interest .
away from this.,height=336 ] in early data analysis applications , cross - correlated noise is likely to produce considerable spurious contributions to the integral ( [ eq : ccstat ] ) . a useful diagnostic tool will thus be the _ cross - correlation spectrum _
@xmath26 which is simply the integrand of ( [ eq : ccstat ] ) ; this may enable us to see directly the impact of these cross - correlated noise sources on the cross - correlation statistic .
a two - week engineering run for the ligo ifos is planned for around the end of 2001 , involving the kilometer - scale ifos in livingston , louisiana @xcite and hanford , washington @xcite .
the 600 meter geo ifo @xcite in hannover , germany and the allegro @xcite resonant bar detector at louisiana state university are also planning to operate during the same two - week period .
four groups have been formed to use the data taken from ligo and the other instruments to set upper limits on various types of gw signals , including stochastic signals @xcite .
the stochastic group plans to set an upper limit on the strength of a stochastic background , assuming it has the form @xmath27 , with the goal of improving on the existing best upper limit of @xmath28 @xcite .
the roles to be played by various pairs of detectors in this effort are largely driven by their separation and alignment relative to one another , as quantified by the overlap reduction function ( see figure [ fig : overlap ] ) .
the distance between the two ligo sites is approximately 3000 km , which makes the light - travel time between the two sites about 10 ms .
thus the two sites are separated by half a wavelength for waves with a frequency of 50 hz , and as shown in figure [ fig : overlap ] the overlap reduction function first crosses zero at a slightly higher frequency .
so the overlap reduction function limits the sensitivity of this pair of detectors at high frequencies ; below around 40 hz , the seismic noise in the detectors will squash the optimal filter ( [ eq : optfilt ] ) .
the net effect ( as illustrated in figure 21 of @xcite ) is that for ligo initial design sensitivity , most of the support of the optimal filter lies between 50 and 250 hz .
the allegro bar detector is far closer to the ligo livingston site than the ligo hanford site is , with only about 40 km separating the two and a `` half - wavelength '' frequency of 3750 hz . thus the observing geometry allows for observations of correlations out to much higher frequencies . on the other hand ,
the sensitivity of allegro is concentrated in two narrow frequency bands in the vicinity of 900 hz , so correlations between allegro and ligo livingston probe a different part of the frequency domain than correlations between the two ligo detectors .
since the geo-600 detector is rather distant from the ligo sites ( over 7500 km from livingston , corresponding to a `` half - wavelength '' frequency of only 20 hz ) , the small overlap reduction function will render the geo-600/ligo - la pair ( the better of the two ) considerably less sensitive to gravitational waves than the ligo - la / ligo - wa pair .
the primary interest in performing this correlation is thus the information it will provide about cross - correlated noise rather than a contribution to the upper limit on stochastic background strength .
the data analysis technique described in section [ sec : techniques ] will be implemented within the ligo data analysis system ( ldas ) @xcite using c routines from the ligo numerical algorithms library ( lal ) @xcite .
we have written and tested lal routines to perform the various parts of the analysis ( calculating the overlap reduction function , constructing the optimal filter , etc . ) .
care has been taken to make these routines general enough to be applied to both ifo data ( e.g. , from ligo and geo ) and data from resonant bar detectors such as allegro .
two major issues have required some care in this regard : first , the treatment of detector geometry ( used in constructing the optimal filter ) needed to be general enough to describe both interferometric and resonant detectors .
this was accomplished by defining a data structure within lal which described an idealized earthbound detector in terms of its location and tensor response to gravitational waves @xcite .
second , since the sensitivity band of allegro is at a relatively high frequency compared to its bandwidth , its gravitational wave signal is heterodyned before being discretely sampled @xcite . by multiplying the time - domain signal by a complex exponential oscillating at a base frequency , one effectively shifts the frequency band represented in the discrete signal ( whose full width is equal to the sampling frequency ) so that it is centered at the base frequency rather than at dc ( 0 hz ) . to allow for this , the lal routines had to be written to deal with complex as well as real time series .
the interface between the c@xmath29 ldas environment and the lal c library is known as lalwrapper @xcite .
lalwrapper contains a number of dynamically - linked shared objects which can be used to `` drive '' various search algorithms in lal .
we have written two lalwrapper shared objects : ` libldasstochastic.so ` calculates the cross - correlation spectrum ( [ eq : ccspec ] ) between two interferometric detectors , and ` libldasstochasticbar.so ` does the same for correlations between an ifo and a resonant bar detector .
( eventually , we plan to integrate the functionality into a single , generalized search engine . )
the general behavior of lalwrapper code is to execute parallel searches on one or more nodes of a beowulf cluster @xcite , with each `` slave '' node reporting to the `` master '' node at least ten times .
both stochastic background search engines use the following algorithm to calculate cross - correlation spectra for a sequence of short consecutive time intervals : 1 .
equal - length stretches of data from a pair of detectors are input , along with power spectra and response functions , and some search parameters
the data streams are each divided into ten or more shorter - length segments .
an optimal filter is constructed using the auxiliary inputs and parameters describing the choice of detectors , etc .
4 . in turn , each corresponding pair of data segments is fourier - transformed and the cross - correlation spectrum calculated using the optimal filter .
enhancements to be made for the scientific data runs which will begin in 2002 include : ( i ) rather than constructing a single optimal filter based on the @xmath27 model , we will choose a set of points in the parameter space of stochastic background models@xcite and filter the data in parallel with a `` grid '' of optimal filters , one optimized according to ( [ eq : optfilt ] ) for each @xmath2 model @xcite ; ( ii ) to set an upper limit or make a measurement of the strength of the stochastic background , we will calculate the cross - correlation statistic for the whole data stream(s ) by integrating the cross - correlation spectrum for a given segment over frequency and then adding the contributions from all the segments .
the shared objects ` libldasstochastic.so ` and ` libldasstochasticbar.so ` were among the data analysis routines tested at the burst - stochastic mock data challenge@xcite , held at the massachusetts institute of technology 4 - 10 september 2001 .
the shared object for ifo - ifo correlations was tested with trivial and non - trivial synthetic data and produced the expected results in each case ; it was also used to analyze 15 minutes of data taken by the two ligo ifos during a recent engineering run , to verify that it could do so without failing .
the ifo - bar shared object , being slightly less mature , was not tested as extensively .
however , some of the tests which were run produced unexpected results which appear to be due to the use of single precision arithmetic in several lal routines .
the finer frequency resolution required by the sharp spectral features in the bar response function may require the use of double precision .
we would like to thank our colleagues in ligo and the larger gravitational wave community who have made this work possible , especially : b. allen and the authors of the grasp routines from which many lal routines are derived ; s. drasco and .
flanagan , who wrote early versions of several lal stochastic background routines ; lsc software cordinator a. wiseman and lal librarian j. creighton , without whom lal would not exist ; the other members of the site structure development team including p. brady , d. chin , j. creighton , c. cutler , k. riles and a. lazzarini , who came up with the idea of using the response tensor to describe a generic gw detector ; our fellow participants in the burst - stochastic mdc including a. searle on the stochastic side and s. finn , e. daw , s. marka , p. saulson and j. zweizig on the burst side and especially e. katsavouinidis and j. sylvestre for hosting and k. blackburn , p. charlton , p. shawhan , m. barnes , p. ehrens , m. lei and i. salzman for providing ldas support ; and finally the other members of the stochastic sources upper limits group , including s. bose , n. chistensen , r. drever , p. fritschel , j. giaime , w. hamilton , w. johnson , m. landry , t. nash , a. ottewill , b. whiting and r. weiss .
this work was supported by the national science foundation under grants phy-9981795 ( utb ) and phy-9970742 ( lsu ) , and by nasa contract jpl1219731 ( utb ) .
figure [ fig : overlap ] was made using slightly modified versions of lal routines @xcite .
9 allen b 1997 in _ proceedings of the les houches school on astrophysical sources of gravitational waves _ , eds marck j a and lasota j p , cambridge , 373 allen b and romano j d 1999 102001 ; allen b and romano j d 1997 _ preprint _ gr - qc/9710117 kolb e w and turner m s 1990 _ the early universe _ ( addison - wesley , 1990 ) christensen n 1992 5250 flanagan 1993 2389 flanagan 1993 _ preprint _ astro - ph/9305029 http://www.ligo-la.caltech.edu/ http://www.ligo-wa.caltech.edu/ http://www.geo600.uni-hannover.de/ http://gravity.phys.lsu.edu/allegro/ allen b et al 2001 ligo technical document ligo - t010017-z http://feynman.utb.edu/ joe / research / stochastic / upperlimits/ astone p , ferrari v , maggiore m and romano j d 2000 _ int . j. mod . phys._*d9 * 361 finn l s and lazzarini a _
gr - qc/0104040 http://www.ldas-sw.ligo.caltech.edu/ http://www.lsc-group.phys.uwm.edu/lal/ allen b et al 1999 ligo technical document ligo - t990030-e lal software documentation , version 0.9 , section 9.2 ; available from @xcite mauceli e et al 1996 1264 mauceli e et al 1996 _ preprint _ gr - qc/9609058 blackburn k 1999 ligo technical document ligo - t990097-e http://www.lsc-group.phys.uwm.edu/mpigroup/ http://www.beowulf.org/ ungarelli c and vecchio a 2001 in preparation maggiore m 2000 _ phys .
rep._*331 * 28 maggiore 1999 _ preprint _ gr - qc/9909001 authors tbd 2001 ligo technical document number tbd entitled `` burst / stochastic mock data challenge '' | section review data analysis technique detect stochastic background gravitational radiation details in @xcite limit attention to backgrounds cosmological assumed isotropic unpolarized gaussian stationary stochastic gravitational - wave gw background described by power spectrum conventional express spectrum gw contribution to cosmological parameter @xmath0 : @xmath1 @xmath2 dimensionless represents contribution to @xmath3 per logarithmic frequency interval not equivalent to @xmath4 critical density @xmath5 normalization @xmath2 proportional to hubble constant @xmath6 @xcite convenient to work with @xmath7 independent of observationally value @xmath8 stochastic signal random impractical to in output single gravitational wave detector @xcite effects detected by cross - correlating outputs of two independent detectors if output @xmath9 first detector consists term @xmath11 ) due to stochastic gravitational waves due to instrument noise noise uncorrelated with gw signal noise only surviving term averaged correlation is term @xmath16 due to gravitational wave background defines cross - correlation statistic _ @xmath17 weighted by filter @xmath18stochastic gravitational wave signal mean @xmath19 variance @xmath20 cross - correlation statistic grow with time signal - to - noise ratio @xmath21 as square root observation time gw background spectrum pair of detectors signal - to - noise ratio maximized optimal filter @xmath22 denominator power spectral densities of noise in two detectors contributions cross - correlation from frequencies where detectors noisy correlations accidental numerator represents average unweighted cross - correlation between outputs two detectors depends on spectrum gravitational wave background locations orientations detectors described by overlap reduction function _ @xmath24 equal to unity for interferometers suppressed as detectors rotated of or separated frequency dependence if wavelength wave comparable to or smaller than separation between two detectors detectors see different phases wave phase difference on direction propagation wave stochastic gw background isotropic averaging over different propagation directions suppresses sensitivity detectors to high - frequency waves wave wavelength twice distance between two detectors drive them out of phase if along line separating in phase if perpendicular to linefigure overlap shows overlap reduction functions for detectors early data analysis cross - correlated noise likely contributions to integral ccstat useful diagnostic tool cross - correlation spectrum _ @xmath26 integrand of ccstat impact of cross - correlated noise sources on cross - correlation statistic two - week engineering run for ligo ifos planned end of 2001 kilometer - scale ifos in livingston louisiana hanford washington 600 meter geo ifo in hannover germany allegro @xcite resonant bar detector at louisiana state university operate four groups formed to use data from ligo set upper limits on gw signals including stochastic signals stochastic group plans to set upper limit on strength of stochastic background goal improving on best upper limit of @xmath28 @xcite roles by detectors driven by separation alignment quantified by overlap reduction function distance between two ligo sites approximately 3000 km light - travel time about 10 ms sites separated by half a wavelength for waves frequency 50 hz overlap reduction function crosses zero at higher frequencyoverlap reduction function limits sensitivity detectors at high frequencies below 40 hz seismic noise optimal filter in figure 21 for ligo initial sensitivity support optimal filter between 50 and 250 hz allegro bar detector closer to ligo livingston site than hanford site 40 km separating and half - wavelength frequency 3750 hz observing geometry allows correlations to higher frequencies sensitivity of allegro concentrated in two narrow frequency bands in 900 hz correlations between allegro and ligo livingston probe different part frequency domain than between two ligo detectors geo-600 detector distant from ligo sites over 7500 km from half - wavelength frequency 20 hz small overlap reduction function geo-600/ligo - la pair less sensitive to gravitational waves than primary interest in correlation is information about cross - correlated noise than contribution to upper limit stochastic background strength data analysis technique in section implemented within ligo data analysis system using c routines from ligo numerical algorithms library written and tested lal routines analysis overlap reduction function optimal filter routines general to both ifo datafrom ligo geo data from resonant bar detectors allegro two major issues required care treatment of detector geometry optimal filter needed interferometric resonant detectors accomplished by defining data structure within lal idealized earthbound detector location tensor response to gravitational waves sensitivity band of allegro high frequency gravitational wave signal heterodyned before sampled multiplying time - domain signal by complex exponential at base frequency shifts frequency band discrete signal sampling centered at base frequency lal routines deal with complex real time series interface between c@xmath29 ldas environment lal c library as lalwrapper @xcite lalwrapper contains dynamically - linked shared objects drive search algorithms in lal written two lalwrapper shared objects libldasstochastic.so ` calculates cross - correlation spectrum between two interferometric detectors libldasstochasticbar.so ` for correlations between ifo resonant bar detector plan to integrate functionality into single generalized search engine lalwrapper code parallel searches on nodes of beowulf cluster @xcite each slave node reporting to master node ten timesstochastic background search engines use algorithm cross - correlation spectra for short time intervals equal - length data from detectors input power spectra response functions search parameters data streams divided into shorter - length segments optimal filter constructed using auxiliary inputs parameters detectors pair data segments fourier - transformed cross - correlation spectrum calculated using optimal filter enhancements for scientific data runs 2002 include single optimal filter choose points in stochastic background models filter data parallel with grid of optimal filters for each @xmath2 model stochastic background calculate cross - correlation statistic for data stream by integrating cross - correlation spectrum for segment adding contributions from all segments shared objects libldasstochastic.so and libldasstochasticbar.so tested at burst - stochastic mock data challenge@xcite massachusetts institute of technology 4 - 10 september 2001 shared object for ifo - ifo correlations tested with trivial non - trivial synthetic data produced expected results used to analyze 15 minutes data by two ligo ifos during engineering run ifo - bar shared object less mature not tested extensivelytests produced unexpected results due to single precision arithmetic in lal routines finer frequency resolution bar response may require double precision colleagues in ligo gravitational wave community b. allen grasp routines s. drasco lal stochastic routines lsc software cordinator a. wiseman lal librarian j. creighton lal site structure development team p. brady d. chin j. creighton c. cutler k. riles a. lazzarini response tensor generic gw detector participants burst - stochastic mdc a. searle s. finn e. daw s marka p. saulson j zweizig e. katsavouinidis j. sylvestre k. blackburn p. charlton shawhan m. barnes ehrens m. lei i. salzman support members stochastic sources upper limits group s. bose n. chistensen r. drever p. fritschel j. giaime w hamilton johnson m. landry t. nash a. ottewill b. whiting r. weiss work supported by national science foundation grants phy-9981795 phy-9970742 nasa contract jpl1219731figure overlap modified lal routines allen b 1997 proceedings les houches school astrophysical sources gravitational waves marck j lasota j cambridge 373 allen romano j 1999 102001 1997 qc/9710117 kolb e turner m 1990 early universe addison wesley christensen 1992 5250 flanagan 1993 astro - ph/9305029-la.caltech.uni-hannover.de.phys.lsu allen b 2001 ligo technical document t010017-z.utb.edu research stochastic upperlimits astone p ferrari v maggiore m romano j 2000. 361 finn l s lazzarini qc/0104040-sw.ligo.caltech-group.phys.uwm.edu allen b 1999 document t990030-e software documentation version 0.9 section 9.2 @xcite mauceli e 1996 1264 preprint qc/9609058 blackburn k 1999 ligo document t990097-e-group.phys.uwm.edu.beowulf.org ungarelli c vecchio a 2001 maggiore m 2000 rep28 1999 preprint qc/9909001 authors tbd 2001 document burst stochastic mock data challenge | 0.566504 | 3,502 | 1,738 | , Saving $0.1 in GPT-4. |
20,695 | with the anticipated launch of glast , the existing x - ray telescopes , and the enhanced capabilities of the new generation of tev telescopes , developing tools for modeling the variability of high energy sources such as blazars is becoming a high priority .
we point out the serious , innate problems one zone synchrotron - self compton models have in simulating high energy variability .
we then present the first steps toward a multi zone model where non - local , time delayed synchrotron - self compton electron energy losses are taken into account . by introducing only one additional parameter , the length of the system , our code can simulate variability properly at compton dominated stages , a situation typical of flaring systems . as a first application , we were able to reproduce variability similar to that observed in the case of the puzzling ` orphan ' tev flares that are not accompanied by a corresponding x - ray flare .
address = department of physics , university of maryland , baltimore county address = department of physics , university of maryland , baltimore county , altaddress = nasa , goddard space flight center address = department of physics and space sciences , florida institute of technology address = nasa , goddard space flight center | in blazars , radio loud active galaxies with their relativistic jets pointing close to the line of sight , the observed radiation is dominated by relativistically beamed emission from the sub - pc base of the jet . due to the small size of the emitting region and the large distance of such sources ,
currently it is not possible to spatially resolve the emitting region .
because of this , we can only obtain information about its structure through multiwavelength variability studies . when modeling blazar radiation , past models have mostly incorporated some form of a one zone model .
such models are limited because high - energy electrons cool faster than the source light crossing time and also because time delay effects must be considered .
when neither of these are accounted for , the model produces variability on timescales less than the light crossing time , which is incorrect .
one must consider the time that it takes for light to be transmitted from different parts of the source , and as chiaberge and ghisellini ( 1999 ) showed , the shortest observable variability that can be trusted is the light crossing time of the zone .
the basic limitation of one zone models stems from the fact that , by construction , all the high energy variability is produced by electrons with cooling times faster than the light crossing time .
one , therefore , can not use one zone models to infer the source structure from high energy variability .
kirk et al . ( 1998 ) developed an analytical model in which electrons , after being accelerated , propagate in a pipe geometry and cool through synchrotron radiation .
this results in a frequency dependent source size , in agreement with the fact that the variability timescale of synchrotron radiation increases with decreasing frequency . here
we present a numerical code that in addition incorporates the important inverse compton ( ic ) process , and in particular the most complicated synchrotron - self compton ( ssc ) losses and emissivity .
we assume that a power law of relativistic electrons is injected at the base of a flow , and that the electrons flow downstream and cool radiatively .
variations in the injected plasma parameters propagate downstream and manifest themselves as frequency dependent variability .
synchrotron and ic losses from photons external to the source are local processes , in the sense that at a given point in the flow , the energy loss rate only depends on the magnetic field and external photon field energy densities , and not on the conditions throughout the source .
this is not the case with ssc losses , because synchrotron photons produced throughout the source in past times - to take into account the light travel time from one point of the source to another - contribute to the photon energy density responsible for the ssc losses and emissivity at a given point and time in the source . to incorporate this , we record the synchrotron emissivity throughout the source as a function of time
in most cases , tev and x - ray variability are correlated ( e.g. fossati et al . 2000 ) .
a variability pattern that can not be explained by one zone models is the so - called orphan flares , tev flares that are not accompanied by x - ray flares ( e.g. krawczynski et al .
2004 , blazejowski et .
al . 2005 ) .
this can not be simulated by one zone models because the same electrons that produce x - rays through synchrotron raditation also ic - scatter lower energy photons to produce tev emission .
it follows then , that flares in these spectra would be linked .
we present ( fig .
1 ) two types of orphan flares produced by our model using a ` pipe ' flow geometry : a tev flare unaccompanied by an x - ray flare and a combined flare followed by an `` echo '' flare in the tev range . we produce both flares with the system being in or near a compton - dominated state , consistent with observations .
the first type of flare occurs when there is an additional injection of low - energy electrons ; through synchrotron radiation , these provide additional low energy seed photons for ic radiation , resulting in the tev flare .
the dip in x - ray radiation is due to the additional ic cooling of the high - energy electrons .
the second type of flare occurs in the case of an increase in high - energy electron injection .
the system responds with an initial flare in both x - ray and tev energies .
as the high - energy electrons travel down the jet and cool , they emit extra radiation in the ir .
this travels back toward the base of the jet and allows for an increase in tev emission by providing additional seed photons .
this project is part of the senior thesis of philip graff , under the supervision of markos georganopoulos .
the authors acknowledge support from a chandra theory grant and a nasa long - term space astrophysics grant . | blazars active galaxies relativistic jets close to sight observed radiation dominated by relativistically beamed emission from sub - pc base jet to small size region large distance sources possible to spatially resolve region obtain information structure through multiwavelength variability studies modeling blazar radiation past models incorporated one zone model models limited high - energy electrons cool faster than source light crossing time time delay effects model produces variability less than light crossing time incorrect time for light from source shortest observable variability trusted is light crossing time zone limitation of one zone models high energy variability produced by electrons with cooling times faster than light crossing time use to infer source structure from high energy variability kirk et al. 1998 developed model electrons accelerated propagate pipe cool through synchrotron radiation results in frequency dependent source size variability timescale of synchrotron radiation increases with decreasing frequency numerical code incorporates inverse compton ( ) process complicated synchrotron - self compton ) losses emissivityassume power law of relativistic electrons injected base flow electrons flow downstream cool radiatively variations in injected plasma parameters propagate downstream manifest as frequency dependent variability synchrotron ic losses from photons external to source local processes energy loss rate depends on magnetic field external photon field energy densities not conditions source not case with ssc losses synchrotron photons produced contribute to photon energy density responsible for ssc losses emissivity record synchrotron emissivity as function of time tev x - ray variability correlated variability pattern not explained one zone models orphan flares tev flares not accompanied by x - ray flares not simulated electrons produce x - rays ic - scatter lower energy photons produce tev emission flares in spectra linked present fig . 1 ) two types of orphan flares produced model flow geometry : tev flare unaccompanied by x - ray flare combined flare followed by echo '' flare in tev range produce both flares with system in or near compton - dominated state consistent with observationsfirst flare occurs additional injection low - energy electrons provide seed photons ic radiation tev flare dip x - ray radiation due additional ic cooling high - energy electrons second flare increase high - energy electron injection system responds initial flare x - ray tev energies high - energy electrons travel cool emit extra radiation travels increase tev emission additional seed photons project senior thesis philip graff supervision markos georganopoulos authors support chandra theory grant nasa long - term space astrophysics grant | 0.183057 | 1,013 | 480 | , Saving $0.0 in GPT-4. |
20,699 | high - resolution data of online chats are studied as a physical system in laboratory in order to quantify collective behavior of users .
our analysis reveals strong regularities characteristic to natural systems with additional features .
in particular , we find self - organized dynamics with long - range correlations in user actions and persistent associations among users that have the properties of a social network .
furthermore , the evolution of the graph and its architecture with specific k - core structure are shown to be related with the type and the emotion arousal of exchanged messages .
partitioning of the graph by deletion of the links which carry high arousal messages exhibits critical fluctuations at the percolation threshold . | quantitative analysis of human collective dynamics has recently become available based on the high - resolution empirical data from online communication systems , which can be studied as complex dynamical systems in physics laboratory .
the self - organized dynamics , common to the online interactions , is crucial for the emergence of collective behaviors of users @xcite . the kind of social structures built in the dynamics , however , may depend on the technology features of the communication platform and other details such as visibility of user - to - user messages .
such details may affect actions of individual users and thus influence the course of events .
for instance , the dynamics of message exchange along `` friendship '' links in the social network ` myspace ` has been shown @xcite to yield structures much different from the conventional social networks @xcite . apart from the online social networks , the online games @xcite , blogs , diggs , forums @xcite etc , can also lead to recognizable user associations . in the online games @xcite , for example
, the users can mark friend / enemy relationship or undertake collective actions towards other users . on the other hand , indirect interactions on blogs provide hidden mechanisms in which the subjects of the posts and negative emotion ( critique ) dominate , leading to user communities centered around certain popular posts @xcite . furthermore , frequency of interactions , subject of the communication , and amount of emotion conveyed in these messages may become critical for collective dynamics and for building specific associations among users over time .
these aspects of the online communication dynamics represent a major challenge for quantitative analysis and theoretical modeling .
ideal empirical systems where this can be studied are the irc ( internet - relay - chat ) channels , where no _ a priori _ relationship among users exists . in this paper
we combine statistical physics with computer science methods of text analysis to perform a quantitative analysis of human collective behavior in online chats . a large dataset from irc ` ubuntu channel `
is analysed considering user - to - user communications with full text of messages and assessing emotion contents in the text .
in addition to humans , the data contains a web robot ( bot ) , which serves predefined text messages upon users request . by mapping the data onto directed weighted network , and analyzing its topology in connection with temporal patterns of user actions , we find that a new type of techno - social structure emerges , which persists for a long time .
we further analyze its architecture by testing the validity of the `` social '' hypothesis and study the percolation transition that is connected with the amount of emotional arousal on the links .
both the graph architecture and its social ties suggest that a specific type of online social network is assembled based on the emotion - carrying communications .
_ data structure .
_ we consider data from the publicly accessible irc channel related to the development of the ` ubuntu ` operating system @xcite .
the data contain both humans and a bot .
users , identified by ids , typically exchange short text messages in seeking information on software or services , joining group discussions , or conducting open - domain chats .
the data considered in this study are for one year period ( year 2009 ) , and contain texts of messages and time resolution of one minute .
after retrieval of all the logs for a given time window , anonymization by substituting user ids by random number references , and removing the spam , the utterances are annotated with appropriate sentiment detecting tools .
in particular , in this study we use the lexicon - based annotation , by which emotion carrying words in a message are detected @xcite .
such words carry emotional arousal
degree of reactivity , and valence pleasure or displeasure , listed in the range @xmath0 in the emotional dictionary @xcite , from which then the average arousal and valence are computed for the whole message .
moreover , the content of each message is determined according to one ( or more ) dialog act classes . in total , twelve such classes can be determined @xcite .
for instance , three classes that we consider here _ ( yes - no question , why - question , statement ) _ , in common `` question''-type messages , appear to be characteristic to a certain group of users .
the data are filtered to identify unique user ids ( the i d labeled by `` 35177 '' belonging to the bot is treated equally ) . the subset where user - to - user communications are clearly identified is selected and mapped onto a directed network .
the network nodes @xmath1 represent users and the directed link @xmath2 indicates that at least one message from user @xmath3 to user @xmath4 occurred within the considered time window .
multiple messages along the link increase its width , while the message emotion and message category are considered as _ properties _ of that link . starting from the beginning of the dataset
, the network evolves in time by addition of new users and new links and increasing the widths of the existing links .
[ fig - net - evolution ] shows how the network grows within first three days .
after one year ( data limit ) the network consists of @xmath5 85185 users with the link density @xmath6 and link reciprocity @xmath7 . cc in the temporal patterns of user actions we noticed that some users appear for a short time and disappear from the channel , while others are constantly or occasionally active within a long time period .
we examine the _ lifetime of links _ that the users establish between each other .
the distribution of the lifetimes of all links in the dataset is shown in fig .
[ fig - lifetime - links ] .
the three curves are , respectively , for all kinds of links , and for the links that carry an overall positive and negative emotion valence . apart from the frequency
, the patterns are similar : for lifetimes shorter than one day , the distribution decays faster than for the lifetimes longer than one day .
the power - law with a small exponent of the distribution in the latter region suggests that the links that survive the first day after the appearance are likely to persist for a longer time ( till data limit ) . the network that contains only such persistent links , here termed the _ persistent network _ , or `
ubunetp ` , is of our interest and can be considered as a potential social structure .
thus , we remove the links which do not survive over the first day after their appearance . note that in this way a number of users are also removed .
( the user survival distribution follows a similar pattern , not shown ) . in the following we explore the emergent network with the persistent links and
study how the above mentioned link properties , in particular the message types and their emotion contents , are related with the network architecture .
cc + cc these network connections emerge through a self - organized dynamics , in which users play different functions and over time they `` settle '' in the evolving environment .
the cooperative behavior is quantified by analysis of correlated fluctuations in the time series of user actions .
specifically , we construct the time series of the number of messages per one minute time bin . in fig .
[ fig - timeseries ] an example of the time series is shown in the bottom panel .
for better view , only a small part corresponding to eight days is shown , while the length of the considered time series is @xmath8 bins .
the power spectrum of the time series is given in the top panel in fig .
[ fig - timeseries ] .
the inset shows the fluctuations @xmath9 , scaled by the standard deviation @xmath10 , of the integrated time series @xmath11 $ ] as a function of varying time window @xmath12 .
the slope determines the hurst exponent @xmath13 .
apart from the peak corresponding to daily periodicity , the spectrum shows long - range correlations of the gaussian noise type with @xmath14 in a wide range of frequency index @xmath15 , which is indicated by the straight line .
the correlation range , considered in the time domain , agrees well with the persistence time window of the fluctuation , i.e. , @xmath16 $ ] .
note also that the values of the exponents @xmath17 and @xmath18 satisfy ( within error bars ) the scaling relation @xmath19 expected for the colored noise signals .
furthermore , the value of the hurst exponent @xmath20 suggests that a _ persistent _ type of fluctuations occurs in the system as a whole .
t note that the hurst exponent in this range is also found for the time series of messages carrying emotional content along a particular link or by a particular user @xcite .
the network with persistent links is further examined as a `` social '' structure .
we recall that `` weak - ties '' hypothesis has recently been confirmed in the data of mobile phone , online games , and ` myspace ` dialogs networks @xcite . in analogy to the social networks ,
the test is performed by computing the overlap @xmath21the number of common neighbors of two nodes which share a link , as a function of the width @xmath22 and of the betweenness centrality @xmath23 of that link . in particular , the overlap increases with the intensity of communications @xmath22 between the nodes as @xmath24 , while it decreases with the centrality of the links as @xmath25 .
both laws hold in the above mentioned online systems , all of which have a typical _ community structure_. the corresponding exponents are close to the universal values @xmath26 and @xmath27 , which are recently conjectured in @xcite .
cc + + in the emergent networks of chats , the situation is somewhat different in that no classical community structure is found .
moreover , the network shows hierarchical organization of links with a large `` core '' , but the links between nodes at different hierarchy levels also occur .
thus the network remains connected even if the central core is removed @xcite .
consequently , we find that the overlap has a different dependence of betweenness centrality , as shown in fig .
[ fig - ubu - overlaps ] .
the inset shows the overlap as function of @xmath22 .
the three curves are for our persistent network ` ubunetp ` , and for two subnetworks described below .
one subnetwork is obtained by removing the bot and all its links .
the other is a reduced network shown in the top fig .
[ fig - ubu - overlaps ] , whose structure reflects the type of messages that the users exchange : the outer layer ( blue nodes ) represents the users who send question - type of messages twice more often than the average frequency on the channel , and no other message types .
they are attached to the central core ( red nodes ) , which consists of the bot and the users who frequently use all types of messages .
qualitatively , the `` social ties '' behaviors are present , but the corresponding exponents are twice larger , @xmath28 and @xmath29 , except for a small parts of the curves ( marked by the red dashed line ) , which agree with the other online social systems . the origin of such behavior as well as the sudden increase of the overlap at high @xmath23 values needs theoretical modeling and analysis of more similar systems , which is left out of the present work .
it is interesting to note that qualitatively same behavior is found in the network from which the bot is removed .
ccc + the chat network has a structure with a power - law distribution for in- and out - degree with a similar exponent @xmath30 and disassortative mixing .
a detailed analysis is left for a separate study @xcite .
hereafter we focus on the resilience of the persistent chats network by analyzing the _ percolation transition _ that occurs in relation to the arousal of the messages carried along the directed links .
the rationale is that the links persist over a long time due to certain attribute ( or `` excitement '' , properly quantified by the emotional arousal ) of the messages exchanged along them .
it is worth noting that the weights and the averaged arousal of the links obey a power - law and a skewed gaussian distribution , respectively @xcite .
starting from the whole persistent network ( with @xmath31 users and @xmath32 links ) we cut the directed links whose average arousal exceeds a threshold value , i.e. , @xmath33 .
for convenience it is normalized by @xmath34 , the largest average arousal found on the network .
note that , depending on the actual value of the threshold @xmath35 , the number of links that satisfy the condition as well as their position on the network can vary considerably .
together with the network topology , directedness , and reciprocity of the links , this fact of the content - based topology determines the features of the percolation transition @xcite . for each threshold
@xmath35 the largest cluster ( as weakly connected component ) is found and its size @xmath36 is plotted against the threshold @xmath35 in fig .
[ fig - ubu - arousal - percolation ] . also shown is the susceptibility @xmath37 , which is defined as @xmath38 .
it determines the fluctuations of the cluster size in response to cutting the links with a given arousal .
the distribution of cluster size satisfies the power law @xmath39 . in agreement with the percolation theory ,
@xmath37 exhibits a peak at the transition point @xmath40 , where the @xmath36 tends to vanish .
what kind of geometry and/or complexity change @xcite in the network structure occurs at the transition point @xmath41 ?
we examine the fraction of retained links @xmath42 , and the average degree @xmath43 of the giant cluster , for the range of values of @xmath44 above the transition point . in the inset of fig .
[ fig - ubu - arousal - percolation ] we show the scaling region @xmath45 , where the index indicates @xmath46 . the critical value @xmath47 and @xmath48 .
the exponent @xmath49 is close to one , i.e. , @xmath50 , @xmath51 , @xmath52 for @xmath53 , respectively .
applying physics methods we have studied the self - organized dynamics of user - to - user interactions in the online chats , and have shown that both the type of the messages and their emotional arousal play an essential role in building a persistent techno - social structure . the social graph is hierarchically organized having a central core with some very active users and the web bot .
in contrast to more familiar user grouping into communities , this kind of spontaneous organization marks a new class of structures in the zoo of emergent online communication networks . the observed percolation transition on this graph
is smooth with the enhanced fluctuations in response to a targeted deletion of links which carry the emotional arousal over a threshold value .
our analysis reveals that the contents of the exchanged messages and the role that users assume in the dynamics need to be considered as an integral parts of the topology in the online social graphs . in general ,
the idea to consider features beyond conventional interactions , that we pursued here for the analysis of human online interactions , can also be useful for studying physical systems which are assembled from complex molecules or other objects in laboratory .
* acknowledgments : * we thank the support from the program p1 - 0044 by the research agency of the republic of slovenia and from the european community s program fp7-ict-2008 - 3 under grant n@xmath54 231323 .
m. mitrovi and b. tadi . .
in handbook of optimization in complex networks theory and applications , eds .
m.t . thai and p. pardalos , springer `` optimization and its applications '' , vol.57 , part 2 , pp . 209227 , doi:10.1007/978 - 1 - 4614 - 0754 - 6@xmath558 , january 2012 , isbn 978 - 0 - 19 - 953304 - 6 j. onela , j. saramaki , j. hyvnen , g. szabo , m. de menezes , k. kaski , a .-
barabsi , and j. kertesz . , 9(6):176 , 2007 .
considered as an undirected graph , the size of @xmath56-core is found to decreases as power @xmath57 .
the strongest core is @xmath58 , which also contains the bot . | quantitative analysis of human collective dynamics available based on high - resolution data from online communication systems studied complex in self - organized dynamics common to online interactions crucial for collective behaviors social structures may depend on technology features communication platform visibility of messages details affect actions influence course events dynamics of message exchange friendship links in myspace yield structures different from conventional social networks online games blogs diggs forums can lead to recognizable user associations in online games users can mark friend / enemy relationship or undertake collective actions indirect interactions on blogs provide hidden mechanisms subjects negative emotion critique dominate to user communities centered around popular posts frequency of interactions subject communication emotion conveyed in messages critical for collective dynamics building associations over time online communication dynamics represent challenge for quantitative analysis theoretical modeling ideal empirical systems are irc channels no a priori relationship users in paper statistical physics with computer science methods text analysis quantitative analysis of human collective behavior in online chats large dataset from irc ` ubuntu channel ` analysed considering communications with full text emotion contents textto humans data contains web robot bot ) serves predefined text messages request mapping data directed weighted network analyzing topology temporal patterns user actions new techno - social structure emerges persists long time analyze architecture validity `` social '' hypothesis study percolation transition connected with emotional arousal on links graph architecture social ties suggest specific online social network based on emotion - carrying communications data structure data from irc channel ubuntu operating system data contain humans and bot users identified by ids exchange short text messages seeking information software services joining group discussions conducting open - domain chats data for one year period 2009 contain texts messages time resolution one minute after retrieval logs anonymization user ids random number removing spam utterances annotated with sentiment detecting tools use lexicon - based annotation emotion carrying words detected words carry emotional arousal reactivity valence pleasure displeasure listed in range emotional dictionary average arousal valence computed for whole message content each message determined according to dialog act classes twelve classes determinedthree classes yes - no question why - question statement in common messages characteristic to certain group users data filtered to identify unique user ids labeled 35177 bot treated equally subset where user - to - user communications identified selected mapped onto directed network network nodes @xmath1 represent users directed link @xmath2 indicates one message from @xmath3 to @xmath4 occurred within time window multiple messages link increase width message emotion category properties of link network evolves by new users links increasing widths existing fig - net - evolution shows network grows first three days after one year network consists of @xmath5 85185 users link density @xmath6 link reciprocity @xmath7 temporal patterns some users appear short time disappear others constantly or active long examine _ lifetime of links establish distribution of lifetimes links shown in fig . lifetime - links three curves for all links links positive negative emotion valence patterns similar for lifetimes shorter than one day distribution decays faster than longer than one day power - law small exponent suggests links survive first day after likely persist longer till data limitnetwork contains persistent links persistent network ubunetp interest potential social structure remove links survive first day after appearance users removed user survival distribution follows similar pattern explore emergent network with persistent links study link properties message types emotion contents with network architecture network connections emerge through self - organized dynamics users play different functions settle in evolving environment cooperative behavior quantified by analysis correlated fluctuations in time series user actions construct time series of number of messages per one minute time bin fig . [ - timeseries ] example time series bottom panel small part to eight days shown length of time series is @xmath8 bins power spectrum time series top panel inset shows fluctuations @xmath9 scaled by standard deviation @xmath10 integrated time series @xmath11 function of varying time window @xmath12 slope determines hurst exponent @xmath13 spectrum shows long - range correlations of gaussian noise type with @xmath14 frequency index @xmath15 indicated by straight line correlation range agrees with persistence time window of fluctuation @xmath16 $values exponents @xmath17 and @xmath18 satisfy scaling relation @xmath19 for colored noise signals value hurst exponent @xmath20 suggests persistent fluctuations in system hurst exponent found for time series of messages carrying emotional content particular link user network with persistent links examined as social structure `` weak - ties '' hypothesis confirmed in data mobile phone online games myspace dialogs networks test by computing overlap @xmath21the common neighbors of two nodes link as function of width @xmath22 betweenness centrality @xmath23 of link overlap increases with intensity of communications nodes decreases with centrality of links @xmath25 laws hold in online systems typical community structure exponents close to universal values @xmath26 and @xmath27 conjectured in @xcite in emergent networks of chats different no classical community structure found network shows hierarchical organization of links with large core links between nodes at different hierarchy levels occur network remains connected even if central core removed overlap of betweenness centrality shown in fig . shows overlap as function of @xmath22three curves for persistent network ubunetp ` two subnetworks one subnetwork obtained by removing bot links other reduced network in top fig . ubu overlaps structure reflects type messages users exchange outer layer ( blue nodes represents users send question - type messages twice often average no other message types attached to central core ( red nodes consists of bot users use all types messages social ties behaviors present exponents twice larger @xmath28 @xmath29 except small parts of curves red dashed line agree with other online social systems origin behavior sudden increase overlap at high @xmath23 values needs theoretical modeling analysis similar systems left out work qualitatively same behavior in network bot removed chat network structure power - law distribution in- out - degree similar exponent @xmath30 disassortative mixing detailed analysis for separate study focus on resilience of persistent chats network analyzing _ percolation transition to arousal of messages along directed links links persist due to excitement emotional arousal of messages weights averaged arousal of links obey power - law skewed gaussian distributionpersistent network @xmath31 users @xmath32 links cut directed links average arousal exceeds threshold @xmath33 normalized by @xmath34 largest average arousal value threshold @xmath35 number links condition position network vary network topology directedness reciprocity content - topology determines percolation transition @xcite each threshold @xmath35 largest cluster weakly connected size @xmath36 plotted against threshold @xmath35 in fig . [ fig - ubu - arousal - percolation ] susceptibility @xmath37 defined as @xmath38 determines fluctuations cluster size cutting links arousal distribution cluster size satisfies power law @xmath39 @xmath37 peak at transition point @xmath40 @xmath36 geometry complexity change network structure at transition point @xmath41 ? examine of retained links @xmath42 average degree @xmath43 of giant cluster range values @xmath44 above transition point fig . scaling region @xmath45 index indicates @xmath46 critical value @xmath47 and @xmath48 exponent @xmath49 close to one@xmath50 physics methods studied self - organized dynamics user interactions online chats shown type messages emotional arousal essential persistent techno - social structure social graph hierarchically organized central core active users web bot contrast spontaneous organization marks new class structures emergent online communication networks percolation transition smooth enhanced fluctuations targeted deletion links emotional arousal over threshold analysis reveals contents exchanged messages role users dynamics integral parts topology online social graphs consider features beyond conventional interactions useful studying physical systems complex molecules objects support program p1 - 0044 research agency slovenia european community s program fp7-ict-2008 - 3 grant@xmath54 231323 m. mitrovi b. tadi handbook optimization complex networks theory applications m thai p. pardalos optimization applications vol.57 part 2 pp . 209227 doi:10.1007/978 - 1 - 4614 - 0754 - 6@xmath558 january 2012 isbn 978 - 0 - 19 - 953304 - 6 j. onela j. saramaki j. hyvnen g. szabo m.menezes kaski barabsi j. kertesz 9(6):176 2007 undirected graph size @xmath56-core decreases power @xmath57 strongest core @xmath58 contains bot | 0.568156 | 3,449 | 1,692 | , Saving $0.1 in GPT-4. |
20,704 | we present an upper limit of galactic diffuse gamma - ray flux , measured with the gamma experiment at energy about 175 tev .
the results were obtained using selection of muon poor extensive air showers at mountain level ( 700 g/@xmath0 , mt . aragats , armenia ) for 5 gev muon energy threshold .
an upper limit for the differential flux at energy @xmath1 tev is equal to @xmath2 @xmath3 for @xmath4 confidence level .
cosmic rays , gamma ray , energy spectrum | ultra - high energy ( @xmath5 tev ) galactic gamma - radiation is an important source of information about the origin of cosmic rays and their propagation . according to conventional model of cosmic rays the expected flux of ultra - high energy @xmath6-rays in the energy range of @xmath7 pev has presumably to have hadronic origin from sources distributed within radii @xmath8 mpc @xcite respectively .
+ in the very - high ( tev ) energy region @xmath6-ray flux was measured by ground - based systems and single cherenkov telescopes ( whipple @xcite , cangaroo @xcite , magic @xcite , h.e.s.s .
@xcite , milagro @xcite ) .
the measurements in the ultra - high energy region ( @xmath5 tev ) are still poor and made only by extensive air showers ( eas ) arrays at chacaltaya @xcite , msu @xcite , tien - shan @xcite , casa - mia @xcite , eas - top @xcite , kascade @xcite and grapes - iii @xcite .
+ this paper is devoted to measurements of the diffuse gamma - rays with gamma eas array @xcite at aragats mountain observatory , where correlation of observable shower parameters with primary energy is about @xmath9 .
the primary nuclei ( predominantly @xmath10 ) and @xmath6-showers were discriminated using no - muon signal from underground muon scintillation carpet .
gamma experiment @xcite is the ongoing study of primary energy spectra in the range of @xmath11 ev using eas array with 33 concentric disposed scintillator stations ( each of 3@xmath121m@xmath13 ) and underground scintillation carpet with 150 scintillators ( each of 1m@xmath13 ) to detect the shower muon component with energy @xmath14 , where @xmath15 is the shower zenith angle .
layout of gamma facility is presented in fig . 1 . + the detailed description of experiment , results of exploration of @xmath16-like , and @xmath17-like primary nuclei energy spectra derived from parametrized eas inverse problem solution are presented in @xcite .
the all - particle primary energy spectrum obtained from event - by - event analysis is published in @xcite . + here
, the discrimination of @xmath6-showers from primary nuclei induced showers is performed on the basis of following 6 conditions : + 1 ) the reconstructed shower core coordinates is distributed within radius of @xmath18 m ; + 2 ) shower zenith angles @xmath19 ; + 3 ) reconstructed shower size @xmath20 ; + 4 ) reconstructed shower age parameters ( @xmath21 ) is distributed within @xmath22 ; + 5 ) goodness - of - fit test for reconstructed showers @xmath23 ; + 6 ) no - muon signal is recorded for detected showers satisfying the previous 5 conditions . + the selection criteria and @xmath6-shower discrimination rule ( 6 ) above were obtained using corsika shower simulation code @xcite for the nkg and egs modes in the frameworks of the sibyll @xcite interaction model .
simulations were done for 4 nuclear species : @xmath24 using united energy spectral index @xmath25 @xcite .
simulated samples were equal @xmath26 , @xmath27 , @xmath28 and @xmath29 for @xmath30 nuclei and nkg mode of corsika .
the samples for the egs mode of corsika were equal to @xmath31 for primary @xmath6-quanta , @xmath32 for primary protons and @xmath33 for @xmath34 and @xmath17 primary nuclei .
the simulation strategy and reconstruction method for shower size ( @xmath35 ) , age parameter ( @xmath21 ) , core coordinates ( @xmath36 ) and shower zenith angle ( @xmath15 ) were the same as in @xcite .
+ the shower trigger efficiency and shower size reconstruction errors ( @xmath37 and @xmath38 ) are presented in figs .
2,3 respectively .
the observed differences of reconstructed shower size biases @xmath37 for different primary particles ( fig .
3 , lower panel ) stems from differences of corresponding lateral distribution functions of shower particles .
+ , where @xmath39 and @xmath40 parameters depend on primary particle ( @xmath41).,width=288 ] ) for different primary particles ( symbols ) . @xmath42 and @xmath43 is an estimation of shower size @xmath35.,width=288 ] the distribution of detected shower age parameters ( gamma data ) in comparison with expected distributions for primary @xmath30 nuclei are presented in fig . 4 ( left panel ) .
the primary elemental ) distribution for all showers ( left panel ) and no - muon detected showers ( @xmath44 , right panel ) .
simulated data for the primary mixed composition ( @xmath45 @xmath46 , left panel ) and primary gamma ray ( right panel ) are normalized to the corresponding gamma experimental data.,width=288 ] composition and energy spectra were taken from solution of parametrized eas inverse problem @xcite in the frameworks of the sibyll interaction model ( fig .
5 , shaded area @xcite ) and were extrapolated up to the 100 tev energy region .
the reliability of this extrapolations stems from data @xcite .
+ the right panel of fig .
4 shows distribution of shower age parameters for selected no - muon signal events ( shades gamma data area ) in comparison with corresponding expected distributions from the simulated @xmath6-showers and background proton showers .
it is well seen , that the eas age parameter is also a data carrier about primary particle ( @xmath6-showers are younger ) .
however , we did not include yet the age parameter in the @xmath6-showers selection criteria .
results in fig .
4 we use only as indication of an agreement between simulated and corresponding detected distributions .
+ the detected muon number spectra in the normalization of probability density function for different shower size thresholds ( @xmath47 and @xmath48 ) are presented in fig .
6 ) spectra for different shower size thresholds ( @xmath49 ) .
hollow symbols ( circle , square and triangle ) are gamma experimental data . the symbols in @xmath50 , @xmath51 and @xmath52 columns correspond to simulated data for the primary @xmath6 and mixed composition ( @xmath30 @xcite ) computed using the egs and nkg modes of corsika.,width=288 ] ( hollow symbols ) in comparison with the corresponding expected spectra from different primary particles ( @xmath53 ) and different simulation modes ( nkg , egs ) of corsika .
energy spectra and elemental composition of primary nuclei ( fig . 5 ) used in the figs .
4,6 were taken from @xcite and applied for the energy region @xmath5 tev .
+ the energy of primary particle is estimated using event - by - event method @xcite according to the empirical expression : @xmath54 , where @xmath55 is in gev , parameters @xmath56 and @xmath57 are determined using goodness - of - fit test for simulated database and depend on primary particle @xmath41 .
+ the corresponding accuracies providing for @xmath58 is described by the log - linear functions + @xmath59 , where @xmath60 , @xmath61 , @xmath62 and @xmath63 , @xmath64 , @xmath65 for primary @xmath6 , @xmath66 and proton induced no - muon detected shower ( @xmath67 ) respectively .
+ the primary energy reconstruction efficiencies are presented in fig . 7 for different primary particles .
the lines represent the log - linear empiric expression above for primary proton showers ( dashed line ) and @xmath6-showers ( solid line ) respectively .
it is seen , that the proton produced no - muon signal events ( hollow circles ) practically similar to @xmath6-showers .
the inset histogram shows the distribution of @xmath68 ratio for @xmath66 and @xmath6 primary particles respectively . + ) and corresponding shower size ( @xmath35 ) distributions at observation level for 5000 primary protons ( bold dot symbols ) and 5000 @xmath6 ( gray dot symbols ) .
hollow circles correspond to the proton showers with no - muon signal from underground muon carpet ( @xmath44 ) .
solid and dashed lines are the log - linear approximations ( see text ) for primary @xmath6 and @xmath66 correspondingly .
inset histograms are @xmath68 distributions for primary proton and @xmath6 ( shaded area).,width=288 ]
@xmath69 shower events were selected for operation time @xmath70 h of gamma experiment .
number of detected shower events versus number of detected muons ( @xmath71 ) for different shower size thresholds @xmath72 and @xmath48 are presented in fig .
8 ( histograms with shaded area ) . the symbols in fig
. 8 are the corresponding expected number of events simulated using the corsika code for primary energy spectra @xcite presented in fig .
the simulations were carried out for two modes of corsika to get high accuracy of simulation ( egs mode ) and large simulated sample ( nkg mode ) .
+ the agreement of simulated and detected muon spectra in all measurement region and lack of statistically significant excess of no - muon signal events ( fig .
8) allowed us to estimate only an upper limit of @xmath6-ray flux according to the expression ) spectra for different shower size thresholds ( @xmath49 ) and different mode ( nkg , egs ) of corsika.,width=288 ] @xmath73 where @xmath74 is the number of no - muon detected showers , @xmath75 is the probability to detect no - muon signal for @xmath6-showers ( see fig .
6 ) , @xmath76 is the trigger efficiency ( see fig .
2 ) , @xmath77 is the average shower zenith angle , @xmath78 and @xmath79 are the eas core detection area and corresponding solid angle . + the obtained upper limit of differential @xmath6-ray flux in the energy range 100 - 300 pev are presented in fig
. 9 ( black downward triangle symbol ) in comparison with casa - mia @xcite , kascade @xcite and eas - top @xcite data .
an upper limit of @xmath6-ray differential flux at energy @xmath80 tev obtained with gamma experiment is equal to @xmath2 @xmath3 for @xmath4 confidence level and it is in close agreement with the casa - mia data @xcite .
+ the lower limit for the primary energy spectra and elemental composition obtained with the gamma experiment @xcite can be extended to the lower energy region up to about 100 tev energies .
+ we are going to increase the underground muon carpet area up to 250 m@xmath13 to improve @xmath6/proton showers discrimination efficiency .
this work has been partly supported by the research grant no . 090 from the armenian government , the rfbr grant 07 - 02 - 00491 in russia , by the moscow p.n .
lebedev physical institute and the hayastan all - armenian fund
. + 99 f.a .
aharonyan , `` very high energy cosmic gamma radiation '' , published by world scientific ( 2004 ) .
m. schroedter et al . , astrophys.j . 634 ( 2005 ) 947 .
m. ohishi et al . , astropart .
phys . , 30 ( 2008 ) 47 .
e. aliu , et al . , arxiv:0810.3561v1 [ astro - ph ] ( 2008 ) .
f. aharonian f. et al . , ( h.e.s.s .
collaboration ) a&a , 477 , ( 2008 ) 481 .
abdo et al.,astrophys.j . 688
( 2008 ) 1078 .
a. castellina et al.,proceedings of the 27th icrc , hamburg , germany ( 2001 ) 2735 .
a.m. anokhina at al . ,
astrophysics and space science , 209 , 1 ( 1993 ) 19 .
nikolsky , j.n .
stamenov , s.z .
ushev , 1984 , jetp ( journal of experimental and theoretical physics ) , 60 , 10 .
mohanty , s.k .
gupta et al . , 29th icrc , pune 6 ( 2005 ) 21 .
chantell , m. c. et al .
1997 , phys .
79 , 1805./ a. borione et al .
, astrophys .
j. , 493:175 - 179 ( 1998 ) .
aglietta , m. et al .
1996 , astropart .
, 6 , 71 . g. schatz et al .
28@xmath81 icrc , tsukuba ( 2003 ) 2293 .
garyaka , r.m .
martirosov , s.v .
ter - antonyan et al . , astropart .
phys . , 28 , 2 ( 2007 ) 169/ arxiv:0704.3200v1 [ astro - ph ]
garyaka , r.m .
martirosov , s.v .
ter - antonyan et al . , j of phys , g : nucl . and part .
phys . , 35 ( 2008 ) 115201/ arxiv:0704.3200v1 [ astro - ph ] .
d. heck , j. knapp , j.n .
capdevielle , g. schatz , t. thouw , forschungszentrum karlsruhe report , fzka 6019 ( 1998 ) .
fletcher , t.k .
gaisser , p. lipari , t. stanev , phys.rev .
* d50 * ( 1994 ) 5710 .
kochanov , t.s .
sinegovskaya and s.i .
sinegovsky , astropart .
phys . 30 ( 2008 )
aharonian , f. a. , a. m. atoyan 2000 , astron .
362 , 937 . | - high energy @xmath5 tev galactic gamma - radiation origin cosmic rays propagation conventional model flux high energy-rays energy range @xmath7 pev hadronic origin from sources @xmath8 @xcite very - high energy region-ray flux measured by ground systems cherenkov telescopes measurements ultra - high energy region @xmath5 tev poor air showers arrays at chacaltaya msu tien - shan casa - mia eas - top kascade grapes - iii paper measurements diffuse gamma - rays gamma eas array @xcite aragats mountain observatory correlation shower parameters with primary energy @xmath9 primary nuclei @xmath10 @xmath6-showers discriminated using no - muon signal underground muon scintillation carpet gamma experiment @xcite primary energy spectra range @xmath11 ev eas array 33 scintillator stations underground scintillation carpet 150 scintillators detect shower muon component energy @xmath14 @xmath15 shower zenith angle layout gamma facility fig . 1 detailed description experiment results @xmath16-like primary nuclei energy spectra from eas inverse problem solution presented in @xcite all - particle primary energy spectrum from event - by - event analysis published in @xcite discrimination of @xmath6-showers from primary nuclei induced showers 6 conditions shower core coordinates within radius @xmath18 m shower zenith angles shower size age parameters within @xmath22 goodness - of - fit test for no - muon signal recorded for showers satisfying selection criteria @xmath6-shower discrimination rule obtained using corsika shower simulation code @xcite nkg egs modes sibyll interaction model simulations for 4 nuclear species energy spectral index samples equal @xmath26 for @xmath30 nuclei nkg mode corsika samples egs mode corsika equal to @xmath31 @xmath6-quanta protons primary nuclei simulation strategy reconstruction method for shower size age parameter core coordinates shower zenith angle same as in @xcite shower trigger efficiency shower size reconstruction errors presented in figs 2,3 differences shower size biases @xmath37 for primary particles from lateral distribution functions shower particles @xmath39 @xmath40 parameters depend on primary particle @xmath42 @xmath43 estimation of shower size @xmath35 distribution of detected shower age parameters with expected distributions for primary @xmath30 nuclei presented in . 4 primary elemental distribution for all showers no - muon detected showers simulated data for mixed composition @xmath45 @xmath46 gamma ray normalized to gamma experimental data composition energy spectra taken from parametrized eas inverse problem @xcite sibyll interaction model extrapolated up to 100 tev energy region reliability stems from data @xcite fig . 4 distribution shower age parameters for no - muon signal events expected distributions from simulated @xmath6-showers background proton showers eas age parameter data carrier about primary particle @xmath6-showers younger include age parameter in @xmath6-showers selection criteria results fig . 4 agreement between simulated detected distributionsdetected muon number spectra probability density function shower size thresholds @xmath47 presented in fig . 6 hollow symbols circle triangle are gamma experimental data symbols @xmath50 columns correspond simulated data primary @xmath6 mixed composition @xmath30 computed egs nkg modes corsika expected spectra primary particles @xmath53 simulation modes energy spectra elemental composition primary nuclei fig . 5 ) taken from @xcite applied energy region @xmath5 tev energy primary particle estimated event - by - event method @xcite expression @xmath54 @xmath55 gev parameters @xmath56 @xmath57 determined goodness - of - fit test depend on primary particle @xmath41 accuracies @xmath58 described log - linear functions @xmath59 primary @xmath6 @xmath66 proton induced no - muon detected shower @xmath67 ) primary energy reconstruction efficiencies presented in fig . 7 primary particles lines represent log - linear empiric expression primary proton showers @xmath6-showersproton produced no - muon signal events similar to @xmath6-showers histogram shows @xmath68 ratio for @xmath66 @xmath6 primary particles shower size distributions for 5000 primary protons 5000 @xmath6 hollow circles proton showers with no - muon signal from underground muon carpet @xmath44 solid dashed lines log - linear approximations for primary @xmath6 @xmath66 histograms are @xmath68 distributions for primary proton @xmath6 @xmath69 shower events selected for operation time @xmath70 h gamma experiment detected shower events versus muons @xmath71 ) for different shower size thresholds @xmath72 @xmath48 in fig . 8 symbols expected number of events simulated using corsika code for primary energy spectra simulations for two modes of corsika agreement of muon spectra lack of excess no - muon signal events estimate upper limit of @xmath6-ray flux for different shower size thresholds corsika number of no - muon detected showers probability to detect no - muon signal for @xmath6-showers@xmath76 trigger efficiency fig 2 average shower zenith angle eas core detection area solid angle upper limit differential @xmath6-ray flux 100 - 300 pev presented fig . 9 casa - mia kascade eas - top data upper limit @xmath6-ray differential flux @xmath80 tev equal to @xmath2 @xmath3 @xmath4 confidence level with casa - mia data lower limit primary energy spectra elemental composition gamma experiment extended lower energy region 100 tev energies increase underground muon carpet area 250 m@xmath13 improve @xmath6/proton showers discrimination efficiency supported research grant 090 armenian rfbr grant 07 - 02 - 00491 russia moscow lebedev physical institute hayastan all - armenian fund aharonyan high energy cosmic gamma radiation world scientific 2004 schroedter astrophys.j . 634 ( 2005 ) 947 ohishi astropart . . 30 2008 ) 47 aliu arxiv:0810.3561v1 astro - 2008 aharonian collaboration a&a 477 2008 481 abdo,astrophys.j688 2008 ) 1078 castellina 27th icrc hamburg germany 2001 ) 2735 anokhina astrophysics space science 209 1 1993 ) 19 nikolsky stamenov ushev 1984 journal experimental theoretical physics 60 10 mohanty gupta 29th icrc pune 6 2005 ) 21 chantell 1997 . 79 1805 borione astrophys . 493:175 - 179 1998 ) aglietta 1996 astropart . 6 , 71 schatz 28@xmath81 icrc tsukuba 2003 ) 2293 garyaka martirosov ter - antonyan astropart phys . 28 2 2007 ) 169 arxiv:0704.3200v1 - garyaka martirosov - antonyan phys nucl part 35 2008 ) 115201 arxiv:0704.3200v1 heck knapp capdevielle schatz thouw forschungszentrum karlsruhe report 6019 1998 ) fletcher gaisser lipari stanev phys.rev d50 1994 ) 5710 kochanovsinegovskaya sinegovsky astropart. 30 2008 aharonian atoyan 2000 astron . 362 937 | 0.494159 | 3,209 | 1,607 | , Saving $0.1 in GPT-4. |
20,711 | solutions for the amplitudes that give accurate description of pp and p@xmath0 scattering at high energies are investigated , with particular attention given to the properties of their zeros and slopes , whose determination is required for the study of the coulomb interference region .
proper extrapolations of these quantities to the lhc energies are important for the analysis of the forthcoming experiments . | the detailed knowledge of t - dependence of the real and imaginary amplitudes in pp and @xmath1p scattering is important for the study of the coulomb interference and the determination of the collision parameters . in general , the behavior of the real amplitude is oversimplified , with assumption that real and imaginary slopes are identical and that the ratio @xmath2 is independent of t. however , the present work , based on our analysis@xcite of the isr , sps and fermilab data , shows that the variation of the real amplitude in the coulomb interference region is very fast , with a slope about twice that of the imaginary part . the forthcoming experiments at rhic and lhc,@xcite
respectively at 200 - 500 gev and 14 tev , make urgent the investigation of the structure of the amplitudes in the low @xmath3 region.@xcite the real forward slope is obviously connected with the value of t at which this amplitude passes through zero .
we find this connection , fulfilling the expectation from the theorem by a. martin@xcite that the real part has a zero that approaches the forward direction as the energy increases .
our work covers the full t - range , obtaining also the zeros of imaginary amplitudes and the shapes of the dips in the cross sections of pp and @xmath1p scattering .
we have constructed an accurate description of the t - dependence of the amplitudes , with smooth energy dependence in the parameters , with particular attention to the determination of slopes and zeros .
the amplitudes have peculiar shapes as functions of @xmath3 , with approximate exponential behaviour only for very small @xmath3 , and then turning fast towards zero .
the results for some energies ( 52.8 gev in pp , and 541 and 1800 gev for @xmath0p ) are shown in fig .
[ ampl ] .
theoretical arguments@xcite about the existence of a real zero close to @xmath4 at high energies are confirmed by our construction .
the real pp amplitude has a second zero , while in @xmath1p only the first zero at small @xmath3 is observed.@xcite the pp case is drawn in fig .
( [ real52fig ] ) , where the real amplitude at @xmath5 gev data is shown for small and for large @xmath6 , so that we can see where the second zero in the pp channel occurs .
this second zero , being close to the zero of the imaginary amplitude , stresses the dips observed in the pp differential cross sections , but not those of the @xmath1p channel,@xcite due to a difference in the sign of the real tail . with real and imaginary amplitudes described at low @xmath3 by exponential slopes ,
we write for each channel , @xmath7 , @xmath8 and the experimental slope @xmath9 is given by @xmath10 we determine separately @xmath11 and @xmath12 for each channel , and observe the approximate relation @xmath13 above 30 gev . the exponential slopes at @xmath14 for 52.8 gev
are shown in fig .
[ ampl ] , right hand side , where we plot the pp amplitudes normalized to one at @xmath14 .
we observe that the concept of exponential slope is limited to a small @xmath3 range , with strong characteristic curvatures of the real and imaginary amplitudes appearing very soon in the @xmath3 scale .
it is remarkable that the imaginary amplitude has a positive ( going upwards ) curvature at the beginning , before it starts running fast to zero .
these results for the slopes are shown in fig .
[ br_bi ] .
for @xmath12 we have a simple parametrization @xmath15 in @xmath16 , valid for both pp and @xmath1p channels , with a prediction @xmath17 for the lhc energy @xmath18 14 gev .
the description is not so simple for the real slope . here
the pp and @xmath19 cases are different at the lower energies , and an extrapolation to higher energies is not neat for pp . for @xmath19 , using the 541 and 1800 gev points , we guess a linear @xmath20 behavior @xmath21 in @xmath16 . the role of @xmath22 in the determination of the collision parameters can be seen in the work of gauron , nicolescu and selyugin,@xcite who find a point @xmath23 where the real amplitude is compared to the coulomb part .
they study pp scattering at 52.8 gev , suggesting a lower value for @xmath2 than 0.077 determined experimentally .
however , to write the real amplitude at the matching point @xmath23 , they use the exponential factor @xmath24 with the experimental slope @xmath25 .
however , according to our work , they should use the faster slope @xmath26 of the real amplitude .
then the new value that they suggest for @xmath2 is not properly evaluated .
a theorem by a. martin@xcite predicts that in pp and @xmath1p scattering the real part has a change of sign at a point @xmath27 which moves to the forward direction as the energy increases .
the paper suggests real zeros of the form @xmath28 as we find with our description of amplitudes , as shown in fig .
[ zeros ] .
0 f. pereira and e. ferreira , _ phys .
_ * d59 * ( 1999 ) 014008 ; _ phys .
rev . _ * d61 * ( 2000 ) 077507 . g. anelli et al . , `` totem physics '' , hep - ex/0602025 .
c. bourrely et al . , `` why the real part of the proton - proton forward scattering amplitude should be measured at the lhc '' , hep - ex/0511135 .
a. martin , _ phys .
* b404 * ( 1997 ) 137 .
p. gauron , b. nicolescu , o.v .
selyugin , _ phys .
lett . _ * b629 * ( 2005 ) 83 . | knowledge t - dependence real imaginary amplitudes in pp @xmath1p scattering important for study coulomb interference collision parameters behavior real amplitude oversimplified imaginary slopes identical ratio independent of t present work shows variation real amplitude in coulomb interference region fast slope twice imaginary part experiments at rhic lhc at 200 - 500 gev 14 tev investigation structure amplitudes in low @xmath3 region real forward slope connected with value t amplitude passes through zero real part zero approaches forward direction as energy increases work covers full t - range zeros of imaginary amplitudes shapes of dips in cross sections pp @xmath1p scattering constructed accurate description of t - dependence amplitudes smooth energy dependence parameters attention to determination slopes zeros amplitudes have peculiar shapes functions of @xmath3 approximate exponential behaviour for small @xmath3 turning fast towards zero results for energies 52.8 541 1800 gev shown in fig . arguments about real zero close to @xmath4 at high energies confirmed by constructionreal pp amplitude second zero @xmath1p first zero at small @xmath3 observed pp case in fig . real amplitude at @xmath5 gev small large @xmath6 second zero pp channel second zero close to zero imaginary amplitude stresses dips pp differential cross sections not @xmath1p channel difference sign real tail real imaginary amplitudes described low @xmath3 by exponential slopes each channel experimental slope determine @xmath11 @xmath12 observe relation @xmath13 above 30 gev exponential slopes at @xmath14 for 52.8 gev shown in fig . pp amplitudes normalized to one @xmath14 concept exponential slope limited small @xmath3 range strong characteristic curvatures real imaginary amplitudes in @xmath3 scale imaginary amplitude positive curvature beginning zero results slopes shown in fig . @xmath12 simple parametrization @xmath15 @xmath16 valid for both pp @xmath1p channels prediction @xmath17 for lhc energy @xmath18 14 gev description not simple real slopepp @xmath19 cases different at lower energies extrapolation to higher energies not for pp @xmath19 541 1800 gev points guess linear @xmath20 behavior @xmath21 @xmath16 role @xmath22 collision parameters in work gauron nicolescu selyugin point @xmath23 real amplitude compared to coulomb part study pp scattering at 52.8 gev lower value for @xmath2 than 0.077 real amplitude matching point @xmath23 use exponential factor @xmath24 experimental slope @xmath25 use faster slope @xmath26 real amplitude new value for @xmath2 not properly evaluated theorem by a. martin predicts pp @xmath1p scattering real part change sign at point @xmath27 moves forward direction energy increases paper suggests real zeros form @xmath28 description amplitudes fig . pereira ferreira phys 1999 ) 014008 2000 ) 077507 anelli al totem physics/0602025 bourrely real proton - proton forward scattering amplitude measured at lhc ex/0511135 a martin phys 1997 ) 137 gauronnicolescu selyugin phys. lett b629 2005 ) 83 | 0.25772 | 1,330 | 682 | , Saving $0.0 in GPT-4. |
20,717 | using the recent measurement of the total production rate for @xmath0 and @xmath1 mesons in electron - positron annihilation to determine the strange quark mass parameter in the @xmath2 fragmentation functions we calculate the momentum distributions of the @xmath0 and @xmath1 mesons . 6.5 in # 1#20=1=to0@xmath30=1=to0#2 # 1 - 01 # 1#2 + * beauty quark fragmentation into strange b mesons * + kingman cheung and robert j. oakes + _ department of physics and astronomy , northwestern university , + evanston , illinois 60208 , usa _
recently the probability for a heavy quark to hadronize into a strange @xmath4 or @xmath5 meson has been measured at lep by the delphi collaboration @xcite .
these data can be used to determine the strange quark mass parameter , @xmath6 , in the heavy quark fragmentation functions and , therefore , to predict the momentum distributions of the @xmath0 and @xmath1 mesons produced in the fragmentation process . using the heavy quark fragmentation functions calculated in perturbative qcd at the scale of the heavy quark @xcite for the initial condition we have numerically integrated the altarelli - parisi equation to obtain the momentum distributions of the @xmath0 and @xmath1 mesons at the scale of the @xmath7 mass .
the resulting distributions are only moderately sensitive to the value of @xmath6 , which we find to be about 300 mev@xmath8
. the fragmentation function for the process @xmath9 at the scale of the @xmath10 mass is easily obtained from the @xmath11 fragmentation functions calculated using perturbative qcd by braaten , cheung , and yuan @xcite : @xmath12\ , .
\label{dz1}\end{aligned}\ ] ] here @xmath13 is the fraction of the @xmath14 quark momentum carried by the @xmath0 meson , @xmath15 , and @xmath16 is the @xmath0 meson @xmath17-wave radial wavefunction at the origin .
the corresponding fragmentation function for @xmath18 is @xmath19\ , , \label{dz2}\end{aligned}\ ] ] the wavefunction at the origin , @xmath16 , can be determined from the @xmath0 meson decay constant , @xmath20 , from the van royen - weisskopf @xcite relation , modified for color : @xmath21 in eqs .
( [ dz1 ] ) and ( [ dz2 ] ) the running coupling constant @xmath22 , where @xmath23 , @xmath24 is the number of flavors at the scale @xmath25 , @xmath26 , and we have chosen the scale to be @xmath27 @xcite .
the scale @xmath28 of the fragmentation functions is of the order of @xmath29 and we choose it to be @xmath30 @xcite .
integrating the altarelli - parisi equation for the fragmentation functions from the initial scale @xmath28 to the @xmath7 mass gives the momentum distributions of the @xmath0 and @xmath1 mesons produced by the fragmentation process in @xmath31 annihilation at the @xmath7 resonance . to determine the strange quark mass parameter , @xmath6
, we use the recent measurement by the delphi collaboration @xcite of the probability , @xmath32 , that a weakly decaying strange heavy meson is produced during the hadronization of a heavy quark .
in addition to the direct production @xmath9 , the @xmath0 meson can also result from the decays of excited states of @xmath0 and @xmath33 mesons .
however , these have been excluded in the measurement of @xmath34 , except for the radiative decay of the first excitation : @xmath35 .
since the @xmath36 mass difference is so small we can determine @xmath6 from the integrated fragmentation functions , which gives the measured probability @xmath34 : @xmath37\,.\ ] ] since the integrated fragmentation functions are independent of the scale , we can choose the scale to be @xmath28 in eq .
( [ fws ] ) and first determine @xmath6 and then calculate the distributions @xmath38 and @xmath39 at the scale of the @xmath7 mass by numerically integrating the altarelli - parisi equation . as input we fix the heavy @xmath10 quark mass at @xmath40 gev@xmath8 . to determine @xmath16 we use eq .
( [ fbs ] ) with the value @xmath41 mev from the lattice calculations of bernard , labrenz , and soni @xcite .
then from the measured value @xcite @xmath42 we find ( adding the two errors in quadrature ) @xmath43 which is in a reasonable range @xcite .
the estimated error in @xmath6 reflects only the experimental uncertainty in @xmath32 .
using this value of @xmath6 we have calculated the momentum distributions @xmath44 and @xmath45 , which are shown in figs .
[ fig1 ] .
figure [ fig1](a ) shows the fragmentation functions at the initial scale @xmath46 , while fig . [
fig1](b ) shows the fragmentation functions at the scale @xmath47 .
we have verified that the shapes of these distributions are only moderately sensitive to the value of @xmath6 , although the peak does shift towards larger values of @xmath13 if @xmath6 is decreased , as one might expect .
in fact , the maxima in both @xmath38 and @xmath39 occur at @xmath48 the average momentum fraction @xmath49 at the @xmath7 mass scale is also not too different for the @xmath0 and @xmath1 contributions to the spectrum : @xmath50 the spectrum also shows some evidence for approximate heavy quark spin symmetry at the level of about 50% .
the ratio @xmath51 is about 2.1 compared to 3 , as expected in the heavy quark limit from the spin independence . in terms of the ratio @xmath52 , which heavy quark symmetry predicts to be 0.75 , we find 0.68 . to compare the @xmath9 and @xmath18 spectra individually with experiments , the two contributions @xmath9 and @xmath53 have to be experimentally separated by observing the rather low energy photon , which has so far not been done . to explore the sensitivity of our calculation to the assumptions we also considered the fragmentation of a charm quark into a @xmath54 or @xmath55 meson . following ref .
@xcite we repeated the calculation assuming the probability that a @xmath54 or @xmath55 meson is produced in the hadronization of a @xmath56 quark is the same as the probability that a @xmath0 or @xmath1 meson is produced in the hadronization of a @xmath10 quark ; that is , @xmath32 is independent of the heavy quark flavor .
using the value @xmath57 @xcite , the lattice calculation @xmath58 mev @xcite , and choosing @xmath59 gev@xmath8 @xcite , we found @xmath60 which is in reasonable agreement with the value of @xmath6 obtained above in the @xmath0 case . with this value of @xmath6
we also calculated the momentum spectra for @xmath61 , for @xmath62 , and their sum at the initial scale @xmath63 as well as at the scale @xmath47 .
they are shown , respectively , in figs .
[ fig2](a ) and [ fig2](b ) .
as expected , the @xmath54 spectra are softer than the @xmath0 spectra shown in figs .
[ fig1](a ) and [ fig1](b ) because of the smaller charm quark mass .
the peaks of the spectra occur at 0.68 for @xmath54 and at 0.75 for @xmath55 and the mean momentum fractions are @xmath64 for @xmath54 and @xmath65 for @xmath55 at the scale @xmath47 [ fig .
[ fig2](b ) ] .
the ratio of the two fragmentation probabilities for the @xmath54 case also differs significantly more from the heavy quark spin symmetry prediction of 3 , being about 1.6 , than in the @xmath0 case , as one expects . also , the ratio @xmath66 is 0.62 , rather than 0.75 . to further test the sensitivity of our results to the input parameters assumed we have repeated the calculation of the @xmath54 and @xmath55 spectra using the measurement of the decay constant
@xmath67 recently reported by the cleo collaboration @xcite .
although this measured value @xmath68 mev @xcite differs substantially from the lattice result @xmath69 mev @xcite , we found @xmath70 which is not significantly different from the value above [ eq . ( [ 267 ] ) ] , and the changes in the fragmentation spectra shown in fig . [ fig2 ] are negligible . to summarize , we have calculated the momentum distributions of @xmath0 and @xmath1 mesons produced by @xmath14 quark fragmentation .
the measured total production probability was used to determine the strange quark mass parameter .
the result for @xmath6 was quite reasonable , being well above the current quark mass value , but lower than values often used for the constituent quark mass .
we note that at the scale @xmath27 that @xmath71 is uncomfortably large , which could mean there are significant corrections , which have not been calculated , to the fragmentation function at the heavy quark scale , which was calculated using perturbative qcd . but
this only affects the normalization and we determined the parameter @xmath6 from the experimental value of the total production probability .
indeed , we repeated the calculation for @xmath54 and @xmath55 meson production by fragmentation and found substantially the same value for @xmath6 and verified that the momentum distributions were only moderately sensitive to the value of @xmath6 .
the results were also not sensitive to the values assumed for the heavy quark masses or the meson decay constants .
essentially , the experimental production probability fixes the normalization , the perturbative qcd result determines the shape of the momentum spectrum at the heavy quark mass scale , and the altarelli - parisi equation governs the evolution up to the scale @xmath72 ; so perhaps it is not so surprising that the results are not too sensitive to the particular values of the parameters .
this work was supported by the u. s. department of energy , division of high energy physics , under grant de - fg02 - 91-er40684 .
99 p. abrau , _ et al . _
( delphi collaboration ) z. phys * c61 * , 407 ( 1994 ) .
e. braaten , k. cheung , and t. c. yuan , phys . rev . *
d48 * , r5049 ( 1993 ) .
r. van royen and v. weisskopf , nuovo cimento * 50 * , 617 ( 1967 ) ; * 51 * , 583 ( 1967 ) . c. bernard , j. labrenz , and a. soni , phys
. rev .
* d49 * , 2536 ( 1994 ) .
particle date group , phys . rev .
* d45 * , 51 ( 1992 ) .
talk give by m.s .
witherell , in proceedings of _ lepton and photon xvi international symposium _ ,
ithaca , ny 1993 , p.198 ; + cleo collaboration , cleo preprint cleo 93 - 14 ; + the first two errors are the statistical and systematic errors , respectively , while the last error reflects the uncertainty in the absolute @xmath73 branching ratio . | [ fig1 ] ( a ) momentum distributions for the fragmentation processes @xmath74 , @xmath75 , and their sum at the scale @xmath46 .
+ ( b ) momentum distributions for the fragmentation processes @xmath74 , @xmath75 , and their sum at the scale @xmath47 .
[ fig2 ] ( a ) momentum distributions for the fragmentation processes @xmath76 , @xmath77 , and their sum at the scale @xmath63 .
+ ( b ) momentum distributions for the fragmentation processes @xmath76 , @xmath77 , and their sum at the scale @xmath47 . | fig1 ] a momentum distributions fragmentation processes @xmath74 sum scale @xmath46 . b ) momentum distributions sum scale @xmath47 . fig2 ] a ) momentum distributions fragmentation processes @xmath76 @xmath77 sum scale @xmath63 . b ) momentum distributions sum scale @xmath47 | 0.06685 | 133 | 66 | , Saving $0.0 in GPT-4. |
20,719 | the eddington - inspired - born - infeld ( eibi ) gravity , which is formulated within the palatini formalism , is characterized by its ability to cure the big bang singularity in the very beginning of the universe .
we further analyze the eibi phantom model , and investigate the possible avoidance or alleviation of other dark energy related singularities .
we find that except for the big rip singularity and little rip event , most of the cosmological singularities of interest can be partially alleviated in this model .
furthermore , we generalize the eibi theory by adding a pure trace term to the determinant of the action .
this amendment is the most general rank - two tensor composed of up to first order of the riemann curvature .
we find that this model allows the occurrence of primitive bounces and some smoother singularities than that of big bang .
most interestingly , for certain parameter space , the big bang singularity can be followed naturally by an inflationary stage in a radiation dominated universe . | in 1934 born and infeld proposed a non - linear action for classical electrodynamics , which is characterized by its success in solving the divergence of the self - energy of point - like charges @xcite . since their proposal , modified theories of gravity with a born - infeld - inspired action have received much attention ( initiated in ref . ) .
these theories not only maintain the properties of general relativity ( gr ) for small curvatures , but provide various interesting deviations from gr at high curvature regimes . some of these amendments are hoped to shed some light on smoothing the singularities in gr .
for example , recently a theory dubbed eddington - inspired - born - infeld theory ( eibi ) @xcite has been studied from both astrophysical and cosmological points of view .
the eibi theory is shown to be able to cure the big bang singularity for a radiation dominated universe .
therefore , in refs .
and we carry out a thorough analysis on the possible avoidance of other cosmological singularities such as big rip , sudden , big freeze , type iv singularities , and little rip event in the eibi theory .
we find that though the big rip singularity and little rip can not be avoided , most of the cosmological singularities can be alleviated in the eibi theory @xcite . in ref . , we extend the eibi theory by adding a trace term into the determinantal action .
it corresponds to the most general action constructed from a rank two tensor that contains up to first order terms in curvature .
this model can provide smooth bouncing solutions which were not allowed in the eibi model for the same eibi coupling .
most interestingly , for a radiation filled universe there are some regions of the parameter space that can naturally lead to a de sitter inflationary stage without any exotic matter field .
we start reviewing the eibi model whose action in terms of the metric @xmath0 and the connection @xmath1 reads @xcite @xmath2+\mathcal{s}_\textrm{m } ( g,\psi ) .
\label{action}\ ] ] the theory is formulated within the palatini approach and the connection is assumed to be torsionless . within this setup ,
the connection @xmath1 and the metric @xmath0 are treated as independent variables .
the parameter @xmath3 is a constant with inverse dimensions to that of a cosmological constant ( @xmath4 and @xmath5 ) , @xmath6 is a dimensionless constant parametrizing the cosmological constant and @xmath7 stands for the matter lagrangian in which matter is assumed to be coupled covariantly to the metric @xmath8 only . for a flrw universe filled with a perfect fluid with energy density @xmath9 and pressure @xmath10
, the friedmann equation reads @xcite @xmath11(1+\bar\rho)(1-\bar p)^2}{[(1-\bar p)(4+\bar\rho-3\bar p)+3\frac{d\bar p}{d\bar\rho}(1+\bar\rho)(\bar\rho+\bar p)]^2 } , \label{field equation}\ ] ] where @xmath12 , @xmath13 is the hubble parameter as defined from the physical metric , @xmath14 , @xmath15 . for simplicity
, we will also use a dimensionless cosmic time : @xmath16 where @xmath17 corresponds to the cosmic time of the physical metric @xmath0 .
it can be easily verified that the big bang singularity can be avoided in this theory for a radiation dominated universe @xcite . in general a universe filled with a perfect fluid with a constant and positive equation of state @xmath18
bounces in the past for @xmath19 or has a loitering behavior in the infinite past for @xmath20 @xcite .
aside , the auxiliary metric @xmath21 which is compatible with the connection is@xcite : @xmath22 where @xmath23 from the auxiliary metric @xmath21 we can define as well a rescaled dimensionless auxiliary hubble parameter @xmath24 as follows @xmath25 where @xmath26 and @xmath27 .
then we have @xmath28 we then analyze the possible avoidance of phantom dark energy related singularities in the eibi setup .
these singularities are characterized by possible divergence of the hubble parameter and its cosmic time derivatives at some finite cosmic time ( see ref . and references therein ) .
note that we have two possible hubble rate : @xmath13 defined by @xmath0 and @xmath29 defined by @xmath21 . for the sake of completeness
, we assume the universe is filled with radiation , dark and baryonic matter , and dark energy : @xmath30 note that @xmath31 means that the equation of state of dark energy is purely a function of the dark energy density . to analyze the big rip singularities , we assume a phantom dark energy with constant equation state @xmath32 to be the component of dark energy . for the big freeze , sudden , and type iv singularities , we regard the phantom generalized chaplygin gas ( pgcg ) as the dark energy component in this model @xcite .
its equation of state takes the form : @xmath33 where @xmath34 and @xmath35 are two dimensionless constants .
the energy density in terms of the scale factor for @xmath36 is @xcite : @xmath37^{\frac{1}{1+\alpha } } , \label{rpp}\ ] ] where @xmath38 is the scale factor corresponding to the past singularity .
a subscript @xmath39 stands for quantities evaluated today .
the phantom energy of this kind will drive past sudden or type iv singularities in gr @xcite . on the other hand , if @xmath40 , the energy density of the pgcg , which drives a future big freeze singularity in gr , reads @xcite : @xmath41^{\frac{1}{1+\alpha } } , \label{rfs}\ ] ] where @xmath42 is the scale factor corresponding to the future singularity . for the little rip event
, we consider the simplest and the most common - used dark energy equation of state driving the little rip in gr ( see ref . and references therein )
@xmath43 where @xmath44 is a positive dimensionless constant . in each cases
, we not only consider the asymptotic behaviors of the physical metric , but also consider the asymptotic behaviors of the auxiliary metric by analyzing the behaviors of @xmath29 and its derivatives with respect to @xmath45 near the singular events .
the results are summarized in table .
[ aba : tbl1 ] @xcite .
[ aba : tbl1 ] besides , we also use a cosmographic approach to constrain the parameters present in our model @xcite . as a result
, it turns out that the cosmographic analyses pick up the physical region preferring the occurrence of a type iv singularity in the finite past or the loitering effect in an infinite past . while it is necessary to impose more conditions to forecast the future doomsdays of the universe in this model @xcite .
in ref . we add a pure trace term , which takes the form of @xmath46 , to the eibi determinantal lagrangian .
the action becomes @xmath47|}-\lambda\sqrt{-g}\big]+s_m . \label{action}\ ] ] the dimensionless constants @xmath34 and @xmath48 should satisfy @xmath49 to ensure the recovering of einstein gr at the low curvature limit @xcite . the field equations with respect to @xmath0 and the connection are @xmath50=0 , \label{eq2}\end{aligned}\ ] ] respectively , where @xmath51 and @xmath52 is its inverse .
note that the covariant derivative @xmath53 is defined through the affine connection . in ref .
we consider a flat flrw universe filled with radiation in this type of model . to analyze the behavior of the cosmological solutions , we rewrite the field equations in an algebraic form and express the energy density and hubble rate using a single variable @xmath54 . a parametrized hubble function
can be obtained @xcite .
the cosmological solutions in the presence of the trace term , i.e , @xmath55 , are summarized in table .
[ aba : tbl2 ] .
we highlight the following interesting results concerning this model : @xcite bouncing solutions which were not allowed in the eibi model for the same eibi coupling @xmath56 a de sitter inflationary stage after the big bang singularity when @xmath57 a quasi - sudden singularity where @xmath58 becomes very large but finite at a finite cosmic time [ aba : tbl2 ]
we give a summarized analysis on the avoidance of dark energy related singularities in eibi type of theories by investigating the asymptotic behaviors of the hubble rate and its cosmic time derivatives defined by the physical metric @xmath0 , which is coupled to matter , and by the auxiliary metric @xmath21 compatible with the physical connection .
for the physical metric @xmath0 we find that though the big rip singularity and the little rip event driven by phantom dark energy are not cured in the eibi theory , this theory to some extent smoothes the other phantom dark energy related singularities present in gr @xcite .
the behavior of the auxiliary metric @xmath21 is even more regular ( see table . [ aba : tbl1 ] for a summary ) .
the fate of a bound structure near the singularities in the eibi theory and some constraints on the model from a cosmographic approach are also analyzed in ref . .
furthermore , a modification of the eibi theory by adding a pure trace term into the determinantal lagrangian and its cosmological solutions are investigated @xcite .
we assume a homogeneous and isotropic universe filled with radiation , and analyze the behaviors of the cosmological solutions using a parametric friedmann equation .
we summarize our results in table .
[ aba : tbl2 ] .
the interesting cosmological solutions quoted in this theory are all stemmed from pure geometrical effects .
the work of mbl is supported by the portuguese agency fundao para a cincia e tecnologia through an investigador fct research contract , with reference if/01442/2013/ cp1196/ct0001 .
c .- y.c . and p.c .
are supported by taiwan national science council under project no .
nsc 97 - 2112-m-002 - 026-my3 and by taiwan s national center for theoretical sciences ( ncts ) .
is in addition supported by us department of energy under contract no .
de - ac03 - 76sf00515 .
m. born and l. infeld , proc .
a * 144 * , 425 ( 1934 ) .
s. deser and g. w. gibbons , class .
grav . *
15 * , l35 ( 1998 ) .
m. baados and p. g. ferreira , phys .
lett . * 105 * , 011101 ( 2010 ) erratum : [ phys . rev . lett . * 113 * , no . 11 , 119901 ( 2014 ) ] . m. bouhmadi - lpez , c. y. chen and p. chen , eur .
j. c * 74 * , 2802 ( 2014 ) .
m. bouhmadi - lpez , c. y. chen and p. chen , eur .
j. c * 75 * , 90 ( 2015 ) . c. y. chen , m. bouhmadi - lpez and p. chen , eur .
j. c * 76 * , 40 ( 2016 ) .
m. bouhmadi - lpez , c. y. chen and p. chen , phys .
d * 90 * , 123518 ( 2014 ) .
avelino and r. z. ferreira , phys .
d * 86 * , 041501 ( 2012 ) .
j. h. c. scargill , m. baados and p. g. ferreira , phys .
d * 86 * , 103533 ( 2012 ) .
m. bouhmadi - lpez and j. a. jimenez madrid , jcap * 0505 * , 005 ( 2005 ) . m. bouhmadi - lpez , p. f. gonzlez - daz and p. martn - moruno , int .
j. mod .
d * 17 * , 2269 ( 2008 ) .
i. albarran , m. bouhmadi - lpez , c. kiefer , j. marto and p. v. moniz , arxiv:1604.08365 [ gr - qc ] . | 1934 born infeld proposed non - linear action for classical electrodynamics solving divergence self - energy point - like charges modified theories gravity with born - infeld - inspired action received attention initiated theories maintain properties general relativity for small curvatures provide deviations at high curvature regimes amendments hoped smoothing singularities gr eddington - inspired - born - infeld theory ( eibi ) studied from astrophysical cosmological points eibi theory big bang singularity radiation dominated universe analysis avoidance cosmological singularities big rip sudden big freeze type iv singularities little rip in eibi theory big rip little rip avoided most cosmological singularities alleviated in eibi theory eibi theory trace term into determinantal action general action rank two tensor first order terms curvature model smooth bouncing solutions not allowed eibi model radiation filled universe regions parameter space lead to de sitter inflationary stage without exotic matter field reviewing eibi model action metric @xmath0 connection @xmath1 reads @xcite @xmath2+\mathcal{s}_\textrm{m } ( g,\psi ) .{action theory palatini approach connection torsionless connection @xmath1 metric @xmath0 independent variables parameter @xmath3 constant inverse dimensions cosmological constant @xmath6 dimensionless constant cosmological constant @xmath7 matter lagrangian coupled covariantly to metric @xmath8 universe perfect fluid energy density @xmath9 pressure @xmath10 friedmann equation reads @xcite @xmath11(1+\bar\rho)(1-\bar p)^2}{[(1 p)(4+\bar\rho-3\bar p)+3\frac{d\bar p}\rho}(1+ p)]^2{field equation @xmath12 @xmath13 hubble parameter physical metric @xmath14 @xmath15 dimensionless cosmic time @xmath16 @xmath17 corresponds cosmic time physical metric @xmath0 big bang singularity avoided radiation dominated universe @xcite universe perfect fluid constant positive equation state @xmath18 bounces past @xmath19 loitering behavior infinite past @xmath20auxiliary metric @xmath21 compatible with connection@xcite @xmath22 @xmath23 rescaled dimensionless hubble parameter @xmath24 @xmath25 @xmath26 @xmath27 @xmath28 avoidance of phantom dark energy singularities in eibi setup divergence hubble parameter cosmic time derivatives at finite cosmic time two hubble rate @xmath13 @xmath0 @xmath29 @xmath21 universe filled with radiation dark baryonic matter dark energy @xmath30 @xmath31 equation of state dark energy function of dark energy density big rip singularities phantom dark energy with constant equation state @xmath32 component of dark energy big freeze sudden type iv singularities phantom generalized chaplygin gas ) as dark energy component @xcite equation of state @xmath33 @xmath34 @xmath35 dimensionless constants energy density scale factor for @xmath36 is @xcite @xmath37 @xmath38 scale factor past singularity @xmath39 for quantities evaluated today phantom energy past sudden type iv singularities in @xcite@xmath40 energy density pgcg drives future big freeze singularity in gr reads @xcite @xmath41^{\frac{1}{1+\alpha } }{rfs @xmath42 scale factor future singularity little rip event consider simplest dark energy equation driving little rip gr @xmath43 @xmath44 positive dimensionless constant consider asymptotic behaviors physical metric auxiliary metric analyzing behaviors @xmath29 derivatives @xmath45 near singular events results summarized in table . tbl1 use cosmographic approach to constrain parameters model @xcite cosmographic analyses physical region type iv singularity in finite past or loitering effect infinite past necessary to impose more conditions to forecast future doomsdays universe model add pure trace term @xmath46 to eibi determinantal lagrangian action becomes @xmath47|}-\lambda\sqrt{-g}\big]+s_m dimensionless constants @xmath34 and @xmath48 satisfy @xmath49 recovering of einstein gr at low curvature limitfield equations @xmath0 connection @xmath50=0{eq2\end{aligned @xmath51 @xmath52 inverse covariant derivative @xmath53 defined through affine connection flat flrw universe radiation analyze cosmological solutions rewrite field equations express energy density hubble rate single variable @xmath54 parametrized hubble function obtained @xcite cosmological solutions trace term @xmath55 summarized in table . results @xcite bouncing solutions not allowed in eibi model @xmath56 de sitter inflationary stage after big bang singularity @xmath57 quasi - sudden singularity @xmath58 large finite at finite cosmic time summarized analysis avoidance dark energy singularities in eibi theories asymptotic behaviors hubble rate cosmic time derivatives defined physical metric @xmath0 coupled to matter auxiliary metric @xmath21 compatible with physical connection physical metric @xmath0 big rip singularity little rip event phantom dark energy not cured eibi theory smoothes other phantom dark energy singularities @xcite behavior auxiliary metric @xmath21 regular see table .fate bound structure near singularities eibi theory constraints model cosmographic approach analyzed modification eibi theory adding trace term determinantal lagrangian cosmological solutions investigated assume homogeneous isotropic universe radiation analyze behaviors cosmological solutions parametric friedmann equation summarize results table cosmological solutions from geometrical effects work supported portuguese agency fundao cincia tecnologia research contract reference if/01442/2013/ cp1196/ct0001 supported taiwan national science council project nsc 97 - 2112-m-002 - 026-my3 taiwan s national center theoretical sciences supported department energy contract de - ac03 - 76sf00515 m. born l. infeld proc . * 144 *, 425 ( 1934 ) deser w. gibbons class 15 *, l35 ( 1998 ) baados ferreira . lett. * 105 * 011101 ( 2010 ) 113 * 11 119901 ( 2014 ) bouhmadi - lpez c y. chen p. chen * 74 * 2802 ( 2014 ) 75 * , 90 ( 2015 )chen bouhmadi - lpez p chen 76 * 40 2016 ) bouhmadi - lpez chen p. chen phys d 90 * 123518 2014 ) avelino z ferreira d 86 041501 2012 ) scargill baados ferreira d 86 * 103533 2012 ) bouhmadi - lpez jimenez madrid 0505 * 005 ( 2005 ) -. gonzlez - daz p. martn - moruno d * 17 * 2269 ( 2008 ) albarran bouhmadi - lpez kiefer marto v. moniz arxiv:1604.08365 | 0.427488 | 2,865 | 1,476 | , Saving $0.1 in GPT-4. |
20,741 | we prove new estimates for the volume of a lorentzian manifold and show especially that cosmological spacetimes with crushing singularities have finite volume . | let @xmath0 be a @xmath1-dimensional lorentzian manifold and suppose that @xmath0 can be decomposed in the form @xmath2 where @xmath3 has finite volume and @xmath4 resp .
@xmath5 represent the critical past resp .
future cauchy developments with not necessarily a priori bounded volume .
we assume that @xmath5 is the future cauchy development of a cauchy hypersurface @xmath6 , and @xmath4 the past cauchy development of a hypersurface @xmath7 , or , more precisely , we assume the existence of a time function @xmath8 , such that @xmath9),&\qq m_2=\{x^0=t_2\ } & , \end{aligned}\ ] ] and that the lorentz metric can be expressed as @xmath10 where @xmath11 are local coordinates for the space - like hypersurface @xmath6 if @xmath5 is considered resp .
@xmath7 in case of @xmath4 .
the coordinate system @xmath12 is supposed to be future directed , i.e. the directed unit normal @xmath13 of the level sets @xmath14 is of the form @xmath15 if we assume the mean curvature of the slices @xmath16 with respect to the past directed normal cf . for a more detailed explanation of our conventions is strictly bounded away from zero , then , the following volume estimates can be proved suppose there exists a positive constant @xmath17 such that @xmath18 then @xmath19 these estimates also hold locally , i.e. if @xmath20 , @xmath21 , are measurable subsets and @xmath22 the corresponding future resp .
past directed cylinders , then , @xmath23
in the following we shall only prove the estimate for @xmath5 , since the other case @xmath4 can easily be considered as a future development by reversing the time direction .
let @xmath24 be an embedding of a space - like hypersurface and @xmath13 be the past directed normal .
then , we have the gau formula @xmath25 where @xmath26 is the second fundamental form , and the weingarten equation @xmath27 we emphasize that covariant derivatives , indicated simply by indices , are always tensors . the slices @xmath16 can be viewed as special embeddings of the form @xmath28 where @xmath29 are coordinates of the slice @xmath30 .
hence , the slices @xmath16 can be considered as the solution of the evolution problem @xmath31 with initial hypersurface @xmath30 , in view of . from the equation we can immediately derive evolution equations for the geometric quantities @xmath32 , and @xmath33 of @xmath16 , cf .
e.g. , where the corresponding evolution equations are derived in riemannian space . for our purpose , we are only interested in the evolution equation for the metric , and we deduce @xmath34 in view of the weingarten equation .
let @xmath35 , then , @xmath36 and thus , the volume of @xmath37 , evolves according to @xmath38 where we shall assume without loss of generality that @xmath39 is finite , otherwise , we replace @xmath30 by an arbitrary measurable subset of @xmath30 with finite volume
. now , let @xmath40 be arbitrary and denote by @xmath41 the cylinder @xmath42 then , @xmath43 where we omit the volume elements , and where , @xmath44 . by assumption , the mean curvature @xmath45 of the slices is bounded from below by @xmath17 , and we conclude further , with the help of , @xmath46 letting @xmath47 tend to @xmath48
gives the estimate for @xmath49 . to prove the estimate , we simply replace @xmath30 by @xmath50 .
if we relax the conditions and to include the case @xmath51 , a volume estimate is still possible .
if the assumptions of are valid with @xmath51 , and if in addition the length of any future directed curve starting from @xmath30 is bounded by a constant @xmath52 and the length of past any directed curve starting from @xmath53 is bounded by a constant @xmath54 , then , @xmath55 as before , we only consider the estimate for @xmath5 . from we
infer that the volume element of the slices @xmath16 is decreasing in @xmath56 , and hence , @xmath57 furthermore , for fixed @xmath58 and @xmath59 @xmath60 because the left - hand side is the length of the future directed curve @xmath61 let us now look at the cylinder @xmath41 as in and .
we have @xmath62 & \le \ga_1\int_{m(t_1)}\sqrt{g(t_1,x)}=\ga_1\abs{m(t_1 ) } \end{aligned}\ ] ] by applying fubini s theorem and the estimates and .
a cosmological spacetime is a globally hyperbolic lorentzian manifold @xmath0 with compact cauchy hypersurface @xmath63 , that satisfies the timelike convergence condition , i.e. @xmath64 if there exist crushing singularities , see or for a definition , then , we proved in that @xmath0 can be foliated by spacelike hypersurfaces @xmath65 of constant mean curvature @xmath66 , @xmath67 , @xmath68}m(\tau)\uu{\msc c}_0,\ ] ] where @xmath69 consists either of a single maximal slice or of a whole continuum of maximal slices in which case the metric is stationary in @xmath70 .
but in any case @xmath69 is a compact subset of @xmath0 . in the complement of @xmath69
the mean curvature function @xmath66 is a regular function with non - vanishing gradient that can be used as a new time function , cf . for a simple proof .
thus , the lorentz metric can be expressed in gaussian coordinates @xmath71 with @xmath72 as in .
we choose arbitrary @xmath73 and define @xmath74 then , @xmath3 is compact , and the volumes of @xmath75 can be estimated by @xmath76 hence , we have proved a cosmological spacetime @xmath0 with crushing singularities has finite volume .
let @xmath0 be a spacetime with compact cauchy hypersurface and suppose that a subset @xmath77 is foliated by constant mean curvature slices @xmath65 such that @xmath78 and suppose furthermore , that @xmath72 is a time function which will be the case if the timelike convergence condition is satisfied so that the metric can be represented in gaussian coordinates @xmath71 with @xmath72 .
consider the cylinder @xmath79 for some fixed @xmath66 .
then , @xmath80 and we obtain in view of @xmath81 and conclude further @xmath82\abs{m(\tau)}\le \tau_2\abs{n_-}+\abs{m(\tau_2)},\ ] ] i.e. @xmath82\abs{m(\tau)}=\un\im \abs{n_-}=\un.\ ] ]
suppose that @xmath0 is a riemannian manifold that is decomposed as in with metric | @xmath0 @xmath1-dimensional lorentzian manifold decomposed form @xmath2 @xmath3 finite volume @xmath4 @xmath5 represent past future cauchy developments not a priori bounded volume @xmath5 future development hypersurface @xmath6 @xmath4 past development hypersurface @xmath7 existence time function @xmath8 @xmath9),_2={x^0=t_2{aligned lorentz metric expressed as @xmath10 @xmath11 local coordinates for space - like hypersurface @xmath6 @xmath5 @xmath7 @xmath4 coordinate system @xmath12 future directed directed unit normal @xmath13 level sets @xmath14 form @xmath15 mean curvature of slices @xmath16 past directed normal bounded from zero volume estimates proved positive constant @xmath17 @xmath18 @xmath19 estimates hold locally @xmath20 @xmath21 measurable subsets @xmath22 future past directed cylinders @xmath23 prove estimate for @xmath5 other case @xmath4 future development reversing time direction @xmath24 embedding space - like hypersurface @xmath13 past directed normalgau formula @xmath25 @xmath26 second fundamental form weingarten equation @xmath27 covariant derivatives indices tensors slices @xmath16 special embeddings of form @xmath28 @xmath29 coordinates slice @xmath30 slices @xmath16 solution of evolution problem @xmath31 initial hypersurface @xmath30 equation derive evolution equations for geometric quantities @xmath32 @xmath33 @xmath16 evolution equations derived riemannian space interested in evolution equation for metric deduce @xmath34 weingarten equation @xmath35 @xmath36 volume @xmath37 evolves according to @xmath38 assume @xmath39 finite replace @xmath30 by arbitrary subset @xmath30 finite volume @xmath40 arbitrary denote @xmath41 cylinder @xmath42 @xmath43 omit volume elements @xmath44 mean curvature @xmath45 of slices bounded by @xmath17 conclude @xmath46 @xmath47 tend @xmath48 estimate for @xmath49 prove estimate replace @xmath30 by @xmath50 relax conditions include case @xmath51 volume estimate possibleassumptions valid with @xmath51 future curve from @xmath30 bounded by constant @xmath52 past curve from @xmath53 constant @xmath54 @xmath55 consider estimate for @xmath5 volume element of slices @xmath16 decreasing in @xmath56 @xmath57 fixed @xmath58 @xmath59 @xmath60 left - hand side length future curve @xmath61 look at cylinder @xmath41 @xmath62\ga_1\int_{m(t_1)}=\ga_1\abs{m(t_1 applying fubini s theorem estimates cosmological spacetime globally hyperbolic lorentzian manifold @xmath0 compact cauchy hypersurface @xmath63 satisfies timelike convergence condition @xmath64 crushing singularities @xmath0 foliated by spacelike hypersurfaces @xmath65 constant mean curvature @xmath66 @xmath67 @xmath68 @xmath69 single maximal slice or continuum slices metric stationary in @xmath70 @xmath69 compact subset of @xmath0 mean curvature function @xmath66 regular function non - vanishing gradient new time functionsimple proof lorentz metric expressed gaussian coordinates @xmath71 @xmath72 choose arbitrary @xmath73 define @xmath74 @xmath3 compact volumes @xmath75 estimated @xmath76 proved cosmological spacetime @xmath0 singularities finite volume @xmath0 spacetime compact cauchy hypersurface subset @xmath77 foliated constant curvature slices @xmath65 @xmath78 @xmath72 time function timelike convergence condition satisfied metric represented gaussian coordinates @xmath71 @xmath72 cylinder @xmath79 fixed @xmath66 @xmath80 @xmath81 conclude @xmath82\abs{m(\tau)}\tau_2\abs{n_-} @xmath82\abs{m\abs{n_-} @xmath0 riemannian manifold decomposed metric | 0.258829 | 1,670 | 905 | , Saving $0.0 in GPT-4. |
20,744 | the entanglement of two atoms is studied when the two atoms are coupled to a single - mode thermal field with different couplings .
the different couplings of two atoms are in favor of entanglement preparation : it not only makes the case of absence entanglement with same coupling appear entanglement , but also enhances the entanglement with the increasing of the relative difference of two couplings .
we also show that the diversity of coupling can improved the critical temperature . if the optical cavity is leaky during the time evolution , the dissipative thermal environment is benefit to produce the entanglement .
pacs numbers : 03.67-a , 03.67.mn , 42.50.-p | environment noise can result in decoherence of quantum system . in order to overcome the decoherence , some authors proposed a number of approaches such as a quantum error correction@xcite and a decoherence - free subspace@xcite .
another direct method is to employ the environment noise to produce entanglement [ 3 - 12 ] . in solid state system ,
the thermal entanglement in heisenberg model is a example of using the thermal environment [ 3 - 6 ] .
it has been found that the anisotropy in xy heisenberg chain can be used to increase the critical temperature@xcite . in cavity qed , the cavity loss or the thermal field
can also induce entangment between two atoms [ 7 - 11 ] . in ref.@xcite , using quantum jump approach , the authors described a scenario where entanglement between two atomic systems can be induced via continous observation of the cavity loss . working in different initinal state ( different from ref .
@xcite ) and employing a master equation , in ref .
@xcite , the author proposed a sheme to prepare two atomic entanglement .
two two - level atoms can be entangled through interacting with a single - mode thermal field @xcite @xcite .
but in the studies on cavity qed@xcite@xcite @xcite , for simplicity of the calculation , the couplings @xmath0 between the two atoms and the field are assumed equal .
although , in ref.@xcite@xcite , their calculation also applied to @xmath1 , they just had discussed the equal couplings in their plots and had not discussed the effect of two atoms with different couplings .
in fact , the coupling rate @xmath2 depends on the atom position @xmath3 ; due to the randomness of the atom position @xmath3 , it is very difficult to control the same couplings between different atom @xcite .
the truth that the anisotropy in two - qubit heisenberg model can increase the critical temperature motivate us to consider whether different couplings @xmath2 of the two atoms result in some novel properties in cavity qed . in this paper , we study the system that the two atoms with different coupling constants are coupled to the optical cavity which is initinally in thermal field state . when the atoms are initially in @xmath4 and the cavity is initially in a single - mode thermal field , with the same coupling constant they are not entangled @xcite but if with different coupling constant they can be entangled .
we analyse the dependence of the relative difference of two couplings on the entanglement .
we find that there is a possiblity to obtian more entanglement through increasing the relative difference of two couplings .
we show that the diversity of coupling can improved the critical temperature at which entanglement disappears .
we also study the case of a cavity with a steady leak during the time evolution , it is found that the dissipative cavity field is benefit to produce the entanglement .
in order to make clear the function of different couplings , we first consider a quantum system composed of two two - level atoms interacting with a single - mode thermal field , which can be produced by a leaky cavity in thermal equilibrium at temperature @xmath5 .
after preparation the initial cavity field , the cavity stop to leak .
the cavity mode is assumed to be resonant with the atomic transition frequency . under the rotating wave approximation , the hamiltonian in the interacting picture is @xmath6 where @xmath7 @xmath8 denotes annihilation ( creation ) operator of the cavity field ; the atomic transition operators are @xmath9 and @xmath10 @xmath11 and @xmath12 are the two coupling constants for the atom @xmath13 @xmath14 . for expression the diversity of the two couplings ,
we let @xmath15 where @xmath2 denote average coupling and @xmath16 express the relative difference of the two atomic couplings . because of the identity of the two atoms , the range of @xmath16 is between @xmath17 and @xmath18 .
the hamiltonian can be rewritten as @xmath19 the single - mode thermal field with its mean photon number @xmath20 is a mixture of fock states .
it take the form @xmath21 and @xmath22 relates to the environment temperature .
entanglement of two atoms can be measured by concurrence @xmath23 which is written as@xcite@xcite @xmath24 where @xmath25 is the square roots of the eigenvalues of the matrix @xmath26 , @xmath27 is the density matrix , @xmath28 and @xmath29 stands for complex conjugate .
the concurrence is available no matter what @xmath27 is pure or mixed .
we will numerical calculate the entanglement between the two atoms . in ref .
@xcite@xcite , for simplicity they assumed the atoms 1 and 2 couple to a single - mode thermal field with the same coupling constant .
they found that if the two atoms initially are both in excited states , the two atoms could not be entangled neither in one - photon process nor in two - photon process . in our simulation , if @xmath30 we also can not find entanglement .
but when we chose @xmath31 , the result is reversed which is shown in fig .
it is obvious that under the condition of different couplings the two atoms ( initial in @xmath4 ) can be entangled .
this is contrary to the case of the same coupling constants@xcite@xcite . when the two atoms initially are both in excited state , it had been shown that the atom and the field are always entangled despite of the presence of the other atom@xcite .
it is the atom - field entanglement that result in the decoherence between the two atoms
. we can understand it from the same couplings : becasuse their couplings are the same , their rabi oscillation also have the same steps , so the two atoms have no correlation , i.e. , there is no entanglement between them
. if the couplings are not the same , when they resonate with the cavity , they are no longer with the same step . through interacting with cavity ,
the two atoms are entangled . as we state above ,
the different couplings are more closely to experiments .
so , the existing entanglement with different couplings maybe more easy achieved .
also show that the vacuum state ( @xmath32 ) induces the maximum entanglement . with the increase of the mean photon number , the entanglement between two atoms decreases . at a certain value of @xmath20
, the entanglement disappears ; we call that @xmath20 as critical average photon number and the temperature ( corresponding to the @xmath33 is called as critical temperature . increasing @xmath16 to @xmath34 , we plot the entanglement of two - atom in fig.1b . comparing fig.1a with fig.1b
, one can find that the maximum entanglement of fig .
1b is larger than that of fig .
1a , that is , one can obtain much more strong entanglement with @xmath35 . on the other hand , in fig .
1a the critical average photon number is below 4 and in fig.1b when @xmath36 , the maximum entanglement is about 0.15 , that is , the critical average photon number is larger than 5 .
so , the critical temperature of fig .
1b is larger than that of fig.1a .
whether can we conclude from fig.1 that the larger the relative difference of two couplings , the larger the critical average photon number , the larger the critical temperature ?
the answer is no . because the rabi oscillation of the atoms is relevant to the couplings , the entanglement is not a monotonic function of @xmath16 .
it will be shown in fig . 2 and we will analyse it laterly .
but at least in some ranges we can say that through increasing @xmath16 , the critical temperature can be improved .
the conclusion , in a certain extent , is similar to that the anisotropy in heisenberg xy chain can increase critical temperature of the thermal entanglement @xcite .
the difference is that there , the critical temperature of the thermal entanglement is monotonously increased with the increasing of the anisotropy parameter , here due to the rabi oscillation , the critical temperature is not a monotonous function of the two couplings . but with the increasing of @xmath16 , there are the possiblity and the tendency to increasing the critical temperature ( we will explain it next ) .
because every atom periodically entangles and disentangles with the cavity field , and the periodicity relates to the coupling , so the entanglement between two atoms also exhibit periodicity and its periods are also related to the two couplings .
we can observe this periodicity in fig.1 .
for make clear the dependence of the entanglement on @xmath37 we plot the entanglement as a function of @xmath16 and @xmath38 in fig . 2 for the atomic initial state @xmath4 , where the average photon number @xmath39 .
we can see that for @xmath40 no entanglement exist which is coincident with the ref.@xcite .
but with the increasing of @xmath16 , in som time intervals one can see the large entanglement . of cause , due to existing many peaks ( existing up the hill and down the hill ) , there are either the extent in which entanglement is increased with the increasing of @xmath16 or the extent in which the entanglement is decreased with the increasing of @xmath41 although the relation between the entanglement and @xmath16 is not monotonous , the main trend of entanglement is increased with the increasing of @xmath16 . in other word , the maxima values of entanglement
is increased with the increasing of @xmath42 thus , if we can keep the relative large @xmath43 , even the two atoms interact with thermal field , we can also obtain relative large entanglement . on the other hand ,
since the average photon number is a measure of the cavity classicality , the larger it is , the smaller the entanglement , i.e. , with the increase of the mean photon number , the entanglement gradually decreases to zero .
if in the extent in which entanglement is increased with the increasing of @xmath37 we can reckon the critical average photon number is increased with the increasing of @xmath43 , thus the critical temperature is increased . if the span of @xmath44 is relative large such as @xmath16 from 0.4 to 0.8 , the maxima entanglement is increased , the critical temperature , of cause , is increased .
the two atoms with initial state @xmath4 , which can not be entangled if they have the same coupling constants @xcite@xcite , can be entangled when they are different in coupling constant .
the initial state latexmath:[$% relative best but is far smaller than 1@xcite@xcite . if the coupling constants are different , what will happen ?
we directly numerical simulate the entanglement as a function of @xmath16 and @xmath38 in fig .
3 . in fig . 2 and fig .
3 , we both can see that when @xmath46 the entanglement do not exist . for @xmath46 ,
only the one atom interact with the cavity and the other atom has no coupling with the field , so , there is no entanglement between them . when the two - atom is initially in the state @xmath47 with @xmath48 is exact the case which had been discussed in ref.@xcite . from fig .
3 , we can observe entanglement when @xmath30 and it is different from the case of initial state @xmath4 with @xmath40 .
but the entanglement in fig .
3 when @xmath40 is not the largest . with the increasing of @xmath37 the maximum values of entanglement
are also increased .
there are still existing many
hills , that is , there are either the extent in which entanglement is increased with the increasing of @xmath43 or the extent in which the entanglement is decreased with the increasing of @xmath42 therefore , we can also obtain strong entanglement and improve the critical temperature through increasing @xmath16 in some extent .
thus , although fig .
3 is obvirously different from fig .
2 , the effects of @xmath16 on increasing critical temperature and improved entanglement are exactly the same in these two kinds of initial states .
when the single - mode cavity is in thermal equilibrium with its environment due to the leakage , the cavity is in thermal field @xcite . in section @xmath49 and in ref.@xcite@xcite ,
the thermal field is just as initial cavity state ; in the process of evolution , the cavity stop leaking . since the cavity is leaky , why we assumed it stop to leaking ?
now , we let it continue to leak in the later time evolution .
the master equation governing the time evolution of the global system is given by ( @xmath50 ) @xmath51+{\cal l}(\rho ) , \ ] ] where the hamiltonian still have the form of eq.(3 ) .
the liouvillean is given by @xmath52 we chose @xmath53 , @xmath31 and simulate the case of which the two atoms are initially in @xmath4 and the cavity is initially in the thermal state .
figure 4 show the entanglement as a function of average photon number and time @xmath38 .
we notice that the critical average photon number is larger than 5 , and in fig
. 1 @xmath20 is smaller than 5 . to a thermal field state
, the dissipative environment can increase the critical average photon number .
furthermore , with the time evolution , the entanglement is no longer periodic appearing zero but gradually reaches their asymptotic values , and the asymptotic values decrease with the increasing of @xmath20 .
when the cavity field initially is in thermal state , the dissipation of the cavity virtually can be considered as a driver of the cavity , so , with the evolution the entanglement is gradually increased .
this is contrary to the general case in which the entanglement is destroyed by environment due to the decoherence .
after all the system is a open one , so the entanglement is no longer periodic appearing zero but gradually reaches their asymptotic values .
so , after long time evolution , we can obtain a strong and steady entanglement .
therefore , as long as the evolution time is enough , it is not necessary to precise control the interaction time . in experiment , precise control of the interacting time is very difficult to achieve @xcite .
we summarize that the dissipative environment not only improved th critical temperature but also provide us a steady entanglement .
we discuss the entanglement induced by a single - mode heat environment when the two atoms are coupled to the optical cavity with different coupling constants . when the atoms are initially in @xmath4 , with the same coupling constant they could not be entangled @xcite but if with different coupling constant they can be entangled
. even to the initial state @xmath47 in which the two atoms can be entangled with the same coupling , the different couplings are avail to produce atom - atom entanglement . through the analysis about the dependence of entanglement on the relative difference of two couplings @xmath16
, we find that by increasing @xmath16 we can obtain strong entanglement .
we also show that the diversity of coupling can improved the critical temperature , which is very similar to that the anisotropy in heisenberg xy chain can increase the critical temperature of thermal entanglement @xcite .
we study the leak of the cavity on the entanglement during the time evolution .
it is found that the keeping on leaking is benefit to produce the entanglement .
it not only can improves the critical temperature but also provid us a relative steady and strong entangled state .
entanglement as a function of average photon number @xmath20 and @xmath38 when the pair of atoms is initially prepared in the state @xmath4and the field is initially in the thermal state , where ( a ) : @xmath54 ( b):@xmath55 ; for all plots @xmath56 . | environment noise in decoherence quantum system authors proposed approaches quantum error correction decoherence - free subspace@xcite method environment noise entanglement [ 3 - 12 ] in solid state system thermal entanglement in heisenberg model example thermal environment [ 3 - 6 ] anisotropy in xy heisenberg chain critical temperature in cavity qed cavity loss thermal field induce entangment between two atoms [ 7 - 11 ] in ref.@xcite quantum jump approach authors described scenario entanglement between two atomic systems induced via continous observation of cavity loss author proposed sheme two atomic entanglement two level atoms entangled single - mode thermal field studies on cavity qed couplings between two atoms field assumed equal ref.@xcite calculation applied to @xmath1 discussed equal couplings not effect of two atoms with different couplings coupling rate depends on atom position @xmath3 difficult to control same couplings between different atom @xciteanisotropy in two - qubit heisenberg model critical temperature different couplings novel properties in cavity qed study two atoms with different coupling constants coupled to optical cavity in thermal field state atoms in cavity in single - mode thermal field same coupling constant not entangled different can be entangled dependence relative difference couplings on entanglement more entanglement increasing difference diversity of coupling critical temperature entanglement cavity with steady leak during time evolution dissipative cavity field entanglement function different couplings consider quantum system two - level atoms interacting with single - mode thermal field produced by leaky cavity in thermal equilibrium at temperature @xmath5 cavity leak cavity mode resonant with atomic transition frequency rotating wave approximation hamiltonian in interacting picture is @xmath6 @xmath7 @xmath8 annihilation ) operator of cavity field atomic transition operators are @xmath9 and @xmath10 @xmath11 @xmath12 two coupling constants for atom @xmath13 @xmath14expression diversity two couplings @xmath15 @xmath2 average coupling @xmath16 express relative difference atomic couplings identity two atoms range @xmath16 between @xmath17 and @xmath18 hamiltonian rewritten @xmath19 single - mode thermal field mean photon number @xmath20 mixture fock states form @xmath21 @xmath22 environment temperature entanglement two atoms measured by concurrence @xmath23 written as@xcite @xmath24 @xmath25 roots eigenvalues matrix @xmath26 @xmath27 density matrix @xmath28 @xmath29 complex conjugate concurrence available @xmath27 pure or mixed numerical calculate entanglement two atoms @xcite assumed atoms 1 2 couple single - mode thermal field same coupling constant two atoms excited states entangled one - photon process two - photon process simulation @xmath30 find entanglement chose @xmath31 result reversed shown fig . different couplings two atoms @xmath4 entangled contrary same coupling constants@xcite two atoms excited state atom field always entangled despite presence other atom atom - field entanglement decoherence between two atomsunderstand from same couplings rabi oscillation same steps two atoms no correlation no entanglement if couplings not same resonate with cavity no longer same step interacting with cavity two atoms entangled different couplings closely to experiments existing entanglement with different couplings easy achieved vacuum state ( @xmath32 induces maximum entanglement increase mean photon number entanglement decreases at value @xmath20 entanglement disappears critical average photon number temperature @xmath33 critical temperature increasing @xmath16 to @xmath34 plot entanglement two atom in fig.1b maximum entanglement of fig 1b larger than 1a more strong entanglement with @xmath35 fig . 1a critical average photon number below 4 fig.1b @xmath36 maximum entanglement about 0.15 critical average photon number larger than 5 critical temperature of fig . 1b larger than fig.1a conclude larger relative difference of couplings larger critical average photon number larger critical temperature no rabi oscillation relevant to couplings entanglement not monotonic function of @xmath16 shown in fig .analyse laterly some ranges increasing @xmath16 critical temperature conclusion similar to anisotropy in heisenberg xy chain critical temperature thermal entanglement @xcite difference critical temperature monotonously increased with increasing anisotropy parameter due to rabi oscillation critical temperature not monotonous function of two couplings increasing @xmath16 possiblity tendency increasing critical temperature explain next atom periodically entangles disentangles with cavity field periodicity relates to coupling entanglement between two atoms periodicity related to couplings observe periodicity in fig.1 dependence entanglement on @xmath37 plot entanglement as function of @xmath16 and @xmath38 in fig . 2 atomic initial state @xmath4 average photon number @xmath39 for @xmath40 no entanglement coincident with ref.@xcite increasing @xmath16 intervals large entanglement due to many peaks extent entanglement increased with increasing @xmath16 or decreased with increasing @xmath41 relation @xmath16 not monotonous main trend of entanglement increased with increasing @xmath16maxima values entanglement increased with @xmath42 if keep large @xmath43 two atoms interact with thermal field obtain large entanglement average photon number measure cavity classicality larger smaller entanglement increase mean photon number entanglement decreases to zero if entanglement increased with @xmath37 critical average photon number increased with @xmath43 critical temperature increased if span @xmath44 large @xmath16 from 0.4 to 0.8 maxima entanglement increased critical temperature increased two atoms with initial state @xmath4 not be entangled same coupling constants can be entangled when different coupling constant initial state latexmath:[$% smaller than 1@xcite@xcite if coupling constants different numerical simulate entanglement function of @xmath16 and @xmath38 in fig . 3 @xmath46 entanglement one atom with cavity other no coupling field no entanglement two - atom initially state @xmath47 with @xmath48 case discussed in ref.@xcite fig . 3 entanglement when @xmath30 different from initial state @xmath4 with @xmath40 entanglement in fig @xmath40 not largest increasing @xmath37 maximum values entanglement increased many hills entanglement increased @xmath43 or decreased @xmath42 obtain strong entanglement improve critical temperature increasing @xmath16 fig . 3 different from fig 2 effects @xmath16 critical temperature improved entanglement same in initial states single - mode cavity in thermal equilibrium due to leakage thermal field @xcite @xmath49 ref.@xcite thermal field initial cavity state evolution cavity leaking cavity leaky continue leak later time evolution master equation time evolution global system given by @xmath50 ) @xmath51+{\cal l}(\rho ) hamiltonian form eq.(3 ) liouvillean by @xmath52 chose @xmath53 @xmath31 simulate case two atoms initially in @xmath4 cavity in thermal state figure 4 entanglement function of average photon number time @xmath38 critical average photon number larger than 5 fig . 1 @xmath20 smaller than 5 thermal field state dissipative environment increase critical average photon numberwith time evolution entanglement no longer periodic zero gradually reaches asymptotic values decrease with increasing @xmath20 cavity field in thermal state dissipation driver cavity with evolution entanglement gradually increased contrary to case entanglement destroyed by environment due to decoherence system is open entanglement no longer periodic zero reaches asymptotic values after long time evolution obtain strong steady entanglement evolution time enough not necessary to precise control interaction time precise control time difficult to achieve dissipative environment improved critical temperature steady entanglement entanglement induced by single - mode heat environment when two atoms coupled to optical cavity with different coupling constants atoms initially in @xmath4 same coupling constant not entangled different coupling constant can be entangled even initial state @xmath47 different couplings produce atom - atom entanglement analysis on difference couplings @xmath16 by increasing @xmath16 obtain strong entanglement diversity of coupling critical temperature similar to anisotropy in heisenberg xy chain critical temperature of thermal entanglementstudy leak cavity entanglement during time evolution found leaking entanglement improves critical temperature steady strong entangled state entanglement function average photon number @xmath20 @xmath38 atoms initially prepared state @xmath4and field thermal state ( a ) : @xmath54 b):@xmath55 all plots @xmath56 | 0.563735 | 3,572 | 1,850 | , Saving $0.1 in GPT-4. |
20,749 | the argoneut experiment features a 175 liter liquid argon ( lar ) time projection chamber ( tpc ) that was located upstream of the minos near detector in the numi neutrino beam at fermilab .
the project is part of the lar tpc development program in the us and has helped initiate the development of simulation and reconstruction tools for lar tpcs .
in addition to its development goals , argoneut will perform several cross - section measurements on ar in the few - gev energy range .
a total of 1.35e20 protons on target were accumulated and data analysis is ongoing .
i will review the experiment and its status , as well as preliminary results from the data analysis . | liquid argon ( lar ) time projection chambers ( tpcs ) seem to be a very promising technology for the next generation of neutrino detectors .
this next generation will require multi - kilo - ton detectors , and lar tpcs are an attractive option .
these detectors offer high detection efficiencies and excellent background rejection and they appear to be scalable .
however , the scalability has not yet been fully demonstrated .
the argoneut experiment , a @xmath0 lar tpc was a first step to a @xmath1 program in the united states .
the main goal of argoneut was to gain experience building and operating lar tpcs .
in addition , the experiment collected data in an energy range relevant to neutrino oscillation physics . these data were used to develop simulation and analysis tools and the first physics results extracted from the data were recently presented .
a review of the argoneut experiment will be given as well as a description of the data analysis and an overview of the first results on @xmath2 charged - current ( cc ) inclusive differential cross - section measurement by argoneut .
the argoneut tpc is a 90 cm @xmath3 40 cm @xmath3 48 cm rectangular volume containing @xmath0 of lar active volume .
the tpc was positioned in a vacuum jacketed cryostat .
left picture in figure [ fig1 ] shows the tpc about to enter the cryostat .
the tpc consists of three wire planes with 240 wires each , separated by 4 mm .
the innermost wire plane has vertical wire orientation and was not instrumented .
the second plane , the instrumented induction plane , has wires oriented at @xmath4 and the third plane , the collection plane , has wires oriented at @xmath5 .
the wire planes can be seen in the left panel of figure [ fig1 ] on the right of the tpc .
a 25kv voltage was applied to the cathode , creating a 500v / cm electric field .
more details on the detector , its electronics and the cryogenics can be found in @xcite .
+ [ cols="^,^ " , ]
argoneut ran stably for over five months , collecting an impressive sample of neutrino and anti - neutrino data .
the two weeks of neutrino data has started to be analyzed and the simulations predictions for the @xmath2 cc - inclusive differential cross section were reported here .
more results will be presented in a paper in preparation .
future analysis will allow channel - specific cross - section measurements .
the sample of anti - neutrino data , containing about fifteen times more statistics , will also be analyzed in the near future .
+ this measurement also demonstrates that the larsoft software is operational and ready for further analyses .
the argoneut data will also be used to look in details at fsis to study nuclear effects .
+ 9 m. soderberg , argoneut : a liquid argon time projection chamber test in the numi beamline , dpf-2009 , detroit , mi , july 2009 , econf c090726 .
i. ambats et al . , the minos detectors technical design report , numi - l-337 , 1998 .
josh spitz , measuring muon - neutrino charged - current differential cross sections with a liquid argon time projection chamber , phd dissertation , yale university , 2011 . | liquid argon lar ) time projection chambers promising technology for next generation neutrino detectors multi - kilo - ton detectors lar tpcs attractive option detectors offer high detection efficiencies excellent background rejection scalable scalability not fully demonstrated argoneut experiment @xmath0 lar tpc first step to @xmath1 program united states goal gain experience building operating lar tpcs experiment collected data relevant to neutrino oscillation physics data used simulation analysis tools first physics results recently presented review of argoneut experiment description data analysis overview first results on @xmath2 charged - current inclusive differential cross - section measurement by argoneut argoneut tpc 90 cm 40 cm 48 cm rectangular volume containing @xmath0 lar active volume tpc in vacuum jacketed cryostat tpc three wire planes 240 wires each separated by 4 mm innermost wire plane vertical wire orientation not instrumented second plane at @xmath4 third plane at @xmath5 25kv voltage applied to cathode creating 500v / cm electric fielddetails on detector electronics cryogenics in @xcite argoneut ran five months neutrino anti - neutrino data two weeks neutrino data analyzed predictions @xmath2 cc - inclusive differential cross section reported more results in paper future analysis channel - specific cross - section measurements anti - neutrino data fifteen times more statistics analyzed measurement larsoft software operational ready for analyses argoneut data used nuclear effects soderberg argoneut liquid argon time projection chamber test numi beamline dpf-2009 detroit july 2009 ambats minos detectors technical design report - 1998 josh spitz measuring muon - neutrino - current differential cross sections liquid argon time projection chamber phd dissertation yale university 2011 | 0.138461 | 726 | 360 | , Saving $0.0 in GPT-4. |
20,753 | recently nikodem , rajba and wsowicz compared the classes of @xmath0-wright - convex functions and @xmath0-jensen - convex functions by showing that the first one is a proper subclass of the latter one , whenever @xmath0 is an odd natural number . till now
the case of even @xmath0 was an open problem . in this paper
the complete solution is given : it is shown that the inclusion is proper for any natural @xmath0 .
the classes of strongly @xmath0-wright - convex and strongly @xmath0-jensen - convex functions are also compared ( with the same assertion ) . | let @xmath1 be an interval and @xmath2 . the _ difference operator _
is given by @xmath3 for any @xmath4 and @xmath5 @xmath6 . in 1954
wright @xcite considered the class of functions @xmath7 , for which the function @xmath8 is non - decreasing for any @xmath9 small enough to guarantee that all the involved arguments belong to @xmath10 .
this monotonicity property is equivalent to the condition @xmath11\big).\ ] ] nowadays a function fulfilling is called to be _
wright - convex_. for @xmath12 we get immediately @xmath13 so any wright - convex function is necessarily jensen - convex .
it was shown by ng @xcite that any wright - convex function is a sum of the convex function ( in the usual sense ) and the additive one .
such representation leads to the simple ( and now classical ) proof that the inclusion in question is proper .
namely , if @xmath14 is a discontinuous additive function and @xmath15 , then @xmath16 is discontinuous and jensen - convex .
hence the graph of @xmath16 is not dense on the whole plane .
but it is well - known ( @xcite ) that the graphs of discontinuous additive functions ( and also by ng s representation the graphs of discontinuous wright - convex functions ) mapping @xmath17 into @xmath17 are dense on the whole plane .
that is why @xmath16 is not wright - convex .
recently nikodem , rajba and wsowicz in the paper @xcite considered the related comparison problem for the classes of @xmath0-wright - convex functions and @xmath0-jensen - convex functions ( the explanation is given below in this section ) .
if @xmath18 is odd , they have shown that the first class is the proper subclass of the latter one .
the example of @xmath0-jensen - convex function which is not @xmath0-wright - convex was @xmath19 @xmath20 , where @xmath14 is the certain discontinuous additive function .
next ples @xcite proved that this example works for any discontinuous additive function .
the authors of @xcite did not solve the problem , if @xmath0 is even .
the objective of the present paper is to give the complete solution by showing that the inclusion is proper for all @xmath18 .
moreover , the construction of our example does not depend on the parity of @xmath0 .
then for odd @xmath0 we have a solution different from the ones given in the papers @xcite .
let us recall the concepts of higher - order jensen - convex ( coming from popoviciu @xcite ) and wright - convex functions ( introduced by gilnyi and ples @xcite ) .
the iterates of difference operator are we defined as usual : for @xmath18 @xmath21 provided that all the involved arguments belong to @xmath10 . if @xmath22 , then we write @xmath23 in particular , @xmath24 .
let @xmath18 and @xmath2 be a function .
we say that @xmath16 is _ wright - convex of order @xmath0 _ ( or @xmath0-_wright - convex _ for short ) , if @xmath25 for any @xmath4 and @xmath26 @xmath27 . if the above inequality is required only for @xmath28 , precisely , if @xmath29 for all @xmath30 and @xmath9 such that @xmath31 , then @xmath16 is called _ jensen - convex of order @xmath0 _ ( @xmath0-_jensen - convex _ for short ) .
it is easy to see that @xmath32-wright - convexity and @xmath32-jensen - convexity are equivalent to wright - convexity and jensen - convexity , respectively . by the above definitions any @xmath0-wright - convex function is necessarily @xmath0-jensen - convex . as we announced above , in this paper we show that this inclusion is proper .
let us start with two preparatory notes .
the following equations hold for a power function function @xmath33 ( ( * ? ? ?
* lemma 15.9.2 ) ) : @xmath34 the formula below was given in @xcite ( as a part of the proof of corollary 2.2 ) .
[ lemat2 ] let @xmath14 be an additive function , @xmath35 and @xmath18 .
then @xmath36 for any @xmath37 .
now we are in a position to state our main result .
[ th_main ] for any @xmath18 the class of @xmath0-wright - convex functions is a proper subclass of the class of @xmath0-jensen - convex functions . because every @xmath0-wright - convex function is @xmath0-jensen - convex
, it is enough to construct an @xmath0-jensen - convex function @xmath19 which is not @xmath0-wright convex . to this end
consider a hamel basis @xmath38 of the linear space @xmath17 over @xmath39 @xmath40 .
any @xmath41 is uniquely represented as a linear combination @xmath42 where @xmath43 , @xmath44 .
obviously the functions @xmath45 and @xmath46 are additive .
define the function @xmath19 by @xmath47 since the difference operator is additive , we infer by lemma [ lemat2 ] and the equation that @xmath48 in particular , @xmath49 let @xmath9 . because of the representation @xmath50 , it is easy to see that @xmath51 .
therefore @xmath52 which shows that @xmath16 is @xmath0-jensen - convex . to prove that @xmath16 is not @xmath0-wright - convex take now @xmath53 , @xmath54 and ( if @xmath55 ) @xmath56 .
then @xmath57 , @xmath58 .
moreover @xmath59 , @xmath60 , @xmath61 and @xmath62 .
by we obtain @xmath63 it means that @xmath16 is not @xmath0-wright - convex .
let @xmath18 and @xmath64 . a function @xmath2 is called _ strongly wright - convex of order @xmath0 with modulus @xmath65 _ ( or strongly @xmath0-wright - convex with modulus @xmath65 ) if @xmath66 for all @xmath30 and @xmath26 such that @xmath27 . if the above inequality is required only for @xmath28 , precisely , if @xmath67 for all @xmath30 and @xmath9 such that @xmath31 , then @xmath16 is called _ strongly jensen - convex of order @xmath0 with modulus @xmath65 _ ( or strongly @xmath0-jensen - convex with modulus @xmath65 ) ( @xcite ) . for @xmath68 we arrive at standard @xmath0-wright - convex and @xmath0-jensen - convex functions , respectively .
let @xmath18 and @xmath64 . the class of strongly @xmath0-wright - convex functions with modulus @xmath65 is a proper subclass of the class of strongly @xmath0-jensen - convex functions with modulus @xmath65 .
it is enough to show , that there exists a strongly @xmath0-jensen
convex function with modulus @xmath65 , which is not strongly @xmath0-wright convex with modulus @xmath65 . by theorem [ th_main ] there exists a function @xmath71 , which is @xmath0-jensen - convex and @xmath71 is not @xmath0-wright - convex .
let @xmath73 .
since @xmath71 is @xmath0-jensen - convex , it follows by theorem [ tw5 ] that @xmath16 is strongly @xmath0-jensen - convex with modulus @xmath65 . by theorem [ tw4 ]
@xmath16 is not strongly @xmath0-wright - convex with modulus @xmath65 ( because the function @xmath70 is not @xmath0-wright - convex ) .
this concludes the proof . | @xmath1 interval @xmath2 difference operator given by @xmath3 @xmath4 @xmath5 @xmath6 1954 wright @xcite considered class functions @xmath7 function @xmath8 non - decreasing for @xmath9 arguments belong to @xmath10 monotonicity property equivalent to condition @xmath11 function fulfilling wright - convex_ @xmath12 @xmath13 wright - convex function jensen - convex @xcite wright - convex function sum of convex function additive one representation leads proof inclusion proper if @xmath14 discontinuous additive function @xmath15 @xmath16 discontinuous jensen - convex graph of @xmath16 not dense whole plane graphs of discontinuous additive functions wright - convex functions mapping @xmath17 into @xmath17 dense whole plane @xmath16 not wright - convex nikodem rajba wsowicz @xcite considered comparison problem for @xmath0-wright - convex @xmath0-jensen - convex functions explanation if @xmath18 odd first class proper subclass of latterexample @xmath0-jensen - convex function not-wright - convex was @xmath19 @xmath20 @xmath14 discontinuous additive function @xcite proved example works for any discontinuous additive function authors solve problem if @xmath0 even objective present paper complete solution inclusion for all @xmath18 construction example depend on parity @xmath0 for odd @xmath0 solution different from papers recall concepts higher - order jensen - convex wright - convex functions by gilnyi ples iterates of difference operator defined @xmath18 @xmath21 arguments belong to @xmath10 if @xmath22 write @xmath23 @xmath24 @xmath18 @xmath2 function @xmath16 is wright - convex of order @xmath25 for @xmath4 @xmath26 @xmath27 if inequality required for @xmath28 @xmath29 for @xmath30 @xmath9 @xmath16 jensen - convex of order @xmath32-wright - convexity @xmath32-jensen - convexity equivalent to wright - convexity jensen - convexitydefinitions @xmath0-wright - convex function @xmath0-jensen - convex paper inclusion proper start preparatory notes equations hold power function function @xmath33 @xmath34 formula in @xcite proof corollary 2.2 @xmath14 additive function @xmath35 @xmath18 @xmath36 @xmath37 state main result @xmath18 @xmath0-wright - convex functions subclass @xmath0-jensen - convex functions-wright - convex function @xmath0-jensen - convex construct @xmath0-jensen - convex function @xmath19 not @xmath0-wright convex consider hamel basis @xmath38 linear space @xmath17 over @xmath39 @xmath40 @xmath41 represented linear combination @xmath42 @xmath43 @xmath44 functions @xmath45 @xmath46 additive define function @xmath19 by @xmath47 difference operator additive infer lemma equation @xmath48 @xmath49 @xmath9 representation @xmath50 see @xmath51 @xmath52 @xmath16 is @xmath0-jensen - convexprove @xmath16 not-wright - convex take @xmath53 @xmath57 @xmath59 obtain @xmath63 @xmath16 not @xmath0-wright - convex let @xmath18 @xmath64 function @xmath2 called strongly wright - convex @xmath0 modulus @xmath65-wright - convex modulus @xmath65 @xmath66 @xmath30 @xmath26 inequality required @xmath28 @xmath67 @xmath30 @xmath9 @xmath16 strongly jensen - convex @xmath0 modulus @xmath65-jensen - convex modulus @xmath68 standard @xmath0-wright - convex-jensen - convex functions @xmath18 @xmath64 class strongly @xmath0-wright - convex functions with modulus @xmath65 subclass @xmath0-jensen - convex functions modulus @xmath65 strongly @xmath0-jensen convex function with modulus @xmath65 not strongly @xmath0-wright convex modulus @xmath65theorem function @xmath71 @xmath0-jensen - convex not-wright - convex @xmath73 @xmath71-jensen - convex theorem @xmath16 @xmath0-jensen - convex modulus @xmath65 theorem tw4 @xmath16 not-wright - convex @xmath65 function @xmath70 not @xmath0-wright - convex concludes proof | 0.311774 | 1,905 | 1,010 | , Saving $0.1 in GPT-4. |
20,762 | to study the evolution of binary star clusters we have imaged 7 systems in the small magellanic cloud with soar 4-m telescope using b and v filters .
the sample contains pairs with well - separated components ( @xmath0pc ) as well as systems that apparently merged , as evidenced by their unusual structures . by employing isochrone fittings to their cmds
we have determined reddening , age and metallicity and by fitting king models to their radial stellar density profile we have estimated core radius .
disturbances of the density profile are interpreted as an evidence of interaction .
circunstances as distances between components and their age difference are addressed in terms of the timescales involved to access the physical connection of the system . in two cases
the age difference is above 50myr , which suggests chance alignment , capture or sequential star formation . | differently from the galaxy , the low - density environment of the magellanic clouds favors the gravitational interaction of single clusters , which form binary ( or multiple ) groupings and eventually will merge or disrupt .
pieces of evidence on this subject are : ( i ) many binary clusters in the magellanic clouds have been identified and recognized as true physical systems ( ( * ? ? ?
* bhatia & hatzidimitriou 1988 ) , ( * ? ? ?
* bhatia et al . 1991 ) ) ; ( ii ) deep and precise colour - magnitude diagrams ( cmds ) for isolated clusters in the large magellanic cloud revealed multi - population sequences ( ( * ? ? ?
* mackey et al . 2008 ) ) .
if these facts are connected their results indicate that we are witnessing different stages of cluster evolution in a scenario where cluster groupings and mergers are not uncommon .
evidence of cluster binarity and mergers has been reported for both clouds which are very rich in cluster pairs and multiplets ( ( * ? ? ?
* dieball et al . 2002 ) , ( * ? ? ?
* bica et al . 2008 ) , ( * ? ? ?
* carvalho et al . 2008 ) ) .
we determine astrophysical and structural parameters of smc binary clusters by fitting isochrones to the cmds and king functions to the radial density profiles with the purpose of providing constraints on the evolution of such systems .
the soar optical imager ( soi ) mounted in a bent - cassegrain configuration to the 4.1-m soar telescope ( cerro pachn , chile ) was employed to observe a sample of 7 merger candidates .
images in bessel _ bv _ filters were obtained on photometric nights ( nov / oct 2007 ) , using the soi mini - mosaic of two e2v 2x4k ccds ( 1 pixel = 15@xmath1 m ) to cover a 5.2x5.2(@xmath2)@xmath3 field of view .
the ccds were binned to 2x2 pixels yielding a scale of 0.154/pixel .
the average seeing was @xmath40.95 in _ b _ and @xmath40.8 in _ v_. two images in each filter were obtained , with single exposure times of 480s in _ b _ and 195s in _
v_. the spatial scale is 1=0.3pc .
the ccd frames were reduced with iraf software .
starfinder ( ( * ? ? ?
* diolaiti et al . 2000 ) ) was used to perform psf photometry .
table 1 presents the cluster sample and includes the parameters derived here as well as literature ones .
image of the interesting non - coeval pair ic1641/ngc422 is shown in fig
. 1 . -0.3 cm .cluster sample and derived parameters . [ cols="<,^,^,^,^,^,^,^,^,^ " , ]
we built the radial density profile ( rdp ) for each merger candidate component by employing star counts in circular rings around the cluster center . for the sake of uniformity , all rdps
are based on rings 5 wide . in order to get an optimized rdp , we limit the data to a threshold magnitude for which the difference between the cluster central density and the adjacent field density is maximum .
the cluster center is tweaked in the process .
all stars fainter than this threshold are then disregarded in the analysis .
we employed two - parameter ( central density @xmath5 , core radius @xmath6 ) king profiles ( ( * ? ? ?
* king 1966 ) ) to derive structural properties .
the background was determined by fitting a constant to the outermost 4 rings , which was then subtracted from the overall stellar density before the king profile fitting was performed .
1 shows the rdp for clusters ic1641/ngc422 .
we have obtained astrophysical parameters by fitting padova isochrones ( ( * ? ? ? * girardi et al . 2002 ) ) to the _ v _ @xmath7 ( _ b v _ ) cmds . in order to minimize the contamination of background stellar fields , a circular region defined by a limiting radius ( @xmath8 ) covering the cluster central area was selected .
@xmath8 is defined as the radius where the cluster stellar density begins to stand out from the background one .
for all clusters in the sample good fittings were obtained for ( _ m - m_)@xmath9 .
reddening as well as age and metallicity corresponding to the best matching isochrones are presented in table 1 .
1 shows isochrone fittings to the observed cmds of ic1641 and ngc422 .
in general , our age estimates agree with cmd based age determinations in the literature ( see table 1 ) .
the ages of l51/b78 and ngc422/ic1641 have been determined using isochrone fittings for the first time .
because of a disturbed surface brightness profile and isophotal features , ( * ? ? ? * carvalho ( 2008 ) ) classified k50 and ngc376 as mergers and the pairs ic1612e / w and ngc241/242 as true binary clusters .
except for ngc422/ic1641 and ic1612e / w , all cluster pairings in our sample are nearly coeval and have ages close to @xmath10yr , consistent with the time scales of dynamical models . ngc422/ic1641 and ic1612e / w have ages that differ by 390myr and 60myr respectively , implying that , if they are interacting , a capture process may have occurred .
another possibility is that the oldest cluster in the pair triggered the star formation that originated the youngest one via supernovae explosions and shell expansion .
such mechanism has also been proposed as a plausible explanation by ( * ? ? ? * vallenari ( 1998 ) ) for the lmc pair ngc1903/sl357 , although their age difference is 550myr . concerning the time scales involved
, the expected effect of the gravitational interaction for clusters having a mass of @xmath11m@xmath12 and half mass radius of 10pc , is that they will merge in @xmath10yr if their separation is @xmath425pc ( ( * ? ? ?
* sugimoto & makino 1989 ) ) .
models for less massive clusters in the lmc tidal field give a few @xmath10yr for the pair lifetime , with separations varying from @xmath46pc for cluster masses @xmath13m@xmath12 and @xmath414pc for @xmath14m@xmath12 ( ( * ? ? ?
* bhatia 1990 ) ) . according to these models , after this time scale either the cluster pair merges due to mutual gravitational interaction or disrupts by the galactic tidal field and , in a less probable event , by passing giant molecular clouds
. clusters resulting from pairs that have merged are expected to show structural properties which differ from those of single clusters .
the plot of @xmath8 against age indicates that k50 , ngc376 and ic1611 have a larger @xmath8 than the mean locus of single clusters , which may be an additional evidence for their merger status ( ( * ? ? ?
* mackey & gilmore 2003 ) ) .
we thank the brazilian institutions cnpq and fapemig for financial support . | from galaxy low - density environment magellanic clouds favors gravitational interaction of single clusters form binary groupings merge or disrupt evidence many binary clusters in magellanic clouds identified recognized as physical systems deep colour - magnitude diagrams for clusters in magellanic cloud revealed multi - population sequences mackey et al 2008 results indicate different stages of cluster evolution cluster groupings mergers not uncommon evidence of cluster binarity mergers reported for both clouds rich in cluster pairs multiplets dieball 2002 bica 2008 carvalho determine astrophysical structural parameters of smc binary clusters by fitting isochrones to cmds king functions to radial density profiles constraints on evolution soar optical imager soi ) 4.1-m soar telescope cerro employed sample of 7 merger candidates images in bessel _ filters obtained on photometric nights ( nov / oct 2007 using soi mini - mosaic of two e2v 2x4k ccds 1 pixel = 15@xmath1 m 5.2x5.2(@xmath2)@xmath3 field of view ccds binned to 2x2 pixels scale of 0.154/pixel .average seeing @xmath40.95 in _ b _ @xmath40.8 v_ two images each filter single exposure times 480s _ b _ 195s v_ spatial scale 1=0.3pc ccd frames reduced with iraf software starfinder used psf photometry table 1 presents cluster sample parameters derived literature ones image non - coeval pair ic1641/ngc422 fig . 1 -0.3 cm .cluster sample derived parameters built radial density profile each merger candidate component star counts circular rings cluster center rdps based on rings 5 wide optimized rdp limit data to threshold magnitude difference cluster central density adjacent field density maximum cluster center tweaked stars fainter than threshold disregarded employed two - parameter central density core radius @xmath6 king profiles structural properties background determined fitting constant to outermost 4 rings subtracted from overall stellar density before king profile fitting. 1 shows rdp for clusters ic1641/ngc422 obtained astrophysical parameters fitting padova isochrones to _ v @xmath7 b v cmds minimize contamination stellar fields circular region limiting radius @xmath8 cluster central area selected @xmath8 radius cluster stellar density out background good fittings obtained for@xmath9 reddening age metallicity best matching isochrones presented in table 1 isochrone fittings to cmds ic1641 ngc422 age estimates agree with cmd age determinations ages of l51/b78 ngc422/ic1641 determined using isochrone fittings first time disturbed surface brightness isophotal features carvalho 2008 classified k50 ngc376 as mergers ic1612e / w ngc241/242 true binary clusters ngc422/ic1641 ic1612e / w all cluster pairings nearly coeval ages close to @xmath10yr consistent with time scales dynamical models ngc422/ic1641 ic1612e / ages differ 390myr 60myr capture process may occurred possibility oldest cluster triggered star formation youngest supernovae explosions shell expansion mechanism proposed explanation by vallenari 1998 for lmc pair ngc1903/sl357 age difference 550myrtime scales gravitational interaction for clusters mass @xmath11m@xmath12 half mass radius 10pc merge in @xmath10yr if separation @xmath425pc sugimoto makino 1989 models for less massive clusters lmc tidal field give @xmath10yr pair lifetime separations varying @xmath46pc @xmath13m@xmath12 @xmath414pc @xmath14m bhatia 1990 after time scale cluster pair merges gravitational interaction or disrupts by galactic tidal field giant molecular clouds clusters pairs merged show structural properties single clusters plot @xmath8 against age k50 ngc376 ic1611 larger @xmath8 than mean locus single clusters evidence merger status mackey gilmore 2003 brazilian institutions cnpq fapemig for financial support | 0.264974 | 1,646 | 856 | , Saving $0.0 in GPT-4. |
20,769 | a monte carlo simulation of the @xmath0 @xmath1 theory in the broken phase is performed on a hypercubic lattice in search of an @xmath2 , @xmath3 resonance .
we investigate the region of the cutoff theory where the interaction is strong as it is there that a resonance would be expected to have a better chance to form . in that region
the presence of an @xmath2 , @xmath3 resonance with mass below the cutoff is excluded .
y epsf \1 .
truein 1.5 truein the simplest effective theory describing low energy pion nucleon interactions is described by a chiral langrangian , and involves the three pion fields @xmath4 , the scalar field @xmath5 , the two nucleons , their interactions , and the dimensionful pion decay constant @xmath6 .
the @xmath5 , although it has never really been seen , may exist in nature as a broad and quite heavy ( @xmath7 ) isospin i=0 , spin j=0 resonance . in that setting
, it was natural to ask whether the presence of the @xmath8 resonance was a consequence of the pion interactions or of some more fundamental interactions .
since the earliest such theory ( gell mann levy sigma model ) becomes an o(4 ) @xmath1 theory if the nucleon fields are neglected , this question reduces to whether or not the o(4 ) @xmath1 theory can sustain an @xmath2 , @xmath3 resonance in the broken phase . today , in a very different setting
, this same question is of interest again .
the scalar sector of the minimal standard model is also a four component @xmath1 theory in the broken symmetry phase .
the equivalent of the @xmath5 resonance is the higgs particle , the three pions are the goldstone bosons , the pion decay constant is the weak scale @xmath9 , and the scattering of longitudinally polarized vector bosons behaves exactly the same way as a scaled up version of @xmath10@xmath10 scattering ( equivalence theorem @xcite ) .
it is possible that the higgs , like the @xmath5 , is quite heavy and broad and may avoid detection at ssc . on the other hand , if the four component @xmath1 theory does indeed contain an @xmath2 , @xmath3 resonance , then since the @xmath8 resonance , as it appears in nature , is quite strong , its equivalent in the scalar sector of the minimal standard model may have a good chance to be detected at ssc .
therefore , it becomes very important to know if another `` signature '' of the scalar sector , besides the higgs resonance , is awaiting discovery at ssc .
earlier attempts to answer this question used analytical approximations such as pade approximants and found that an @xmath2 , @xmath3 resonance is present in the theory .
it is not clear , however , whether or not the presence of this resonance is an artifact of the approximations used and therefore the question of the existence of an @xmath2 , @xmath3 resonance in the theory has yet to receive a conclusive answer .
we report here on a monte carlo simulation to shed new light on this old question .
the o(4 ) @xmath1 theory has the lattice action @xmath11 and was simulated on the lattice in the @xmath12 limit . in that limit ,
the theory has the strongest interactions and a resonance probably has a better chance to form .
the simulation was done on a hypercubic lattice @xmath13 of spatial extension @xmath14 and time extension @xmath15 using an incomplete heat bath algorithm on the cm-2 machine at scri . since , on a finite lattice ,
the direction of the symmetry breaking changes from configuration to configuration , the same approach as in @xcite was used to disentangle the massive scalar field @xmath5 from the goldstone modes .
the mass @xmath16 of the @xmath5 field was measured using standard techniques . to measure the energy of the lowest laying state in the @xmath2 , @xmath3 channel , an operator carrying these quantum numbers needs to be constructed .
the simplest such operator is : @xmath17 where summation over repeated indices is assumed , @xmath18 are the isospin indices , @xmath19 is the totally antisymmetric tensor , @xmath20 is the @xmath21th euclidean unit vector of the time slice @xmath22 , and @xmath23 $ ] is the z - component spin index .
unfortunately , the time slice connected correlation function of this operator gives a very weak signal . to get a better signal
an operator that extends over several lattice spacings needs to be constructed using a trial wave function for the two pion state .
the `` bag '' and `` bound state '' meson wave functions were considered .
the latter gave a better signal with an affordable cost in computer time and was therefore used for the simulation .
using as a trial wave function for the two pions at relative position @xmath24 , the hydrogen wave function @xmath25 , an @xmath2 , @xmath3 operator with total 3-momentum zero was constructed : @xmath26 where @xmath27 is a three - dimensional cubic box centered at the origin and contained in @xmath22 , @xmath28 is the `` bohr radius '' in lattice units , @xmath29 and @xmath30 are the azimuthal and polar angles of @xmath31 , and @xmath32 is the @xmath33 spherical harmonic ( up to a multiplicative normalization constant ) .
the parameter @xmath28 can be given any value .
a very large @xmath28 will cause the exponential to decrease very slowly and then the sum over @xmath24 will have to be carried out over a box @xmath27 as large as @xmath22 .
since this can increase the computer time significantly , a smaller @xmath28 has to be used so that the size of the box that contains the important contribution from the exponential can be made smaller .
however , @xmath28 can not be made very small because the signal becomes weaker as @xmath28 decreases .
an optimal choice of @xmath28 and @xmath27 was found to be @xmath34 , and @xmath27 extending from @xmath35 to @xmath36 in each of the three directions .
the operator @xmath37 couples to the @xmath2 , @xmath3 states . the energy of the lowest laying state in this channel can be found by looking at the time slice connected correlation function @xmath38 of this operator .
the simulation was done on an @xmath39 and @xmath40 lattice and for three values of the hopping parameter @xmath41 .
these values were chosen so that a wide range of @xmath16 will be covered ( they also coincide with some of the values used in @xcite ) .
the expectation values of @xmath42 , were measured .
.energy @xmath43 in the @xmath2 , @xmath3 channel .
[ cols="^,^,<,^,^",options="header " , ] @xmath44 was fitted for a few different ranges of @xmath45 , and the resulting energies @xmath43 , together with the @xmath46 per degree of freedom for each fit are given in table 1 for the @xmath47 and @xmath48 lattices .
the effective energy @xmath49 , for the two time slices at t-1 and t , is plotted versus t for the three values of @xmath50 in figures 1 ( @xmath47 ) and 2 ( @xmath48 ) .
the values of t omitted from those plots had an @xmath49 with very large error .
in a two - pion @xmath2 , @xmath3 state with zero total 3momentum , the lowest 3momentum a pion can have has one component equal to @xmath51 and two equal to @xmath52 .
the next one has two components equal to @xmath51 and one equal to @xmath52 .
the energy spectrum in the @xmath2 , @xmath3 channel is therefore expected to contain levels with energies close to the energies of these states , but slightly different because of the interaction . for the @xmath47 lattice ,
these levels have energies @xmath53 and @xmath54 respectively , and are denoted by the dotted lines y = in figure 1 . from this figure , it is clear that the observed energy levels are very close to the levels of two free pions .
in fact , for smaller @xmath45 , the levels are close to @xmath55 , and for larger @xmath45 they are close to @xmath56 . because the free two pion levels are not very well separated at @xmath47 , the observed levels are probably a mixture of the two lowest ones . in that sense , the energies in table 1 for the @xmath47 lattice are probably a mixture as well . the fact that the observed levels correspond to a two pion state and not to a resonance is also greatly supported by the fact that these levels change only slightly from @xmath57 to @xmath58 .
after all , in that range @xmath16 varies from @xmath59 to @xmath60 . therefore , if a resonance is present it must have energy larger than @xmath61 and is either too heavy ( for example , heavier than @xmath62 ) to be observed with this method , y = or is `` hiding '' between @xmath63 and @xmath62 . that the latter is not the case
can be seen by looking at the energy levels obtained for the @xmath48 lattice . from figure 2
it is clear that there are no energy levels below @xmath64 .
in fact , since the two lowest free two pion states are well separated in this lattice , the lowest energy level is clearly visible in this figure .
it is true that the limited statistics give @xmath49 only up to @xmath65 , but because the lowest level is well separated from the next one , the correlated @xmath46 fit gives a good estimate of the energy of this level .
the numbers given in table 1 are indeed very close to the lowest free two pion energy @xmath66 and do not seem to change much for the different values of @xmath50 . from this analysis it is evident that if an @xmath2 , @xmath3 resonance exists for @xmath67 , it is unphysical since it must have energy larger than the cutoff .
of course there is always the possibility that a resonance with energy below the cutoff does exist , but , because the overlap of the operator in eq .
@xmath68 with an @xmath69 , @xmath3 resonance may have been much smaller than its overlap with the free two pion states present in that channel , it was not possible to observe with the statistics and lattices used .
however , it is unlikely that this is the case since the operator in eq .
@xmath68 is more appropriate for a resonance than for a free two particle state .
for the theory of gell - mann and levy of low energy pion processes the @xmath67 region corresponds to the region where the @xmath5 particle mass is greater than approximately @xmath70 ( or equivalently the cutoff is less than approximately @xmath71 .
( or @xmath72 ) is made through the renormalized coupling constant @xmath73 .
the value of @xmath74 for this value of @xmath50 was taken from . ] since @xmath16 is expected to be much larger than @xmath70 , it is clear from the results that the existence of the @xmath8 resonance in nature can not be accounted for by this low energy theory alone ( in contrast to ) . for the scalar sector of the minimal standard model
the @xmath75 region corresponds to where the higgs mass is greater than approximately @xmath76 ( or equivalently the cutoff is less than approximately @xmath77 ) .
the higgs mass is of course not known but its upper bound is placed at around @xmath78 ( hypercubic lattice triviality bound ) .
thus the existence of an @xmath2 , @xmath3 resonance can be excluded with confidence for values of the higgs mass above @xmath79 . for @xmath80 ( higgs mass @xmath81 )
the strength of the interaction becomes weaker and hence it is safe to say that if a resonance could not form for @xmath67 , where the interaction is stronger , it is unlikely that it will in this region either .
for this reson it was not deemed necessary to investigate the @xmath80 region . a numerical simulation for @xmath80 not only is not necessary , but it would also be very costly since larger lattices will have to be used ( the correlation length becomes larger than approximately @xmath82 ) .
it must be emphasized that these conclusions are valid only within the realm of the scalar sector of the minimal standard model .
it is of course still possible that the physics that enters at higher energies may be able to produce such a resonance in very much the same way the physical @xmath8 particle owes its existence to qcd .
this resonance if it exists due to some higher energy theory it would have energy determined by that theory . in fact , it is possible that the energy of this resonance is determining the cutoff energy of the minimal standard model . | * figure @xmath83 : * * figure @xmath84 : *
this research was partially funded by the u.s .
department of energy through contract no . de - fg05 - 92er40742 and de - fc05 - 85er250000 .
9 m. gell mann , m. levy , * il nuovo cimento vol .
xvi , n. 4 * ( 1960 ) 705 ; h. georgi , * weak interactions and modern particle theory * ( 1984 ) addison - wesley pub .
chanowitz , m.k .
gaillard * nucl .
phys . b261 * ( 1985 ) 379 .
see for example : j.l .
basdevant , b.w .
lee * phys .
2 , no 8 * ( 1970 ) 1680 ; lh .
chan , r.w .
haymaker * phys .
12 * ( 1974 ) 4170 .
bitar , p.m. vranas * phys .
b 284 * ( 1992 ) 366 .
k. fredenhagen , m. marcu * phys .
b 193 * ( 1987 ) 486 .
a. hasenfratz , k. jansen , j. jersak , c.b .
lang , t. neuhaus , h. yoneyama * nucl .
phys . b317 * ( 1989 ) 81 .
degrand , r.d .
loft * comp .
65 * ( 1991 ) 84 .
m. lscher , p. weisz * nucl .
phys . b318 * ( 1989 ) 705 . | @xmath83 research funded u.s department energy contract - fg05 - 92er40742 fc05 - 85er250000. gell mann levy nuovo cimento vol xvi n 4 1960 ) 705 georgi weak interactions modern particle theory 1984 ) addison - wesley chanowitz gaillard . b261 1985 ) 379 j.l. basdevant b.w lee . 2 8 1970 ) 1680 chan haymaker phys . 12 1974 ) 4170 bitar vranas phys . 284 1992 ) 366 fredenhagen marcu . b 193 1987 ) 486 hasenfratz jansen jersak lang neuhaus yoneyama phys . b317 1989 ) 81 degrand r.d loft 65 1991 ) 84 lscher weisz . b318 1989 ) 705 | 0.071789 | 377 | 196 | , Saving $0.0 in GPT-4. |
20,770 | the aim of the na61/shine strong interaction program is to explore the phase diagram of strongly interacting matter .
the main physics goals are the study of the onset of deconfinement and the search for the critical point of strongly interacting matter .
these goals are pursued by performing a beam momentum ( 13@xmath0 - 158@xmath0 gev / c ) and system size ( p+p , p+pb , be+be , ar+sc , xe+la ) scan .
this contribution presents results on transverse momentum and multiplicity fluctuations from the be+be and p+p energy scan .
also , results on two - particle correlations in pseudorapidity and azimuthal angle obtained in p+p interactions will be shown .
the influence of conservation laws and resonance decays on multiplicity and chemical fluctuations of identified particles in p+p interactions will be discussed .
obtained results will be compared with data from other experiments and with model predictions .
proton - proton interactions , beryllium - beryllium interactions , multiplicity , chemical and transverse momentum fluctuation , onset of deconfinement , two - particle correlations | it is a well established fact that matter exists in different states . for strongly interacting matter
at least three states are expected : nuclear matter , hadron gas ( hg ) and a system of deconfined quarks and gluons ( often called the quark - gluon plasma - qgp ) . one of the most important goals of high - energy heavy - ion collisions is to establish the phase diagram of strongly interacting matter by finding the possible phase boundaries and critical points . in particular , we want to produce the quark - gluon plasma and analyze its properties and the transition between qgp and hg .
this contribution presents experimental results on event - by - event fluctuations of multiplicity , particle type and transverse momentum of charged and identified hadrons produced in inelastic p+p interactions at 20 , 31 , 40 , 80 and 158 gev / c , and centrality selected @xmath1be+@xmath2be collisions at 20@xmath0 , 30@xmath0 , 40@xmath0 , 75@xmath0 and 150@xmath0 gev / c .
the measurements were performed by the multi - purpose na61/shine @xcite experiment at the cern super proton synchrotron ( sps ) .
they are part of the strong interaction program devoted to study of properties of the onset of deconfinement ( od ) and search for the critical point ( cp ) of strongly interacting matter . within this program
a two dimensional scan in collision energy and size of colliding nuclei is in progress .
the expected signal of a critical point is a non - monotonic dependence of various fluctuation / correlation measures in such a scan .
evidence for the onset of deconfinement at lower sps energies suggests the possibility of detecting the cp in heavy ion collisions . a specific property of the cp - the increase in the correlation length - makes fluctuations its basic signal .
in order to observe the cp , created matter should freeze - out near its location .
so , if at all , the cp signals are expected at energies higher than those of the onset of deconfinement .
indications of enlarged event - by - event fluctuations in multiplicity and transverse momentum as well as an intermittency signal were found by na49 in si+si collisions at 158@xmath0 gev @xcite .
when comparing fluctuations in systems of different size ( volume ) one should use quantities which are insensitive to volume and volume fluctuations ( unavoidable in a+a collisions ) .
two families of such quantities were used .
first , the intensive quantity scaled variance , @xmath3 = \frac{\langle a^{2}\rangle-\langle a\rangle^{2}}{\langle a \rangle}$ ] where @xmath0 stands for an event quantity . in the grand canonical ensemble
it is independent of volume but it depends on volume fluctuations .
second , the strongly intensive quantities @xmath4 $ ] and @xmath5 $ ] defined as : @xmath6 = \frac{1}{c_{\delta}}\big [ \langle b \rangle\omega_{a } - \langle a \rangle\omega_{b } \big ] , \quad \sigma[a , b ] = \frac{1}{c_{\sigma}}\big [ \langle b \rangle\omega_{a } + \langle a \rangle\omega_{b } - 2 \big ( \langle ab \rangle - \langle a \rangle \langle b \rangle \big ) \big],\ ] ] where @xmath0 and @xmath7 are two different event quantities .
there is an important difference between @xmath5 $ ] and @xmath4 $ ] .
only the first two moments of a and b distributions are required to calculate @xmath5 $ ] , whereas @xmath4 $ ] includes the correlation term @xmath8 .
thus they can be sensitive to various physics effects in different ways . with the proper normalization they are dimensionless and have a common scale
they are 0 for absence of e - b - e fluctuations ( @xmath9 , @xmath10 ) and one for the independent particle production model . here
a and b stand for multiplicities of different particle types or multiplicity and sum of transverse momenta . for comparison with the na49 experiment also the @xmath11 quantity was used . for fluctuations of @xmath12
it reads @xmath13 } \big [ \sqrt{\sigma[p_{t},n]}-1 \big]$ ] . for particle type fluctuations of two particle types @xmath14 and @xmath15 it is defined as @xmath16}-1]$ ] .
enhanced fluctuations of multiplicity and transverse momentum were suggested as a possible signature of the cp of strongly interacting matter @xcite .
transverse momentum fluctuations were measured by the na49 experiment using the @xmath17 quantity @xcite . for system size dependence at the top sps energy na49
finds a maximum of @xmath17 for c+c and si+si collisions @xcite .
the energy dependence of @xmath18 and @xmath19 $ ] in p+p and @xmath1be+@xmath2be interactions is shown in fig .
the values in @xmath1be+@xmath2be interactions exhibit no centrality dependence and are close to p+p results .
no structures which could be connected to the cp or od in p+p and @xmath1be+@xmath2be collisions are visible .
bose - einstein statistics and @xmath20 , @xmath21 correlations probably introduce differences between @xmath22 $ ] and @xmath19 $ ] .
details on this analysis can be found in ref .
@xcite .
$ ] ( left ) and @xmath19 $ ] ( right ) in p+p and @xmath1be+@xmath2be interactions.,title="fig:",scaledwidth=45.0% ] $ ] ( left ) and @xmath19 $ ] ( right ) in p+p and @xmath1be+@xmath2be interactions.,title="fig:",scaledwidth=45.0% ] charge fluctuations may be sensitive to numerous effects including the qgp phase , where the charge carriers are quarks with fractional charge , number of resonances at the chemical freeze - out , conservation laws and long range correlations .
recently , they were studied in @xmath1be+@xmath2be interactions at 150@xmath0 gev / c as a function of the size of pseudorapidity window with center at @xmath23 .
the size of the @xmath24 window was varied from 0.2 to 3.4 .
figure [ charge ] shows results on @xmath25 $ ] , where @xmath26 and @xmath27 are multiplicities of positively and negatively charged hadrons .
there is no centrality dependence and the data agrees well with the epos model @xcite predictions .
$ ] in @xmath1be+@xmath2be interactions at 150a gev / c as a function of the size of pseudorapidity window for different size of centrality bins ( _ left _ ) and with predictions from the epos model ( _ right_).,title="fig:",scaledwidth=45.0% ] $ ] in @xmath1be+@xmath2be interactions at 150a gev / c as a function of the size of pseudorapidity window for different size of centrality bins ( _ left _ ) and with predictions from the epos model ( _ right_).,title="fig:",scaledwidth=45.0% ] ( _ left _ ) and @xmath28 ( _ right _ ) in p+p ( na61/shine ) ( preliminary ) and central pb+pb ( na49 ) interactions @xcite.,title="fig:",scaledwidth=45.0% ] ( _ left _ ) and @xmath28 ( _ right _ ) in p+p ( na61/shine ) ( preliminary ) and central pb+pb ( na49 ) interactions @xcite.,title="fig:",scaledwidth=45.0% ] several mechanisms could lead to specific event - by - event particle ratio ( chemical ) fluctuations . among them
are overheating - supercooling fluctuations due to a first order phase transition with considerable latent heat , or fluctuations due to coexistence of confined and deconfined matter ( mixed phase ) @xcite .
comparison of p+p reaction with pb+pb interactions from na49 is shown in fig .
[ figphi ] .
the results for @xmath29 and @xmath28 in both reactions are similar .
@xmath29 is probably dominated by conservation laws and resonance decays ( agreement with urqmd model supports this interpretation ) .
no structures can be connected with the cp but there is a systematic difference in @xmath28 possibly connected to od .
further studies are needed in order to understand the observed discrepancy .
in inelastic p+p interactions for all charged particle pairs . the correlation function is mirrored around @xmath30.,title="fig:",width=172 ] in inelastic p+p interactions for all charged particle pairs .
the correlation function is mirrored around @xmath30.,title="fig:",width=172 ] in inelastic p+p interactions for all charged particle pairs .
the correlation function is mirrored around @xmath30.,title="fig:",width=172 ] in inelastic p+p interactions for all charged particle pairs .
the correlation function is mirrored around @xmath30.,title="fig:",width=172 ] in inelastic p+p interactions for all charged particle pairs .
the correlation function is mirrored around @xmath30.,title="fig:",width=172 ] two - particle correlations in @xmath31 were studied extensively at rhic and lhc .
they may allow to disentangle different sources of correlations : jets , flow , resonance decays , quantum statistics effects , conservation laws , etc .
correlations are calculated as a function of the difference in pseudo - rapidity ( @xmath24 ) and azimuthal angle ( @xmath32 ) between two particles in the same event .
the correlation function @xmath33 is obtained as the ratio of distributions of pairs from data and mixed events of the same multiplicity , respectively .
details can be found in ref .
the energy dependence of the correlation functions from p+p reactions for all charged pair combinations is presented in fig .
[ deltaphi ] .
two structures can be seen in the plots : * a maximum at @xmath34 , probably a result of resonance decays and momentum conservation .
it is strongest for unlike - sign pairs and significantly weaker for same charge pairs * a weak enhancement at @xmath35 , likely due to coulomb interactions ( unlike - sign pairs ) and quantum statistics ( same charge pairs ) .
this work was supported by the the national science centre , poland grant 2012/04/m / st2/00816 .
00 n. antoniou _ et al . _
[ na49-future collaboration ] , cern - spsc-2006 - 034 , cern - spsc - p-330 , cern - spsc-2012 - 022 , spsc - p-330-add-6 .
n. abgrall _ et al . _
[ na61 collaboration ] , jinst * 9 * , p06005 ( 2014 ) [ arxiv:1401.4699 [ physics.ins-det ] ] . k. grebieszkow ( for the na49 collab . ) , nucl
. phys . * a 830 * , 547c ( 2009 ) and ref . therein .
t. anticic _ et al . _
[ na49 collaboration ] , arxiv:1208.5292 [ nucl - ex ] .
m. a. stephanov , k. rajagopal and e. v. shuryak , phys .
d * 60 * , 114028 ( 1999 ) [ hep - ph/9903292 ] .
m. gazdzicki , m. i. gorenstein and m. mackowiak - pawlowska , phys .
c * 88 * , no . 2 , 024907 ( 2013 ) doi:10.1103/physrevc.88.024907 [ arxiv:1303.0871 [ nucl - th ] ] .
t. anticic et al .
( na49 collab . ) , phys .
* c 70 * , 034902 ( 2004 ) ; t. anticic et al .
( na49 collab . ) , phys . rev .
* c 79 * , 044904 ( 2009 ) a. aduszkiewicz _ et al .
_ [ na61/shine collaboration ] , arxiv:1510.00163 [ hep - ex ] .
k. werner , f. m. liu and t. pierog , acta phys . hung . a * 24 * , 329 ( 2005 ) doi:10.1556/aph.24.2005.1 - 4.45 [ hep - ph/0405274 ] .
l. p. csernai and i. n. mishustin , phys .
lett . * 74 * , 5005 ( 1995 ) ; j. i. kapusta , a. mekjian , phys . rev . *
d 33 * , 1304 ( 1986 ) ; r. stock , nature * 337 * , 319 ( 1989 ) .
a. rustamov , pos cpod * 2013 * , 005 ( 2013 ) .
b. maksiak , arxiv:1503.02470 [ nucl - ex ] ( 2015 ) . | matter exists in states strongly interacting matter three states expected nuclear matter hadron gas hg deconfined quarks gluons quark - gluon plasma collisions establish phase diagram matter phase boundaries critical points produce quark - gluon plasma analyze properties transition between qgp hg contribution presents results on fluctuations multiplicity particle type transverse momentum of charged hadrons in inelastic p+p interactions at 20 31 40 80 158 gev / c+ collisions at 20 30 40 75 150 gev / c measurements performed by na61/shine @xcite experiment cern super proton synchrotron part of strong interaction program onset of deconfinement critical point cp of interacting matter two dimensional scan in collision energy size of colliding nuclei progress expected signal critical point non - monotonic dependence of fluctuation / correlation measures evidence onset of deconfinement at lower sps energies detecting cp in heavy ion collisions cp increase correlation length makes fluctuations basic signal observe cp matter should freeze - out near location cp signals expected at energies higher than onset deconfinementenlarged event fluctuations in multiplicity transverse momentum intermittency signal found by na49 in si+si collisions at 158@xmath0 gev @xcite comparing fluctuations in systems different size use quantities insensitive to volume fluctuations in a+a collisions two families quantities used intensive quantity scaled variance @xmath3 = \frac\langle a\rangle- a @xmath0 event quantity independent of volume depends on volume fluctuations strongly intensive quantities @xmath4 @xmath5 $ ] as @xmath6 = \frac{1}{c_\delta}}\big b -\sigma = \frac{1}{c\big @xmath0 and @xmath7 two different event quantities important difference between @xmath5 $ @xmath4 $ ] first two moments of a and b distributions required to calculate @xmath5 $ ] @xmath4 $ includes correlation term @xmath8 sensitive to physics effectsnormalization dimensionless common scale 0 absence e - b - e fluctuations @xmath9 @xmath10 one independent particle production model a b multiplicities particle types sum transverse momenta comparison na49 experiment @xmath11 quantity used fluctuations @xmath12 @xmath13 } particle type fluctuations @xmath14 @xmath15 @xmath16}-1]$ ] enhanced fluctuations multiplicity transverse momentum signature cp interacting matter transverse momentum fluctuations measured na49 experiment @xmath17 quantity system size dependence top sps energy na49 maximum @xmath17 c+c si+si collisions energy dependence @xmath18 @xmath19 p+p @xmath1be+@xmath2be interactions shown fig . values no centrality dependence close to p+p results no structures connected cp od collisions visible bose - einstein statistics @xmath20 @xmath21 correlations introduce differences @xmath22 @xmath19 $ details analysis ref @xcite @xmath19 p+p @xmath1be+@xmath2be interactionsleft @xmath19 right p+p @xmath1be+@xmath2be interactions=45.0% charge fluctuations sensitive effects qgp phase charge carriers quarks charge resonances chemical freeze - out conservation laws long range correlations studied @xmath1be+@xmath2be interactions 150@xmath0 gev / c size pseudorapidity window center @xmath23 size @xmath24 window varied 0.2 to 3.4 figure results @xmath25 @xmath26 @xmath27 multiplicities positively negatively charged hadrons no centrality dependence data agrees with epos model predictions @xmath1be+@xmath2be interactions 150a gev / c size pseudorapidity window different centrality bins left predictions epos model right=45.0% @xmath1be+@xmath2be interactions 150a gev / c size pseudorapidity window centrality bins left predictions epos model right=45.0% left @xmath28 right p+p na61/shine preliminary central pb+pb na49 interactions.0% left @xmath28 right p+p na61/shine central pb+pb na49 interactions,title="fig:",scaledwidth=45.0% mechanisms could lead to event particle ratio chemical fluctuations overheating - supercooling fluctuations due to first order phase transition with latent heat fluctuations to coexistence of confined and deconfined matter ( mixed phase ) @xcite comparison of p+p reaction with pb+pb interactions from na49 shown in fig . results for @xmath29 and @xmath28 similar @xmath29 dominated by conservation laws resonance decays agreement with urqmd model supports interpretation no structures connected with cp systematic difference in @xmath28 possibly connected to od further studies needed to understand discrepancy inelastic p+p interactions for pairs correlation function mirrored around @xmath30.=172 mirrored around two - particle correlations in @xmath31 studied at rhic and lhc may allow to disentangle sources of correlations : jets flow resonance decays quantum statistics effects conservation lawscorrelations calculated pseudo - rapidity @xmath24 azimuthal angle @xmath32 between particles same event correlation function @xmath33 ratio distributions pairs data mixed events same multiplicity details ref energy dependence correlation functions p+p reactions charged pair combinations presented in fig . two structures plots maximum at @xmath34 resonance decays momentum conservation strongest for unlike - sign pairs weaker same charge pairs weak enhancement at @xmath35 due coulomb interactions - quantum statistics same charge pairs work supported national science centre poland grant 2012/04/m / st2/00816 antoniou al na49-future collaboration cern - spsc-2006 - 034 spsc - p-330 - 022-add-6 abgrall na61 collaboration jinst 9 p06005 ( 2014 arxiv:1401.4699.ins-det grebieszkow na49 nucl phys . 830 547c ( 2009 ) anticic na49 collaboration arxiv:1208.5292 nucl -. stephanov rajagopal shuryak . 60 114028 ( 1999 ) hep - ph/9903292 gazdzickigorenstein mackowiak - pawlowska phys . 88 2 024907 2013 ).1103/physrevc.88.024907 arxiv:1303.0871 nucl - th anticic na49 . 70 034902 ( 2004 ) anticic . 79 044904 2009 ) aduszkiewicz na61/shine collaboration arxiv:1510.00163 hep - ex werner liu pierog acta phys . 24 329 ( 2005 ) doi:10.1556/aph.24.2005.1 - 4.45 hep - ph/0405274 csernai mishustin . 74 5005 ( 1995 ) kapusta mekjian . rev. 33 1304 ( 1986 ) stock nature 337 319 ( 1989 ) rustamov cpod 2013 005 ( 2013 maksiak arxiv:1503.02470 nucl - ex 2015 ) | 0.428269 | 2,996 | 1,517 | , Saving $0.1 in GPT-4. |
20,783 | it is well known , that there is temporal relationship between cmes and associated flares .
the duration of the acceleration phase is related to the duration of the rise phase of a flare .
we investigated cmes associated with long duration flares of slow rise phase ( slow lde ) .
these cmes are characterized by high velocity ( @xmath0 km / s ) during the propagation phase and low average acceleration during the main , prolongated , acceleration phase .
the cmes are accelerated until the height @xmath1 , which is higher value than in typical impulsive cmes .
cmes associated with slow ldes have characteristics of both classes of cmes , i.e. cmes associated with flares and cmes associated with eruptive prominences . | the first coronal mass ejection ( cme ) was clearly identified by r. tousey in 1971 @xcite in the _ orbiting solar observatory-7 _ observations . later on cmes
were observed by _
@xcite and since that time have become one of the most interesting topics in solar physics .
it is well known that cmes are associated with flares and eruptive prominences @xcite .
basing on small samples of events @xcite and @xcite showed that the flare - associated cmes are characterized by high velocities ( approximately constant ) and rapid acceleration . on the other hand cmes associated with the prominences
are accelerated over larger height - range and attain lower velocities than flare - associated cmes . using _
soho_/lasco data @xcite separated cmes into two classes : 1 ) gradual cmes with lower speed ( @xmath2@xmath3 km / s ) , associated with eruptive prominences ; 2 ) impulsive cmes with higher speed ( @xmath4 km / s ) , associated with flares and showing clear evidence of deceleration . the existence of two classes of cmes was confirmed by many authors @xcite . however , there are also observations that not confirm differences between the two types of cmes @xcite .
@xcite presented statistical analysis of 545 flare - associated cmes and 104 non - flare cmes .
they demonstrated that both data sets show very similar characteristics : in both samples significant fraction of cmes was accelerated or decelerated and both samples include a comparable ratio of fast and slow cmes .
this directly opposes the concept of two distinct ( flare / non - flare ) classes of cmes . on average , mean cme
velocities gradually decrease , following the sequence : strong - flare events @xmath5 weak - flare events @xmath5 non - flare events with eruptive prominences @xmath5 no - flare / no - prominence events .
@xcite investigated the temporal relationship between cmes and associated solar flares .
they found that kinematic evolution of cmes can be described in three - phase scenario : the initiation phase , the impulsive acceleration phase , and the propagation phase . the impulsive acceleration phase is often synchronized with the rising phase of the associated soft x - ray flare ( @xcite and references therein ) , i.e. , during the flare impulsive phase ( @xcite and references therein ) .
typically , cmes are accelerated over short distance range , usually below the occulting disc of white - light space - borne coronagraphs
. however , sometimes they accelerate up to heights of of several solar radii , in some cases even until the height of 7 @xmath6 @xcite . during the acceleration phase
, a cme has an acceleration of a few hundred m / s@xmath7 @xcite .
the value of the acceleration of a cme depends on a duration of the main acceleration phase @xcite .
a fast cme ( @xmath8 km / s ) can be strongly accelerated over short time , or weakly accelerated over an extended time interval of several hours @xcite . since cmes are associated with flares , a duration of the acceleration phase is related to a duration of the rise phase of a flare .
cmes associated with flares of slow rising phase should be accelerated slowly but for a long time . here , we investigate the kinematic evolution of cmes associated with flares of slow rising phase .
these cmes were accelerated as long as sxr flux of associated flares increased . during the main acceleration phase cmes reached heights @xmath1 , which is a higher value than for a typical cme .
these cmes are characterized by a high velocity during the propagation phase and a low average acceleration during the main , prolongated , acceleration phase .
height - time profiles for the cmes are similar to profiles of cmes associated with eruptive prominences , which means that those cmes may have characteristics both groups of cmes , i.e. cmes associated with flares and cmes associated with eruptive prominences .
we selected five cmes associated with slow ldes .
we were interested only in near - the limb events to avoid projection effects .
basic information about associated slow ldes are given in table [ flares ] .
we determined heights of the cmes ( in the plane of the sky ) using running - differences images from the _ soho_/eit , mlso / mk4 and _
soho_/lasco c2 and c3 instruments .
we fitted second - degree polynomial to the points on the height - time profile to obtain average acceleration during the main acceleration phase .
time range of data used for quadratic fitting estimates the acceleration phase duration .
we used linear fit to obtain average velocity during the propagation phase . here , we present two events in detail .
the results for all the analysed cmes are shown in table [ cmes ] .
[ flares ] .data for flares associated with the cmes [ cols="^,^,^,^,^,^,^ " , ]
, lower curve : @xmath9 ) . _
right : _ _ soho_/eit 171 @xmath10 image showing the flare of 2004 july 31 during the rise phase .
contours show the emission in the @xmath11 kev range observed with _
rhessi_. the contours are for 10% to 90% ( with increment 10 % ) of maximum emission.,title="fig:",width=257 ] , lower curve : @xmath9 ) . _
_ soho_/eit 171 @xmath10 image showing the flare of 2004 july 31 during the rise phase .
contours show the emission in the @xmath11 kev range observed with _
rhessi_. the contours are for 10% to 90% ( with increment 10 % ) of maximum emission.,title="fig:",width=200 ] the cme was associated with the slow lde , which occurred in the active region noaa 10652 , @xmath12 behind the western solar limb .
soft x - ray ( sxr ) flux started to increase at 05:16 ut and reached its maximum ( _ goes _ class c8.4 ) at 06:57 ut ( fig .
[ goes31jul04 ] , left panel ) .
the flare was observed by many instruments : _
soho_/eit ( after 05:24 ut ) , _ goes_/sxi and _ rhessi _ ( between 06:3007:15 ut ) . the hard x - ray ( hxr ) emission was observed above a loop seen in the eit images .
hxr source was observed at high altitude , @xmath13 mm , ( fig .
[ goes31jul04 ] , right panel ) .
the cme was observed by lasco / c2 and c3 coronographs ( fig .
[ lasco31jul04 ] ) .
unfortunately , the eit telescope observed the cme after 05:24 ut , 8 minutes after the flare start , so it was not possible to analyse cme at the very beginning .
a leading edge of the cme was visible in the c2 images during at least 1 hour . over
the next few hours cme was in the fov of the c3 coronograph .
we measured the height of the cme ( fig .
[ ht_v31jul04 ] , left panel ) and used it to calculate an average acceleration during the main acceleration phase and a velocity profile ( fig .
[ ht_v31jul04 ] , right panel ) .
velocity increased until @xmath1407:40 ut and reached @xmath15 @xmath14 @xmath16 km / s ( the average velocity during the propagation phase was about @xmath17 km / s ) .
the velocity profile is similar to the _ goes _ light curve , which suggests a connection between the cme and the flare .
the main acceleration phase was characterized by the average acceleration of about @xmath18 m / s@xmath19 .
if we take into account the poor temporal resolution of the observations , we can assume that the cme is accelerated to the height @xmath14 @xmath20 @xmath21 .
this value is greater than a value during a typical impulsive cme .
the high velocity of the cme is caused by low , but prolongated acceleration , which lasted more than 2 hours .
the 2005 september 6 cme was associated with a flare that occurred in the active region noaa 10808 , @xmath22 behind the eastern limb .
the flare began at 19:32 ut and reached its maximum ( _ goes _ class m1.4 ) at 22.02 ut ( fig .
[ goes6sep05 ] , left panel ) .
the rise phase of the flare lasted @xmath23 minutes .
the flare was well observed by _
rhessi _ and _ goes_/sxi instruments .
the hxr emission source was observed above the loop seen in the sxi images ( fig .
[ goes6sep05 ] , right panel ) . a loop - top source seen in _ rhessi _
images was observed at the height @xmath24 mm and had size @xmath25 mm .
only thermal emission was observed , however footpoints were occulted . the detailed analysis of this flare @xcite revealed that thermal energy was released very slowly and decreased extremely slow after reaching the maximum value ( characteristic time @xmath26 of heating rate decrease was @xmath27 s ) which can explain long rise phase .
did the energy release slowly also in the associated cme ?
, lower curve : @xmath9 ) .
_ right : _ _ goes_/sxi image showing flare of the 2005 september 6 during the rise phase .
contours show the emission in the @xmath11 kev range observed with _
rhessi_. the contours are for 10% to 90% ( with increment 10 % ) of maximum emission.,title="fig:",width=253 ] , lower curve : @xmath9 ) . _
right : _ _ goes_/sxi image showing flare of the 2005 september 6 during the rise phase .
contours show the emission in the @xmath11 kev range observed with _
rhessi_. the contours are for 10% to 90% ( with increment 10 % ) of maximum emission.,title="fig:",width=207 ] the cme was observed by mlso mk4 ( fig .
[ mk4 ] ) and lasco / c2 and c3 coronographs ( fig .
[ lasco6sep05 ] ) .
mk4 observations allowed us to investigate early stage of the evolution . using all available observations , we measured height of the cme ( h - t profile is shown in fig .
[ ht_v6sep05 ] , left panel ) .
velocity increased until @xmath14 22:00 ut ( fig .
[ ht_v6sep05 ] , right panel ) , to the flare maximum . to that time
cme was accelerated with the average value of @xmath28 m / s@xmath19 .
average velocity during the propagation phase was equal to @xmath29 km / s , maximum value was greater than @xmath30 km / s .
this high value is due to the very long acceleration phase , which lasted more than 2 hours .
the cme was accelerated up to the height @xmath14 @xmath31 @xmath21 , which is a much greater value than a value in a typical cme .
height - time profile is similar to the profile for a cme associated with an eruptive prominence , however the cme was associated with the flare , in which slow energy release was also observed .
we analysed five cmes associated with slow lde flares .
we measured heights of the cmes , estimated acceleration phase duration and calculated average velocity during the propagation phase and average acceleration during the main acceleration phase .
the results for all the cmes are given in table [ cmes ] .
[ cmes ] date&phase duration&phase duration@xmath32&velocity@xmath7&acceleration@xmath33 & height@xmath34 + & [ min]&[min]&[km / s]&[m / s@xmath19]&[r@xmath36 + 21-may-02&76&78&1250&280&5.26.9 + 28-jul-04&217&168&900&94&7.410.0 + 29-jul-04&82&108&1400&140&4.89.3 + 31-jul-04&101&132&1300&230&6.011.9 + 06-sep-05&150&178&1460&170&12.016.80 + + + + + we can summarize our analysis as follows : * cmes associated with slow lde flares are characterized by high velocity ( in most cases @xmath0 km / s ) .
this high velocity was caused by low , but prolongated acceleration .
* duration of the acceleration phase was longer than during typical cmes .
duration of this phase is related to duration of the rising phase of associated flares . *
average value of acceleration during the main acceleration phase was low ( @xmath37 m / s@xmath19 for analysed cmes ) .
we obtained the lowest value of acceleration for the cme connected with the flare with the slowest rise phase .
* cmes were accelerated to the heights @xmath38@xmath39 ( table [ cmes ] ) , which is greater value than value in typical impulsive cmes . *
all analysed cmes were accelerated for at least 1 hour .
this prolongated acceleration caused that the h - t profiles are more similar to cmes associated with eruptive prominences , although they are connected with the slow lde flares and their evolution can be described in three - phase scenario .
our results regarding this type of events indicate that there is no clear distinction flare - associated and prominences - associated cmes .
we thank _ rhessi _ , _ soho _ , _ goes _ and mlso teams for their open data policy .
we thank the referee for useful comments and suggestions .
we also thank barbara cader - sroka for editorial remarks . this investigation has been supported by a polish ministry of science and high education ,
n203 1937 33 . | first coronal mass ejection cme identified by r. tousey in 1971 in orbiting solar observatory-7 observations later cmes observed by @xcite interesting in solar physics cmes associated with flares eruptive prominences small samples flare - associated cmes high velocities rapid acceleration cmes associated with prominences accelerated over larger height range lower velocities soho/lasco data @xcite separated cmes into two classes gradual with lower speed associated with eruptive prominences 2 impulsive cmes with higher speed km s associated with flares deceleration existence of two classes cmes confirmed by authors observations differences between types cmes presented statistical analysis of 545 flare - associated cmes 104 non - flare cmes both data sets show similar characteristics significant fraction of cmes accelerated or decelerated comparable ratio of fast and slow cmes opposes of two distinct classes of mean cme velocities decrease strong - flare weak - flare non - flare events no - @xcite investigated temporal relationship between cmes solar flaresfound kinematic evolution of cmes described three - phase scenario initiation impulsive acceleration propagation phase impulsive acceleration phase synchronized with rising phase soft x - ray flare cmes accelerated over short distance range below occulting disc of white - light space - borne coronagraphs sometimes accelerate to several solar radii 7 @xmath6 during acceleration phase cme acceleration few hundred m / s value acceleration depends on duration main acceleration phase fast cme ( km / s strongly accelerated short time or weakly accelerated over extended several hours cmes associated with flares duration acceleration phase related to rise phase flare cmes associated with flares slow rising phase accelerated slowly long time kinematic evolution of cmes associated with flares cmes accelerated sxr flux flares increased during main acceleration phase cmes reached heights @xmath1 higher value than typical cme characterized high velocity during propagation phase low average acceleration during main acceleration phase height - time profiles similar to eruptive prominences may characteristics groupscmes associated with flares eruptive prominences selected five cmes slow ldes interested in near - limb events projection effects information slow ldes in table [ flares determined heights cmes using running - differences images from soho_/eit mlso / mk4 soho_/lasco c2 c3 instruments fitted second - degree polynomial to height - time profile average acceleration main acceleration phase time range data quadratic fitting estimates acceleration phase duration used linear fit average velocity propagation phase present two events results for analysed cmes in table [ cmes flares for flares associated cmes @xmath9 soho_/eit 171 @xmath10 image flare 2004 july 31 rise phase contours show emission @xmath11 kev range 10% to 90% maximum emission soho_/eit 171 @xmath10 image flare 2004 july 31 rise phase contours show emission @xmath11 kev range maximum emission cme associated with slow lde occurred in active region noaa 10652 @xmath12 behind western solar limbsoft x - ray flux at 05:16 reached maximum 06:57 ut goes31jul04 left flare observed by instruments soho_/eit goes_/sxi rhessi _ between 06:3007:15 hard x - ray hxr emission observed above loop eit images hxr source observed at high altitude @xmath13 mm goes31jul04 cme observed by lasco / c2 c3 coronographs lasco31jul04 eit telescope observed cme after 05:24 ut 8 minutes after flare start possible analyse cme leading edge cme visible in c2 images 1 hour next hours cme in fov of c3 coronograph measured height cme_v31jul04 average acceleration phase velocity profile velocity increased until @xmath1407:40 reached @xmath15 @xmath14 @xmath16 km / s average velocity propagation phase @xmath17 km / s velocity profile similar to _ goes _ light curve suggests connection between cme flare main acceleration phase average acceleration @xmath18 m / s@xmath19 poor temporal resolution cme accelerated to height @xmath14 @xmath20 @xmath21 value greater than during typical impulsive cme high velocity caused by low prolongated acceleration lasted more than 2 hours 2005 september 6 cme associated with flare in active region noaa 10808 @xmath22 behind eastern limb flare began at 19:32 ut reached maximum at 22.02 ut rise phase lasted @xmath23 minutes observed by _ rhessi _ _ goes_/sxi instruments hxr emission source observed above loop in sxi images loop - top source in rhessi images at height @xmath24 mm size @xmath25 mm thermal emission observed footpoints occulted analysis thermal energy released slowly decreased slow after maximum value heating rate decrease was @xmath27 s long rise phase energy release slowly in associated cme ?/sxi image showing flare 2005 september 6 during rise phase contours show emission in @xmath11 kev range with for 10% to 90% % of maximum emission/sxi image showing flare of 2005 september 6 during rise phase contours show emission in kev range for 10% to 90% of maximum emissioncme observed by mlso mk4 lasco c2 c3 coronographs mk4 observations early evolution measured height cme profile fig velocity increased until @xmath14 22:00 to flare maximum cme accelerated with average @xmath28 m / s@xmath19 velocity propagation phase equal to @xmath29 km / s maximum greater than @xmath30 km / s high value due to long acceleration phase 2 hours cme accelerated to height @xmath14 @xmath31 @xmath21 greater value than typical cme height - time profile similar to cme eruptive prominence associated with flare slow energy release analysed five cmes slow lde flares measured heights estimated acceleration phase duration calculated average velocity propagation phase acceleration main acceleration phase results for in table [ cmes date&phase duration&velocity&acceleration@xmath33 height@xmath34 min[km / s][m / s@xmath19@xmath36 + 21-may-02&76&78&1250&280&5.26.9 + 28-jul-04&217&168&900&94&7.410.0 29-jul-04&82&108&1400&140&4.89.3 31-jul-04&101&132&1300&230&6.011.9 06-sep-05&150&178&1460&170&12.016.80 analysis cmes slow lde flares high velocity @xmath0 km / s caused by prolongated acceleration duration acceleration phase longer typical related to rising phase flares average acceleration during phase low @xmath37 m / s@xmath19 lowest for cme flare slowest rise phase cmes accelerated to heights @xmath38@xmath39 greater than typical impulsive cmes all cmes accelerated for 1 hour prolongated acceleration h - t profiles similar to cmes eruptive prominences connected with slow lde flares three - phase scenario results no clear distinction flare - prominences - associated cmes thank _ rhessi soho goes mlso teams for open data policy referee comments suggestions barbara cader - sroka for editorial remarks investigation supported by polish ministry of science high education 1937 33 | 0.433465 | 3,137 | 1,566 | , Saving $0.1 in GPT-4. |
20,786 | we investigate the energy dependence of a single - channel effective potential between the @xmath0 and the `` @xmath1''-core nucleus , which can be obtained as an @xmath0-``@xmath1 '' equivalent local potential from a coupled - channel model for @xmath2-@xmath3 systems .
it turns out that the imaginary part of the resultant potential near the @xmath4 decay threshold can well approximate the phase space suppression factor of @xmath5 decay modes .
the effects on the pole position of the @xmath3 state in the @xmath4 channel are also discussed .
example.eps gsave newpath 20 20 moveto 20 220 lineto 220 220 lineto 220 20 lineto closepath 2 setlinewidth gsave .4 setgray fill grestore stroke grestore | in order to elucidate the nature of antikaon - nucleon ( @xmath6 ) interaction in high density nuclear matter , it is an important subject to clarify whether the `` deeply - bound kaonic nucleus '' exists or not . in particular , the @xmath7_{i=1/2}$ ] , @xmath8 bound state , which is called `` @xmath9 '' here ,
is suggested to be the lightest and most fundamental kaonic nucleus @xcite . a new experimental search of @xmath9 via the @xmath10he(in - flight @xmath0 , @xmath11 )
reaction has been planned at j - parc as e15 experiment @xcite .
we theoretically have discussed the expected inclusive and semi - exclusive spectra for the @xmath10he(in - flight @xmath0 , @xmath11 ) reaction within the framework of the distorted - wave impulse approximation ( dwia ) .
we have employed a phenomenological single - channel @xmath0-``@xmath1 '' ( complex ) effective potential between the @xmath0 and the `` @xmath1''-core nucleus @xcite , which has the form of @xmath12 , \label{eq1}\ ] ] where @xmath13 and @xmath14 are strength parameters in real and imaginary parts , respectively , and @xmath15 is the range parameter .
@xmath16 is the phase space suppression factor of the @xmath5 decay modes @xcite where @xmath17 is the energy measured from the @xmath9 threshold . in the single - channel framework
, we have shown that behavior of the @xmath10he(in - flight @xmath0 , @xmath11 ) spectrum can be understood in the `` moving pole '' picture @xcite ; a pole of the @xmath9 bound state moves in the complex energy plane as a function of the energy @xmath17 on the real axis because the @xmath0-``@xmath1 '' effective potential is considerably energy - dependent .
a trajectory of its moving pole governs the shape of the spectrum .
thus , the validity of our calculations relies in part on whether the energy dependence of eq.([eq1 ] ) is appropriate near the @xmath4 decay threshold or not . to examine this subject ,
we have extended the previous @xmath9 single channel description to the @xmath2-@xmath3 coupled - channel ( cc ) description because the energy dependence of eq.([eq1 ] ) should originate from eliminating the @xmath4 channel in such a cc scheme . in this article , we investigate a single - channel @xmath0-``@xmath1 '' equivalent local potential which is derived from cc green s functions for @xmath18-@xmath3 systems , and evaluate the energy dependence of this effective potential to be compared with a phenomenological one @xcite which is determined within eq.([eq1 ] ) .
let us consider the following cc green s function @xmath19 for @xmath18-@xmath3 systems ; @xmath20 \left [ \begin{array}{cc } g^{(l)}_{11}(r , r ' ) & g^{(l)}_{12}(r , r ' ) \\ g^{(l)}_{21}(r , r ' ) & g^{(l)}_{22}(r , r ' ) \\ \end{array } \right ] = \delta(r'-r ) { \mbox{1}\hspace{-0.25em}\mbox{l } } , \label{eq2}\end{aligned}\ ] ] where the channel 1 ( 2 ) refers to the @xmath0 and @xmath1-core system ( the @xmath21 and @xmath22-core system ) : @xmath23 denote the kinetic energies for 1 , 2 , and the energy @xmath24 where @xmath25 @xmath26 mev below the @xmath9 threshold .
@xmath27 , @xmath28 and @xmath29 denote the diagonal and coupling potentials , respectively . for simplicity , we assume an energy - independent gaussian form as @xmath30 = \left [ \begin{array}{cc } v_1 + i \ , w_1 & v_{\rm c } \\
v_{\rm c } & v_2 + i \
, w_2 \\ \end{array } \right ] \exp[-(r / b)^2 ] , \label{eq3}\ ] ] where @xmath31 and @xmath32 denote the strength parameters of the diagonal and coupling potentials , respectively , and @xmath33 describe effects of the other decay modes .
these parameters in eq.([eq3 ] ) should be determined so as to reproduce the proper values of the binding energy and width which are obtained by single - channel calculations with @xmath34 in eq.([eq1 ] ) .
if we use cc green s functions in eq.([eq2 ] ) , we expect to confirm the previous results @xcite for the @xmath35(in - flight @xmath0 , n ) spectrum within the dwia calculation .
now we define the @xmath0-``@xmath1 '' equivalent local potential for channel 1 , @xmath36 as @xmath37 where @xmath38 is the ( 1,1 ) component of a solution of eq.([eq2 ] ) .
then , we get @xmath39 by multiplying the initial wave function @xmath40 as a bound state in channel 1 , and by integrating over @xmath41 , we obtain the expression of the equivalent local potential as @xmath42 which is defined under the boundary condition at _ every _
@xmath17 on the physical axis , because green s functions can be calculated numerically at every points in the complex energy plane .
this is the advantage that we used green s functions rather than wave functions . for the @xmath0-``@xmath1 '' channel .
potential parameters of eq.([eq3 ] ) are used as @xmath43 mev , @xmath44 mev , @xmath45 mev , @xmath46 mev and @xmath47 fm .
the energy @xmath17 , which is measured from the @xmath9 threshold , is varied form @xmath48 mev to @xmath49 mev . ]
figure 1 shows the calculated equivalent local potential @xmath36 with @xmath43 mev , @xmath44 mev , @xmath45 mev , @xmath46 mev and @xmath47 fm , in which the potential parameters are determined so that the binding energy and width of @xmath9 bound state are consistent with those calculated by yamazaki and akaishi @xcite .
we find that the strength of the imaginary part of @xmath36 becomes shallower as @xmath17 goes to @xmath25 .
this behavior is approximately equivalent to that of the energy dependence obtained from the phase space factor in eq.([eq1 ] ) , whereas the shape of @xmath36 is not a gaussian function exactly .
therefore , we estimate the potential strength of @xmath50 which is obtained with the help of the volume integrals ; @xmath51 \ , r^2 dr}. \label{eq7}\ ] ] figure 2 shows values of @xmath52 as a function of the energy @xmath17 , together with the fitted curve of @xmath53 in eq.([eq1 ] ) .
the results are as follows : * if no bound state exists in the @xmath3 channel ( @xmath54 mev ) , i m @xmath36 approximates to the phase space factor @xmath16 that is used in eq.([eq1 ] ) near the @xmath5 decay threshold . *
if a bound state exists in the @xmath3 channel ( @xmath55 mev ) , the energy dependence of i m @xmath36 considerably differs from that of eq.([eq1 ] ) due to modification of the phase volume via a pole which is located near the @xmath56 threshold . in summary
, we have investigated the energy dependence of the single - channel @xmath0-@xmath57 equivalent local potential @xmath36 derived from the @xmath2-@xmath3 model . it has turned out that the imaginary part of @xmath36 near the @xmath4 decay threshold can well approximate the phase space suppression factor of @xmath5 decay modes .
if potential parameters in eq.([eq3 ] ) are replaced by the energy - dependent ones , as obtained from chiral @xmath6-@xmath58 dynamics @xcite , their energy dependence would additionally contribute to the effective potential .
the further detailed investigation is now in progress . of the @xmath0-@xmath59 equivalent local potential @xmath36 .
the solid and dashed lines denote the cases with @xmath54 mev and @xmath60 mev in eq.([eq3 ] ) , respectively .
the dotted line denotes values obtained by @xmath53 with @xmath61 mev . ]
9 t. yamazaki and y. akaishi , phys .
b535 * , 70 ( 2002 ) .
m. iwasaki @xmath62 @xmath63 , j - parc proposal , t. koike and t. harada , phys . lett . *
b652 * , 262 ( 2007 ) .
t. koike and t. harada , phys . rev . * c80 * , 055208 ( 2009 ) .
j. mare@xmath64 , e. friedman and a. gal , phys .
b606 * , 295 ( 2005 ) .
d. jido _ et al_. , nucl
* a725 * , 181 ( 2003 ) . | elucidate antikaon - nucleon @xmath6 interaction in high density nuclear matter clarify deeply - bound kaonic nucleus @xmath7_{i=1/2}, @xmath8 bound state called @xmath9 lightest fundamental kaonic nucleus @xcite new search @xmath9 via @xmath10he(in - flight reaction planned at j - parc e15 experiment @xcite discussed inclusive semi - exclusive spectra for @xmath10he reaction distorted - wave impulse approximation employed phenomenological single - channel @xmath0-``@xmath1 '' effective potential between @xmath0 @xmath1''-core nucleus @xcite form @xmath12 @xmath13 @xmath14 strength parameters @xmath15 range parameter @xmath16 phase space suppression factor of @xmath5 decay modes @xmath17 energy from @xmath9 threshold behavior @xmath10he(in - flight spectrum understood in moving pole '' picture @xcite pole @xmath9 bound state moves energy plane energy @xmath17 @xmath0-``@xmath1 effective potential energy - dependent trajectory moving pole governs shape spectrumvalidity of calculations relies on energy dependence of eq.([eq1 ] ) near @xmath4 decay threshold extended @xmath9 single channel description to @xmath2-@xmath3 coupled - channel description energy dependence of eq.( from eliminating @xmath4 channel investigate single - channel @xmath0-``@xmath1 equivalent local potential derived from cc green s functions for @xmath18-@xmath3 systems evaluate energy dependence with phenomenological one @xcite determined within eq.([eq1 ] ) consider cc green s function @xmath19 for @xmath18-@xmath3 systems ; @xmath20{11}} channel 1 ( 2 ) refers to @xmath0 and @xmath1-core system @xmath21 and @xmath22-core system : @xmath23 denote kinetic energies for 1 2 and energy @xmath24 where @xmath25 @xmath26 below @xmath9 threshold@xmath27 denote diagonal coupling potentials assume energy - independent gaussian form @xmath30 ={array v_1 + i w_1 v_ c_2 + i{array[-(r b)^2{eq3 @xmath31 denote strength parameters diagonal coupling potentials @xmath33 effects decay modes parameters eq.[eq3 determined reproduce values binding energy width obtained single - channel calculations with @xmath34 use green s functions eq.[eq2 confirm previous results @xmath35(in - flight @xmath0 spectrum dwia calculation define @xmath0-``@xmath1 '' equivalent local potential for channel 1 @xmath36 @xmath37 @xmath38 1,1 ) component solution eq.([eq2 get @xmath39 multiplying initial wave function @xmath40 bound state channel 1 integrating over @xmath41 obtain equivalent local potential as @xmath42 defined under boundary condition at @xmath17 physical axis green s functions calculated points complex energy plane advantage used green s functions rather wave functions @xmath0-``@xmath1 '' channelparameters of eq.([eq3 used as @xmath43 fm energy @xmath17 measured from @xmath9 threshold varied form @xmath48 to mev figure 1 shows calculated equivalent local potential @xmath36 with fm determined binding energy width of @xmath9 bound state consistent with yamazaki akaishi @xcite strength of imaginary part @xmath36 shallower as @xmath17 to @xmath25 behavior equivalent to energy dependence from phase space factor in eq.( ) shape of @xmath36 not gaussian function estimate potential strength of @xmath50 obtained with volume integrals ; @xmath51 figure 2 shows values of @xmath52 as function of energy @xmath17 fitted curve of @xmath53 in eq results if no bound state in @xmath3 channel mev @xmath36 approximates to phase space factor @xmath16 near @xmath5 decay threshold if bound state in @xmath3 channel ( mev ) energy dependence of @xmath36 differs from eq.modification phase volume pole near @xmath56 threshold investigated energy dependence single - channel @xmath0-@xmath57 local potential @xmath36 from @xmath2-@xmath3 model imaginary part @xmath36 near @xmath4 decay threshold approximate phase space suppression factor @xmath5 decay modes if potential parameters replaced by energy - dependent ones @xmath6-@xmath58 dynamics energy potential further investigation progress @xmath0-@xmath59 equivalent local potential @xmath36 solid dashed lines denote cases @xmath54 mev @xmath60 mev eq dotted line denotes values @xmath53 @xmath61 mev. yamazaki akaishi phys . b535 70 ( 2002 ) iwasaki @xmath62 @xmath63. koike harada phys b652 262 ( 2007 ) 055208 ( 2009 ) mare@xmath64 friedman a. gal phys . b606 295 ( 2005 ). jido a725 181 ( 2003 ) | 0.311228 | 2,139 | 1,110 | , Saving $0.1 in GPT-4. |
20,789 | the aperture of the fluorescence detector ( fd ) of the pierre auger observatory is evaluated from simulated events using different detector configurations : mono , stereo , 3-fd and 4-fd .
the trigger efficiency has been modeled using shower profiles with ground impacts in the field of view of a single telescope and studying the trigger response ( at the different levels ) by that telescope and by its neighbours .
in addition , analysis cuts imposed by event reconstruction have been applied .
the hybrid aperture is then derived for the auger final extension .
taking into account the actual surface detector ( sd ) array configuration and its trigger response , the aperture is also calculated for a typical configuration of the present phase . | one of the main objectives of the pierre auger observatory @xcite is the measurement of the flux of cosmic rays above 10@xmath0 ev .
an accurate knowledge of the aperture is an important piece of this measurement , in particular in its different detection configurations : sd - only , fd - only and hybrid . for this purpose
it is mandatory to use a detailed simulation to exploit dependences on the numerous parameters . the detector aperture , integrated over solid angle , is given by @xmath1 where @xmath2 and @xmath3 are the differential and total solid angles respectively .
excluding inclined showers , @xmath4 varies from 0 to @xmath5 .
@xmath6 is the effective area , @xmath7 is the overall trigger efficiency and @xmath8 is the reconstruction efficiency . @xmath9 and @xmath10 are the differential and total areas where shower events hit ground level . equation ( [ apert ] ) allows the calculation of aperture for different conditions and detector configurations .
single detector ( sd or fd ) apertures are obtained using the relevant detector efficiency functions for @xmath7 and @xmath8 . on the other hand , a hybrid aperture calculation has to include efficiencies of both detectors and therefore each efficiency function has to be read in ( [ apert ] ) as the product @xmath11 .
under the assumption that all six telescopes forming an fd eye have identical characteristics one can restrict the generation area to a slice with a vertex at a telescope , radius from 0 up to a maximum distance and azimuthal width equal to the telescope field of view ( i.e. @xmath12 around the telescope axis ) .
efficiency is calculated as a function of polar coordinates of shower impact points .
because of left - right symmetry , efficiency is a function of @xmath13 and @xmath14 , the absolute azimuthal distance of the impact point from the telescope axis azimuth .
the assumption of identical telescopes is needed to extend the evaluation of @xmath15 for @xmath16 , using the efficiency measured by the neighbouring telescopes .
the simulation fdsim @xcite was used for generating showers from two primary types , protons and iron nuclei , in 8 energy bins from 10@xmath17 to 10@xmath18 ev .
two different atmospheres were used with vertical aerosol optical depth ( vaod ) above the fd level of 0.03 and 0.06 , respectively .
these atmospheres will be referred to as `` clean '' and `` dirty '' , having respectively lower and higher vaod with respect to the typical malarge atmosphere@xcite . in total 700,000 shower profiles
have been generated .
trigger efficiency has been calculated in cells of @xmath19 km times @xmath20 for four separate cases : protons and iron nuclei in `` clean '' and `` dirty '' atmospheres .
the basic trigger function @xmath15 has been regularized and parameterised , at each energy , through a fit to an empirical function . during the current period of data
taking only some telescopes are fully equipped with a corrector ring ( an optical device for increasing aperture ) @xcite . therefore , for a selected configuration of primary and atmosphere ( protons in `` clean '' atmosphere ) , we simulated 20,000 shower profiles with this optics , to study the differences between telescopes with or without a corrector ring .
using the basic trigger function discussed in the previous section it is straightforward to evaluate the efficiency @xmath21 to trigger a shower by _ at least one telescope in the eye_. this function is mapped onto the ground . for this purpose
the ground area is divided into a grid of 0.5 @xmath22 0.5 km@xmath23 cells and the corresponding polar coordinates @xmath24 relative to each eye is calculated .
this allows the evaluation of @xmath21 at the cell center as seen by each eye for the different detection configurations , i.e. _ mono _ , _ stereo _ , _ 3-fd _ or _
4-fd_. the corresponding aperture follows directly from the surface integral in eq .
( [ apert ] ) .
the integral has been performed over a domain fully including the sd array in its final extension and assuming @xmath25 within the array boundary and 0 outside .
this represents the _ effective hybrid area in the ideal case of a fully efficient sd at any energy_. since a hybrid trigger requires only a sd single station trigger ( sd - t1)@xcite , this condition turns out to be fulfilled as soon as the energy exceeds 10@xmath0 ev .
then the aperture is @xmath26 where the bar signs stand for the average over solid angle , with a @xmath27 @xmath28 distribution .
p.45 p.45 , title="fig : " ] & , title="fig : " ] in fig .
[ hyb ] the aperture calculated for different fd configurations is shown .
the mono - hybrid aperture is practically independent of primary type , but is slightly sensitive to atmosphere parameters .
the relative difference between the two atmospheres is about 15% at @xmath29 ev and vanishes at higher energies . for more demanding
detection configurations ( stereo , 3-fd and 4-fd ) , apertures tend to be more sensitive to both composition and atmosphere .
apertures shown in fig .
[ hyb ] are trigger apertures . since the only requirement applied to showers
is the fullfillment of trigger conditions .
other conditions have to be applied if showers are to be reconstructed . in fig . [ rec ]
we show how the aperture changes when the shower detector plane ( sdp ) ( _ geometry level _ ) and the shower profile ( _ profile level _ ) are reconstructed . in this figure ,
sdp reconstruction was assumed to be fulfilled with at least 5 pixels .
the profile level requires a minimum number of 10 pixels and the shower maximum in the telescope field of view .
the aperture shown so far is based on fd efficiency only and for the full auger site as the generation area . however the auger detector is not static , but is continually changing : at present , since new tanks are added and put into operation and , at completion , since individual detectors ( sd tanks and/or fd telescopes ) can be occasionally out of operation or excluded because of non standard performance .
therefore the actual hybrid aperture in the construction phase and also in the final configuration has to be calculated using both fd and sd trigger responses .
the sd trigger efficiency has been extensively studied separately and in particular the use of a `` lateral trigger probability '' ( ltp ) function has been proposed @xcite for a complete description of the trigger features .
these ltp functions have been parameterised as a function of distance at various energies .
because in the current phase of data taking fd telescopes have different optical configurations ( with and without corrector ring ) , the basic trigger functions were calculated accordingly . as shown in fig [ crnocr ]
, the configuration with corrector ring allows more distant viewing at any angle . in fig .
[ hmono ] the mono hybrid trigger efficiency is shown as a function of energy , for the site configuration of mid october 2004 . at lower energies
the mono hybrid trigger accesses only the area covered by sd tanks just in front of each eye .
in particular at @xmath30 ev the single tank are easily visible as red spots in the trigger patterns . at higher energies ,
when the ltp range exceeds the distance between contiguous tanks , the area accessed by the hybrid trigger is roughly the intersection of fd patterns with sd domain .
p.45 p.50 & in fig .
[ aeffnow ] the hybrid trigger aperture is shown as a function of energy for mono and stereo detection as compared with sd - t1 detection ( sd - only single station trigger efficiency ) .
it can be seen that mono detection approaches sd - t1 as energy increases .
stereo detection becomes sizable above @xmath31 ev and then rapidly increases .
the use of basic trigger functions for both sd and fd detectors allows a detailed study of the local trigger probability and aperture in the case of hybrid detection .
this approach is particularly appealing since it limits the intensive cpu simulation to the generation of these functions , that can then be used very easily for aperture calculations .
the advantage is still more evident if one takes into account that simulation studies are largely detector specific .
finally we have shown that this approach allows the continuous evaluation of apertures as new detectors are added in the field and put into operation . 99 j. abraham
_ , the auger collaboration , nucl .
instr . meth .
a523 ( 2004 ) 50 - 95 l. prado jr , _ et al .
instr . meth .
a545 ( 2005 ) 632 - 642 .
r. cester _
et al . _ ,
contribution to this conference .
d. allard _ et al .
_ , contribution to this conference . | main pierre auger observatory @xcite flux cosmic rays above 10@xmath0 ev accurate knowledge aperture important detection configurations sd fd hybrid mandatory detailed simulation parameters detector aperture solid angle given by @xmath1 @xmath2 differential total solid angles @xmath4 varies from 0 to @xmath5 @xmath6 effective area @xmath7 overall trigger efficiency @xmath8 reconstruction efficiency @xmath9 @xmath10 differential total areas where shower events hit ground level equation allows calculation aperture for different conditions detector configurations single detector apertures obtained using detector efficiency functions for @xmath7 @xmath8 hybrid aperture calculation efficiencies both detectors each efficiency function read product @xmath11 six telescopes restrict generation area to slice with vertex at telescope radius from 0 maximum distance azimuthal width equal to telescope field of view @xmath12 around axis efficiency calculated function of polar coordinates of shower impact points efficiency function of @xmath13 @xmath14 azimuthal distance impact point from telescope axisassumption identical telescopes evaluation @xmath15 for @xmath16 efficiency neighbouring simulation fdsim @xcite used generating showers from two primary types protons and iron nuclei in 8 energy bins from 10@xmath17 to 10@xmath18 ev two atmospheres used with vertical aerosol optical depth above fd 0.03 and 0.06 atmospheres referred clean '' and `` dirty '' lower higher vaod 700,000 shower profiles generated trigger efficiency calculated in cells @xmath19 km times @xmath20 for four cases protons iron nuclei in clean dirty atmospheres basic trigger function @xmath15 regularized parameterised at each energy function some telescopes equipped with corrector ring @xcite protons simulated 20,000 shower profiles differences between telescopes with or without corrector ring trigger function efficiency @xmath21 shower by one telescope in the eye function mapped ground ground area divided into grid 0.5 @xmath22 0.5 km@xmath23 cells polar coordinates @xmath24 relative to each eye calculated allows evaluation of @xmath21 at cell center for different detection configurations mono stereoaperture follows from surface integral in eq . integral performed over domain including sd array in assuming @xmath25 within array boundary 0 outside represents effective hybrid area in efficient sd at any energy hybrid trigger requires sd single station trigger condition fulfilled energy exceeds 10@xmath0 ev aperture is @xmath26 bar signs for average over solid angle with @xmath27 @xmath28 distribution fig . aperture for different fd configurations mono - hybrid aperture independent of primary type slightly sensitive to atmosphere parameters relative difference between atmospheres about 15% at @xmath29 ev vanishes at higher energies for demanding detection configurations ( apertures more sensitive to composition and atmosphere apertures in fig are trigger apertures requirement fullfillment of trigger conditions other conditions if showers reconstructed fig . aperture changes when shower detector plane shower profile reconstructed sdp reconstruction assumed fulfilled with 5 pixels profile level requires minimum 10 pixels shower maximum in telescope field of view aperture based on fd efficiency for full auger site as generation areaauger detector not static continually changing new tanks added completion individual detectors sd tanks fd telescopes occasionally out of operation or excluded non standard performance hybrid aperture construction phase final configuration calculated using fd and sd trigger responses sd trigger efficiency studied separately lateral trigger probability ltp function proposed @xcite for trigger ltp functions parameterised function distance at various energies fd telescopes different optical configurations with without corrector ring basic trigger functions calculated accordingly configuration with corrector ring allows more distant viewing fig hybrid trigger efficiency function of energy site configuration mid october 2004 lower energies accesses area covered by sd tanks at @xmath30 ev single tank visible as red spots in trigger patterns higher energies ltp range exceeds distance between tanks area accessed trigger intersection of fd patterns with sd domain fig . hybrid trigger aperture function of energy for mono stereo detection compared with sd - t1 detection mono detection approaches sd - t1 as energy increases stereo detection sizable above @xmath31 ev increasesuse of basic trigger functions for sd fd detectors allows detailed study local trigger probability aperture hybrid detection approach limits intensive cpu simulation to generation functions used for aperture calculations advantage simulation studies detector specific approach allows continuous evaluation apertures as new detectors added. abraham auger collaboration nucl instr . meth. a523 ( 2004 ) 50 - 95 l prado jr instr. meth. a545 ( 2005 ) 632 - 642 r. cester _ contribution conference d. allard _ contribution | 0.310114 | 1,905 | 938 | , Saving $0.1 in GPT-4. |
20,793 | compact binary mergers , with neutron stars or neutron star and black - hole components , are thought to produce various electromagnetic counterparts : short gamma - ray bursts ( grbs ) from ultra - relativistic jets followed by broadband afterglow ; semi - isotropic kilonova from radioactive decay of r - process elements ; and late time radio flares ; etc .
if the jets from such mergers follow a similar power - law distribution of lorentz factors as other astrophysical jets then the population of merger jets will be dominated by low-@xmath0 values .
the prompt gamma - rays associated with short grbs would be suppressed for a low-@xmath0 jet and the jet energy will be released as x - ray / optical / radio transients when a shock forms in the ambient medium . using monte carlo simulations , we study the properties of such transients as candidate electromagnetic counterparts to gravitational wave sources detectable by ligo / virgo . approximately 78% of merger - jets result in failed grb with optical peaks 14 - 22 magnitude and an all - sky rate of 2 - 3 per year . | the merger of a binary system due to gravitational wave ( gw ) emission where the binary components are either neutron stars ( ns ) or stellar mass black holes ( bh ) is a potential source of gws detectable by current gravitational wave detectors ( e.g. ligo / virgo ) . in 2015 ,
advanced ligo made two unambiguous detections at 5-@xmath1 ( gw150914 and gw151226 ) and a third possible detection at 87@xmath2 confidence ( lvt151012 ) of the inspiral and merger of binary stellar mass bh systems ( * ? ? ?
* ( the ligo scientific collaboration et al . 2016 ) ) .
an electromagnetic ( em ) counterpart to the inspiral and merger of a binary bh system is not generally expected and no em counterpart was detected for any of the observed gw signals ( * ? ? ?
* ( copperwheat et al . 2016 ) )
. the next gw breakthrough will be the detection of the inspiral and merger of ns - bh or ns - ns systems .
for such mergers we expect various em counterparts , where the nature of the counterpart depends on the viewing angle ( * ? ? ?
* ( metzger & berger 2012 ) ) . when a ns - bh or ns - ns system merge , the merger ejecta forms an accretion disk , tidal tail and disk wind .
the rapid accretion from a disk onto a newly formed bh is thought to power bipolar ultrarelativistic jets responsible for short grbs and their afterglow , the typical timescale for a short grb is seconds and hours to days for the associated afterglow (;
* nakar 2007 ) ; ) .
the more isotropic ejecta can power r - process nucleosynthesis where radioactive decay gives rise to macronova emission , these occur on a day to weeks timescale ( ( * ? ? ?
* tanvir et al . 2013 ) ; ( * ? ? ?
* berger et al . 2013 ) ; ( * ? ? ?
* tanaka 2016 ) ) . at later times ,
approximately a month , the merger ejecta can interact with the ambient medium giving rise to radio flares ( ( * ? ? ?
* nakar & piran 2011 ) ; ( * ? ? ?
* hotokezaka et al . 2016 ) ) .
except for two cases ( ( * ? ? ?
* cenko et al . 2013 ) , @xcite ) we currently detect grbs and their afterglow by a high energy @xmath3-ray trigger .
the variability timescale of the prompt @xmath3-ray emission indicates a dependence for the emiting region , @xmath4 , where @xmath5 is the dissipation radius and @xmath0 the bulk lorentz factor .
the photospheric radius @xmath6 , the point at which the fireball becomes optically thin , can be conservatively estimated by considering the density of electrons associated with a baryonic outflow , @xmath7 .
these conditions require grbs to be produced in ultrarelativistic jets with @xmath8 . for jets with @xmath9
the dissipation radius falls below the photoshere and the prompt @xmath3-rays would be suppressed .
such a failed grb will still have a broadband orphan afterglow .
the observation of afterglows associated with the relativistic jets from binary ns mergers is currently biased by the requirement of a high energy @xmath3-ray trigger , gw triggered searches may reveal a hidden population of low-@xmath0 merger jets by their on - axis orphan afterglow .
astrophysical jets from other accreting bh systems ( e.g. agn , blazars ) follow a simple power law distribution with a negative index , @xmath10 , where @xmath11 is @xmath12 for blazars ( * ? ? ?
* ( saikia et al . 2016 ) ) .
if the jets from mergers follow a similar distribution then failed grbs would outnumber short grbs .
m , dash - dotted ska1-low , and dashed 48 lofar ) , and the dashed magenta line in the bottom panel shows the distribution of bright events @xmath13 ( * ? ? ?
* ( lamb & kobayashi 2016)).,width=326 ]
to test whether the on - axis orphan afterglow of a failed grb from a compact stellar merger would be observable , given a gw detection , a monte carlo of @xmath14 events was generated .
the events followed the ( * ? ? ? * wanderman & piran ( 2015 ) ) redshift and luminosity distribution for short grbs .
the bulk lorentz factor followed a simple power law with index @xmath15 in the limits @xmath16 , and independent of a bursts energy . for details of the monte carlo and radiative transfer
see ( * ? ? ?
* lamb & kobayashi ( 2016 ) ) .
we found that for a sample in the range @xmath17 mpc , the ligo / virgo detection limit for face - on ns - ns mergers , that events below a line given by @xmath18 always result in failed grbs , where @xmath19 is the isotropic equivalent kinetic energy .
for the parameters used we found that 78@xmath2 of mergers resulted in failed grbs ; the peak flux and time for the on - axis orphan afterglows from the failed grbs is shown in figure [ f1 ] .
the model parameters used for the afterglows are : @xmath20 protons @xmath21 , microphysical parameters @xmath22 and @xmath23 , the index of the power - law distribution of random electrons accelerated at shock @xmath24 , and the jet half - opening angle @xmath25 ensuring that the jet break time is later than the deceleration time for our sample and is within the limits @xmath26 found by ( * ? ? ?
* fong et al . ( 2015 ) ) for short grbs .
the jet opening angle plays a role only when we estimate the jet break time . .
the beaming factor @xmath27 ( black dashed line ) , given a gw detection the simple approximation @xmath28 ( blue dash - dot line ) , and the numerical results ( red solid line).,width=326 ]
gw emission is strongest on - axis from a merger event ( * ? ? ?
* ( kochanek & piran 1993 ) ) , thus the probability of an on - axis merger is higher , given a gw detection , than if we consider only the isotropic case ( see figure [ f2 ] ) . if the lorentz factor of a jet is correlated with the opening angle , where lower @xmath0 gives a wider half - opening angle ( as indicated for long grbs by ( * ? ? ?
* ghirlanda et al . 2013 ) ) , then the rates of on - axis orphan afterglow following a gw trigger will be higher than those given here . by using the rate of short grbs within the ligo detection volume predicted from the swift short grb rate by (
* metzger & berger ( 2012 ) ) , 0.03 per year , and considering the all sky rate - we find approximately 2.6(26 ) on - axis orphan afterglows from ns - ns(ns - bh ) mergers per year when assuming a jet half - opening angle of @xmath29 .
em counterparts to gw emission from ns - ns and ns - bh mergers include radio flares , macronova , short grbs , and on and off axis ( orphan ) afterglow . a gw triggered search for em counterparts could reveal a hidden population of low-@xmath0 merger jets from such events and on - axis orphan afterglow of failed grbs are therefore a strong candidate for em follow - up searches .
the ( non)detection of such sources will help constrain the lorentz factor distribution of such jets ( e.g. clustered at high-@xmath0 , a power - law , a log - normal , or multiple populations ) , and provide constraints on the acceleration process of relativistic jets . | merger binary system due to gravitational wave emission neutron stars or stellar mass black holes potential source gws detectable by gravitational wave detectors ligo / virgo 2015 ligo made two detections at 5-@xmath1 gw150914 gw151226 third detection at 87@xmath2 lvt151012 of inspiral merger binary stellar mass bh systems ligo scientific collaboration 2016 electromagnetic counterpart to inspiral merger binary bh system not expected no em counterpart detected for observed gw signals copperwheat al 2016 next gw breakthrough detection inspiral merger ns - bh or ns - ns systems mergers expect em counterparts depends on viewing angle metzger & berger 2012 ns - bh or ns - ns system merger ejecta forms accretion disk tidal tail disk wind rapid accretion disk onto bh power bipolar ultrarelativistic jets short grbs afterglow typical timescale seconds hours to days afterglow nakar 2007 isotropic ejecta power r - process nucleosynthesis radioactive decay macronova emission day to weeks timescale tanvir 2013 berger 2013 tanaka 2016later times month merger ejecta interact ambient medium radio flares nakar piran 2011 hotokezaka 2016 two cases cenko 2013 detect grbs afterglow by high energy @xmath3-ray trigger variability emission indicates emiting region @xmath5 dissipation radius @xmath0 bulk lorentz factor photospheric radius @xmath6 fireball thin estimated density of electrons baryonic outflow @xmath7 conditions require grbs in ultrarelativistic jets with @xmath8 jets @xmath9 dissipation radius below photoshere @xmath3-rays suppressed failed grb broadband orphan afterglow observation afterglows relativistic jets from binary mergers biased by high energy @xmath3-ray trigger searches reveal hidden population low-@xmath0 merger jets on - axis orphan afterglow astrophysical jets from other accreting systems blazars follow simple power law distribution negative index @xmath10 @xmath11 @xmath12 for blazars saikia al . 2016 jets mergers similar distribution failed grbs outnumber short grbsdash - dotted ska1-low dashed 48 lofar magenta line bottom shows bright events @xmath13 lamb & kobayashi test on - axis orphan afterglow failed grb compact stellar merger monte carlo @xmath14 events generated events followed wanderman & piran 2015 redshift luminosity distribution short grbs bulk lorentz factor followed power law index @xmath15 limits @xmath16 independent bursts energy details monte carlo radiative transfer see lamb & kobayashi 2016 sample range @xmath17 mpc events below line @xmath18 result failed grbs @xmath19 isotropic equivalent kinetic energy 78@xmath2 mergers resulted failed grbs peak flux time on - axis orphan afterglows failed grbs shown figure f1 model parameters afterglows @xmath20 protons @xmath21 microphysical parameters @xmath22 @xmath23 power - law distribution random electrons accelerated shock @xmath24 jet half - opening angle @xmath25 jet break time later than deceleration time within limits @xmath26 found fong et al 2015 short grbs jet opening angle jet break timebeaming factor @xmath27 black line gw detection approximation @xmath28 blue line numerical results red gw emission strongest on - axis from merger event kochanek piran 1993 probability on - axis merger higher gw detection isotropic case figure f2 lorentz factor jet correlated with opening angle lower @xmath0 wider half - opening angle ghirlanda 2013 rates on - axis orphan afterglow gw trigger higher rate short grbs ligo detection volume swift short grb rate metzger berger 2012 0.03 per year all sky rate 2.6(26 ) on - axis orphan afterglows from ns - ns - bh mergers per year jet half - opening angle @xmath29 counterparts gw emission from ns ns - bh mergers include radio flares macronova short grbs on off axis orphan afterglow gw triggered search reveal hidden population low-@xmath0 merger jets on - axis orphan afterglow failed grbs strong candidate follow - up searches non)detection sources lorentz factor distribution jets high-@xmath0 normal populations acceleration process relativistic jets | 0.26538 | 1,780 | 944 | , Saving $0.1 in GPT-4. |
20,797 | we present results following from the coupled channel model of two @xmath0 resonances decaying into the @xmath1 and @xmath2 mesons .
the @xmath3 resonance can be described by _ two _ distinct _ poles_. it is shown that the discrepancy in the @xmath3 mass position between the crystal barrel collaboration and the e852 group can be explained and removed . in our model with parameters fixed by the present experimental data the @xmath3 _ can not _ be interpreted as a bound @xmath2 state although the @xmath2 forces in the s - wave isovector channel are attractive . | among the scalar mesons there are only two isovector resonances @xmath3 and @xmath4 @xcite .
physical properties of both mesons are , however , not well known , for example the widths and branching ratios are poorly determined .
also the mass determination is problematical .
the proper interpretation of the resonant states can be obtained in terms of the s - matrix poles but the data are rarely analysed in such a way .
two @xmath2 thresholds at @xmath5 mev and at @xmath6 mev lie close to @xmath3
. therefore a description of the @xmath3 line shape using the simple breit - wigner formula leads to a _ distortion _ of the @xmath3 mass and width ( compare , for example differences between the results given in @xcite and @xcite ) . in @xcite montanet
indicated important differences of the @xmath4 masses and widths found in the @xmath7 annihilation by the crystal barrel collaboration @xcite and by the obelix collaboration @xcite .
recently the wa102 collaboration has observed the @xmath3 production in the pp central collisions at 450 gev , however the @xmath4 resonance has not been seen @xcite .
the main decay channels of the @xmath0 resonances are @xmath1 and @xmath2 . in the first part of this paper the @xmath3 production and decays
will be discussed using the flatt model @xcite . next we shall present the results following from the coupled channel model formulated in 1996 @xcite and further developed in @xcite
two experimental collaborations have used the flatt model to analyse their data [ 2 , 3 ] .
two different @xmath2 thresholds have not been distinguished . in the flatt model the effective mass distribution in the @xmath8channel ( @xmath9 )
is given by @xmath10 where c is a constant and the production amplitudes are given by @xmath11 the @xmath1 partial width is a product of the coupling constant @xmath12 and the @xmath1 momentum @xmath13 : @xmath14 . similarly above the @xmath2 threshold @xmath15 .
however , below the threshold this width becomes imaginary : @xmath16 , where @xmath17 and @xmath18 denotes the kaon mass . near
the @xmath2 threshold @xmath19 varies slowly and @xmath20 varies _
rapidly_. therefore the @xmath1 distribution is _ narrowed _ on both sides of the @xmath2 threshold as compared to the breit - wigner shape characteristic for the constant width ( independent on @xmath21 ) .
the second important feature of the flatt model is the existence of _ two _ complex poles of the production amplitudes @xmath22 at different energies corresponding to the same meson @xmath3 .
let us derive the pole positions following from the parameter values used by the e852 group in their analysis of the reaction @xmath23 @xcite . at the beginning one
has to correct the coupling constants for the finite experimental energy resolution .
then the @xmath1 coupling constant reduces to the value @xmath24 from which we have found two poles at @xmath25 mev on sheet ii and @xmath26 mev on sheet iii .
these values are in very good agreement with the pole positions of the crystal barrel collaboration recalculated and _ corrected _ by us : @xmath27 mev on sheet ii and @xmath28 mev on sheet iii .
the energy differences are indeed quite large : @xmath29 mev ; the corresponding difference of the total widths is @xmath30 mev .
mass determination from @xcite supplemented by our determination , indicated by crosses and described in the text ] in fig .
1 we show crosses with errors corresponding to our two pole determination of the real energy parts for the experiments @xcite and @xcite .
one can notice that about 3@xmath31 discrepancy between the e852 value , based on the breit - wigner form , and the mean value determination by the particle data group @xcite disappears when one considers the pole value on sheet ii at 988 mev . by looking at fig .
[ fig : pdg ] one can try to answer often asked question : where is located the @xmath3 , below or above the @xmath2 threshold ?
the most probable answer is : the first @xmath3 pole on sheet iii is located below and the second @xmath3 pole on sheet ii is located above the @xmath32 threshold .
one remark is relevant here : it would be useful to print in the review of particle physics the @xmath3 masses and its widths in the form of the complex energy values corresponding to different determinations of the t - matrix poles .
such a presentation is already given for the @xmath33 and the @xmath34 [ 1 , 10 ] .
the @xmath3 coupling to the @xmath2 channel is responsible for the appearance of two @xmath3 poles . if @xmath35 then @xmath36 . if , however , @xmath37 then @xmath38 and @xmath39 in these equations @xmath40 is the @xmath1 relative momentum at the @xmath2 threshold and @xmath41 denotes the @xmath2 momentum at the @xmath3 pole on sheet iii .
below we present results of the simple two - channel model of @xmath0 resonances applied in @xcite to the analysis of experimental results . in this model four reactions : @xmath42 , @xmath43 , @xmath44 and @xmath45 are described simultaneously using the separable interactions in the form @xmath46 here @xmath47 are the real coupling constants and @xmath48 are the yamaguchi form factors inversely proportional to @xmath49 , where @xmath50 is the c.m .
momentum and @xmath51 are constants .
the model has altogether only five independent parameters : the @xmath1 coupling constant @xmath52 , the @xmath2 coupling constant @xmath53 , the interchannel coupling @xmath54 and two range parameters @xmath55 and @xmath56 . the t - matrix satisfies the lippmann
schwinger equation @xmath57 , where g is the propagator matrix .
we fix four model parameters by choosing the @xmath3 pole of the t - matrix on sheet ii at @xmath58 mev and the @xmath4 pole on sheet iii at @xmath59 mev .
the fifth model parameter is constrained by the experimental @xmath2/@xmath1 branching ratio near the @xmath2 threshold @xmath60 where @xmath61 .
we have chosen @xmath62 and @xmath63 , corresponding to the branching ratio for the neutral @xmath3 decays .
the branching ratio depends very strongly on the upper integration limit @xmath64 as shown in fig .
the crystal barrel result @xcite @xmath65 corresponds to large value of @xmath66 mev . the upper limit of the wa102 group @xcite is lower as it is equal to @xmath67 mev since this group studied the @xmath68 decay into @xmath69 .
their measured value @xmath70 agrees well with the theoretical value 0.19 shown in fig .
the obelix ratio @xmath71 @xcite is also in a good agreement with the theoretical curve .
decay branching ratio on the effective mass upper limit ]
according to @xcite we take the @xmath4 mass equal to @xmath72 mev and its total width @xmath73 mev . then the @xmath2/@xmath1 branching ratio calculated in the limits of @xmath21 between @xmath74 and @xmath75 is equal to 0.98 .
when this ratio is evaluated in the slightly larger limits between @xmath76 mev and @xmath77 mev then it decreases to @xmath78 .
both numbers stay well within the experimental value @xmath79 found by the crystal barrel collaboration @xcite .
thus the model provides us with the theoretical branching ratios consistent with the experimental findings for both @xmath0 mesons .
there are many other predictions coming from the coupled channel model described previously .
one can calculate the elastic and transition amplitudes , the effective mass distributions in two channels , positions of other t - matrix poles not initially imposed in the procedure of fixing the model parameters and the coupling constants in two channels at each pole .
some of these predictions have been already published in @xcite .
we have calculated rather important energy difference of the @xmath3 poles : @xmath80 mev and the width difference @xmath81 mev .
these numbers are in qualitative agreement with the results found in sect .
2 using the flatt model .
three interesting physical quantities are related to the diagonal s - matrix elements in two channels : @xmath82 the inelasticity @xmath83 is common for both channels , what follows from the unitarity condition , however the phase shifts @xmath84 are different in both channels , as shown in fig .
[ fig : phase ] .
the sudden rise of the @xmath1 phase shifts near the @xmath2 threshold is due to the presence of the @xmath3 pole on sheet ii .
rather weak variation of the @xmath2 phase shifts near @xmath85 gev is a result of the destructive interference between two related @xmath3 poles lying on sheets ii and iii .
the interference effects are also seen near @xmath86 mev where in the @xmath1 channel the @xmath4 pole interferes with the @xmath87 zero related to the second @xmath4 pole . in the @xmath2 channel ,
however , the interference is constructive and the @xmath2 phase shifts rise very quickly in vicinity of @xmath88 mev .
the presence of the @xmath0 resonances leads to two characteristic dips of the inelasticity coefficient @xmath83 .
the present model is suitable to study the question whether a @xmath2 pair can form a bound s - wave isospin @xmath85 state .
this question is relevant since the coupling between kaons in that state is negative so the forces in the isovector state are attractive . in order to answer this interesting question
we have studied the evolution of the t - matrix poles in the limit of vanishing interchannel coupling constant @xmath54 . in this limit
one can find that the @xmath2 resonance exists at the complex effective mass equal to @xmath89 mev .
if the interchannel coupling is switched on then this resonance evolves into the @xmath4 pole and not into the @xmath3 pole .
the energy of the bound state would be real and smaller than the sum of two kaon masses . since this is not a case in our model the answer about
the presence of the bound @xmath2 isovector state is _
we have found that the @xmath3 meson can be described in terms of two distinct poles lying near the @xmath2 threshold on sheets ii and iii .
this result is a common feature of the two models , namely the flatt model , often used to describe a single @xmath0 resonance , and the coupled channel model of two @xmath0 resonances , constructed recently by us .
the second model has been constrained in the @xmath3 mass range and then successfully applied at higher effective @xmath1 or @xmath2 masses where the @xmath4 resonance is present .
the two channel model can be extended to treat more decay channels . | scalar mesons two isovector resonances @xmath3 @xmath4 @xcite physical properties not known widths branching ratios poorly determined mass determination problematical interpretation resonant states s - matrix poles data rarely analysed @xmath2 thresholds @xmath5 mev close to @xmath3 description @xmath3 line shape breit - wigner formula leads distortion @xmath3 mass width montanet indicated differences @xmath4 masses widths in @xmath7 annihilation crystal barrel collaboration obelix collaboration wa102 collaboration observed @xmath3 production in central collisions at 450 gev @xmath4 resonance not seen @xcite main decay channels @xmath0 resonances @xmath1 and @xmath2 @xmath3 production decays discussed flatt model @xcite results from coupled channel model 1996 @xcite developed @xcite two experimental collaborations used flatt model data two different @xmath2 thresholds not distinguishedflatt model mass distribution @xmath8channel given by @xmath10 c constant production amplitudes by @xmath11 @xmath1 partial width product of coupling constant @xmath12 @xmath1 momentum @xmath13 @xmath14 above @xmath2 threshold @xmath15 below width becomes imaginary @xmath16 @xmath17 kaon mass near @xmath2 threshold @xmath19 varies slowly @xmath20 rapidly @xmath1 distribution narrowed both sides threshold breit - wigner shape constant width independent @xmath21 two complex poles production amplitudes @xmath22 different energies same meson @xmath3 pole positions from parameter values e852 group reaction @xmath23 @xcite correct coupling constants for finite experimental energy resolution @xmath1 coupling constant reduces to @xmath24 two poles at @xmath25 @xmath26 values with pole positions crystal barrel collaboration recalculated corrected @xmath27 @xmath28 energy differences large @xmath29 mev difference total widths @xmath30 mev mass determination from @xcite supplemented by our determination indicated crosses fig .show crosses errors two pole determination energy parts experiments @xcite 3@xmath31 discrepancy between e852 value wigner mean value determination particle data group @xcite disappears pole value sheet ii at 988 mev fig . @xmath3 below or above @xmath2 threshold ? first @xmath3 pole sheet iii below second @xmath3 pole above @xmath32 threshold print review particle physics @xmath3 masses widths complex energy values determinations t - matrix poles presentation given for @xmath33 @xmath34 @xmath3 coupling to @xmath2 channel appearance two @xmath3 poles if @xmath35 @xmath36 @xmath37 @xmath38 @xmath39 @xmath40 @xmath1 relative momentum at @xmath2 threshold @xmath41 @xmath2 momentum at @xmath3 pole on sheet iii present results two - channel model of @xmath0 resonances applied in @xcite analysis experimental resultsmodel four reactions @xmath42 described separable interactions coupling constants yamaguchi form factors proportional to @xmath49 c.m momentum constants model five independent parameters @xmath1 coupling constant coupling constant interchannel coupling @xmath54 range parameters @xmath55 t - matrix satisfies lippmann schwinger equation @xmath57 g propagator matrix four parameters choosing @xmath3 pole matrix ii @xmath58 mev @xmath4 pole @xmath59 mev fifth model parameter constrained by experimental @xmath2/@xmath1 branching ratio threshold chosen @xmath62 @xmath63 branching ratio neutral @xmath3 decays branching ratio depends on upper integration limit @xmath64 crystal barrel result @xmath65 corresponds @xmath66 mev upper limit wa102 group @xcite lower equal to @xmath67 mev studied @xmath68 decay measured value @xmath70 agrees with theoretical value 0.19 obelix ratio @xmath71 @xcite theoretical curvedecay branching ratio effective mass upper limit @xcite @xmath4 mass equal @xmath72 mev total width @xmath73 mev @xmath2/@xmath1 branching ratio limits @xmath21 between @xmath74 @xmath75 equal 0.98 ratio limits @xmath76 @xmath77 decreases to @xmath78 numbers within experimental value @xmath79 model provides theoretical branching ratios consistent experimental findings @xmath0 mesons predictions from coupled channel model calculate elastic transition amplitudes effective mass distributions two channels positions t - matrix poles coupling constants two channels predictions published @xcite calculated energy difference @xmath3 poles @xmath80 mev width difference @xmath81 mev qualitative agreement with results sect . 2 flatt model three physical quantities diagonal s - matrix elements two channels @xmath82 inelasticity @xmath83 common both channels phase shifts @xmath84 different both channels fig . rise @xmath1 phase shifts near @xmath2 threshold due @xmath3 pole on sheet iiweak @xmath2 phase shifts near @xmath85 gev destructive interference between @xmath3 poles on sheets ii and iii interference effects seen near @xmath86 mev @xmath4 pole interferes with @xmath87 zero @xmath2 channel interference constructive phase shifts rise quickly @xmath88 mev @xmath0 resonances leads to two dips inelasticity coefficient @xmath83 model @xmath2 pair form bound s - wave isospin @xmath85 state coupling between kaons negative forces in isovector state attractive studied evolution t - matrix poles in limit vanishing interchannel coupling constant @xmath54 @xmath2 resonance exists at effective mass equal to @xmath89 mev if coupling switched on resonance evolves into @xmath4 pole not @xmath3 pole energy bound state smaller than sum two kaon masses not in our model answer bound @xmath2 isovector state @xmath3 meson two distinct poles near @xmath2 threshold on sheets ii and iii common feature of two models flatt coupled channel model two resonancessecond model constrained @xmath3 mass range applied @xmath1 masses @xmath4 resonance two channel model extended more decay channels | 0.419125 | 2,504 | 1,312 | , Saving $0.1 in GPT-4. |
20,833 | in this letter we investigate the information provided by the compass rose " ( crack , t.f . and ledoit , o. ( 1996 ) , journal of finance , 51(2 ) , pg .
751 - 762 ) patterns revealed in phase portraits of daily stock returns .
it has been initially suggested that the compass rose is just a manifestation of price clustering and discreteness and the tick size , factors that can affect the unbiasedness of an array of statistical tests based on stock returns .
we show that this may not entirely be the case .
price clustering and discreteness , microstructure , compass rose , nonlinear and complex dynamics , surrogate data analysis .
+ ( jel g10 ; g12 ; g22 ; z00 ) .
02.50.-r , 02.50.tt , 05.45.tp , 05.45.ac . | price clustering and discreteness is considered an important chapter of the market microstructure " literature with serious implications for tests of market efficiency , risk evaluation techniques and optimal design of securities procedures .
the _ compass rose _ " , introduced dy @xcite , is simply the manifestation of price clustering and discreteness in two or three dimensional phase portraits .
more precisely , a pattern is usually revealed in scatter diagrams of daily percentage returns against their lagged values : rays emanating from the center of the portrait ( the origin of the cartesian axis system ) , generating a compass rose like formation of clusters of points .
@xcite identify three conditions necessary for the appearance of the compass rose pattern :
1 . the daily price changes of the stock should be small relative to the price level ; 2 .
daily price changes should also be effected in discrete jumps of a small number of ticks and 3 .
the price of the stock should vary over a relatively wide range .
if any of the above points is violated , compass rose patters will be very weak or fail to realize altogether .
several papers have recently appeared on the compass rose theme .
some mainly confirm the compass rose as a result of the tick size , price clustering and discreteness and investigate or relax the above conditions under which it appears ( see @xcite , @xcite , @xcite , @xcite , @xcite , @xcite and @xcite ) .
others ( see @xcite , @xcite , @xcite and @xcite ) also concentrate on how the above factors affect the validity of statistical tests based on stock return sequences that reveal this nanostructure " @xcite .
the compass rose should have appeared in several other works ( such as @xcite , @xcite , @xcite , @xcite , @xcite , @xcite and @xcite ) .
however this was not achieved due to an unfortunate choice of graphical representation style or resolution and of the length ( history ) of the sequences examined .
@xcite suggested originally that the compass rose may not offer any help in predicting returns series .
this is due to an apparent absence of any strong temporal continuity pattern in the phase portraits ( see fig . [
fig : fig1]a ) . in this letter
we show that from a small manipulation of the information that appears in the compass roses , we can gain additional information on the dynamics of stock returns processes in a very costless algorithmically way .
an initial analysis of this information suggests the presence of strong nonlinear and possibly deterministic dynamics .
price clustering and discreteness , as manifested in the compass rose , appears in the form of rays that emanate from the center of the phase portraits ( see fig . [
fig : fig1](a ) were we have a detail of the compass rose for the returns of the tesco stock , chosen randomly from stocks forming to the ftse100 index ) .
it has been suggested that there is no predictability in such phase portraits @xcite .
this is due to temporal structure and information being concealed in the pase portraits of return sequences .
one could initially adopt the view that the distribution of the points in the compass rose in fig .
[ fig : fig1 ] could have been generated by a suitably discretized random gaussian process .
however @xcite showed that there may be a more delicate temporal pattern hidden under noise in the compass rose .
[ insert fig .
[ fig : fig1 ] about here . ] in this letter we follow the approach of @xcite and instead of examining directly the distribution of points in the phase portraits , we choose to model the value of the arcs @xmath0 formed by the line joining the point with the origin ( 0,0 ) and the horizontal axis ( fig .
[ fig : fig1]a ) .
we measure this in degrees for simplicity .
an advantage of such recoding " of the compass rose information is that all prices will range between 0 and 2@xmath1 or 0 and 360 degrees .
it will also allow us to observe more efficiently the clustering and obtain a non - subjective view of the compass rose pattern ( refer to @xcite and @xcite for discussions ) .
indeed , the investigation of the distribution of the compass rose arc values for tesco ( fig .
[ fig : fig1]b ) , shows that there is concentration of points across the horizontal and vertical axis ( 0 , 90 , 180 and 270 ( -90 ) degrees ) and the diagonal rays ( close to 45 , 45 + 90 , 45 + 180 and -45 degrees i.e. , the ne - sw and nw - se directions of the compass rose ) .
this is referred to as _ x - skewness _ in @xcite and attributed to the presence of big - players and herding ( see @xcite and @xcite ) .
they also suggested that x - skewing may be inconsistent with arch effects and demonstrate this with simulations.(b ) may also be an indication of more complex dynamics .
see in @xcite and @xcite for discussions . ]
@xcite also revealed a similar result through a different approach .
judging from the above literature , it seems that several differently fabricated nonlinear and non - random sequences can pass as ( g)arch processes .
there is also the case that stock returns may be characterized by more complex dynamics , not excluding deterministic or chaotic structures ( refer to @xcite and @xcite for more details ) .
if we plot the arc values @xmath0 as a time series we also obtain a pattern that confirms this clustering of points ( see fig .
[ fig : fig1]c ) .
plotting a sorted version of the sequence in fig .
[ fig : fig1](c ) , provides us with the pattern observed in [ fig : fig1](d ) , where the plateaus indicate more clearly where the clustering occurs .
the intensity of the compass rose patterns may differ between stocks , however the clustering along the main directions as discussed above , usually prevails .
searching for some type of temporal dependency , we looked into the phase portraits of the arcs . in fig .
[ fig : fig2](a ) we have the first lag phase portrait and in fig .
[ fig : fig2](b ) the second lag one .
we can see very curious patterns arising in both diagrams .
there are grids ( fig .
[ fig : fig2]b ) and rays ( fig .
[ fig : fig2]a ) which correspond to the main rays of the compass rose as seen in fig .
[ fig : fig1](c ) and ( d ) .
these patterns need more analysis .
[ insert fig .
[ fig : fig2 ] about here . ] in order to provide evidence of some kind of dependency in the above patterns we calculated the bds test ( see @xcite and @xcite ) for the tesco sequence of arc values @xmath0 . in table
[ tab : tab1 ] we see clearly that for a range of dimensions @xmath2 ( 2 to 10 ) , the bds test reports non iid dynamics for neighborhood area ranges between 0.5 and 1.5 times the standard deviation of arcs .
when the size becomes twice the standard deviation ( which is regarded as a large radius ) , we can accept independence only up to dimension 3 .
this is an initial indication that the dynamics as observed from the sequence of arcs , may contain some dependency that could be used for forecasting purposes .
[ insert table [ tab : tab1 ] about here . ] to back up this result , we calculated the bds test within a surrogate data analysis framework ( sda : @xcite ) .
sda is a permutation test framework , similar to bootstrapping , and is used to test specific nulls that exclude certain dynamics .
every null comes with its own tailor made simulation procedures for creating surrogate data sets from the original sequences ( refer to @xcite , @xcite , @xcite and @xcite for more details ) . in nutshell , via sda one searches for large discrepancies between the statistic values on the original and surrogate sequences .
when this occurs , one can safely reject the hypothesis that the original sequence stems from a process that is in compliance with the null .
sda is appropriate here as it allows us to bypass the limitations of chaotic invariance measures ( such as lyapunov exponents and dimension statistics ) due to small data sets and noisy information .
although the bds is a test for independence and not nonlinear determinism and complexity , under the sda framework it can be used to test for absence of stochastic randomness via the exclusion of the null hypothesis .
for our case we used the null of the @xmath0 sequences being _ a monotonic nonlinear transformation of linearly filtered noise _
, which is also regarded as the most interesting " .
this implies strongly , the absence of stochastic ( random ) dynamics .
we analyzed data from 53 ftse 100 index stocks , spanning the period 01/01/1970 to 5/30/2003 ( a maximum of 8717 observations ) . in tables
[ tab : tab2 ] and [ tab : tab3 ] we present the results for 5% and 2.5% levels of statistical significance , for neighborhood sizes @xmath3 ranging from 0.5 to 2 times the standard deviation of @xmath0 .
we clearly see that for up to 1.5 times the standard deviation , the null can be safely rejected ( observe the large biases ) .
however , for @xmath3 equalling twice the standard deviation , we can not always reject the null ( but this is only for a few cases ) .
our sda results do not reject the case of forecastability on the basis of arc values .
they also suggest the presence of complex dynamics and the possibility of some level of determinism .
[ insert table [ tab : tab2 ] about here . ] [ insert table [ tab : tab3 ] about here . ]
by adopting an approach similar to @xcite we showed that the compass rose @xcite can provide useful information for understanding further the dynamics of stock return sequences .
moreover , we do not reject the case of these dynamics being complex and forecastable .
we also do not exclude the possibility of nonlinear determinism .
however more research is needed on this controversial area ( see @xcite , @xcite , and @xcite ) .
an interesting area for future research would be to experiment with differently simulated processes ( such as purely chaotic , stochastic , ( g)arch and mixtures of these ) and examine similarities or dissimilarities with the results for the framework we follow in this letter .
it would also be interesting to see how inhomogeneous sampling of such processes could alter the results we report here . that could provide useful information on how our view of the stock return dynamics changes as we move from high - frequency data to lower frequencies , which is also an issue that has been concerning the general compass rose literature .
antoniou , a. , vorlow , c. e. , 2004 .
price clustering and discreteness : is there chaos behind the noise ? under review in physica
a : statistical mechanics and its applications .
http://arxiv.org/abs/cond-mat/0407471 chen , p. , 1993 .
searching for economic chaos : a challenge to econometric practice and nonlinear tests . in : day , r. h. , chen , p. ( eds . ) , nonlinear dynamics and evolutionary economics .
oxford university press , oxford ; new york
; toronto and melbourne , pp . 21753 .
kyrtsou , c. , terraza , m. , 2003 .
is it possible to study chaotic and arch behaviour jointly ?
application of a noisy mackey - glass equation with heteroskedastic errors to the paris stock exchange returns series .
computational economics 21 ( 3 ) , 257276 .
rrrrr + @xmath3 : & @xmath4 & @xmath5 & @xmath6 & @xmath7 + @xmath2 & 0.897 & 1.794 & 2.691 & 3.588 + 2 & 745.31 & 241.38 & 65.39 & 3.45 + 3 & 846.67 & 223.27 & 56.12 & 2.84 + 4 & 1010.50 & 220.27 & 49.64 & 1.68 + 5 & 1243.98 & 222.22 & 44.97 & 0.95 + 6 & 1577.48 & 229.07 & 41.54 & 0.39 + 7 & 2053.94 & 239.21 & 38.59 & @xmath80.18 + 8 & 2743.93 & 252.22 & 36.10 & @xmath80.72 + 9 & 3741.75 & 269.69 & 34.16 & @xmath81.08 + 10 & 5201.71 & 291.13 & 32.85 & @xmath81.16 + + 2 & 0.00 & 0.00 & 0.00 & 0.00 + 3 & 0.00 & 0.00 & 0.00 & 0.00 + 4 & 0.00 & 0.00 & 0.00 & 0.09 + 5 & 0.00 & 0.00 & 0.00 & 0.34 + 6 & 0.00 & 0.00 & 0.00 & 0.69 + 7 & 0.00 & 0.00 & 0.00 & 0.86 + 8 & 0.00 & 0.00 & 0.00 & 0.47 + 9 & 0.00 & 0.00 & 0.00 & 0.28 + 10 & 0.00 & 0.00 & 0.00 & 0.25 + .surrogate data analysis results on arc values @xmath0 for 53 companies in the ftse100 . discriminating statistic : bds test ( embedding dimension 2 ) .
neighborhood size @xmath9 , @xmath10 , @xmath11 and @xmath12 , where @xmath13 standard deviation of @xmath0 sequence .
biases and standard errors reported for significance level @xmath14 . [ cols= "
< , > , > , > , > , > , > , > , > , > , > , > , > " , ] | price clustering discreteness important chapter market microstructure literature implications for tests market efficiency risk evaluation techniques optimal design securities procedures compass rose _ introduced dy @xcite manifestation of price clustering discreteness in two three dimensional phase portraits pattern revealed in scatter diagrams of daily percentage returns against lagged values : rays emanating from center portrait origin cartesian axis system generating compass rose like formation of clusters points @xcite identify three conditions for compass rose pattern : daily price changes stock small relative to price level in discrete jumps small number ticks 3 price stock vary over wide range if violated compass rose patters weak or fail to realize papers appeared on compass rose theme some confirm compass rose result tick size price clustering discreteness investigate conditions see @xcite @xcite others concentrate on factors affect validity of statistical tests based on stock return sequences nanostructure compass rose should have appeared in other works not achieved due to unfortunate choice of graphical representation style resolution length history of sequences examined@xcite suggested compass rose may help predicting returns series due to absence strong temporal continuity pattern in phase portraits ( see fig fig1]a ) letter show small manipulation of information compass roses can gain additional information on dynamics stock returns costless initial analysis suggests strong nonlinear possibly deterministic dynamics price clustering discreteness manifested in compass rose appears in rays from center phase portraits ( see fig (a ) detail of rose for returns tesco stock chosen randomly from ftse100 index suggested no predictability in phase portraits due to temporal structure information concealed in portraits return sequences view distribution of points in compass rose could have been generated by discretized random gaussian process @xcite showed more delicate temporal pattern hidden under noise in compass rose letter follow approach @xcite instead examining distribution of points phase portraits model value of arcs formed by line joining point with origin and horizontal axis fig measure in degrees for advantage of recoding compass rose information prices range between 0 and 2 or 0 and 360 degreesobserve clustering obtain non - subjective view compass rose pattern refer @xcite @xcite discussions investigation distribution compass rose arc values for tesco fig shows concentration points across horizontal vertical axis ( 0 90 180 270 ( -90 ) degrees diagonal rays close to 45 45 + 90 45 + 180 -45 degrees ne - sw nw - se directions of compass rose referred as x - skewness _ in @xcite attributed to big - players herding x - skewing inconsistent with arch effects simulations indication complex dynamics @xcite discussions @xcite revealed similar result different approach differently fabricated nonlinear non - random sequences can pass as g)arch processes stock returns characterized complex dynamics deterministic chaotic structures refer @xcite @xcite details plot arc values as time series pattern confirms clustering points see fig . fig1]c ) plotting sorted version sequence in fig .(c ) provides pattern in fig : fig1](d ) plateaus indicate clustering intensity compass rose patterns differ between stocks clustering along main directions prevailssearching for temporal dependency looked into phase portraits of arcs fig(a ) first lag phase portrait(b ) second lag curious patterns in both diagrams grids ) rays ( ) correspond to main rays of compass rose in fig fig1](c ) and d ) patterns need more analysis insert fig fig2 ] evidence of dependency calculated bds test @xcite for tesco sequence of arc values @xmath0 table for range of dimensions @xmath2 ( 2 to 10 ) bds test reports non iid dynamics for neighborhood area ranges between 0.5 and 1.5 times standard deviation of arcs when size becomes twice standard deviation large radius accept independence up to dimension 3 initial indication dynamics may contain dependency for forecasting purposes table calculated bds test within surrogate data analysis framework ( sda : @xcite ) sda permutation test framework similar test specific nulls exclude certain dynamics every null tailor made simulation procedures for creating surrogate data sets from original sequences refer to @xcite for detailsvia sda searches for large discrepancies between statistic values original surrogate sequences reject hypothesis original sequence from process with null sda allows bypass limitations of chaotic invariance measures due to small data sets noisy information bds test for independence not nonlinear determinism complexity under sda framework can test for absence of stochastic randomness via exclusion null hypothesis used null of @xmath0 sequences monotonic nonlinear transformation of linearly filtered noise interesting implies absence of stochastic dynamics analyzed data from 53 ftse 100 index stocks 01/01/1970 to 5/30/2003 maximum 8717 observations tables tab2 tab3 results for 5% 2.5% levels of statistical significance for neighborhood sizes @xmath3 from 0.5 to 2 times standard deviation of @xmath0 for up to 1.5 times standard deviation null can be rejected for @xmath3 equalling twice standard deviation not always reject null few cases sda results do not reject forecastability arc values suggest complex dynamics possibility of determinism insert table tab2 table tab3adopting approach similar @xcite showed compass rose @xcite useful information understanding dynamics stock return sequences reject dynamics complex forecastable exclude nonlinear determinism more research needed controversial area see @xcite future research experiment with simulated processes chaotic stochastic mixtures examine similarities dissimilarities results framework interesting see inhomogeneous sampling alter results information view stock return dynamics high - to lower frequencies issue compass rose literature antoniou a. vorlow c. e. 2004 price clustering discreteness chaos behind noise ? review physica a statistical mechanics applications http://arxiv/cond-mat/0407471 chen p. 1993 searching for economic chaos challenge econometric practice nonlinear tests day r. h. chen p. nonlinear dynamics evolutionary economics oxford university press 21753 kyrtsou c. terraza m. 2003 study chaotic arch behaviour jointly ? application noisy mackey - glass equation with heteroskedastic errors paris stock exchange returns series computational economics 21 ( 3 ) , 257276@xmath3 + @xmath2 0.897 1.794 2.691 3.588 + 2 745.31 241.38 65.39 3.45 + 3 846.67 223.27 56.12 2.84 + 4 1010.50 220.27 49.64 1.68 + 5 1243.98 222.22 44.97 0.95 + 6 1577.48 229.07 41.54 0.39 + 7 2053.94 239.21 38.59 @xmath80.18 + 8 2743.93 252.22 36.10 @xmath80.72 + 9 3741.75 269.69 34.16 @xmath81.08 + 10 5201.71 291.13 32.85 @xmath81.16 + + 2 0.00 0.00 + 3 0.00 + 4 0.09 + 5 0.34 + 6 0.69 + 7 0.00 0.86 + 8 0.47 + 9 0.28 + 10 0.25 + .surrogate data analysis results arc values @xmath0 53 companies in ftse100 discriminating statistic bds test embedding dimension 2 ) neighborhood size @xmath9 @xmath13 standard deviation of @xmath0 sequence biases standard errors significance level @xmath14 | 0.43669 | 3,144 | 1,558 | , Saving $0.1 in GPT-4. |
20,834 | we argue that _ emergent behavior _ is inherent to cybersecurity . | the human - created cyberspace is a very large - scale complex system of cybersystems .
its security properties are difficult to understand and characterize .
we attribute this difficulty to its _ complexity _ , a manifestation of which is the so - called _ emergent behavior _ in complexity science @xcite .
although there is no universally accepted definition of emergent behavior @xcite , the basic idea is intuitive .
the simplest example of emergent behavior may be the well known @xmath0 " effect . for our purpose , it is sufficient to use the following informal definition of emergent behavior in cybersecurity domain .
[ def : emergent - behavior ] a security property of a cybersystem exhibits emergent behavior if the property is _ not _ possessed by the underlying lower - level components of the cybersystem .
a direct consequence of emergent behavior is that at least some security properties can not be understood by solely considering the lower - level components ; instead , we must explicitly consider the _ interactions _ between the lower - level components .
although emergent behavior of cybersystems has been discussed from a function or constructional perspective @xcite , emergent behavior in cybersecurity is not systematically examined until now .
we demonstrate emergent behavior in cybersecurity through three examples .
* example 1 : emergent behavior exhibited by cybersecurity dynamics .
* we refer to @xcite for an exposition of the emerging field of cybersecurity dynamics . in order to explain how emergent behavior is exhibited by cybersecurity dynamics
, we consider the perhaps simplest model @xcite . for our purpose
, it suffices to consider two cybersystems that respectively induce attack - defense structures ( or graphs ) @xmath1 , where @xmath2 is the node ( vertex ) set and @xmath3 is the edge set for @xmath4 .
let @xmath5 denote the largest eigenvalue of the adjacency matrix of a graph @xmath6 , @xmath7 denote the defense capability in detecting and cleaning compromised nodes ( e.g. , the probability that a compromised node gets cleaned at a time step ) , and @xmath8 denote the attack capability in compromising secure nodes ( e.g. , the probability that this event occurs over an edge at a time step ) . for simplicity ,
suppose @xmath9 is a complete graph with @xmath10 nodes for @xmath4 .
it is well known that @xmath11 and @xmath12 . for @xmath4 , if @xmath13 , the attacks will eventually be wiped out in the cybersystem that induces @xmath9 ; if @xmath14 , the attacks can not be wiped out @xcite .
now we consider a new cybersystem that is obtained by interconnecting the aforementioned two cybersystems that induced @xmath15 and @xmath16 .
consider the simplest case that any node can attack any other node in the interconnected cybersystem , which effectively induces attack - defense structure , a complete graph , @xmath17 with @xmath18 nodes and @xmath19 . in many ( if not all ) cases ,
the defense capability @xmath20 and the attack capability @xmath21 associated to @xmath17 are respectively the same as the defense capability @xmath7 and the attack capability @xmath8 associated to @xmath15 and @xmath16 .
since @xmath13 for @xmath4 do not imply @xmath22 , we conclude that the attacks can be wiped out in the two underlying component cybersystems , but can not be wiped out in the interconnected cybersystem as long as @xmath23 .
this phenomenon can be naturally extended to more sophisticated settings ( e.g. , @xcite ) .
this implies that cybersecurity dynamics can not be determined by looking at the component cybersystems alone .
rather , we need to look into how the component cybersystems interact with each other .
* example 2 : emergent behavior exhibited by security properties in the extended trace - property framework .
* in the field of program verification , it was known that specifications that are sufficient for _ sequential _ programs are not sufficient for _ concurrent _ programs . for dealing with concurrent programs , lamport proposed the safety - liveness framework of trace properties @xcite .
intuitively , a _ trace _ is a finite or infinite sequence of states corresponding to an execution of a program .
trace property _ is a set of traces such that every trace , _ in isolation _ , satisfies the same predicate .
a _ safety _
property says that no bad thing " happens during the course of a program execution , while a _ liveness _ property says that good thing " will eventually happen during the course of a program execution .
both safety and liveness are trace properties .
a beautiful result is that every trace property is the intersection of a safety property and a liveness property @xcite . given the above history , it is appealing to specify a cybersystem as a set of traces , and therefore as a subset of a security property that is also specified as _ a set of traces_. unfortunately , security properties are not trace properties as shown in @xcite and refreshed below .
first , _ noninterference _ is a security property that captures the intuition that system security is preserved as long as high - clearance ( or high - privilege ) processes can not influence the behavior of low - clearance ( low - privilege ) processes .
it is no trace property because it can not be verified without examining the other traces .
second , _ information - flow _ captures some kind of correlation between the values of variables in multiple traces .
it is no trace property because it can not be verified by examining each trace alone .
average service response time _ is an availability property .
it is no trace property because it depends on the response time in all traces . in an effort to overcome the above limitation of the safety - liveness framework , clarkson and schenider extended the notion of _ trace properties _ to the notion of _ trace hyperproperties _
basically , hyperproperties are _ sets of trace properties_. in parallel to the safety - liveness framework , a hyperproperty is also the intersection of a safety hyperproperty and a liveness hyperproperty . it is now known that information - flow , integrity and availability can be hypersafety or hyperliveness @xcite . exactly because hyperproperties capture that the verification procedure _ must _ examine across _ multiple _ traces , which may accommodate interactions between the component systems , we say that hyperproperties exhibit the emergent behavior .
this means that we need to study the emergent behavior in cybersecurity , which may explain why it took so long to realize the importance of hyperproperties .
* example 3 : emergent behavior exhibited by cryptographic security properties . *
cryptographic secure multiparty computation allows multiple parties @xmath24 , each having a respective secret @xmath25 , to compute a function @xmath26 such that no information about the @xmath27 s is leaked except for what is implied by the output of the function .
this manifests a confidentiality property .
a beautiful feasibility result is that any polynomial - time computable function @xmath28 can be securely computed @xcite , as long as the protocol executes _ in isolation _ ( the stand - alone setting ) and trapdoor permutations exist .
when such cryptographic protocols are used as building - blocks in larger applications / systems , they may execute concurrently ( rather than in isolation ) .
this leads to a natural question : are the cryptographic protocols , which are provably secure when executed in isolation , still secure when they are concurrently called by larger applications / systems ? intuitively , concurrent executions offer the attacker the leverage ( for example ) to schedule the messages in a way that is to the attacker s advantage , which does not have a counterpart in the stand - alone setting .
quite similar to what happened in the field of program verification , where specific properties ( e.g. , partial correctness and mutual exclusion ) were investigated before the introduction of the unifying safety - liveness framework @xcite , the same kind of development was made in the field of cryptographic protocols .
that is , specific cryptographic security properties were investigated before the introduction of the unifying notion called _ universal composability _
@xcite , or its equivalent ( but perhaps more intuitive ) version called _ concurrent general composition _
( arbitrarily many instances run possibly together with arbitrary other protocols ) @xcite .
it is now known that there are cryptographic multiparty computation protocols , which are provably secure when executed in isolation , but are _ not _ secure when they are concurrently called by larger applications / systems .
for example , there exist classes of functions that can not be computed in the universally composably secure fashion @xcite . in other words ,
these functions can be securely computed by running some cryptographic protocols in isolation , but can not be securely computed when the protocols execute concurrently . in order to make cryptographic multiparty computation protocols secure when they are used as building - blocks for constructing larger cybersystems , we need to make extra assumptions , such as that majority of the parties @xmath24 are not compromised @xcite .
this manifests emergent behavior .
( it is interesting to note that whether or not it is reasonable to assume that majority of the parties are not compromised may be addressed by the cybersecurity dynamics framework @xcite . ) | human - created cyberspace is large - scale complex system cybersystems security properties difficult to understand difficulty to complexity manifestation emergent behavior in complexity science no universally accepted definition of emergent behavior basic idea intuitive simplest example @xmath0 " effect sufficient to informal definition of emergent behavior in cybersecurity domain security property exhibits emergent behavior if not possessed by lower - level components consequence some security properties not understood by solely considering lower - level components must consider interactions between lower - level components emergent behavior discussed from function constructional perspective in cybersecurity not systematically examined until now demonstrate emergent through three examples example 1 : emergent behavior by cybersecurity dynamics refer to for emerging field cybersecurity dynamics explain consider simplest model consider two cybersystems attack - defense structures @xmath1 @xmath2 is node vertex set @xmath3 edge set for @xmath4 @xmath5 denote largest eigenvalue of adjacency matrix graph @xmath6 @xmath7 defense capability in detecting cleaning compromised nodesprobability compromised node cleaned time step @xmath8 denote attack capability compromising secure nodes event edge time step suppose @xmath9 complete graph with @xmath10 nodes for @xmath4 @xmath11 @xmath12 @xmath4 if @xmath13 attacks wiped out in cybersystem @xmath9 @xmath14 attacks not wiped out @xcite new cybersystem interconnecting two cybersystems @xmath15 @xmath16 any node attack other node interconnected cybersystem induces attack - defense structure complete graph @xmath17 with @xmath18 nodes @xmath19 defense capability @xmath20 attack capability @xmath21 @xmath17 same as defense @xmath7 attack @xmath8 @xmath15 @xmath16 @xmath13 for @xmath4 imply @xmath22 attacks wiped out in two underlying cybersystems not wiped interconnected cybersystem long @xmath23 extended to sophisticated settings @xcite cybersecurity dynamics not determined component cybersystems alone need cybersystems interact example 2 emergent behavior security properties in extended trace - property frameworkin program verification specifications sufficient for sequential programs not sufficient for concurrent programs for lamport proposed safety - liveness framework of trace properties @xcite a trace _ is finite or sequence of states to execution program trace property is set of traces every satisfies same predicate safety _ property says no bad thing happens during program execution liveness property says good thing will both safety and liveness are trace properties every trace property is intersection of safety and liveness property appealing to specify cybersystem as set of traces subset of security property security properties are not trace properties in @xcite noninterference _ is security property system security preserved high - clearance processes influence behavior low - processes no trace property verified without examining other traces _ information - flow _ captures correlation between values variables in multiple traces no trace property not verified by examining each trace alone average service response time _ is availability property no trace property depends on response time in all traces safety clarkson and schenider extended notion of trace properties to _ trace hyperproperties _ hyperproperties are sets of trace propertiesparallel to safety - liveness framework hyperproperty is intersection of safety and liveness information - flow integrity availability can be hypersafety or hyperliveness hyperproperties capture verification procedure examine across multiple traces accommodate interactions between component systems exhibit emergent behavior need to study emergent behavior in cybersecurity explain importance of hyperproperties example 3 : emergent behavior by cryptographic security properties cryptographic secure multiparty computation allows multiple parties @xmath24 secret @xmath25 to compute function @xmath26 no information @xmath27 leaked except implied output manifests confidentiality property any polynomial - time computable function @xmath28 can be securely computed @xcite protocol executes in isolation trapdoor permutations exist cryptographic protocols used in larger applications systems may execute concurrently question are cryptographic protocols secure when in isolation secure concurrently called by larger applications / systems ? concurrent executions offer attacker leverage to schedule messages attacker s advantage stand - alone setting similar to program verification specific properties partial correctness mutual exclusion investigated before unifying safety - liveness framework same development made in cryptographic protocolsspecific cryptographic security properties investigated before unifying notion _ universal composability _ @xcite or _ concurrent general composition _ arbitrarily many instances run with protocols @xcite now known cryptographic multiparty computation protocols provably secure when executed in isolation not secure concurrently called by larger applications systems classes of functions not be computed universally composably secure @xcite functions can securely computed by running cryptographic protocols in not when concurrently make cryptographic multiparty computation protocols secure larger cybersystems need extra assumptions majority of parties @xmath24 not compromised @xcite manifests emergent behavior majority parties not compromised addressed by cybersecurity dynamics framework @xcite | 0.315558 | 1,937 | 961 | , Saving $0.1 in GPT-4. |
20,840 | we extend a previous bispectrum analysis of the cobe - dmr 4 year maps , allowing for the presence of correlations between all possible angular scales .
we find that the non - gaussian signal found earlier for bispectrum components including adjacent modes does not extend to triplets of modes with larger separations . indeed for all separations
@xmath0 we find that the cobe - dmr data is very gaussian .
the implication seems to be that the previously detected non - gaussian scale - scale correlation falls off very quickly with mode separation . | the issue of whether or not the primordial fluctuations are gaussian is of fundamental importance in theories of structure formation .
gaussian large - scale primordial fluctuations are believed to be the hallmark of the simplest inflationary scenarios . in these theories
vacuum quantum fluctuation , with gaussian statistics , are stretched by a period of exponential expansion to give rise to the large scale fluctuations we observe today .
in contrast , topological defect theories , which are also possible candidates for structure formation @xcite , are believed to produce non - gaussian density fields .
non - gaussianity is also believed to be associated with some non - minimal inflationary models @xcite , such as those generating isocurvature fluctuations @xcite .
several recent @xcite papers have shown indications of non - gaussianity in the cosmic microwave background ( cmb ) temperature fluctuations , as measured by the cobe - dmr experiment .
the first of these detections + @xcite was later found to be due to an experimental systematic @xcite , and the second @xcite to an error of method @xcite .
the third claim @xcite , however , remains unassailable . in that work a bispectrum analysis was carried out , using adjacent scales @xmath1 as components of the bispectrum .
a significant deviation from gaussianity was found , with a high confidence level .
an obvious question is whether this non - gaussianity extends to components correlating scales with a larger separation , say @xmath2 , with @xmath0 .
the purpose of this paper is to answer this question .
as we shall see the non - gaussian signal found in @xcite does not appear to extend to larger separations @xmath3 .
in this letter we examine the possibility of deviations from gaussianity in terms of the bispectrum , which should be zero for a gaussian process .
let us expand the temperature fluctuations @xmath4 in spherical harmonic functions : @xmath5 a gaussian probability distribution function is completely described by the first two moments , @xmath6 and @xmath7 i.e. all higher moments can be obtained from the mean value and the variance .
this does not hold for non - gaussian functions .
the bispectrum , a rotationally invariant cubic form associated with the third moment , @xmath8 , is therefore a possible test of gaussianity .
it is given by @xmath9 where the @xmath10 is the wigner @xmath11 symbol .
in @xcite correlations between multipoles separated by @xmath12 , @xmath13 were investigated . in this letter
we complete this work by using the other possible separations , @xmath14 , which fit within the signal dominated range of @xmath15 probed by the cobe - dmr experiment .
selection rules require @xmath16 to be even , @xmath17 and @xmath18 . from these constraints
we see that a suitable chain of correlators is @xmath19 , with @xmath15 even .
we then follow precisely the same procedure documented in @xcite and @xcite , apart from considering the more general ratio @xmath20 where @xmath21 .
this quantity is dimensionless , and is invariant under rotations and parity .
the same data is examined ; the inverse noise - variance - weighted average maps of the 53a,53b,90a and 90b _ cobe_-dmr channels , with monopole and dipole removed , at resolution 6 , in galactic and ecliptic pixelization .
we use the extended galactic cut of @xcite , and @xcite to remove most of the emission from the plane of the galaxy . to estimate the @xmath22s we set the value of the pixels within the galactic cut to 0 and the monopole and dipole of the cut map to zero .
we then integrate the map multiplied with spherical harmonics to obtain the estimates of the @xmath23s and apply equations [ bispec ] and [ defi ] .
the observed @xmath22s are to be compared with their distributions @xmath24 as inferred from monte carlo simulations in which gaussian maps are subject to dmr noise and galactic cut . in simulating dmr noise
we take into account the full noise covariance matrix , as described in @xcite .
this includes correlations between pixels @xmath25 degrees apart .
the comparison between @xmath24 and the observed @xmath22 for various @xmath26 is displayed in figs .
[ fig1 ] - [ fig4 ] .
the observed values agree well with the probability distributions from the monte carlo distributions : implying that as far as @xmath27 bispectrum components are concerned the data supports gaussianity .
to quantify just how well the observations agree with gaussianity , we do a goodness of fit statistic on the data .
we calculate the @xmath28 value for the different @xmath15 separations , and compare with the respective distributions @xmath29 .
the rationale behind considering separate @xmath28 was explained in @xcite , and for the distributions @xmath30 found ( which are very close to gaussians ) it makes sense to use the standard @xmath28 : @xmath31 where @xmath32 is the number of different observables and @xmath33 and @xmath34 are the expectation value and the variance of the @xmath35th observable respectively .
this reduces to the expression proposed in @xcite in our case .
the @xmath29 are then calculated through further monte carlo simulations , so that possible correlations between the different @xmath15 may be taken into account .
the @xmath28 distributions and corresponding observed values are shown in fig [ fig5 ] . as we can see @xmath36 for all separations @xmath0 , with deviations well within the spread expected from @xmath29 .
this is to be contrasted with the results obtained for @xmath37 , that is bispectrum components measuring correlations between adjacent modes . for these @xmath38 , meaning that the observed @xmath39 do not display the scatter around zero expected from a gaussian distribution .
we now found that this distinctive non - gaussian signal does not extend to bispectrum components with @xmath27 .
the implication seems to be that the inter - mode correlations suggested by the findings of @xcite fall off quickly with the scale separation , something which conforms to theoretical prejudice . in a recent paper @xcite a variety of components of the bispectrum
was studied in the context of a semi - analytical texture model .
such a model is at the very least oversimplified , and probably has little to do with real - life textures . in any case
that work stresses the importance of analysing off - diagonal components , as we have done in this paper .
c + [ cols="^,^ " , ]
h.b . sandvik would like to thank the research council of norway for financial support .
this work was performed on cosmos , the origin 2000 supercomputer owned by the uk - ccc and supported by hefce and pparc .
99 banday , a.j .
_ et al _ _
_ , * 475 * , 393 ( 1997 ) .
banday , a.j . ,
zaroubi , s. , and grski , k.m .
, astro - ph/9908070 .
bennet , c.l .
_ et al _ _ ap . j. _ * 464 * , 1 ( 1996 ) .
r. barreiro et al , astro - ph/0004202 .
bromley and tegmark , 1999 , apj , 524 , l79 . c. r. contaldi , r. bean and j. magueijo , phys .
* b468 * ( 1999 ) 189 [ astro - ph/9910309 ] . c. r. contaldi , astro - ph/0005115 contaldi , hidmarsh and magueijo , phys.rev.lett .
82 ( 1999 ) 679 - 682 .
n. phillips and a. kogut , astro - ph/0010333 a. linde and v. mukhanov , phys .
* d56 * ( 1997 ) 535 [ astro - ph/9610219 ] . c. h. lineweaver , g. f. smoot , c. l. bennett , e. l. wright , l. tenorio , a. kogut , p. b. keegstra , g. hinshaw , a. j. banday , astro - ph/9403021 .
ferreira , p.g . ,
magueijo , j. and grski , k.m.g .
_ * 503 * l1(1998 ) magueijo , j. , ap.j.lett .
528 , l57-l60 ( 2000 ) j. magueijo , p. g. ferreira and k. m. gorski , astro - ph/9903051 .
pando , j. , valls - gabaud and fang , l.z . ,
_ phys.rev.lett _ * 81 * 4568 ( 1998 ) .
peebles , p.j .
_ * 483 * , l1 ( 1997 ) .
d. s. salopek , phys .
* d45 * ( 1992 ) 1139 . | issue primordial fluctuations gaussian in theories structure formation gaussian large - scale primordial fluctuations hallmark of inflationary scenarios vacuum quantum fluctuation gaussian statistics stretched by exponential expansion large scale fluctuations topological defect theories possible candidates for structure formation produce non - gaussian density fields non - gaussianity associated with non - minimal inflationary models isocurvature fluctuations recent papers shown non - gaussianity in cosmic microwave background temperature fluctuations measured by cobe - dmr experiment first detections due to experimental systematic @xcite second error of method third remains unassailable bispectrum analysis carried adjacent scales @xmath1 significant deviation from gaussianity found high confidence level question non - gaussianity extends to correlating scales with larger separation purpose of paper to answer question non - gaussian signal in @xcite extend to larger separations @xmath3 possibility of deviations from gaussianity in bispectrum should be zero for gaussian processexpand temperature fluctuations @xmath4 in spherical harmonic functions @xmath5 gaussian probability distribution function described by first moments higher moments obtained from mean value variance for non - gaussian functions bispectrum rotationally invariant cubic form third moment @xmath8 possible test of gaussianity given by @xmath9 @xmath10 wigner @xmath11 symbol correlations between multipoles by investigated separations @xmath14 signal range @xmath15 cobe - dmr experiment selection rules require @xmath16 even suitable chain correlators @xmath19 @xmath15 even follow procedure @xcite general ratio @xmath20 @xmath21 quantity dimensionless invariant under rotations parity same data examined inverse noise - variance - weighted average maps of 53a,53b cobe_-dmr channels monopole dipole removed at resolution 6 in galactic ecliptic pixelization use extended galactic cut of @xcite remove emission from plane galaxy estimate @xmath22s set value pixels galactic cut to 0 monopole dipole to zerointegrate map with spherical harmonics estimates @xmath23s apply equations bispec defi ] observed @xmath22s compared with distributions @xmath24 monte carlo simulations gaussian maps dmr noise galactic cut noise full noise covariance matrix described @xcite includes correlations between pixels @xmath25 degrees apart comparison between @xmath24 observed @xmath22 @xmath26 displayed in figs fig1 - fig4 observed values agree with probability distributions monte carlo distributions data supports gaussianity quantify goodness of fit statistic calculate @xmath28 value for @xmath15 separations compare with distributions @xmath29 rationale separate @xmath28 explained @xcite distributions @xmath30 close gaussians use standard @xmath28 : @xmath31 @xmath32 number observables expectation value variance of @xmath35th observable reduces expression @xcite @xmath29 calculated through monte carlo simulations correlations between @xmath15 @xmath28 distributions observed values shown in fig fig5 @xmath36 for all separations @xmath0 deviations within spread expected from @xmath29contrasted with results @xmath37 bispectrum components correlations between modes @xmath38 @xmath39 display scatter around zero gaussian distribution non - gaussian signal extend to bispectrum components with @xmath27 implication inter - mode correlations @xcite fall off with scale separation conforms to theoretical prejudice recent paper @xcite components bispectrum studied semi - texture model model oversimplified little to with real - life textures work stresses importance of analysing off - diagonal components sandvik research council of norway for financial support work performed on cosmos origin 2000 supercomputer owned by uk - ccc supported by hefce pparc banday , a.j . al 475 , 393 ( 1997 ) zaroubi grski astro - ph/9908070 bennet 464 *, 1 ( 1996 ) barreiro et al astro - ph/0004202 bromley tegmark , 1999 524 , l79 contaldi bean j. magueijo phys b468 * ( 1999 ) 189 [ astro - ph/9910309 ]contaldi astro - ph/0005115 hidmarsh magueijo.rev.lett 82 1999 ) 679 - 682 phillips kogut astro - ph/0010333 linde mukhanov phys d56 1997 ) 535 astro - ph/9610219 lineweaver smoot bennett wright tenorio kogut keegstra hinshaw banday astro - ph/9403021 ferreira magueijo grski 503 l1(1998 ) magueijo j ap.j.lett 528 l57-l60 ( 2000 ) magueijo ferreira gorski astro - ph/9903051 pando j valls - gabaud fang.rev.lett 81 4568 ( 1998 ) peebles p.j l1 1997 ) salopek phys d45 1992 ) 1139 | 0.313406 | 2,030 | 1,036 | , Saving $0.1 in GPT-4. |
20,842 | we present preliminary results for the curvature of the pseudocritical line and susceptibilities in @xmath0 flavor qcd .
the computations are carried out on lattice sizes of @xmath1 , at matching parameters of early work of the bielefeld group .
emphasis is placed on the control of systematic errors , by cross - validating results obtained by use of the taylor expansion and measurements at imaginary chemical potential . to this end
, we generalize the magnetic equation of state to the analysis of the number density , and we extend it to imaginary values of the chemical potential . | most likely , a large fraction of the participants in this meeting are familiar with the challenges faced by lattice qcd at non - zero baryon density@xcite : the region of quark chemical potential @xmath2 exceeding @xmath3 remains prohibitively difficult , due to a severe sign problem@xcite . for smaller chemical potential a variety of methods
have been proposed to circumvent these difficulties , which have been quite successful in many respects . however , even in the easy region @xmath4 there are challenges to face : for instance , it has been noted that most current estimates of the qcd endpoint lie dangerously close to the forbidden region @xmath5 ; the qualitative agreement among different results on the slope of the critical line still needs to be corroborated by some convincing continuum estimates obtained with different methods . and
, there is a growing interest in the values of higher order susceptibilities , particularly at larger values of the chemical potential , while the current results are limited to the lowest order ones @xcite .
so there is room for improvement , which can perhaps be achieved by mixing and matching different approaches .
one way to handle the physics of small chemical potential relies on taylor expansion@xcite .
this approach is well understood mathematically , however it only works inside its convergency circle , whose radius is unknown .
moreover even inside the circle it is hard to judge when convergence has been achieved . alternatively , analytic continuation from imaginary chemical potential where conventional simulations are possible to real chemical potential can in principle be extended to the entire analytic domain , but the danger here is a possible lack of control over the continuation itself : for instance , terms which are subleading for imaginary chemical potential might well become leading for real chemical potential .
the strengths of the two methods are somewhat complementary , and their weaknesses different : this suggests to develop a hybrid method which combines the two , and this note reports on our ongoing effort in this direction@xcite .
-12 truecm -2.0 truecm ( right ) . note the ( approximate ) universal behaviour of the condensate .
, title="fig:"]-2.0truecm ( right ) .
note the ( approximate ) universal behaviour of the condensate .
, title="fig : " ] -1.0 truecm
our study for staggered 2 + 1 flavors on lattices of size @xmath1 clearly is exploratory .
however , the quark mass values are close to the ones realized in nature .
the goldstone pion mass is tuned to about @xmath6 mev and the kaon acquires its physical mass .
this corresponds to a ( degenerate ) light to strange quark mass ratio of @xmath7 .
the action utilized is the p4fat3 action for which taylor coefficients computed at @xmath8 are available in our range of temperatures and at our quark masses@xcite .
the computations were carried out at three temperatures above the critical temperature for two massless flavors for the considered lattice spacings : @xmath9 mev . in fig.1 , left we sketch the simulations trajectories , alongside the critical line in the chiral limit , in the @xmath10 plane .
the sketch evidences the @xmath11 periodicity as well as the symmetry under reflection @xmath12 ; the actual simulations were performed only at positive three - flavor degenerate @xmath11 , and within the first half period .
the blue points mark the endpoint of the roberge weiss phase transition .
the light quark chiral condensate , plotted in fig.1 , right , is showing a fairly rapid rise when @xmath11 is increased , signaling the pseudocritical point for the transition to the broken phase ( marked @xmath13 in the diagram ) .
we plot the results against a rescaled chemical potential to confirm the feasibility of a scaling study based on the magnetic equation of state .
consider now the magnetic equation of state@xmath14 where @xmath15 and @xmath16 are identified with the light chiral condensate and the light bare mass : @xmath17 , @xmath18 respectively , @xmath19 is a scaling function which only depends on the universality class of the transition , and @xmath20 defines the critical line .
our light mass does not change : if scaling ( i.e. eq .
1 ) holds true the chiral condensate results plotted versus @xmath21 , where @xmath22 is the critical chemical potential for each temperature @xmath23 , should fall on a single curve . in fig.1
we see that this is indeed happening , demonstrating the universal behaviour of the chiral condensate at least in the range of parameters we are exploring .
@xmath22 is found by searching for the optimal match between results for different temperatures , and compares favorably with alternative estimates . for small chemical potential @xmath24
can be explicitely written as @xmath25 where @xmath26 is the critical temperature in the chiral limit . the standard condition @xmath27 fixes the normalization @xmath28 and @xmath29 which have been determined in ref . @xcite . as a working hypothesis
we assume that @xmath28 and @xmath29 do not depend on the temperature , and we check a posteriori that this indeed is the case .
@xmath30 the curvature of the critical line
can then be computed from the equation of state , using the explicit forms available for the two plausible universality classes @xmath31 and @xmath32 , and the results contrasted with those obtained in ref.@xcite . -10 truecm -0.8truecm -3.1 truecm -3.1 truecm -1truecm .the results for the curvature of the critical line @xmath30 ( eq.(2 ) ) from the fits to the magnetic eos for the chiral condensate ( @xmath33 ) , and from the fits to the number density ( @xmath34 ) . [ cols="^,^,^,^,^,^,^",options="header " , ] from the eos for the order parameter one can derive analogous relations for the magnetic susceptibility , as well as for the number density @xmath35 and its susceptibility@xcite , leading to alternative estimates of @xmath30 . to fix the notation ,
let us write the taylor expansion of the pressure in terms of the quark chemical potentials around a generic @xmath36 for our @xmath37 system : @xmath38 with the abbreviation @xmath39 and the coefficients @xmath40 evaluated at @xmath41 .
we will in particular discuss @xmath42 and @xmath43 , which are related to the particle number and its susceptibility .
in fig.2 and fig.3 we show the results of the eos fits and in table 1 we record the @xmath30 values returned by the fits for the chiral condensate and the quark number ( the coefficient @xmath44 ) .
while the fit results , at least for o(2 ) , do not seem to be inconsistent , it is clear that at values for @xmath11 close to the roberge - weiss limit other scaling studies need to be performed , work for the future .
-7.5 truecm -0.8truecm and @xmath45 ( related to quark number and its susceptibility respectively , see text ) superimposed to the data points.,title="fig : " ] -3.1 truecm and @xmath45 ( related to quark number and its susceptibility respectively , see text ) superimposed to the data points.,title="fig : " ] -3.1 truecm and @xmath45 ( related to quark number and its susceptibility respectively , see text ) superimposed to the data points.,title="fig : " ] -1truecm how far in @xmath46 we can trust the taylor expansion ?
a direct comparison of the number density of e.g. the u quark given by the partial sums of its taylor series : @xmath47 with results obtained directly at imaginary @xmath46 might give the answer .
we show the results for increasing temperatures in fig.4 : for @xmath48 mev a good convergence is achieved indicating a small contribution from the @xmath49 order terms and superior ones . at the highest temperature instead
the results suggest that a @xmath50 order expansion is not capable to describe the data at @xmath11 @xmath51 .
the simultaneous consideration of taylor expansion and imaginary @xmath46 results naturally suggests a strategy to extract higher order terms of the expansions or , equivalently , higher order susceptibilities
withouth having to implement the cumbersome higher order derivatives . at the same time
one can place lower bounds on the radius of convergence simply by assessing by eye the extent of the region where actual convergence is achieved .
-7.5 truecm -0.8truecm -3.1 truecm -3.1 truecm -1truecm
we have presented an overview of results obtained by performing simulations at zero and non zero imaginary chemical potential . we have noted that an analysis based on the magnetic equation of state for the chiral condensate , the number density and their susceptibilities at imaginary chemical potential gives information on the universality class of the transition for two massless flavors , and on the slope of the chiral line .
we have argued that the simultaneous consideration of results from the taylor expansion at @xmath52 and at imaginary @xmath46 can place a lower bound on the radius of convergence at @xmath52 , and facilitate the computation of higher order coefficients .
99 g. aarts , pos lattice * 2012 * 017 ( 2012 ) .
k. splittorff , pos lattice * 2006 * , 023 ( 2006 ) .
m. p. lombardo , pos lattice * 2012 * , 016 ( 2012 ) .
m. cheng et al . , phys .
d79 ( 2009 ) 074505 .
f. meyer , master thesis , bielefeld 2012 .
s. ejiri et al . , phys .
d80 ( 2009 ) 094505 .
o. kaczmarek et al . , phys .
d83 ( 2011 ) 014504 .
j. engels and f. karsch , phys .
d85 ( 2012 ) 094506 . | large participants meeting familiar with challenges by lattice qcd at non - zero baryon density@xcite region of quark chemical potential @xmath2 exceeding @xmath3 difficult due to severe sign problem for smaller chemical potential methods proposed to circumvent difficulties successful even in easy region @xmath4 challenges current estimates of qcd endpoint close to forbidden region @xmath5 qualitative agreement among results needs by continuum estimates growing interest in higher order susceptibilities larger values chemical potential current results limited to lowest order ones room for improvement by mixing and matching approaches small chemical potential taylor expansion@xcite approach understood works inside convergency circle radius unknown hard to judge when convergence achieved analytic continuation from imaginary chemical potential to real chemical potential can be extended to entire analytic domain danger is possible lack of control over continuation terms subleading for imaginary chemical potential might become leading for real chemical potential strengths of two methods complementary weaknesses different suggests hybrid method note reports on ongoing effort direction -12 truecm -2.0 truecm note universal behaviour of condensate .title="fig-2.0truecm right universal behaviour condensate -1.0 truecm study for staggered 2 + 1 flavors on lattices @xmath1 exploratory quark mass values close to nature goldstone pion mass @xmath6 mev kaon acquires physical mass light to strange quark mass ratio @xmath7 action p4fat3 action taylor coefficients computed @xmath8 temperatures quark masses computations at three temperatures above critical temperature two massless flavors lattice spacings @xmath9 mev.1 simulations trajectories critical line chiral limit @xmath10 plane evidences @xmath11 periodicity symmetry under reflection @xmath12 simulations performed at positive three - flavor degenerate @xmath11 first half period blue points endpoint roberge weiss phase transition light quark chiral condensate rapid rise when @xmath11 increased pseudocritical point transition to broken phase @xmath13 plot results against rescaled chemical potential scaling study magnetic equationmagnetic equation state@xmath14 @xmath15 identified light chiral condensate light bare mass @xmath17 scaling function depends on universality class transition @xmath20 defines critical line light mass change if scaling eq . 1 chiral condensate results versus @xmath21 @xmath22 critical chemical potential each temperature @xmath23 fall single curve.1 universal behaviour chiral condensate range parameters @xmath22 found optimal match results temperatures compares with alternative estimates small chemical potential @xmath24 written as @xmath25 @xmath26 critical temperature chiral limit standard condition @xmath27 fixes normalization @xmath28 @xmath29 determined in ref. @xcite @xmath28 @xmath29 depend on temperature check posteriori @xmath30 curvature critical line computed from equation of state explicit forms two universality classes @xmath31 @xmath32 results contrasted with ref.@xcite -10 -0.8truecm -3.1 results curvature critical line @xmath30 eq.(2 ) ) from fits to magnetic eos chiral condensate @xmath33 fits number density @xmath34 )cols="^,^,^",options="header " from eos order parameter derive analogous relations magnetic susceptibility number density @xmath35 susceptibility@xcite alternative estimates @xmath30 fix notation write taylor expansion pressure quark chemical potentials around generic @xmath36 @xmath37 system @xmath38 abbreviation @xmath39 coefficients @xmath40 @xmath41 discuss @xmath42 @xmath43 related to particle number susceptibility fig.2.3 show results eos fits table 1 @xmath30 values returned fits chiral condensate quark number coefficient @xmath44 ) fit results o(2 ) not inconsistent values @xmath11 close to roberge - weiss limit other scaling studies need future -7.5 truecm -0.8truecm @xmath45 related quark number susceptibility text superimposed data points.,title="fig : -3.1 truecm @xmath45 related quark number susceptibility superimposed data points -3.1 truecm @xmath45 superimposed data points -1truecm @xmath46 trust taylor expansion ? direct comparison number densityu quark partial sums taylor series @xmath47 results imaginary @xmath46 answer results increasing temperatures fig.4 @xmath48 good convergence small contribution @xmath49 terms superior highest temperature results suggest @xmath50 expansion describe data @xmath11 @xmath51 simultaneous consideration taylor expansion @xmath46 results suggests strategy extract higher order terms susceptibilities derivatives lower bounds radius convergence assessing extent region convergence -7.5 -0.8truecm -3.1 overview results simulations zero non zero imaginary chemical potential analysis magnetic equation state chiral condensate number density susceptibilities at imaginary chemical potential gives information universality class transition massless flavors slope chiral line simultaneous consideration results taylor expansion at @xmath52 imaginary @xmath46 lower bound on radius convergence at @xmath52 facilitate computation higher order coefficients. aarts , pos lattice * 2012 * 017 ( 2012 splittorff , pos lattice 2006 * 023 lombardo pos lattice 2012 * 016 2012 cheng et al . phys . d79 ( 2009 ) 074505meyer master thesis bielefeld 2012 ejiri phys d80 2009 ) 094505 kaczmarek phys d83 2011 014504 engels karsch d85 2012 094506 | 0.349353 | 2,246 | 1,171 | , Saving $0.1 in GPT-4. |
20,845 | up to present date , no circumbinary planet around contact binaries were discovered neither by transit method nor by the minima times variation , although they are known having third component stars around .
we thus ask : where are the circumbinary planets of contact binaries ? by considering the physical and geometrical parameters we simulated the light curves of contact binaries with possible transiting circumbinary jovian planets .
it seems either the circumbinary jovian planets are not formed around contact binaries , probably due to dynamical effects of the binary and third component stars , or they are present but the discovery of such planets were not possible so far due to larger distortions then expected in the photometric data and in the minima times . | the disks around stars provide the reservoirs of material out of which planets are formed .
the disks are formed by lost or accreted material around single stars , but they form at different stages of the evolution if the star is a member of a close binary system : ( a ) during formation of binary , ( b ) during roche lobe overflow in semi detached binaries , ( c ) during high rate of mass loss stage , ( d ) during mass loss from the second lagrangian point in overcontact binaries , ( e ) during merging process of contact binaries , ( f ) during post common envelope stage , and ( g ) during evaporation of the secondary in x ray binaries .
it is thus expected first , second , third , ... generation planets around close binary systems , or around the components of wider binary systems .
planets in s - type orbits around a component of 44 binaries were found . for these systems
the semi - major axis a of the orbit is generally greater than 20 au ( see e.g. * ? ? ?
circumbinary planets around 11 post common envelope binaries were also discovered by ground based eclipse timing ( et ) observations ( see e.g. * ? ? ? * ) .
circumbinary planets in p - type orbits around relatively short period 7 unevolved eclipsing binaries ( 0.08 < @xmath0 < 0.23 au ) were discovered by kepler s transit observations ( see e.g. * ? ? ?
all these circumbinary planets ( except two outer planets of kepler 47 ) were found reside close to the dynamical stability limit boundary ( in average _ _ p__@xmath1 / _ _ p__@xmath2 = 6.88 ) which is very close to inner edge of the circumbinary disks where _ _
p__@xmath1 / _ _ p__@xmath2 = 4.5 . due to the difficulty of forming planets in such close orbits
, it is believed that they have formed further out in the disk and migrated to their present locations ( see e.g. * ? ? ?
altough the mass loss may form disks around contact or overcontact binaries , there is no claim of any circumbinary planet around these systems .
either there are no circumbinary planets around contact binaries , or the observations are not sufficient to detect any such planets , if they exist . in the present contribution we first simulate the light curves of contact binaries with possible transiting circumbinary jovian planets by considering the physical and geometrical parameters of the circumbinary planets .
we than reconsider the ground based et observations to see if they are capable to detect lite effects of the circumbinary planets around contact binaries .
if exist , the circumbinary planets around contact binaries should have similar properties with those of either unevolved eclipsing binaries ( ueb ) or those of post common envelope binaries ( pceb ) .
we thus expect orbital periods _
_ p__@xmath1 of the circumbinary planets around contact binaries would be around 6.88 @xmath3 _
_ p__@xmath2 , i.e. not more than a few days in the case of ueb , or not less then ten years in the case of pceb . for the simulation we considered a jovian circumbinary planet in one year orbit around the contact binary w uma system itself .
transit geometry is shown in fig.1 , where the semi - major axis _ _ a__@xmath2 and _ _ a__@xmath1 of the binary and planet becomes 2.45 and 260.21 r@xmath4 .
the duration of a planetary transit in front of a single star was given by @xcite which was used by @xcite to derive the following formula for the crossing time _
_ t__@xmath5 of a circumbinary planet in front of a binary system : @xmath6 where @xcite introduced the quantity _ _ r__@xmath7 representing the maximum extent of the binary s orbit projected on to the sky .
the other symbols in equation ( 1 ) have their usual meaning for planetary orbit . for contact binaries
, it can be estimated that ( _ _ r__@xmath7 @xmath8 _ _ a__@xmath2 + _ _ r__@xmath9 + _ _ r__@xmath10 ) where the summation of the mean radii ( _ _ r__@xmath9 + _ _ r__@xmath10 ) of the component stars for marginal contact binaries is about 0.75__a__@xmath2 which may increase up to 0.8__a__@xmath2 for overcontact binaries .
thus , for the circular coplanar planetary orbits with 90inclination , it can be found that @xmath11 such transit times are usually much longer than the orbital periods of contact binaries .
implications are shown in fig .
2 as a simulation on the contact binary system w uma itself , where if we substitute _ _ a__@xmath2 = 2.45 r@xmath4 , _ _
a__@xmath1 = 260.21 r@xmath4 and _ _
p__@xmath1 = 1 yr , then we obtain _ _
t__@xmath5 = 0.985 d. it is seen that during the transit of such circumbinary jovian planet , the contact binary inside revolve about three times ( 0.985 / 0.3336 @xmath8 3 ) . in the case of coplanar orbits of the planet and binary
, the shape of the transit light curve would be like 3 times @xmath12 s side by side where each @xmath12 representing the transit light curve in one binary orbital period . in fig .
2 , the light variation of the transit ( 2a ) superimposed on the contact binary light curve ( 2b ) as shown in ( 2c ) would be really difficult to disentangle especially in the presence of magnetic activity disturbances ( see fig .
this may be the reason that no transit observations of any circumbinary planet around contact binaries were claimed so far .
however , in the light of fig .
2c , some small changes in the ground - based photometric observations of contact binaries may well be related with transiting circumbinary planets ( see e.g. fig .
it is well known that the cyclic variations in the _ o - c _ diagrams formed by et are the signatures of light time effect ( lite ) of unseen objects around binary systems ( see e.g. * ? ? ?
the et precision ( @xmath13__t _ _ ) depends not only the precision of a single observation but also on the shape , duration and depth of the eclipse minima ( see e.g. * ? ? ?
the detectable timing amplitude of a cyclic variation ( @xmath13__t _ _ ) , on the other hand , is a function of the mass , orbital period and orbital inclination of the circumbinary third body .
the ( @xmath13__t _ _ ) should be at least as good as ( @xmath13__t _ _ ) .
it was estimated that @xmath13__t _ _
@xmath13__t _ _ is usually between 10 - 20 seconds for contact binaries .
the orbital periods
_ _ p__@xmath15 of the circumbinary third bodies ( with masses of 1 and 10 jupiter masses ) were estimated for the total systemic mass of 1 and 2 solar mass , and for @xmath13__t _ _ = 4 , 10 , 20 , and 40 seconds . resulting
_ _ p__@xmath15 values ( in year ) were listed in table 1 .
ccccc m@xmath16 & @xmath13__t _ _ = 4 sec & @xmath13__t _ _ = 10 sec & @xmath13__t _ _ = 20 sec & @xmath13__t _ _ = 40 sec + 10 & 0.45 & 1&3&9 + 1&14.4&34&97&273 +
the light changes of contact binaries with a possible transiting circumbinary jovian planet was simulated .
it is concluded that if the transiting circumbinary planets are present around contact binaries , the relatively large intrinsic light variability , long exposure times and the complexity of the combined low amplitude light variation formed by the transits of the circumbinary planets do not allow yet the disentangling of the transit light curves .
on the other hand , the et precision decreases by the intrinsic variability caused by the enhanced magnetic activity of late type g , k , and m- type binary components , and thus the scatter in the _ o - c _ diagrams increases . the major scatter in the light curves and the _ o - c _ diagrams are caused by the starspot activity .
the time shift in et ( which can be as large as 30 minutes ) caused by starspots is proportional to the photometric wave amplitude , while inversely proportional to the depth of the eclipse minima .
the starspot effect in the _ o - c _ diagram can be minimised by using the average of the minimum i and minimum ii times , or before forming the _ o - c _ diagram , the time shifts in et s can be corrected by spot modelling . due to relatively large scatter in the _ o - c _ diagrams of contact binaries , the detectable lite amplitudes ( @xmath13__t _ _ ) would be relatively large and thus , as table 1 implies , in practice , only the very high mass planets about @xmath17 = 10 m@xmath16 are detectable around contact binaries by using et . | disks around stars provide reservoirs material planets formed disks formed by lost material around stars form at different stages evolution if star member close binary system formation roche lobe overflow high mass loss mass loss overcontact merging process post common envelope stage evaporation secondary in x ray binaries expected first second third generation planets around close binary systems or components wider binary systems planets in s - type orbits around 44 binaries found semi - major axis a orbit greater than 20 au circumbinary planets around 11 post common envelope binaries discovered by eclipse timing observations circumbinary planets in p - type orbits around 7 unevolved eclipsing binaries ( 0.08 < @xmath0 < 0.23 au discovered by kepler s transit observations planets two close to dynamical stability limit boundary _ p__@xmath1 / p__@xmath2 = 6.88 ) close to inner edge circumbinary disks where /@xmath2 = 4.5 due to difficulty forming planets in close orbits believed they formed further in disk migrated to present locationsmass loss may form disks around contact binaries no claim of circumbinary planet around systems no planets or observations not sufficient detect planets simulate light curves of contact binaries with possible transiting circumbinary jovian planets physical geometrical parameters reconsider ground observations to detect effects of circumbinary planets around if exist planets should have similar properties with unevolved eclipsing binaries ueb or post common envelope binaries pceb expect orbital periods of planets around around 6.88 @xmath3@xmath2 not more few days ueb or ten years pceb simulation considered jovian circumbinary planet in one year orbit around contact binary system transit geometry shown in fig.1 semi - major axis and of binary planet becomes 2.45 and 260.21 r@xmath4 duration of planetary transit in front of single star given by @xcite formula for crossing time of circumbinary planet in front of binary system maximum extent of binary s orbit other symbols in equation ( 1 have usual meaning for planetary orbitcontact binaries estimated r__@xmath7 @xmath8 a__@xmath2 r__@xmath9 +@xmath10 ) mean radii r__@xmath9 +@xmath10 ) component stars marginal binaries 0.75__a__@xmath2 increase 0.8__a__@xmath2 overcontact binaries circular coplanar planetary orbits 90inclination @xmath11 transit times longer than orbital periods contact binaries implications . 2 contact binary system substitute a__@xmath2 = 2.45 r@xmath4 a__@xmath1 = 260.21 r@xmath4 p__@xmath1 = 1 yr obtain t__@xmath5 = 0.985 d transit circumbinary jovian planet contact binary three times ( 0.985 / 0.3336 @xmath8 3 ) coplanar orbits transit light curve 3 times @xmath12 s side by side transit light curve orbital period fig . 2 light variation transit ( 2a ) contact binary light curve ( 2b ) difficult to disentangle magnetic activity disturbances no transit observations circumbinary planet around contact binaries claimed small changes ground photometric observations binaries related transiting circumbinary planetsfig . cyclic variations in o - c diagrams are signatures of light time effect of unseen objects around binary systems precision ( @xmath13__t _ depends observation on shape duration depth of eclipse minima detectable timing amplitude of cyclic variation @xmath13__t function of mass orbital period inclination of circumbinary third body @xmath13__t _ _ should be good as estimated @xmath13__t between 10 - 20 seconds for contact binaries orbital periods of circumbinary third bodies 1 10 jupiter masses estimated for total systemic mass of 1 and 2 solar mass @xmath13__t _ _ = 4 10 20 40 seconds resulting values in year listed in table 1 @xmath16 @xmath13__t = 4 sec = 10 sec 20 sec 40 sec 1&3&9&14.4&34&97&273 light changes of contact binaries with transiting circumbinary jovian planet simulatedconcluded if transiting circumbinary planets present around contact binaries large intrinsic light variability long exposure times complexity low amplitude light variation allow disentangling transit light curves et precision decreases by variability enhanced magnetic activity of late type g k m- type binary components scatter in o - c diagrams increases major scatter in caused by starspot activity time shift in et 30 minutes caused by starspots proportional to photometric wave amplitude inversely proportional to depth eclipse minima starspot effect minimised by using average of minimum i and ii times time shifts corrected by spot modelling large scatter in diagrams contact detectable lite amplitudes ( @xmath13__t _ _ large only high mass planets about @xmath17 = 10 m@xmath16 detectable around contact binaries using et | 0.313549 | 2,081 | 1,053 | , Saving $0.1 in GPT-4. |
20,849 | the goal of this document is to illustrate that teaching the concepts of magnitudes is a needless complication in introductory astronomy courses , and that use of monochromatic luminosities , rather than arbitrarily defined magnitudes , leads to a large gain in transparency .
this illustration is done through three examples : the hertzsprung - russell diagram , the cosmic distance ladder , and interstellar reddening .
i provide conversion equations from the magnitude - based to the luminosity - based system ; a brief discussion ; and a reference to sample lecture notes .
i suggest that we , astronomers in the 21st century , * abolish magnitudes and instead use ( apparent ) monochromatic luminosities * in non - specialist teaching . given the large gain in transparency
i further propose that we seriously consider using ( apparent ) monochromatic luminosities also in research papers , bringing optical astronomy in line with astronomy at other wavelengths .
i would appreciate your opinion , comments and suggestions ( see sect.5 ) . | as a function of colour @xmath0 of 20305 stars with an hipparcos distance more accurate than 10% . the corresponding @xmath1 and @xmath2 are shown at the right hand side and top .
[ f : newhr ] _ ] in analogy with the total or bolometric luminosity , we define the _ monochromatic luminosity _ @xmath3 as @xmath4 here @xmath5 and @xmath6 are fluxes per unit wavelength , i.e. monochromatic fluxes , at the stellar surface and at earth , respectively , @xmath7 and @xmath8 are the radius of and distance to the star . for the sun , the flux at 550 nm in a band of 1 nm is @xmath9 , and thus with @xmath10au , @xmath11 .
( in fact , this flux is averaged over the johnson v filter , and one may prefer to denote it @xmath12 . ) the hipparcos catalogue provides distances @xmath8 and monochromatic fluxes @xmath13 which with eq.[e : monlumin ] give the monochromatic luminosity @xmath14 and colour @xmath15 .
( for the moment we assume that interstellar absorption is negligible . )
we combine these in a colour - luminosity diagram figure[f : newhr ] .
this figure tells us immediately that a b main - sequence star emits about 100 times more photons at 550 nm than the sun , and a k dwarf about 10 times fewer .
it also tells us directly that an early g star emits about as much flux at 550 nm as at 440 nm .
such information is also present in the diagram labeled with @xmath1 and @xmath2 , but very much hidden . ) and _ apparent _ luminosities of stars in the pleiades ( @xmath16 ) , for an assumed distance @xmath17pc .
[ f : pleiades ] _ ]
for distances beyond the reach of parallax measurements , several methods are available that are based on monochromatic luminosities . for these
we define the _ apparent monochromatic luminosity _
@xmath18 , based on an _ assumed distance _
@xmath19 , as : @xmath20 we allow any choice for @xmath19 , which means that _ the apparent luminosity must be given in tandem with an assumed distance_. the logarithms of the real and apparent monochromatic luminosities differ , with eqs.[e : monlumin ] and [ e : amonlumin ] , by : @xmath21 when we can determine this shift , the distance follows as @xmath22 as an example i show in figure[f : pleiades ] the apparent luminosity at 550 nm for stars in the pleiades for an assumed distance @xmath17pc , together with the lower envelope of the main - sequence for nearby stars , from figure[f : newhr ] .
a vertical shift of @xmath23 will put the main sequence of the pleiades just above the lower envelope for nearby stars .
the distance of the pleiades follows with eq.[e : dism ] as about 160pc . for assumed distance @xmath17mpc for three cepheids in ic4182 .
the periods of the cepheids in ic4182 are for decreasing luminosity : 42d , 9.24d and 4.26d .
[ f : cepheid ] _ ]
figure[f : cepheid ] shows the period - luminosity relation for cepheids in the small magellanic cloud ( left frame ; the luminosity shown for each cepheid is the average of the maximum and minimum value of @xmath24 persei by main - sequence fitting , and the distance to the smc by demanding that the period - luminosity relation passes through the values for the cepheids in h and @xmath25 persei ] ) and lightcurves of three cepheids in the galaxy ic4182 ( right frame ; apparent luminosities for an assumed distance @xmath17mpc ) .
the brightest cepheid in ic4182 shown has a period of 42d and an average apparent luminosity @xmath26 . from the left frame of figure[f :
cepheid ] we see that the correct luminosity at 42d is @xmath27 , i.e. @xmath28 . with eq.[e :
dism ] a distance of about 3.3mpc follows for ic4182 .
mpc , the lightcurve for coma is a composite of several supernovae ) .
[ f : sn1a ] _ ] in 1937 a supernova of type ia was observed in ic4182 . with the distance of this galaxy derived from the cepheids in it
, we can compute the monochromatic luminosities of this supernova .
the resulting lightcurve is shown in figure[f : sn1a ] . in figure[f : sn1a ] i also show the lightcurve for a supernova ia in the virgo cluster , and a composite lightcurve of several supernovae ia in the coma cluster , computing apparent luminosities for an assumed distance @xmath17mpc .
we see from the maxima ( at @xmath29d ) that @xmath30 for virgo and @xmath31 for coma , giving distances with eq.[e : dism ] of about 18mpc and 100mpc , respectively .
absorption and scattering of radiation by interstellar matter is characterized by the optical depth @xmath32 , where the subscript indicates the wavelength dependence .
the flux @xmath33 corrected for the absorption is given by @xmath34 _
it is this corrected flux which should be entered in eqs._[e : monlumin],[e : amonlumin ] , when @xmath35 . because of the wavelength dependence of the absorption , the colours of the star change . with eq.[e : absflux ] we can relate the change in colour to the optical depths at the different wavelengths . for the visual - to - blue flux ratio ,
we find @xmath36 generalizing this equation to arbitrary @xmath37 , we see that by taking the ratio of the intrinsic colour @xmath38 and the observed colour @xmath39 we can derive the ratio @xmath40 .
figure[newred ] shows how this works . from the line strengths we can derive the spectral types of the stars as o5 and a0v , even if the spectrum is affected by interstellar absorption
the dependence of @xmath32 on wavelength is found by dividing the observed , absorbed spectrum by the correct , standard spectrum .
the result is shown in figure[newred ] ( right ) . ) .
right : interstellar absorption as a function of wavelength , normalized on the absorption at 550 nm .
the overall smooth decline towards longer wavelength is due to absorption by gas , the extra opacity near 220 nm is due to dust .
[ newred ] _ ]
if we know the intrinsic colour of the star , we can derive the interstellar absorption even if we do not know the distance .
for example , from the data used to make figure[newred ] we have @xmath41 .
hence @xmath42 where the right hand side is found from eq.[e : redcol ] . .
[ newcoldis ] _ ] if we have measurements at three wavelengths , i.e. two independent colours , we can also determine the reddening , as illustrated in figure[newcoldis ] .
the left frame shows the main sequence colours in the absence of reddening , the right frame shows a reddened star that due to reddening no longer lies on this sequence .
corrections to the colours have been applied for assumed values of @xmath43 in steps of 0.15 .
it is seen that a correction for @xmath44 puts the star on the main - sequence colour - colour relation , close to a g5v star . with the intrinsic luminosity @xmath45 ,
the observed flux @xmath46 ws@xmath47nm@xmath47 , and the determined reddening @xmath44 the distance of the star can be determined .
the figure also shows that the solution is not unique : an early f star reddened by @xmath48 , or an even more reddened b star , is also possible . often the use of other colours allow selection of a unique solution .
absolute visual magnitudes to luminosity at 550 nm : @xmath49= 0.4\left[4.81-m_v\right ] \label{e : mvtolv}\ ] ] colour to flux ratio ( for @xmath50 ) : @xmath51 \\ \log{f_{806}\over f_{550 } } & = & \log{f_{806}\over f_{550 } } = 0.4\left[v - i-1.35\right ] \\ \log{f_{440}\over f_{365 } } & = & \log{f_{440}\over f_{365 } } = 0.4\left[u - b-0.54\right ] \label{e : fvofb}\end{aligned}\ ] ] the constants follow from the definitions of zero magnitude for @xmath52 , @xmath53 , @xmath54 , and @xmath55 .
apparent visual magnitudes to monochromatic luminosity : @xmath56 \label{e : vdtolv}\ ] ] absorption in magnitudes to optical depth : @xmath57 hence colour excess to difference in optical depth : @xmath58
* what is wrong with magnitudes ?
* the definition of the absolute magnitude as used in the original form of the colour - magnitude diagram @xmath59 has no less than four arbitrary elements , which we have to explain to students : 1 . the minus - sign , which causes high magnitudes to correspond to low fluxes 2 .
the factor 2.5 3 . the flux at zero apparent magnitude @xmath60 , as expressed in @xmath61 ( which has to be defined for each filter or @xmath37 in each filter system ! ) 4 .
the distance of 10pc monochromatic luminosity has no arbitrariness .
i have no quarrel with claudius ptolemaios for his assigning magnitudes between 1 and 7 to the stars that are visible to the naked eye , but i somehow think that he will forgive us if we switch to a more convenient notation .
* apparent luminosities . *
the advantage of leaving the assumed distance @xmath19 free , is that a suitable value can be chosen for each application . when determining the distance to the coma cluster a reference distance of 10pc is less suitable than a reference distance measured in mpc .
the disadvantage is that each apparent luminosity must be accompanied explicitly by a mention of the assumed distance @xmath19 .
this disadvantage can be removed if we agree to limit choices of @xmath19 to integer powers of 10 .
the power used can then be added to the index : @xmath62 would denote the apparent luminosity for @xmath17mpc , and @xmath63 the apparent luminosity for @xmath64pc .
the numbers 1 - 9 would suffice for any practical purpose . in practice a flux often represents a weighted average over a filter .
one can then replace the wavelength in the index with the filter name . thus @xmath65 is the apparent luminosity derived from the average flux in the @xmath52 filter , for @xmath64kpc .
* sample lecture notes . * to test the viability of using monochromatic luminosities , i have re - written sections of my lecture notes for a course on _ life of the stars : from dust to black holes _ that i teach to first year physics & astronomy students in utrecht .
the course is in dutch and uses si units .
english versions of the re - written sections can be found on my website : it may be argued that an advantage of the magnitude system is that it makes explicit that any flux measurement is based on flux ratios with calibrated sources .
it appears to me that this should be obvious anyway : remember that radio and x/@xmath67 astronomers do very well without magnitudes .
if a miracle happens , i.e. if reason prevails , and enough astronomers agree that we should stop using magnitudes , we may wish to avoid a proliferation of different notations and choices by asking the appropriate committees of the _ international astronomical union _ to define the method and notation of preference .
: ( 1 ) perryman et al
. 1995 a&a 304 , 69 .
( 2 ) johnson et al .
1952 apj 117 , 313 .
( 3 ) arp 1960 aj 65 , 404 ; saha et al .
1994 apj 425 , 14 .
( 4 ) baade & zwicky 1938 , apj 88 , 411 ( 1937c ) ; kimeridze & tsvetkov 1987 soviet astr . 25 , 513 ; barbon et al . 1989 a&a 220 , 83 ( 1984a in ngc4419 ) ; zwicky 1961 pasp 73 , 185 ( 1961d ) , zwicky & barbon 1967 aj 72 , 1366 ( 1962a ) , barbon 1978 aj 83 , 13 ( 1963f , m ) , kohoutek & kowal 1978 pasp 90 , 565 ( 1975f ) ( 5 ) gunn & stryker 1983 apjs 52 , 121 , seaton 1979 mnras 187 , 75p ( 6 ) bessell et al
. 1998 a&a 333 , 231 | function of colour @xmath0 of 20305 stars hipparcos distance accurate than 10% @xmath1 @xmath2 shown right side top luminosity monochromatic luminosity @xmath3 @xmath4 @xmath5 @xmath6 fluxes per unit wavelength monochromatic at stellar surface earth @xmath7 @xmath8 radius distance to star sun flux at 550 nm is @xmath9 @xmath10au @xmath11 flux averaged over johnson v filter @xmath12 hipparcos catalogue provides distances @xmath8 monochromatic fluxes @xmath13 monochromatic luminosity @xmath14 colour @xmath15 interstellar absorption negligible combine in colour - luminosity diagram figure a b main - sequence star emits 100 times more photons at 550 nm sun k dwarf 10 times fewer early g star emits as much flux at 550 nm as 440 nm information present in diagram labeled @xmath1 @xmath2 apparent luminosities of stars in pleiades @xmath16 ) assumed distance @xmath17pc distances beyond parallax measurements methods based on monochromatic luminositiesdefine apparent monochromatic luminosity @xmath18 based on assumed distance @xmath19 @xmath20 choice @xmath19 apparent luminosity tandem with assumed distance logarithms real apparent monochromatic luminosities differ with eqs.[e : monlumin amonlumin @xmath21 shift distance follows @xmath22 figure[f : pleiades apparent luminosity at 550 nm for stars pleiades assumed distance @xmath17pc lower envelope main - sequence nearby stars vertical shift of @xmath23 main sequence pleiades above lower envelope nearby stars distance follows with eq.[e : dism ] about 160pc assumed distance @xmath17mpc three cepheids in ic4182 periods for decreasing luminosity : 42d 9.24d 4.26d figure shows period - luminosity relation for cepheids in small magellanic cloud luminosity average maximum minimum value @xmath24 persei distance to smc @xmath25 persei lightcurves of three cepheids in galaxy ic4182 luminosities assumed distance @xmath17mpc brightest cepheid period 42d average apparent luminosity @xmath26 left frame figure[f : cepheid correct luminosity at 42d @xmath27 @xmath28 eq.[e : dism distance 3.3mpc ic4182 lightcurve coma composite several supernovae sn1a 1937 supernova type ia observed ic4182 distance from cepheids compute monochromatic luminosities lightcurve figure[f : sn1a ] figure : sn1a lightcurve supernova ia virgo cluster composite lightcurve supernovae ia coma cluster luminosities assumed distance @xmath17mpc maxima @xmath29d ) @xmath30 virgo @xmath31 coma distances. dism 18mpc 100mpc absorption scattering radiation interstellar matter optical depth @xmath32 subscript indicates wavelength dependence flux @xmath33 corrected absorption @xmath34 corrected flux entered in eqs.[e : monlumin, : amonlumin @xmath35 wavelength dependence colours star change eq.[e : absflux relate change colour optical depths different wavelengthsvisual - to - blue flux ratio @xmath36 to @xmath37 ratio intrinsic colour @xmath38 observed colour @xmath39 derive ratio @xmath40 figure[newred ] shows line strengths derive spectral types stars o5 a0v affected by interstellar absorption dependence of @xmath32 on wavelength dividing observed absorbed spectrum by standard spectrum result in figure interstellar absorption function of wavelength normalized at 550 nm decline towards longer wavelength due to absorption by gas extra opacity near 220 nm to dust know intrinsic colour star derive interstellar absorption distance data @xmath41 @xmath42 right hand side from eq.[e : redcol measurements at three wavelengths two independent colours determine reddening in figure left frame shows main sequence colours absence reddening right frame reddened star due to reddening corrections to colours applied for assumed values @xmath43 in steps of 0.15 correction for @xmath44 puts star on main - sequence colour relation close to g5v starintrinsic luminosity @xmath45 observed flux @xmath46 determined reddening @xmath44 distance star determined shows solution not unique early f star reddened by @xmath48 more reddened b star possible other colours allow unique solution visual magnitudes to luminosity at 550 nm @xmath49= 0.4\left[4.81-m_v\right colour to flux ratio @xmath50 : @xmath51 \log{f_{806}\over f_{550 } 0.4[v - i-1.35{f_{440}\over f_{365 } 0.4 b-0.54 constants follow definitions zero magnitude @xmath52 @xmath53 @xmath54 @xmath55 visual magnitudes to monochromatic luminosity : @xmath56 vdtolv absorption in magnitudes to optical depth : @xmath57 colour excess to difference optical depth : @xmath58 magnitudes definition absolute magnitude colour - magnitude diagram @xmath59 four arbitrary elements minus - sign high magnitudes low fluxes factor 2.5flux at zero magnitude @xmath60 expressed in @xmath61 defined for each filter @xmath37 distance of 10pc monochromatic luminosity no arbitrariness no quarrel with claudius ptolemaios assigning magnitudes 1 7 to stars visible switch to convenient notation apparent luminosities advantage assumed distance @xmath19 free suitable value chosen for each application determining distance to coma cluster reference distance 10pc less suitable than mpc disadvantage each luminosity assumed distance @xmath19 disadvantage choices of @xmath19 to integer powers of 10 power used added to index @xmath62 luminosity @xmath17mpc @xmath63 @xmath64pc numbers 1 - 9 suffice flux represents weighted average over filter replace wavelength index with filter name @xmath65 luminosity from average flux in @xmath52 filter for @xmath64kpc sample lecture notes test monochromatic luminosities re - written lecture notes for course on life of stars dust to black holes first year physics astronomy students utrecht course dutch uses si unitsenglish versions sections website advantage magnitude system flux measurement based ratios calibrated sources radio astronomers well without magnitudes if astronomers stop using magnitudes avoid asking committees international astronomical union define method notation preference perryman et al . 1995 a&a 304 , 69 johnson et 1952 apj 117 , 313 arp 1960 aj 65 , 404 saha et 1994 apj 425 , 14 baade zwicky 1938 apj 88 , 411 ( 1937c kimeridze tsvetkov 1987 soviet astr . 25 , 513 barbon 1989 a&a 220 , 83 zwicky 1961 pasp 73 , 185 1967 aj 72 , 1366 1978 aj 83 , 13 kohoutek kowal 1978 pasp 90 , 565 1975f gunn 1983 apjs 52 , 121 seaton 1979 mnras 187 , 75p bessell et al . 1998 a&a 333 , 231 | 0.430052 | 3,011 | 1,588 | , Saving $0.1 in GPT-4. |
20,865 | we present a preliminary analysis of public epic data of one of the otelo targets , the groth - westphal strip , gathered from the xmm - newton science archive ( xsa ) .
epic images are combined with optical @xmath0 data from our broadband survey carried out with the 4.2 m wht at la palma . | otelo @xcite will search for emission line objects using osiris tunable filters at the 10 m gtc telescope in la palma in selected atmospheric windows ( centred at the h@xmath1 line at z = 0.24 and z = 0.4 ) that are relatively free of sky emission lines . a total area of more than one square degree will be observed .
the survey technique will allow for separation of the h@xmath1 and [ nii ] lines and therefore agns from starburst galaxies .
a 5@xmath2 depth of @xmath3 erg @xmath4 s@xmath5 will make otelo the deepest emission line survey to date .
otelo science includes the evolution of galaxies , the evolution of star formation in the universe , chemical evolution of galaxies , qso spatial density determination , agn evolution , the low - end of the galaxy luminosity function , galactic astronomy etc .
a complementary optical broadband survey is currently on - going ( morphology , photometrical redshifts ) .
it is intended to complement the optical survey with other ground or space - based facilities , from x - rays ( xmm - newton , chandra ) to fir and sub - mm ( herschel , gtm ) .
the present work is being performed in the framework of the multiwavelength study of the otelo fields .
xmm - newton observations of the groth - westphal strip @xcite were collected from the xmm science archive ( xsa ) .
the epic observations were reprocessed using sas v6.0.0 emproc and epproc standard procedures .
high - radiation intervals were removed by inspecting the count rate curves .
gti files were created by means of the tabgtigen sas task .
these files were further included in standard filtering expressions .
the observations were co - added by means of the sas merge procedure .
attitude files were also merged .
final exposure times were about 82 ksec and 70 ksec in mos and pn , respectively .
three energy bands were defined : ( soft band ) , ( medium band ) and ( hard band ) .
sources were detected by means of the edetect_chain sas procedure .
we imposed a likehood parameter @xmath6 , were @xmath7 is the probability that the source exists .
furthermore , we considered only those sources for which @xmath8 , lying within the optical fov .
we have detected 75 sources fulfilling these conditions .
two hardness ratios have been also computed by this sas procedure .
optical @xmath0 observations were carried out at the prime focus of the 4.2 m william herschel telescope ( wht ) of the observatorio del roque de los muchachos ( la palma ) .
image size is with a plate scale of 0.24 arcsec / pixel .
total exposure time at each of the three pointing directions is 9000 sec per filter .
reduction process followed standard steps using iraf packages .
photometric calibration was obtained with several landolt standard fields .
absolute astrometry was performed using the usno b1 catalogue .
sources were extracted using sextractor 2.2 @xcite .
the 50% detection level is 25.3 in b , 25.3 in v , 24.7 in r and 23.5 in i. details on the optical observations can be found in @xcite .
we have matched x - rays and optical sources by searching in boxes centred in the x - rays sources coordinates . upon comparison with published chandra
coordinates @xcite we found a bulk shift of @xmath9 and @xmath10 . after correcting for this global image shift , we found : _ ( a ) _ unique match in all photometric bands for 43 sources ( 57.3% ) ; _ ( b ) _ partial match ( not detected in all bands ) for 10 sources ( 13.3% ) ; _ ( c ) _ multiple match for 10 sources ( 13.3% ) ; _ ( d ) _ partial match + multiple match for 2 sources ( 2.7% ) ; and _ ( e ) _ no optical counterpart found for 10 sources ( 13.3% )
since photometric redshift values are still not available , we have concentrated our analysis in distance - independent parameters , as described below . the x - rays to optical flux ratio ( x / o )
is a powerful means to discriminate between different classes of x - rays sources ( fiore et al . , 2003 ; della ceca et al . ,
figure [ fig1 ] shows a diagnostic diagram combining the 0.5 - 4.5 kev to optical r - band flux ratio and one of the hardness ratios .
the dashed - dotted line corresponds to x / o = 0.1 , typical of coronal - emitting stars , normal galaxies and heavily absorbed agns .
five of our sources ( 11.6% ) lie below this line , and therefore are likely either normal galaxies or compton - thick agns .
the dashed - line box corresponds to the region containing the 85% of the optically identified broad - line agns in the xmm - newton bright source sample ( bss ) @xcite .
29 of our sources ( 67.4% ) fall within the broad - line agn region defined by this box .
objects with harder hr and large x / o ( i.e. those to the right of the broad - line box ) are likely narrow - line agns , according to the bss and hbss diagnostics from @xcite .
we have derived the b / t ( bulge - to - total luminosity ) relation for the sample objects , comparing its distribution with a sample of active galaxies in the local universe @xcite , as shown in figure [ fig2 ] .
while not incompatible , b / t distributions are likely different , as proven by means of a k - s test : null hypothesis ( that both samples are drawn from the same distribution ) significance is only 17% . if we exclude liners from the local universe sample ( i.e. considering only seyfert galaxies ) , the null hypothesis significance is higher but still reduced , 39% .
the x - rays selected sample tends to higher b / t values and lacks very low b / t objects ( generally present in latest hubble types , t @xmath11 4 ) .
finally , we do nt find any correlation between x - rays hardness ratios and the b / t ratio . | otelo @xcite emission line objects osiris filters 10 m gtc telescope la palma atmospheric windows h@xmath1 line z = 0.24 = 0.4 free sky emission lines more than one square degree observed separation h@xmath1 nii ] lines agns from starburst galaxies 5@xmath2 @xmath3 @xmath4@xmath5 deepest emission line survey science includes evolution galaxies star formation chemical evolution qso spatial density determination agn evolution low - end galaxy luminosity function galactic astronomy complementary optical broadband survey on - going photometrical complement optical survey facilities x - rays fir sub - mm work multiwavelength study otelo fields xmm - newton observations groth - westphal strip collected from xmm science archive observations reprocessed sas v6.0.0 emproc standard procedures high - radiation intervals removed count rate curves gti files created sas task included standard filtering expressions observations co - added sas merge procedure attitude files merged final exposure times 82 ksec 70 ksec mos pn energy bands defined soft medium hard band sources detected edetect_chain sas procedure imposed likehood parameter @xmath6 @xmath7 probability source exists considered sources @xmath8 within optical fov detected 75 sources fulfilling conditions two hardness ratios computed optical @xmath0 observations prime focus 4.2 m william herschel telescope observatorio del roque de los muchachos image size 0.24 arcsec / pixel exposure time 9000 sec per filter reduction process iraf packages photometric calibration landolt standard fields absolute astrometry usno b1 catalogue sources extracted sextractor 2.2 @xcite 50% detection level 25.3 in b 25.3 v 24.7 r 23.5 in i details optical observations in @xcite matched x - rays optical sources boxes x - rays sources coordinates found bulk shift of @xmath9 @xmath10 unique match in photometric bands for 43 sources3% partial match not bands 10 sources ( 13.3% ) multiple match 10 sources 13.3% d partial match + multiple match 2 sources ( 2.7% ) no optical counterpart 10 sources 13.3% photometric redshift values not available concentrated analysis in distance - independent parameters x - rays to optical flux ratio x sources fiore 2003 figure diagnostic diagram 0.5 - 4.5 kev to optical r - band flux ratio hardness ratios dashed - dotted line x / o = 0.1 typical coronal - emitting stars normal galaxies heavily absorbed agns five sources ( 11.6% ) below line likely normal galaxies or compton - thick agns dashed - line box region 85% identified broad - line agns xmm - newton bright source sample @xcite 29 sources ( 67.4% ) within broad - line agn region objects harder hr large x / o right likely narrow - line agns derived b / t - total luminosity ) relation sample objects with active galaxies local universe @xcite figure not incompatible b / t distributions likely different proven k - s test null hypothesis significance 17%exclude liners local universe sample seyfert galaxies null hypothesis significance higher reduced 39% x - rays sample higher b / t values lacks low b / t objects hubble types correlation x - rays hardness ratios b / t ratio | 0.262979 | 1,391 | 707 | , Saving $0.0 in GPT-4. |
20,869 | by the weight of a boolean function @xmath0 , denoted by @xmath1 , we mean the number of inputs for which @xmath0 outputs @xmath2 . given a promise that an @xmath3-variable boolean function ( available in the form of a black box and the output is available in constant time once the input is supplied ) is of weight either @xmath4 or @xmath5 @xmath6 , we present a detailed study of quantum algorithms to find out which one actually it is . to solve this problem we apply the grover s operator .
first we consider the restricted problem . given a promise that an @xmath3-variable boolean function is of weight either @xmath7 or @xmath8 ( @xmath9 means the nearest integer corresponding to the real value @xmath10 ) ,
we show that one can suitably apply grover s operator for @xmath11-many iterations to decide which case this is with a probability almost unity for large @xmath3 and @xmath11 in @xmath12 . on the other hand ,
the best known probabilistic classical algorithm has a success probability close to @xmath13 ( from above ) after @xmath11 many steps when @xmath11 is large .
we further show that the best known probabilistic classical algorithm can achieve a success probability almost unity only after @xmath14 many iterations where @xmath15 .
this indicates a quadratic speed up ( and also agrees to the quadratic speed up by the use of grover s algorithm in database search ) on time complexity in the quantum domain with respect to the best known result in the classical domain .
second , we modify the basic randomized algorithm into a sure success algorithm , which can distinguish boolean functions of weights @xmath4 or @xmath5 for any @xmath16 , @xmath17 . to do that we have exploited a sure success grover search algorithm , which modifies the very last operation . for the weight decision problem , we show that the very last two operations should be changed to distinguish any weight with certainty and found the phase conditions for the last two operations . as quantum counting methods exist , which can count the number of solutions , here we compare our method with that . since the quantum counting method needs to exploit period information , which requires many grover operations , we have found that our method is faster than the quantum counting method .
* keywords : * quantum algorithm , boolean function , grover s operator , weight decision problem . | david deutsch designed a quantum algorithm , which evaluates whether the two outputs of a boolean function is the same or not using only one function evaluation @xcite .
deutsch - jozsa generalized the deutsch s algorithm for more general case such as whether the boolean function is constant or balanced @xcite .
deutsch - jozsa algorithm had been proposed to show an exponential speed - up in quantum machine than the classical machines .
the most important contribution in this area has been achieved when peter shor discovered a polynomial - time quantum algorithm for factoring and computing discrete logarithms , which is exponentially faster than the classical best known algorithm @xcite . because this quantum factoring algorithm works in polynomial time compared to exponential time in classical method
, many researchers started to find other applications .
lov grover discovered a quantum database search algorithm , which is quadratic faster than classical database search algorithm @xcite .
meanwhile , database search algorithm is one of the most widely used algorithm in the computer applications , the impact is so huge and more researchers are interested in other applications of quantum database search algorithm .
this research also focuses on the applications of grover database search algorithm . in this research
, we assume that there is a promise in the boolean function as it has a weight @xmath4 or @xmath5 .
initially , for better understanding , we formulate the problem for some special weight cases such as @xmath18 or @xmath19 , by applying @xmath11 many grover operations . here the quantum algorithm presented is of randomized in nature , but the success probability is arbitrarily close to unity .
next we consider the general case to distinguish functions having weight either @xmath4 or @xmath5 and we also consider a deterministic algorithm .
this requires changes in the very last two grover operations with phase conditions .
briefly , from @xmath20 to @xmath21 steps , the original grover operators are used , but , in the last two steps , two different grover operators are used with the phase conditions .
also we found that the phase conditions depending on the required number of grover operations .
meanwhile , because quantum counting algorithm was already proposed , we compared two methods . since the quantum counting method requires more grover iterations to find period information , we can conclude that our method is faster than this method .
a boolean function @xmath22 on @xmath3 variables may be viewed as a mapping from @xmath23 into @xmath24 .
a boolean function @xmath0 is constant if @xmath25 for all @xmath26 , @xmath27 .
that means @xmath1 is either @xmath28 or @xmath29 .
a boolean function @xmath0 is balanced if @xmath30 .
given a promise that the function @xmath0 is either constant or balanced , one may ask for an algorithm , that can exactly answer which case it is . note
that throughout this document we consider that any boolean function @xmath0 is available in the form of an oracle ( black box ) only , where one can apply an input to the black box to get the output .
a classical algorithm needs to check the function for @xmath31 inputs in the worst case to decide whether the function is constant or balanced .
( 25,9 ) ( 1.3,7.6)(0,0)[bl ] @xmath32 ( 1.3,4.6)(0,0)[bl ] @xmath33 ( 10,4)(5,5 ) ( 3,5)(1,0)2 ( 3.6,7.6)(1,1)0.8 ( 3.4,8.5)(0,0)[bl ] @xmath3 ( 5,4.5)(1,1 ) ( 4.8,4.6)(0,0)[bl ] @xmath34 ( 6,5)(1,0)4 ( 15,5)(1,0)7 ( 3,8)(1,0)2 ( 5,7.5)(2,1 ) ( 4.8,7.6)(0,0)[bl ] @xmath35 ( 7,8)(1,0)3 ( 15,8)(1,0)2 ( 17,7.5)(2,1 ) ( 16.8,7.6)(0,0)[bl ] @xmath35 ( 19,8)(1,0)1 ( 20,7.5)(1,1 ) ( 19.75,7.6)(0,0)[bl ] @xmath36 ( 21,8)(1,0)1 ( 9.9,4.9)(0,0)[bl ] @xmath37 ( 9.9,7.9)(0,0)[bl ] @xmath38 ( 14,7.9)(0,0)[bl ] @xmath38 ( 12.2,4.9)(0,0)[bl ] @xmath39 ( 11.7,6.5)(0,0)[bl ] @xmath40 ( 4,2)(0,0)[bl ] @xmath41 ( 9,2)(0,0)[bl ] @xmath41 ( 16,2)(0,0)[bl ] @xmath41 ( 19.5,2)(0,0)[bl ] @xmath41 ( 4,0.5)(0,0)[bl ] @xmath42 ( 9,0.5)(0,0)[bl ] @xmath43 ( 16,0.5)(0,0)[bl ] @xmath44 ( 19.5,0.5)(0,0)[bl ] @xmath45 it is known that given a classical circuit for @xmath0 , there is a quantum circuit of comparable efficiency which performs a transformation @xmath40 that takes input like @xmath46 and produces the output @xmath47 .
given such a @xmath40 , deutsch - jozsa @xcite provided a quantum algorithm that can solve this problem in constant time , indeed , in a single evaluation of @xmath40 . the circuit for their algorithm is given in fig . [ fig1 ] .
_ deutsch - jozsa algorithm @xcite _ [ algo : deutsch_jozsa_algorithm ] [ cols= " < ,
< " , ] we analyze the quantum counting algorithm for three purposes .
first , when a weight @xmath16 is given , how we can check the correctness of the given weight with how much time complexity .
second , when two weights are given , how we can decide the real weight with how much time complexity .
third , when @xmath3 possible weights are given , how we can find the real weight with how much time complexity .
if a weight is given as @xmath16=sin@xmath48 , how can we check whether this is correct or incorrect and what about time complexity ? let @xmath49 . if @xmath0 were an integer , there would be two possibilities : either @xmath50 ( which happens if @xmath51 or @xmath52 ) , in which case @xmath53 , or @xmath54 , in which case @xmath55 , where @xmath56 and @xmath57 are complex numbers of norm @xmath58 @xcite .
in other words , if we assume the value of @xmath59 as @xmath60 , the measured value of @xmath61 should be @xmath11 . as a result
, we can easily check whether the given weight is correct or not by measuring @xmath62 .
if the measured value is @xmath11 , the given weight is correct , otherwise incorrect . in this case , the time complexity is @xmath63 .
meanwhile , because we already know the value of @xmath64 in the initial time , we can find the smallest integer value of @xmath59 as @xmath65 .
if @xmath64 is @xmath66 , then @xmath59 is just @xmath67 .
therefore , the time complexity of this case is @xmath68 when @xmath64 is @xmath66 . from the previous section
, we can know that when @xmath71(@xmath72 ) is given , we can find the required number @xmath59 for @xmath69(@xmath70 ) and the expected measured value as @xmath73(@xmath74 ) .
however , when we want to decide which weight is real one , @xmath73 and @xmath74 should be different because they are the clues to distinguish . therefore , we need to find the integer value @xmath59 , which will be used for both two cases .
two values of @xmath59 are @xmath75 and @xmath76 .
because we need to execute the algorithm only once , @xmath77 and @xmath78 should be the same .
hence , @xmath79 should be equal to @xmath80 and our job is to find suitable @xmath73 and @xmath74 . meanwhile ,
if @xmath81 and @xmath82 , @xmath59 should be @xmath83 . considering the time complexity , we need to find the smallest integer value of @xmath59 by @xmath84 , where @xmath85 is the least common integer multiplier
. then , @xmath86 and @xmath87 .
finally , with the value of @xmath88 , we evaluate the algorithm , and if the measured value of @xmath61 is @xmath73 , we can say that the weight is @xmath69 .
if the measured value of @xmath61 is @xmath74 , we can say that the weight is @xmath70 .
time complexity of this case is @xmath89 , where @xmath90 is @xmath91 and @xmath92 is @xmath93 .
now we need to compare our method and the above method based on quantum counting when @xmath94 .
if @xmath95 and @xmath96 , our method , algorithm [ algo : randomized_weight_decision_algorithm ] and [ algo : sure_success_weight_decision_algorithm ] can decide which one is real one in @xmath11 many steps . on the other hand ,
the method based on quantum counting needs @xmath59 as @xmath97 and if @xmath61 is @xmath11 , the real weight is @xmath69 and if @xmath98 @xmath99 , the real weight is @xmath70 . from this analysis , we can know that the method based on quantum counting is less efficient than our method because the quantum counting method needs to exploit some period , which requires several grover operations .
we can extend the previous result to a more general case .
when @xmath3 possible weights are given .
can we decide which weight is real one ?
the most important thing in this problem is to find the smallest integer value of @xmath100 .
then , as the same method in the previous section , @xmath101 should be @xmath102 . from this analysis
, we can find the real weight from the measured value @xmath61 as if @xmath61 is @xmath101 then the real weight is @xmath103 . in this case , the time complexity is @xmath104 .
in this work , we have investigated the application of grover operators to distinguish weight of boolean function when two weights are given .
firstly , when we assume that the weight is the number of solutions , we found that algorithm [ algo : randomized_weight_decision_algorithm ] can find the exact weight with almost certainty .
secondly , by exploiting the sure success grover search method , we found a sure success weight decision algorithm , algorithm [ algo : sure_success_weight_decision_algorithm ] , with modification of the last two grover operations with phase conditions .
lastly , we have compared the proposed method to the quantum counting algorithm . because the quantum counting algorithm needs period information , which requires more grover operations ,
our method is more efficient than the quantum counting method . on the other hand , until this work
, we assume that two weights are given such as @xmath4 and @xmath5 .
how about other cases such as @xmath69 and @xmath70 , where @xmath105 ?
moreover , when three or more weights are given , can we find the exact weight with the similar approach ?
this may be attempted using the brief idea presented in subsection [ difwt ] .
+ * acknowledgment : * samuel l. braunstein currently holds a royal society - wolfson research merit award .
byung - soo choi is supported by the ministry of information & communication of korea ( it national scholarship program ) .
l. grover . a fast quantum mechanical algorithm for database search . in _ proceedings of 28th annual symposium on the theory of computing ( stoc )
_ , may 1996 , pages 212219 .
available at xxx.lanl.gov/quant-ph/9605043 . | david deutsch designed quantum algorithm evaluates two outputs of boolean function same using one evaluation deutsch - jozsa generalized algorithm for case boolean function constant or balanced algorithm proposed exponential speed - up in quantum machine important contribution peter shor discovered polynomial - time quantum algorithm for factoring computing discrete logarithms exponentially faster than classical algorithm @xcite researchers other applications lov grover discovered quantum database search algorithm quadratic faster than classical algorithm @xcite database search algorithm widely used more researchers interested in other applications research focuses on applications grover database search algorithm assume promise in boolean function weight @xmath4 or @xmath5 formulate problem for special weight cases @xmath18 or @xmath19 applying @xmath11 many grover operations quantum algorithm randomized success probability arbitrarily close to unity consider general case to distinguish functions having weight @xmath4 or @xmath5 deterministic algorithm requires changes in last two grover operations with phase conditions @xmath20 to @xmath21 steps original operators used last two steps different operators with phase conditions phase conditions depending on required number of grover operationsquantum counting algorithm proposed compared two methods iterations our method faster boolean function @xmath22 on @xmath3 variables mapping from @xmath23 into @xmath24 boolean function @xmath0 constant if @xmath25 @xmath26 @xmath1 @xmath28 or @xmath29 @xmath0 balanced if @xmath30 algorithm boolean function @xmath0 available oracle black box ) classical algorithm for @xmath31 inputs constant 25,9 ) 1.3,7.6)(0,0) @xmath32 ( 1.3,4.6)(0,0 @xmath33 ( 10,4)(5,5 ) ( 3,5)(1,0)2 ( 3.6,7.6)(1 3.4,8.5)(0,0 @xmath3 ( 5,4.5)(1,1 4.8,4.6)(0,0 @xmath34 ( 6,5)(1,0)4 ( 15,5)(1,0)7 ( 3,8)(1,0)2 ( 5,7.5)(2,1 4.8,7.6) @xmath35 ()3 ( 15,8)(1,0)2 ( 17,7.5)(2,1 ) 16.8,7.6)(0,0 @xmath35 ( 19,8)(1,0)1 ( 20,7.5)(1 19.75,7.6)(0,0 @xmath36 ( 21,8)(1,0)1 ( 9.9,4.9)(0,0 @xmath37 ( 9.9,7.9)(0,0 @xmath38 ( 14,7.9)(0,0 12.2,4.9)(0,0 @xmath39 ( 11.7,6.5)(0,0 @xmath40 ( 4,2)(0,0 @xmath41 ( 9,2)(0,0 @xmath41 ( 16,2)(0,0 @xmath41 ( 19.5,2)(0,0 4,0.5)(0,0 @xmath42 ( 9,0.5)(0,0 @xmath43 ( 16,0.5)(0,0 @xmath44 ( 19.5,0.5)(0,0 @xmath45 classical circuit @xmath0 quantum circuit comparable transformation @xmath40 input output @xmath47 deutsch - jozsa @xcite quantum algorithm constant time single evaluation circuit algorithm fig . deutsch - jozsa algorithm @xcite quantum counting algorithm three purposes @xmath16 check correctness two weights decide real weight @xmath3 weights find real weightweight @xmath16=sin@xmath48 check correct or incorrect time complexity ? @xmath49 @xmath0 integer two possibilities : @xmath50 or @xmath53 @xmath54 @xmath55 @xmath56 @xmath57 complex numbers norm @xmath58 @xcite assume value @xmath59 as @xmath60 measured value @xmath61 be @xmath11 check weight measuring @xmath62 if measured value @xmath11 weight correct otherwise incorrect time complexity is @xmath63 know value @xmath64 find smallest integer value @xmath59 as @xmath65 if @xmath64 is @xmath66 @xmath59 @xmath67 time complexity @xmath68 when @xmath64 is @xmath66 @xmath71(@xmath72 ) given find required number @xmath59 for @xmath69(@xmath70 ) expected measured value @xmath73(@xmath74 ) decide weight real @xmath73 and @xmath74 different clues distinguish need find integer value @xmath59 used for two cases values @xmath59 are @xmath75 and @xmath76 execute algorithm once @xmath77 and @xmath78 same@xmath79 equal to @xmath80 find suitable @xmath73 @xmath74 if @xmath81 @xmath82 @xmath59 be @xmath83 find smallest integer value of @xmath59 by @xmath84 @xmath85 least common integer multiplier @xmath86 @xmath87 value of @xmath88 evaluate algorithm if value @xmath61 is @xmath73 weight is @xmath69 if @xmath61 is @xmath74 weight is @xmath70 time complexity is @xmath89 @xmath90 is @xmath91 @xmath92 is @xmath93 compare our method method quantum counting @xmath94 if @xmath95 and @xmath96 our method decide real one in @xmath11 method quantum counting needs @xmath59 as @xmath97 if @xmath61 is @xmath11 real weight is @xmath69 if @xmath98 @xmath99 real weight @xmath70 method based quantum counting less efficient period requires operations extend result general case @xmath3 possible weights weight real important find smallest integer value of @xmath100 same method previous section @xmath101 should be @xmath102 find real weight from value @xmath61 @xmath61 is @xmath101 real weight is @xmath103 time complexity is @xmath104 investigated grover operators distinguish weight boolean function two weights given weight number of solutions algorithm randomized_weight_decision_algorithm find exact weight certainty sure success grover search method found sure success weight decision algorithm modification last two grover operations phase conditions compared proposed method quantum counting algorithm needs period information more grover operations our method more efficient assume two weights @xmath4 and @xmath5 other cases @xmath69 @xmath70 @xmath105 ? three or more weights find exact weight similar approach ? attempted brief idea subsection [ difwt ] samuel l. braunstein holds royal society - wolfson research merit award byung - soo choi supported ministry of information & communication korea national scholarship program. grover fast quantum mechanical algorithm for database search proceedings 28th annual symposium theory computing may 1996 pages 212219 available xxx.lanl.gov/quant-ph/9605043 | 0.356155 | 2,790 | 1,563 | , Saving $0.1 in GPT-4. |
20,872 | phase - sensitive detection ( psd ) is an important experimental technique that allows signals to be extracted from noisy data .
psd is also used in modulation spectroscopy and is used in the stabilization of optical sources .
commercial lock - in amplifiers that use psd are often expensive and host a bewildering array of controls that may intimidate a novice user .
low - cost microcontrollers such as the arduino family of devices seem like a good match for learning about psd ; however , making a self - contained device ( reference signal , voltage input , mixing , filtering , and display ) is difficult , but in the end the project teaches students `` tricks '' to turn the arduino into a true scientific instrument . | lock - in amplification and phase - sensitive detection are important techniques in experimental physics , see @xcite for pedagogical uses of these techniques .
commerical devices are expensive and can be intimidating for new users .
building a home - made instrument can be instructive @xcite ; however , doing so requires advanced electronics skills that a student may not already have , putting the emphasis on the electronics and not the method .
conversely , it is possible to perform the mixing and filtering on a computer using a computer s sound card or other low - cost data acquisition devices to handle the input and output , see for instance @xcite , but in the author s experience computer driver issues often get in the way of relatively easy implementation . in this paper
i describe a psd that uses the popular arduino microcontroller and the processing programming environment @xcite .
the major design goal was to make the device as self - contained as possible , a task made difficult by the memory and hardware constraints of a typical microcontroller .
while these devices are fantastic for controlling robots and basic data - logging , turning them into scientific instruments requires techniques that go beyond what is normally found in the literature .
an added benefit is that these techniques can be used to build other types of instrumentation such as function generators and fast dmms out of a low - cost microcontroller that typically only cost a few tens of dollars .
arduino @xcite is a catch - all term for a family of open - source hardware based on atmel micro - controllers with a pre - loaded bootloader so that instead of programming the arduino with the more difficult , but more powerful , avr instruction set , users can program with a c@xmath0-like language .
programming and communication can be done via usb or through a set of on - board communication pins .
there have been sixteen different arduino - labeled boards produced to date , each of which has its own unique hardware and memory specifications . in this paper ,
the uno r3 is used , simply because it was what was on hand and is one of the cheaper and more basic arduinos .
the heart of the uno r3 is the atmel atmega328 microcontroller @xcite .
the uno r3 has 14 digital input / output pins , six of which provide pulse width modulation ( pwm ) output .
it also has six analog inputs , and an on - board @xmath1mhz oscillator .
the atmega 328 has 32 kb of flash memory to store programs , and 2 kb of sram for variable storage .
the program given in this paper is easily stored in the flash memory , but the limited sram places severe restrictions on the amount of data that can be taken and manipulated on - board the arduino .
and @xmath2 are the input and reference voltages respectively .
_ lpf _ is a low - pass filter and _ a _ is an amplifier . a phase - shifter is shown after @xmath2 , but this and the amplifier are not used in this project . see the text for more details . ] a block diagram for phase - sensitive detection is shown in fig .
[ fig : block ] .
the input stage of a psd consists of two signals , the input of interest and the reference signal .
typically , these two signals are oscillating at the same frequency , but in the derivations to follow this will only be enforced at the end . after the signals
are acquired , the input signal may be amplified and filtered . after the input stage comes the mixing stage where the input and reference signals are combined .
there are a number of ways of doing this , but in this implementation the signals will simply be multiplied .
finally , the output of the mixing stage is heavily filtered by a low - pass filter .
once again this can be done using analog circuitry , but here it will be done mathematically .
let us assume that the input signal @xmath3 and the reference @xmath4 are given by @xmath5 a dc offset is included with @xmath3 because arduino , like most microcontrollers , only input or output positive voltage ; therefore the arduino outputs an offset reference voltage .
this offset is removed during mixing , so @xmath4 is centered on zero in what follows . upon multiplication of the two signals in eq .
[ eq : inputs ] , and making use of a trig identity , the output of the mixing stage @xmath6 is @xmath7-\cos\left [ \left(\omega_2+\omega_1 \right)t-\left(\phi_2+\phi_1\right ) \right]\rbrace\ ] ] at this stage eq .
[ eq : mix ] is nothing more than what we find in heterodyne detection in radio and optical engineering .
the power of coherent detection is that small signals get amplified by a larger local oscillator signal is evident in eq .
[ eq : mix ] . upon mixing of the two signals
there are contributions to the signal at the original frequencies and at the sum and difference frequencies .
this is the heart of psd ; the output filter is set to reject all frequencies other than the difference frequency of the inputs .
furthermore , if @xmath3 and @xmath4 are made to oscillate at the same frequency before entering the psd , the output @xmath8 of the psd simply depends on the phase difference between the two signals @xmath9 the final effect of this filtering is to move our output from @xmath10 to dc .
the stronger the filtering , the more noise is rejected and the signal - to - noise increases .
however , in effect , the psd is performing signal averaging , so with each factor of two increase in the signal - to - noise , the collection time increases by a factor of four .
the other disadvantage of psd is that by making the measurements at dc we are placing our signal where @xmath11 noise dominates@xcite .
once again we can see similarities with homodyne detection , and even though we have added no active amplification , the presence of a strong local oscillator can boost a weak experimental signal .
as stated in sec . [ sec : intro ] , the ultimate goal of this project was to make the project as self - contained as possible .
it would need only a computer for display , the arduino , and as few passive circuit components as possible .
these turned out to be fairly severe constraints and required some exploits that are not commonly presented in introductions to microcontrollers .
the structure of this section is to discuss the implementation of each of the sub - systems shown in fig .
[ fig : block ] .
briefly , and following fig .
[ fig : flowchart ] , the arduino synthesizes a sine wave output from a wavetable . as
the output is updated from the wavetable , the input signal is quickly read .
finally after the arduino has cycled through the complete wavetable , mixing is performed mathematically onboard the arduino , and the resulting waveforms are sent to the host computer running processing for filtering and display . creating the reference signal is in many ways the most significant constraint on the project .
the arduino does not have a true analog output .
the only way to approximate an analog signal is to use pwm with a low pass filter circuit on the output to smooth things out .
additional tricks are needed to generate something resembling a sine wave . to generate a reference signal under these limitations
a technique sometimes called `` bit - banging''@xcite is employed .
bit - banging takes advantage of the timers on the arduino and pwm . in pwm
, the duty - cycle ( ratio of ` on ' to ` off ' times ) of a square wave is modulated .
the greater the duty - cycle , the longer the pwm pin on the arduino is held high . integrating that square - wave produces a dc voltage that increases with the duty - cycle .
the final step is to rapidly change this dc voltage so that the desired waveform , in this case a sine - wave , is synthesized .
this paper will only briefly sketch the relevant idea . for details , refer to the comments in the source code for this project , the `` bit - banging '' paper @xcite , or the atmega328 datasheet @xcite .
this technique relies on two counters on the arduino .
timer 1 , tcnt1 , is a 16-bit timer / counter that is configured as an 8-bit counter .
tcnt1 runs at the full arduino clock frequency , @xmath12 , and produces the pwm output signal .
timer 2 , tcnt2 , runs slower than tcnt1 by a factor of eight , and is used to step through the pre - generated wavetable .
[ fig : bitbang ] output compare registers ( ocrna ) set a flag when the value of tcntn matches ocrna .
when the ocrna flag is set each timer can exhibit different behavior depending on how the arduino is programmed .
when the ocr1al flag is set , the pwm pin is made to go low , but tcnt1 keeps incrementing until it overflows .
at which point tcnt1 starts counting again , the pwm pin goes high again , and ocr1la is reset . so changing the value of ocr1al determines the duty - cycle of the pwm signal and consequently the dc voltage that is used to construct @xmath4 .
the pre - programmed wavetable for @xmath2 is what updates ocr1al . to do this tcnt2 ,
running eight times slower than tcnt1 , updates ocr1al with the next value from the wavetable upon reaching ocr2a .
ocr2a is pre - set in the arduino in the code such that the frequency of @xmath2 is given by : @xmath13 when tcnt2 reaches ocr2a , an interrupt is called which does two things , update ocr1al to the next value of the wavetable and get an input voltage for the adc .
the last step in generating @xmath2 is to place a simple low - pass rc filter on the pwm pin to get a smooth sine - wave .
the arduino adc is a 10-bit successive approximation circuit connected to an 8-channel multiplexer .
the measurements are single - ended and normally referenced to a @xmath14 on - board reference voltage , although it is possible to use an external reference voltage . because interrupts are used to create an analog output for the reference signal , it is important to ensure that anything that happens during the interrupt is quick , which is why in fig .
[ fig : flowchart ] , the data is sent for filtering after the wavetable has been completely cycled through .
serial calls are slow . according to atmel the ultimate sampling rate for single - ended measurements
is limited by the adc clock speed .
the adc clock speed is derived from the main system clock , and for maximum resolution should be between @xmath15 and @xmath16 , however , for purposes of this project satisfactory resolution and accuracy are obtained by setting the adc clock to a speed of @xmath17 .
the successive approximation circuit requires @xmath18 clock cycles , giving a sampling rate of @xmath19 @xcite , which is much faster than this project needs . since the values of the reference signal are pre loaded into the arduino memory , changing the phase while data is being taken is difficult .
often , however , the ability to shift the relative phase between @xmath20 and @xmath2 is need to maximize the signal .
additionally , sometimes what is needed in a measurement is the phase shift between the reference and input signals . to meet these needs a cue
is taken from dual - phase , lock - in amplifiers@xcite and in addition to our reference waveform a quadrature signal , which is just the reference waveform shifted by @xmath21 , is generated .
both the in - phase , @xmath22 and quadrature @xmath23 mixed waveforms are used so that the phase ambiguity is removed by calculating the magnitude signal @xmath24 and the phase difference between the input and reference signals is @xmath25 .
the @xmath26 and @xmath27 signals are what are sent from the arduino to the host computer for filtering and display .
once the arduino finishes running through the waveform , it takes the collected data and sends it via usb to a computer running the processing ide @xcite .
processing is a programming language and development environment that was designed to make it easier for the arts community to become software literate . like the arduino
, it has a vibrant community that has produced numerous tutorials , examples , and books that make learning processing relatively easy .
processing is also a direct forebear of the arduino development system and therefore makes it a natural fit with the arduino side of this project . in this project
the processing sketch ( processing - talk for program ) initiates serial communications with the arduino and once communications are established , the arduino sends the @xmath28 and @xmath29 data to the processing sketch
. the processing sketch applies a recursive , single - pole , low - pass filter[bib : dsp_guide ] to the array that is to be displayed ( @xmath30 ) . in general a recursive filter is given by @xmath31=a_0 x[n]&+a_1x[n-1]+a_2x[n-2]+\dots\\ & + b_1 y[n-1]+b_2y[n-2]+\dots,\end{aligned}\ ] ] where @xmath32 is the output of the filter and @xmath33 is the input data .
for this project a single - pole filter was used , so the only coefficients used are given by a single parameter @xmath34
@xmath35 where @xmath36 is the time constant of the filter .
mathematically this filter is identical to a single - pole rc - filter in electronics .
the stronger the filtering , the better it is for lock - in detection , since inadequate filtering causes the output to oscillate at the reference frequency . ref .
@xcite has algorithms for calculating coefficients for higher - order filters with faster roll - off , single - pole filters are limited to attenuations of @xmath37db / decade .
to test this project a all - pass phase - shifter was built .
[ fig : phase_shift ] shows the circuit used to test the arduino .
the voltage gain for this circuit is @xmath38 , so the amplitude of the output is unchanged with respect to the input voltage .
the high - pass rc filter at the non - inverting terminal of the op - amp controls the amount of phase - shift at the output . at @xmath39
the phase - shift is @xmath21 and changes by @xmath40 . by varying @xmath41 , the @xmath21 point shifts and moves our operating frequency along the phase plot . to test the arduino psd , the reference signal generated by the arduino
is sent to the input of the phase - shifting circuit .
the output of the phase - shifting circuit becomes the input signal to the psd .
[ fig : results ] shows the @xmath26 signal , as displayed in the processing sketch , as @xmath41 is varied . [
fig : phase_shift ] shifts the phase through @xmath42 degrees.,title="fig : " ] [ fig : results ] of the phase - shift test circuit is varied.,title="fig:",width=326 ]
a phase - sensitive detector using an arduino microcontroller has been described .
the main design goal of the project was to make the psd as self - contained as possible , namely that an external reference signal was not needed and that any phase corrections were unnecessary . aside from the arduino and a computer the only other hardware needed
are a resistor and a capacitor .
furthermore the techniques used for this project can be used in other projects involving using the arduino microcontroller as a standalone , low - cost , scientific instrument .
the `` bit - banging '' technique is not limited to sine - waves , but can be used to output any synthesized waveform that may be required .
more complicated waveforms can be created outside of the arduino environment and loaded into the eeprom allowing faster execution and freeing up regular memory.@xcite | lock - in amplification phase - sensitive detection important techniques in experimental physics see @xcite for pedagogical uses commerical devices expensive intimidating for new users building home - made instrument instructive @xcite requires advanced electronics skills emphasis on electronics not method possible to perform mixing filtering on computer using sound card or low - cost data acquisition devices input output see @xcite computer driver issues easy implementation paper describe psd uses arduino microcontroller processing programming environment @xcite design goal to make device self - contained difficult by memory hardware constraints typical microcontroller devices fantastic for controlling robots data - logging turning into scientific instruments requires techniques beyond literature techniques can used to build other instrumentation function generators fast dmms out of low - cost microcontroller arduino @xcite for family open - source hardware based on atmel micro - controllers with pre - loaded bootloader program with c@xmath0-like language programming communication via usb or on - board communication pins sixteen different arduino - labeled boards produced each unique hardware memory specifications in paper uno r3 used on hand cheaper basic arduinosheart of uno r3 is atmel atmega328 microcontroller @xcite has 14 digital input / output pins six provide pulse width modulation output six analog inputs on - board @xmath1mhz oscillator atmega 328 has 32 kb flash memory 2 kb sram for variable storage program stored in flash memory limited sram on data on arduino. @xmath2 are input and reference voltages _ lpf _ is low - pass filter _ a _ is amplifier phase - shifter shown after @xmath2 amplifier not used in project see text for details block diagram for phase - sensitive detection in fig . input stage of psd two signals interest reference signal oscillating at same frequency derivations only enforced at end after signals acquired input signal amplified and filtered mixing stage input reference signals combined signals multiplied output mixing stage filtered by low - pass filter done using analog circuitry mathematically input signal @xmath3 and reference @xmath4 given by @xmath5 dc offset included with @xmath3 because arduino input output positive voltage outputs offset reference voltageoffset removed during mixing @xmath4 centered on zero multiplication of two signals in eq . trig identity output mixing stage @xmath6 is @xmath7-\cos\left [ (\omega_2+\omega_1-\left(\phi_2+\phi_1\right ) \rbrace eq . mix heterodyne detection in radio optical engineering coherent detection small signals amplified by larger local oscillator signal mixing two signals contributions to at original frequencies sum and difference frequencies heart of psd output filter reject all frequencies other than difference frequency if @xmath3 and @xmath4 oscillate at same frequency before psd output @xmath8 depends on phase difference signals @xmath9 filtering output from @xmath10 to dc stronger filtering more noise rejected signal - to - noise increases psd signal averaging each two increase in signal - to - noise collection time increases by four disadvantage of psd measurements at dc placing signal where @xmath11 noise dominates@xcite similarities with homodyne detection strong local oscillator can boost weak experimental signalsec intro ultimate goal project was self - contained only computer for display arduino few passive circuit components severe constraints required exploits not presented in introductions to microcontrollers structure section implementation of sub - systems in fig . fig block ] flowchart arduino synthesizes sine wave output from wavetable output updated input signal quickly read after arduino wavetable mixing performed onboard arduino resulting waveforms sent to host computer for filtering and display creating reference signal significant constraint arduino true analog output only approximate analog signal use pwm with low pass filter circuit output additional tricks needed to generate sine wave reference signal technique bit - banging''@xcite employed bit - banging takes advantage of timers on arduino and pwm in pwm duty - cycle of square wave modulated greater duty - cycle longer pwm pin on arduino held high integrating square - wave produces dc voltage increases with duty - cycle final step to rapidly change dc voltage desired waveform sine - wave synthesized paper briefly sketch relevant ideadetails refer comments source code project bit - banging paper @xcite atmega328 datasheet @xcite technique relies on two counters arduino timer 1 tcnt1 16-bit timer configured 8-bit counter tcnt1 runs full arduino clock frequency produces pwm output signal timer 2 tcnt2 runs slower tcnt1 factor eight step through pre - generated wavetable bitbang output compare registers ocrna set flag value tcntn matches ocrna set each timer different behavior arduino ocr1al flag set pwm pin low tcnt1 incrementing until overflows tcnt1 starts counting pwm pin high ocr1la reset changing value ocr1al determines duty - cycle pwm signal dc voltage construct @xmath4 pre - programmed wavetable @xmath2 updates ocr1al tcnt2 eight times slower tcnt1 updates ocr1al next value reaching ocr2a ocr2a pre - set arduino frequency @xmath2 given @xmath13 tcnt2 reaches ocr2a interrupt called ocr1al next value input voltage adclast step generating @xmath2 place low - pass rc filter on pwm pin smooth sine - wave arduino adc 10-bit approximation circuit connected to 8-channel multiplexer measurements single - ended referenced to @xmath14 on - board reference voltage possible use external reference voltage interrupts create analog output for reference signal important interrupt quick fig . data sent for filtering after wavetable cycled through serial calls slow ultimate sampling rate for limited by adc clock speed derived from main system clock maximum resolution between @xmath15 and @xmath16 satisfactory resolution accuracy setting to speed @xmath17 successive approximation circuit requires @xmath18 clock cycles sampling rate of @xmath19 @xcite faster than project needs values reference signal pre loaded into arduino memory changing phase while data taken difficult shift phase between @xmath20 and @xmath2 maximize signal phase shift between reference input signals cue from dual - phase lock - in amplifiers@xcite quadrature signal shifted by @xmath21 generatedin - phase @xmath22 quadrature @xmath23 mixed waveforms used phase ambiguity calculating magnitude signal @xmath24 phase difference between input reference signals @xmath25 @xmath26 @xmath27 signals sent from arduino to host computer for filtering display arduino collected data sends via usb to computer running processing processing programming language development environment arts community software literate arduino vibrant community tutorials examples books learning processing easy forebear of arduino development system natural fit with arduino project processing sketch initiates communications with arduino arduino sends @xmath28 @xmath29 data to processing sketch applies recursive single - pole low - pass filter to array displayed @xmath30 ) recursive filter given by @xmath31=a_0 x[n]+a_1x[n-1]+a_2x[n-2]+[n-1]+b_2y[n-2]+\dots @xmath32 output filter @xmath33 input data single - pole filter used coefficients given by single parameter @xmath34 @xmath35 @xmath36 time constant of filter filter identical to single - pole rc - filter in electronicsstronger filtering better for lock - in detection inadequate filtering causes output at reference frequency @xcite algorithms calculating coefficients for higher - order filters faster roll - off single - pole filters limited to attenuations of @xmath37db / decade all - pass phase - shifter built fig phase_shift circuit arduino voltage gain is @xmath38 amplitude output unchanged high - pass rc filter at non - inverting terminal controls phase - shift @xmath39 phase - shift is @xmath21 changes by @xmath40 varying @xmath41 @xmath21 point shifts moves operating frequency test arduino psd reference signal sent to input phase - shifting circuit output becomes input signal to psd fig results @xmath26 signal @xmath41 varied phase_shift shifts phase through @xmath42 degrees phase - shift test circuit phase - sensitive detector using arduino microcontroller described design goal make psd self - contained external reference signal not needed phase corrections unnecessary arduino other hardware needed resistor capacitortechniques used project can other projects arduino microcontroller standalone low - cost scientific instrument bit - banging technique not limited to sine - waves any synthesized waveform complicated waveforms created outside arduino loaded into eeprom faster execution freeing memory | 0.569524 | 3,454 | 1,741 | , Saving $0.1 in GPT-4. |
20,879 | among the direct searches for wimp - type dark matter , the dama experiment is unique in that it has consistently reported a positive signal for an annual - modulation signal with a large ( 9.3@xmath0 ) statistical significance .
this result is controversial because if it is interpreted as a signature for wimp interactions , it conflicts with other direct search experiments that report null signals in the regions of parameter space that are allowed by the dama observation .
this necessitates an independent verification of the origin of the observed modulation signal using the same technique as that employed by the dama experiment , namely low - background nai(tl ) crystal detectors . here ,
we report first results of a program of nai(tl ) crystal measurements at the yangyang underground laboratory aimed at producing nai(tl ) crystal detectors with lower background levels and higher light yields than those used for the dama measurements . dark matter , wimp , kims , nai(tl ) crystal 29.40.mc , 85.60.ha | numerous astronomical observations have led to the conclusion that most of the matter in the universe is invisible , exotic , and nonrelativistic dark matter @xcite . however , in spite of a considerable amount of experimental effort , the nature of the dark matter remains unknown . weakly interacting massive particles ( wimps )
are one of the most attractive candidates for dark matter particles @xcite . in supersymmetric models for beyond the standard model physics ,
the lightest supersymmetric particle ( lsp ) is a natural candidate for wimp - type dark matter .
a number of experiments have made direct searches for a wimp component of our galaxy by looking for evidence for wimp - nucleus scattering by detecting the recoiling nucleus in ultra - sensitive low - background detectors @xcite .
the dama experiment searches for an annual modulation in the detection rate of nuclear recoils in an array of ultra - low - background nai(tl ) crystals caused by the earth s orbital motion through our galaxy s dark - matter halo @xcite .
this experiment , which has been operating for over 15 years , has consistently reported a positive signal for an annual modulation with a phase that is consistent with expectations for motion of the earth relative to the galactic rest frame .
the statistical significance of the dama annual modulation signal has now reacehed @xmath1 .
other experiments , including cogent @xcite , cresst @xcite , and cdms @xcite , have also reported signals that could be interpreted as being possibly due to wimp interactions .
however , these signals have marginal significance at the < 3@xmath0 confidence level and some inconsistencies with null results using similar detectors that were already reported @xcite .
nevertheless , these marginal signals have motivated the same experimental groups and , in some cases , independent groups to devise experiments to check these signals using similar techniques but with higher sensitivity .
in contrast , to date , no independent verification of the dama signal in an experiment using the same technique has been done .
the dama signal , in particular its interpretation as being due to wimp - nucleus scattering , has been the subject of a continuing debate that started with the first dama report 15 years ago .
this is because the wimp - nucleon cross sections inferred from the dama modulation are in conflict with limits from other experiments that directly measure the total rate of nuclear recoils , such as xenon100 @xcite , lux @xcite , and supercdms @xcite .
however , room remains for explaining all of these experimental results without conflict in terms of nontrivial systematic differences in detector responses @xcite and the commonly used astronomical model for the wimp distribution @xcite .
an unambiguous verification of the dama signal by an independent experiment using similar nai(tl ) crystals is mandatory .
reproducing the dama measurement requires the independent development of ultra - low - background nai(tl ) crystals .
this is because the crystal - growing company that supplied the dama nai(tl ) crystals will not produce the same crystals for other experimental groups .
recently , the anais group has been developing ultra - low - background nai(tl ) crystals with the goal of reproducing the dama result @xcite and the dm - ice group reports background measurements of nai(tl ) crystals @xcite .
however , to date , no experimental group has produced nai(tl ) crystals with background levels at or below than those used in the dama experiment . the korea invisible mass search ( kims ) collaboration has been performing direct wimp searches using ultra - low - background csi(tl ) crystal detectors , which are similar experimental devices to the nai(tl ) crystals used in the dama experiment .
this implementation required extensive research and development aimed at identifying and reducing internal backgrounds in the csi crystals @xcite .
null results from kims reject wimp - iodine nuclei interactions as the source of the dama signal with very little model dependence @xcite .
however , because the dama results can be interpreted as being primarily due to wimp - sodium nuclei interactions , which , for example , would be the case for low - mass wimps , it remains necessary to confirm the dama observations with nai(tl ) crystal detectors .
this motivated a program to develop ultra - low - background nai(tl ) crystal detectors with lower background levels and and higher light yields ( and , thus , a lower energy threshold ) than those of the dama experiment in order to identify unambiguously the origin of the dama modulation signature .
to evaluate the nai(tl ) crystals , we use the experimental setup that was used for the kims csi(tl ) detector measurements at the yangyang underground laboratory ( y2l ) , which has a 700 m earth overburden ( 2400 m water equivalent ) . this includes a 12-module array of csi(tl ) detectors ( total mass of 103.4 kg ) inside a shield that consists , from inside out , of 10 cm of copper , 5 cm of polyethylene , 15 cm of lead , and 30 cm of liquid - scintillator - loaded mineral oil to stop external neutrons , gamma rays , and veto cosmic ray muons .
each detection module consists of a low - background csi(tl ) crystal ( @xmath2 ) with a photomultiplier tube ( pmt ) mounted at each end .
two nai(tl ) crystals were mounted inside the csi(tl ) detector array as shown in fig .
[ experimentalsetup ] . in november 2013 , we installed the first nai(tl ) crystal , nai-001 , in the csi array ; the second nai(tl ) crystal , nai-002 , was added in february 2014 , soon after that crystal was delivered and assembled .
the two nai(tl ) crystals were produced by the alpha spectra company .
the crystals were grown from nai powder that was not of the highest attainable purity ; the initial purification was carried out by alpha spectra .
the crystals have a cylindrical shape and were cut from a 32 inch ingot that was grown by the kyropoulos method .
the detailed sizes of the crystals are listed in table [ crystalspec ] .
after the crystal surfaces were polished they were wrapped with a teflon reflector and inserted into an oxigen - free electronic ( ofe ) copper cylinder and encapsulated in a nitrogen gas environment .
there is a 12.7-mm - thick quartz - plate window at each end of the cylinder , with optical grease between the crystal and the quartz windows .
a 3@xmath3 pmt is mounted at each end of the cylinder . [ cols="<,^,^,^,^,^",options="header " , ]
the dama experiment has been consistently showing a significant annual modulation with two different experimental arrangements that has persisted over the past 15 years . to check these results
, it is necessary to use the same target material and preferable to have a lower threshold and reduced background levels . achieving a software energy threshold of 1
kev seems feasible because of the high crystal light output and the high quantum efficiency of the new pmts .
background levels can be significantly improved from the current measured background level of @xmath43 counts / kev / kg / day ( at 6 kev ) by growing the crystals from purer nai powder and in a rn - free environment . because the main internal backgrounds are due to and , we are attempting to reduce these contaminations below the background level of 0.2 counts / kev / kg / day for each source , and the prospects of starting with purer powder from sigma aldrich are promising .
in addition to the internal backgrounds , the external backgrounds need to be controlled well below 0.5 counts / kev / kg / day .
low - background , metal - housed pmts with lower radioactivity specifications are commercially available and it is possible to use high efficiency sba photocathodes with these tubes .
we are working closely with the hamamatsu company to develop a pmt that is better suited for a low - background nai(tl ) crystal detector module .
further , we expect a significant reduction in the internal or external backgrounds by the immersion of the nai(tl ) crystal array inside a liquid scintillator box that provides an active veto capability .
a naive simulation shows that @xmath470% of the pmt - initiated backgrounds below 10 kev can be vetoed .
a performance test with a single nai crystal is in progress .
we tested the performance of two large nai(tl ) crystals as part of a program to develop ultra - low - background nai crystals for wimp searches .
we developed selection requirements that are effective for reducing pmt - noise induced background signals .
based on this effort , we achieved a background level of @xmath43 counts / kev / kg / day at 6 kev and a < 2 kev energy threshold .
a number of efforts are being pursued that are aimed at further reduction of these backgrounds .
the successful development of ultra - low - background and low - energy - threshold nai crystals with much reduced pmt background will guarantee a definitive and unambiguous test of the dama experiment s annual modulation signal .
we thank the korea hydro and nuclear power ( khnp ) company for providing the underground laboratory space at yangyang .
this research was funded by grant no .
ibs - r016-d1 and was supported by the basic science research program through the national research foundation of korea ( nrf ) funded by the ministry of education ( nrf-2011 - 35b - c00007 ) .
r. bernabei _ et al . _
( dama collaboration ) , eur .
j. c 56 ( 2008 ) 333 .
r. bernabei _ et al . _
( dama collaboration ) , eur .
j. c 67 ( 2010 ) 39 .
r. bernabei _ et al . _
( dama collaboration ) , eur .
j. c 73 ( 2013 ) 2648 . c. e. aalseth _
_ ( cogent collaboration ) , phys .
( 2011 ) 141301 . c. e. aalseth _
( cogent collaboration ) , phys .
d 88 ( 2013 ) 012002 . c. e. aalseth _ et al . _
( cogent collaboration ) , arxiv:1401.3295 .
r. agnese _ et al .
_ ( cdms collaboration ) , phys .
( 2013 ) 251301 .
( cdex collaboration ) , arxiv:1404.4946 .
_ , arxiv:1407.2238
. g. angloher _
et al . _ ( cresst collaboration ) , arxiv:1407.3146 .
g. plante _ et al .
_ , phys rev .
c 84 ( 2011 ) 045805 .
d. barker and d. m. mei , astropart .
38 ( 2012 ) 1
. y. y. mao , l. e. strigari , and r. h. wechsler , phys .
d 89 ( 2014 ) 063513 .
et al . _ ,
methods phys .
res . , sect .
a 742 ( 2014 ) 187 .
et al . _ ,
j. cherwinka _
( dm - ice collaboration ) , arxiv:1401.4804 .
y. d. kim _ et al .
_ , j. korean .
40 ( 2002 ) 520 ; y. d. kim _ et al .
methods phys .
res . , sect .
a 552 ( 2005 ) 456 .
t. y. kim _ et al .
methods phys .
res . , sect .
a 500 ( 2003 ) 337 ; h. s. lee _ et al . _ ,
methods phys .
res . , sect .
a 571 ( 2007 ) 644 .
h. s. lee _ et al .
_ , ( kims collaboration ) , phys .
99 ( 2007 ) 091301 .
s. c. kim _ et al .
_ , ( kims collaboration ) , phys . rev . lett . 108 ( 2012 ) 181301
. j. j. zhu _ et al .
_ , high energ .
( 2005 ) 8 .
r. bernabei _ et al .
_ , jinst 7 ( 2012 ) p03009
. h. s. lee _ et al . _
( kims collaboration ) , phys .
b 633 ( 2006 ) 201 .
r. bernabei _ et al .
methods phys .
res . , sect .
a , 592 ( 2008 ) 297 . s. agostinelli _
et al . _ ,
methods phys .
res . , sect .
a 506 ( 2003 ) 250 ; j. allison _ et al . _ ,
ieee tran .
53 ( 2006 ) 270 .
c. cuesta _
_ , arxiv:1403.3580
. m. robinson _
. instrum .
methods phys .
res . , sect .
a , 545 ( 2005 ) 225 . | astronomical observations most matter universe is invisible exotic nonrelativistic dark matter @xcite experimental effort nature dark matter remains unknown weakly interacting massive particles ( wimps ) attractive candidates for dark matter supersymmetric models lightest supersymmetric particle ( lsp ) natural candidate for wimp - type dark matter experiments for wimp component evidence wimp - nucleus scattering detecting recoiling nucleus in ultra - sensitive low - background detectors dama experiment searches for annual modulation in detection rate nuclear recoils in low earth s orbital motion through dark - matter halo experiment over 15 years reported positive signal for annual modulation phase consistent with motion earth galactic rest frame statistical significance of dama annual modulation signal reacehed other experiments cogent cresst cdms reported signals due to wimp interactions signals marginal significance at < 3@xmath0 confidence level inconsistencies with null results using similar detectors marginal signals motivated experimental groups to check signals similar techniques higher sensitivity no independent verification of dama signal in experiment same technique donedama signal due to wimp - nucleus scattering subject debate first dama report 15 years ago wimp - nucleon cross sections with limits experiments nuclear xenon100 lux supercdms room for explaining results without conflict differences in detector responses astronomical model for wimp distribution verification of dama signal by independent experiment using similar nai(tl ) crystals mandatory reproducing dama measurement requires independent development ultra - low - background nai(tl crystals crystal - growing company supplied dama crystals produce same for groups anais group developing ultra - low - background nai crystals reproducing dama result dm - ice group reports background measurements crystals no group produced nai(tl crystals with background levels at or below dama experiment korea invisible search collaboration direct wimp searches using ultra - low - background csi(tl ) crystal detectors similar to required research identifying reducing internal backgrounds in csi crystals null results reject wimp - iodine nuclei interactions source dama signal little model dependencedama results due to wimp - sodium nuclei interactions necessary to confirm dama observations with nai(tl ) crystal detectors motivated program to develop ultra - low - background nai(tl crystal detectors with lower background levels higher light yields lower energy threshold than dama experiment identify origin dama modulation signature nai crystals experimental setup for kims csi(tl ) detector measurements at yangyang underground laboratory 700 m earth overburden 2400 m water equivalent includes 12-module array of csi detectors total mass 103.4 kg ) shield 10 cm copper 5 cm polyethylene 15 cm lead 30 cm liquid - scintillator - loaded mineral oil to stop neutrons gamma rays cosmic ray muons each detection module low - background csi(tl ) crystal with photomultiplier tube at each end two nai(tl crystals inside detector array november 2013 installed first nai crystal nai-001 second nai-002 added february 2014 crystals produced by alpha spectra companycrystals grown from nai powder not highest purity initial purification by alpha spectra crystals cylindrical shape cut from 32 inch ingot grown by kyropoulos method detailed sizes listed in table crystalspec after surfaces polished wrapped with teflon reflector inserted into oxigen - free copper cylinder encapsulated in nitrogen gas environment 12.7-mm - thick quartz - plate window at each end cylinder optical grease between crystal quartz 3@xmath3 pmt mounted at each end dama experiment significant annual modulation 15 years same target material lower threshold reduced background levels software energy threshold 1 kev feasible high crystal light output high quantum efficiency of new pmts background levels @xmath43 counts / kev / kg / day growing crystals from purer nai powder rn - free environment to reduce contaminations below 0.2 counts / kev / kg / day starting with purer powder from sigma aldrich external backgrounds below 0.5 counts / kev / kg / day low - background metal - housed pmts with lower radioactivity available use high efficiency sba photocathodes with tubesworking with hamamatsu company develop pmt for low - background nai(tl crystal detector module expect reduction in backgrounds by nai(tl crystal array inside liquid scintillator box active veto capability simulation shows pmt - initiated backgrounds below 10 kev be vetoed performance test with single nai crystal progress tested performance two large nai(tl crystals ultra - low - background nai crystals for wimp searches developed selection requirements for reducing pmt - noise background signals achieved background level of counts / kev / kg / day at 6 kev < 2 kev energy threshold efforts pursued further reduction backgrounds successful development of ultra - low - background low - energy - threshold nai crystals with reduced pmt background test of dama experiment s annual modulation signal thank korea hydro nuclear power for underground laboratory space at yangyang research funded by grant ibs - r016-d1 basic science research program national research foundation korea funded by ministry of education - - c00007 bernabei et al dama collaboration 56 2008 ) 333 67 2010 39 73 2013 ) 26482011 ) 141301 aalseth cogent collaboration . 88 2013 ) 012002 cogent collaboration arxiv:1401.3295 agnese cdms collaboration . 2013 ) 251301 cdex collaboration arxiv:1404.4946 arxiv:1407.2238 angloher cresst collaboration arxiv:1407.3146 plante rev . 84 2011 ) 045805 barker. mei astropart . 38 ( 2012 ) 1 mao strigari h. wechsler phys . d 89 ( 2014 ) 063513 al methods phys . res . sect . a 742 2014 ) 187 cherwinka dm - ice collaboration arxiv:1401.4804 kim korean . 40 ( 2002 ) 520 kim methods . res sect . 552 ( 2005 ) 456 kim sect . 500 ( 2003 ) 337 lee methods phys . res sect . 571 ( 2007 ) 644 kims collaboration phys . 99 ( 2007 ) 091301 kim kims collaboration phys . rev . lett . 108 ( 2012 ) 181301zhu high energ . 2005 ) 8 bernabei jinst 7 2012 ) p03009 lee kims collaboration . b 633 ( 2006 ) 201 bernabei methods phys . sect . a 592 ( 2008 ) 297 agostinelli methods . sect 506 ( 2003 ) 250 allison ieee tran 53 ( 2006 ) 270 cuesta arxiv:1403.3580 robinson methods . sect . a 545 ( 2005 ) 225 | 0.434209 | 2,955 | 1,418 | , Saving $0.1 in GPT-4. |
20,888 | the efficient use of near - ir data in studies of external stellar populations depends on our ability to recognize the nature of the predominant sources of light , and to interprete these findings in terms of age and metallicity .
here we focus on elderly post - starburst populations , with ages of @xmath0yrs .
new models confirm that they are indeed expected to display specific spectral signatures in the near - ir , due to variable m stars of the asymptotic giant branch and to carbon stars .
the signatures depend on age and metallicity .
we summarize the status of current quantitative predictions and emphasize the importance of an empirical calibration of the spectral synthesis models . | near - ir data has been used very often to confirm the presence of an evolved stellar component " in starburst galaxies . the need to study stellar populations
deeply embedded in dust is also frequently invoked to justify near - ir observations .
but what exactly are these cool stellar populations seen between 1 and 3@xmath1 m ? population synthesis models tell us that red supergiants are the main sources @xmath2yrs after a star formation episode ; during the following @xmath3yrs , stars of the upper , thermally pulsing asymptotic giant branch ( tp - agb ) become predominant ; and finally giants of the early agb ( e - agb ) and of the first red giant branch ( rgb ) outweigh other populations .
the efficient use of near - ir data depends on our ability to distinguish these categories from each other and to interprete them in terms of age and metallicity .
very strong co absorption ( 2.3@xmath1 m ) is a clear signature of a red supergiant population ( @xcite initiated the systematic exploration of co in cool stars ; @xcite analysed early observations in starbursts ; @xcite discussed the diagnostic power of the feature and its limitations ) .
but the dependence of the red supergiant population on metallicity is strong @xcite and not reproduced by stellar evolution models unless ad hoc corrections are invoked ( @xcite , @xcite ) .
therefore the threshold above which the co absorption becomes an unambiguous signature of a supergiant population is uncertain . on the other hand , it has recently been claimed that high quality , high spectral resolution near - ir observations allow for _ relative _ age - dating of young post - starburst populations to within @xmath4myr , at least at solar metallicity ( gilbert , this workshop ) .
the focus of our work is on somewhat older post - starburst populations , such as we are likely to find in evolved mergers or e+a " galaxies , i.e. after the most obvious signs of a recent starburst have faded away . between @xmath5 and @xmath3yrs of age , of the order of 50% of the k band light of a coeval population originates from tp - agb stars .
large stellar samples show that essentially all the tp - agb stars are long period variables @xcite .
spectral libraries as well as model atmospheres demonstrate that this pulsation results in deeper h@xmath6o absorption bands than seen in e - agb or rgb stars .
this provides us with a means of identifying strong contributions of oxygen - rich ( e.g. m type ) tp - agb stars .
in addition , the convective dredge - up of carbon - rich material from the stellar core into the envelope progressively turns some tp - agb stars into carbon stars .
those again have unmistakable spectral signatures in the near - ir .
finally , tp - agb stars may become dust enshrouded mid - ir sources , and it has been suggested that the presence of these modifies the location of stellar populations in near - ir versus mid - ir two - colour diagrams @xcite .
all the mentioned processes depend sensitively on the initial stellar mass ( and thus on post - starburst age ) as well as on metallicity .
this was the motivation for our work .
we summarize its status below .
our population synthesis models are based on the code structure of pgase @xcite , the stellar tracks of the padova group " ( @xcite,@xcite ) up to the end of the e - agb , and the spectral library of lejeune et al .
we have completed this set of inputs with synthetic evolutionary tracks for the tp - agb that incorporate the effects of mass loss , dredge - up and envelope burning of dredged - up materials ( @xcite , @xcite ) ; we also use a new library of cool stellar spectra @xcite . various global properties can be determined without invoking spectral libraries , on the basis of the tp - agb tracks alone @xcite .
the values given here are based on our currently prefered internal model parameters ( e.g. mass loss prescription , convective mixing length , dredge - up efficiency ) , which have been tested against statistical properties in nearby stellar populations ( see below ) .
the stars with the longest tp - agb durations ( and thus the largest luminous energy output , integrated along the tp - agb ) are found to have initial masses ( @xmath7 ) between 2 and 2.5m@xmath8 ( i.e. ages of 0.7 to 1.1gyr ) ; stars with @xmath7 in the @xmath9m@xmath8 range last @xmath10yrs or more in that phase .
mass loss has the strongest direct influence on these durations , and itself depends on various other model parameters through the instantaneous masses , luminosities and radii . as the tp - agb stars are the coolest luminous stars present , the duration of that phase essentially controls the ( v - k ) colour of the stellar population @xcite .
the fraction of the tp - agb duration spent as a carbon star is found to be maximal for objects with initial masses between 2.5 and 3m@xmath8 ( i.e. ages of 0.4 to 0.8gyr ) , where it typically amounts @xmath11 at solar metallicity and about @xmath12 at the metallicity of the large magellanic cloud ( fig.[fractions.fig ] ) . due to strong mass loss
, stars above @xmath13m@xmath8 have relatively short tp - agb durations ( @xmath14yrs ) and are embedded in dust for a significant fraction of this time .
this reduces the predicted contribution of tp - agb stars to the light at ages of @xmath15gyr , when compared to our earlier models @xcite .
as mentioned , our trust in the model prediction is based on a variety of tests .
for instance , the new evolutionary models for the tp - agb successfully reproduce the empirical initial to final mass relation for stars in the solar neighbourhood .
the predicted contributions of c - stars to the near - ir light of intermediate age populations are consistent with those observed in the magellanic clouds ( e.g. @xcite ) .
one of us ( m.m . ) has coupled the tp - agb models with chemical evolution models for galaxies ( including the yields of intermediate mass stars as given by the new tracks ) .
the tight relations observed in the local group galaxies between metallicity and the mean luminosity of c - stars , or between metallicity and the c - star / m - star number ratio can be reproduced very naturally .
these results will be discussed in detail elsewhere .
stellar libraries are required for the prediction of near - ir colours or spectral features .
based on the extremely diverse agb star spectra of @xcite ( @xmath16 between 1 and 2.5@xmath1 m ) , we have constructed a regular sequence of averages that can be more conveniently coupled with population synthesis models @xcite . as a result , we predict that the signatures of upper agb stars are seen in the integrated light of a stellar populations between @xmath17 and @xmath18gyr after a starburst ( @xcite , @xcite ) .
the main signatures are extremely broad , and best detected with low resolution data with very wide spectral coverage ( the k band spectrum alone is not sufficient ) .
as the features reach accross telluric absorption bands , observations at good ir sites are essential .
the features are due to h@xmath6o if the tp - agb stars are o - rich , to c@xmath6 and cn if they are c - rich .
narrow band indices can be defined that detect tp - agb stars independently of their chemical nature , and others that then separate the two classes @xcite .
photometry with an accuracy better than 5% must be aimed for .
we are exploring the feasibility of surveys with the available filters on the nicmos camera onboard the hubble space telescope .
once tp - agb evolutionary tracks have been constructed and tested , only a part of the work is done : associating a spectrum with a given point of the theoretical hr diagram is difficult , as illustrated in fig.[hr.fig ] .
individual stars vary significantly in luminosity and colour temperature .
while optical colours and temperature indicators such as ( v - k ) correlate well , the near - ir properties ( colours , strength of the h@xmath6o or co bands ) are very dispersed at a given colour temperature .
in addition , no model atmospheres are available to reliably relate an instantaneous colour temperature to the effective temperature of the simple static stars on which evolutionary tracks are based .
the regularity of the spectral sequence obtained by @xcite gives some confidence in the evolution of the spectra with temperature ; but only empirical calibrations , based on stellar populations with ages and metallicities known from optical observations , can constrain absolute relationships . despite the more numerous near - ir instruments now available on large telescopes , appropriate targets for the model calibration remain hard to find .
the lmc and m33 clusters , of which several have adequate ages , tend to be too small : they contain a handful of tp - agb stars and their integrated properties are affected by the stochastic nature of this subpopulation . in @xcite , it was shown that stellar populations with total masses of @xmath19m@xmath8 are needed to obtain significant constraints on the tp - agb models .
massive stellar clusters around merging galaxies are well suited , but only a few are bright enough for near - ir spectroscopy .
we have observed the 500myr old cluster w3 in ngc7252 with sofi ( ntt , eso ) , between 1 and 2.4@xmath1 m .
illustrative adjustments with models are shown in fig.[fits.fig ] .
the figures illustrate the need for broad wavelength coverage and excellent relative calibrations of the various atmospheric windows ( with sofi the overlap between grisms ensures this ) .
our preliminary conclusions are the following .
the data reject models that have few c - stars _ and _ use a cool effective temperature scale for the tp - agb spectra : the water bands in those are too strong .
this suggests that the metallicity is not more than solar and that the coolest of the mira temperature scales in the literature is inappropriate for our purpose .
good fits to the spectral shape are obtained at solar and lmc metallicities , with extinction values compatible with optical estimates @xcite .
more calibration points will be required , but this first test already restricts the allowed range of parameters and gives us confidence in the new tool .
based on new synthetic evolution models for the tp - agb and on a purpose designed library of stellar spectra , we have shown that near - ir spectra of post - starburst populations indeed carry information about post - starburst age and metallicity .
elderly post - starbursts with ages of @xmath5 to @xmath3yrs can be distinguished from younger or older ones , using the specific spectral signatures of o - rich and c - rich tp - agb stars .
the chemical nature of the predominant tp - agb stars is a metallicity indicator .
the population of dust - obscured tp - agb stars depends both on age and metallicity .
we have pointed out that the relevant spectral features can be detected at low spectral resolution as long as a broad continuous spectral coverage is obtained .
high quality narrow band photometry can also be used .
the quantitative prediction of the strength of the near - ir features must be based on empirical calibrations .
we have described the current status of these .
applications of this work include studies of the duration of star formation episodes in starburst galaxies or mergers , of the survival timescale of young starburst clusters in the galaxy environment , and of the nature of
e+a " galaxies .
first observations of
e+a " galaxies will be available to us soon .
baldwin , j.a .
frogel , s.e .
persson : apj * 184 * , 427 ( 1973 ) a. bressan , f. fagotto , g. bertelli , c. chiosi : a&as * 100 * , 647 ( 1993 ) a. bressan , g.l .
granato , l. silva : a&a * 332 * , 135 ( 1998 ) f. fagotto , a. bressan , g. bertelli , c. chiosi : a&as * 105 * , 29 ( 1994 ) m. fioc , b. rocca - volmerange : a&a ... l. girardi , g. bertelli : mnras * 300 * , 533 ( 1998 ) a. lanon , c. boily ( eds . ) : _ massive stellar clusters _ , asp conf . ser .
* 211 * ( asp , san francisco 2000 ) a. lanon , m. mouhcine . in preparation ( 2001 )
a. lanon , b. rocca - volmerange : new astron .
* 1 * , 215 ( 1996 ) a. lanon , m. mouhcine , m. fioc , d. silva : a&a * 344 * , l21 ( 1999 ) a. lanon , p.g .
wood : a&as * 146 * , 217 ( 2000 ) n. langer , a. maeder : a&a * 295 * , 685 ( 1995 ) t. lejeune , f. cuisinier , r. buser : a&as * 125 * , 229 ( 1997 ) p. massey : aj * 501 * , 153 ( 1998 ) m. mouhcine , a. lanon : in @xcite , p. 144
( 2000 ) m. mouhcine , a. lanon . in preparation ( 2001 )
l. origlia , j.d .
goldader , c. leitherer , d. schaerer , e. oliva : apj * 514 * , 96 ( 1999 ) s.e .
persson , m. aaronson , j.g .
cohen , j.a .
frogel , k. matthews : apj * 266 * , 105 ( 1983 ) g.h .
rieke , m.j .
lebofsky , r.i .
thompson , f.j .
low , a.t .
tokunaga : apj * 238 * , 24 ( 1980 ) f. schweizer , p. seitzer : aj * 116 * , 2206 ( 1998 ) p.g .
wood : in _ agb stars _ , iau symp . | near - ir data evolved stellar component in starburst galaxies need study stellar populations in dust observations cool stellar populations between 1 and 3@xmath1 m ? synthesis models red supergiants main sources @xmath2yrs after star formation following @xmath3yrs stars of upper giant branch tp - agb ) predominant giants of early agb ( e - agb ) first red giant branch ( rgb ) outweigh populations efficient use of near - ir data depends on categories age and metallicity strong co absorption ( 2.3@xmath1 m ) signature of red supergiant population exploration in cool stars red supergiant population on metallicity strong not reproduced by stellar evolution models unless ad hoc corrections invoked threshold co absorption signature of supergiant population uncertain high quality high spectral resolution near - ir observations allow relative age - dating of young post - starburst populations to within @xmath4myr solar metallicity focus work on older post - starburst populations in evolved mergers or e+a galaxiessigns recent starburst faded between @xmath5 and @xmath3yrs 50% k band light coeval population originates from tp - agb stars large stellar samples show tp - agb stars are long period variables pulsation results in deeper h@xmath6o absorption bands than e - agb or rgb stars identifying strong contributions oxygen - rich tp - agb stars convective dredge - up of carbon - rich material turns tp - agb stars into carbon stars spectral signatures in near - ir tp - agb stars become dust enshrouded mid - ir sources modifies location stellar populations in near - ir mid - ir diagrams processes depend on initial stellar mass post - starburst age metallicity motivation for work population synthesis models based on code structure of pgase stellar tracks padova group spectral library of lejeune et al completed with synthetic evolutionary tracks for tp - agb effects mass loss dredge - up envelope burning materials new library of cool stellar spectraglobal properties determined without spectral libraries tp - agb tracks @xcite values based on model parameters mass loss convective mixing length dredge - up efficiency tested against stellar populations stars with longest tp - agb durations largest luminous energy output initial masses @xmath7 between 2 and 2.5m@xmath8 ages 0.7 to 1.1gyr stars with @xmath7 in @xmath9m@xmath8 range last @xmath10yrs more mass loss durations depends on model parameters tp - agb stars coolest luminous duration controls colour stellar population @xcite fraction tp - agb duration spent as carbon star maximal for objects initial masses between 2.5 and 3m@xmath8 0.4 to 0.8gyr amounts @xmath11 at solar metallicity @xmath12 at metallicity large magellanic cloud strong mass loss stars above @xmath13m@xmath8 have short tp - agb durations @xmath14yrs embedded in dust reduces predicted contribution of tp - agb stars to light at ages @xmath15gyr trust in model prediction based on testsnew evolutionary models for tp - agb reproduce initial to final mass relation for stars in solar neighbourhood contributions of c - stars to near - ir light intermediate age populations consistent with magellanic clouds coupled tp - agb models with chemical evolution models galaxies intermediate mass stars tight relations between metallicity mean luminosity of c - stars c - star / m - star number ratio reproduced discussed stellar libraries required for near - ir colours spectral features diverse agb star spectra @xcite constructed regular sequence of averages with population synthesis models @xcite predict signatures of upper agb stars seen in integrated light stellar populations between @xmath17 and @xmath18gyr after starburst main signatures broad detected with low resolution data wide spectral coverage features reach accross telluric absorption bands observations at good ir sites essential features due to h@xmath6o o - rich c@xmath6 and cn c - rich narrow band indices tp - agb stars two classes photometry with accuracy better than 5% aimed forexploring feasibility surveys with filters on nicmos camera hubble space telescope tp - agb evolutionary tracks constructed tested work done associating spectrum with point theoretical hr diagram difficult illustrated in fig. stars vary in luminosity colour temperature optical colours temperature indicators correlate near - ir properties dispersed at temperature no model atmospheres relate instantaneous colour temperature to temperature static stars regularity spectral sequence by @xcite gives confidence in evolution with temperature empirical calibrations on stellar populations constrain relationships near - ir instruments on telescopes targets for model calibration hard to find lmc and m33 clusters too small contain handful of tp - agb stars properties affected by stochastic subpopulation @xcite stellar populations with masses of @xmath19m@xmath8 needed constraints on tp - agb models massive stellar clusters around merging galaxies suited few bright enough for near - ir spectroscopy observed 500myr old cluster w3 in ngc7252 with sofi between 1 and 2.4@xmath1 m adjustments with models in fig.figures illustrate broad wavelength coverage calibrations atmospheric windows overlap preliminary conclusions data reject models few c - stars cool temperature scale tp - agb spectra water bands strong suggests metallicity not more than solar coolest mira temperature scales inappropriate good fits spectral shape at solar lmc metallicities extinction values compatible with optical estimates more calibration points required first test restricts range parameters confidence in new tool synthetic evolution models tp - agb stellar spectra near - ir spectra post - starburst carry information post - starburst age metallicity elderly post - starbursts ages to distinguished younger using spectral signatures o - c - rich tp - agb stars chemical nature predominant tp - agb stars metallicity indicator population of dust - obscured tp stars depends on age metallicity relevant spectral features detected at low spectral resolution broad continuous spectral coverage high quality narrow band photometry used prediction strength near - ir features empirical calibrations described current status applications include star formation episodes starburst galaxies survival timescale young starburst clusters nature of e+a galaxiesobservations galaxies soon baldwin frogel persson apj 184 427 ( 1973 ) bressan fagotto bertelli chiosi a&as 100 * 647 1993 ) bressan granato silva a&a 332 135 ( 1998 ) fagotto bressan bertelli chiosi 105 29 ( 1994 ) fioc rocca - volmerange girardi bertelli 300 * 533 ( 1998 ) lanon boily massive stellar clusters 211 * san francisco 2000 lanon mouhcine preparation 2001 ) lanon rocca - volmerange astron 1 215 ( 1996 ) lanon mouhcine fioc silva a&a 344 * l21 ( 1999 ) lanon wood a&as 146 * 217 ( 2000 ) langer maeder a&a 295 * 685 ( 1995 ) lejeune cuisinier buser a&as * 125 * 229 ( 1997 ) massey * 501 * 153 ( 1998 ) mouhcine lanon @xcite 144 ( 2000 preparation 2001 ) origlia goldader leitherer schaereroliva apj * 514 * 96 ( 1999 ) persson aaronson cohen frogel matthews apj 266 * 105 ( 1983 ) rieke lebofsky thompson low tokunaga apj 238 * 24 ( 1980 ) seitzer 116 * 2206 ( 1998 ) wood agb stars symp | 0.499814 | 3,177 | 1,604 | , Saving $0.1 in GPT-4. |
20,900 | power grids and other delivery networks has been attracted some attention by the network literature last decades . despite the power grids dynamics
has been controlled by computer systems and human operators , the static features of this type of network can be studied and analyzed .
the topology of the brazilian power grid ( bpg ) was studied in this work .
we obtained the spatial structure of the bpg from the ons ( electric system s national operator ) , consisting of high - voltage transmission lines , generating stations and substations .
the local low - voltage substations and local power delivery as well the dynamic features of the network were neglected .
we analyze the complex network of the bpg and identify the main topological information , such as the mean degree , the degree distribution , the network size and the clustering coefficient to caracterize the complex network .
we also detected the critical locations on the network and , therefore , the more susceptible points to lead to a cascading failure and even to a blackouts .
surprisely , due to the characteristic of the topology and physical structure of the network , we show that the bpg is resilient against random failures , since the random removal of links does not affect significantly the size of the largest cluster .
we observe that when a fraction of the links are randomly removed , the network may disintegrates into smaller and disconnected parts , however , the power grid largest component remains connected .
we believe that the even a static study of the network topology can help to identify the critical situations and also prevent failures and possible blackouts on the network .
keywords : power grid , complex systems , complex networks , resilience analysis , computer modeling and simulation . | power grids are part of what have been called technological networks , the physical insfrastructure network for the transportation of energy and information @xcite .
the internet and telephone networks are some examples of these type of networks .
this field has received some attention from the researches and the correspondent literature last decades mainly due to the problems caused by the possibility of large scale failures leading to major blackouts @xcite . despite the fact that the power grids dynamics has been controlled by computer systems and human operators @xcite , the static features of this type of networks
can be studied and provide valuable information for private or public investments and prevention of failures @xcite .
the topology of the brazilian power grid ( bpg ) was studied in this work .
we obtained the spatial structure of the bpg from the brazilian electric system s operator , ons , consisting of the high voltage transmission lines , generating stations and substations .
the local low - voltage substations were neglected in this work .
we also neglected the dynamic features of the power grid , such as the electromagnetic processes and human interference by controlling the energy flux on the network .
we analyze the bpg as a complex network in which all the stations and substations were represented by nodes and the transmission lines were represented by links among the nodes . by identifying the main topological information and use these features to find out the network s weakness , one can give indications for improvements of the bpg in order to prevent failures and blackouts .
this paper is organized as follows . in sec . 2 we present our methodology and the details for data analysis . in sec .
3 we describe our main results and discussions . finally , in sec . 4
, we summarize the results and present our main conclusions .
the bpg data was obtained from the ons database , regarding to the 2015 update .
the network analysis was performed by using the methodos of networks theory , focusing on some topological properties , namely the mean degree and the degree distribution , betweenness centrality , transitivity and the connectivity of the network .
the degree @xmath0 , is the number of links attached to a node @xmath1 , in a total of @xmath2 nodes , given by the number of transmission lines connected to the station or substation . on the other hand ,
the mean degree of the network is a global measure , given by @xmath3 the degree distribution @xmath4 is the probability that a node chosen uniformly at random within the network has a degree @xmath5 : @xmath6 where @xmath2 is the total number of nodes in the network and @xmath7 is the average number ( or fraction ) of nodes with degree @xmath5 on the network @xcite .
the distribution is given by the equation [ equation_dd ] and , once it is known , much information can be obtained by the calculation of moments of this distribution .
the @xmath8-th moment of the distribution is @xmath9 the first moment @xmath10 is the mean degree , while the second moment is a mesure of the degree fluctuations of the distribution .
if @xmath11 diverges , structure and function dramatically changes in the network , in contrast to those for finite @xmath10 @xcite .
an exponential degree distribution has the form @xmath12 where @xmath13 is an arbitrary constant .
networks with exponential degree distributions were reported in some real - world situations such as the power grids @xcite .
we also can visualize the network structure with the corresponding positions to the generating stations and substations and the transmission lines connecting them .
as one can see in the fig .
[ fig1 ] , we show the brazilian power grid and it s corresponding complex network . this picture was created by using the large network analysis software pajek@xmath14@xcite . 0.2 cm .,title="fig : " ] + .,title="fig : " ] + as mentioned before
, in this work we do not take into account the electromagnetic processes , computer controlling and human interference in the energy flux of the network .
despite the fact that we neglect the dynamic features of the network , our analysis perhaps it is not so naive after all .
as we will show in next section , the bpg is strongly dependent of the topological structure . in case of some large structural failures ,
the network can be lead to cascading failures that may not be corrected by human interference or computer systems .
we believe that the this approach can provide powerful information about the bpg s vulnerability .
the bpg has evolved greatly in the last decades but so far does not show itself as a complex as other power grids worldwide , in comparison to countries with approximately same size or population than brazil . only for sake of comparison ,
the usa s power grid is represented by a network of 14099 nodes and 19657 links among them @xcite .
the main results of the characterization of the bpg are shown in the table 1 .
one can see that these measures represent the most basic topological information of the network .
[ table ] @lcc characteristic & numerical value ( brazil ) & numerical value ( us ) + mean degree & 6.487 & 1.394 + number of nodes & 230 & 14099 + number of links & 1492 & 19657 + clustering coefficient & 0.174 & 0.080 + betweenness centrality & 0.438 & - + + the degree distribution is a very important network measure .
much information about a network is related to degree distribution .
networks with heavy tail scale free degree distribution have been investigated over a decade . a network with a power - law degree distribution , @xmath15 with @xmath16 , for instance , is expected to be resilient to a random removal of links , _
i.e. _ resilient against random failures of the grid @xcite .
numerous real world networks , including the internet , have a power - law degree distribuition with @xmath17 and in all of them were observed the resilience phenomena @xcite .
networks may have the degree distribution with exponent even greater the 3 .
for the case where @xmath18 , this type of networks are kwnown as ultra - resilient against random failures . in this work
we use the cumulative degree distribution , @xmath19 which is the probability that the degree is greater or equal to k. by using the cumulative degree distribution to represent the data , we reduce the noise in the tail of the distribution . as one can see in fig .
[ fig2 ] , the cumulative degree distribution is shown and it has the exponential form . , of the brazilian power grid . as shown in the picture of semilogarithmic scale
, the degree distribution has an exponential form . ]
another important measure is the mean clustering coefficient of the network , @xmath20 .
the concept of clustering reflects how the first neighbors of a node are connected to each other .
this feature is a statistic of the number of multiple connections among the nodes in the network . in other words ,
clustering coefficient refers to the statistics of the number of triangles ( loops of length 3 ) in the network .
the clustering coefficient of the entire network , or the _ mean clustering coefficient _ , is the average of the local clustering coefficient over all nodes : @xmath21 and may vary between 0 and 1 . in the bpg we found @xmath22 , a very small value for the clustering coefficient .
this indicate that the bpg is , in average , poorly connected and highly susceptible to the formation of isolated clusters in the network . in order to investigate the connectivity of the network and how it should be affected by random removal of links and nodes we studied the number of connected components on the network , @xmath13 and also the size of the largest cluster on the network , @xmath23
. as one can see by our simulations , both measurements of connectivity presents different and complementary results .
for the first measurement , when all the links are present , the fraction of removed links , @xmath24 , is equal zero and the network is a single connected component . for the case , @xmath25 . as the number of removed links increases , the network is divided into disconnected parts and @xmath26 .
when the fraction of removed links , @xmath27 , @xmath28 .
we performed simulations to investigate the network connectivity .
by randomly removal of links from the network , we calculate the number of connected components on the newtork , for different realizations .
our results are shown in the figure [ fig3 ] , where one can see on the left axis the plot 1/@xmath13 versus @xmath24 .
we show that for a small fraction of links removed , @xmath29 , the brazilian powed grid turn into a set of several disconnected parts ( small value of @xmath30 ) .
the power grid rupture is significantly important for the brazilian case , since the number of generating stations is relatively small . and the size of the largest cluster on the network , @xmath23 . in the plot
we show on the left axis the 1/@xmath13 versus @xmath24 and , on the rigth axis , s versus @xmath24 .
the results were obtained averaging 100 different realizations .
this result suggest that even the network turn into a set of disconnected parts , for small @xmath24 , the brazilian power grid is resilient against random failures . ]
the second connectivity measurement that were studied is @xmath23 .
as one can see on the right axis of the figure [ fig3 ] , the decaing of @xmath23 is not as fast as @xmath30 . for small values of @xmath24 ,
@xmath23 is almost linear , following by an exponential regime .
as shown , even for small values of @xmath24 , the network turn into a set of several disconnected parts
. however , most of these disconnected parts are isolated nodes , not affecting substantially the connected component of the network . only for larger values of @xmath24
the size of largest cluster @xmath23 shows a rapidly decreasing .
this result show that the topological structure of the brazilian power grid prove itself , in some sense , resilient against random failues . analyzing the bpg
one can see that there are few nodes playing an important role by keeping the network connected .
the most important ( and susceptible ) nodes are shown in figure [ fig1 ] as two black circles .
these points consists in large substation and hidrelectric power plants , called _ furnas centrais eltricas_. these critical locations on the network make the connection between the south - southeast and the center of the country and then to the north and northeast regions .
the importance of the nodes are due to the failures in these major electric complex will lead the bpg system to a cascading failure and even to a major blackouts .
in this work we studied and analyzed the topology of the brazilian power grid ( bpg ) . despite the fact that the bpg is comparatively much smaller than power grids from other countries , by analyzing its spatial structure , we show that the bpg is poorly connected and highly dependent on a few nodes to maintenance the network s connected component .
we calculated and summarize most basic topological information of the bpg .
the degree distribution of the network has an exponential form and , according to the newtork literature , are not resilient against random failures .
we also simulate random failures on the network by removing links at random .
our results show that with a small fraction of missing connections ( @xmath29 ) , the network turns into a set of several disconnected parts , leading parts of the country isolated .
however , these isolated parts of the network does not represent the larger part of the power grid .
we show that the random removing of links does not affect significantly , for small values of @xmath24 , the size of the largest cluster on the network , @xmath23 , keeping the most of the country connected by power grid . in conclusion , even with a static analysis of the network
, one can see the main topological information and it s weakness . to improve the network and become more resilient against random failues ,
bpg need some improvements regarding to the number of connections to prevent failures and future blackouts .
the authors would like to thank the federal university of southern bahia , ufsb , for the financial support by project 005/2015-prosis . | power grids part of technological networks for transportation of energy information internet telephone networks examples these field received attention from due to problems large scale failures to major blackouts power grids dynamics controlled by computer systems human operators static features can be studied provide valuable information for investments prevention of failures topology of brazilian power grid ( bpg ) studied obtained spatial structure bpg from brazilian electric system s operator ons high voltage transmission lines generating stations substations local low - voltage substations neglected neglected dynamic features electromagnetic processes human interference controlling energy flux analyze bpg as complex network stations substations represented by nodes transmission lines by links among nodes identifying main topological information to network s weakness can give indications for improvements bpg prevent failures blackouts paper organized in sec . 2 methodology details for data analysis sec. 3 describe main results discussions sec. 4 summarize results present main conclusions bpg data obtained from ons database 2015 updatenetwork analysis performed networks theory focusing on topological properties mean degree degree distribution centrality transitivity connectivity degree @xmath0 is number links attached to node @xmath1 total @xmath2 nodes given by number transmission lines connected to station substation mean degree is global measure given by @xmath3 degree distribution @xmath4 probability node chosen random has degree @xmath5 : @xmath6 @xmath2 total number nodes @xmath7 average number nodes with degree @xmath5 network distribution given by equation [ equation_dd ] information obtained by calculation of moments distribution @xmath8-th moment is @xmath9 first moment @xmath10 is mean degree second moment degree fluctuations if @xmath11 diverges structure function changes contrast to finite @xmath10 @xcite exponential degree distribution form @xmath12 @xmath13 arbitrary constant networks with exponential degree distributions reported in real - world situations power grids visualize network structure with positions generating stations substations transmission lines fig brazilian power grid complex network picture created large network analysis software pajek@xmath14@xcite,title="fig : " work electromagnetic processes computer controlling human interference in energy flux network neglect dynamic features network analysis not naive bpg dependent of topological structure large structural failures network can lead to cascading failures not corrected by human interference or computer systems approach can information about bpg s vulnerability bpg evolved as complex as other power grids comparison to countries same size population brazil usa s power grid represented by 14099 nodes 19657 links main results characterization bpg shown in table 1 measures represent basic topological information network table ] @lcc characteristic numerical value ( brazil ) us ) mean degree 6.487 1.394 number of nodes 230 14099 number of links 1492 19657 clustering coefficient 0.174 0.080 betweenness centrality 0.438 degree distribution important network measure information network related to degree distribution networks with heavy tail scale free degree distribution investigated over a decade network with power - law degree distribution @xmath15 expected resilient to random removal of links random failures real world networks including internet have power - law degree distribuition with @xmath17 observed resilience phenomenanetworks have degree distribution with exponent greater 3 @xmath18 networks ultra - resilient against random failures use cumulative degree distribution , @xmath19 probability degree greater or equal to k reduce noise in tail distribution in fig . fig2 cumulative degree distribution shown exponential form brazilian power grid semilogarithmic scale degree distribution exponential form important measure mean clustering coefficient network , @xmath20 clustering reflects first neighbors node connected statistic of number multiple connections among nodes clustering coefficient refers to statistics number of triangles ( loops length 3 ) in network clustering coefficient network mean clustering coefficient average local clustering coefficient over all nodes @xmath21 vary between 0 and 1 in bpg found @xmath22 small value for clustering coefficient bpg poorly connected susceptible to formation isolated clusters investigate connectivity network by random removal of links nodes studied number of connected components network @xmath13 size of largest cluster @xmath23 both measurements connectivity presents different complementary resultsfirst measurement all links present fraction removed links @xmath24 equal zero network single connected component case @xmath25 removed links increases network divided into disconnected parts @xmath26 fraction removed links @xmath27 @xmath28 performed simulations network connectivity randomly removal links calculate number connected components newtork different realizations results shown figure [ fig3 ] plot 1/@xmath13 versus @xmath24 small fraction of links removed @xmath29 brazilian grid several disconnected parts small value @xmath30 ) power grid rupture important brazilian case generating stations small size largest cluster network @xmath23 plot left axis 1/@xmath13 versus @xmath24 s versus @xmath24 results obtained averaging 100 realizations network disconnected parts small @xmath24 brazilian power grid resilient against random failures second connectivity measurement @xmath23 decaing @xmath23 not as fast as @xmath30 small values @xmath24 @xmath23 almost linear exponential regime small values @xmath24 network several disconnected partsmost disconnected parts are isolated nodes not affecting connected component network for larger values of @xmath24 size of largest cluster @xmath23 rapidly decreasing topological structure brazilian power grid resilient against random failues bpg few nodes important role keeping network connected most important susceptible nodes shown in figure fig1 ] as two black circles points large substation hidrelectric power plants _ furnas centrais eltricas_ critical locations connection between south - southeast center country north and northeast regions importance failures in major electric complex lead to cascading failure major blackouts studied analyzed topology brazilian power grid bpg bpg smaller than other poorly connected dependent on few nodes connected component calculated basic topological information bpg degree distribution network has exponential form not resilient against random failures simulate random failures by removing links at random results show with small fraction of missing connections ( @xmath29 ) network turns into several disconnected parts leading parts country isolated isolated parts not represent larger part of power gridshow random removing links affect small values @xmath24 size largest cluster @xmath23 keeping most country connected by power grid static analysis network see main topological information weakness improve network resilient against random failues bpg need improvements number connections prevent failures future blackouts authors thank federal university of southern bahia financial support project 005/2015-prosis | 0.429011 | 2,648 | 1,308 | , Saving $0.1 in GPT-4. |
20,904 | gravity is a second generation instrument for the vlt interferometer , designed to enhance the near - infrared astrometric and spectro - imaging capabilities of vlti . combining beams from four telescopes
, gravity will provide an astrometric precision of order 10 micro - arcseconds , imaging resolution of 4 milli - arcseconds , and low and medium resolution spectro - interferometry , pushing its performance far beyond current infrared interferometric capabilities . to maximise the performance of gravity
, adaptive optics correction will be implemented at each of the vlt unit telescopes to correct for the effects of atmospheric turbulence . to achieve this ,
the gravity project includes a development programme for four new wavefront sensors ( wfs ) and nir - optimized real time control system .
these devices will enable closed - loop adaptive correction at the four unit telescopes in the range 1.4 - 2.4 @xmath0 m .
this is crucially important for an efficient adaptive optics implementation in regions where optically bright references sources are scarce , such as the galactic centre .
we present here the design of the gravity wavefront sensors and give an overview of the expected adaptive optics performance under typical observing conditions .
benefiting from newly developed selex / eso saphira electron avalanche photodiode ( eapd ) detectors providing fast readout with low noise in the near - infrared , the ao systems are expected to achieve residual wavefront errors of @xmath1400 nm at an operating frequency of 500 hz . | gravity@xcite is a second generation 4-beam combiner instrument for eso s very large telescope interferometer ( vlti ) at paranal observatory .
operating in the near - infrared ( nir ) k - band ( 2.2 @xmath0 m ) , it will provide of order 10-micro - arcsec astrometric precision and phase - referenced imaging with 4 milli - arcsec resolution . with fibre - fed integrated optics , wavefront sensors , fringe tracker , beam stabilisation and a novel metrology concept , gravity s performance will bring a dramatic improvement on current nir interferometric instrumentation in sensitivity as well as astrometric precision .
first astronomical light at vlti is planned for 2014 .
key science goals for gravity include : * probing the physics in the immediate vicinity of the galactic centre black hole@xcite ; * detecting and measuring the masses of black holes in massive star clusters ; * study the details of mass accretion and jets in young stellar objects ; and * characterize the detailed motions of exoplanets , binary stars and young stellar discs .
crucial to achieving its ambitious science goals is the implementation of adaptive optics ( ao ) correction to provide seeing - improved beams to the instrument .
the gravity project includes a development programme for 4 new wavefront sensors ( wfs ) for the vlt unit telescopes , which will allow the measurement of the atmospheric turbulence correction in the nir h and k bands ( 1.4 - 2.4 @xmath0 m ) .
these wfs will be installed in the coud space for each ut , and interface directly with the existing multiple application curvature adaptive optics ( macao ) systems at vlt . in this paper
, we present an overview of the design and development of these wfs units and their real time control , as well as the expected on - sky performance in the various operational modes of the ao system .
the following sections will provide basic descriptions of the gravity instrument , its subsystems and operational concept , and the ao architecture for the instrument .
the gravity instrument consists of three major components : the beam - combiner instrument , the laser metrology system , and the wavefront sensors ( see fig .
[ fig : grav_diagram ] ) . in the beam combiner instrument
, the four input beams from the unit or auxiliary telescopes are combined to produce interferograms from six baselines simultaneously .
fringe tracking is implemented internally to the instrument to compensate in real time for longitudinal vibrations and atmospheric piston@xcite .
the input beams are fed via optical fibres into two integrated optics beam combiners@xcite .
the fringes of both science and fringe tracking targets are spectrally dispersed by the spectrometers inside the beam combiner cryostat , with the science spectrometer providing spectral resolutions of r@xmath220 , @xmath2500 or @xmath24000@xcite .
the fringe tracking spectrometer provides low - resolution dispersion only .
the laser metrology system tracks a reference beam through the observatory , allowing for a precise measurement of the optical path .
when observing with the vlt uts , gravity s ao system will use parts of the existing adaptive optics ( ao ) hardware to correct for atmospheric turbulence effects .
wavefront sensing will however be performed with dedicated wfs operating at nir wavelengths , interfacing with the macao dms already present at the uts .
light is fed to the wfs from the telescopes via the star separator systems ( sts)@xcite , installed at the coud foci of each ut as part of the prima instrument , which can pick up 2 stars anywhere inside a 2 field .
the top - level functional requirements for the wfs are : * provide real - time processing of wavefront distortions ; * allow for on - axis and off - axis wavefront sensing ; * provide sensing in the h and k bands ( 1.45 - 2.45 @xmath0 m ) ; * allow use by other vlti instruments ; * interface to the existing macao deformable mirrors . in on - axis mode ,
the science target itself is used as ao reference source , in this case a beam splitter redirects @xmath250% of the light to the wfs , with the rest passing to the vlti tunnels towards the gravity beam combiner . in off - axis mode , where a dedicated reference source is used
, the wfs receives all the light from the ao reference .
performance requirements for the wfs are as follows : * mk=7 : strehl 25% for 7 off - axis ao reference source at 30@xmath3zenith angle * mk=7 : strehl 35% for on - axis ao reference source at 30@xmath3zenith angle
* mk=10 : strehl 10% on - axis at zenith * rms tip / tilt at the sts in the coud room of @xmath4 10 mas on sky for mk = 7 , @xmath4 20 mas on sky for mk = 10 * piston fluctuations of @xmath4 25 nm rms over 48 ms , @xmath4 125 nm rms over 290 ms , @xmath4 2000 nm rms over 10 min .
the main new hardware under development for the ciao system is the nir wavefront sensor . in addition
, a new real time control system will be implemented for the operation of the ao system .
this section describes the optical and cryo - mechanical design of the full wfs unit , as well as the control architecture and detector characteristics .
the wavefront sensors follow a straightforward shack - hartmann design .
the incoming light is passed through a lenslet array , with each spot refocused onto the detector , from where the distorted wavefront is characterized .
each wfs unit comprises a warm and cold assembly .
the warm optics , shown in fig . [ fig : wfs_warm ] , relay the light from the telescope and the sts units to the cryostat .
the entrance into the wfs is the ao mode selector consisting of a dichroic and three flat mirrors mounted onto a translation stage , which chooses one of three operational modes for the wfs : ( i ) on - axis mode , splitting the beam 50/50 between the wfs and the vlti laboratory ; ( ii ) off - axis mode , 100% of the light into the wfs , and ( iii ) visible ao mode , using the existing visible - light macao system .
the light is subsequently folded by parabolic and flat mirrors to the pupil derotator , which maintains pupil orientation throughout an observation . the parabolic mirror
will be mounted in a motorized gimbal for optical alignment and focus adjustment .
the derotator follows a simple k - mirror design in protected gold - coated aluminium .
the cold optics are kept at inside a bath cryostat ( see fig . [
fig : wfs_cold ] ) , the cooling concept is described in section [ sec : wfs_cryomech ] . inside the cryostat ,
a field lens compensates for pupil motion using a piezoelectric actuator .
after this , the light is relayed to the lenslet array via a filter wheel , holding bandpass and neutral density filters .
the lenslet array focuses 9 x 9 spots onto the detector from 192 @xmath0 m diameter micro - lenslets , each with a focal length of 2.09 mm .
further details of the optical design are presented by yang et al.@xcite in these proceedings .
table [ tab : selex_req ] summarises the key requirements for detector performance for the wfs . the gravity wfs will for the first time use recently developed high - speed ir electron avalanche photodiode ( eapd ) arrays , developed by eso and selex - galileo .
the selex / eso saphira arrays have been shown to reach read noise levels of @xmath43 e- and quantum efficiency ( qe ) of @xmath240% at 80k@xcite .
the arrays measure 320 @xmath5 256 24@xmath0 m pixels , of which the wfs requires a read - out window of 72 @xmath5 72 pixels for the spot array ( 9 subapertures , 8 pixels / subaperture ) .
the detector noise characteristics , qe and cosmetics are all significantly improved at lower temperatures , and to benefit from this a cryostat development programme is under way to lower the wfs operating temperature to 40k ( see section [ sec : wfs_cryomech ] ) .
the detector read - out is managed by eso s next generation controller ( ngc)@xcite , a new high - speed data acquisition system , to be housed inside the coud lab close to the detector array .
.requirements for the performance of the ir detectors for the gravity wfs . [ cols="<,^,^",options="header " , ]
the gravity wfs development programme will enable adaptive optics wavefront sensing at ir wavelengths for the first time at the vlt interferometer , enabling the use of ir - bright reference sources for the gravity instrument .
detailed performance simulations show that the wfs with a shack - hartmann optical design scheme will provide 360 nm rms residual wfe for the case of a mk=7 ao guide star observed on - axis , in line with the required performance .
the wfs will benefit greatly from the use of the new selex / eso saphira eapd detector arrays , which will for the first time allow frame rates of @xmath2500 hz with low read noise ( 6.5 e- rms ) compared with current state - of - the - art nir detectors .
part of the instrument will be cooled to ln2 temperatures , with an upgrade path for pulse tube cooling to 40k under active development
. the wfs will be comprehensively tested in the adaptive optics lab at mpia , where closed loop performance will be achieved with the aid of an atmospheric seeing generator and a macao deformable mirror identical to the units installed at vlt . the first gravity wfs is expected to see first light at paranal in 2014 .
eisenhauer , f. et al .
, `` gravity : observing the universe in motion '' , the messenger 143 , pp .
16 - 24 ( 2011 ) .
paumard , t. et al , `` science with gravity , the nir interferometric imager '' , proc .
s. 439 , p. 267
vincent , f. h. et al .
, `` performance of astrometric detection of a hotspot orbiting on the innermost stable circular orbit of the galactic centre blck hole '' , mnras 412(4 ) , pp .
2653 - 2664 ( 2011 ) choquet , e. et al .
, `` gravity : design and performance of the fringe tracker '' , proc .
spie 7734 , 77341z-1 ( 2010 ) .
jocou , l. et al .
, `` the integrated optics beam combiner assembly the gravity / vlti instrument '' , proc .
spie 8445 , in press ( 2012 ) .
fischer , s. et al .
, `` the gravity spectrometers : system design '' , proc .
spie 8445 , in press ( 2012 ) .
straubmeier , c. et al , `` the gravity spectrometers : optical design and principle of operation '' , proc .
spie 7734 , 773432 - 11 ( 2010 ) .
bartko , h. et al . , `` the fringe detection laser metrology for the gravity interferometer at the vlti '' , proc .
spie 7734 , 773421 ( 2010 ) .
delplancke , f. et al . , `` star separator system for the dual - field capability ( prima ) of the vlti '' , proc .
spie 5491 , pp .
1528 - 1535 ( 2004 ) .
yang , p. et al .
, `` optimizing the transmission of the gravity / vlti near - infrared wavefront sensor '' , proc .
spie 8445 , in press ( 2012 ) .
finger , g. et al .
, `` evaluation and optimization of nir hgcdte avalanche photodiode arrays for adaptive optics and interferometry '' , proc .
spie 8453 , in press ( 2012 ) .
suarez - valles , m. et al .
, `` sparta for the vlt : status and plans '' , proc .
spie 8447 , in press ( 2012 ) .
clnet , y. et al .
, `` dimensioning the gravity adaptive optics wavefront sensor '' , proc .
spie 7736 , 77364a ( 2010 ) . | gravity@xcite second generation 4-beam combiner instrument eso s large telescope interferometer vlti paranal observatory operating near - infrared k - band 2.2 @xmath0 m ) 10-micro - arcsec astrometric precision phase - referenced imaging 4 milli - arcsec resolution fibre - fed optics wavefront sensors fringe tracker beam stabilisation novel metrology concept performance nir interferometric instrumentation astrometric precision first astronomical light vlti planned 2014 science goals probing galactic centre black hole@xcite detecting masses black holes massive star clusters mass accretion jets young stellar objects characterize motions exoplanets binary stars young stellar discs adaptive optics ao ) correction improved beams project includes 4 new wavefront sensors vlt telescopes measurement atmospheric turbulence correction nir h k bands ( 1.4 - 2.4 @xmath0 m ) installed coud space interface with multiple application curvature adaptive optics systems vlt overview design development wfs units real time control expected on - sky performance operational modes ao system descriptions gravity instrument subsystems operational concept ao architecturegravity instrument consists three components beam - combiner instrument laser metrology system wavefront sensors fig . in beam combiner instrument four input beams from telescopes interferograms from six baselines fringe tracking implemented for longitudinal vibrations atmospheric piston input beams fed via optical fibres into two beam combiners fringes science fringe tracking targets spectrally dispersed by spectrometers science spectrometer resolutions of@xmath220 fringe tracking spectrometer provides low - resolution dispersion laser metrology system tracks reference beam precise measurement of optical path with vlt uts gravity s system adaptive optics hardware for atmospheric turbulence effects wavefront sensing performed with dedicated wfs at nir wavelengths interfacing with macao dms light fed to wfs from telescopes via star separator systems sts@xcite at each ut up 2 stars fieldrequirements wfs real - time processing wavefront distortions on - off - axis wavefront sensing sensing in h and k bands ( 1.45 - 2.45 use by other vlti instruments interface to macao deformable mirrors on - axis mode science target as reference source beam splitter redirects light to wfs rest vlti tunnels gravity beam combiner off - axis mode dedicated reference source wfs receives light from ao reference performance requirements mk=7 : strehl 25% off - axis source at angle=7 35% on - axis mk=10 strehl 10% on - axis at zenith rms tip / tilt at sts room 10 mas 7 20 10 piston fluctuations of 25 nm rms over 48 ms 125 over 290 ms 2000 nm over 10 min new hardware nir wavefront sensor new real time control system for describes optical cryo - mechanical design wfs unit control architecture detector characteristics wavefront sensors follow shack - hartmann design incoming light through lenslet array each spot refocused onto detector distorted wavefront characterizedwfs unit comprises warm cold assembly warm optics in fig . relay light from telescope sts units to cryostat entrance wfs is ao mode selector dichroic three flat mirrors translation stage chooses three operational modes for on - axis mode off - axis mode light into visible ao mode light folded by parabolic flat mirrors to pupil derotator maintains pupil orientation parabolic mirror in motorized gimbal for optical alignment focus adjustment derotator follows k - mirror design in gold - coated aluminium cold optics inside bath cryostat cooling concept in field lens compensates for pupil motion using piezoelectric actuator light relayed to lenslet array via filter wheel bandpass neutral density filters lenslet array focuses 9 x 9 spots onto detector from 192 m diameter micro - lenslets focal length 2.09 mm details optical design presented by yang et al.@xcite table selex_req summarises key requirements for detector performance for wfsgravity wfs use high - speed ir electron avalanche photodiode arrays by eso selex - galileo selex / saphira arrays reach read noise @xmath43 e- quantum efficiency qe @xmath240% at 80k@xcite arrays measure 320 @xmath5 256 24@xmath0 m pixels wfs requires read - out window 72 @xmath5 pixels for array 9 subapertures 8 pixels / detector noise characteristics qe cosmetics improved at lower temperatures cryostat development programme wfs operating temperature to 40k detector read - out managed by eso s next generation controller)@xcite high - speed data acquisition system coud lab detector array.requirements for performance ir detectors for gravity wfs wfs development adaptive optics wavefront sensing at ir wavelengths vlt interferometer ir - bright reference sources performance simulations wfs 360 nm rms residual wfe for mk=7 ao guide star observed on - axis from new selex / eso saphira eapd detector arrays allow frame rates @xmath2500 hz low read noise ( 6.5 e- rmsinstrument cooled to ln2 temperatures upgrade for pulse tube cooling to 40k wfs tested in adaptive optics lab at mpia closed loop performance atmospheric generator macao deformable mirror identical to units vlt first gravity wfs light at paranal in 2014 eisenhauer f gravity observing universe in motion messenger 143 pp 16 - 24 2011 paumard t science with gravity nir interferometric imager s. 439 p. 267 vincent f performance astrometric detection hotspot orbit galactic centre blck hole mnras 412(4 ) pp 2653 - 2664 2011 choquet e. gravity design performance fringe tracker spie 7734 , 77341z-1 2010 jocou l integrated optics beam combiner assembly gravity / vlti instrument 8445 2012 fischer s gravity spectrometers system design 8445 2012 straubmeier gravity spectrometers optical design principle operation spie 7734 773432 - 11 2010 bartko fringe detection laser metrology for gravity interferometer at vlti spie 7734 , 773421 2010 delplanckestar separator system dual - field capability vlti spie 5491 pp 1528 - 1535 2004 ) yang p. optimizing transmission gravity vlti near - infrared wavefront sensor spie 8445 2012 ) finger g. evaluation optimization nir hgcdte avalanche photodiode arrays adaptive optics interferometry spie 8453 2012 ) suarez - valles. sparta vlt status plans spie 8447 2012 ) clnet y dimensioning gravity adaptive optics wavefront sensor spie 7736 77364a 2010 ) | 0.431055 | 2,768 | 1,390 | , Saving $0.1 in GPT-4. |
20,907 | we study the disorder driven superconductor to insulator transition in amorphous films of high carrier - concentration indium - oxide . using thin films with various sizes and aspect ratios
we show that the ` critical ' sheet - resistance @xmath0 depends systematically on sample geometry ; superconductivity disappears when @xmath0 exceeds @xmath1k@xmath2 in large samples . on the other hand , wide and sufficiently short samples of the same batch exhibit superconductivity ( judged by conductivity versus temperature ) up to @xmath0 which is considerably larger .
these results support the inhomogeneous scenario for the superconductor - insulator transition . | sufficiently strong spatial disorder may precipitate an insulating phase in an otherwise metallic system .
when the ground state of the metal under study is a superconductor , the disorder - driven transition is called the superconductor to insulator transition ( sit ) .
the sit has been studied over many decades , especially in thin films where tuning the disorder is readily achieved by varying the film thickness @xcite .
many common features of the sit in such 2d ( two - dimensional ) systems were found in different superconducting materials . on the other hand some details of the resistive transitions such as re - entrant behavior , prompted researchers to focus on systems that are structurally ` homogeneous ' presumably with the hope of being able to characterize the `` clean '' physics of the sit .
it was however found out that even purely amorphous materials , that structurally show no irregularities on scales of @xmath310 , often exhibit features that are suggestive of ( transport ) inhomogeneities .
it was argued that some degree of inhomogeneity is unavoidable as this is an inherent effect of disorder @xcite . in addition , recent results , most notably a non - monotonous magneto - resistance has been reported by a number of groups @xcite .
this magnetoresistance involves conductance variation of many orders of magnitude and is difficult to explain except through a percolation picture that explicitly treats the system as being inhomogeneous @xcite . in this note
we give further evidence for the inhomogeneous nature of transport near the sit of amorphous indium - oxide ( @xmath4 ) films .
using samples with lateral sizes of 1 mm to 0.45 @xmath5 m , and aspect ratios in the range 1 to @xmath310@xmath6 we show that the conductivity at liquid helium temperatures is strongly scale dependent .
the results are interpreted as evidence for long - range potential - fluctuations , and it is suggested that their presence be taken into account in trying to explain details of the sit .
the present study was done on thin films of amorphous indium oxide @xmath4 .
these were prepared in a vacuum chamber capable of achieving a base pressure of better than 3@xmath710@xmath8 mbar and equipped with electron - beam and knudsen evaporation sources .
oxygen , cleaned by cold trap , could be bled into the chamber at controlled rates though a needle valve .
standard glass slides were used as substrates .
these were chemically cleaned before being placed in the chamber . a setup for plasma cleaning of the substrates
was used prior to deposition using pure oxygen .
after plasma cleaning for a few minutes in an atmosphere of 2@xmath710@xmath9 mbar , the chamber was pumped down to better than 5@xmath710@xmath10 mbar .
films were prepared by e - beam evaporation of 99.999% pure pressed in@xmath11o@xmath12 pieces manufactured by cerac as described in detail elsewhere @xcite . a quartz crystal thickness monitor ( calibrated against a tolanski interferometer )
was used to determine deposition rates and film thickness .
the @xmath4 films used in this study were deposited at a rate of 1 / s to a nominal thickness of 200 - 250 .
these conditions yield the relatively high carrier concentration @xmath13 ( @xmath4 samples with @xmath14 smaller than @xmath15 are not superconducting down to 1 k ) .
the as - prepared films were studied by high resolution transmission electron microscopy and exhibited microstructure and diffraction patterns that are typical for amorphous structures ( c.f .
, figure 1 ) .
samples for measuring large scale material properties and hall - effect measurements were defined by placing an appropriate aluminum mask over the substrate during the deposition . to study the scale dependence , we compared the properties of batches of simultaneously prepared samples of common width and different lengths down to less than a micron .
these were prepared by a liftoff technique ; strips of amorphous indium - oxide , 0.5@xmath16 mm wide and few millimeters long were made using a mechanical mask .
thin fibers were then pulled from a drop of 7031-ge varnish and placed across the strip .
the thickness of the fibers could be controlled in the range of @xmath1750@xmath18 m down to less than 0.2@xmath18 m by proper dilution of the varnish in a 1:1 solution of ethanol and toluene ( varnish thinner ) , and by the speed at which the fibers were pulled .
the slides were then returned to the vacuum chamber and a @xmath3300 layer of gold was deposited over the @xmath4 . finally the slides were immersed in varnish thinner to dissolve the fibers and wash away the gold that was on them .
this left a series of gaps in the gold layer which defined indium oxide samples all `` cut '' from the same thin film , having a common width but different lengths .
longer samples could be made by using commercial copper wire instead of the varnish fibers .
electrical connections were made by soldering wires to small pieces of indium metal pressed onto the gold contacts . before depositing the gold and after completing sample preparation
the fibers / gaps were viewed through an optical microscope at x400 magnification to see the range of sample lengths obtained and their geometrical quality .
figure 2 shows a sem micrograph of a 0.9@xmath18 m long sample .
the borders were checked to bee smooth and equidistant over the entire width of the sample ( 500@xmath18 m ) to within @xmath310% .
[ ptb ] fig1.eps
as prepared films of @xmath4 are usually insulating and exhibit negative temperature coefficient of resistivity from room temperatures down , and some form of variable range hopping may be observed at the 10 - 100 k range @xcite .
samples that are near the sit on the insulating side tend to show simple activation at liquid helium temperatures .
namely , @xmath19 is seen to follow @xmath20 $ ] with typical activation energy @xmath21 of order 3 - 15 k depending on disorder .
this is true in 3d @xcite as well as in the 2d regime @xcite and is still one of the important unsolved issues of transport near the sit @xcite .
figure 3 illustrates this behavior for a large - area film measured at several stages of annealing ( disorder ) .
note that as the sheet - resistance @xmath22 at @xmath23 @xmath24 4 k is reduced by annealing , so does the activation energy , and for @xmath256 k@xmath2 the temperature dependence becomes rather weak .
in fact , as shown in the inset , samples with this value of @xmath22 are superconducting below @xmath30.3 k. a ` critical ' @xmath22 may be thus identified at the value of @xmath266 k@xmath2 for the sit in this system .
a similar value for @xmath27 has been reported by other researchers .
a somewhat different situation is encountered in the small - area samples .
this is illustrated in figures 4a and 4b that show @xmath28 data for a single batch of samples at two stages of annealing and for samples with different length @xmath29 note that @xmath21 has a marked dependence on @xmath30 ( and/or the aspect ratio ) and this leads to a significant variation of @xmath22 as the temperature is lowered ( figure 4a ) .
upon further annealing ( causing a decrease in the average disorder ) , this disparity makes for a sharper division of the @xmath28 data ; the two longest samples in the batch are still activated with @xmath21 somewhat reduced relative to the previous stage .
however , the two shortest samples of the same batch now show a clear tendency to become superconducting at low temperatures ( figure 4b ) .
this is a striking result as the samples are all made from the same batch and in fact show very similar resistances at higher temperature . moreover ,
associating , as previously done , @xmath27 with the value of the sheet resistance that separates these two groups we now get a value that is at least twice larger than that of the large - area samples of the same material .
it is emphasized that the length - controlling fibers were placed on the common @xmath4 strip in close proximity to one another and at no particular order ( _ i.e_. , no hierarchy of @xmath30 along the strip ) , and therefore the @xmath30 dependence can not be related to a ` technological ' inhomogeneity .
these dependencies of the transport properties on sample scale and aspect ratio is suggestive of a percolation phenomena in an inhomogeneous media .
as argued before @xcite the inhomogeneity is a natural consequence of disorder , and inasmuch as superconductivity is concerned , the effect of disorder may be highly accentuated .
the argument is based essentially on the phenomenon of universal conductance fluctuation : superconductivity usually vanishes when the average conductance of the 2d system is comparable with @xmath31 at this degree of disorder the conductance of the underlying diffusive metal fluctuates considerably on scales of order @xmath32 the phase - coherent length .
in fact , the local conductivities @xmath33 are naturally distributed over a @xmath34 range , which means that at the transition to strong localization the distribution - width of the local conductances is of order unity .
therefore as the average critical disorder is attained , there are perforce some regions of the sample where the local disorder is smaller than critical and these regions may be locally superconducting .
it is easy to see that from the point of view of conductivity the system may appear ` granular ' with small superconducting regions embedded in an insulating matrix ( and a complementary situation on the ` weakly localized ' side of the transition ) .
this may happen even in a structurally featureless amorphous structure as are the samples in the present case ( as mentioned above , c.f .
, figure 1 ) .
such a scenario is generic and should be pertinent for all materials where sub - critical disorder does not appreciably decrease @xmath35 .
these then include all systems that obey the anderson theorem as well as metals that show ` enhancement ' phenomena .
if one allows for just short - range disorder , these arguments can not however account for the spatial scale over which the results in figures 4a and 4b are still size dependent .
gaussian disorder could conceivably lead to an inhomogeneous situation on scale of the order of @xmath36 evaluated at the measurements temperatures ( or the percolation radius @xmath37 of the hopping problem on the insulating side of the transition ) .
neither of these lengths is larger than 1 @xmath5 m at liquid helium temperatures whereas significant size dependence is observed in our experiments up to @xmath30 that is at least one order of magnitude larger .
the relevant scale for the actual ` granularity ' may of course be larger than the size of the individual superconducting island ; several such islands may be josephson - coupled to form a larger cluster .
but this still leaves unanswered the scale dependence of the activation energy observed in figure 4a to extend over tens of microns .
it would therefore seem necessary to consider the existence of long - range potential - fluctuations in these samples , which is not an uncommon phenomenon in amorphous materials @xcite .
in fact , long - range fluctuations in the disorder may be expected to be present when the disorder is large as must be the case here ( strong disorder is actually required when a high carrier - concentration system is to be rendered anderson localized ) .
this is due to the cost of creating a potential gradient which would favor large potentials drops on the longer spatial scales . naturally then
, long range potential fluctuations should be a consideration especially when the sample is prepared or is later exposed to room ( or higher ) temperatures .
the presence of long - range potential - fluctuations in these films is also reflected in the current - voltage ( @xmath38 ) characteristics of the studied samples . in their normal state all our samples showed deviations from ohmic behavior above a characteristic voltage @xmath39 .
these deviations were always such that @xmath40which is presumably related to the fact that the temperature coefficient of the resistance is negative at helium temperatures . as a rule , the critical field @xmath41at which non - linearity became important was noticeably larger at small @xmath42this trend
is illustrated in figure 5 for all the relevant samples used in our study . despite the scatter in the data ,
it is clear that long samples are `` softer '' than short ones as may be expected from long - range fluctuations ; samples with larger @xmath30 are associated with with larger local fields hence they are more susceptible to non - linearity at any macroscopic field .
note that @xmath43 at which measurable deviations from ohm s law are observed is scale - dependent over the same range of @xmath44where the data in figure 4a and 4b exhibit scale - dependence . in summary
, we have presented experimental evidence for the inhomogeneous nature of transport in an amorphous system near the sit .
the results seem to suggest the relevance of long - range potential - fluctuations which may result from thickness and/or composition variations to which characterization methods such as diffraction , stm , tem , as well as other form of microscopy may not be as sensitive as charge transport . these issues clearly deserve further experimental and theoretical elucidation . | strong spatial disorder may insulating phase in metallic system ground state metal is superconductor disorder - driven transition called superconductor to insulator transition ( sit ) sit studied decades especially in thin films tuning disorder by varying film thickness common features of sit in 2d systems found in different superconducting materials resistive transitions re - entrant behavior prompted researchers to focus on systems structurally homogeneous characterize clean physics of sit amorphous materials no irregularities exhibit suggestive of transport inhomogeneities argued inhomogeneity unavoidable inherent effect of disorder recent results non - monotonous magneto - resistance reported involves conductance variation orders magnitude difficult to explain except through percolation picture system inhomogeneous further evidence for inhomogeneous nature of transport near sit of amorphous indium - oxide ( ) films samples with lateral sizes 1 mm to 0.45 m aspect ratios 1 to @xmath310 conductivity at liquid helium temperatures scale dependent results evidence for long - range potential - fluctuations suggested presence in details sitstudy on thin films amorphous indium oxide @xmath4 prepared in vacuum chamber base pressure 3@xmath710 mbar equipped with electron - beam knudsen evaporation sources oxygen cleaned by bled into chamber though needle valve standard glass slides as substrates chemically cleaned before chamber plasma cleaning used using pure oxygen after 2 mbar chamber pumped to 5@xmath710 mbar films prepared by e - beam evaporation of 99.999% pure pressed in@xmath12 pieces cerac quartz crystal thickness monitor used deposition rates film thickness @xmath4 films deposited at 1 / s to nominal thickness 200 - 250 high carrier concentration @xmath13 samples with @xmath14 smaller than not superconducting down to 1 k films studied by high resolution transmission electron microscopy exhibited microstructure diffraction patterns typical for amorphous structures figure 1 samples for defined aluminum mask over substrate during deposition compared properties of batches prepared samples of width different lengths less micronprepared by liftoff technique strips of amorphous indium - oxide 0.5@xmath16 mm wide millimeters long made using mechanical mask thin fibers pulled from 7031-ge varnish placed across strip thickness controlled m to 0.2@xmath18 m by dilution varnish in 1:1 ethanol toluene speed slides returned to vacuum chamber @xmath3300 layer of gold deposited over @xmath4 slides immersed in varnish thinner dissolve fibers gold left gaps in gold layer defined indium oxide samples film common width different lengths longer samples commercial copper wire instead electrical connections by soldering wires to indium metal gold contacts before fibers / gaps viewed through optical microscope at x400 magnification sample lengths geometrical quality figure 2 shows micrograph of 0.9@xmath18 m long sample borders checked smooth equidistant over entire width to @xmath310% films of @xmath4 insulating exhibit negative temperature coefficient resistivity room variable range hopping at 10 - 100 k range samples near insulating side show activation at liquid helium temperatures@xmath19 @xmath20 with typical activation energy @xmath21 3 - 15 k depending on disorder true in 3d 2d regime @xcite unsolved transport near sit @xcite figure 3 illustrates behavior for large - area film at several stages annealing sheet - resistance @xmath22 at @xmath23 @xmath24 4 k reduced by annealing activation energy @xmath256 k@xmath2 temperature dependence weak samples with value @xmath22 superconducting below @xmath30.3 k critical @xmath22 identified at value @xmath266 k@xmath2 for similar value for @xmath27 reported by other researchers different situation in small - area samples illustrated in figures 4a and 4b @xmath28 data for at two stages annealing different length @xmath21 on @xmath30 leads to variation of @xmath22 as temperature lowered figure upon further annealing average disorder disparity sharper division of @xmath28 data two longest samples activated with @xmath21 reduced two shortest samples superconducting at low temperatures figure 4b samples same batch show similar resistances at higher temperatureassociating @xmath27 with sheet resistance groups value twice larger than large - area samples same material length - controlling fibers placed on common @xmath4 strip close at no order no hierarchy of @xmath30 @xmath30 dependence not related to technological inhomogeneity dependencies of transport properties on sample scale aspect ratio suggestive of percolation phenomena in inhomogeneous media inhomogeneity is natural consequence of disorder superconductivity effect disorder may accentuated argument based on universal conductance fluctuation superconductivity vanishes when average conductance 2d system comparable with @xmath31 conductance of underlying diffusive metal fluctuates considerably on scales @xmath32 phase - length local conductivities @xmath33 distributed over @xmath34 range at transition to strong localization distribution - width of local conductances of order unity average critical disorder attained regions where local disorder smaller than critical regions may be locally superconducting system may appear granular with small superconducting regions in insulating matrix complementary situation on weakly localized side of transition even in structurally featureless amorphous structure samples present case figure 1scenario generic pertinent for materials where sub - critical disorder decrease @xmath35 include systems anderson theorem metals enhancement phenomena short - range disorder arguments account for spatial scale results in figures 4a and 4b size dependent gaussian disorder could lead to inhomogeneous situation scale @xmath36 temperatures or percolation radius @xmath37 neither lengths larger than 1 @xmath5 m at liquid helium temperatures significant size dependence experiments up to @xmath30 one order magnitude larger relevant scale for granularity may be larger than size superconducting island several islands - coupled larger cluster leaves unanswered scale dependence of activation energy figure 4a over tens of microns necessary consider long - range potential - fluctuations in samples not uncommon in amorphous materials long - range fluctuations expected present when large strong disorder required high carrier - concentration system due to cost creating potential gradient large potentials drops longer spatial scales long range potential fluctuations consideration especially when sample prepared exposed to room or higher temperatures long - fluctuations reflected in current - voltage ( @xmath38 ) characteristics samplesnormal state samples showed deviations from ohmic behavior above voltage @xmath39 deviations @xmath40which related to temperature coefficient resistance negative at helium temperatures critical field @xmath41at non - linearity important larger at small @xmath42this trend illustrated in figure 5 samples long samples softer than short samples larger @xmath30 associated larger local fields more susceptible to non - linearity macroscopic field @xmath43 deviations from ohm s law observed scale - dependent range @xmath44where data figure 4a 4b exhibit scale - dependence presented experimental evidence for inhomogeneous transport in amorphous system near sit results suggest relevance long - range potential - fluctuations from thickness composition variations characterization methods not sensitive transport issues deserve further experimental theoretical elucidation | 0.436187 | 2,916 | 1,459 | , Saving $0.1 in GPT-4. |
20,913 | we use a two - component model to describe the production mechanism of @xmath0 mesons in @xmath1 collisions .
the model combines the usual qcd processes plus fragmentation and recombination of charm that has been produced by nonperturbative qcd mechanisms . a hard charm component in the pion
must be responsible of the particle anti - particle production asymmetries observed . 6.5 in -.2 in .25 in .25 in | recent measurements of charm meson production in @xmath1 collisions @xcite indicate that there are important nonperturbative qcd phenomena in the production process .
+ the so called `` leading effect '' or particle anti - particle production asymmetry is much bigger than predicted by next - to - leading ( nlo ) , qcd predictions .
+ in this letter we study the @xmath2 distribution of @xmath3 and @xmath4 mesons in the framework of a two - component model .
the production of @xmath0 mesons in the model is assumed to take place _ via _ two different processes , namely qcd parton fusion with the subsequent fragmentation of quarks in the final state and conventional recombination of valence and sea quarks present in the initial state .
+ the asymmetry obtained with the conventional soft charm component does not reproduce the experimental results .
we show that with the presence of a hard charm component in the pion , the sum of these two processes gives rise to an enhancement at large values in the @xmath2 distribution of leading mesons .
+ unlike the @xmath5 @xcite where the asymmetry is generated by the presence of the @xmath6 diquark in the initial hadron , the meson production asymmetry seems to imply the presence of a hard charm component in the pion .
+ to quantify the difference in the production of leading and non - leading particles , an asymmetry @xmath7 is defined @xmath8 in the reaction @xmath9 - @xmath10 the @xmath11 and @xmath12 are leading mesons while @xmath13 and @xmath14 are non - leading .
+ several models have been proposed although none fully account for the difference in production between leading and non - leading particles .
the intrinsic charm model @xcite postulates the existence of quantum fluctuations in the beam particle that bring about fock states containing @xmath15 pairs .
the @xmath15 quarks having the same velocity as the original valence quarks are likely to coalesce forming leading particles .
although the shape of the asymmetry as a function of @xmath2 predicted by this model is similar to the experimental measurement , the prediction is to low for the whole @xmath2 interval .
+ in @xcite it is pointed out that the annihilation of the @xmath16 from the proton and the @xmath17 in the pion would liberate the @xmath18 of the pion which can then recombine to form a @xmath19 and certainly not a @xmath14 .
the @xmath4 will not be enhanced because it can not be formed with this simple annihilation diagram .
the recent measurement of significant asymmetry in @xmath5 production in @xmath20 collisions @xcite however , shows that there must be other production mechanisms at work that must account for the asymmetry .
+ in two component models @xcite the production of charm mesons by parton fusion is the same for @xmath3 and @xmath4 . in section 2 the formalism to obtain the cross section by parton fusion is shown for charged and neutral charm mesons .
the recombination of charm as the second component is discussed in section 3 .
recombination , is different for the charged mesons where a parton process favors the formation of @xmath19 enhancing the @xmath3 asymmetry .
this will be discussed in section 4 .
in the parton fusion mechanism the @xmath3 @xmath21 mesons are produced via the @xmath22 with the subsequent fragmentation of the @xmath23 quark .
the inclusive @xmath2 distribution of the charm mesons in @xmath1 collisions , is given by @xcite @xmath24 where @xmath25 and @xmath26 and @xmath27 being the parton momentum fractions , @xmath28 and @xmath29 the quark and gluon distribution in colliding hadrons , @xmath30 the energy of the produced @xmath31-quark and @xmath32 the appropriated fragmentation function . in eq .
[ sig - qcd ] , @xmath33 is the squared transverse momentum of the produced @xmath31-quark , @xmath34 is the rapidity of the @xmath35 quark and @xmath36 is the momentum fraction of the charm quark carried by the @xmath0 .
the sum in eq .
[ int - qcd ] runs over @xmath37 .
+ we use the lo results for the elementary cross - sections @xmath38 and @xmath39 : @xmath40 ^3 } \label{q - antiq}\ ] ] @xmath41 ^ 3 } \ : \left [ cosh \left ( \delta y \right)+ \frac{2m_c^2}{\hat{m}_c^2}+\frac{2m_c^4}{\hat{m}_c^4}\right ] , \label{gg}\ ] ] where @xmath42 is the rapidity gap between the produced @xmath31 and @xmath43 quarks and @xmath44 . in order to be consistent with the lo calculation of the elementary cross sections , we use the grv - lo parton distribution functions @xcite , and apply a global factor @xmath45 in eq .
[ sig - qcd ] to take into account nlo contributions @xcite .
+ we take @xmath46 for the @xmath31-quark mass and fix the scale of the interaction at @xmath47 @xcite . following @xcite
, we use a delta fragmentation function @xmath48 , which seems to describe experimental data better than the peterson fragmentation function .
sometime ago v. barger
_ et al . _
@xcite explained the spectrum enhancement at high @xmath2 in @xmath5 production assuming a hard momentum distribution of charm in the proton . according with them charm anti - charm pairs which give rise to the flavor excitation diagrams ( see fig
. 1 ) are not intrinsic but generated by qcd evolution of the structure functions . + in this framework , following the gluon scattering process of fig .
1 the charm quarks will fragment into charm hadrons .
when the @xmath43 is scattered , the spectator quark could recombine with the @xmath17 valence quark of the pion to form a @xmath49 or less frequently with an antiquark from the pion s sea . on the other hand , when the @xmath31 quark is scattered
, the spectator quark @xmath43 could recombine with @xmath18 valence quark from the pion to form a @xmath50 meson . +
as pointed out in @xcite the charm hadrons resulting from the scattered charm quark , populate the low @xmath2 region of the cross section , while those originating from the spectator quark dominate at high @xmath2 .
+ here we assume a qcd evolved charm distribution , of the form proposed by v. barger _
@xcite @xmath51 with a normalization @xmath52 fixed to @xmath53 and @xmath54 . with this values for @xmath55 and
@xmath56 one tries to resemble the distribution of valence quarks .
in contrast with the parton fusion calculation , in which the scale @xmath57 of the interaction is fixed at the vertices of the appropriated feynman diagrams , in recombination the value of the parameter @xmath57 should be used to give adequately the content of the recombining quarks in the initial hadron .
we used @xmath58 and therefore , the integrated charm quark distribution would take the value given in eq.7 . + the production of leading mesons at low @xmath59 was described by recombination of quarks long time ago @xcite .
+ in recombination models it is assumed that the outgoing hadron is produced in the beam fragmentation region through the recombination of the maximum number of valence quarks and the minimum number of sea quarks of the incoming hadron .
the invariant inclusive @xmath2 distribution for leading mesons is given by @xmath60 where @xmath61 , @xmath62 are the momentum fractions and @xmath63 is the two - quark distribution function of the incident hadron .
@xmath64 is the two - quark recombination function . +
the two - quark distribution function is parametrized in terms of the single quark distributions @xmath65 with @xmath66 .
we use the grv - lo parametrization for the single quark distributions in eqs .
[ 3-quark ] .
the charm contribution however , is parametrized with the distribution of eq .
it must be noted that since the grv - lo distributions are functions of @xmath67 and @xmath57 , then our @xmath68 also depends on @xmath57 .
the recombination function is given by @xmath69 with @xmath70 fixed by the condition @xmath71 .
the inclusive production cross section of the @xmath0 is obtained by adding the contribution of recombination eq .
[ rec - cs ] to the qcd processes of eq .
[ sig - qcd ] , @xmath72 the resulting inclusive @xmath0 production cross section @xmath73 is used then to construct the asymmetry defined in eq .
+ in the @xmath1 reaction however , the recombination of a @xmath74 gets a higher probability than the recombination of a @xmath4 @xmath75 @xcite .
+ the process that releases a @xmath18 quark from the incident @xmath76 to form a @xmath74 ( after recombination with a c - quark from the sea ) is also present in the formation of a @xmath4 . in this case
the released quark should be @xmath17 . in the @xmath19 case
however , there is an additional mechanism that releases the @xmath18 quark from the pion , namely the fusion of a @xmath16-quark from the proton with the @xmath16 from the pion .
this is correctly incorporated by the fact that a higher contribution from the recombination is needed to describe the @xmath19 asymmetry while @xmath77 ) production and its production asymmetry requires a small contribution from the recombination process .
2 shows the model prediction for the @xmath19 and @xmath14 cross section and experimental results from the wa82 collaboration @xcite .
the corresponding production asymmetry is shown in fig .
3 together with all the experimental measurements available . fig .
4 shows the cross section for @xmath4 and @xmath13 mesons produced in @xmath78 collisions and the experimental points from the wa92 experiment @xcite . the corresponding production asymmetry for @xmath4 and @xmath13
is shown in fig .
by looking at the fit of the cross section in fig.4 one can see that the asymmetry obtained for @xmath79 could actually change drastically by slightly changing the curves .
more statistics is needed to improve the confidence of the predicted asymmetry .
in an earlier work @xcite the production asymmetry of @xmath5 was described using the same recombination scheme used here .
the presence of a diquark in the initial state , plays an important role in @xmath5 production .
however , it is not possible to describe the asymmetry in charm mesons using the grv structure functions for the charm in the pion . a hard charm component must be present .
this fact has been pointed out before @xcite but the asymmetry in the intrinsic charm model @xcite does not seem to fit experimental results very well .
the hard charm component proposed in @xcite and a recombination scheme , give a good description of particle anti - particle production asymmetries for mesons .
99 e769 collaboration ( g. a. alves _ et al .
_ ) , _ phys . rev .
lett . _ * 77*(1996)2388 e769 collaboration ( g. a. alves _ et al .
_ ) , _ phys . rev .
lett . _ * 72*(1994)812 wa82 collaboration ( m. adamovich _ et al .
_ ) , _ phys
* b305*(1993)402 e791 collaboration ( e.m .
_ ) , _ phys . lett . _ * b371*(1996)157 wa92 collaboration ( m. adamovich _
_ ) , _ nucl . phys .
_ * b495 * ( 1997)3 j. dos anjos , g. herrera , j. magnin , f.r.a .
simo , _ phys .
_ * d56 * ( 1997)394 g. herrera and j. magnin , submitted to _ phys .
b _ r. vogt , s.j . brodsky and p.
hoyer , _ nucl .
* b383 * ( 1992)643 v. a. bednyakov , _ mod .
lett . _ * a10 * ( 1995)61 r. vogt and s.j . brodsky , _ nucl .
* b478 * ( 1996)311 j. babcock , d. sivers and s. wolfram , _ phys .
* d 18 * ( 1978)162 ; b.l .
combridge , _ nucl .
* b 151 * , 429 ( 1979 ) ; p. nason , s. dawson and r.k .
ellis , _ nucl .
phys . _ * n 327 * ( 1989)49 ; m. glck , e. reya and a. vogt , _ z. phys . _
* c53 * ( 1992)127 r. vogt , _ z. phys . _
* c71 * ( 1996)475 v. barger , f. halzen , w. y. keung , _ phys . rev . _ * d24*(1981)1428 ; + _ phys
* d25 * ( 1982)112 k.p .
das and r.c .
hwa , _ phys .
* b 68 * ( 1977)459
* heavy flavor excitation diagram . this and similar diagrams are included in a nlo qcd calculation .
however , in oder to reproduce particle anti - particle asymetries a hard charm component in the pion is needed . * production cross section for @xmath19 and @xmath14 mesons in @xmath78 collisions .
the dashed line represents the parton fusion contribution with a delta fragmentation function .
the dotted line gives the contribution from recombination for @xmath19 .
the solid line is the sum of the two contributions . *
[ fig . 3 : ] measured production asymmetry for @xmath19 and @xmath14 and the model incorporating a hard charm component in the pion ( solid line ) .
the dotted line shows the intrinsic charm model prediction . *
production cross section for @xmath4 and @xmath13 mesons produced in @xmath78 collisions .
experimental points @xcite and two component model prediction ( solid line ) . *
[ fig . 5 : ] measured production asymmetry for @xmath4 and @xmath13 and the model incorporating a hard charm component in the pion .
the horizontal line at @xmath80 is for reference only , | measurements charm meson production in @xmath1 collisions indicate important nonperturbative qcd phenomena in production process leading effect particle - particle production asymmetry bigger than predicted next - to - leading qcd predictions study @xmath2 distribution of @xmath3 mesons two - component model production @xmath0 mesons assumed via two processes qcd parton fusion fragmentation conventional recombination of valence sea quarks asymmetry with conventional soft charm component reproduce experimental results presence hard charm component in pion sum processes enhancement large values in @xmath2 distribution of leading mesons @xmath6 diquark meson production asymmetry hard charm component in pion difference production leading non - leading particles asymmetry @xmath7 defined @xmath8 in reaction @xmath9 - are leading mesons non - leading models proposed none account for difference production between leading non - leading particles intrinsic charm model postulates quantum fluctuations in beam particle fock states containing @xmath15 pairs @xmath15 quarks same velocity as original valence likely coalesce forming leading particlesasymmetry @xmath2 predicted model similar experimental measurement prediction low for @xmath2 interval annihilation @xmath16 proton @xmath17 pion @xmath18 @xmath19 not @xmath14 @xmath4 enhanced formed annihilation diagram recent measurement asymmetry in @xmath5 production @xmath20 collisions shows other production mechanisms account asymmetry two component models production charm mesons parton fusion same for @xmath3 @xmath4 section 2 formalism cross section fusion for charged neutral charm mesons recombination charm second discussed section 3 different for charged mesons parton process favors formation @xmath19 enhancing @xmath3 asymmetry discussed section 4 parton fusion @xmath3 @xmath21 mesons produced via @xmath22 fragmentation @xmath23 quark inclusive @xmath2 distribution charm mesons in @xmath1 collisions given by @xcite @xmath24 @xmath25 @xmath27 parton momentum fractions @xmath28 @xmath29 quark gluon distribution colliding hadrons @xmath30 energy @xmath31-quark @xmath32 fragmentation functionsig - qcd @xmath33 squared transverse momentum @xmath31-quark @xmath34 rapidity @xmath35 quark @xmath36 momentum fraction charm quark @xmath0 sum eq int - qcd runs over @xmath37 use lo results elementary cross - sections @xmath38 @xmath39 @xmath40 - @xmath41{2m_c^2 @xmath42 rapidity gap between @xmath31 @xmath43 quarks @xmath44 consistent lo calculation use grv - lo parton distribution functions @xcite apply global factor @xmath45 eq sig - qcd nlo contributions take @xmath46 for @xmath31-quark mass fix scale interaction at @xmath47 @xcite use delta fragmentation function @xmath48 experimental data better than peterson fragmentation function v. barger @xcite explained spectrum enhancement at high @xmath2 @xmath5 production hard momentum distribution charm in proton charm anti - charm pairs flavor excitation diagrams fig 1 not generated by qcd evolution structure functionsframework gluon scattering process charm quarks fragment into hadrons @xmath43 scattered spectator quark recombine with @xmath17 valence quark form @xmath49 or antiquark pion s sea @xmath31 quark scattered spectator quark @xmath43 recombine with @xmath18 valence quark form @xmath50 meson charm hadrons from scattered quark populate low @xmath2 region cross section from spectator quark dominate high @xmath2 assume qcd evolved charm distribution normalization @xmath52 to @xmath53 @xmath54 values for @xmath55 @xmath56 resemble distribution valence quarks contrast with parton fusion calculation scale @xmath57 fixed at vertices feynman diagrams recombination value parameter @xmath57 content recombining quarks initial hadron used @xmath58 integrated charm quark distribution value in eq.7 production of leading mesons at low @xmath59 described by recombination quarks outgoing hadron produced in beam fragmentation region through recombination maximum valence quarks minimum sea quarks incoming hadroninvariant @xmath2 distribution mesons by @xmath60 momentum fractions two - quark distribution function incident hadron @xmath64 two - quark recombination function two - quark distribution parametrized single quark distributions @xmath65 @xmath66 grv - lo parametrization single quark distributions 3-quark charm contribution parametrized with distribution eq grv - lo distributions functions of @xmath67 @xmath57 @xmath68 depends on @xmath57 recombination function by @xmath69 @xmath70 fixed by @xmath71 inclusive production cross section @xmath0 obtained adding contribution recombination eq to qcd processes resulting inclusive @xmath0 production cross section @xmath73 asymmetry eq @xmath1 reaction recombination @xmath74 higher probability than @xmath4 @xmath75 process @xmath18 quark from @xmath76 form @xmath74 recombination - quark present in formation @xmath4 released quark @xmath17 @xmath19 case additional mechanism releases @xmath18 quark from pion fusion @xmath16-quark proton with @xmath16 pionhigher contribution recombination @xmath19 asymmetry @xmath77 production asymmetry requires small contribution 2 shows model prediction @xmath19 @xmath14 cross section experimental results wa82 collaboration @xcite production asymmetry shown in fig . 3 experimental measurements fig 4 shows cross section @xmath4 @xmath13 mesons @xmath78 collisions experimental points wa92 experiment @xcite production asymmetry @xmath4 @xmath13 shown fig . asymmetry @xmath79 could change changing curves more statistics needed improve confidence predicted asymmetry earlier work production asymmetry @xmath5 described same recombination scheme presence diquark initial state important role in @xmath5 production not possible describe asymmetry charm mesons using grv structure functions hard charm component must pointed out @xcite asymmetry intrinsic charm model @xcite fit experimental results hard charm component @xcite recombination scheme good description particle anti - particle production asymmetries mesons 99 e769 collaboration g. a. alves _ et al . phys . rev . lett 77*(1996)2388 e769 collaboration 72*(1994)812 wa82 collaborationadamovich al b305(1993)402 e791 collaboration phys b371(1996)157 wa92 collaboration adamovich phys b495 1997)3 j. dos anjos g. herrera j. magnin simo d56 1997)394 g. herrera j. magnin r. vogt s.j brodsky p. hoyer b383 1992)643 bednyakov lett a10 1995)61 r. vogt s.j brodsky b478 ( 1996)311 j. babcock d. sivers s. wolfram 18 1978)162 b.l combridge nucl 151 429 ( 1979 ) p. nason s. dawson r.k ellis phys 327 ( 1989)49 ; m. glck e. reya a. vogt c53 1992)127 r. vogt phys c71 ( 1996)475 v. barger f. halzen w y. keung phys d24(1981)1428 phys d25 1982)112 k.p . das r.c hwa 68 1977)459 heavy flavor excitation diagram similar diagrams nlo qcd calculationparticle particle asymetries hard charm pion needed production cross section @xmath19 mesons @xmath78 collisions dashed line parton fusion contribution delta fragmentation function dotted line contribution recombination solid line sum two contributions fig . 3 production asymmetry model incorporating hard charm pion dotted line intrinsic charm model prediction production cross section @xmath4 @xmath13 mesons @xmath78 collisions experimental points two component model prediction solid line fig . 5 production asymmetry @xmath4 model incorporating hard charm component pion horizontal line @xmath80 reference | 0.499352 | 3,346 | 1,770 | , Saving $0.1 in GPT-4. |
20,919 | we reconsider the problem of the formation of a large - scale magnetic field in the accretion disks around black holes .
in contrast with previous work we take into account the nonuniform vertical structure of the disk . the high electrical conductivity of the outer layers of the disk prevents the outward diffusion of the magnetic field .
this implies a stationary state with a strong magnetic field in the inner parts of the accretion disk close to the black hole .
keywords : accretion , accretion disks magnetohydrodynamics black hole physics | early work on disk accretion to a black hole argued that a large - scale magnetic field of , for example , the interstellar medium would be dragged inward and greatly compressed by the accreting plasma ( bisnovatyi - kogan & ruzmaikin 1974 , 1976 ; lovelace 1976 ) .
subsequently , analytic models of the field advection and diffusion in a turbulent disk suggested , that the large - scale field diffuses outward rapidly ( lubow , papaloizou , & pringle 1994 ; lovelace , romanova , & newman 1994 ) and prevents a significant amplification of the external poloidal field by electrical current in the accretion disk .
this has led to the suggestion that special conditions ( non - axisymmetry ) are required for the field to be advected inward ( spruit & uzdensky 2005 ) .
we reconsider the question of the advection / diffusion of a large - scale magnetic field in a turbulent plasma accretion disk , taking into account its nonuniform vertical structure .
the high electrical conductivity of the surface layers of the disk , where the turbulence is suppressed by the radiation flux and the relatively high magnetic field , prevents outward diffusion of the magnetic field .
this leads in general to a strong magnetic field in the inner parts of accretion disks around black holes .
there are two limiting accretion disk models which have analytic solutions for a large - scale magnetic field structure .
the first was constructed by bisnovatyi - kogan and ruzmaikin ( 1976 ) for a stationary non - rotating accretion disk .
a stationary state in this disk ( with a constant mass flux onto a black hole ) is maintained by the balance between magnetic and gravitational forces , and thermal balance ( local ) is maintained by ohmic heating and radiative radiative conductivity for an optically thick conditions .
the mass flux to the black hole in the accretion disk is determined by the finite conductivity of the disk matter and the diffusion of matter across the large - scale magnetic field as sketched in figure 1 .
the value of the large - scale magnetic field in stationary conditions is determined by the accretion disk mass , which in turn is determined by the magnetic diffusivity of the matter . for a laminar disk with coulomb conductivity ( which is very large ) ,
the mass of the stationary disk is also very large making the disk s self - gravity important .
correspondingly , the magnetic field needed to support a mechanical equilibrium is also very large , reaching in the central parts of the disk @xmath0 g for a stellar mass black hole , for a temperature and density at infinity @xmath1 k and for @xmath2 g/@xmath3 .
the stationary magnetic field increases with the black hole mass as @xmath4 , and with a mass flux @xmath5 , where @xmath6 ( bisnovatyi - kogan and ruzmaikin 1976 ) .
it is widely accepted that the laminar disk is unstable to different hydrodynamic , magnetohydrodynamic , and plasma instabilities which implies that the disk is turbulent . in x - ray binary systems
the assumption about turbulent accretion disk is necessary for construction of a realistic models ( shakura and sunyaev 1973 ) .
therefore the turbulent accretion disks had been constructed also for non - rotating models with a large - scale magnetic field .
a turbulent magnetic diffusivity was considered by parker ( 1971 ) , and by bisnovatyi - kogan and ruzmaikin ( 1976 ) . in the last paper the turbulent diffusivity
was scaled by the parameters of the turbulent motion , similar to the scaling of the shear @xmath7-viscosity in turbulent accretion disk in binaries ( shakura & sunyaev , 1973 ) where the viscous stress tensor component @xmath8 , with @xmath9 a dimensionless constant and @xmath10 the pressure in the disk midplane . in a more consistent representation ,
the coefficient of turbulent kinematic viscosity @xmath11 in the navier - stokes equation is taken in the form @xmath12 , where @xmath13 is the isothermal sound speed and @xmath14 is the midplane density of the disk .
using this representation , the expression for the turbulent electrical conductivity @xmath15 is @xmath16 here , @xmath17 .
the characteristic turbulence scale is @xmath18 , where @xmath19 is the half - thickness of the disk , the characteristic turbulent velocity is @xmath20 . the mass of the turbulent magnetized disk is orders of magnitude less that in the laminar disk case . the evolution of a large - scale magnetic field threading a turbulent keplerian disk
can be estimated easily .
this field arises from two sources : external electrical currents and currents in the accretion disk .
evidently , the field generated by the currents in the disk can be much larger than that due to the external currents .
the magnetic field may become dynamically important , influencing the accretion disk structure and leading to powerful jet formation , only if it is strongly amplified during the radial inflow of the disk matter .
this amplification is possible only when the radial accretion speed of matter in the disk is larger than the outward diffusion speed of the poloidal magnetic field due to the turbulent diffusivity @xmath21 .
estimates by lubow et al .
( 1994 ) shown that for a turbulent conductivity ( [ eq1 ] ) , the outward diffusion speed is larger than the accretion speed .
thus it appears that there is no large - scale magnetic field amplification during keplerian disk accretion .
this conclusion is discouraging because the most plausible models of many phenomena observed in the the systems with black holes inside the galaxy , as well as in the extragalactic supermassive black holes are connected with a large values of a large - scale magnetic fields .
nevertheless , this result directly follows from the equations of the standard disk structure , with the turbulent electric conductivity ( [ eq1 ] ) lubow et al .
( 1994 ) did numerical calculations for a simplified situation with constant relative disk thickness @xmath22 , constant kinematic viscosity @xmath11 , and turbulent conductivity @xmath15 .
it is easy to show that the same result follows analytically for the standard accretion disk structure which can be written as @xmath23 @xmath24 where @xmath25 is the keperian velocity ( e.g. , bisnovatyi - kogan & lovelace 2001 ) . for regions
far from the inner disk boundary , the specific angular momentum @xmath26 .
the characteristic time @xmath27 of the accretion disk matter advection due to the shear viscosity is @xmath28 . from the first three relations in equation ( [ eq2 ] )
, we obtain @xmath29 to estimate the time - scale of outward magnetic field diffusion , we use @xmath30 ( lubow et al . 1994 ) , where @xmath31 and @xmath32 are the large scale field components evaluated at the top surface of the disk . here the coefficient of the magnetic turbulent diffusivity @xmath33 is obtained from ( [ eq1 ] ) @xmath34 for stationary conditions , the large - scale magnetic field in the accretion disk is determined by the equality @xmath35 which implies @xmath36 in contrast , the coronal poloidal field solutions typically have @xmath37 at the disk surface ( bisnovatyi - kogan and blinnikov 1972 ; ustyugova et al .
1999 ) , which implies that @xmath38 .
this inequality indicates that the magnetic field is not amplified during accretion for these physical conditions .
near the surface of the disk , in the region of low optical depth , the turbulent motion is suppressed by the radiative flux , similar to the suppression of the convection over the photospheres of stars with outer convective zones .
the presence of the outer radiative layer does not affect the estimate of the characteristic time @xmath27 of the matter advection in the accretion disk because it is determined by the main turbulent part of the disk .
the time of the field diffusion , on the contrary , is significantly changed , because the electrical current is concentrated in the radiative highly conductive regions , which generate the main part of the magnetic field .
the structure of the magnetic field with outer radiative layers is shown schematically in figure 2 . inside the turbulent disk
the electrical current is negligibly small so that the magnetic field there is almost fully vertical , with @xmath39 , according to ( [ eq6 ] ) . in the outer radiative layer ,
the field diffusion is very small , so that matter advection is leading to strong magnetic field amplification .
the field amplification will last until the magnetic forces in the region over the photosphere become of the order of the gravitational ones , and start to participate in the equilibrium balance .
in such conditions the mhd and plasma instabilities are developed , decreasing the effective electrical conductivity .
we suppose , that in the stationary state the magnetic forces could support the optically thin regions against gravity . in the nonrotating magnetized disk
magnetic forces support the whole disk against the gravity , so they should be much higher . when the magnetic force balances the gravitational force on the outer optically thin part of the disk of surface density @xmath40 one finds the following relation takes place @xmath41 ( bisnovatyi - kogan & ruzmaikin 1976 ) .
the surface density over the photosphere corresponds to a layer with effective optical depth close to @xmath42 ( e.g. , bisnovatyi - kogan 2001 ) .
we estimate the lower limit of the magnetic field strength , taking @xmath43 ( instead of the effective opacity @xmath44 ) .
writing @xmath45 we obtain @xmath46 ( g/@xmath47 for the opacity of the thomson scattering , @xmath48 @xmath49/g .
the absorption opacity @xmath50 is much less than @xmath43 in the inner regions of a luminous accretion disk . thus using in equation ( [ eq8 ] ) @xmath40 from equation ( [ eq9 ] )
, we estimate the lower bound on the large - scale magnetic field of a keplerian accretion disk as @xmath51 where @xmath52 and @xmath53 . for comparison ,
the surface density @xmath54 of the disk in the inner radiation dominated region , where we may expect the largest values of the magnetic fields , is @xmath55 @xmath56 ( see bisnovatyi - kogan 2001 ) .
the maximum magnetic field is reached when the outward magnetic force balances the gravitational force on the disk of surface mass density @xmath40 . in equilibrium , @xmath57 .
we find that @xmath32 in a keplerian accretion disk is about @xmath58 times less than its maximum possible value for @xmath59 and @xmath60 .
an important question is the energy density of the large - scale stationary magnetic field in comparison with the rotational or gravitational energy density of the disk . for a non - rotating magnetized accretion disk ,
the energy density of the field is of the order of the gravitational one so that the magnetic field strength is very large in the vicinity of a black hole ( bisnovatyi - kogan & ruzmaikin 1976 ) , and may be many order of magnitude larger than the external seed field . in the case of the fully turbulent keplerian disk
the poloidal magnetic field tends to drift outward ( lubow et al .
1994 ; lovelace et al .
1994 ) so that its value can not significantly exceed the strength of the large - scale seed magnetic field .
the suggestion of a fully turbulent accretion disk with a small turbulent conductivity is violated in the outer surface layer of the accretion disk where the optical depth is small and the turbulence is suppressed by the strong radiative flux .
this is similar to the radiative layer above the main body of a convective star ( similar to the sun ; e.g. , bisnovatyi - kogan 2001 ) . in the radiative layer ,
the magnetic field diffusion is much slower than in the region of fully developed turbulence . in the radiative layer
the diffusion is determined by the classical coulomb conductivity which is very large .
the diffusion in this layer is practically negligible .
the electric current is concentrated in the radiative layer , and the main body of the turbulent disk is almost current - free and thus force - free .
the magnetic field lines in this region are almost straight as shown in figure 2 .
because of the negligible diffusion in the radiative layer , the large - scale field drifts inward until the dynamical action of the magnetic field on the photosphere becomes comparable with that of centrifugal and gravitational forces . at this point the inward drift of the field will be halted and a the stationary state formed where the the magnetic , centrifugal , and gravitational forces on the optically thin region will be comparable , and deviations from the keplerian angular velocity will be significant .
the strength of the magnetic field for such conditions is smaller than in the nonrotating disk of bisnovatyi - kogan and ruzmaikin ( 1976 ) , but still it is very large in the vicinity of the black hole . in this situation
we may expect a nonuniform distribution of the angular velocity over the disk thickness : the main body of the turbulent disk is rotates with the velocity close to the keplerian one , and outer optically thin layers rotate substantially slower .
self - consistent models of the rotating accretion disks with a large - scale magnetic field requires solution the equations of magnetohydrodynamics .
we expect two different self - consistent solutions for the same set of the input parameters . in the case of a fully turbulent disk without radiative surface layers ,
the large - scale magnetic field will remain close to the value of the seed field and the disk s angular velocity will be close to keplerian . in the second solution
the strength of the magnetic field is large , and it may greatly exceed the strength of the seed field . in this solution the angular velocity distribution may deviate considerably from the keplerian law . in presence of the radiative layers the solution with a small field will not be stationary , and a transition to the strong field solution will take place .
we conclude therefore that the strong field solution is the only stable stationary solution for a rotating accretion disk .
further investigation of the build up of strong large - scale fields by accretion is underway ( rothstein et al . 2007 ) .
the work of g.b .- k . was partially supported by rbfr grants 05 - 02 - 17697 , 06 - 02 - 90864 , and ran program p-04 .
the work of r.l . was supported in part by nasa grants nag5 - 13220 , nag5 - 13060 , by nsf grant ast-0507760 . | work on disk accretion to black hole argued large - scale magnetic field dragged inward compressed by accreting plasma kogan ruzmaikin 1974 1976 lovelace 1976 models turbulent disk suggested large - scale field diffuses outward rapidly lubow pringle 1994 1994 prevents amplification external poloidal field electrical current disk special conditions non - axisymmetry for field inward ( spruit uzdensky 2005 ) reconsider question advection diffusion large - magnetic field in turbulent plasma accretion disk nonuniform vertical structure high electrical conductivity of surface layers prevents outward diffusion leads to strong magnetic field in inner parts accretion disks around black holes two limiting accretion disk models solutions for large - scale magnetic field structure first by bisnovatyi - kogan ruzmaikin 1976 for stationary non - rotating accretion disk state constant mass flux onto black hole maintained by balance magnetic gravitational forces thermal balance by ohmic heating radiative conductivity mass flux to black hole determined by finite conductivity disk matter diffusion across large - magnetic field in figure 1value large - scale magnetic field in stationary conditions determined by accretion disk mass by magnetic diffusivity laminar disk with coulomb conductivity large mass large self - gravity important magnetic field mechanical equilibrium large central parts disk @xmath0 g stellar mass black hole temperature density infinity @xmath1 k @xmath2 g stationary magnetic field increases with black hole mass mass flux @xmath5 laminar disk unstable to hydrodynamic magnetohydrodynamic plasma instabilities implies disk turbulent in x - ray binary systems assumption turbulent accretion disk necessary for realistic models turbulent disks constructed for non - rotating models with large - scale magnetic field turbulent magnetic diffusivity considered by parker ( 1971 bisnovatyi - kogan ruzmaikin ( 1976 diffusivity scaled by parameters turbulent motion similar to shear @xmath7-viscosity in turbulent accretion disk viscous stress tensor component @xmath8 dimensionless constant pressure in disk midplanecoefficient turbulent kinematic viscosity @xmath11 navier - stokes equation @xmath12 @xmath13 isothermal sound speed @xmath14 midplane density disk expression turbulent electrical conductivity @xmath15 @xmath16 @xmath17 characteristic turbulence scale @xmath18 @xmath19 half - thickness disk turbulent velocity @xmath20 mass turbulent magnetized disk less laminar disk evolution large - scale magnetic field threading turbulent keplerian disk estimated field arises from external electrical currents currents accretion disk field currents disk larger magnetic field important accretion disk structure powerful jet formation amplified during radial inflow disk matter amplification radial accretion speed larger than outward diffusion speed poloidal magnetic field turbulent diffusivity @xmath21 estimates lubow et al. turbulent conductivity outward diffusion speed larger than accretion speed no large - scale magnetic field amplification during keplerian disk accretion conclusion discouraging plausible models of phenomena systems black holes extragalactic supermassive black holes connected with large large - scale magnetic fieldsresult follows from equations standard disk structure turbulent electric conductivity lubow et 1994 calculations constant disk thickness kinematic viscosity @xmath11 turbulent conductivity @xmath15 same result follows standard accretion disk structure written as @xmath23 @xmath24 @xmath25 keperian velocity regions from inner disk boundary specific angular momentum @xmath26 characteristic time @xmath27 accretion disk matter advection due to viscosity @xmath28 first three relations equation obtain @xmath29 time - scale outward magnetic field diffusion use @xmath30 @xmath31 @xmath32 large scale field components top surface disk coefficient magnetic turbulent diffusivity @xmath33 obtained from eq1 @xmath34 stationary conditions large - scale magnetic field accretion disk determined by equality @xmath35 implies @xmath36 coronal poloidal field solutions have @xmath37 at disk surface implies @xmath38 inequality indicates magnetic field not amplified during accretionnear surface disk in low optical depth turbulent motion suppressed by radiative flux similar to convection over photospheres stars outer radiative layer affect characteristic time matter advection in accretion disk determined by main turbulent part disk time field diffusion significantly changed electrical current concentrated in radiative highly conductive regions generate main magnetic field structure magnetic field with outer radiative layers shown in figure 2 inside turbulent disk electrical current negligibly small magnetic field almost fully vertical in outer radiative layer field diffusion small matter advection to strong magnetic field amplification field amplification until magnetic forces in over photosphere become gravitational in equilibrium balance mhd and plasma instabilities developed decreasing effective electrical conductivity in stationary state magnetic forces support optically thin regions against gravity in nonrotating magnetized disk forces support whole disk against gravity higher when magnetic force balances gravitational force on outer optically thin part disk of surface density relation surface density over photosphere corresponds to layer with effective optical depth close to @xmath42estimate lower limit magnetic field strength taking @xmath43 instead opacity @xmath44 writing @xmath45 obtain @xmath46/@xmath47 opacity thomson scattering @xmath48 @xmath49/g absorption opacity @xmath50 less than @xmath43 inner regions luminous accretion disk using equation @xmath40 estimate lower bound large - scale magnetic field keplerian accretion disk @xmath51 @xmath52 @xmath53 surface density @xmath54 inner radiation dominated region largest values magnetic fields @xmath55 @xmath56 maximum magnetic field reached when outward magnetic force balances gravitational force disk surface mass density @xmath40 equilibrium @xmath32 keplerian accretion disk @xmath58 times less than maximum value for @xmath59 and @xmath60 important question energy density large - scale stationary magnetic field with rotational gravitational energy density disk non - rotating magnetized accretion disk energy density field gravitational magnetic field strength large vicinity black hole larger than external seed field fully turbulent keplerian disk poloidal magnetic field outward1994 value not exceed strength large - scale seed magnetic field suggestion fully turbulent accretion disk small turbulent conductivity violated in outer surface layer optical depth small turbulence suppressed by strong radiative flux similar to radiative layer above main body convective star sun - kogan 2001 radiative layer magnetic field diffusion slower than fully developed turbulence determined by classical coulomb conductivity large diffusion negligible electric current concentrated main body turbulent disk almost current - free force - free magnetic field lines almost straight figure 2 negligible diffusion large - scale field drifts inward until comparable with centrifugal gravitational forces drift halted stationary state formed magnetic centrifugal gravitational forces comparable deviations from keplerian angular velocity significant strength magnetic field smaller than nonrotating disk of bisnovatyi - kogan and ruzmaikin 1976 large in vicinity of black hole nonuniform distribution of angular velocity over disk thickness main body turbulent disk rotates with velocity close to keplerian outer thin layers rotate slower models of rotating accretion disks large - scale magnetic field requires equations of magnetohydrodynamicsexpect two solutions same input parameters turbulent disk without radiative layers large magnetic field close to seed field disk angular velocity close to keplerian second solution magnetic field large exceed seed field angular velocity deviate from keplerian law radiative layers small field stationary transition to strong field strong field solution only stable stationary for rotating accretion disk investigation strong large - scale fields accretion underway rothstein et al . 2007 ) work g.b . k . supported by rbfr grants 05 - 02 - 17697 06 - 02 - 90864 ran program p-04 r.l . supported nasa grants - 13220 - 13060 nsf grant ast-0507760 | 0.499841 | 3,234 | 1,599 | , Saving $0.1 in GPT-4. |
20,920 | we study the oscillatory behaviour of a gene regulatory network with interlinked positive and negative feedback loop .
frequency and amplitude are two important properties of oscillation . studied network produces two different modes of oscillation . in one mode ( mode 1 ) frequency
remains constant over a wide range amplitude and in other mode ( mode 2 ) the amplitude of oscillation remains constant over a wide range of frequency .
our study reproduces both features of oscillations in a single gene regulatory network and show that the negative plus positive feedback loops in gene regulatory network offer additional advantage .
we identified the key parameters / variables responsible for different modes of oscillation .
the network is flexible in switching between different modes by choosing appropriately the required parameters / variables .
department of physics , akpc mahavidyalaya , subhasnagar , bengai , hooghly , pin-712611 , w.b . , | the fundamental unit of life is the cell . organisms may consist of just one cell or they may be multicellular . the multicellular organisms organized into tissues which are groups of similar cells arranged so as to perform a specific function .
one may view cell life as a collection of networks interacting through proteins , rna , dna and small molecules involved in signaling and energy transfer .
these networks process environmental signals , induce cellular responses and execute internal events such as gene expression , thus allowing cells and entire organisms to perform their basic functions .
these control and communication networks can be relatively simple ( in bacteria ) or they may be incredibly sophisticated ( in higher organisms ) .
in addition to their own needs for survival and reproduction , cells in multicellular organisms need additional levels of complexity in order to enable communication among cells and overall regulations . in living organism , proteins are the functional molecules .
they are synthesized in a regulated processes known as gene expression ( ge ) .
so , gene expression and regulation are of fundamental importance in cell .
again , proteins from one gene regulate the expression from other . in this way
, gene regulatory networks have grown inside the cell .
there can be other type of networks like metabolic networks , protein - protein interaction networks etc@xmath0 .
in general , the structure or architecture of the networks determines the function of the networks@xmath1 .
it is observed that positive and negative feedback loops are very common motifs in biological networks .
they occur frequently in different gene regulatory and cell signaling circuits . in general
, it is known that positive feedback loop induces a switch like behaviour and bistability and that negative feedback loop produces oscillations , suppresses noise / fluctuation effects etc .
the loops are often coupled to perform various functions in the networks acting as bistable switches , oscillators , excitable devices etc.@xmath2 .
rhythmic phenomena represent one of the most striking manifestations of dynamic behaviour in biological systems@xmath3 .
cellular rhythms are generated by complex interactions among genes , proteins and metabolites .
they are used to control signaling , motility , growth , division and death .
these rhythms appear in many regulatory mechanisms that control the dynamics of living system .
for example , neural and cardiac rhythms are associated with the regulation of voltage dependent ion - channels , metabolic oscillations originate from the regulation of enzyme activity and intracellular calcium oscillations involve the control of transport process while regulation of gene expression underlies circadian rhythms at the cellular level .
there are some essential requirements for biochemical oscillations@xmath4 .
in the course of time , open systems that exchange matter and energy with their environment generally reach a stable steady state . however , once the system operates sufficiently far from equilibrium and when its kinetics acquires a sufficient nonlinearity , the steady state may become unstable .
feedback processes and cooperativity are two main sources of nonlinearity that favour the occurrence of instabilities in biological systems .
when the steady state becomes unstable , the system moves away from it , often bursting into sustained oscillations around the unstable steady state .
theoretical analysis shows that a negative feedback networks with sufficient amount of time delay and nonlinearity produces oscillations@xmath5 . the time delay in the networks can be created by a long chain of intermediate reactions or by an extra positive feedback loop .
different types of interlinked positive and negative feedback loops are observed in cellular systems with different number of nodes and links ( fig .
1 ) . such coupled loops play a variety of roles , acting as bistable switches , oscillators etc . , although a single positive and negative feedback loop can also perform these functions under certain conditions .
it is demonstrated that coupled or interlinked feedback loops are superior to single feedback loops as oscillators@xmath6 .
a superior oscillator has the property of constant amplitude over a wide range of frequency .
there may be another type of oscillation in which frequency remains constant though amplitude of oscillations may varied .
constant amplitude oscillations are important in heart beat , cell cycle etc .
for circadian oscillations frequency should remain constant in different environmental conditions . here
we study a gene regulatory network which show both type oscillations depending on the variation of appropriate parameter .
we consider a gene regulatory network consist of three genes x , y and z which synthesizes three proteins @xmath7 , @xmath8 and @xmath9 respectively .
three genes form a closed loop structure and the product of each gene represses the synthesis process from other in a cyclic way starting from x to y to z. in addition to that there is a autocatalytic positive feedback loop in x. the network architecture is identical to the module considered by tsai et al.@xmath6 .
only difference is that , the network module considered by tsai et al .
is regulated at the degradation level but in our network the regulation is achieved at the synthesis level .
the notwork is shown in fig .
the dynamics of the network is driven by the following coupled nonlinear differential equation .
@xmath10 @xmath11 @xmath12 the equations contain basically three different kinds of terms , viz . , degradation , negative transcription or repression and autocatalysis .
the oscillatory behaviour ( figs . 2 and 3 ) of the interlinked gene transcription regulatory network is studied by varying the different rate constants .
we solve the coupled nonlinear equations to observe oscillation numerically by runge - kutta 4 technique . to verify the stability of the network , we consider the random parameter values in the range given in table 1 .
we observe that 500 out of 8404 parameter sets ( 5.94% ) yielded the oscillations in presence of positive feedback loop . but in absence of autocatalytic positive feedback loop we observe that 500 out of 7937 parameter sets ( 6.3% ) yielded the oscillations .
the last result ( network without autocatalytic feedback loop ) is completely different from the result of tsai et al.@xmath6 . to study the role of positive feedback loop in the network we measure frequency and amplitude of oscillations from the 500 oscillatory data sets with different rate constants as variable .
then we take a particular set of rate constant chosen from the 500 sets of parameter values for which oscillations are observed .
when we vary the repression strength ( k@xmath13 ) on x , we observe that frequency remains constant over a wide range of amplitude . as the positive
feedback strength increases the range of amplitude of oscillation over which the frequency remains constant increases ( fig .
but if we vary the degradation rate constant ( k@xmath14 ) , we observe that amplitude remains constant over a wide range of frequency .
the autocatalytic loop in x increases that behaviour further ( fig .
same behaviour is observed when both k@xmath13 and k@xmath14 varies simultaneously ( fig .
this observation shows that the degradation rate has more impact on the oscillatory behaviour of the network .
7 shows that amplitude remains constant over a wide range of frequency when varied the autocatalytic positive feedback strength k@xmath15 .
1 . gene transcription regulatory network with interlinked positive and negative feedback loop .
time dependent oscillatory behaviour of x , y and z for the parameter values k@xmath13=266.152 ,
k@xmath14=5.730 , k@xmath16=331.660 , k@xmath17=3.681 , k@xmath18=494.232 , k@xmath15=9.168 , k@xmath13=1.280 , k@xmath14=0.982 , k@xmath16=0.959 , k@xmath17=18.882 , n@xmath13=2.658 , n@xmath14=2.048 , n@xmath16=2.512 , n@xmath17=3.940 .
3 . different phase plots corresponding to the oscillatory behaviour in fig . 2 fig .
4 . amplitude versus frequency plot when varied @xmath19 with different values of positive feedback strength shown in the graph .
the other rate constants are fixed at @xmath20=5.730 , @xmath21=331.660 , @xmath22=3.681 , @xmath23=494.232 , @xmath24=9.168 , @xmath25=1.280 , @xmath26=0.982 , @xmath27=0.959 , @xmath28=18.882 , @xmath29=2.658 , @xmath30=2.048 , @xmath31=2.512 , n@xmath17=3.940 .
5 . frequency versus amplitude plot with @xmath32 variation for different values of positive feedback strength shown in the graph .
the other rate constants are fixed at @xmath33=266.152 , @xmath21=331.660 , @xmath22=3.681 , @xmath23=494.232 , @xmath24=9.168 , @xmath25=1.280 , @xmath26=0.982 , @xmath27=0.959 , @xmath34=18.882 , @xmath29=2.658 , @xmath30=2.048 , @xmath31=2.512 , n@xmath17=3.940 .
frequency versus amplitude plot with @xmath32 and @xmath33(both ) variation for different values of positive feedback strength shown in the graph .
the other rate constants are fixed at @xmath21=331.660 , @xmath22=3.681 , @xmath23=494.232 , @xmath24=9.168 , @xmath25=1.280 , @xmath26=0.982 , @xmath27=0.959 , @xmath28=18.882 , @xmath29=2.658 , @xmath30=2.048 , @xmath31=2.512 , @xmath35=3.940 .
frequency versus amplitude plot with @xmath24variation .
the other rate constants are fixed at @xmath33=266.152 , @xmath32=5.730 , @xmath21=331.660 , @xmath22=3.681 , @xmath23=494.232 , @xmath24=9.168 , @xmath25=1.280 , @xmath26=0.982 , @xmath27=0.959 , @xmath28=18.882 , @xmath29=2.658 , @xmath30=2.048 , @xmath31=2.512 , @xmath35=3.940 .
.different rate constants with their ranges used to solve the coupled equation numerically . [
cols="^,^,^",options="header " , ]
we study a gene regulatory network with interlinked positive and negative feedback loop .
the loop produces two different modes of oscillation . in one mode ( mode 1 ) frequency
remains constant over a wide range amplitude and in other mode ( mode 2 ) the amplitude of oscillation remains constant over a wide range of frequency . for circadian rhythm , mode 1 oscillation is very important because organisms try to maintain a constant frequency of their daily clocks in spite of the variation of the environmental condition .
mode 2 oscillation is important for heart beat or cell cycle for which fixed amplitude of oscillations is very much crucial in different frequency region .
our study reproduces both features of oscillations in a single gene regulatory network and show that the negative plus positive feedback loops in gene regulatory network offer additional advantage .
we identified the key parameters / variables responsible for different modes of oscillation .
the network is flexible in switching between different modes by choosing appropriately the required parameters / variables .
therefore , gene regulatory networks with interlinked positive and negative feedback loops work as more superior oscillator rather than the signaling networks . | fundamental unit life is cell organisms one cell or multicellular multicellular organisms organized into tissues groups similar cells perform specific function cell life as collection of networks interacting through proteins rna dna small molecules in signaling energy transfer networks process environmental signals induce cellular responses execute internal events gene expression allowing cells organisms perform basic functions control communication networks be simple in bacteria or sophisticated in higher organisms cells multicellular organisms need additional complexity communication cells regulations living proteins are functional molecules synthesized in regulated gene expression gene expression regulation fundamental importance cell proteins from one gene regulate expression other gene regulatory networks grown inside cell other networks like metabolic networks protein - protein interaction networks structure of networks determines function positive negative feedback loops common in biological networks occur in gene regulatory cell signaling circuits positive induces switch like behaviour bistability negative produces oscillations suppresses noise fluctuation loops coupled to perform functions as bistable switches oscillators excitable devices rhythmic phenomena represent dynamic behaviour in biological systemscellular rhythms generated by interactions among genes proteins metabolites control signaling motility growth division death rhythms appear in regulatory mechanisms dynamics living system neural cardiac rhythms with voltage dependent ion - channels metabolic oscillations from enzyme activity intracellular calcium oscillations control transport process gene expression underlies circadian rhythms essential requirements for biochemical oscillations open systems exchange matter energy reach stable steady state once system operates far from equilibrium nonlinearity steady state may become unstable feedback processes cooperativity sources of nonlinearity instabilities steady state unstable system moves away into sustained oscillations around unstable shows negative feedback networks with time delay nonlinearity produces oscillations time delay created by intermediate reactions or extra positive feedback loop interlinked positive negative feedback loops observed in cellular systems nodes links coupled loops play as bistable switches oscillators single can perform functions under conditions coupled interlinked feedback loops superior to single feedback loops as oscillators superior oscillator has constant amplitude over wide range frequencyanother type oscillation frequency constant amplitude constant amplitude oscillations important in heart beat cell cycle circadian oscillations frequency constant environmental conditions study gene regulatory network both oscillations variation parameter gene regulatory network three genes x y z synthesizes three proteins @xmath7 genes form closed loop structure gene represses synthesis process x to y z autocatalytic positive feedback loop in x network architecture identical to module tsai et al.@xmath6 regulated at degradation level synthesis level fig dynamics network driven by coupled nonlinear differential equation @xmath10 equations contain three terms degradation negative transcription repression autocatalysis oscillatory behaviour figs 2 3 interlinked gene transcription regulatory network studied varying rate constants solve coupled nonlinear equations observe oscillation by runge - kutta 4 technique verify stability consider random parameter values range table 1 500 out of 8404 parameter sets ( 5.94% ) yielded oscillations positive feedback loop absence feedback 500 out of 7937 parameter sets ( 6.3% ) yielded oscillationslast result network without autocatalytic feedback loop different from tsai et al.@xmath6 study role positive feedback loop network measure frequency amplitude oscillations from 500 oscillatory data sets with different rate constants rate constant from 500 values vary repression strength k@xmath13 ) on x frequency remains constant over amplitude positive feedback strength increases amplitude increases vary degradation rate constant k@xmath14 amplitude remains constant over wide range frequency autocatalytic loop in x increases behaviour same behaviour when k@xmath13@xmath14 simultaneously degradation rate on oscillatory behaviour shows amplitude remains constant over frequency when varied autocatalytic positive feedback strength k@xmath15 gene transcription regulatory network with interlinked positive negative feedback loop time dependent oscillatory behaviour of x y z for parameter values k@xmath13=266.152=5.730.660=3.681=494.232=9.168.280=0.982=0.959@xmath17=18.882@xmath13=2.658=2.048=2.512=3.940 phase plots oscillatory behaviour in fig fig . 4 amplitude versus frequency plot varied @xmath19 different values positive feedback strength other rate constants fixed @xmath20=5.730 @xmath21=331.660=3.681=494.232=9.168.280=0.982=0.959 @xmath28=18.882.658=2.048=2.512@xmath17=3.940 frequency versus amplitude plot @xmath32 variation different values positive feedback strength other rate constants fixed @xmath33=266.152 @xmath21=331.660=3.681=494.232=9.168.280=0.982=0.959 @xmath34=18.882.658=2.048=2.512=3.940 frequency versus amplitude plot @xmath32 @xmath33 variation different values positive feedback strength other rate constants fixed at @xmath21=331.660=3.681=494.232=9.168.280=0.982=0.959 @xmath28=18.882.658=2.048=2.512=3.940 frequency versus amplitude plot @xmath24variation other rate constants fixed at @xmath33=266.152 @xmath32=5.730 @xmath21=331.660=3.681=494.232=9.168 @xmath25=1.280=0.982=0.959 @xmath28=18.882 @xmath29=2.658=2.048=2.512 @xmath35=3.940 rate constants ranges solve coupled equation numerically study gene regulatory network with interlinked positive negative feedback loop loop produces two modes oscillation one mode 1 frequency constant wide range 2 amplitude constant wide range for circadian rhythm mode 1 oscillation important constant frequency variation environmental condition mode 2 important for heart beat cell cycle fixed amplitude crucial frequency study reproduces both features oscillations in single gene regulatory network negative plus positive feedback loops offer additional advantage identified key parameters variables for different modes oscillation network flexible switching between modes required parameters gene regulatory networks with interlinked positive negative feedback loops work superior oscillator signaling networks | 0.352755 | 2,518 | 1,251 | , Saving $0.1 in GPT-4. |
20,927 | we demonstrate a gigapixel inline digital holographic microscopy using a consumer scanner .
the consumer scanner can maximally scan an a4 size image ( 297 mm @xmath0 210 mm ) with 4800 dpi ( @xmath1 m ) , theoretically achieving a resolution of @xmath2 gigapixels .
the system using a consumer scanner has a simple structure , compared with synthetic aperture digital holography using a camera mounted on a two - dimensional moving stage . in this demonstration
, we captured an inline hologram with @xmath3 pixels ( @xmath4 0.43 gigapixels ) .
in addition , to accelerate the reconstruction time of the gigapixel hologram and decrease the amount of memory for the reconstruction , we applied the band - limited double - step fresnel diffraction to the reconstruction .
99 u.schnars and w. juptner , `` direct recording of holograms by a ccd target and numerical reconstruction , '' appl.opt . , * 33 , * 2 , 179181 ( 1994 ) . m. k. kim , `` principles and techniques of digital holographic microscopy , '' spie reviews * 1 * , 018005 ( 2010 ) . d. j. brady , s. lim,``gigapixel holography , '' 2011 ico international conference on information photonics ( ip ) , 1 - 2 , ( 2011 ) .
j. r. fienup , a. e. tippie,``gigapixel synthetic - aperture digital holography , '' proc .
spie * 8122 * , 812203 ( 2011 ) .
s. o. isikman , a. greenbaum , w. luo , a.f .
coskun , a. ozcan , `` giga - pixel lensfree holographic microscopy and tomography using color image sensors , '' plos one * 7 * , e45044 ( 2012 ) .
g. zheng , x. ou , c. yang , `` gigapixel microscopy using a flatbed scanner , '' arxiv:1209.4045 ( 2012 ) .
j. g. sucerquia , w. xu , s. k. jericho , p. klages , m. h. jericho , and h. j. kreuzer , `` digital in - line holographic microscopy , '' appl .
opt .
45 , 836 - 850 ( 2006 ) .
j. g. sucerquia , w. xu , m. h. jericho , and h. j. kreuzer , `` immersion digital in - line holographic microscopy , '' opt .
lett .
31 , 1211 - 1213 ( 2006 ) .
n. okada , t. shimobaba , y. ichihashi , r. oi , k. yamamoto , m. oikawa , t. kakue , n. masuda , and t. ito , `` band - limited double - step fresnel diffraction and its application to computer generated hologram , '' opt . express * 21 * , 91929197 ( 2013 ) . f. zhang , i. yamaguchi , and l. p. yaroslavsky , `` algorithm for reconstruction of digital holograms with adjustable magnification , '' opt
. lett .
* 29 * , 16681670 ( 2004 ) .
t. shimobaba , j. weng , t. sakurai , n. okada , t. nishitsuji , n. takada , a. shiraki , n. masuda and t. ito , `` computational wave optics library for c++ : cwo++ library , '' comput .
. commun . * 183 * , 11241138 ( 2012 ) . | digital holographic microscopy ( dhm ) captures a hologram with an electronic device such as a cmos or ccd camera , and the captured hologram is reconstructed on a computer using diffraction calculation @xcite . in order to increase the field - of - view , lateral and depth resolution powers of the reconstructed image
, we need to capture a large hologram , for example , the amount of gigapixels achieved in recent researches , @xcite .
there are several methods available to capture a gigapixel hologram . in astronomy ,
a single ccd device with over 0.1 gigapixels has been achieved ; however such a large area ccd is expensive .
another method for acquiring a gigapixel hologram is synthetic aperture digital holography using a camera mounted on a two - dimensional moving stage ( or moving a reference light ) .
@xcite used a green light source and a color cmos sensor , which generally has a bayer filter with one red , two green and one blue pixels for the acquisition of color images . a hologram captured with
the cmos sensor and green light source lacks the pixels corresponding to the red and blue pixel due to the bayer filter .
the absent pixels are interpolated by adjacent pixels . recently
, gigapixel microscopy using a consumer scanner has been proposed @xcite .
the approach is excellent because the microscopy has a wide field - of - view , low - cost and simple structure ; however , it can not observe a sample in the depth direction . in this paper
, we demonstrate a gigapixel inline dhm using a consumer scanner , inherently observing a sample in three - dimensions .
the consumer scanner can maximally scan an a4 size image ( 297 mm @xmath0 210 mm ) with 4800 dpi ( @xmath1 m ) , theoretically achieving a resolution of @xmath2 gigapixels .
the dhm system using a consumer scanner has a simple structure , compared with synthetic aperture digital holography . in this demonstration
, we captured an inline hologram with @xmath3 pixels ( @xmath4 0.43 gigapixels ) .
in addition , to accelerate the reconstruction time of the gigapixel hologram and decrease the memory usage for the reconstruction , we applied the band - limited double - step fresnel diffraction ( bl - dsf ) @xcite to the reconstruction . in section 2 , we explain the gigapixel inline dhm using a consumer scanner and bl - dsf . in section 3 , we present the results .
section 4 concludes this work .
figures [ fig : system ] ( a ) and ( b ) show the outline and photograph of the gigapixel inline dhm using a consumer scanner , respectively .
inline dhm @xcite is capable of obtaining a hologram without using beam splitters and mirrors .
samples are placed between a light source and the hologram .
the diffracted and undiffracted lights by the samples are regarded as the object light and the reference light , respectively , then , the interference fringe of these lights generates a hologram .
inline dhm can control the area of the hologram and the magnification of the object light just by the location of the light source or the numerical aperture ( na ) of an objective lens . if using a holographic setup that requires beam splitters , lens and mirrors , a large aperture is needed because the scanner has a large aperture ; therefore , the inline dhm is suitable for gigapixel holograms .
we used a fiber - output laser with a wavelength of 405 nm .
although the output of the laser has a spherical wave , we used an objective lens with a magnification of @xmath010 and an na of 0.25 , to expand the angle of spread of the fiber - output laser .
we placed samples between the objective lens and the scanner .
the scanner captures an inline hologram by moving the image sensor of the scanner . as shown in fig.[fig : scanner ] , consumer scanners are mainly categorized by two types : `` ccd '' and `` contact image sensors ( cis ) '' scanners . regarding ccd scanners ( fig.[fig : scanner](a ) ) , they have a two - dimensional ccd sensor whose size is smaller than the scan surface ( cover glass ) ; therefore , a reduction optical system composed of some mirrors and a lens is required to reduce the image on the scan surface to the ccd . and
ccd sensors generally have a color filer for scanning color images ; however , the color filter disturbs the capturing of holograms @xcite .
in contrast , cis scanners ( fig.[fig : scanner](b ) ) have a simpler structure than ccd scanners because cis sensors are one - dimensional devices whose size is the same as the scan surface ; therefore , reduction of the optical system in ccd scanners is not required .
in addition , because cis sensors are one dimensional , the electronic circuit is simpler than ccd sensors ; therefore , the cis scanners can capture an image with 16bit / pixel maximally , while most ccd scanners capture an image only with 8bit or 12bit / pixel maximally .
cis sensors also do not need a color filter unlike ccs scanner because cis sensors capture a color image by switching rgb light sources in time - divisions ; therefore , cis scanners are not adversely affected by the color filter .
thus , we adopted a cis scanner in this research because cis scanners are suitable for hologram recording .
the reconstruction of gigapixel holograms is very time - consuming and requires huge memory usage .
we adopted bl - dsf @xcite , which is an effective way to obtain a reconstructed image from gigapixel holograms . in fourier optics ,
diffraction calculations are categorized by two forms : the first is convolution - based diffraction and the second is fourier transform - based diffraction . here
, we show the angular spectrum method ( asm ) as an example of convolution - based diffraction .
asm is expressed as follows : @xmath5 } \exp ( - 2\pi i z \sqrt{1/\lambda^2-f_x^2-f_y^2 } ) \biggl ] } , \label{eqn : conv_form}\ ] ] where @xmath6 is the wavelength of light , the operators @xmath7}$ ] and @xmath8}$ ] are the fast fourier and inverse fast fourier transform respectively , @xmath9 and @xmath10 indicate a source and destination planes , @xmath11 is the coordinate on the frequency domain and @xmath12 is the propagation distance .
the merit of convolution - based diffraction is that the sampling rates on the source and destination planes are the same ; however , the demerit is the need to expand the source and destination planes by zero - padding to avoid aliasing .
it takes large memory usage and long calculation time . to overcome this problem ,
double - step fresnel diffraction ( dsf ) has been proposed @xcite .
it calculates the light propagation between the source to destination planes by twice the fourier transform - based diffraction , via a virtual plane @xmath13 .
dsf does not need zero - padding because dsf is based on fourier transform - based diffraction .
in addition , although most fourier transform - based diffractions change the sampling rates on the source and destination plane , dsf does not change them . due to the original dsf incurring the aliasing noise under certain conditions , we proposed bl - dsf introducing the rectangular function for band - limitation to the original dsf .
bl - dsf is expressed as follows : @xmath14 } \biggl ] } \label{eqn : dsf}\end{aligned}\ ] ] where @xmath15 is the propagation distance from the source plane to the virtual plane , @xmath16 is the propagation distance from the virtual plane to the destination plane , @xmath17 .
the operator @xmath18 means the forward fft when the sign of @xmath12 is positive and the inverse type when it is negative .
@xcite how to determine the band - limiting area @xmath19 .
bl - dsf was implemented in our open - source wave optics library , cwo++ @xcite .
we show a hologram of the usaf1951 test target captured by the dhm and the reconstructed image .
figure [ fig : holo ] shows the hologram with @xmath3 pixels ( 0.43gpixels ) and the hologram is sampled at about 5.29 @xmath20 m .
the hologram is captured under the condition that the distance between the scanner and the sample is 50 cm and the distance between the sample and the light source is 15 cm . because the number of pixels in the hologram is much larger than the resolution of a display ( 1,920 @xmath0 1,080 ) , the decimated hologram by image processing is displayed .
the inset shows the raw hologram in a part of the hologram ( in the red square ) .
we can observe the interference fringe of the hologram .
although the scanner can obtain a hologram maximally in 16bit / pixel , we were unable to recognize the difference between the reconstructed images in 8bit and 16bit / pixel .
therefore , we used a hologram captured by 8bit / pixel in terms of the calculation time and the memory usage .
pixels and the hologram is sampled at about 5.29 @xmath20m.,width=491 ] the reconstructed images are shown in figure [ fig : reconst1 ] .
the observational area of fig.[fig : reconst1 ] ( a ) is about @xmath21 mm@xmath22 .
figure [ fig : reconst1 ] ( b ) shows the details of the red square in fig.[fig : reconst1 ] ( a ) .
figure [ fig : reconst1 ] ( c ) shows the red square in fig.[fig : reconst1 ] ( b ) in greater detail .
we can observe a reconstructed image in the resolution power of about @xmath23 m .
the movie shows the observation of the reconstructed image as magnifying the image . pixels .
( movie),width=491 ] in a large hologram , the calculation time and memory usage for the reconstruction are important issues .
we compared the calculation time and memory usage in the reconstruction of the hologram ( fig.[fig : holo ] ) using asm and bl - dsf . in asm ,
the calculation time and the memory usage were about 355 seconds and 12.6gbytes , while the calculation time and the memory usage of bl - dsf were about 177 seconds and 3.2gbytes , respectively .
the peak signal - to - noise ratio ( psnr ) between the reconstructed images of asm and bl - dsf is over 30 db .
figure [ fig : reconst2 ] shows a hologram that , records an ant and water - flea placed at 30 cm and 50 cm from the scanner , respectively , and the reconstructed images .
the number of pixels of the hologram is @xmath3 .
when focusing on the water - flea , the ant is unfocused . while , when focusing on the ant , the water - flea is unfocused .
we demonstrated gigapixel inline digital holographic microscopy using a consumer scanner . in the demonstration , we showed holograms with @xmath3 pixels ( @xmath4 0.43 gigapixels ) captured by a consumer scanner and the reconstruction images .
we believe that the scanner is a promising imaging device for holography . in the reconstruction of a large hologram
, we showed the effectiveness of bl - dsf in terms of the calculation time and the memory usage , compared with the convolution - based diffraction .
the consumer scanner can maximally scan an a4 size image ( 297 mm @xmath0 210 mm ) with 4800 dpi ( @xmath1 m ) , theoretically achieving the resolution of @xmath24 gigapixels .
we are attempting to obtain a larger hologram with our system .
this work is supported by japan society for the promotion of science ( jsps ) kakenhi ( grant - in - aid for scientific research ( c ) 25330125 ) 2013 , and kakenhi ( grant - in - aid for scientific research ( a ) 25240015 ) 2013 . | digital holographic microscopy captures hologram with camera hologram reconstructed on computer using diffraction calculation increase field - of - view lateral depth resolution reconstructed image need capture large hologram gigapixels several methods to capture gigapixel hologram in astronomy single ccd device with over 0.1 gigapixels achieved large expensive another method gigapixel hologram is synthetic aperture digital holography using camera on two - dimensional moving stage @xcite used green light source color cmos sensor bayer filter with for color hologram captured with cmos sensor green light source lacks pixels red blue pixel due bayer filter absent pixels interpolated by adjacent pixels gigapixel microscopy using consumer scanner proposed wide field - of - view low - cost simple structure can not observe sample in depth direction paper gigapixel inline dhm using consumer scanner observing sample in three - dimensions scanner can scan a4 size image with 4800 dpi achieving resolution of @xmath2 gigapixels dhm system simple structure synthetic aperture digital holography captured inline hologram with @xmath3 pixels43 gigapixels accelerate reconstruction time gigapixel hologram decrease memory usage applied band - limited double - step fresnel diffraction ( bl - dsf ) @xcite reconstruction section 2 gigapixel inline dhm using consumer scanner bl - dsf section 3 results section 4 concludes work figures ( a b ) show outline photograph gigapixel inline dhm using consumer scanner inline dhm @xcite hologram without beam splitters mirrors samples placed between light source hologram diffracted undiffracted lights regarded as object light reference light interference fringe generates hologram area hologram magnification object light by location light source numerical aperture objective lens holographic setup beam splitters lens mirrors large aperture needed scanner inline dhm suitable for gigapixel holograms used fiber - output laser wavelength 405 nm used objective lens with magnification @xmath010 na 0.25 expand angle spread laser placed samples between objective lens scanner scanner captures inline hologram by moving image sensor consumer scanners categorized by two types ccd contact image sensors cis ) '' scanners ccd scanners fig[fig : scanner](a ) two - dimensional ccd sensor smaller than scan surface reduction optical system mirrors lens required to reduce image scan to ccd ccd sensors have color filer for color color filter disturbs capturing holograms cis scanners(b ) simpler structure than ccd one - dimensional size same as scan surface reduction optical not required cis sensors one dimensional electronic circuit simpler than ccd cis capture image with 16bit / pixel maximally most ccd scanners 8bit or 12bit / pixel cis sensors need color filter capture color image by switching rgb light sources not affected by color filter adopted cis scanner suitable for hologram recording reconstruction of gigapixel holograms time - consuming requires memory usage adopted bl - dsf @xcite effective to obtain reconstructed image from in fourier optics diffraction calculations categorized convolution - based diffraction fourier transform - based diffraction angular spectrum method ( ) example of convolution - based diffractionasm expressed @xmath5 - 2\pi z\lambda^2-f_x^2-f_y^2\biggl conv_form @xmath6 wavelength light operators @xmath7 @xmath8 fast fourier inverse fast fourier transform @xmath9 @xmath10 source destination planes @xmath11 coordinate frequency domain @xmath12 propagation distance merit convolution - based diffraction sampling rates source destination planes same demerit need expand source destination planes zero - padding avoid aliasing large memory usage long calculation time double - step fresnel diffraction dsf ) proposed calculates light propagation source destination planes twice fourier transform - based diffraction virtual plane @xmath13 dsf need zero - padding based on fourier transform - based diffraction diffractions change sampling rates source destination dsf change original dsf aliasing noise proposed bl - dsf rectangular function band - limitation dsf dsf expressed @xmath14 }\biggl{eqn : dsf{aligned @xmath15 propagation distance source virtual plane @xmath16 propagation distance destination plane @xmath17operator @xmath18 means forward fft sign @xmath12 positive inverse type negative @xcite determine band - limiting area @xmath19 bl - dsf implemented in open - source wave optics library @xcite hologram usaf1951 test target captured by dhm reconstructed image figure fig holo shows hologram with @xmath3 pixels ( 0.43gpixels ) sampled at 5.29 @xmath20 m distance scanner sample 50 cm sample light source 15 cm number pixels hologram larger than resolution display ( 1,920 @xmath0 1,080 ) decimated hologram image processing displayed inset shows raw hologram red square ) observe interference fringe hologram scanner obtain hologram 16bit / pixel recognize difference between reconstructed images 8bit 16bit / pixel used hologram captured by 8bit / pixel calculation time memory usage hologram sampled at 5.29 @xmath20m reconstructed images shown in figure fig : reconst1 ] observational area fig. @xmath21 mm@xmath22 figure b shows details red square figure shows red square detail observe reconstructed image resolution power @xmath23 m movie shows reconstructed image magnifying pixels=491 large hologram calculation time memory usage important compared memory usage reconstruction hologram using asm bl - dsf asm calculation time memory usage 355 seconds 12.6gbytes bl - dsf 177 seconds 3.2gbytes peak signal - to - noise ratio between images asm bl - dsf over 30 db figure reconst2 shows hologram records ant water - flea at 30 cm 50 cm from scanner reconstructed images number pixels hologram @xmath3 focusing on water - flea ant unfocused demonstrated gigapixel inline digital holographic microscopy using consumer scanner showed holograms with @xmath3 pixels ( 0.43 gigapixels ) captured consumer scanner reconstruction images scanner promising imaging device for holography reconstruction large hologram showed effectiveness of bl - dsf calculation time memory usage compared with convolution - based diffraction consumer scanner can scan a4 size image with 4800 dpi achieving resolution @xmath24 gigapixels attempting to obtain larger hologram with systemwork supported by japan society promotion science kakenhi grant - aid scientific research c 25330125 ) 2013 kakenhi grant 25240015 ) 2013 | 0.429099 | 2,616 | 1,316 | , Saving $0.1 in GPT-4. |
20,933 | we fit the hires ultra - high energy cosmic ray ( uhecr ) spectrum measurements with broken power laws in order to identify features .
these fits find the previously observed feature known as the ankle at @xmath0 ev , as well as evidence for a suppression at higher energies , above @xmath1 ev .
we use the integral spectrum and the @xmath2 test to identify this high energy suppression with the gzk suppression .
finally , we use a model of uniformly distributed extragalactic proton sources together with a phenomenological model of the galactic cosmic ray spectrum to compare the hires spectra to what should be expected from the gzk suppression , and to measure how the extragalactic sources must evolve and what the input spectral slope must be to fit the hires data .
fits using updated spectra will be presented in pune . | the hires collaboration has recently released two measurements of the uhecr flux using monocular observations from its two sites@xcite .
broken power law ( bpl ) fits made to these measurements can be used to identify features in the spectrum and to estimate their statistical significance in the data .
all the fits presented in this paper are performed using the normalized binned maximum likelihood method@xcite .
this method requires comparing the numbers of events expected in a given model to the number actually observed .
the number of expected events is obtained from the predicted flux divided by the same exposure used to calculate the observed flux .
the numbers of events observed by hires are shown in figure [ fig : events ] .
the binned maximum likelihood method also allows one to use bins in which there were no observed events , but in which events were expected .
the result of the fits are expressed in terms of a quality - of - fit parameter @xmath3 which approaches a true @xmath3 in the limit of large numbers of events .
the bpl fits are only made to bins above @xmath4 ev . while hires - ii does have three bins below this energy , these bins have poor statistics .
we do nt fit these bins in order to avoid biases due to an expected change in the spectral slope , the second knee , at about this energy .
parameters from fits to the hires monocular spectra to a bpl , @xmath5 , with zero , one and two floating break points are shown in table [ brokenpowerlawfits ] . the result of the fit in the two floating break point case is shown in figure [ hr12_trilinear_m05 ] .
the hires spectra fit to a bpl with two break points .
the parameters of the fit are given in table [ brokenpowerlawfits ] .
the red line is used to calculate the significance of the second break point and to calculate the expected integral spectrum in the @xmath2 calculation . ] the hires spectra fit to a bpl with two break points .
the parameters of the fit are given in table [ brokenpowerlawfits ] .
the red line is used to calculate the significance of the second break point and to calculate the expected integral spectrum in the @xmath2 calculation . ]
.[brokenpowerlawfits ] parameters found in broken power law fits to the hires monocular spectra . [ cols="^,^,^,^,^,^,^",options="header " , ] the simple power law fit is clearly not a good fit . adding one floating break point gives a much better fit , and the break point finds the feature known as the ankle with a very high degree of statistical significance . adding a second floating break point improves the fit further .
the break point is found to be at approximately the energy expected of the gzk suppression@xcite .
the statistical significance of the second break in the spectrum can be estimated by looking at the reduction in the @xmath3 achieved by adding the break point , or by comparing the number of expected events above the second break point to what would be expected if the spectrum continued unabated above the second break point .
the reduction of 16 in the @xmath3 for two additional degrees of freedom corresponds to just under 4@xmath6 significance in a gaussian fit . using the red line in figure [ hr12_trilinear_m05 ]
, one would expect to see 28.0 events , where 11 events were actually observed .
the poisson probability of observing 11 events or fewer when expecting 28.0 is @xmath7 . for comparison ,
the area in one tail of a gaussian distribution outside of 4@xmath6 ( 3@xmath6 ) is @xmath8 ( @xmath9 ) .
so the significance of the high energy suppression is between 3@xmath6 and 4@xmath6 .
berezinsky _ et al._@xcite have suggested measuring the energy of a break in the uhecr spectrum finding the energy , @xmath2 , at which the observed integral spectrum is half of what one would expect with no break . the integral spectrum measured by hires , along with the expected integral spectrum using the red line in figure [ hr12_trilinear_m05 ] , is shown in figure [ integral - spectrum ] .
the ratio of the observed to the expected integral spectra is shown in figure [ integral - ratio ] . by interpolating between the hires - i points in figure [ integral - ratio ]
, we find an experimental value of @xmath10 ( @xmath11 in ev ) .
et al._@xcite have determined @xmath12 as what is to be expected for the gzk suppression for @xmath13 .
the observed - to - expected ratio for integral spectra .
the value of @xmath2 is obtained by a simple interpolation between the hires - i points as indicated by the blue lines . ]
the observed - to - expected ratio for integral spectra .
the value of @xmath2 is obtained by a simple interpolation between the hires - i points as indicated by the blue lines . ]
berezinsky _ et al._@xcite also calculate the energy loss rate for uhe protons traveling through the cosmic microwave background radiation , assuming continuous energy loss through electron pair production , pion production and universal expansion .
this can be used to predict the observed spectrum of extragalactic protons for a given input spectrum . since protons with energies above the pion production threshold lose energy in highly inelastic interactions
, we have used the monte carlo method of demarco _
et al._@xcite to model this part of the propagation of extragalactic protons .
protons from a shell at a given redshift are propagated from generation to observation ( at @xmath14 ) .
each input energy corresponds to a distribution of observed energies after such a propagation .
this @xmath15@xmath16 distribution convolved with the input spectrum gives the observed spectrum for protons from sources at a given redshift .
if the distribution of sources is assumed to be uniform at any given redshift , but to evolve with redshift as @xmath17 , one can combine the spectra from different shells as shown in for a coarse , logarithmic series of shells in figure [ shell - series ] . because the hires - ii spectrum extends to fairly low energies
, one should take into account an additional galactic component when fitting to this usm model .
we made the simple phenomenological assumption that the extragalactic and galactic components of the spectrum are expressed , respectively , in the light ( protons ) and heavy ( iron ) components of a composition measurement .
we use fits to the hires prototype / mia@xcite and hires stereo@xcite composition measurements with respect to qgsjet proton and iron expectations to determine the relative sizes of the extragalactic and galactic components .
then , for a given set of parameters for the extragalactic uhecr spectrum , we can add the appropriate galactic cosmic ray flux .
we varied the input spectral slope , @xmath18 , and evolution parameter , @xmath19 , to find the best fit of the hires monocular data to this usm - plus - galactic model .
the best fit spectrum is shown in figure [ usm - best ] . the final result for
the extragalactic usm model is @xmath20 , @xmath21 , where the systematinc uncertainties come from varying the extragalactic / galactic ratio within the limit allowed by the composition measurements .
best fit of the usm - plus - galactic model to the hires monocular spectra .
the red line shown the extragalactic component , the green line shows the galactic component and the black line shows the sum . ]
best fit of the usm - plus - galactic model to the hires monocular spectra .
the red line shown the extragalactic component , the green line shows the galactic component and the black line shows the sum . ] the fit works best in the region on either side of the ankle , with the fall at lower energies , into the ankle , primarily determining @xmath19 , and the rise at higher energies , out of the ankle , primarily determining @xmath18 .
the fit does not work as well in the region just below the gzk suppression .
this could be due to some sources having a maximum energy below the gzk threshold .
the is also little sign of a second knee at around @xmath4 ev , though the hires - ii data has little statistical power in this region .
the hires detector has observed the ankle and has evidence for a suppression at higher energies above @xmath1 ev .
the energy for this high energy suppression agrees with what is expected from the gzk suppression according to the @xmath2 test .
the observed spectra are well fit by a usm - plus - galactic model , which finds an input spectral slope for extragalactic protons of @xmath20 , and an evolution parameter @xmath22 .
this work is supported by us nsf grants phy-9321949 , phy-9322298 , phy-9904048 , phy-9974537 , phy-0098826 , phy-0140688 , phy-0245428 , phy-0305516 , phy-0307098 , and by the doe grant fg03 - 92er40732 .
we gratefully acknowledge the contributions from the technical staffs of our home institutions .
the cooperation of colonels e. fischer and g. harter , the us army , and the dugway proving ground staff is greatly appreciated .
99 r. abbasi _ et .
al . _ , to appear in phys .
b. , astro - ph/0501317 ; see also http://www.physics.rutgers.edu/~dbergman/hires-monocular-spectra.html .
s. eidleman _ et al .
_ , phys . lett . *
b592 * , 1 ( 2004 ) .
k. greisen , phys .
* 16 * , 748 ( 1966 ) ; g.t .
zatsepin and v.a .
kuzmin , pisma zh .
. fiz . * 4 * , 114 ( 166 ) [ jetp lett . * 4 * , 78 ( 1966 ) ] .
v. berezinsky , a.z .
gazizov and s.i .
grigoreva , hep - ph/0204357 .
d. de marco , p. blasi and a.v .
olinto , astropart . phys . *
20 * 53 ( 2003 ) . t. abu - zayyad _ et al .
84 * , 4276 , ( 2000 ) .
r. abbasi _ et al .
_ , astrophys .
j. * 622 * , 910 ( 2005 ) . | hires collaboration released two measurements uhecr flux using monocular observations from two sites@xcite broken power law bpl ) fits measurements identify features spectrum estimate statistical significance fits performed using binned maximum likelihood method events expected to observed expected events from predicted flux divided by exposure observed flux events observed shown in figure events allows bins no observed events but events expected result fits expressed quality - of - fit parameter @xmath3 approaches true @xmath3 large bpl fits made to bins above @xmath4 ev hires - ii three bins below this energy poor statistics fit bins avoid biases due expected change spectral slope parameters from fits to hires monocular spectra bpl @xmath5 with zero two floating break points shown in table [ brokenpowerlawfits result of fit in two floating break point case shown in figure hr12_trilinear_m05 hires spectra fit to bpl with two break points parameters fit in table [ brokenpowerlawfits red line significance second break point expected integral spectrum in @xmath2 calculation hires spectra fit to bpl with two break points parametersred line significance second break point expected integral spectrum in @xmath2 calculation[brokenpowerlawfits parameters monocular spectra cols="^^",options="header " simple power law fit not good adding one floating break point better fit finds ankle high statistical significance second break point improves fit break point approximately energy expected gzk suppression@xcite statistical significance second break estimated reduction in @xmath3 adding break point comparing expected events above second break point to spectrum reduction of 16 in @xmath3 for two additional degrees freedom under 4@xmath6 significance in gaussian fit red line in figure hr12_trilinear_m05 28.0 events 11 events observed probability observing 11 events or fewer 28.0 is @xmath7 area one tail gaussian distribution outside 4@xmath6 ( 3 is @xmath8 ( @xmath9 ) significance high energy suppression between 3@xmath6 and 4@xmath6 berezinsky et al. measuring energy break in uhecr spectrum energy @xmath2 observed integral spectrum half of no breakintegral spectrum measured by hires expected spectrum red line figure hr12_trilinear_m05 shown in figure integral - spectrum ratio observed to expected spectra shown in figure integral - ratio interpolating hires - i points experimental value of @xmath10 ( @xmath11 determined @xmath12 expected for gzk suppression @xmath13 observed - to - expected ratio for integral spectra value @xmath2 obtained by interpolation between hires - i points blue lines calculate energy loss rate for uhe protons through cosmic microwave background radiation continuous energy loss through electron production expansion observed spectrum of extragalactic protons input spectrum protons above pion production threshold lose energy in inelastic interactions used monte carlo method of model propagation of extragalactic protons protons from shell at redshift propagated from generation to observation @xmath14 each input energy corresponds to distribution of observed energies after propagation @xmath15@xmath16 distribution spectrum gives observed spectrum for protons from sources at redshiftdistribution sources uniform at redshift with redshift @xmath17 combine spectra from shells logarithmic series figure [ shell - series hires - ii spectrum extends low energies additional galactic component fitting usm model assumption extragalactic galactic components expressed in light protons heavy ( iron ) components composition measurement use fits hires prototype / mia@xcite stereo@xcite composition measurements proton iron expectations determine relative sizes extragalactic galactic components extragalactic uhecr spectrum add galactic cosmic ray flux varied input spectral slope @xmath18 evolution parameter @xmath19 best fit hires monocular data usm - plus - galactic model best fit spectrum shown in figure [ usm - best ] final result extragalactic usm model @xmath20 , @xmath21 systematinc uncertainties from varying extragalactic / galactic ratio within limit allowed composition measurements best fit usm - plus - galactic model to hires monocular spectra red line extragalactic component green line galactic black line shows sum fit monocular spectrafit works best either side ankle fall lower energies into ankle determining @xmath19 rise higher energies out ankle determining @xmath18 fit work below gzk suppression due sources maximum energy below gzk threshold little sign second knee around @xmath4 ev hires - ii data little statistical power hires detector observed ankle evidence suppression higher energies above @xmath1 ev high energy suppression agrees with gzk suppression @xmath2 test observed spectra fit by usm - plus - galactic model input spectral slope for extragalactic protons @xmath20 evolution parameter @xmath22 work supported by nsf grants doe grant fg03 - 92er40732 acknowledge contributions technical staffs cooperation of colonels e. fischer g. harter us army dugway proving ground staff appreciated r. abbasi et . al . phys . b. astro - ph/0501317.physics.rutgers.edu/hires-monocular-spectra eidleman et al . phys . lettb592 1 2004 ) greisen . 16 748 ( 1966 ) zatsepin kuzmin pisma fiz 4 114 ( 166 ) jetp lett 4 78 ( 1966 ) berezinsky gazizov grigoreva hep - ph/0204357 de marco blasi olinto astropart . 20 53 ( 2003 ) abu - zayyad 84 4276 2000 ) abbasi astrophys. 622 910 ( 2005 ) | 0.349848 | 2,331 | 1,157 | , Saving $0.1 in GPT-4. |
20,934 | the symmetry group analysis is applied to classify the phonon modes of @xmath0-stacked graphene layers ( nsgl s ) with ab- and aa - stacking , particularly their infra - red and raman properties .
the dispersions of various phonon modes are calculated in a multi - layer vibrational model , which is generalized from the lattice vibrational potentials of graphene to including the inter - layer interactions in nsgl s .
the experimentally reported red shift phenomena in the layer number dependence of the intra - layer optical c - c stretching mode frequencies are interpreted .
an interesting low frequency inter - layer optical mode is revealed to be raman or infra - red active in even or odd nsgl s respectively .
its frequency shift is sensitive to the layer number and saturated at about 10 layers . | in recent years , the ultrathin graphite films , i.e. @xmath0 ( @xmath1 ) stacked graphene layers ( nsgl s ) , are successfully fabricated.@xcite extensive studies have been devoted to these systems due to their competitive capability for the design of novel nano - devices .
it s desirable to carry out a through symmetry analysis to figure out the corresponding electron and phonon spectra , and to reveal the relevant selection rules and the optical activities .
moreover , to our knowledge , although there are some studies on the electronic structure,@xcite no existing works concern the phonon dispersions for the nsgl s .
recently it is reported that the frequencies of optical c - c stretching mode in the nsgl s decrease with increasing @xmath0.@xcite the amplitude of this red shift is about @xmath2 , @xmath3 , and 8 @xmath4 in these three experiments respectively .
theoretical explanation of this red shift is required . in this paper ,
the symmetry analysis is referred for both the ab- and aa - stacked lattice structures , while the latter has the point group @xmath5 irrespective of the even - oddness of @xmath0 .
we classify the phonon normal modes at @xmath6 point and determine their raman and infra - red ( ir ) properties .
we generalize the force constant model of graphene@xcite into nsgl s and it can be applied to calculate the phonon dispersions for the nsgl s in ab- or aa - stacking with arbitrary layer number @xmath0 . for the intra - layer optical c - c stretching mode with frequency around 1600 @xmath4 , it is raman active for all nsgl s and the calculated frequencies exhibit layer - number dependence as that the frequency decreases with @xmath0 increasing .
the red shift values for the ab- ( aa- ) stacked systems are about 2 @xmath4 ( 4 @xmath4 ) which are in consistent with the experimental measurements . in the medium frequency range around 800 @xmath4 ,
the out - of - plane optical mode is ir active for ab - stacked structure but neither raman nor ir active in aa - stacking .
its frequency behaviors a blue shift as layer number increasing .
there is an interesting inter - layer optical mode in the low frequency region which is raman active in the nsgl s with @xmath0 even ( ensgl s ) while ir active in nsgl s with @xmath0 odd ( oensgl s ) .
its frequency value depends on the layer number @xmath0 more sensitively and increases from 106 @xmath4 ( 94.5 @xmath4 ) to 149.8 @xmath4 ( 133.6 @xmath4 ) for the ab- ( aa- ) stacked nsgl s , which is of an order less than those intra - layer optical modes .
phonon dispersions for the aa - stacked 3-dimensional ( 3d ) graphite are also discussed .
the present paper is organized as follows . in sec .
ii , the lattice configuration is illustrated for the nsgl s .
iii is devoted to the symmetry analysis for the phonon modes .
the vibrational potential energy is discussed in subsec .
iv a , while the main results and relevant discussions on phonon spectrum calculations are presented in subsec .
iv b. the paper ends with a brief summary in sec .
[ 1.4 ] it is known that graphene is a single layer of carbon atoms with the honeycomb lattice configuration which is characterized by the @xmath5 symmetry.@xcite the 3d graphite is ab - stacked honeycomb lattice , where the b layers are achieved by shifting the a layers along one of its first - nearest carbon - carbon bonds in the horizontal plane as shown in fig . [ fig : unitcell](a ) .
the space group of the 3d graphite is non - symmorphic group @xmath7 with non - primitive translation @xmath8 ( primitive translation @xmath9).@xcite the distance between two adjacent layers is about @xmath10 which is much larger than the bond length between two nearest - neighbor atoms in the plane , @xmath11 .
the ngsl s are also constructed by ab - stacked honeycomb lattice , but with limited number of layers . although the structures of each layer of 3d graphite and nsgl s are the same , the corresponding symmetry groups are different since the displacement symmetry along @xmath12 axis no longer exists for nsgl s , so does the symmetry associated with @xmath13 .
now the symmetry group becomes a direct product of a 3d point group and a 2d translational group .
and the point groups are different for ensgl s and onsgl s as mentioned in ref . .
for the ensgl s , a center of inversion @xmath14 occurs in the middle of atom 4 in the @xmath15-th layer and 5 in the @xmath16-th layer as shown in fig .
[ fig : unitcell](a ) .
there are one 3-fold main axis in the direction perpendicular to the layers and three 2-fold axes , @xmath17 , perpendicular to the main axis and at angles of @xmath18 to each other .
all these symmetry operations together constitute the point group @xmath19 . in the onsgl s , instead of a center of inversion
there is a reflection symmetry @xmath20 with the middle layer as its reference plane .
a 3-fold main axis exists in the direction of @xmath21-axis and three 2-fold axes , @xmath22 , are perpendicular to it .
we notice that these three 2-fold axes @xmath22 are one to one perpendicular to those of @xmath17 .
consequently , the symmetry group for the onsgl s is @xmath23 .
the environments for an atom in the graphite and nsgl s are different from that in 2d graphene . for each carbon atom in the graphene layer , there are three nearest - neighbor carbon atoms and six next - nearest- neighbors
. there are four carbon atoms 1 , 2 , 3 , 4 in a unit cell of graphite , as represented in fig .
[ fig : unitcell](a ) . for atom 4 in a layer , there are two inter - layer nearest - neighbors in each of the two adjacent layers , with the distance @xmath24 .
also , in each of the two adjacent layers , there are three inter - layer next - nearest - neighbor atoms around atom 4 with distance @xmath25 . as illustrated in fig .
[ fig : unitcell](a ) , the adjacent environment for atom 3 is quite different from that of atom 4 .
it has no inter - layer first - neighbors with the same distance as @xmath26 . however , atom 3 has six inter - layer neighbors in each of the two adjacent layers with the same distance as that of the second - nearest - neighbor of atom 4 .
the aa - stacked nsgl s ( aa - nsgl s ) are constructed by aa - stacked honeycomb lattice where all layers have the same configuration . in the aa - stacked system , the ensgl s , onsgl s and the 3d graphite have the same point group @xmath5 which is the symmetry of the graphene . as shown in fig .
[ fig : unitcell](b ) , the environments for carbon atoms in the aa - stacked nsgl s are quite different from those in ab - stacked systems . for each atom , there are two inter - layer nearest - neighbors in each of the two adjacent layers with the distence @xmath26 . and
each atom has six inter - layer second - nearest - neighbors with distance @xmath25 in its two adjacent layers .
we notice here that in the aa - stacked 3d graphite , there are only two atoms in the unit cell and the primitive translation along @xmath24-axis is @xmath27 , which is half of the correspondence in the ab - stacked 3d graphite .
the dynamical representation @xmath28 can be decomposed into the irreducible representations of the symmetry group , with the lattice displacements as the bases , where @xmath29 is the vector representation and @xmath30 is the permutation representation of the group . by applying the projection operator technique ,
we carry out the decomposition of the dynamical representation into the irreducible representations for the nsgl s with @xmath0 even and odd respectively . according to elliott,@xcite
the ir active phonon modes belong to the irreducible representations decomposed from the vector representation @xmath29 , while the raman active modes correspond to the irreducible representations shown up in the decomposition of a six - dimensional representation with bases as the quadratic forms : @xmath31 , @xmath32 , @xmath33 , @xmath34 , @xmath35 , and @xmath36 .
the three acoustic modes with zero frequency at the @xmath6 point , which correspond to the vector representation @xmath29 , are excluded in the consideration of ir and raman active modes . for comparison
the corresponding results for the graphene and 3d graphite are also listed in the following .
the symbol for the irreducible representations we used here is the notation used in ref . which is the most commonly used in the treatment of molecules .
[ cols="<,^,^,^,^,^",options="header " , ] [ 1.0 ] .
lines are draw to guide eyes.,title="fig:",width=287 ] [ 1.0 ] .
this mode is ir active in the ab stacking while it is neither ir nor raman active in the aa stacking .
lines are draw to guide eyes.,title="fig:",width=287 ] [ 1.2 ] .
datas for the ab and aa stacked nsgl s are designated by pentagrams and circles , respectively .
the raman and infra - red activities for this mode are displayed by the full and empty symbols , respectively .
the broken and dashed lines correspond to the frequencies for the ab - stacked and aa - stacked graphite , respectively.,title="fig:",width=287 ] [ 1.0 ] with the above force constants , we can calculate the phonon dispersion for nsgl s of the aa- or ab - stacking with an arbitrary layer number @xmath0 . in fig .
[ fig : modeintralayer ] , the calculated frequency of the intra - layer optical c - c stretching mode is represented with different stacked styles and layer number @xmath0 .
the layer dependence of the frequency shows up a red shift behavior which is in agreement with the experimental measurements .
the frequency value for this mode is about 1588 @xmath4 in the single graphene layer and decreases with increasing @xmath0 and almost saturates at @xmath37 .
the limit is 1586.7 @xmath4 ( 1584.7 @xmath4 ) in the ab- ( aa- ) stacked system respectively .
the amount of red shift value in our calculation corresponds excellently with that measured by experiments within the ranges @xmath2 , @xmath38 , and 8 @xmath4 in refs . , and , respectively .
the out - of - plane optical mode , belonging to the @xmath39 ( @xmath40 ) irreducible representation in the ab ( aa ) stacking , is ir active in the ab - stacking yet inactive in the aa - stacking irrespective of the even - oddness of the layer number @xmath0 and is useful in determining whether the nsgl s is of ab or aa stacking . as shown in fig .
[ fig : modeout ] , frequencies for this mode depend on the layer number @xmath0 and increase from 864.8 @xmath4 to 872.6 @xmath4 in both the ab and aa stacking .
in contrast to the c - c stretching optical mode , this mode frequency exhibits a blue shift type layer dependence which could be identified with the development of the experimental technique . for the inter - layer optical mode ,
the layer number dependence of the frequency value is shown in fig .
[ fig : modeinterlayer ] .
this mode takes the greatest advantage of the inter - layer interation and is considerably dependent on the layer number @xmath0 and the stack style ab or aa . in case of @xmath41
, the @xmath42 mode has the frequency values 106 @xmath4 and 94.5 @xmath4 for the ab- and aa - stacked nsgl s respectively .
the frequencies of the @xmath43 mode increase with increasing @xmath0 and almost come to the limit values at @xmath37 .
the limit values are 149.8 @xmath4 and 133.6 @xmath4 for the ab- and aa - stacked nsgl s respectively .
the frequency differences as well as the raman versus ir ( see sec .
iii ) of the @xmath42 mode in nsgl s with different layers might inspire considerably experimental interest in the @xmath43 mode .
we then calculate the phonon dispersion for the 2-layered ab stacking , in comparison with that of the 3d graphite .
the most significant difference between the 2-layer graphene and the 3d graphite lies in the low - frequency region around the @xmath6 point as shown in fig .
[ fig:2layer3dlow ] .
the frequencies of the low - frequency optical modes in the 2-layer graphene are much smaller than their counterparts in the 3d graphite .
the frequencies of the raman and ir active modes are shown in the third line of table [ tab : raman ] among which the two @xmath44 modes have the frequency value @xmath42=106 @xmath4 and @xmath45=867.4 @xmath4 .
in fact , @xmath42 and @xmath45 modes are the above mentioned @xmath46 modes of the ensgl s specified to @xmath41 . [ 1.0 ] we further calculate the phonon dispersion curve for the aa - stacked 3d graphite as shown in fig .
[ fig : aa3dgraphite ] .
since the unit cell contains only two atoms in contrast to that of the ab - stacked graphite , there are six branches of phonon dispersion . along @xmath47 in the brillouin zone , the lowest and highest branches which correspond to the in - plane acoustic and optical vibrational modes are doubly degenerate , while the remaining two branches describing the out - of - plane vibration are non - degenerate . at the @xmath48 point in the brillouin zone
, there is a phase factor difference of @xmath49 between two adjacent layers for the out - of - plane motion , among which the @xmath42 mode has the frequency value of 133.6 @xmath4 . in the fourth line of table [ tab : raman ] , the raman and ir active modes for the aa - stacked 3d graphite are listed which are to be confirmed in future experiments .
based upon a thorough investigation of the lattice symmetry of the nsgl s , the raman and ir properties , in particular its layer dependence , is systematically studied . with a proposed generalized vibrational potential ,
we further calculate the phonon dispersion of various modes of the ab- or aa - stacked nsgl s where the layer dependence is also stressed .
the calculated frequencies of optical c - c stretching mode exhibit a red shift as layer number increasing in both the ab- and aa - stacked nsgl s , and the shift value 2 @xmath4 ( 4 @xmath4 ) for ab- ( aa- ) stacked nsgl s is in good consistent with the experimental measurements . the out - of - plane optical mode with frequency
around about 800 @xmath4 is ir active in ab - stacked structure yet neither raman nor ir active in aa - stacking .
its frequency shows a blue shift layer dependence .
we also predict that the frequency of the inter - layer optical mode increases with @xmath0 increasing .
since this mode is more sensitive to the layer number @xmath0 , it should be experimentally interesting in determining the lattice structure properties of the nsgl s . | ultrathin graphite films ) stacked graphene layers fabricated extensive studies devoted systems competitive capability design novel nano - devices desirable symmetry analysis electron phonon spectra reveal selection rules optical activities no works concern phonon dispersions nsgl s frequencies optical c - c stretching mode decrease increasing @xmath0 amplitude red shift about @xmath2 @xmath3 8 @xmath4 experiments theoretical explanation required symmetry analysis for ab- aa - stacked lattice structures point group @xmath5 classify phonon normal modes at @xmath6 point determine raman infra - red properties generalize force constant model into nsgl s calculate phonon dispersions ab- or aa - stacking arbitrary layer number intra - layer optical c - c stretching mode frequency around 1600 @xmath4 raman active for all nsgl s calculated frequencies exhibit layer - number dependence decreases with @xmath0 increasing red shift values ab- ( aa- ) stacked systems about 2 @xmath4 ( 4 @xmath4 ) consistent with experimental measurementsmedium frequency range 800 @xmath4 out - of - plane optical mode ir active ab - stacked structure neither raman nor ir active aa - stacking frequency layer number increasing inter - layer optical mode low frequency region raman active nsgl s @xmath0 even ir active @xmath0 odd frequency value depends on layer number increases from 106 to 149.8 ab- ) stacked nsgl s less than intra - layer optical modes phonon dispersions aa - stacked 3-dimensional graphite discussed paper organized sec ii lattice configuration nsgl s iii symmetry analysis phonon modes vibrational potential energy discussed subsec iv a main results discussions phonon spectrum calculations subsec iv b ends summary 1.4 graphene single layer carbon atoms honeycomb lattice configuration @xmath5 symmetry 3d graphite ab - stacked honeycomb lattice b layers achieved shifting a layers first - nearest carbon bonds horizontal plane fig .space group 3d graphite non - symmorphic @xmath7 non - primitive translation @xmath8 primitive @xmath9). distance between layers @xmath10 larger than bond length between atoms @xmath11 ngsl s constructed by ab - stacked honeycomb lattice limited layers structures 3d graphite nsgl s same symmetry groups different displacement symmetry @xmath12 axis exists nsgl s symmetry @xmath13 symmetry product 3d point group 2d translational group point groups different for ensgl s onsgl s ensgl s center of inversion @xmath14 middle atom 4 @xmath15-th 5 @xmath16-th layer fig one 3-fold main axis perpendicular layers three 2-fold axes @xmath17 perpendicular main axis angles @xmath18 symmetry operations constitute point group @xmath19 onsgl s center inversion reflection symmetry @xmath20 middle layer reference plane 3-fold main axis @xmath21-axis three 2-fold axes @xmath22 perpendicular axes @xmath22 perpendicular to @xmath17 symmetry group onsgl s @xmath23 environments atom in graphite nsgl s different 2d graphene each carbon atom in graphene layer three nearest - neighbor six next - nearest neighbors four carbon atoms 1 2 3 4 in unit cell graphite atom 4 two inter - layer nearest - neighbors in adjacent layers distance @xmath24 three layer next - nearest - neighbor atoms around atom 4 distance @xmath25 adjacent environment atom 3 different from atom 4 no inter - layer first - neighbors same distance as @xmath26 3 six inter - layer neighbors in layers same distance as - 4 aa - stacked nsgl s constructed by honeycomb lattice layers same configuration 3d graphite same point group @xmath5 symmetry of graphene environments for carbon atoms in aa - stacked nsgl s different from ab - stacked systems each atom two inter - layer nearest - neighbors in each adjacent layers with distence @xmath26 each atom six inter - layer second - nearest - neighbors with distance @xmath25 in two layersaa - stacked 3d graphite two atoms in unit cell primitive translation along @xmath24-axis is @xmath27 half correspondence ab - stacked 3d graphite dynamical representation @xmath28 decomposed into irreducible representations symmetry group lattice displacements as bases @xmath29 vector @xmath30 permutation projection operator technique decomposition into irreducible representations for nsgl s with @xmath0 even and odd ir active phonon modes to irreducible representations from vector representation @xmath29 raman active modes correspond to irreducible representations six - dimensional representation with bases quadratic forms : @xmath31 @xmath32 three acoustic modes with zero frequency at @xmath6 point to vector representation @xmath29 excluded ir and raman active modes corresponding results for graphene and 3d graphite listed in following symbol for irreducible representations is notation in ref . treatment molecules lines guide eyes mode ir active in ab stacking neither ir nor raman active in aa stacking1.2 ab aa stacked nsgl s designated by pentagrams circles raman infra - red activities displayed by full empty symbols broken dashed lines correspond frequencies ab - aa - stacked graphite calculate phonon dispersion for nsgl aa- ab - stacking with arbitrary layer number @xmath0 fig . modeintralayer calculated frequency intra - layer optical c - stretching mode represented with stacked styles layer number @xmath0 layer dependence frequency red shift behavior with experimental measurements frequency value 1588 @xmath4 in single graphene layer decreases with @xmath0 saturates at @xmath37 limit 1586.7 @xmath4 ( 1584.7 @xmath4 ) in ab- aa- ) stacked system red shift value corresponds with experiments ranges @xmath2 @xmath38 8 @xmath4 out - of - plane optical mode @xmath39 ) active in ab - stacking inactive in aa - stacking number useful determining nsgl ab or aa stacking fig . frequencies depend on layer number @xmath0 increase from 864@xmath4 to 872.6 ab and aa stacking contrast c - stretching optical mode frequency blue shift layer dependence with experimental technique inter - layer optical mode layer number dependence frequency in fig . dependent on layer number @xmath0 stack style ab or aa @xmath42 mode frequency values 106 94.5 ab- aa - stacked nsgl s frequencies @xmath43 mode increase with @xmath0 limit values at @xmath37 limit values 149.8 @xmath4 and 133.6 @xmath4 for ab- aa - stacked nsgl s frequency differences raman versus ir @xmath42 mode different layers inspire experimental interest in @xmath43 mode calculate phonon dispersion 2-layered ab stacking 3d graphite significant difference low - frequency region around @xmath6 point fig . frequencies low - frequency optical modes in 2-layer graphene smaller than 3d graphite frequencies raman and ir active modes shown in third line of table two @xmath44 modes frequency value @xmath42=106 and @xmath45=867.4 @xmath4@xmath42 @xmath45 modes are @xmath46 modes ensgl @xmath41 calculate phonon dispersion curve for aa - stacked 3d graphite in fig . unit cell contains two atoms ab - stacked graphite six branches phonon dispersion @xmath47 zone lowest highest branches in - plane acoustic optical modes doubly degenerate two branches non - degenerate @xmath48 point phase factor difference of @xmath49 between two layers out - of - plane motion @xmath42 mode frequency 133.6 @xmath4 fourth line table raman ir active modes for - stacked 3d graphite confirmed in future experiments lattice symmetry raman and ir properties layer dependence studied calculate phonon dispersion of modes ab- or aa - stacked nsgl s layer dependence stressed calculated frequencies of optical c - c stretching mode exhibit red shift as layer number increasing ab- aa - stacked nsgl shift value 2 @xmath4 ( for ab- stacked nsgl s consistent with experimental measurementsout - of - plane optical mode frequency 800 @xmath4 active ab - stacked neither aa - stacking frequency shows blue shift layer dependence frequency inter - layer mode increases @xmath0 increasing sensitive layer number @xmath0 lattice structure properties nsgl | 0.559847 | 3,513 | 1,739 | , Saving $0.1 in GPT-4. |
20,936 | we introduce a dynamical fermion algorithm which is based on the hybrid monte carlo ( hmc ) algorithm , but without pseudofermions .
the molecular dynamics steps in hmc are retained except the derivatives with respect to the gauge fields are calculated with the @xmath0 noise .
the determinant ratios are estimated with the pade - @xmath0 method .
finally , we use the kennedy - kuti linear accept / reject method for the monte carlo step which is shown to respect detailed balance .
we comment on the comparison of this algorithm with the pseudofermion algorithm . | the partition function of the gauge field theory with dynamical fermions in the euclidean path - integral formalism has the following generic form @xmath1 det m[u ] e^{- s_g}\ ] ] where @xmath2 is the pure gauge part of the action and u is the link variable .
the grassmann numbers representing the fermions are formally intergrated out to give a fermion determinant .
since the determinant is very non - local , the monte carlo algorithms with dynamical fermions is known to be very time - consuming .
the best algorithm so far is the hybrid monte carlo @xcite which introduces a monte carlo step to the hybrid ( molecular / hybrid ) algorithm to remove the discretization error from the molecular dynamics step .
this algorithm avoids calculating the determinant directly by introducing pseudofermion variables .
the partition then takes the form @xmath3[d\chi^*][d\chi ] e^{-s_g -\chi^*(m^{\dagger}m)^{-1}\chi}\ ] ] where @xmath4 is the pseudofermion field variable . in this paper
, we shall present an algorithm which estimates the determinant ratio directly without using pseudofermions . to begin with ,
we note that the determinant can be written as @xmath5 and the trace can be estimated by the @xmath0 noise @xcite which is known to yield the minimum variance @xcite .
thus the partition in eq .
( [ partition ] ) takes the form @xmath6e^{-s_g + \int dz_2\ , z_2 ( \log m ) z_2}\ ] ] where @xmath7 is the @xmath0 noise for estimating the trace of the matrix @xmath8 .
it has been shown that there is an efficient algorithm to estimate the determinant by expanding the @xmath8 using the pad approximation .
the trace is then estimated by the @xmath0 noise .
this is reported by chris thron @xcite .
the expectation value of the @xmath9 can be written as @xmath10 = e[tr \log m ] \sim \frac{1}{n } \sum_{i , j } z_2^j ( \frac{b_i}{m + c_i i } ) z_2^j\ ] ] where @xmath11 and @xmath12 are constants .
it is worthwhile noting that since we will use this estimate for the determinant ratios in the monte carlo step for the two matrices @xmath13 and @xmath14 whose eigenvalues are close to each other , we can do the pad expansion aroud 1 with a small set of @xmath11 and @xmath12 .
as long as there is enough memory , the matrix inversion with the set of @xmath11 and @xmath12 can be done with the minimum residual with about 8% overhead in cpu time as compared with inversion of the one with the smallest diagonal term ( the equivalent of the quark mass ) @xcite .
furthermore , @xmath12 are real and positive so that the matrix @xmath15 is better conditioned then m itself .
finally , we should caution that the pz method does not work if there are real negative eigenvalues , because the @xmath16 has a cut on the negative real axis .
the algorithm we propose involves two steps : \1 . * molecular dynamics with determinant force : * the molecular dynamics step is retained as in hmc , except the pseudofermion force is replaced with the determinant force in the evolution equations .
@xmath17\frac{\delta\tau}{2 } \nonumber \\ & \!\!\!+\!\!\ ! & z_2 t[\frac{\partial \log m } { \partial u}(0)u(0)]z_2\frac{\delta\tau}{2 } \nonumber \\
u(\delta\tau ) & \!\!\ ! = \!\!\ !
& exp[\pi(\frac{\delta\tau}{2})\delta\tau ] u(0 ) \nonumber \\
\pi(\delta\tau ) & \!\!\ ! = \!\!\ !
& \pi(\frac{\delta\tau}{2 } ) - t[\frac{\partial s_g}{\partial u}(\delta\tau)u(\delta\tau ) ] \frac{\delta\tau}{2 } \nonumber\\ & \!\!\!+\!\!\ ! & z_2 t[\frac{\partial \log m } { \partial u}(\delta\tau)u(\delta\tau)]z_2\frac{\delta\tau}{2}\end{aligned}\ ] ] several comments are ready : the leapfrog scheme is adopted for this step as in hmc to make it reversible and area preserving .
there is a discretization error @xmath18 as in hmc . like the pseudofermion variable in hmc
, @xmath7 is kept fixed during the md trajectory .
monte carlo : * in lieu of the metropolis accept / reject criteria , we shall us the kennedy - kuti linear test @xcite @xmath19 + where u and v are two configurations and @xmath20 .
h is molecular dynamics hamiltonian .
we note that this satisfies detailed balance provided @xmath21 is an unbiased estimate @xcite .
since the estimate of @xmath22 is unbiased with the @xmath0 noise , we can use the bhanot - kennedy algorithm @xcite to turn @xmath21 into an unbiased estimate : + suppose x is an unbiased estimate , @xmath23
we shall point out several apparent differences between the pseudofermion algorithm and the current determinant algorithm and will make some preliminary comparison of the two approaches .
the pseudofermion approach requires two matrix inversions @xmath24 in each step . whereas , the determinant approach reqires only one such inversion in @xmath25 .
accuracy of the residual of matrix inversion is needed for reversibility for both approaches , i.e. @xmath26 where the average is over all the trajectories .
this allows one to test the thermal equilibrium and make a comparison of the two approaches . the bhanot - kennedy unbiased estimate in eq .
( [ bk ] ) , needs on the average e terms in the matrix inversion to have independent estimate of @xmath22 .
it may be desirable to have more than one @xmath0 noise in the estimate of the determinant ratio . in the kennedy - kuti linear
accept / reject test , the ratio @xmath21 should lie between 0 and 2 to be a probability .
this , however , can be violated in practice @xcite . to make sure that the systematic errors due to the violation of @xmath27 is small , say less than 0.1%
, one needs to have small enough step size in the md trajectories .
the ultimate test between different algorithms will be conducted by comparing how fast observables decorrelate .
this will be done next . in the meantime
, we shall concentrate on the issue of intrinsic noisiness . while the pseudofermion method gives the joint probability of gauge fields and pseudofermion fields , the determinant method has built in it the fluctuation of the @xmath0 noise .
we shall explore the intrinsic noisiness by computing @xmath28 for a fixed pair of @xmath13 and @xmath14 separated by heatbath update of 20 links on a @xmath29 lattice with @xmath30 with the two different methods .
we plot in fig . 1 and fig . 2 the distribution of @xmath34 for 50 gaussian and @xmath0 noises respectively .
we see that the distribution of the pseudofermion method is wider than that of the pz method .
we also plot in fig .
3 the jackknife errors of @xmath35 due to the two method as a function of the number of noise .
again we see that the pz method yields smaller error than that of the pseudofermion method .
9 s. duane , a. d. kennedy , b. pendleton , and d. roweth , phys . lett .
* b195 * ( 1987 ) 216 . s. j. dong and k. f. liu , phys .
b328 * ( 1994 ) 130 .
s. bernardson , p. mccarty , and c. thron , comp .
* 78 * ( 1994 ) 256 . c. thron , k. f. liu , and s. j. dong , this proceeding .
a. frommer , v. hannemann , b. nckel , th .
lippert , and k. schilling , int .
c5 * ( 1994 ) 1073 .
u. glssner , s. gsken , th .
lippert , g. ritzenhfer , k. schilling , and a. frommer , hep - lat/9605008 .
h. p. ying , s. j. dong , and k. f. liu , this proceeding .
d. kennedy and j. kuti , phys .
* 54 * ( 1985 ) 2473 .
g. bhanot and a. d. kennday , phys . lett .
* b157 * ( 1985 ) 70 . | partition function of gauge field theory with dynamical fermions euclidean path - integral formalism form @xmath1 det m[u ] e s_g}\ @xmath2 gauge part u link variable grassmann numbers fermions intergrated fermion determinant determinant non - local monte carlo algorithms time - consuming best algorithm hybrid monte carlo @xcite introduces monte carlo step discretization error avoids calculating determinant pseudofermion variables partition form @xmath3[d\chi^*][d\chi ] e^{-s_g -\chi^*^{-1}\chi}\ @xmath4 pseudofermion field variable algorithm estimates determinant ratio without pseudofermions determinant written as @xmath5 trace estimated by @xmath0 noise @xcite minimum variance partition @xmath6e^{-s_g + dz_2\ z_2_2}\ @xmath7 is @xmath0 noise for estimating trace of matrix @xmath8 efficient algorithm to estimate determinant by expanding @xmath8 using pad approximation trace estimated by @xmath0 noise reported by chris thron @xciteexpectation value @xmath9 written as @xmath10 = e[tr \log m ] \frac{1}{n } \sum_{i j } z_2^j (\frac{b_i}{m + c_i i } z_2^j @xmath11 @xmath12 constants estimate determinant ratios monte carlo step matrices @xmath13 @xmath14 eigenvalues close pad expansion aroud 1 with set @xmath11 @xmath12 enough memory matrix inversion with @xmath11 minimum residual 8% overhead cpu time compared inversion smallest diagonal term @xcite @xmath12 real positive matrix @xmath15 better conditioned m pz method work negative eigenvalues @xmath16 cut negative real axis algorithm involves two steps molecular dynamics with determinant force step retained pseudofermion force replaced with determinant force evolution equations@xmath17\frac{\delta\tau}{2 } \nonumber z_2\frac{\partial \log m u]z_2\frac\delta\tau}{2 \nonumber u(\delta\tau )\pi(\delta\tau\delta\tau u(0 ) \nonumber (\delta\tau )(\delta\tau{2[\frac{\partial s} u}(\delta\tau)u(\delta\tau )\delta\tau{2 } \nonumber\\ z_2[\frac{\partial \log m } u}(\delta\tau)u]z_2\frac{\delta\tau}{2}\end{aligned} comments leapfrog scheme adopted step reversible area preserving discretization error @xmath18 pseudofermion variable @xmath7 fixed md trajectory monte carlo metropolis accept / reject criteria kennedy - kuti linear test @xmath19 + u and v two configurations @xmath20 molecular dynamics hamiltonian satisfies detailed balance @xmath21 unbiased estimateestimate @xmath22 unbiased with @xmath0 noise bhanot - kennedy algorithm to turn @xmath21 into unbiased estimate x unbiased estimate differences between pseudofermion algorithm determinant algorithm preliminary comparison pseudofermion requires two matrix inversions @xmath24 each step determinant approach one inversion in @xmath25 accuracy residual matrix inversion needed for reversibility approaches @xmath26 average trajectories allows test thermal equilibrium comparison bhanot - kennedy unbiased estimate needs average e terms matrix inversion independent estimate of @xmath22 more than one @xmath0 noise in estimate determinant ratio kennedy - kuti linear accept / reject test ratio @xmath21 should between 0 and 2 probability can be violated @xcite systematic errors due violation @xmath27 small less than 0.1% small step size in md trajectories ultimate test between algorithms comparing observables decorrelate concentrate on intrinsic noisiness pseudofermion method joint probability of gauge fields pseudofermion fields determinant method fluctuation of @xmath0 noiseexplore intrinsic noisiness computing @xmath28 separated heatbath update 20 links @xmath29 lattice @xmath30 two methods plot fig 1 2 distribution @xmath34 50 gaussian @xmath0 noises distribution pseudofermion method wider pz method plot 3 jackknife errors @xmath35 method number noise pz method smaller error pseudofermion s. duane d. kennedy pendleton d. roweth . lett b195 1987 ) 216 s j. dong k. f. liu phys . b328 1994 ) 130 bernardson p. mccarty c. thron 78 1994 ) 256 thron f. liu s j. dong proceeding. frommer. hannemann nckel lippert k. schilling c5 1994 ) 1073 glssner gsken lippert ritzenhfer schilling frommer. ying s j. dong k f. liu proceeding d. kennedy j. kuti phys . 54 1985 ) 2473 g. bhanot a d. kennday phys . lett b157 1985 ) 70 | 0.310481 | 2,032 | 1,085 | , Saving $0.1 in GPT-4. |
20,940 | gravitational lenses are presently playing an important role in astrophysics . by means of these lenses the parameters of the deflector such as its mass , ellipticity , etc . and
hubble s constant can be determined .
using c , _ xforms _ , _ mesa _ and _ imlib _ a computer program to visualize this lens effect has been developed .
this program has been applied to generate sequences of images of a source object and its corresponding images .
it has also been used to visually test different models of gravitational lenses .
epsf
= 11 = 12 | computer visualization is nowadays an important field of scientific research , because it often allows a better understanding of natural phenomena . gravitational lenses ( hereafter gl ) , predicted by einstein s general relativity theory ,
deflect the light rays coming from a distant object .
gl s allow the formation of mutiple images of the same source object .
moreover this gravitational effect opens the possiblity of determining the mass of the deflector and the computation of the age of the universe ( hubble s constant ) , as well as other lens parameters and cosmological parameters .
an interactive computer program to visualize this gravitational effect has been developed .
the program was written in the c programming language and uses the _ xforms _ ( tool kit ) , _ mesa _ ( graphic library ) and _ imlib _ ( image library ) .
the program runs on systems with linux or unix .
an sgi version of this program is also available . in the first section the gravitational lens effect is briefly reviewed .
the visualization program is presented in the second section . in the third section
some applications are shown : image sequences and easy visual modelling .
improved versions of this program can be produced and the author wishes to invite the interested reader to participate in this process .
the author is preparing a website for downloading this program : http://lia.efis.ucr.ac.cr/~frutos/
due to the curvature of space - time , light rays coming from a distant source object ( quasar , star or galaxy ) are deflected when passing close to a lens or deflector ( star , galaxy or cluster of galaxies ) . for weak gravitational fields
the _ post - newtonian _ approximation is applicable . under this approximation
the deflection angle or einstein s angle does not depend on the direction of propagation and the trayectory are approximated by straight lines .
this gravitational light deflection is depicted in figure 1 . from this figure
an equation , the _ gl equation _ or the _ ray tracing equation_which is obeyed by light rays passing near a lens object can be deduced : @xmath0 where @xmath1 @xmath2 and @xmath3 is the length scale on the lens surface . the angle @xmath4 is called the einstein s angle .
the surface density of the lens is given by @xmath5 where @xmath6 represents the _ critical density _ : @xmath7 a generalization of equation ( [ ecgl ] ) which considers the perturbation of a _ macrolens _ ( a galaxy or cluster of galaxies ) is given by @xmath8 where the matrix @xmath9 is given by the expression @xmath10 the components of the matrix depend upon the parameters @xmath11 and @xmath12 and @xmath13 which are respectively the dimensionless constant surface mass density , the dimensionless shear of the macrolens and the shear angle .
the microlens ( a star in a galaxy or a galaxy in a cluster of galaxies ) is represented by means of the einstein s angle @xmath4 .
[ fig1 ] = 0.9 to simulate einstein s angle there are several models for the lens mass distribution .
the gravitational lens models are divided as follows : * parametric models * non - parametric models . at the present time the program has only parametric models@xmath14 .
the non - parametric models are lately used to model gravitational lenses@xmath14 . for weak lensing the _ kaiser - squires _ method can be applied to model lenses @xmath14 . in the program
these techniques have not been implemented .
the program includes the following parametric models : * chang refsdal model , * double plane lens , * transparent sphere , * singular isothermal sphere , * nonsingular isothermal sphere , * elliptical model , * king model , * truncated king model , * hubble model , * de vaucouleur model , * spiral model , * multipole lens , * rotation lens and * uniform ring .
the interested reader is refered to the literature for a discussion of these models@xmath14 .
the program was written in c and the _ mesa graphic libraries _ ( free version of the open gl ) have been used .
these graphic subroutines are available for unix or linux systems .
the _ xforms library _ was used to designed the control panel program(see fig .
the _ image library _ permits to load an image file on the program .
these libraries can be found at the following addresses : + http://bragg.phys.uwm.edu/xform + http://www.mesa3d.org/ + ftp://ftp.enlightenment.org/pub/enlightenment/imlib/ the program creates a window : the _ control panel _ ( see fig .
2 ) . the user can control all item on it by clicking . a second window , the _ image window _
, appears when the image window button is clicked on this panel ( see fig .
the help button on the panel give the user a concise program guide .
a version of this program which employs the sgi _
graphic libraries_is also available .
[ general ] [ images2 ] all the objects are adjusted interactively with the mouse . this control panel interface has the following items : * source menu ( filled circle , coloured rings and image file ) * model menu ( the aforementioned models ) * source positioner * three counters , two for source positioning ( x , y ) and one for adjusting the pixel resolution ( n ) * two inputs for positioning the observed lense images ( x , y ) * buttons ( model paramters , image window , source , images , smooth , grid , r plot , r_e , file , add , replace , remove , clear , obip , save , help , exit ) , * sliders ( if one chooses a model , then the corresponding sliders appear ) * a browser for showing the observed lense images for a detailed discussion of these items see the manual program ( to get the program manual send an email to the author or click the help button on the control panel ) . on this window the events ( images , ray plot , etc .
all variations of the parameters on the control panel are shown immediately on this window . to show the images of a chosen gravitational lens model
the _ ray tracing method _ has been used .
the caustic is represented by means of the _ ray plot method_.
in this section two of many applications of the program are discussed : * _ image sequences _ * _ easy visual modeling _ an image sequence for an elliptical model and a visual modeling of the gravitational lens @xmath15 will be shown .
an image sequence is easy to generate with this program .
the structure of the images for different positions of the source can be investigated by means of such an image sequence .
_ an elliptical lens _ a sequence of images is shown in figure 4 .
the sequence begins on the top left - hand - side panel of the sequence .
the source , concentric rings , moves from left to right .
the source is not shown , only the images .
the best possible resolution given by the control panel was used . in this sequence the shear @xmath16 and the constant surface mass density @xmath17 are non zero and the shear angle @xmath18 .
there are two circles : the einstein ring or lens scale and the core radius
. the formation of a small and a large arc can be seen . in the middle of the sequence one can recognize the deformation of the einstein ring in an ellipse . due to the core radius
an image appears in the centre , which fuses with an arc , in the case that the source moves away .
[ ellipt ] ( 11.4 , 17.2)[p]elliptical.gif [ gl2237 ] ( 12.8 , 13.2)[p]gl2237.gif [ m2237el ] ( 12.8 , 13.2)[p]m2237e.gif here an easy visual modeling of the gravitational lens system @xmath15 is shown .
the model images are considered well - fit when they overlap most of the area of the observed images ( represented through concentric circles ) , as visually estimated . at the present time a fitting subroutine is not implemented .
when such a subroutine is included , this program would become an excellent tool to model gravitational lenses . _ visual model for the gravitational lens @xmath15 _ the system @xmath15 [ see fig .
5 ] was discovered in 1985 . due to its unusual shape
this lens is called _
einstein cross_. attempts to model this lens have been carried out@xmath14 the data used in tables [ tab1 ] and [ tab2 ] , was taken from kent and falco , and schneider et al.@xmath14the cosmological distances of table [ tab2 ] were calculated using the equations from schneider et al .
( @xmath19 and @xmath20 ) .
[ tab1 ] .image positions for the lens @xmath15 [ cols="^,^,^ " , ]
from both the didactical and scientific point of view a program to visualize the gravitational lenses is useful .
this versatile program works quickly and interactively with the mouse .
the user can produce sequences of images for a chosen gravitational lens model . through the variation of model parameters
he or she can investigate the structure of the images . the user can also attempt to visually model observed gravitational lenses .
the observed position data can be used as input data and the model parameters can be easily varied in order to approximate the observed images .
so the user can quickly obtain model parameter estimations .
some observed lenses have already been modeled and the user can compare those results with the output from a chosen model of the control panel . * _ contour subroutine for the isochrones ( time delay ) _ * _ light curves subroutine ( dependence of brightness with time ) _ * _ subroutine for computing the image magnification _ * _ subroutine to calculate critical curves and caustics _ * _ fitting subroutine _ * _ root finder subroutine _ * _ subroutine to load images of observed gravitational lenses _ * _ subroutine with more complex ( elliptical ) models _ * _ subroutine for superposition of models in different lens planes _ * _ subroutine with cosmic string lens models _ * _ subroutine for non - parametric reconstruction _ * _ kaiser - squires subroutine _ | computer visualization important scientific research allows understanding natural phenomena gravitational lenses gl predicted by einstein s general relativity theory deflect light rays from distant object allow mutiple images of same source object effect opens determining mass deflector age universe hubble s constant other lens parameters cosmological parameters interactive computer program to visualize gravitational effect developed written in c programming language uses xforms mesa imlib runs on linux unix sgi version available first section gravitational lens effect reviewed visualization program presented in second section third section applications shown image sequences visual modelling improved versions can be produced author to participate preparing website for downloading program : http://lia.efis.ucr.ac.cr/~frutos/ due to curvature space - time light rays from distant source object deflected when close to lens or deflector for weak gravitational fields post - newtonian approximation applicable deflection angle einstein s angle depend on direction propagation trayectory approximated by straight lines gravitational light deflection depicted in figure 1 figure equation ray tracing equation obeyed by light rays near lens object deduced : @xmath0 @xmath1 length scale lens surface angle @xmath4 einstein s angle surface density lens by @xmath5 @xmath6 critical density @xmath7 generalization equation perturbation of macrolens galaxy by @xmath8 matrix @xmath9 expression @xmath10 components matrix depend parameters @xmath11 @xmath12 @xmath13 dimensionless constant surface mass density dimensionless shear macrolens shear angle microlens star galaxy represented einstein s angle @xmath4 fig1 ] = 0.9 simulate einstein s angle several models for lens mass distribution gravitational lens models divided parametric non - parametric models program parametric models@xmath14 non - parametric models used model gravitational lenses@xmath14 weak lensing kaiser - squires method model lenses @xmath14 program techniques not implementedprogram includes parametric models chang refsdal model double plane lens transparent sphere singular isothermal sphere nonsingular isothermal sphere elliptical model king model truncated king model hubble model de vaucouleur model spiral model multipole lens rotation lens uniform ring reader refered to literature for discussion models@xmath14 program written in c mesa graphic libraries free open gl used subroutines available for unix or linux systems xforms library used control panel program image library permits load image file program libraries found at addresses http://bragg.phys.uwm.edu/xform http://www.mesa3d.org/ ftp://ftp.enlightenment/pub/enlightenment/imlib/ program creates control panel _ fig 2 user control item by clicking second window image window appears when image window button clicked help button concise program guide version employs sgi libraries available objects adjusted interactively with mousecontrol panel interface has source menu ( filled circle coloured rings image file model menu ( models source positioner three counters two for source positioning x y one for adjusting pixel resolution n two inputs for positioning observed lense images y buttons ( model paramters image window source images smooth grid r plot r_e file add replace remove clear obip save help exit sliders model sliders appear browser for observed lense images detailed discussion see manual program send email to author or click help button events ( images ray plot variations of parameters control panel shown images gravitational lens model _ ray tracing method used caustic represented _ ray plot method two applications image sequences easy visual modeling image sequence for elliptical model visual modeling gravitational lens @xmath15 image sequence easy to generate structure images for different positions source investigated elliptical lens sequence of images shown in figure 4 begins top left - hand side panel source concentric rings moves from left to right source not shown only images best resolution control panel usedsequence shear @xmath16 constant surface mass density @xmath17 non zero shear angle @xmath18 two circles einstein ring lens scale core radius formation small large arc middle deformation einstein ring ellipse core radius image centre fuses with arc source moves away ellipt ] 11.4 , 17.2)]elliptical.gif gl2237 ] 12.8 , 13.2)]gl2237.gif m2237el ] 12.8 13.2)]m2237e.gif easy visual modeling gravitational lens system @xmath15 model images well - fit overlap area observed images concentric circles fitting subroutine not implemented subroutine included program excellent tool model gravitational lenses visual model gravitational lens @xmath15 system fig . 5 discovered 1985 unusual shape lens einstein cross_ attempts model carried out@xmath14 data tables tab1 tab2 from kent falco schneider et al. cosmological distances table tab2 ] calculated equations schneider et al @xmath19 @xmath20 tab1 ].image positions lens @xmath15 program visualize gravitational lenses useful versatile program works quickly interactively mouseuser produce sequences images for chosen gravitational lens model variation model parameters investigate structure images visually model observed gravitational lenses observed position data used as input model parameters varied approximate observed images obtain model parameter estimations some observed lenses modeled compare results with output chosen model control panel contour subroutine for isochrones ( time delay light curves subroutine brightness with time computing image magnification calculate critical curves caustics fitting subroutine root finder subroutine load images of observed gravitational lenses complex elliptical models superposition of models in different lens planes cosmic string lens models non - parametric reconstruction kaiser - squires subroutine | 0.348058 | 2,209 | 1,092 | , Saving $0.1 in GPT-4. |
20,945 | grlite and grtensorj are first and second generation graphical user interfaces to the computer algebra system grtensorii .
current development centers on grtensorj , which provides fully customizable symbolic procedures that reduce many complex calculations to elementary functions " .
although still in development , grtensorj , which is now available ( free of charge ) over the internet , is sufficiently advanced to be of interest to researchers in general relativity and related fields . | grlite@xcite and grtensorj are first- and second- generation graphical user interfaces to the computer algebra system grtensorii @xcite .
they allow students and researchers in the area of general relativity and related fields to perform symbolic calculations either locally or through the web .
grlite is a calculator - style tool for evaluating more common tensors and scalars , reducing them to elementary functions .
grtensorj provides fully customizable symbolic procedures without recompilation .
these interfaces , which are open source , are written in java and will run on any platform with browser support for jdk1.1 .
any user can initiate a grlite or grtensorj session remotely on the web by logging onto the grtensorii home page and following the links .
all that is needed is a browser with support for jdk1.1 .
the user initiates a session by clicking on the java applet that opens the grlite or grtensorj graphical user interfaces .
they appear as in figures 1 and 2 . behind the scenes , a computer algebra session in grtensorii
is started automatically .
grlite and grtensorj use maplev@xcite as the algebraic engine .
they are in fact compatible with any engine that can output an ascii stream .
grlite is a proof of concept study .
it is perhaps most appropriate for beginning students as it is restricted to classical tensor analysis .
the commands are performed through a predefined menu and selection of buttons .
the first step is to select a spacetime .
this choice automatically displays the components of the metric tensor .
the second step is to simply click on the object to be calculated .
grlite includes the 35 pre - defined functions as shown in figure 1 . after the calculation
is displayed the user can apply to it eight simplification procedures from the menu .
there is also some support , like buffer clearing and a help system .
grlite comes with a set of pre - defined spacetimes but also offers a graphical sub - interface for entering new spacetimes defined for a given session .
the development of grlite is complete but it will be maintained in the interim .
our current development effort concentrates on grtensorj .
* figure 1 * grlite graphical user interface .
calculation of the differential invariant @xmath0 for the kerr metric . at the time of writing
, this calculation executes in about one second on a contemporary pc with maplev release 5.1 and grtensorii 1.76 .
grtensorj provides all the functionality of grlite and much more .
it has a different architecture that allows it to be expanded and programmed by the user .
the commands are accessed through menu and sub - menu selections .
the first step is to select a spacetime in coordinates or in tetrads .
then , the user can select the object(s ) to be calculated from the corresponding menus .
after the result of the calculation is displayed the user can apply to it any simplification procedure supported by the engine .
the help system is built into the menu system as shown in figure 2 .
grtensorj comes with a selection of spacetimes and a set of pre - defined commands .
further , the user can define new metrics , tetrads and procedures through the grtensorii definition facilities .
these are entered through a sub - interface with the option of saving them on the users disk space ( locally ) or in temporary space when used over the internet .
when a session begins , grtensorj reads a directory on the server named textsheets ( not to be confused with worksheets " ) and builds the menu and sub - menus for the interface from the underlying structure .
all sub - directories to textsheets will appear as primary menu bar items .
the names of ascii files contained within these sub - directories will be displayed as menu selections .
these files contain a sequence of commands written in the syntax of the computer algebra engine being used@xcite . by selecting a menu item
the user sends these commands to the engine .
items to be displayed in the interface window are distinguished simply by an asterisk in the file .
in other words , creating new menu items and calculation commands is as simple as creating and editing very simple ascii files .
yet , the file , and resultant menu item , can be the equivalent of an entire worksheet with only the result chosen for display . * figure 2 * grtensorj graphical user interface .
an example of the embedded help system is shown . by the nature of the interface , the help " system is more of the nature of an information system .
grlite and grtensorj are written in java and have been designed using an object - oriented approach .
in addition , all communication between the user and the server are object - based .
grtensorj has a multi - layer architecture that allows the generic functionality described above .
this architecture is outlined below .
a number of useful features are easily added to grtensorj . for example , recently automatic latex , fortran and c output has been added@xcite .
more involved is the current development of a dynamic database of solutions to einstein s field equations .
lake , k 1997 algebraic computing " in _ gravitation and relativity : at the turn of the millennium " proceedings of gr-15 _ , iucaa , pune , edited by naresh dadhich and jayant narlikar pp 421 - 427 ( gr - qc/9803072 ) grtensorii is a package that runs within maplev .
it is entirely distinct from packages distributed with maplev and must be obtained independently .
the grtensorii software and documentation is distributed freely on the world - wide - web from the address http://130.15.26.63/~grtensor/newgrtensorii/ maplev is copyright waterloo maple software and use of maplev through grlite and grtensorj on the web is by special arrangement with waterloo maple software .
for example ,
qload(test ) in grtensorii under maplev is equivalent to qload[test ] in grtensorm under mathematica . at present
, grtensorj linked to grtensorm is not public .
a latex call in maplev produces output suitable for use with latex 2e that appears in the applet window .
we are serious when we say that you can , for example , go to the web , do your calculation , and paste the answer into the paper you are writing ! | grlite@xcite grtensorj are first second- generation graphical user interfaces to computer algebra system grtensorii @xcite allow students researchers general relativity perform symbolic calculations locally or web grlite calculator tool evaluating common tensors scalars elementary functions grtensorj provides customizable symbolic procedures without recompilation interfaces open source written in java run any platform support jdk1.1 initiate grlite grtensorj session remotely logging grtensorii home page links needed browser support jdk1.1 initiates session clicking java applet opens grlite interfaces figures 1 2 algebra session started automatically grlite grtensorj use maplev@xcite algebraic engine compatible with any engine ascii stream grlite proof of concept study for beginning students restricted to classical tensor analysis commands performed through predefined menu buttons first select spacetime displays components metric tensor click on object calculated grlite includes 35 pre - defined functions figure 1 after calculation apply eight simplification procedures from menu buffer clearing help system grlite comes pre - defined spacetimes graphical sub - interface for entering new spacetimesdevelopment grlite complete maintained interim current development concentrates on grtensorj figure 1 grlite graphical user interface calculation differential invariant @xmath0 for kerr metric calculation executes one second on pc with maplev release 5.1 grtensorii 1.76 grtensorj provides functionality grlite more different architecture allows expanded programmed by user commands accessed through menu sub - menu selections first step select spacetime coordinates or tetrads select object(s calculated from menus result calculation displayed apply simplification procedure help system built into menu system figure 2 grtensorj comes with selection spacetimes pre - defined commands user define new metrics tetrads procedures through grtensorii definition facilities sub - interface saving on users disk space or temporary space session begins grtensorj reads directory named textsheets builds menu sub - menus structure sub - directories to textsheets appear as primary menu bar items ascii files sub - directories displayed as menu selections files contain commands in syntax computer algebra engine@xcite selecting menu item user sends commands to engine items displayed interface distinguished by asterisk filecreating menu items calculation commands simple as editing ascii files file menu item equivalent of worksheet result chosen for display figure 2 grtensorj graphical user interface embedded help system help system information system grlite grtensorj written in java designed object - oriented approach communication between user server object - based grtensorj multi - layer architecture allows generic functionality outlined useful features added to grtensorj automatic latex fortran c output added development dynamic database of solutions to einstein s field equations lake , k 1997 algebraic computing gravitation relativity millennium proceedings gr-15 pune narlikar pp 421 - 427 grtensorii package runs within maplev distinct from packages maplev obtained independently software documentation distributed freely web http://130.15.26.63/~grtensor/newgrtensorii/ maplev copyright waterloo maple software use maplev through grlite grtensorj web by special arrangement with waterloo maple qload(test ) in grtensorii under maplev equivalent to qload[test ] in grtensorm under mathematica grtensorj linked to grtensorm not publiclatex call in maplev produces output suitable with latex 2e in applet window serious can go to web do calculation paste answer into paper writing | 0.256814 | 1,350 | 669 | , Saving $0.0 in GPT-4. |
20,947 | we study the production of non - strange mesons in proton - proton scattering reactions within an effective model of mesons and baryons with global chiral symmetry .
the model includes the nucleon ( @xmath0 ) and its chiral partner ( @xmath1 ) and their couplings to the @xmath2 multiplets of ( pseudo)scalar and ( axial-)vector mesons .
results for the production of @xmath3 , @xmath4 mesons are presented . | phenomenological models of qcd are suitable for the study of hadrons and their interactions at energies of a few gev , where perturbative methods are not applicable , see e.g. @xcite . in this work we use an effective chiral model of qcd ,
denoted as the extended linear sigma model ( elsm ) susanna1,denis2,denis3,susanna2,otherselsm , in order to study proton - proton scattering in which a meson is produced in the final state . in the elsm
, the nucleon ( @xmath0 ) and its chiral partner ( @xmath1 ) are introduced according to the mirror assignment @xcite . to this end
, we first introduce the baryonic fields @xmath5 and @xmath6 which transform under chiral transformations @xmath7 as follows : @xmath8where @xmath9 and @xmath10 denote the right- and left - handed components of the baryon field @xmath11 , respectively .
this configuration allows to write down a chirally symmetric mass term of the baryons of the form : @xmath12the physical fields ( @xmath0 ) and ( @xmath1 ) are given by : @xmath13the interaction of the baryons with the ( pseudo)scalar fields is given by the term @xmath14 : @xmath15where @xmath16 represents the matrix of states with zero spin : @xmath17with @xmath18 and @xmath19 as the identity and pauli matrices divided by 2 respectively and @xmath20 as the non - strange component of the @xmath4 meson .
the interaction with the ( axial-)vector states is obtained via the term @xmath21 : @xmath22with : @xmath23and @xmath24 the values of the parameters @xmath25 , @xmath26 , @xmath27 , @xmath28 and @xmath29 are fixed using the following physical quantities : the masses of the nucleon and nucleonic resonance , the decay width @xmath30 , the axial coupling constant of the nucleon and the axial coupling constant of the nucleonic resonance .
further details on the determination of parameters and assignment of the fields to physical particles are presented in refs .
using the model described above , we study processes of the type where @xmath31 is a single @xmath3 or @xmath4 meson . in figure
[ fig : feyn ] , two sample diagrams of the process are presented .
the diagrams represent either the nucleonic current ( fig .
fig : sub1 ) or the resonant current ( fig .
[ fig : sub2 ] ) . to each current 36 different diagrams
contribute ( 9 exchange mesons , @xmath32 and @xmath33channels and early and late emission of the produced meson ) .
the results presented in section [ sec : res ] are obtained using either only the nucleonic current or the coherent sum of both the nucleonic and resonant currents .
the corresponding curves are labelled ` nucleonic ' or ` resonant ' respectively .
.5 : a ) the nucleonic current and b ) the resonant current .
m represents an exchanged @xmath34 , @xmath35 , @xmath36 , @xmath37 , @xmath38 , @xmath39 , @xmath40 , @xmath41 or @xmath42 meson.,title="fig : " ] .5 : a ) the nucleonic current and b ) the resonant current .
m represents an exchanged @xmath34 , @xmath35 , @xmath36 , @xmath37 , @xmath38 , @xmath39 , @xmath40 , @xmath41 or @xmath42 meson.,title="fig : " ]
the total cross section for the reaction @xmath43 is shown in as a function of the center - of - mass momentum of the incoming protons .
the ( black ) dashed curve shows the calculations using only the nucleonic current and the ( blue ) solid one shows the calculations using the resonant and nucleonic currents . as seen in figure [ fig : omg ] , calculations with the nucleonic current only underestimate the data points near the production threshold . including the resonant contribution leads to a significant improvement of the description of the data near threshold .
it is worth mentioning that the data points shown in the figure were not used for the determination of the model parameters . at higher energies , both calculations ( nucleonic and resonant ) tend to overestimate the data points . .
the data points show experimental data and are referenced in @xcite.,scaledwidth=80.0% ] in the case of the @xmath44 meson , the difference between the nucleonic and resonant calculations is not significant as can be seen in figure fig : eta .
again , the ( black ) dashed curve shows the calculated cross section considering the nucleonic current only and the ( blue ) solid curve shows the sum of the resonant and nucleonic currents . both curves overestimate the data considerably even at threshold . .
the data points show experimental data and are referenced in @xcite.,scaledwidth=80.0% ]
the overestimation of the production cross section at high energy ( especially the case of the @xmath4 meson ) hints at a missing ingredient in the calculation .
so far we have only considered the production from a baryon - baryon - meson vertex . the possibility for the produced meson to come from a three - meson vertex
was not included .
including such vertices is expected to have a considerable effect on the results .
in addition , vertex form factors might be needed for a better agreement with the data at higher energies . | models qcd study hadrons interactions at energies few gev perturbative methods not applicable @xcite use chiral model qcd extended linear sigma model elsm study proton - proton scattering meson produced final state nucleon @xmath0 chiral partner @xmath1 ) introduced mirror assignment introduce baryonic fields @xmath5 @xmath6 transform under chiral transformations @xmath9 @xmath10 denote right- left - handed components of baryon field @xmath11 configuration allows chirally symmetric mass term baryons physical fields @xmath0 @xmath1 ) given by interaction baryons with pseudo)scalar fields term @xmath14 @xmath16 matrix states with zero spin @xmath18 @xmath19 identity pauli matrices divided by 2 @xmath20 non - strange component of @xmath4 mesoninteraction with axial-)vector states obtained via term @xmath21 @xmath22with @xmath23and @xmath24 values of parameters @xmath25 @xmath29 fixed using physical quantities : masses of nucleon and nucleonic resonance decay width @xmath30 axial coupling constant of nucleon resonance details on determination parameters assignment fields to physical particles presented in refs model study processes where @xmath31 is single @xmath3 or @xmath4 meson in figure two sample diagrams process presented diagrams represent nucleonic or resonant current 36 diagrams contribute ( 9 exchange mesons @xmath32 @xmath33channels early late emission of produced meson results in section obtained using nucleonic current or coherent sum of nucleonic resonant currents corresponding curves labelled ` nucleonic ' or resonant ' .5 : nucleonic current resonant current represents exchanged @xmath34 @xmath42 meson nucleonic current resonant currentm represents exchanged @xmath34 or meson total cross section for reaction @xmath43 function of center - of - mass momentum incoming protons black dashed curve shows calculations using nucleonic current blue solid resonant and nucleonic currents calculations with nucleonic current underestimate data points near production threshold resonant contribution description data near threshold data points not used for determination model parameters at higher energies both calculations nucleonic and resonant overestimate data points data points show experimental data referenced in @xmath44 meson difference between nucleonic and resonant calculations not significant figure black dashed curve shows calculated cross section considering nucleonic current blue solid curve shows sum of resonant and nucleonic currents both curves overestimate data even at threshold data points show experimental data referenced in overestimation of production cross section at high energy especially @xmath4 meson hints at missing ingredient in calculation only considered production from baryon - baryon - meson vertex possibility for meson from three - meson vertex not included including such vertices on resultsvertex form factors needed better agreement data higher energies | 0.251763 | 1,226 | 586 | , Saving $0.0 in GPT-4. |
20,950 | analytic gravitational collapse and expansion solutions with anisotropic pressure are generated .
metric functions are found by requiring zero heat flow scalar .
it emerges that a single function generates the anisotropic solutions .
each generating function contains an arbitrary function of time which can be chosen to fit various astrophysical time profiles .
two examples are provided : a bounded collapse metric and an expanding cosmological solution .
keywords : anisotropic fluids , collapse , expansion | the general problem of anisotropic fluids which undergo spherical collapse with energy transport by heat flow was formulated some time ago by misner and sharp @xcite .
general relativistic models with anisotropic stress have become increasing useful @xcite .
there has been considerable interest in anisotropic spheres @xcite because of applications to stellar models @xcite .
collapse has been studied in the semiclassical regime by visser et .
@xcite . they found that hawking radiation might prevent formation of a trapped surface and subsequent horizon , although an exponential approach might allow it to form in infinite time .
they proposed a new class of collapsed objects with no horizons .
( in this work , the collapse example does have a trapped surface . )
herrera and santos @xcite have reviewed anisotropy in self - gravitating systems . among the topics in their extensive review are discussions of the energy content of collapsing spheres and the stability of perturbed solutions .
herrera et .
@xcite mention the possibility of a single generating function for non - static anisotropic solutions .
that possibility is developed in this work . in previous work @xcite ,
a method of generating collapse solutions with radial heat flow was presented .
that method took a known spherical perfect fluid solution which was either static or in shear - free collapse and mapped it to a spherical shear - free collapse solution with radial heat flow . in this work ,
a different method of constructing anisotropic solutions is presented .
the metric components are obtained by requiring zero heat flow scalar .
a single function @xmath0 and a parameter `` @xmath1 '' generate anisotropic solutions . not all ( @xmath2 ) pairs are allowed , since the @xmath1 values are restricted .
the fluid volume rate - of - expansion determines the type of anisotropic solution .
there are two possibilities : collapse with shrinking volume and expansion with increasing volume .
the paper is organized as follows : section ii describes the metric and metric functions .
section iii covers energy - momentum components and treats the sectional curvature mass .
generated solutions are given in section iv .
section v has matching conditions which relate the interior and exterior .
the work is summarized in section vi .
appendix a contains a general metric , an associated energy - momentum tensor with heat flow , and trapped surface equations .
the collapse interior has an exterior which is vacuum schwarzschild .
expanding interiors have cosmological exteriors . both collapse and expanding interiors
are covered by the spherically symmetric metric@xmath3 where @xmath4 and where @xmath5 is the metric of the unit sphere .
the velocity is comoving with @xmath6 .
the interior metric is spanned by the tetrad @xmath7 such that@xmath8 metric components are chosen so that the heat flow scalar @xmath9 .
the equation for @xmath10 is written from eq.([q ] ) , where primes denote @xmath11 , and overdots denote @xmath12 , as@xmath13 we find that , @xmath14 satisfies eq.([zero - q ] ) identically for arbitrary @xmath0 . as a metric component @xmath15
@xmath16 allows @xmath17 to be absorbed by redefining @xmath18 , similarly for @xmath19 and @xmath20 .
therefore we set @xmath21 .
the interior metric is now @xmath22 using values from eqs.([new - a ] ) and ( [ new - b ] ) for @xmath23 and @xmath24 , the fluid volume rate - of - expansion and rate - of - shear scalar , given in eq.([exp ] ) and eq.([shear ] ) respectively , are@xmath25 thus restricting @xmath26 @xmath27 .
the value of @xmath28 provides two different interiors : collapse with @xmath29 and expansion with @xmath30 .
we define a dimensionless measure of anisotropy as@xmath31
the energy - momentum given in appendix a is written here with zero heat flow @xmath32 , since the heat scalar is chosen to vanish@xmath33 with fluid velocity @xmath34 such that @xmath35 .
@xmath36 is the mass - energy density , @xmath37 is the radial pressure and @xmath38 is the tangential pressure . substituting in equations
( [ w ] ) , ( [ p - r ] ) , and ( [ p - perp ] ) for @xmath23 and @xmath24 , we have@xmath39 + \frac { 1}{r^{2 } } \label{w - eqn}\]]@xmath40 + ( 1 + 2\alpha)r^{2\alpha-2}+\frac{1}{r^{2 } } \label{pr - eqn}\]]@xmath41 \nonumber\end{aligned}\ ] ] specifying @xmath0 and @xmath1 distinguishes an anisotropic configuration .
equation ( [ sect - curv - mass ] ) provides the sectional curvature mass with @xmath23 and @xmath24@xmath42 when @xmath43 there is a possible trapped surface at @xmath44 . from equations ( [ s1 ] ) and ( [ s2 ] ) the trapping scalars are@xmath45 when @xmath46 and @xmath47 have the same sign ,
a trapped surface will exist . for @xmath43
the trapping scalars are @xmath48 they are both non - negative , therefore during collapse a trapped surface develops at @xmath44 . the trapped surface condition @xmath43 has the integral @xmath49
when the trapped surface condition @xmath43 is substituted into equations ( [ w - eqn],[pr - eqn],[pperp - eqn ] ) then @xmath50
the rate - of - expansion must be negative for collapse , therefore @xmath1 must be less than @xmath52 . here
we assume @xmath51 .
a trapped surface occurs when @xmath43 .
thus @xmath53 or @xmath54 with solution@xmath55^{1/6},\text { \ } h_{1}(t)\text { arbitrary } \label{r - trap - soln}\ ] ] the density and pressure equations for @xmath51 become@xmath56 + r^{-2}\\ 8\pi p_{r } & = -2r^{5}\left [ 3(\frac{r^{\prime}}{r})^{2}+(\frac{r^{\prime}}{r})\frac{\dot{r}^{\prime}}{\dot{r}}\right ] -4r^{-7}+r^{-2}\\ 8\pi p_{\perp } & = -10r^{-7}+r^{5}\left [ \frac{\dot{r}^{\prime\prime}}{\dot{r}}+\frac{17}{2}\frac{\dot{r}^{\prime}}{\dot{r}}(\frac{r^{\prime}}{r})+\frac{7}{2}\frac{r^{\prime\prime}}{r}+15(\frac{r^{\prime}}{r})^{2}\right]\end{aligned}\ ] ] we assume @xmath57 the density and pressures in this case are@xmath58 the mass - energy density @xmath59 is graphed below in figure 1
. w-vs-r-1.eps when @xmath60 the density and pressures revert to the trapped values given in eq.([trapped - values ] ) .
the sectional curvature mass given in eq.([trap - mass ] ) is@xmath61 \label{mass1}\ ] ] the constant @xmath62 and function @xmath63 remain free choices . the dimensionless measure of anisotropy , defined in eq.([delta - a ] ) ,
is given by@xmath64 for @xmath63 constant and fixed @xmath62 , the anisotropy is a maximum at the center @xmath65 .
as @xmath66 increases @xmath67 falls to @xmath68 ( @xmath66 has an upper bound - see the section on matching ) . at the trapped surface ,
where @xmath60 , there is a minimum with @xmath69 .
we choose @xmath70 with @xmath71 for expansion .
we also choose@xmath72 for @xmath70 @xmath73+r^{-2}\\ 8\pi p_{r } & = 2r^{-3}[(\frac{r^{\prime}}{r})^{2}-(\frac{r^{\prime}}{r})\frac{\dot{r}^{\prime}}{\dot{r}}]+4r+r^{-2}\\ 8\pi p_{\perp } & = -r[2\frac{\ddot{r}}{\dot{r}}(\frac{r}{\dot{r}})+3]+r^{-3}[\frac{\dot{r}^{\prime\prime}}{\dot{r}}-\frac{3}{2}(\frac{\dot { r}^{\prime}}{\dot{r}})(\frac{r^{\prime}}{r})-\frac{1}{2}\frac{r^{\prime \prime}}{r}+\frac{3}{4}(\frac{r^{\prime}}{r})^{2}]\end{aligned}\ ] ] with @xmath74 and @xmath75 , the density and pressures in this case are@xmath76+\frac{(r^{2}+r_{0}^{2})^{2}(r_{0}^{2}-3r^{2})}{f^{4}}-\frac{3r^{2}}{f^{2}(r^{2}+r_{0}^{2})^{2}}\ ] ] w-vs-r-2.eps + fig.2 @xmath77 , @xmath78 , w vs r ( unscaled ) the sectional curvature mass is@xmath79 .\ ] ] when @xmath80 and @xmath78 , as in figure 2 , then as @xmath81 @xmath82 @xmath83 .
interior and exterior regions match across a separating boundary @xmath84 if the first and second fundamental forms are continuous across @xmath84 @xcite .
the collapsing interiors match to a schwarzschild exterior @xmath85 the interior metric is @xmath86 ( interior objects will be indicated by a minus and exterior objects by a plus . ) the metric match requires@xmath87 at the boundary , the sectional curvature mass must equal the exterior schwarzschild mass parameter@xmath88 the unit normal to the boundary is@xmath89 such that @xmath90 junction conditions equivalent to continuity of the second fundamental form are @xmath91@xmath92 where three linearly independent @xmath93 span the boundary surface . the schwarzschild exterior is vacuum ,
so that@xmath94 this implies that at the boundary @xmath95 i.e. the density and pressures must vanish .
figure 1 show the density falling off to zero .
the pressures fall off similarly .
expanding solutions have cosmological exteriors .
collins @xcite discussed the global properties of kantowski - sachs @xcite cosmologies .
he dropped their dust condition and showed in detail how the expanding interiors relate to their cosmological exteriors .
analytic anisotropic collapse and expansion solutions have been generated .
the interior collapse metric has been matched to an exterior schwarzschild vacuum .
the collapse solution has a trapped surface where the dimensionless anisotropy measure is minimized . to paraphrase herrera and santos @xcite : `` for dense systems , phase transitions may occur during gravitational collapse , particularly transitions to a pion condensed state .
a softened equation of state can provide a large energy release which is important in the evolution of collapsing configurations . ''
additionally , redshift data from the coma cluster of galaxies @xcite shows a distinct infall region .
expanding solutions have been discussed by collins @xcite .
these solutions allow inhomogeneous cosmologies to be modelled .
for instance , barrow and maartens @xcite have investigated the effect of anisotropic stresses on the late - time evolution of inhomogeneous universes .
they found decay of shear anisotropy to be slowed by the presence of anisotropic stresses .
they also found attractor solutions relating distortion in the expansion anisotropy to the fractional density in anisotropic stress .
this work differs from previous studies such as @xcite which are static , and studies @xcite in which the spherical area radius is simplified to @xmath66 , rather than @xmath0 used here .
we find this is the only work which uses @xmath0 to generate solutions . in summary ,
the solution generating function is @xmath0 with parameter `` @xmath1 '' .
@xmath0 was constructed by requiring zero heat flow scalar . for each valid choice of @xmath1 and @xmath0 ,
one obtains an anisotropic configuration which collapses or expands , and each generating function contains an arbitrary function of time which can be chosen to fit various astrophysical time profiles . * acknowledgment * we thank professor jean krisch for constructive comments .
metric @xmath96 is petrov type * d. * the two principal null vectors , normal to ( @xmath97 ) two - surfaces , are@xmath98 the energy - momentum tensor is given by ( @xmath99 ) @xmath100 where @xmath37 is the radial pressure , @xmath38 is the tangential pressure , @xmath36 is the mass - energy density , and @xmath32 is the radial heat flow vector orthogonal to @xmath101 .
we use notation of taub @xcite for @xmath36 .
taub s purpose was to distinguish mass - energy density from proper mass - density @xmath102 , with @xmath103 .
this allows the first law of thermodynamics to be written in its usual form @xmath104 with specific entropy @xmath105 and specific internal energy @xmath106 .
the kinematics of the fluid are described by [ three]@xmath107,\ \ \sigma = a^{-1}(\dot{r}/r-\dot{b}/b ) \label{shear}\ ] ] where primes denote @xmath108 , and overdots denote @xmath109 .
the rate - of shear @xmath110 is trace - free , and @xmath111 .
the heat flow vector ( @xmath112 ) is given by the sectional curvature mass is @xmath114 .
\label{sect - curv - mass}\ ] ] the mass - energy density and pressures are given , respectively , by@xmath115
-\frac{1}{b^{2}}\left [ 2\frac{r^{\prime\prime}}{r}+(\frac{r^{\prime}}{r})^{2}-2\frac{b^{\prime}}{b}\frac{r^{\prime}}{r}\right ] + \frac{1}{r^{2 } } \label{w}\]]@xmath116 -\frac{1}{a^{2}}\left [ 2\frac{\ddot{r}}{r}+(\frac{\dot{r}}{r})^{2}-2\frac{\dot{r}}{r}\frac{\dot{a}}{a}\right ] -\frac{1}{r^{2 } } \label{p - r}\]]@xmath117+\frac{1}{b^{2}}[\frac{r^{\prime\prime}}{r}+\frac { a^{\prime\prime}}{a}+\frac{r^{\prime}}{r}(\frac{a^{\prime}}{a}-\frac { b^{\prime}}{b})-\frac{a^{\prime}}{a}\frac{b^{\prime}}{b } ] \label{p - perp}\ ] ] the topological two - spheres ( @xmath97 ) nested in an @xmath118 surface at time @xmath119 have outgoing null geodesic normal @xmath120 and incoming null geodesic normal @xmath121 . the two principal null vectors ( [ l - vec ] ) and ( [ n - vec ] )
provide trapping scalars@xmath122 when scalars @xcite @xmath46 and @xmath47 have the same sign , a trapped surface @xcite will exist : @xmath123 | problem of anisotropic fluids spherical collapse with energy transport by heat flow formulated by misner sharp relativistic models with anisotropic stress useful interest in anisotropic spheres applications to stellar models collapse studied by visser . hawking radiation trapped surface horizon exponential approach might in infinite time proposed new class of collapsed objects with no horizons collapse example trapped surface herrera santos reviewed anisotropy in self - gravitating systems energy content of collapsing spheres stability of perturbed solutions single generating function for non - static anisotropic solutions developed in work previous work method generating collapse solutions with radial heat flow presented spherical fluid solution to different method constructing anisotropic solutions presented metric components zero heat flow scalar single function @xmath0 parameter @xmath1 generate anisotropic solutions not all pairs allowed @xmath1 values restricted fluid volume rate - of - expansion determines type of anisotropic solution two possibilities : collapse with shrinking volume expansion with increasing volumepaper organized section ii describes metric functions covers energy - momentum sectional curvature mass solutions section iv section v conditions interior exterior work summarized section vi appendix a contains general metric energy - momentum tensor heat flow trapped surface equations collapse interior exterior vacuum schwarzschild expanding interiors cosmological exteriors covered by spherically symmetric metric@xmath3 metric unit sphere velocity with @xmath6 interior metric spanned by tetrad @xmath7@xmath8 metric components heat flow scalar @xmath9 equation @xmath10 from eq.([q ] primes denote @xmath11 overdots @xmath12 @xmath14 satisfies eq.([zero - q for @xmath0 metric component @xmath15 @xmath16 allows @xmath17 @xmath18 set @xmath21 interior metric @xmath22 values from eqs @xmath23 fluid volume rate - of - expansion rate - of - shear scalar are@xmath25 restricting @xmath26 @xmath27value @xmath28 provides two interiors collapse expansion define dimensionless measure anisotropy@xmath31 energy - momentum appendix a zero heat flow @xmath32 heat scalar@xmath33 fluid velocity @xmath34 @xmath36 mass - energy density radial pressure tangential pressure substituting equations w p - r p - perp @xmath23 @xmath24@xmath39 +\frac 1{r^{2{w - eqn@xmath40 + 1 + 2^{2^{pr - eqn}@xmath41{aligned @xmath0 @xmath1 distinguishes anisotropic configuration equation [ sect - curv - mass ] provides sectional curvature mass @xmath23 @xmath24@xmath42 possible trapped surface @xmath44 trapping scalars@xmath45 @xmath46 @xmath47 same sign trapped surface @xmath43 trapping scalars @xmath48 non - negative collapse trapped surface develops @xmath44trapped surface condition @xmath43 integral @xmath49 substituted into equations - @xmath50 rate - - expansion negative for collapse @xmath1 less than @xmath52 assume @xmath51 trapped surface occurs @xmath43 @xmath53 or @xmath54 solution@xmath55^{1/6} arbitrary - trap - soln density pressure equations @xmath51 become@xmath56 + r^{-2}\\ 8\pi p_{r } = -2r^{5}\left 3(\frac{r^{\prime}}{2}+}{r}}\right -4r^{-7}+r^{-2}\\ 8\pi p^{-7}+r^{5}\left{r\prime+{17}{2}+{7}{2}+15(^{2} assume @xmath57 density pressures are@xmath58 mass - energy density @xmath59 graphed figure 1 w-vs-r-1.@xmath60 density pressures revert to trapped values. - sectional curvature mass is@xmath61 constant @xmath62 @xmath63 free choices dimensionless measure anisotropy[delta -@xmath64 @xmath63 constant fixed @xmath62 anisotropy maximum at center @xmath65 @xmath66 increases @xmath67 falls to @xmath68 @xmath66 upper bound trapped surface @xmath60 minimum with @xmath69 choose @xmath70 with @xmath71 for expansionchoose@xmath72 @xmath70 @xmath73+r^{-2 8\pi{r 2r^{-3}\frac{r\prime}}{2}-\frac{r\prime}}{r}{r{r}}]+4r+r^{-2 8\pi -r[2\frac{r}}{r}})+3]+r^{-3}-\frac{3}{2}\prime}}{r}-\frac{1}{2}\prime+\frac{3}{4}{r{\prime}}{r}^{2}]{aligned @xmath74 @xmath75 density pressures@xmath76+\frac{(r^{2}+r{2}{2}{2}{f^{4}}-\frac{3r^{2}}{f^{2}(r{2+r{0{2}{2}}-vs-r-2.eps.2 @xmath77 @xmath78 sectional curvature mass@xmath79 @xmath80 @xmath78 figure 2 @xmath81 @xmath82 @xmath83interior exterior regions match boundary first second fundamental forms continuous collapsing interiors match schwarzschild exterior interior metric @xmath86 interior objects minus exterior plus metric match requires boundary sectional curvature mass equal exterior schwarzschild mass parameter unit normal to boundary@xmath89 junction conditions continuity second fundamental form three independent @xmath93 span boundary surface schwarzschild exterior vacuum boundary density pressures vanish figure 1 density zero pressures fall off expanding solutions have cosmological exteriors collins discussed properties kantowski - sachs cosmologies expanding interiors cosmological exteriors analytic anisotropic collapse expansion solutions generated interior collapse metric matched to exterior schwarzschild vacuum collapse solution trapped surface dimensionless anisotropy measure minimized herrera santos dense systems phase transitions occur during gravitational collapse condensed state softened equation state large energy release important evolution collapsing configurations redshift data coma cluster galaxies shows distinct infall region expanding solutions discussed by collinssolutions allow inhomogeneous cosmologies barrow maartens investigated effect anisotropic stresses on late evolution inhomogeneous universes found decay anisotropy slowed by anisotropic stresses found solutions relating distortion expansion anisotropy to fractional density anisotropic stress work differs from previous studies static spherical area radius simplified to @xmath66 only work uses @xmath0 generate solutions solution generating function is @xmath0 with parameter @xmath1 zero heat flow scalar each valid choice of @xmath1 obtains anisotropic configuration collapses or expands each generating function contains arbitrary function time astrophysical time profiles professor jean krisch for comments @xmath96 is petrov type two principal null vectors are@xmath98 energy - momentum tensor by @xmath100 @xmath37 radial pressure tangential pressure mass - energy density radial heat flow vector to notation taub @xcite for @xmath36 mass - energy density from proper mass - density @xmath102allows first law thermodynamics written form @xmath104 specific entropy @xmath105 internal energy kinematics fluid described@xmath107 = a{-1}{r-{b} primes denote @xmath108 overdots denote @xmath109 rate shear @xmath110 trace - free @xmath111 heat flow vector sectional curvature mass @xmath114{sect - curv - mass mass - energy density pressures given@xmath115 -\frac{1}{b^{2}} 2\frac{r\prime}-2\frac{b^{\prime}}{r{{r} \frac{1}{r^{2@xmath116 -\frac{1}{a^{2}}\left 2\frac{\ddot{r}}{r}+\dot{r}}{2}-2\frac{{r}}{r}\frac{\dot{a}}{a} -\frac{1}{r^{2 } -@xmath117+\frac{1}{b^{2}}{r{\prime+{\prime{r^{\prime}}{a{\prime}}^{-{a{{b^{\prime}}{b{p - topological two - spheres @xmath97 ) @xmath118 surface outgoing null geodesic normal @xmath120 incoming null @xmath121 null vectors l - n - trapping scalars@xmath122 trapped surface @xcite @xmath123 | 0.60572 | 3,596 | 1,811 | , Saving $0.1 in GPT-4. |
20,951 | the calculation of partial decay rates in @xmath0 decays at next - to - next - to - leading order ( nnlo ) in @xmath1 and to leading order in @xmath2 is described .
new results for the hard function are combined with known results for the jet function and shape - function moments in a numerical analysis which explores the impact of the nnlo corrections on partial decay rates and the determination of @xmath3 . | the inclusive decay @xmath0 is of much interest because of its potential to constrain the ckm element @xmath3 . due to experimental cuts required to suppress charm background , measurements of this decay are available only in the shape - function region , where the hadronic final state is collimated into a single jet carrying a large energy on the order of @xmath4 and a moderate invariant mass squared on the order of @xmath5 .
the theory challenge is to calculate partial decay rates in the presence of these cuts , where a local operator product expansion is insufficient .
a systematic treatment relies on a non - local operator product expansion whose end result can be formulated in terms of a factorization theorem .
different approaches to this factorization have been put forth in the literature , going under the names of blnp @xcite , ggou @xcite , and the dressed - gluon exponentiation @xcite .
the purpose of this talk is to describe the elements that go into the blnp formalism at next - to - next - to - leading order ( nnlo ) in @xmath1 and leading order in @xmath2 .
the main point of discussion is the factorization formula for an arbitrary decay distribution restricted to the kinematics of the shape - function region @xmath6 where the symbol @xmath7 denotes a convolution .
the factorization formula contains a hard function @xmath8 , which is related to physics at the hard scale @xmath4 , a jet function @xmath9 , which is related to physics at the intermediate scale @xmath5 , and a non - perturbative shape function @xmath10 , describing the internal soft dynamics of the @xmath11 meson @xcite .
this factorized form was originally derived in @xcite using diagrammatic techniques , and was rederived in the framework of soft - collinear effective theory ( scet ) in @xcite .
the hard and jet functions to nlo in perturbation theory have been known for some time @xcite , and the nnlo jet function was obtained in @xcite .
very recently , the nnlo contributions to the hard function have also been calculated @xcite , thus completing the perturbative corrections to the factorization formula to this order . in section [ sec : nnlofact ] we review the calculation of @xmath8 and @xmath9 to nnlo , along with the form and solutions of their renormalization - group ( rg ) evolution equations .
then , in section [ sec : numerics ] , we give preliminary results illustrating the numerical impact of the nnlo corrections on the @xmath12 spectrum , comment on the relevance for the determination of @xmath3 , and make some concluding remarks .
in this section we describe the calculation of the hard and jet functions at nnlo , as well as the form and solution of their rg evolution equations .
the hard function arises when integrating out fluctuations at the scale @xmath4 by matching qcd onto scet .
this matching can be done at the level of the @xmath13 transition current , and to leading order in @xmath2 takes the form @xmath14 where we have followed the scet conventions of @xcite , and the @xmath15 are a set of three dirac structures . in practice ,
the matching calculation is carried out in momentum space and yields results for the fourier - transformed coefficients , which read @xmath16 the hard function @xmath8 is derived from the matrix of coefficients @xmath17 . to obtain
the matching coefficients @xmath18 requires to calculate uv - renormalized matrix elements in full qcd and scet .
the calculation is simplest when the external states are chosen as on - shell quarks and both uv and ir divergences are regulated in dimensional regularization in @xmath19 dimensions . in that case
the loop corrections to the scet matrix elements are given by scaleless integrals and vanish , so that the result is just its tree - level value multiplied by renormalization factors from operator and wave - function renormalization .
the main challenge is to calculate the qcd result , which is written in terms of three dirac structures multiplied by scalar form factors @xmath20 . to obtain these scalar form factors at nnlo requires to calculate the two - loop corrections to the @xmath13 current in qcd .
this task has been completed in @xcite .
one then turns the results for the scalar form factors @xmath21 into those for the scet wilson coefficients by evaluating the matching conditions @xmath22 where @xmath23 is a current renormalization factor .
details of the matching calculation and results for the wilson coefficients have been presented in @xcite . in the blnp approach , the hard function
is evaluated at a scale @xmath24 gev and logarithms of the ratio @xmath25 are treated as large .
one can resum these logarithms by deriving and solving the rg equation for the hard coefficients @xmath18 , or equivalently the matrix of coefficients @xmath26 .
the evolution equation has the form ( see , e.g. @xcite ) @xmath27h_{ij}(\bar n \cdot p,\mu ) \,,\ ] ] and its solution is @xmath28 h_{ij}(\bar n \cdot p,\mu_h)\,,\ ] ] where we have defined @xmath29 .
explicit results for the sudakov factor @xmath10 and the anomalous exponents @xmath30 , @xmath31 can be read off from @xcite .
a consistent treatment at nnlo in @xmath1 requires the matching coefficients at two loops , the anomalous dimension @xmath32 at three loops , and the cusp anomalous dimension at four loops .
the anomalous dimensions are both known to one loop lower , which adds a small uncertainty to the analysis .
we now turn to the jet function @xmath9 , which arises when integrating out fluctuations at the intermediate scale @xmath5 by matching scet onto hqet .
the calculation of the jet function can be recast into the evaluation of the imaginary part of a certain vacuum matrix element in qcd , and was calculated to nnlo in @xcite . in light - cone gauge , this matrix element is the quark propagator .
the jet function depends on a renormalization scale @xmath33 , which is usually kept fixed at @xmath34 gev in the blnp approach . however , it is possible to derive and solve the rg equation for @xmath9 , which allows one to also vary the renormalization scale at which the jet function is evaluated .
the starting point is the exact integro - differential evolution equation derived in @xcite : @xmath35 j(p^2,\mu ) - 2\gamma_{\rm cusp } \int_0^{p^2}\!dp^{\prime 2}\ , \frac{j(p^{\prime 2},\mu)-j(p^2,\mu)}{p^2-p^{\prime 2 } } \,.\ ] ] a novel solution to the evolution equation ( [ jrge ] ) was presented in @xcite .
the method is based on the observation that the laplace transformed jet function @xmath36 obeys a local evolution equation analogous to ( [ eq : rgh ] ) . solving as before and inverting the laplace transform gives the resummed @xmath9 in momentum space . alternatively ,
it is always possible to write partial decay rates in terms of an integral over the jet function , defined as @xmath37 using the solution of the rg equation , one can write a compact expression for the resummed integrated jet function @xmath38 \left(\frac{q^2}{\mu_i^2}\right)^\eta \widetilde j\big(\partial_\eta+\ln\frac{q^2}{\mu_i^2},\mu_i\big ) \frac{e^{-\gamma_e\eta}}{\gamma(1+\eta ) } , \ ] ] where expressions for the rg factors @xmath39 and @xmath40 can be found in @xcite . equations ( [ eq : resummedh ] ) and ( [ eq : resummedj ] ) allow one to evaluate integrals of the product @xmath41 at an arbitrary scale @xmath42 at which the shape function is renormalized , and study the dependence of partial decay rates under independent variations of the matching scales @xmath43 and @xmath44 , under which they are formally independent .
detailed numerical results shall be given in @xcite .
[ ] @xmath43 ( gev ) [ ] @xmath45 [ ] [ ] lo [ ] nlo [ ] nnlo in this section we briefly explore the numerical impact of the nnlo corrections , taking as an example the partial rate with a cut @xmath46 .
this partial rate can be written as @xmath47 where @xmath48 , and the hard and integrated jet functions are to be evaluated in their resummed forms ( [ eq : resummedh ] ) and ( [ eq : resummedj ] ) .
the function @xmath49 is related to certain combinations of the @xmath26 and can be deduced from @xcite . to evaluate ( [ eq : p+ ] )
requires a model for the shape function . in what follows ,
we use the two - parameter exponential model @xmath50 it is possible to put model - independent constraints on the normalization factor @xmath51 and the parameters @xmath52 and @xmath53 by studying shape - function moments @xcite .
these moments are defined as @xmath54 and can be obtained as an expansion in local hqet operators as long as @xmath55 .
the zeroth moment fixes the normalization of the shape function and was determined to nnlo in @xcite , and can be used along with the first and second moment to define the parameters @xmath56 and @xmath57 in the shape - function scheme .
numerical values for the hqet parameters @xmath4 and @xmath57 in the shape - function scheme at nnlo can be determined from those obtained from global fits in other schemes using the two - loop conversion relations derived in @xcite .
one then tunes the parameters @xmath52 and @xmath53 such that the moments of the model function ( [ eq : smod ] ) give appropriate results for @xmath4 and @xmath57 . in our analysis here
, we tune the parameters such that @xmath58 gev and @xmath59 at nnlo in the shape - function scheme . in figure
[ fig : scalevar ] , we show results for the partial rate with a cut @xmath46 , for the choice @xmath60 gev , as a function of the hard matching scale @xmath43 . from the figure
, one sees that the dependence on the renormalization scale @xmath43 is reduced when going to higher orders in perturbation theory , and also that there is a fairly large downward shift in the central value between nlo and nnlo .
this would tend to raise the value of @xmath3 compared to the result deduced from the nlo calculation in the blnp approach .
however , we must stress that these numerical results are preliminary , and will be finalized in @xcite , along with uncertainties associated with variations of the matching scale @xmath61 results for other partial rates , such as those with cuts on the hadronic invariant mass and lepton energy , will also be presented .
it is worth emphasizing that the partial decay rates are rather sensitive to the numerical value of @xmath4 .
if , as suggested in @xcite , the @xmath62 moments are excluded from global fits on the grounds that measurements of the photon energy spectrum are made in kinematic regions where shape - function effects are important , then @xmath4 tends to come out closer to @xmath63 gev .
changing @xmath4 by such an amount raises the partial decay rates upwards by approximately 20% .
a precise determination of @xmath3 from inclusive decays will thus require that this point be settled .
s. w. bosch , b. o. lange , m. neubert and g. paz , nucl .
b * 699 * ( 2004 ) 335 [ arxiv : hep - ph/0402094 ] .
b. o. lange , m. neubert and g. paz , phys . rev .
d * 72 * ( 2005 ) 073006 [ arxiv : hep - ph/0504071 ] .
p. gambino , p. giordano , g. ossola and n. uraltsev , jhep * 0710 * ( 2007 ) 058 [ arxiv:0707.2493 [ hep - ph ] ] .
j. r. andersen and e. gardi , jhep * 0601 * ( 2006 ) 097 [ arxiv : hep - ph/0509360 ] .
i. i. y. bigi , m. a. shifman , n. g. uraltsev and a. i. vainshtein , int .
j. mod .
phys . a * 9 * , 2467 ( 1994 ) [ arxiv : hep - ph/9312359 ]
. g. p. korchemsky and g. sterman , phys .
b * 340 * , 96 ( 1994 ) [ arxiv : hep - ph/9407344 ] .
r. akhoury and i. z. rothstein , phys .
d * 54 * ( 1996 ) 2349 [ arxiv : hep - ph/9512303 ] . c. w. bauer and a. v. manohar , phys .
d * 70 * ( 2004 ) 034024 [ arxiv : hep - ph/0312109 ] .
t. becher and m. neubert , phys .
b * 637 * , 251 ( 2006 ) [ arxiv : hep - ph/0603140 ] .
r. bonciani and a. ferroglia , arxiv:0809.4687 [ hep - ph ] .
h. m. asatrian , c. greub and b. d. pecjak , arxiv:0810.0987 [ hep - ph ] . | inclusive decay @xmath0 potential ckm element @xmath3 due to experimental cuts charm background measurements decay available only in shape - function region hadronic final state collimated into single jet large energy @xmath4 moderate invariant mass @xmath5 challenge calculate partial decay rates cuts local operator product expansion insufficient systematic treatment relies on non - local operator product expansion factorization theorem approaches factorization blnp @xcite ggou @xcite dressed - gluon exponentiation @xcite talk describe elements blnp formalism leading order in @xmath1 leading order @xmath2 main discussion factorization formula for arbitrary decay distribution restricted to kinematics shape - function region @xmath6 symbol @xmath7 denotes convolution factorization formula contains hard function @xmath8 scale jet function @xmath9 intermediate non - perturbative shape function @xmath10 soft dynamics @xmath11 meson @xcite factorized form derived in @xcite rederived soft - collinear effective theory in @xcite hard jet functions to nlo in perturbation known @xcite nnlo jet function obtained @xcite nnlo contributions to hard function calculated @xcite completing perturbative corrections to factorization formula section review calculation @xmath8 @xmath9 to nnlo renormalization - evolution equations section numerics preliminary results impact nnlo corrections on @xmath12 spectrum relevance for determination @xmath3 concluding remarks calculation hard jet functions at nnlo form rg evolution equations hard function arises fluctuations at scale @xmath4 matching qcd scet matching at @xmath13 transition current @xmath2 form @xmath14 scet conventions @xmath15 three dirac structures matching calculation in momentum space yields results for fourier - transformed coefficients hard function @xmath8 derived from matrix coefficients @xmath17 matching coefficients @xmath18 uv - renormalized matrix elements in full qcd scet calculation simplest external states as on - shell quarks uv ir divergences regulated in dimensional regularization in @xmath19 dimensionsloop corrections to scet matrix elements by scaleless integrals result tree - level value multiplied by renormalization factors challenge qcd result three dirac structures multiplied by scalar form factors @xmath20 factors nnlo two - loop corrections to @xmath13 current in qcd completed in @xcite turns results scalar form factors @xmath21 into scet wilson coefficients matching conditions @xmath22 @xmath23 renormalization factor matching calculation presented in @xcite blnp approach hard function evaluated at scale @xmath24 logarithms of ratio @xmath25 treated as large logarithms rg equation for hard coefficients @xmath18 matrix of coefficients @xmath26 evolution equation @xmath27h solution @xmath28 defined @xmath29 results for sudakov factor @xmath10 anomalous exponents @xmath30 @xmath31 read from @xcite consistent treatment at nnlo in @xmath1 requires matching coefficients at two loops anomalous dimension @xmath32 three cusp anomalous dimension four loopsanomalous dimensions known loop lower adds uncertainty analysis turn to jet function @xmath9 arises integrating fluctuations intermediate scale @xmath5 matching scet hqet calculation jet function evaluation imaginary vacuum matrix element in qcd calculated to nnlo in @xcite matrix element quark propagator jet function depends on renormalization scale @xmath33 fixed at @xmath34 gev blnp possible derive solve rg equation for @xmath9 vary renormalization scale jet function starting point integro - differential evolution equation derived in @xcite @xmath35 j(p^2,\mu ) - 2\gamma_{\rm cusp^2 2 novel solution to evolution equation jrge presented in @xcite method laplace transformed jet function @xmath36 obeys local evolution equation analogous solving inverting laplace transform gives resummed @xmath9 in momentum spacepossible write partial decay rates integral over jet function @xmath37 solution rg equation write compact expression for resummed integrated jet function @xmath38 \left(\frac{q^2}{\mu_i^2}\right)^\eta j(\partial_\eta+\frac{q^2}_i^2} ){e{-\gamma_e\eta}}\gamma(1+\eta expressions for rg factors @xmath39 @xmath40 in @xcite equations resummedh resummedj allow evaluate integrals product @xmath41 arbitrary scale @xmath42 shape function renormalized study dependence partial decay rates under variations matching scales @xmath43 @xmath44 detailed numerical results in @xcite @xmath43 @xmath45 explore numerical impact of nnlo corrections partial rate with cut @xmath46 partial rate written as @xmath47 @xmath48 hard integrated jet functions evaluated in resummed forms function @xmath49 related to combinations @xmath26 deduced from @xcite evaluate ( [ eq : p+ ] ) requires model for shape functionuse two - parameter exponential model @xmath50 independent constraints on normalization factor @xmath51 parameters @xmath52 @xmath53 studying shape - function moments @xcite moments defined as @xmath54 obtained expansion local hqet operators @xmath55 zeroth moment fixes normalization shape determined to nnlo in @xcite define parameters @xmath56 @xmath57 shape - function scheme numerical values for hqet parameters @xmath4 @xmath57 at nnlo determined from global fits two - loop conversion relations @xcite tunes parameters @xmath52 @xmath53 give appropriate results for @xmath4 @xmath57 tune parameters @xmath58 gev @xmath59 at nnlo shape - function scheme figure results for partial rate cut @xmath46 choice @xmath60 gev function of hard matching scale @xmath43 dependence on renormalization scale @xmath43 reduced higher orders in perturbation theory large downward shift in central value between nlo nnlo value of @xmath3 nlo calculation blnp approachnumerical results preliminary finalized in @xcite uncertainties variations matching scale @xmath61 results partial rates cuts hadronic invariant mass lepton energy presented partial decay rates sensitive to value @xmath4 @xmath62 moments excluded from global fits photon regions @xmath4 closer to @xmath63 changing @xmath4 raises partial decay rates 20% precise determination @xmath3 inclusive decays point. bosch b o. lange m. neubert g. paz nucl . b * 699 * ( 2004 ) 335 arxiv hep - ph/0402094 lange neubert paz . rev . d 72 * ( 2005 ) 073006 hep - ph/0504071 p. gambino giordano. ossola n uraltsev jhep * 0710 * ( 2007 ) 058 arxiv:0707.2493 andersen e. gardi jhep * 0601 * ( 2006 ) 097 hep - ph/0509360 bigi shifman uraltsev vainshtein. j. mod . . * 9 2467 ( 1994 ) hep - ph/9312359 ]. korchemskysterman phys 340 96 1994 ) hep - ph/9407344 akhoury rothstein phys d 54 1996 2349 hep - ph/9512303 bauer manohar phys d 70 2004 034024 hep - ph/0312109 becher neubert . 637 251 2006 ) hep - ph/0603140 bonciani ferroglia arxiv:0809.4687 hep - ph asatrian c. greub d pecjak arxiv:0810.0987 hep - ph | 0.501747 | 3,136 | 1,634 | , Saving $0.1 in GPT-4. |
20,961 | nonlinear dynamics of a bouncing ball moving vertically in a gravitational field and colliding with a moving limiter is considered and the poincar map , describing evolution from an impact to the next impact , is described .
displacement of the limiter is assumed as periodic , cubic function of time . due to simplicity of this function analytical computations
are possible .
several dynamical modes , such as fixed points , @xmath0 - cycles and chaotic bands are studied analytically and numerically .
it is shown that chaotic bands are created from fixed points after first period doubling in a corner - type bifurcation .
equation for the time of the next impact is solved exactly for the case of two subsequent impacts occurring in the same period of limiter s motion making analysis of chattering possible . | in the present paper we study dynamics of a small ball moving vertically in a gravitational field and impacting with a periodically moving limiter ( a table ) .
this model belongs to the field of nonsmooth and nonlinear dynamical systems @xcite . in such systems
nonstandard bifurcations such as border - collisions and grazing impacts leading often to complex chaotic motions are typically present .
it is important that nonsmooth systems have many applications in technology @xcite . in the bouncing ball dynamics it is usually difficult or even impossible to solve nonlinear equation for an instant of the next impact .
we approached this problem assuming a special motion of the table .
recently , we have considered several models of motion of a material point in a gravitational field colliding with a limiter moving periodically with piecewise constant velocity @xcite and velocity depending linearly on time @xcite . in the present work we study the model in which periodic displacement of the table is a cubic function of time , carrying out our project to approximate the sinusoidal motion of the table as exactly as possible but preserving possibility of analytical computations aobr2010c .
the paper is organized as follows . in section 2 a one dimensional dynamics of a ball moving in a gravitational field and colliding with a table
is reviewed and the corresponding poincar map is constructed .
a bifurcation diagram is computed for displacement of the table assumed as cubic and periodic function of time . in the next section dynamical modes shown in the bifurcation diagram such as fixed points , @xmath0 - cycles and chaotic bands as well as the case of @xmath1 impacts in one interval of the limiter s motion
are studied analytically and numerically .
we summarize our results in section 4 .
let a ball moves vertically in a constant gravitational field and collides with a periodically moving table .
we treat the ball as a material point and assume that the limiter s mass is so large that its motion is not affected at impacts .
dynamics of the ball from an impact to the next impact can be described by the following poincar map in nondimensional form aobr2007 ( see also ref .
@xcite where analogous map was derived earlier and ref .
@xcite for generalizations of the bouncing ball model ) : [ tv ] @xmath2where @xmath3 denotes time of the @xmath4-th impact and @xmath5 is the corresponding post - impact velocity while @xmath6 .
the parameters @xmath7 , @xmath8 are a nondimensional acceleration and the coefficient of restitution , @xmath9 @xcite , respectively and the function @xmath10 represents the limiter s motion . the table s motion has been usually assumed in form @xmath11 , cf . @xcite and references therein . in this case
it is basically impossible to solve the eq.([t ] ) for @xmath12 .
accordingly , we have decided to choose the limiter s periodic motion in a polynomial form to make analytical investigations of the dynamics possible . in our previous papers
we have assumed displacement of the table as piecewise linear periodic function of time @xcite as well as quadratic aobr2010b . in this work
we study dynamics for a cubic function of time @xmath13 : [ c1 ] @xmath14with @xmath15 , where @xmath16 is the floor function the largest integer less than or equal to @xmath17 ( i.e. @xmath18 ) . since the period of motion of the limiter is equal to one , the map ( [ tv ] ) is invariant under the translation @xmath19 .
accordingly , all impact times @xmath3 can be reduced to the unit interval @xmath20 $ ] .
the model consists thus of equations ( [ tv ] ) , ( c1 ) with control parameters @xmath8 , @xmath7 . , @xmath21 $ ] . ] in fig .
[ f1 ] above we show the bifurcation diagram with impact times versus @xmath7 computed for growing @xmath7 and @xmath22 .
it follows that dynamical system ( [ tv ] ) , ( [ c1 ] ) has several attractors : two fixed points which after one period doubling give rise to chaotic bands and two other fixed points which go to chaos via period doubling scenario .
there are also several small attractors .
we shall investigate some of these attractors in the next section combining analytical and numerical approach .
general analytical conditions for birth of new modes of motion were given in @xcite .
we shall first study periodic solutions with one impact per @xmath23 periods .
such states have to fulfill the following conditions : @xmath24 where:@xmath25 substituting these conditions into ( [ tv ] ) , ( [ c1 ] ) we obtain two sets of fixed points:@xmath26where the impact occurs in time interval @xmath27 and @xmath28 with impacts taking place in time interval @xmath29 . solutions ( [ fixsta ] ) fulfill physical requirements and are stable in the following interval of @xmath7:@xmath30where lower bound is a consequence of @xmath31 while the upper bound follows from the condition that eigenvalues @xmath32 of the stability matrix obey @xmath33 .
accordingly , for @xmath22 there are only four stable fixed points with @xmath34 , shown in fig .
[ f1 ] they appear at @xmath35 and @xmath36 , respectively . on the other hand , solutions ( [ fixusta ] )
are always unstable and are physical for:@xmath37what is equivalent to the condition @xmath38 .
let us note that @xmath0 - cycles are created from fixed points such that @xmath39 , \ t_{i+1}\in
\left [ k , k+1\right ] $ ] with @xmath40 . therefore equations defining @xmath0 - cycles read:@xmath41where @xmath40 and @xmath42 are given by ( [ c1 ] ) .
we know that @xmath0 - cycles appear for @xmath43 , cf .
eq.([exst1 ] ) .
we have solved the set of equations ( [ 2cycleg ] ) in closed form but we skip these lengthy formulae because of lack of space . the bifurcation diagram shown in fig .
1 suggests that in the case of two fixed points with @xmath44 transition to chaos occurs after the first period doubling when @xmath45 .
therefore , in order to determine values of parameters at which the transition to chaos occurs , we have to solve eq.(2cycleg ) with condition @xmath46 . after this
substitution equations ( 2cycleg ) are easily solved to yield : @xmath47 where @xmath48 .
for example , for @xmath22 and @xmath49 we get from ( [ sol ] ) @xmath50 while for @xmath22 and @xmath51 we obtain @xmath52 and indeed , precisely at this point on the @xmath7 axis , branches of the corresponding @xmath0 - cycles reach values @xmath45 and transform into chaotic bands
. this scenario does not apply for @xmath22 and @xmath53 - in the case @xmath54 there is another period doubling prior to @xmath55 while for @xmath56 physical solutions of ( [ sol ] ) do not exist . in the bouncing ball dynamics chattering and chaotic dynamics
arise typically , see @xcite where chattering mechanism was studied numerically for sinusoidal motion of the table
. due to simplicity of our model analytical computations are possible .
let us assume that @xmath57 impacts , @xmath3 , @xmath12 , occur in the same period .
then the solution @xmath58 of equation ( [ t ] ) is always present .
we thus obtain from eqs.([t ] ) and ( [ ci ] ) : [ dd ] @xmath59 where @xmath60 , @xmath6 .
we can now rewrite eqs.([tv ] ) in simplified form : [ eqs ] @xmath61 where @xmath62 is a relative velocity and the solution @xmath63 , cf .
( [ dd ] ) , was chosen .
equations ( [ eqs ] ) define an _ explicit _ nonlinear map .
this formalism makes possible analysis of chattering and grazing in the @xmath64 limit .
it follows immediately that grazing manifold , @xmath65 , i.e. @xmath58 , exists only for @xmath66 and hence for @xmath67 after computing eigenvalues @xmath68 of the stability matrix @xmath69 on the grazing manifold , @xmath70 , we get after straightforward calculations @xmath71 , @xmath72 .
it thus follows that the grazing manifold is always attracting in one eigendirection and neutral in another ( @xmath73 ) . in the final stage of grazing @xmath74
is very small .
let @xmath75 .
it follows that for @xmath76 , @xmath77 , we get the convergent expansion of the square - root in ( [ del ] ) : @xmath78and the inequality @xmath79
implies:@xmath80 approximate equations describing chattering , obtained from the approximate equation ( [ approxdelta ] ) and exact equations ( [ eqs ] ) , are of form : [ eqsapprox ] @xmath81and to ensure convergence @xmath82 we assume that @xmath83 ( i.e. @xmath84 ) this leads to the following condition for @xmath3 : @xmath85 it is now possible to estimate time of @xmath1-th impact:@xmath86the time @xmath87 can be computed exactly since@xmath88where equations ( [ approxdelta ] ) , ( [ approxv ] ) were used , and thus@xmath89 if the ball impacts at time @xmath90 , performs infinite number of impacts and sticks ( in one interval of limiter s motion ) then the time is @xmath91 .
solving the inequality @xmath92 we get conditions for grazing : @xmath93 finally , the ( approximate ) condition for grazing is : @xmath94 , cf .
( [ ta1 ] ) , ( ta2 ) , ( [ tb ] ) , ( [ tc ] ) .
it has been shown that some chaotic bands are created from fixed points after first period doubling in a corner - type bifurcation and critical values of control parameter @xmath7 have been determined .
equations for @xmath1 impacts in one period of limiter s motion were found and simplified significantly , making analysis of chattering and grazing possible .
approximate equations describing final stage of chattering were obtained and ( approximate ) condition for grazing was computed in analytical form .
a. c. j. luo , y. guo , motion switching and chaos of a particle in a generalized fermi - acceleration oscillator , mathematical problems in engineering , vol . *
2009 * , article i d 298906 , 40 pages , 2009 . doi:10.1155/2009/298906 . | present paper dynamics small ball moving vertically gravitational field impacting periodically moving limiter table model belongs nonsmooth nonlinear dynamical systems @xcite nonstandard bifurcations border - collisions grazing impacts complex chaotic motions present nonsmooth systems applications in technology @xcite bouncing ball dynamics difficult solve nonlinear equation instant next impact assuming special motion table considered models motion material point gravitational field colliding with limiter moving periodically constant velocity depending on time present work study model periodic displacement table cubic function of time approximate sinusoidal motion table preserving analytical computations paper organized section 2 one dimensional dynamics ball moving gravitational field colliding table reviewed poincar map constructed bifurcation diagram computed for displacement table cubic periodic function time next section dynamical modes fixed points - cycles chaotic bands @xmath1 impacts interval limiter s motion studied analytically numerically summarize results section 4 ball moves vertically constant gravitational field collides with periodically moving table treat ball as material point limiter s mass large motion not affected at impactsdynamics ball from impact to described by poincar map nondimensional form ref @xcite bouncing ball model : @xmath2where @xmath3 time impact @xmath5 post - impact velocity parameters @xmath7 @xmath8 are nondimensional acceleration coefficient of restitution @xcite function @xmath10 represents limiter s motion table s motion assumed in @xmath11 impossible to solve eq.([t ] ) for @xmath12 choose limiter s periodic motion in polynomial form previous papers assumed displacement table as linear periodic function of time quadratic study dynamics for cubic function of time @xmath13 @xmath14with @xmath15 @xmath16 floor function largest integer less than or equal to @xmath17 @xmath18 period of motion limiter equal to one map tv invariant under translation @xmath19 all impact times @xmath3 reduced to unit interval @xmath20 model consists of equations ( tv ] c1 ) with control parameters @xmath8 @xmath7 @xmath21 fig .bifurcation diagram impact times versus @xmath7 growing @xmath7 @xmath22 dynamical system tv c1 has attractors two fixed points chaotic bands two chaos small attractors investigate attractors next section combining analytical numerical approach analytical conditions for new modes motion in @xcite study periodic solutions one impact per @xmath23 periods states fulfill conditions @xmath24@xmath25 substituting conditions into tv two sets fixed points impact occurs in time interval @xmath27 @xmath28 interval @xmath29 solutions fixsta fulfill physical requirements stable in interval @xmath7:@xmath30where lower bound consequence of @xmath31 upper bound follows from condition eigenvalues @xmath32 obey @xmath33 @xmath22 four stable fixed points with @xmath34 fig at @xmath35 @xmath36 solutions ( fixusta unstable physical for@xmath37what equivalent to condition @xmath38 @xmath0 - cycles created from fixed points @xmath39 @xmath40 equations defining @xmath0 - cycles read@xmath41where @xmath40 @xmath42 given by ( [ c1 ] ) @xmath0 - cycles for @xmath43 eq.([exst1 ] ) solved equations [ 2cycleg ] ) skip lengthy formulae lack space bifurcation diagram fig . 1 two fixed points with @xmath44 transition to chaos after first doubling @xmath45 solve eq.(2cycleg ) with @xmath46 substitution equations ( 2cycleg ) solved @xmath47 @xmath48 @xmath22 @xmath49 from @xmath50 @xmath22 @xmath51 obtain @xmath52 branches @xmath0 - cycles reach @xmath45 transform into chaotic bands for @xmath22 @xmath53 @xmath54 another period doubling prior to @xmath55 @xmath56 solutions of ( [ sol ] ) bouncing ball dynamics chattering chaotic dynamics arise sinusoidal motion simplicity analytical computations possible @xmath57 impacts @xmath3 @xmath12 same period solution @xmath58 of equation ( [ t ] ) always present obtain from eqs.([t ] ) ) : [ dd @xmath59 @xmath60 @xmath6 rewrite eqs.[tv ] simplified form [ eqs ] @xmath61 @xmath62 relative velocity solution @xmath63 dd chosen equations eqs define nonlinear map formalism analysis chattering grazing @xmath64 limit grazing manifold @xmath65 @xmath58 exists @xmath66 @xmath67 computing eigenvalues @xmath68 stability matrix @xmath69 grazing manifold @xmath70 calculations @xmath71 @xmath72 grazing manifold attracting one eigendirection neutral another @xmath73 final stage grazing @xmath74 small @xmath75 @xmath76 @xmath77 convergent expansion - root del ] inequality @xmath79 implies@xmath80 approximate equations describing chattering equation approxdelta ] exact equations [ eqs ] [ eqsapprox ] ensure convergence @xmath82 assume @xmath83@xmath84 leads condition @xmath3 @xmath85 estimate time @xmath1-th impact time @xmath87 computed since equations approxdelta approxv ] used@xmath89 ball impacts @xmath90 infinite impacts interval limiter s motion time is @xmath91 solving inequality @xmath92 conditions for grazing @xmath93 approximate condition grazing @xmath94 ta1 ta2 tb tc ] chaotic bands created from fixed points after doubling corner bifurcation critical values control parameter @xmath7 determined equations @xmath1 impacts period limiter s motion simplified analysis chattering grazing possible approximate equations final stage chattering obtained condition grazing computed analytical form c. j. luo y. guo motion switching chaos particle generalized fermi - acceleration oscillator mathematical problems engineering 2009 article 298906 40 pages doi:10.1155/2009/298906 | 0.353624 | 2,410 | 1,245 | , Saving $0.1 in GPT-4. |
20,964 | we consider a globally invariant chiral lagrangian that contains vector and axial - vector mesons .
we compute the @xmath0 decay width and @xmath1 scattering lengths and compare with the corresponding results in which the ( axial-)vector degrees of freedom decouple .
we show that the role of vector mesons has a great impact on these quantities . | one way to investigate quantum chromodynamics ( qcd ) at low energies is by using effective models - such as the linear sigma model - that possess the same global symmetries as qcd , most notably the chiral @xmath2 symmetry ( with @xmath3 flavours ) . in these effective models
hadronic degrees of freedom , rather than quark and gluon degrees of freedom , are present ; spontaneous breaking of the chiral invariance leads to the emergence of nearly massless pseudoscalar goldstone bosons ( e.g. , pions for @xmath4 ) .
their chiral partners , the scalar states , remain massless . if the linear sigma model is used , there are at least two possibilities to choose the hadronic degrees of freedom in the corresponding lagrangian : with ( pseudo-)scalar degrees of freedom only , or with the addition of ( axial- ) vector degrees of freedom . in this paper
we present a linear sigma model _ with _ vector and axial - vector mesons for @xmath4 and discuss two important quantities : the @xmath5 scattering lengths and the decay width of @xmath6 ( in our notation : @xmath7 ) into two pions . in particular , we aim to investigate the role of ( axial-)vector mesons for these quantities . to this end
we investigate the limit in which the ( axial-)vector mesons decouple and compare the results with the results of the full model .
the paper is organized as follows : in section 2 the model is introduced , in section 3 the results with and without vector mesons are presented and in section 4 we present our conclusions .
the model is described by a @xmath8 chirally symmetric lagrangian @xcite : @xmath9-m_{0}^{2}\mathrm{tr}(\phi^{\dagger}\phi)-\lambda_{1}[\mathrm{tr}% ( \phi^{\dagger}\phi)]^{2}-\lambda_{2}\mathrm{tr}(\phi^{\dagger}\phi)^{2}% } \nonumber\\ & -\frac{1}{4}\mathrm{tr}[(l^{\mu\nu})^{2}+(r^{\mu\nu})^{2}]+\frac{m_{1}^{2}% } { 2}\mathrm{tr}[(l^{\mu})^{2}+(r^{\mu})^{2}]+\mathrm{tr}[h(\phi+\phi^{\dagger } ) ] \nonumber\\ & + c(\det\phi+\det\phi^{\dagger})-2ig_{2}(\mathrm{tr}\{l_{\mu\nu}[l^{\mu } , l^{\nu}]\}+\mathrm{tr}\{r_{\mu\nu}[r^{\mu},r^{\nu}]\})\nonumber\\ & + \frac{h_{1}}{2}\mathrm{tr}(\phi^{\dagger}\phi)\mathrm{tr}[(l^{\mu}% ) ^{2}+(r^{\mu})^{2}]+h_{2}\mathrm{tr}[\vert l^{\mu}\phi \vert ^{2 } + \vert \phi r^{\mu } \vert ^{2}]\nonumber\\ & + 2h_{3}\mathrm{tr}(\phi r_{\mu}\phi^{\dagger}l^{\mu})+ ... \text { , } \label{lagrangian}%\end{aligned}\ ] ] where @xmath10 ( scalar and pseudoscalar degrees of freedom ) , @xmath11 and @xmath12 ( vector and axial - vector degrees of freedom ; our model is currently constructed for @xmath4 - thus , our eta meson @xmath13 contains only non - strange degrees of freedom ) ; @xmath14 , @xmath15 are the generators of @xmath16 ; @xmath17 $ ] ( @xmath18 denotes the photon field ) , @xmath19-(\partial^{\nu}l^{\mu } -iea^{\nu}[t^{3},l^{\mu}])$ ] , @xmath20-(\partial^{\nu}r^{\mu}-iea^{\nu}[t^{3},r^{\mu}]).$ ] the dots refer to further globally invariant terms which are irrelevant in the following .
the explicit breaking of the global symmetry is described by the term tr@xmath21\equiv h\sigma$ ] @xmath22 .
the chiral anomaly is described by the term @xmath23 @xcite .
note that the model possesses global chiral invariance ; the reasons to use the global invariance ( instead of the local one ) may be found , e.g. , in ref .
@xcite as well as refs . @xcite .
the reason to consider operators up to the fourth order only has been presented in ref .
@xcite . when @xmath24 , spontaneous symmetry breaking @xmath25 takes place .
the sigma field is shifted by its ( constant ) vacuum expectation value @xmath26 : @xmath27 as a result of this shift non - diagonal mixing terms of the form @xmath28 and @xmath29 arise . in order to remove them , the fields @xmath30 and @xmath31 are also shifted as @xmath32 , @xmath33 with @xmath34 and the pseudoscalar fields are renormalized : @xmath35 , @xmath36 where @xmath37 @xcite .
the identification of mesons with particles listed by the pdg @xcite is straightforward in the pseudoscalar and ( axial-)vector sectors : the fields @xmath38 and @xmath13 correspond to the pion and the @xmath39 counterpart of the @xmath40 meson , @xmath41 ( with a mass @xmath42 of about @xmath43 mev , obtained by unmixing the physical @xmath40 and @xmath44 mesons ) .
the fields @xmath45 , @xmath46 represent the @xmath47 and the @xmath48 vector mesons ; the fields @xmath49 , @xmath50 represent the @xmath51 and @xmath52 axial - vector mesons .
unfortunately , the identification of the @xmath7 and @xmath53 fields is controversial with the possibilities being the pairs @xmath54 and @xmath55 . for the purpose of this paper
we chose the assignment @xmath56 , which allows us to study the decay @xmath57 and the @xmath5 scattering lengths .
eleven parameters are present in the lagrangian : @xmath58 , @xmath59 , @xmath60 , @xmath61 , @xmath62 , @xmath63 , @xmath64 , @xmath65 , @xmath66 , @xmath67 , @xmath68 .
using the masses @xmath69 , @xmath42 , @xmath70 , @xmath71 and @xmath72 , the pion decay constant @xmath73 ( via the eq .
@xmath74 ) and the experimentally well - known decay widths @xmath75 mev and @xmath76 mev @xcite , one can eliminate eight parameters .
all quantities of interest can be then expressed in terms of three independent combinations of parameters , which for convenience are chosen to be @xmath77 , @xmath78 and @xmath62 .
explicit expressions and a detailed description of the procedure can be found in ref .
@xcite .
for the present purpose , it is useful to write explicitly the equation for the @xmath79 mass@xmath80 \label{rhomass}%\ ] ] where the decomposition of @xmath81 in terms of two terms is evident : the term @xmath82 $ ] represents the contribution of the chiral condensate ( and vanishes in the limit @xmath83 ) , while @xmath84 is independent of the chiral condensate and can be related to the gluon condensate . note that we made explicit that @xmath85 and @xmath86 are functions of @xmath87 in the following we will consider @xmath88 as varying between 0 and @xmath89 :
in fact , a negative @xmath88 would imply that the vacuum is not stable in the limit @xmath90 ; @xmath91 would imply that the contribution of the chiral condensate is negative .
this is counter - intuitive and at odds with various microscopic approaches such as the njl model .
_ * results with ( axial-)vector mesons : * _ the scattering lengths @xmath92 and @xmath93 and @xmath94 $ ] are functions of @xmath95 , @xmath78 and @xmath62 ( explicit expressions can be found in ref .
we use the experimental values @xmath96 and @xmath97 from the na48/2 collaboration @xcite and @xmath98 mev @xcite .
a standard fit procedure allows us to determine the parameters @xmath95 , @xmath78 , and @xmath62 and their errors : @xmath99 , @xmath100 and @xmath101 mev .
obviously , the errors of @xmath62 and @xmath95 are too large .
a drastic improvement can be obtained via a physical consideration : as we discussed after eq .
( [ rhomass ] ) we impose that @xmath84 is varying between @xmath102 and @xmath81 . as a consequence for z=1.67 , the parameter @xmath103 needs to lie in the interval @xmath104\longleftrightarrow h_{1}=-68_{-15}^{+36}$ ] ( the upper boundary of @xmath62 corresponds to @xmath105 ) .
the value of @xmath95 can be constrained in a similar way using the scattering length @xmath106 : @xmath107 $ ] mev @xmath108 mev . with these parameter values
it is possible to calculate the @xmath7 decay width , @xmath109 which is shown in the following _ fig .
1_. the three lower curves in the plot depict the @xmath7 decay width in the model that contains vector and axial - vector degrees of freedom .
it is clear that the decay width becomes smaller when the vector and axial - vector degrees of freedom are introduced . in that case , the value of the @xmath7 decay width changes very slowly with @xmath95 so that even for @xmath110 mev the width is not larger than approximately 300 mev - a value that is too small when compared , e.g. , with the results from refs .
@xcite . note that the value of @xmath110 mev is actually too large to agree with the scattering length @xmath106 ( as apparent from _
2 ) if the ( axial-)vector fields are considered - then , the @xmath7 mass should not be larger than 450 mev which in turn yields a @xmath7 decay width smaller than about 100 mev .
@xmath111 + _ * results without ( axial-)vector mesons : * _ this limit is obtained for @xmath112 , i.e. @xmath113 , and @xmath114 in this case the decay width @xmath115 and the scattering length @xmath116 are functions of @xmath117 only and are plotted in _ fig . 1 _ and _ fig . 2
_ , respectively .
the scattering length @xmath106 allows for @xmath118 $ ] mev ( see _ fig .
thus , the value of the @xmath6 decay width in this case is _ much larger _ than in the previous cases : given that @xmath119 $ ] mev , one obtains @xmath120 $ ] mev ( see _ fig . 1 _ ) .
these values are in agreement with the result obtained by leutwyler _
( @xmath121 mev ) @xcite and the result @xmath122 mev obtained by pelez _
however , this limit also implies that @xmath123 , which is clearly at odds with experiment @xcite .
the @xmath7 model without ( axial-)vector degrees of freedom can properly describe @xmath5 scattering and @xmath6 phenomenology .
this is also one of the reasons for its success in low - energy hadron physics .
however , the inclusion of vector and axial - vector fields drastically worsens the agreement with data : a simultaneous description of both @xmath5 scattering lengths and @xmath57 is no longer possible .
this is due to the fact that a notable decrease in the decay width of the @xmath6 resonance , @xmath124 mev , takes place and due to the important role of the @xmath79 meson in @xmath5 scattering . in light of these results
one could be led to neglect ( axial-)vector degrees of freedoms .
this is , however , not feasible : vector and axial - vector mesons are well - known quark - antiquark states which should necessarily be part of a realistic low - energy model . neglecting them
would imply that they decouple from the pseudoscalar goldstone bosons , which is obviously at odds with the experimentally measured decay widths @xmath125 and @xmath126 .
a way out of the problem is to change the scenario for scalar mesons by identifying the scalar fields of the model with @xmath55 ( as done - for instance - in refs .
still , a light scalar resonance with mass of about 400 mev is needed to describe @xmath5 scattering : the resonance @xmath6 can be interpreted as an additional tetraquark state . in this way , mixing of tetraquark and quarkonia takes place @xcite .
a detailed study of this possibility and the generalization to @xmath127 ( involving also the scalar glueball ) are possible extensions of the present approach .
a. h. fariborz , r. jora and j. schechter , phys .
d * 72 * ( 2005 ) 034001 [ arxiv : hep - ph/0506170 ] .
m. napsuciale and s. rodriguez , phys .
d * 70 * ( 2004 ) 094043 [ arxiv : hep - ph/0407037 ] .
f. giacosa , phys .
d * 75 * , 054007 ( 2007 ) [ arxiv : hep - ph/0611388 ] .
a. heinz , s. struber , f. giacosa and d. h. rischke , phys .
d * 79 * ( 2009 ) 037502 [ arxiv:0805.1134 [ hep - ph ] ] . | quantum chromodynamics qcd ) at low energies models linear sigma model same global symmetries as qcd notably chiral @xmath2 symmetry with @xmath3 flavours models hadronic degrees freedom quark gluon present breaking chiral invariance leads massless pseudoscalar goldstone bosons pions for @xmath4 chiral partners scalar states remain massless linear sigma model two possibilities choose hadronic degrees freedom with pseudo-)scalar only or axial- ) vector degrees freedom paper present linear sigma model with vector axial - vector mesons for @xmath4 discuss quantities @xmath5 scattering lengths decay width of @xmath6 @xmath7 into two pions investigate role of axial-)vector mesons for these quantities limit mesons decouple compare results with full model paper organized section 2 model introduced section 3 results with and without vector mesons presented section 4 conclusionsmodel described @xmath8 lagrangian @xcite @xmath9-m{2}\mathrm{tr}\phi\dagger\lambda{1}\mathrm{tr{2}-\lambda{2}\mathrm{tr\dagger{2}%\nonumber\\\frac{1}{4}\mathrm{tr}^{2}+\mu^{2}]+\frac{m_{1}{2}%}\mathrm{tr{2}+{\mu}^{2}]+\mathrm{tr}\phi+\phi{\dagger \nonumber\\\phi\dagger})-2ig_{2}\mathrm{tr}{\mu\nu}+\mathrm{tr}{\mu\nu}\nonumber\\\frac{h_{1}}{2}\mathrm{tr}\phi\dagger}\phi\mathrm{tr}{%{2}+(r{\mu}^{2}]+h_{2}\mathrm{tr{\mu}{2{\mu{2}]\nonumber\\ 2h_{3}\mathrm{tr}\mu}\phi^{\dagger}l^{\mu})+ \text { \label{lagrangian}%\end{aligned} @xmath10 scalar pseudoscalar @xmath11 @xmath12 vector axial - vector freedom model for @xmath4 meson @xmath13 contains non - strange degrees freedom @xmath14 @xmath15 generators of @xmath16 @xmath17 @xmath18 photon field @xmath19-(\partial^{\nu}l^{\mu{}[t{3 @xmath20-\nu}r dots refer globally invariant terms irrelevant explicit breaking global symmetry tr@xmath21\equiv h\sigma$ @xmath22 chiral anomaly @xmath23 @xcite model global chiral invariance reasons use global invariance ref . @xcite reason consider operators fourth order ref . @xcite @xmath24 spontaneous symmetry breaking @xmath25 sigma field shifted by vacuum expectation value @xmath26 @xmath27 non - diagonal mixing terms @xmath28 @xmath29 ariseremove fields @xmath30 shifted as @xmath32 @xmath34 pseudoscalar fields renormalized @xmath35 @xmath37 @xcite identification of mesons with particles pdg @xcite straightforward in pseudoscalar axial-)vector sectors fields @xmath38 @xmath13 correspond pion @xmath39 counterpart @xmath40 meson @xmath41 mass @xmath42 @xmath43 mev unmixing @xmath40 @xmath44 mesons fields @xmath45 @xmath46 represent @xmath47 @xmath48 vector mesons @xmath49 @xmath50 @xmath51 @xmath52 axial - vector mesons identification @xmath7 @xmath53 fields controversial possibilities pairs @xmath54 @xmath55 chose assignment @xmath56 study decay @xmath57 @xmath5 scattering lengths eleven parameters present in lagrangian @xmath58 using masses @xmath69 @xmath42 @xmath72 pion decay constant @xmath73 decay widths @xmath75 mev @xmath76 mev @xcite eliminate eight parametersquantities interest expressed three independent parameters @xmath77 @xmath78 @xmath62 expressions detailed description procedure in ref . @xcite write equation for @xmath79 mass@xmath80 decomposition of @xmath81 two terms evident term @xmath82 represents contribution chiral condensate vanishes in limit @xmath83 @xmath84 independent of chiral condensate related to gluon condensate @xmath85 @xmath86 functions of @xmath87 @xmath88 varying between 0 and @xmath89 negative @xmath88 vacuum not stable in limit @xmath90 @xmath91 contribution chiral condensate negative counter - intuitive at odds with microscopic approaches njl model results axial-)vector mesons scattering lengths @xmath92 @xmath93 @xmath94 functions of @xmath95 @xmath78 @xmath62 explicit expressions in ref use experimental values @xmath96 @xmath97 from na48/2 collaboration @xcite @xmath98 standard fit procedure determine parameters @xmath95 @xmath78 @xmath62 errors errors of @xmath62 @xmath95 too large drastic improvement via physical considerationimpose @xmath84 between and z=1.67 parameter @xmath103 in interval @xmath104}=-68_{-15}+36} upper boundary @xmath62 corresponds to @xmath105 value @xmath95 constrained using scattering length @xmath106 : @xmath107 @xmath108 mev parameter values calculate @xmath7 decay width @xmath109 shown fig . 1_ three lower curves depict @xmath7 decay width model vector axial - vector degrees freedom decay width becomes smaller when vector degrees freedom introduced value @xmath7 decay width changes slowly with @xmath95 @xmath110 mev width not larger than 300 mev too small compared results value @xmath110 mev too large agree with scattering length @xmath106 axial-)vector fields considered @xmath7 mass not larger than 450 mev yields @xmath7 decay width smaller than 100 mev @xmath111 results without axial-)vector mesons limit obtained for @xmath112@xmath113 @xmath114 decay width scattering length functions of @xmath117 plotted in fig . 1 fig . 2 scattering length @xmath106 allows @xmath118 $ ] mev @xmath6 decay width larger than previous @xmath119 $ ] mev obtains @xmath120 $ ] mev agreement with result leutwyler @xmath121 mev @xcite @xmath122 mev pelez limit implies @xmath123 at odds with experiment @xcite @xmath7 model without axial-)vector degrees freedom describe @xmath5 scattering @xmath6 phenomenology success in low - energy hadron physics inclusion of vector axial - vector fields worsens agreement simultaneous description of @xmath5 scattering lengths @xmath57 possible due decrease in decay width @xmath6 resonance @xmath124 mev role @xmath79 meson in @xmath5 scattering neglect axial-)vector degrees freedoms not feasible vector axial - vector mesons quark - antiquark states part of low - energy modelneglecting decouple pseudoscalar goldstone bosons odds with decay widths @xmath125 @xmath126 change scenario scalar identifying scalar fields with @xmath55 light scalar resonance 400 mev needed @xmath5 scattering resonance @xmath6 additional tetraquark state mixing tetraquark quarkonia takes @xcite detailed study generalization to @xmath127 scalar glueball extensions approach fariborz jora j schechter phys . d * 72 * 2005 ) 034001 hep - ph/0506170 napsuciale s rodriguez . d * 70 * 2004 ) 094043 hep - ph/0407037 f giacosa . d * 75 * 054007 2007 ) hep - ph/0611388 heinz struber giacosa rischke phys . d * 79 * 2009 ) 037502 arxiv:0805.1134 | 0.498448 | 3,206 | 1,732 | , Saving $0.1 in GPT-4. |
20,967 | we have completed a new fiber array , sparsepak , optimized for low - surface - brightness studies of extended sources on the wiyn telescope . we are now using this array as a measuring engine of velocity and velocity - dispersion fields of stars and ionized gas in disk galaxies from high to low surface - brightness . here
we present commissioning data on the velocity ellipsoids , surface densities and mass - to - light ratios in two blue , high surface - brightness , yet small disks .
if our preliminary results survive further observation and more sophisticated analysis , then ngc 3949 has @xmath0 , implying strong vertical heating , while ngc 3982 s disk is substantially sub - maximal .
these galaxies are strikingly unlike the milky way , and yet would be seen more easily at high redshift .
-0.25 in # 1_#1 _ # 1_#1 _ = # 1 1.25 in .125 in .25 in | -0.05 in sparsepak contains a 76@xmath177 arcsec sparsely - packed hexagonal grid of 75 , 5-arcsec diameter fibers , which pipe light from the `` imaging '' nasmyth focus to the versatile bench spectrograph on the wiyn 3.5 m telescope ( bershady et al .
seven `` sky '' fibers are placed at @xmath2 30 arcsec from two grid sides .
the large anamorphic demagnification in the echelle - mode of the bench spectrograph yields instrumental spectral resolutions of 8000 - 10,000 and as high as 21,000 with sparsepak
comparable to resolutions obtained with smaller fibers . at the same time , sparsepak has @xmath3% throughput for @xmath4 nm , and provides nearly 3@xmath1 the light gathering power over 5@xmath1 the area as densepak , the pre - existing wiyn fiber array .
these attributes make sparsepak well suited for medium - resolution spectroscopy on extended objects down to low surface - brightness .
-0.15 in we have mapped the velocity and velocity dispersion fields in two blue , very high surface - brightness , yet small disks in the ursa major cluster ( b@xmath5r=0.8 - 1 mag , @xmath6(b)=19.3 - 19.5 mag arcsec@xmath7 , h@xmath8=0.9 - 1.7 kpc , verheijen 1996 ) .
ngc 3982 is nearly face - on ( i=26@xmath9 ) , while ngc 3949 is moderately inclined ( i=54@xmath9 ) . as such
, these photometrically similar galaxies offer projections which enhance measurements of different components of their velocity ellipsoids .
each galaxy was observed in three spectral regions near 5130(mgi ) , 6680 ( h@xmath10 ) and 8660 ( caii - triplet ) with fwhm spectral resolutions of 24 , 16 and 37 km / s respectively .
stellar absorption - line observations were taken with a single sparsepak pointing while h@xmath10 was contiguously mapped in three pointings at very high s / n .
the latter provided high - precision kinematic inclinations ( see andersen & bershady , these proceedings ) . in the inner regions of the galaxy ,
the signal - to - noise in the absorption - line spectra of individual fibers is high enough to determine the line centroids and widths and to construct stellar velocity fields ( figure 1 ) .
the layout of the fibers allows for azimuthal averaging of fibers to improve signal - to - noise in the outer regions ( figure 2 ) .
the near face - on orientation of ngc 3982 , for example , permits 18 fibers to be combined at a radius of 3.5 h@xmath11 . before averaging ,
the projected rotational velocities were taken out , using the centroids of the high signal - to - noise h@xmath10 lines . in both galaxies
this may lead to an overestimate of the inner velocity dispersions .
comparison of the h@xmath10 and stellar velocity fields in figure 1 clearly shows the effects of asymmetric drift inside 1.5 disk scale lengths ( h@xmath8 ) . outside this radius the asymmetric drift approaches zero .
two results stand out from these initial measurements , illustrated in figures 2 and 3 .
first , for ngc 3949 , the ratio of major and minor axis velocity dispersions are unity ( @xmath12% ) , while the epicycle approximation yields @xmath13 . from this
we find the surprising result that @xmath14 must be much larger than either @xmath15 or @xmath16 , implying at face value that ngc 3949 has a dynamically unusual ( hot ) disk ( cf gerssen et al . 2000 ) .
second , the velocity dispersion in the disk of ngc 3982 is @xmath2 constant with radius until the last radial datum , where @xmath17 falls to a significantly lower value .
( no corrections were applied for contributions from the projected radial and tangential velocity dispersion components a @xmath1813% effect . )
we derive a _ total _
disk mass surface density for a variety of scale heights , based on results from recent work by kregel et al .
( 2002 ) , and assuming an isothermal ( sech@xmath19 ) vertical density profile .
this mass surface density is at a level @xmath23.5 higher than the surface density of the milky way in the solar neighborhood ( kuijken & gilmore 1991 ) , and is nearly constant with radius until r / h@xmath8=3 .
consequently the mass - to - light ratio of the disk increases with radius for r / h@xmath20 , while the disk color gets bluer a trend opposite that indicated by the models of bell & de jong ( 2000 ) .
however , our color range is limited , and we must also check whether red super - giants are in abundance in this system . finally , a weighted radial average of ( m / l@xmath21)=0.18 is used in a rotation curve decomposition shown in figure 2 which implies a substantially ( @xmath22% ) sub - maximum , yet very high sb disk ( see also verheijen et al .
this research was supported by nsf / ast-9970780 . | sparsepak contains 76@xmath177 arcsec hexagonal grid 75 5-arcsec fibers light from nasmyth focus to bench spectrograph wiyn 3.5 m telescope seven sky fibers at @xmath2 30 arcsec sides large anamorphic demagnification echelle - mode spectrograph yields resolutions 8000 - 10,000 21,000 sparsepak smaller sparsepak has @xmath3% throughput for @xmath4 nm provides 3@xmath1 light gathering power over 5@xmath1 area sparsepak for medium - resolution spectroscopy low surface - brightness mapped velocity dispersion fields in two small disks in ursa major cluster b@xmath5r=0.8 - 1 mag @xmath6)=19.3 - 19.5 mag@xmath8=0.9 - 1.7 kpc ngc 3982 nearly face - on ngc 3949 moderately inclined similar galaxies offer projections measurements velocity ellipsoids galaxy observed in three spectral regions near 5130 6680 8660 with spectral resolutions 24 16 37 km / sstellar absorption - line observations single sparsepak pointing h@xmath10 mapped three pointings high s / n latter provided high - precision kinematic inclinations andersen bershady inner regions galaxy signal - to - noise absorption - line spectra high line centroids widths construct stellar velocity fields figure layout allows azimuthal averaging signal - to - noise outer regions face - ngc 3982 permits 18 fibers radius 3.5 h@xmath11 averaging projected rotational velocities taken centroids high signal - to - noise h@xmath10 lines overestimate inner velocity dispersions comparison h@xmath10 stellar velocity fields shows asymmetric drift inside 1.5 disk scale lengths h@xmath8 outside radius asymmetric drift approaches zero two results measurements figures 2 3 ngc 3949 ratio major minor axis velocity dispersions unity @xmath12% ) epicycle approximation yields @xmath13 @xmath14 larger than @xmath15 @xmath16 ngc 3949 dynamically unusual disk velocity dispersion disk ngc 3982 @xmath2 constant radius until last radial datum @xmath17 lowerno corrections for radial tangential velocity dispersion @xmath1813% effect derive total disk mass surface density for scale heights based kregel et al. 2002 isothermal@xmath19 ) vertical density profile mass surface density @xmath23.5 higher than milky way in solar neighborhood constant with radius until r / h@xmath8=3 mass - to - light ratio increases with radius for r / h@xmath20 disk color bluer bell & de jong 2000 color range limited check red super - giants weighted radial average of ( m / l@xmath21)=0.18 in rotation curve decomposition figure 2 implies @xmath22% ) sub - maximum high sb disk supported by nsf / ast-9970780 | 0.185204 | 1,184 | 622 | , Saving $0.0 in GPT-4. |
20,977 | neural networks have proved to be versatile and robust for particle separation in many experiments related to particle astrophysics .
we apply these techniques to separate gamma rays from hadrons for the magic erenkov telescope .
two types of neural network architectures have been used for the classification task : one is the multilayer perceptron ( mlp ) based on supervised learning , and the other is the self - organising tree algorithm ( sota ) , which is based on unsupervised learning .
we propose a new architecture by combining these two neural networks types to yield better and faster classification results for our classification problem . | many gamma ray experiments have to deal with the problem of separating gammas from hadrons . the experiments usually generate large data sets with many attributes in them .
this multi - dimensional data classification problem offers a daunting challenge of extracting small number of interesting events ( gammas ) from an overwhelming sea of background ( hadrons ) .
many techniques are in active research for addressing this problem .
the list includes classical statistical multivariate techniques to more sophisticated techniques like neural networks , classification trees and kernel functions .
the class of neural networks provides an automated technique for the classification of the data set into given number of classes @xcite .
it is in active research in both artificial intelligence and machine learning communities .
several neural network models have been developed to address the classification problem .
usually , one makes the distinction between supervised and unsupervised classifiers : the former are trained with data for which the classification is known and then used to classify raw data , while the latter attempt to find the best - fitting class structure in the input data by using some measure of merit ( usually an euclidean metric is used @xcite ) . from a mathematical perspective ,
a neural network is simply a mapping from @xmath0 , where @xmath1 is the input data set dimension and @xmath2 is the output dimension of the neural network .
the network is typically divided into various layers ; each layer has a set of neurons also called nodes or information units , connected together by the links .
the artificial neural networks are able to classify data by learning how to discriminate patterns in features ( or parameters ) associated with the data .
the neural network learns from the data set when each data vector from the input set is subjected to it .
the learning or information gain is stored in the links associated with the neurons .
the output structure of the network is dependent on both the problem and the network type . for a gamma / hadron separation problem
the network maps each input vector onto the [ 0,1 ] interval in supervised networks , whereas in unsupervised networks the nodes are adapted to the input vector in such a way that the output of the network represents the natural groups that exist in the data set . the output of the unsupervised network is generally stored in an ascii file .
a visualization technique is then used to view the groups by processing the output file generated by the network .
section 2 describes the data sets used for the classification .
section 3 deals with the multilayer perceptron network and its classification results .
section 4 deals with the self - organizing tree algorithm and its variant along with their classification results .
conclusions and future perspectives are discussed in the section 5 .
the data sets are generated by a montecarlo simulation program , corsika @xcite .
they contain 12332 gammas , 7356 on events ( mixture of gammas and hadrons ) , and 6688 hadron or off events .
these events are stored in different files .
the files contain event parameters in ascii format , each line of 10 numbers being one event @xcite with the parameters defined below . 1
. _ length _ : major axis of ellipse [ mm ] 2 .
_ width _ : minor axis of ellipse [ mm ] 3 . _
size _ : 10-log of sum of content of all pixels 4 .
_ conc _ : ratio of sum of two highest pixels over fsize [ ratio ] 5 . _
conc1 _ : ratio of highest pixel over fsize [ ratio ] 6 .
_ asym _ : distance from highest pixel to centre , projected onto major axis [ mm ] 7 . _
m3long _ : 3rd root of third moment along major axis [ mm ] 8 . _
m3trans _ : 3rd root of third moment along minor axis [ mm ] 9 .
_ alpha _ : angle of major axis with vector to origin [ deg ] 10 .
_ dist _ : distance from origin to centre of ellipse [ mm ] these hillas image parameters @xcite are derived from pixel analysis and are used for classification .
for this approach we used the root analysis package ( v4.00/02 ) and in particular the multilayer perceptron class @xcite which implements a generic layered network .
since this is a supervised network we took two thirds of gamma and off data to train the network and the remaining data to test it.the code of the root package is very flexible and simple to use .
it allowed us to create a network with a 10 nodes input layer , a hidden layer with the same number of nodes and an output layer with just a single neuron which should return `` 0 '' if the data represent hadrons or `` 1 '' if they are gammas .
weights are put randomly at the beginning of the training session and then adjusted from the following runs in order to minimize errors ( back - propagation ) .
errors at cycle @xmath3 are defined as : @xmath4 where @xmath5 is the error of the output node .
data to input and output nodes are transferred linearly , while for hidden layers they use a sigmoid ( usually : @xmath6 ) .
we have tested the same network using different learning methods proposed by the code authors , as for example the so called `` stochastic minimization '' , based on the robbins - monro stochastic approximation , but the default `` broyden , fletcher , goldfarb , shanno '' ( bfgs ) method has proved to be the quickest and with the best error approximation .
figures [ mlp].a and [ mlp].b represent a possible output when using the root package on those data .
the first one depicts the error function for each run of the network , comparing the training and the test data .
note that the greater is the number of runs , the better the network behaves .
the second one shows the distributions of output nodes , that is how many times the network decides to give a value near to `` 0 '' or to `` 1 '' .
the self - organizing tree algorithm @xcite is an unsupervised neural network which implements a growing hierarchical clustering and is based on the self organising map network @xcite .
it hierarchically clusters the data into a binary tree of natural groups that exist in the data set .
initially the tree consists of one root node linked to 2 child cells .
all the input events are randomly distributed between the 2 initial child nodes .
the tree grows by expanding the child node having the most heterogeneous population of associated inputs .
two new descendants are generated from this heterogeneous cell that changes its state from cell to node .
the tree then grows by descending the cells into child nodes until each cell has one single input sequence , producing a complete classification of the sequences .
classification results.,width=302 ] alternatively , the expansion can be stopped at the desired level of heterogeneity in the cells , producing in this way a classification of sequences at a higher taxonomic level .
this kind of classification could be useful for astrophysics when a multi - event separation is needed on the same dataset , that is when multiple particles have been detected simultaneously and the analysis software should assign them a label ( as `` proton '' , `` muon '' , `` gamma '' , etc . ) .
they are also used as a data mining tool to explore the natural groups that exists in data sets .
using the magic datasets the tree has grown up to 10 levels , with the training sets taken from the montecarlo simulations ( figure [ sotares ] ) .
this approach gives a hierarchical view of data , is robust for noisy data and is faster than traditional hierarchical clustering .
flow for a sota - mlp network using montecarlo datasets.,width=340 ] figure [ schema1 ] shows the combination of sota with mlp for the separation task .
the sota method is applied to the initial montecarlo datasets ( gamma , on and off ) to find the natural clusters that exist in the datasets .
the sota tree produced two clusters of gamma and hadron which are used to train the mlp .
preliminary result using an mlp feeded with sota labelled datasets.,width=302 ] the sota cluster emulates the data distribution of the patterns , thus reducing the number of events in the training set .
the use of these clustered data can result in fast training for the mlp .
the trained mlp network is then used to perform testing through the on dataset and producing hadron probability for each event .
the preliminary results for this approach are shown on figure [ sotamlp ] where we can notice a better separation in the histograms respect to the non - treated mlp results ( figure [ mlp].b ) .
in this article we classified the gamma ray data using mlp and sota .
both mlp and sota shown some good classification results .
the algorithms used here suggest that a complex problem could not be solved using standalone methods even if they are suitable for a large part of other data analysis problems .
sota algorithm clusters the data set into groups thus reducing the number of events in the training set
. this can be useful for the magic experiment where there are overwhelming events to be classified .
mlp based on supervised technique identifies the group labels , but the training session could be longer . by combining sota with mlp we can significantly decrease the training period and yield better classification results .
the work can be further extended by using combination of different models in both self - organizing networks and supervised networks .
future experiments can be done using growing cell structures and growing neural gas models @xcite in the unsupervised category . in
supervised networks , tests can be performed by using probabilistic networks and mlp trained with fast back - propagation .
99 a.m. hillas , proc .
19th icrc , la jolla 3 ( 1985 ) 445 d. heck , et al . ,
corsika , _ a monte carlo code to simulate extensive air showers _ , forschungszentrum karlsruhe , report fzka 6019 ( 1998 ) s. lawrence et al .
, _ neural network classification and prior class probabilities _ , lecture notes in computer science state - of - the - art surveys , edited by g. orr et al . , springer verlag , pp .
299 - 314 ( 1998 ) c. delaere , _ multilayer perceptron root class _ , http://root.cern.ch r.k .
bock et al . , _ methods for multidimensional event classification : a case study using images from a cherenkov gamma - ray telescope _ , nuclear instruments and methods in physics research , a 516 ( 2003 ) rajaniemi et al .
, _ classifying gamma - ray bursts using self - organizing maps _
t. kohonen _ self - organizing maps _ , 2nd ed .
, springer ( 1997 ) j. dopazo , j.m .
carazo _ phylogenetic reconstruction using an unsupervised growing neural network that adopts the topology of a phylogenetic tree _
, j. mol .
( 1997 ) 44:226 - 233 b. fritzke , _ a growing neural gas network learns topologies _ , in _ advances in neural information processing systems 7 _ , g. tesauro , d.s .
touretzky and t. k. leen , editors , mit press ( 1995 ) | gamma ray experiments separating gammas from hadrons experiments generate large data sets with many attributes multi - dimensional data classification problem extracting interesting events ( gammas from background ( hadrons techniques in active research for list includes classical statistical multivariate techniques to neural networks classification trees kernel functions class neural networks provides automated technique for classification data into classes in active research in artificial intelligence machine learning communities neural network models developed to address classification problem between supervised and unsupervised classifiers former trained with data classification known classify raw data latter attempt find best - fitting class structure in input data by using merit euclidean metric used neural network is mapping from @xmath0 @xmath1 is input data set @xmath2 output dimension network divided into layers each layer has neurons nodes information units connected by links artificial neural networks classify data by patterns in features data neural network learns from data set learning stored in links neurons output structure network on problem and network typegamma / hadron separation problem network maps input vector 0,1 ] interval in supervised networks unsupervised networks nodes to input vector output represents natural groups data set output unsupervised network stored in ascii file visualization technique groups output section 2 describes data sets for classification section 3 multilayer perceptron network classification results 4 self - organizing tree algorithm variant classification results conclusions future perspectives section 5 data sets generated by montecarlo simulation program corsika @xcite contain 12332 gammas 7356 on events 6688 hadron or off events events stored in different files files contain event parameters in ascii format each line of 10 numbers one event @xcite parameters defined length major axis of ellipse width minor axis ellipse size 10-log of sum content pixels conc ratio highest pixels over fsize conc1 ratio highest pixel over fsize asym distance from highest pixel to centre major axis m3long 3rd root of third moment along major axis m3trans 3rd root third moment along minor axis alpha angle of major axis with vector to origindist distance from origin to centre ellipse [ mm ] hillas image parameters @xcite derived from pixel analysis used for classification used root analysis package ( v4.00/02 ) multilayer perceptron class @xcite generic layered network supervised network took two thirds of gamma off data to train remaining test code root package flexible simple network with 10 nodes input layer hidden layer same output layer single neuron `` 0 '' if data hadrons or `` 1 '' if gammas weights put randomly training session adjusted following runs minimize errors back - propagation errors at cycle @xmath3 defined as @xmath4 @xmath5 error output node data to input output nodes transferred linearly hidden layers sigmoid @xmath6 ) tested network using different learning methods stochastic minimization default `` broyden , fletcher goldfarb shanno '' ) method quickest best error approximation figures.a.b represent possible output root package first error function for each run comparing training test data greater number of runs better network behaves second shows distributions of output nodes times network value near to `` 0 '' or `` 1 ''self - organizing tree algorithm @xcite is unsupervised neural network implements growing hierarchical clustering based on self organising map network @xcite hierarchically clusters data into binary tree of natural groups initially tree one root node to 2 child cells input events randomly distributed between 2 child nodes tree grows child node heterogeneous two new descendants generated from heterogeneous cell changes state from cell to node tree grows cells into child nodes until each cell one single input sequence complete classification sequences expansion can stopped at desired level heterogeneity classification at higher taxonomic level classification useful for astrophysics when multi - event separation multiple particles used as data mining tool to explore natural groups in using magic datasets tree grown to 10 levels training sets from montecarlo simulations approach gives hierarchical view of data robust for noisy data faster than traditional hierarchical clustering sota - mlp network using montecarlo datasets shows combination of sota with mlp for separation task sota method applied to initial montecarlo datasets to find natural clusterssota tree produced two clusters gamma hadron train mlp preliminary result using mlp feeded with sota labelled datasets sota cluster emulates data distribution reducing events training set clustered data fast training for mlp trained mlp network testing hadron probability each event preliminary results shown on figure better separation in histograms non - treated mlp results classified gamma ray data using mlp sota sota good classification results suggest complex problem solved standalone methods sota algorithm clusters data into groups reducing events training set useful for magic experiment overwhelming events mlp identifies group labels training session longer combining sota with mlp training period better classification results work extended models in self - organizing networks supervised networks future experiments using growing cell structures neural gas models unsupervised supervised networks tests probabilistic networks mlp trained with fast back - propagation a. hillas 19th icrc la jolla 3 ( 1985 ) 445 d. heck corsika monte carlo code simulate extensive air showers forschungszentrum karlsruhe report fzka 6019 ( 1998 ) lawrence et alneural network classification prior class probabilities lecture notes computer science surveys orr springer verlag pp 299 - 314 1998 delaere multilayer perceptron root class http://root.cern.ch bock multidimensional event classification study cherenkov gamma - ray telescope nuclear instruments methods research 516 2003 ) rajaniemi classifying gamma - ray bursts self - organizing maps 2nd ed springer 1997 ) dopazo carazo reconstruction unsupervised growing neural network topology phylogenetic tree j mol 1997 ) 44:226 - 233 fritzke growing neural gas network learns topologies advances neural information processing systems tesauro touretzky leen editors mit press 1995 | 0.356984 | 2,431 | 1,179 | , Saving $0.1 in GPT-4. |
21,000 | wimp - nucleon cross sections @xmath0 pb may be probed by ton - scale experiments with low thresholds and background rates @xmath1 events per year .
an array of cryogenic detectors ( `` cryoarray '' ) could perform well enough to reach this goal .
sufficient discrimination and background suppression of photons has already been demonstrated .
reduction of neutron backgrounds may be achieved by siting the experiment deep enough .
removal of the surface - electron backgrounds alone has not yet been demonstrated , but the reductions required even for this troublesome background are quite modest and appear achieveable . | direct detection of supersymmetric wimp dark matter in the coming decade appears possible . as shown in fig .
[ limitplot ] , experiments under construction should probe a large fraction of parameter space allowed by minimal supersymmetric theory and experimental constraints @xcite .
interestingly , if the sign of @xmath2 , the higgs mixing parameter in the superpotential , is positive , the wimp mass must be @xmath3 , and the wimp - nucleon cross section must be @xmath4 pb @xcite , possibly making the entire parameter space accessible to a ton - scale wimp - detection experiment .
cryoarray would be a 1-ton deployment of semiconductor cryogenic detectors , of a type similar to that in use in cdms ii @xcite .
this experiment would be sensitive to a wimp - nucleon cross section @xmath5 pb , corresponding to a signal of a few wimp interactions per 100 kg years .
the detectors of cdms or cryoarray measure phonons and charge carriers separately for each interaction in order to allow rejection of the otherwise dominant electron - recoil background events . the background discrimination of these detectors has already been demonstrated to be so good @xcite that instrumenting a one - ton detector is fully justified .
below , we review in detail the expected performance of cryoarray , and conclude with a short discussion of the challenge to build such an experiment at a reasonable cost .
[ cols= " < , < , > , > , < , > , > , > , > , < " , ] [ backgrounds ]
table [ backgrounds ] lists photon , electron , and neutron background event rates in the energy range 1545 kev for actual exposures of cdms i @xcite and cdms ii at stanford @xcite , and for projections for cdms ii at soudan and for cryoarray .
the goals of cryoarray reflect a factor 20 improvement in backgrounds , and a factor 10 improvement in electron discrimination , over the goals for cdms ii .
the low - energy threshold is at 15 kev ( rather than 5 or 10 kev ) to avoid the contributions of the cosmogenic activation peaks and tritium beta spectrum in ge near 10 kev .
the first columns of the table show the radioactivity event rates and the total numbers of events expected during the tabulated exposures .
the other columns show how event - by - event discrimination can be used not only to reject most of the background , but also to subtract part of the residual background if the detector response is well enough understood @xcite .
the results of this background subtraction include the effects of a 5% systematic uncertainty in the detector discrimination .
a muon veto provides the neutron rejection ( see section [ neutrons ] ) .
however , the neutron subtraction is performed based on the observed population of multiple - detector neutron events . as shown in fig .
[ discrimination ] , the original cdms ii and cryoarray goals @xcite for rejection of photons have already been exceeded . based on the demonstrated rejection
, the original photon - background goal for cdms ii ( 15 events after subtraction ) has already been met , even without the expected 3@xmath6 reduction in rate from increasing the experiment s depth .
furthermore , the rejection efficiencies in table [ backgrounds ] are statistically limited lower limits . and are probably overly conservative . based on the demonstrated rejection , cryoarray s post - subtraction background goal for photons is achievable merely by reducing the raw photon rate to 13 mdru , only @xmath7 lower than that currently achieved by igex @xcite in hpge detectors in the same energy range .
this modest improvement is expected due to the increased detector self - shielding and multiple - scatter rejection from increased mass and many detectors .
the beta electron background represents a greater challenge .
the current observed background rate for cdms ii at suf is 80 mdru , @xmath8 above the cdms ii soudan goal of 20 mdru .
some of the current beta contamination may be due to exposure of the detectors to a leaking calibration source during testing .
use of such sources has been restricted . moving to the deeper site at soudan
will remove essentially all cosmogenically produced betas , potentially the dominant beta background .
furthermore , the discrimination performance of the new detectors ( see fig . [ discrimination]b ) may well be sufficient even if cdms ii does not reach the absolute beta background goal . for cryoarray ,
self - shielding will eliminate the beta backgrounds from materials surrounding the detector stacks .
nevertheless , improving the beta background by a further factor of 20 will require sensitive screening of the detectors and nearby materials .
while screening to the required level of @xmath9 counts kev@xmath10 m@xmath11 day@xmath10 for betas with @xmath12 kev appears achievable @xcite , cleaning the materials may require new techniques .
improving the electron discrimination to 99.5% appears achievable through improvements of the signal - to - noise of the detectors .
% efficient , limited by statistics .
( b ) applying a cut based on phonon risetimes can remove @xmath13% of the low - ionization - yield events ( mostly electrons ) while keeping @xmath14% of the neutron events . ]
the low - energy ( @xmath1510 mev ) neutrons from @xmath16,n ) and fission processes in the rock are trivially stopped by hydrogenous shielding .
such shielding reduces the low - energy neutron flux by @xmath1710@xmath6 per 10 cm of material , so 70 cm of such shielding should make this neutron flux negligible .
other neutrons arise from muons interacting inside the experiment s shielding or in the cavern rock .
the former can be tagged using a local active muon veto directly around the passive shielding .
a muon veto efficiency of 80% is adequate at soudan ( 2080 mwe ) to ensure that these `` shield '' neutrons contribute negligibly to the cdms ii background .
the demonstrated cdms ii muon veto efficiency of 99.95% would even be sufficient to limit this neutron background to a small enough rate at soudan for cryoarray .
a deeper site such as gran sasso ( at @xmath174 kmwe ) would reduce the neutron flux by @xmath1725@xmath6 , making the muon - veto requirement very modest .
the high - energy neutrons generated in the cavern rock by spallation processes of high - energy muons are difficult both to veto and to stop . with the cdms ii polyethylene shielding sandwiching pb shielding , @xmath178 secondary neutrons from these `` punch - through '' neutrons should scatter in single detectors during the 2500 kg - day exposure , based on extensive monte carlo studies @xcite . because this neutron flux for deep sites is approximately proportional to the muon flux @xcite , increased depth is the easiest way to reduce this neutron background .
siting cryoarray at a depth of 4 kmwe would likely reduce this background sufficiently , while siting it at 7 kmwe should essentially remove this background .
if the experiment were sited at @xmath18 mwe ( e.g. wipp or soudan ) , either a much thicker liquid - scintillator buffer would be required around the detector in order to tag high - energy neutrons , or the cavern rock itself ( or an outer heavy shield ) would need to be instrumented with additional veto detectors in order to catch some part of the shower associated with the muon that generated the neutron .
cryoarray should work well enough to probe wimp - nucleon cross sections @xmath0 pb once it is built . the most significant challenge at this point is determining how to build it at a reasonable cost .
for cdms ii , the 42 detectors are being built using university facilities and technicians under the close supervision of cdms physicists .
detectors are tested at least twice in batches of @xmath19 in university dilution refrigerators .
this time - consuming and physicist - intensive procedure is not practical for cryoarray s @xmath17 5002000 detectors .
an industrial approach is needed . in
order to `` out - source '' this effort , we will need to engineer and specify a stable and reliable high - yield process .
since it is not feasible to test each individual module prior to deployment in the array , the yield must be 90% .
this goal is ambitious ; current detector yields are @xmath1750% .
we expect to continue to make gains in understanding pathologies in the processing steps during the remaining cdms ii construction period , which is scheduled for completion at the end of 2003 . in 12 years
we will have substantially more experience in fabricating and operating cdms detectors .
the knowledge gained will be essential for establishing a realistic plan for building cryoarray and assessing its likely performance
. this knowledge will likely point the way to additional laboratory work to be carried out before a full - scale proposal is considered .
we thank our cdms collaborators for innumerable useful discussions . this work is supported by the university of california , berkeley , under cooperative agreement no .
ast-99 - 78911 , and by the national science foundation under grant no .
some of this work was done at the aspen center for physics .
r. abusaidi et al .
84 ( 2000 ) 5699 ; d. abrams et al .
, exclusion limits on the wimp - nucleon cross - section from the cryogenic dark matter search , submitted to phys .
d ( 2002 ) , astro - ph/0203500 ; r. w. schnee et al . , these proceedings .
r. j. gaitskell , in : n. j. c. spooner , v. kudryavtsev ( eds . ) , proceedings of the third international workshop on the identification of dark matter , world scientific , london , 2001 , pp .
606617 , astro - ph/0106200 . | detection of supersymmetric wimp dark matter decade possible in fig . experiments construction probe fraction parameter space by supersymmetric theory experimental constraints if sign of @xmath2 higgs mixing parameter positive wimp mass @xmath3 wimp - nucleon cross section @xmath4 pb entire parameter space accessible to ton - scale wimp - detection experiment cryoarray 1-ton deployment semiconductor cryogenic detectors similar cdms ii @xcite sensitive to wimp - nucleon cross section @xmath5 pb few wimp interactions per 100 kg years detectors measure phonons charge carriers separately rejection electron - recoil background events background discrimination good instrumenting one - ton detector justified review expected performance of cryoarray challenge to build experiment at reasonable cost backgrounds table lists photon electron neutron background event rates energy range 1545 kev for exposures cdms i @xcite cdms ii stanford @xcite projections cdms ii soudan cryoarray goals cryoarray reflect factor 20 improvement in backgrounds factor 10 improvement in electron discrimination over goals cdms iilow - energy threshold 15 kev avoid cosmogenic activation peaks tritium beta spectrum near 10 kev first columns table show radioactivity event rates total events expected exposures other columns show event - by - event discrimination reject background subtract residual background if detector response understood results background subtraction include 5% systematic uncertainty detector discrimination muon veto provides neutron rejection neutron subtraction based population multiple - detector neutron events original cdms ii cryoarray goals for rejection photons exceeded original photon - background goal for cdms ii 15 events after subtraction met without expected 3 reduction increasing experiment depth rejection efficiencies table backgrounds statistically limited overly conservative cryoarray s post - subtraction background goal for photons achievable reducing raw photon rate to 13 mdru lower than igex @xcite detectors improvement due to increased detector self - shielding multiple - scatter rejection increased mass detectors beta electron background greater challenge current observed background rate for cdms ii at suf 80 mdru above cdms ii goal 20 mdrucurrent beta contamination due to exposure to leaking calibration source use restricted moving deeper site at soudan all cosmogenically betas potentially dominant beta background discrimination performance new detectors may be sufficient if cdms ii reach absolute beta background goal for cryoarray self - shielding beta backgrounds from detector improving beta background by factor 20 sensitive screening detectors nearby materials screening to level for achievable cleaning materials may require new techniques improving electron discrimination to 99.5% achievable through improvements signal - to - noise detectors limited by statistics applying cut based on phonon risetimes low - ionization - yield events electrons neutron events low - energy neutrons from fission rock stopped by hydrogenous shielding reduces low - energy neutron flux by per 10 cm material 70 cm shielding neutron flux negligible other neutrons arise from muons interacting inside shielding or in cavern rock active muon veto shielding muon veto efficiency of 80% adequate at soudan neutrons contribute negligibly to cdms ii backgroundcdms ii muon veto efficiency 99.95% limit neutron background at soudan for cryoarray deeper site gran sasso at @xmath174 kmwe neutron flux by @xmath1725@xmath6 muon - veto requirement modest high - energy neutrons in cavern rock by spallation difficult to veto stop with cdms ii polyethylene shielding pb shielding secondary neutrons scatter in single detectors during 2500 kg - day exposure monte carlo studies neutron flux for deep sites proportional to muon flux @xcite increased depth reduce neutron background cryoarray at 4 kmwe background 7 kmwe remove background if experiment at @xmath18 mwe thicker liquid - scintillator buffer required high - energy neutrons or cavern rock heavy shield with additional veto detectors catch shower muon cryoarray should probe wimp - nucleon cross sections once built challenge determining build at reasonable cost for cdms ii 42 detectors built using university facilities technicians under cdms physicists detectors tested twice in in university dilution refrigeratorstime - consuming physicist - intensive procedure not practical for cryoarray s @xmath17 5002000 detectors industrial approach needed need engineer stable reliable high - yield process test each module deployment yield must be 90% goal ambitious current detector yields are @xmath1750% expect gains in understanding pathologies processing steps during cdms ii construction period completion end of 2003 in 12 years more experience fabricating operating cdms detectors knowledge essential for plan for building cryoarray assessing performance knowledge additional laboratory work before full - scale proposal thank cdms collaborators for discussions work supported by university of california national science foundation work done at aspen center for physics abusaidi et al 84 ( 2000 ) 5699 abrams exclusion limits on wimp - nucleon cross - section from cryogenic dark matter search ( 2002 ) astro - ph/0203500 schnee gaitskell spooner kudryavtsev proceedings third international workshop on identification dark matter world scientific london 2001 606617 astro - ph/0106200 | 0.318497 | 2,198 | 1,032 | , Saving $0.1 in GPT-4. |
21,003 | we present the constraints on the parameter space of the lee - wick standard model coming from electroweak precision observables .
the model predicts a large positive @xmath0 and a negative @xmath1 .
we show that it is possible to find some regions in parameter space with a fermionic state as light as 2.4 - 3.5 tev .
we also propose a simple extension of the model including a fourth generation . in this case
it is possible to pass the electroweak constraints with lee - wick fermionic masses of order 0.4 - 1.5 tev and lee - wick gauge masses of order 3 tev . | the standard model ( sm ) describes the electroweak ( ew ) interactions with an incredible precision .
however , the instability of the higgs potential under radiative corrections signals our ignorance over the real mechanism of electroweak symmetry breaking ( ewsb ) and has lead to many extensions beyond the sm that attempt to solve the hierarchy problem .
recently , grinstein , oconnell and wise proposed a new extension of the sm @xcite , based on the ideas of lee and wick @xcite for a finite theory of quantum electrodynamics .
the building block of the lee - wick proposal is to consider that the pauli - villars regulator describes a physical degree of freedom . in the lee - wick standard model ( lwsm )
, this idea is extended to all the sm in such a way that the theory is free from quadratic divergences and the hierarchy problem is solved .
the basic idea can be illustrated discussing the self energy of a scalar field , as shown in fig.[example1 ] .
( 500,80)(0,0 ) ( 140,25)3 ( 170,75)[c]@xmath2 ( 100,35)[c]@xmath2 ( 180,35)[c]@xmath2 ( 90,25)(140,25)3 ( 140,25)(190,25)3 ( 140,49)(24,0,360)3 ( 240,25)[c]@xmath3 ( 340,25)3 ( 370,75)[c]@xmath4 ( 300,35)[c]@xmath2 ( 380,35)[c]@xmath2 ( 290,25)(340,25)3 ( 340,25)(390,25)3 ( 340,49)(24,0,360)3 the propagators for the scalar field @xmath2 and its lw partner @xmath4 differ by an important overall sign @xmath5 an explicit calculation shows that the quadratic divergence cancels and only a mild logarithmic divergence remains @xmath6 it is possible to understand the origin of the lw partner @xmath4 starting from a higher derivative version of the scalar field lagrangian , as discussed in ref .
@xcite .
similarly , in the lwsm every sm field has a lw - partner with an associated lw - mass .
these masses are the only new parameters in the minimal lwsm .
a potential problem in this model is that the lw - states violate causality at the microscopic level due to the opposite sign of their propagators .
however , the authors of ref .
@xcite argued that there is no causality violation on a macroscopic scale provided that the lw - particles are heavy and decay to sm - particles .
the lwsm can also be thought as an effective theory coming from a higher derivative theory .
we refer the reader to refs .
@xcite for more details and only quote here some specific interaction terms that are useful to understand the contributions to the ew observables @xcite .
the gauge - fermion interactions are : @xmath7 + \sum_{\psi = q , u , d}\left[g_1\bar{\tilde\psi}({\not \!\!b}+{\not \!\!{\tilde b}})\tilde\psi + g_2\bar{\tilde\psi}({\not \!\!w}+{\not \!\!{\tilde w}})\tilde\psi\right].\end{aligned}\ ] ] note that the lw - fermions couple to the gauge fields with the opposite sign compared with the sm - fermions .
different to the sm chiral fermions , the lw - fermions combine into dirac spinors of masses @xmath8 .
we will assume that the lw - fermions transforming in the same representation of the gauge symmetries have the same mass , which is compatible with the minimal flavor violation principle .
in the lwsm the mixings between the gauge bosons and their lw partners induce non - canonical couplings with the sm fermions , as noticed in ref .
a shift in the gauge fermion couplings can be reabsorbed into the oblique parameters . therefore , to correctly define the oblique parameters @xmath9 it is necessary to properly normalize the couplings between the fermions and the gauge bosons .
this observation was overlooked in ref .
we find it useful to work in the effective theory obtained after integrating out the heavy lw fields at tree level .
the canonical couplings are obtained after a field redefinition : @xmath10 where @xmath11 are the usual currents of sm fermions and @xmath12 and @xmath13 are identified as the sm fields .
the effective lagrangian written in terms of the sm fields @xmath14 together with the usual definitions of the oblique parameters @xmath0 and @xmath1 @xmath15\ ; , \ ] ] lead to tree level contributions to @xmath0 and @xmath1 : @xmath16 where @xmath17 is the vacuum expectation value of the higgs field and @xmath18 are the mass parameters of the lw fields @xmath19 , respectively . notice that the sign difference between @xmath1 in eq .
( [ ttree ] ) and the result of ref .
@xcite is due to the additional contribution coming from the redefinition of the gauge fields mentioned above .
we can see that for @xmath20 the tree level @xmath1 parameter vanishes as expected , since in this limit we recover a custodial symmetry in the lw - gauge sector .
the contributions from the higgs sector come from the diagrams shown in fig.[figlwhiggs ] and are found to be numerically small both for @xmath0 and @xmath1 , when compared to their tree level values .
the fermionic contributions shown in fig.[figlwfermions ] are more important .
the isospin splitting of the third generation gives a large negative contribution to the @xmath1 parameter that changes the sign of the tree level result .
the resulting @xmath1 parameter is negative . on the other hand
the fermionic contributions to @xmath0 are negative but small compared with the tree level value of @xmath0 .
the result of scanning the parameter space is shown in fig.[planest ] .
a careful examination of the numerical values spanned by the dots that fall into the ellipse allows to find bounds on the lw masses , see ref .
@xcite for details . + ( 300,60)(0,0 ) ( -45,25)3 ( 5,25)3 ( -20,60)[c]@xmath21 ( -20,-10)[c]@xmath21 ( -70,25)(-45,25)33 ( 5,25)(30,25)33 ( -20,25)(25,0,180)4 ( -20,25)(25,180,360)4 ( -65,-10)[c](a ) ( 90,25)3 ( 140,25)3 ( 115,60)[c]@xmath22 ( 115,-10)[c]@xmath23 ( 65,25)(90,25)33 ( 140,25)(165,25)33 ( 115,25)(25,0,180)4 ( 115,25)(25,180,360)36 ( 90,25)(80,50)4 ( 140,25)(150,50)4 ( 70,-10)[c](b ) ( 240,25)3 ( 240,60)[c]@xmath21 ( 215,25)(240,25)33 ( 240,25)(265,25)33 ( 240,38)(12,0,360)4 ( 240,-10)[c](c ) + ( 300,80)(0,0 ) ( -45,45)3 ( 5,45)3 ( -20,80)[c]@xmath24 ( -20,10)[c]@xmath25 ( -70,45)(-45,45)33 ( 5,45)(30,45)33 ( -20,45)(25,0,180 ) ( -20,45)(25,180,360 ) ( -65,0)[c](a ) ( 270,45)3 ( 320,45)3 ( 276,61)3 ( 313,61)3 ( 276,29)3 ( 313,29)3 ( 295,80)[c]@xmath26 ( 295,10)[c]@xmath26 ( 282,45)[c]@xmath27 ( 308,45)[c]@xmath27 ( 245,45)(270,45)33 ( 320,45)(345,45)33 ( 295,45)(25,0,180 ) ( 295,45)(25,180,360 ) ( 278,60)(259,80)4 ( 312,60)(331,80)4 ( 276,30)(259,10)4 ( 312,30)(331,10)4 ( 250,0)[c](c ) ( 120,45)3 ( 170,45)3 ( 145,70)3 ( 145,20)3 ( 95,45)(120,45)33 ( 170,45)(195,45)33 ( 145,45)(25,0,90 ) ( 145,45)(25,90,180 ) ( 145,45)(25,180,270 ) ( 145,45)(25,270,360 ) ( 145,70)(145,90)4 ( 145,20)(145,0)4 ( 130,75)[c]@xmath27 ( 130,15)[c]@xmath27 ( 165,75)[c]@xmath26 ( 165,15)[c]@xmath26 ( 100,0)[c](b ) @xmath28{fig3a.eps } \put(-7,30){\rotatebox{90}{$t$ } } \put(50,-5){$s$ } \end{overpic } & \ \ \ \ \ & \begin{overpic}[width=7.5cm]{fig3b.eps } \put(-7,30){\rotatebox{90}{$t$ } } \put(50,-5){$s$ } \end{overpic } \end{array}$ ] @xmath29{st4ta.eps } \put(-7,50){\rotatebox{90}{$t$ } } \put(50,-5){$s$ } \end{overpic } & \ \ \ \ \ & \begin{overpic}[width=5.5cm]{m4ta.eps } \put(-10,40){\rotatebox{90}{$m_d^4 $ [ tev ] } } \put(50,-7){$m_u^4 $ [ tev ] } \end{overpic } \end{array}$ ] the figures show that a scan of the parameter space gives @xmath0 and @xmath1 values in the lower right quadrant , which is the less favorable one .
this translates into higher bounds for the lw masses . a positive contribution to @xmath1
would bring the dots to the more favorable upper right quadrant and therefore lower the bounds on the lw masses .
this can be achieved by extending the lwsm adding a fourth generation .
the contribution to @xmath1 is proportional to the isospin breaking of the new generation , which allows for more freedom , in contrast to the third generation , where the isospin breaking is fixed by the experimentally known masses .
the new contribution to @xmath0 remains small and the results are shown in fig.[fig4ta ] . for a more thorough discussion
we refer the reader to ref .
we discussed the electroweak precision constraints for the lwsm @xcite .
we find sizeable tree level contributions for @xmath0 and @xmath1 . after including the one loop contributions and scanning the parameter space we see that the lwsm is mostly located in the @xmath30 quadrant .
a careful analysis of the ( @xmath0,@xmath1 ) values compatible with experimental data leads to bounds that imply lw - fermion masses greater than 2.4 -3.5 tev and lw - gauge bosons masses greater than 5 - 8 tev .
extending the lwsm adding a 4th - generation allows to relax these bounds . in this case
we find that lw - fermion masses as small as 0.4 -1.5 tev and lw - gauge bosons masses as small as 3 tev are still compatible with electroweak precision data . | standard model describes electroweak interactions precision instability higgs potential under radiative corrections signals ignorance mechanism electroweak symmetry breaking to extensions beyond sm hierarchy problem grinstein oconnell wise proposed new extension sm @xcite based ideas lee wick for finite theory quantum electrodynamics lee - wick proposal pauli - villars regulator describes physical freedom lee - standard model idea extended to sm theory free from quadratic divergences hierarchy problem solved basic idea illustrated discussing self energy scalar field shown in fig.[example1 ( 500,80) 140,25)3 170,75)@xmath2 100,35) 180,35),25)(140(190,25)3(24,0,360 240,25) 340,25)3 370,75) 300,35) 380,35) 290,25)(340 propagators scalar field @xmath2 lw partner @xmath4 differ by sign @xmath5 shows quadratic divergence cancels mild logarithmic divergence remains @xmath6 understand origin of lw partner @xmath4 starting from higher derivative version scalar field lagrangian@xcite in lwsm every sm field has lw - partner associated lw - mass masses new parameters in minimal lwsm potential problem lw - states violate causality at microscopic level due to opposite sign of propagators authors of no causality violation on macroscopic scale lw - particles heavy decay to sm - particles lwsm effective theory from higher derivative theory refer reader to refs @xcite for details specific interaction terms ew observables gauge - fermion interactions are @xmath7_\psi = q u d lw - fermions couple to gauge fields with opposite sign sm - fermions lw - fermions combine into dirac spinors of masses lw - fermions transforming in same gauge symmetries have same mass compatible with minimal flavor violation principle in lwsm mixings between gauge bosons lw partners induce non - canonical couplings with sm fermions shift in gauge fermion couplings reabsorbed into oblique parametersdefine oblique parameters @xmath9 normalize couplings between fermions gauge bosons observation overlooked in ref useful work in effective theory after integrating heavy lw fields at tree level canonical couplings after field redefinition : @xmath10 currents sm fermions @xmath12 sm fields effective lagrangian sm fields @xmath14 definitions oblique parameters @xmath0 lead to tree level contributions to @xmath0 @xmath1 @xmath16 vacuum expectation value higgs field @xmath18 mass parameters of lw fields @xmath19 sign difference between @xmath1 ttree result ref due to additional contribution from redefinition gauge fields for @xmath20 tree level @xmath1 parameter vanishes custodial symmetry in lw - gauge sector contributions from higgs sector diagrams fig. numerically small for @xmath0 @xmath1 compared to tree level values fermionic contributions important isospin splitting third generation negative contribution to @xmath1 parameter changes sign tree level result parameter negative fermionic contributions to @xmath0 negative small compared with tree level value @xmath0result scanning parameter space shown in fig.[planest examination numerical values dots ellipse find bounds lw masses see ref @xcite details300,60) -45,25)3 5 -20,60)@xmath21 -20,-10 -70,25)(-45,0,180,180,360 -65,-10 90,25)3 140 115,60)@xmath22 115,-10@xmath23 65,25)(90 140,25) 115,25)(25,0,180)4(25,180,360 90,25)(80,50)4 140,25)(150,50)4 70,-10 240,25)3 240,60)@xmath21 215,25)(240,25)33(265 240,38)(12,0,360)4 240,-10) 300,80) -45,45)3 5,80)@xmath24,10)@xmath25,45)(-45 5 -20,45),0,180(25,180,360 -65,0) 270,45)3 320,45)3 276,61)3 313,61 276 313 295,80)@xmath26 295,10) 282,45) 308,45) 245,45)(270 320,45)(345 295,45)(25,0,180 295,180,360 278,60)(259,80)4 312,60)(331 276,30)(2594 ( 312,30)(331,10)4 ( 250,0) 120,45)3 ( 170,45)3 145,70)3 145,20)3 95,45)(120,45)33 ( 170,45)(195,45)33 ( 145,45)(25,0,90 145,180 145,45),270,270,360 ) 145,70)(145,90)4 ( 145,20)(145,0)4 130,75)@xmath27 ( 130,15) 165,75) 165,15) 100,0) @xmath28{fig3a.eps \put(-7,30)\rotatebox{90}$t$(50,-5){$s${overpic{overpic}[width=7.5cm](-7,30){90}$t(50,-5){$s @xmath29.eps(-7,50)\rotatebox{90}$t$(50,-5){$s${overpic{overpic}[width=5.5cm](-10,40)\rotatebox{90}_d^4(50,-7){overpic{array} scan parameter space gives @xmath0 @xmath1 values lower right quadrant less favorabletranslates higher bounds lw masses positive contribution @xmath1 dots favorable upper right quadrant bounds achieved extending lwsm adding fourth generation contribution @xmath1 proportional isospin breaking new generation more freedom contrast third generation fixed masses new contribution @xmath0 small results shown in fig. discussion refer ref discussed electroweak precision constraints lwsm @xcite sizeable tree level contributions @xmath0 @xmath1 one loop contributions lwsm mostly in @xmath30 quadrant analysis @xmath0,@xmath1 values leads bounds lw - fermion masses 2.4 -3.5 tev lw - gauge bosons 5 - 8 tev extending lwsm adding 4th - generation bounds lw - fermion masses 0.4 -1.5 tev lw - gauge bosons 3 tev compatible with electroweak precision data | 0.42362 | 2,779 | 1,440 | , Saving $0.1 in GPT-4. |
21,004 | excitation of a global alfvn wave ( gaw ) is proposed as a viable mechanism to explain plasma heating in the magnetosphere of young stars . the wave and basic plasma parameters are compatible with the requirement that the dissipation length of gaws be comparable to the distance between the shocked region at the star s surface and the truncation region in the accretion disk . a two - fluid magnetohydrodynamic plasma model is used in the analysis .
a current carrying filament along magnetic field lines acts as a waveguide for the gaw .
the current in the filament is driven by plasma waves along the magnetic field lines and/or by plasma crossing magnetic field lines in the truncated region of the disk of the accreting plasma .
the conversion of a small fraction of the kinetic energy into gaw energy is sufficient to heat the plasma filament to observed temperatures . | in the present popular model for classical t tauri stars ( cttss ) , the so called class ii objects @xcite , a central young star is surrounded by a geometrically thin accretion disk .
the disk is disrupted at a given radius by the magnetic field of the star . for smaller radii
the accretion flow follows the magnetic field lines of the star until it impacts onto the stellar surface ghosh & lamb ( 1979 ) ; knigl ( 1991 ) .
this magnetospheric accretion model explains observational signatures seen in some cttss as , for example , the excess of optical and ultraviolet continuum flux ( veiling ) , as well as redshift absorption features in emission line profiles ( muzerolle @xmath0 1998 ; hartmann , hewett & calvet 1994 ) .
the plasma of the accretion disk flows along the dipole magnetic field lines towards the magnetic poles of the star .
it is inhomogeneous in the plane perpendicular to the magnetic field lines , but although somewhat more homogeneous along the field lines , cavities and plasma condensations ( plasmoids ) , aligned by the magnetic field , may be present .
these plasmoids are distributed randomly in the plane perpendicular to the plasma flow .
waves excited by the shock of the plasma flux with the star s surface can propagate upstream and interact with the plasmoids , creating a current , which in turn , will create current - carrying filaments .
an upstream or downstream electric current can be driven by the waves or by the @xmath1-mhd effect in the truncated region of the disk .
such filaments have been observed in the laboratory ( see , e.g. fadeev @xmath0 ( 1965 ) ; rosenberg @xmath0 ( 2001 ) ) .
the occurrence of filaments has also been predicted in simulations of rotating magnetized disks ( machida @xmath0 2000 ) .
hartmann , hewett & calvet ( 1994 ) and , more recently , muzerrole @xmath0 ( 1998 ) carried out calculations on magnetosphere accretion models , solving the radiative transfer equations in the sobolev approximation .
theoretical balmer line profiles obtained were found to be in reasonable agreement with observations .
the major uncertainty in radiative magnetosphere model calculations is the temperature profile along the tube .
martin ( 1996 ) calculated the temperature profile from the energy balance of the gas , including heating by adiabatic compression of magnetic field lines , balmer photoionization , and ambipolar diffusion .
however , the values of the temperature that he obtained were too low to explain the observed spectra .
thus , additional heating mechanisms inside the tube are needed .
the line - averaged electron temperature in the magnetosphere taken from optical measurements , is about @xmath2 ev ( see , e.g. hartmann , hewett & calvet 1994 ; martin 1996 ; muzerrole @xmath0 1998 ) . balancing the heating rate with the black body emission at a temperature @xmath3 ev , we find that about 5@xmath4 of the flux energy is sufficient to heat the downstream flux to a few ev .
the main consumption of the flux energy is spent on plasma ionization . using laboratory experimental data for the ionization - recombination processes ( see delcroix 1965 ) , we can estimate the energy necessary for the plasma ionization rate to be about @xmath5 in the filament for electron temperatures of about @xmath6 ev .
we use here a theoretical temperature profile , varying from @xmath7 k to @xmath8 k , to reproduce the optical line features observed . taking into account the evidence for turbulence in the observations , as well as
the necessity for an additional heating mechanism , vasconcelos , jatenco - pereira & opher ( 2002 ) suggested that alfvn waves may be important in the heating of magnetic flux tubes of cttss .
they studied the possibility that the waves generated at the star s surface due to the shock are produced by the accreting matter , as suggested by scheurwater & kuijpers ( 1988 ) .
various damping mechanisms for alfvn waves have been suggested in the literature , such as alfvn resonant heating of solar loops ( hollweg 1990 ; elfimov @xmath0 1996 ) , wave damping by phase mixing ( also in the solar context ) , and cyclotron heating , occurring as an alfvn wave travels down a magnetic field gradient until its frequency matches that of the cyclotron resonance of helium or some other plasma specie due to the decreasing ion - cyclotron resonance ( magnetic beach ) .
in addition to the more conventional collisional and viscous - resistive alfvn wave damping ( osterbrock 1961 ) , vasconcelos @xmath0 ( 2002 ) concentrate on nonlinear and turbulent damping ( vasconcelos , jatenco - pererira & opher 2000 ) .
nonlinear and turbulent damping have also been studied by jatenco - pereira & opher ( 1989a ) ; jatenco - pereira , opher & yamamoto ( 1994 ) in the solar wind , jatenco - pereira & opher ( 1989b ) in protostellar winds , jatenco - pereira & opher ( 1989c ) in late - type giant stars , dos santos , jatenco - pereira & opher ( 1993a ; 1993b ) in wolf - rayet stars , gonalves , jatenco - pereira & opher ( 1993a ; 1996 ) in quasar clouds and gonalves @xmath0 ( 1993b ) in extragalactic jets .
the ( gaw ) has not been previously analyzed as a possible mechanism for the energy transport from the shock and truncation regions to the magnetosphere of t tauri stars .
they were predicted in numerical calculations ( ross @xmath0 1982 ) using a cylindrical model for magnetized plasmas with an axial current and were observed later in tokamak experiments .
gaws are excited in plasmas in a helical magnetic field , with the poloidal component produced by an axial current , @xmath9 , where @xmath10 is the axial magnetic field component and @xmath11 and @xmath12 are the plasma radius and length , respectively .
the so - called tokamak safety factor , @xmath13 , is considered to be unity . in cold plasmas with temperatures
less than @xmath14 0.1 ev , gaws dissipate via electron - ion collisions .
however , in hot plasmas , when the alfvn velocity , @xmath15 , is of the order of the electron thermal velocity , @xmath16 , gaws dissipate via electron landau damping .
the electron landau damping mechanism is very effective when waves interact with a group of electrons that have velocities approximately equal to the wave phase velocity , @xmath17 , where @xmath18 is the frequency and @xmath19 is the wavenumber parallel to the magnetic field lines .
electromagnetic field components related to gaws are proportional to @xmath20 $ ] , with poloidal @xmath21 and axial @xmath22 wave numbers .
the requirement for these waves to propagate is that @xmath23 and @xmath24 , in the case of current flowing along the magnetic field .
a gaw has a very large quality factor , ratio of frequency to dissipation decrement , @xmath25 and a very long dissipation length in comparison to their wave length , @xmath26 .
thus it is a very good candidate to explain plasma heating far from a source , such as in solar loops @xcite . in addition
a gaw is able to drive the current , which can be calculated by balancing the wave momentum transfer force with the electron - ion friction force ( elfimov @xmath0 1996 ) .
we study here a two - fluid magnetohydrodynamic model of a young star , accreting plasma ( see fig .
[ fig1 ] ) .
the standard parameters of cttss , similar to those used by vasconcelos @xmath0 ( 2002 ) , are assumed : star radius @xmath27 cm , truncated radius @xmath28 cm , truncated depth @xmath29 cm , magnetic field at the star s surface @xmath30 kg and @xmath31 kg in the truncated region , where the electron density is @xmath32 @xmath33 .
the length @xmath34 of the magnetic field lines from the shock at the star s surface to the truncated region is @xmath35 .
the velocity of the shock is @xmath36 cm s@xmath37 at the star s surface and @xmath38 cm s@xmath37 in the truncated region .
the ionization rate parameter @xmath39 for @xmath40 ev is @xmath41 @xmath42 s@xmath37 ( delcroix 1965 ) .
the accretion energy density is defined by the ratio of the energy flux density to the path length @xmath34 . balancing ninety percent of the accretion power density @xmath43 with the ionization power density @xmath44 , we obtain a plasma with an average density along the magnetic field of @xmath45 @xmath33 .
we assume that the magnetic field connecting the truncated disk and the star s surface consists of filaments , along which the gaw propagates .
the filament current can be driven by the wave ponderomotive force ( see discussion below ) and/or by a magnetodynamo in the truncated region of the disk .
the current , @xmath46}=5ab_{\theta , ef}$ ] , and current density , @xmath47 stata @xmath48 , in a filament of radius @xmath11 can be produced by the electric field , @xmath49 statv cm@xmath37 , that is sufficient to create the required poloidal magnetic field @xmath51 for the gaw excitation .
this electric field can be generated by the induction of a perpendicular electric field , @xmath52 statv cm@xmath37 , in the truncated region of the disk due to the magnetodynamo @xmath1 effect ( see fig . [ fig1 ] ) .
a difference of potential on the scale of the filament radius is of the order of @xmath53 statv .
this potential is much higher than is necessary and can generate a filament current with a hot electron tail and a characteristic velocity higher than that of the alfvn velocity @xmath54 , as was found in the plasma focus device .
we expect that the electrons in the tail will remain hot in the filament by their interaction with the gaw . for the parameters of the star assumed above
, the phase velocity of a gaw , @xmath55 , is @xmath56 cm s@xmath37 at the center of the filament .
taking into account the plasma resistance ( see , e.g. ginsbrug 1961 ) , the dissipation length of the alfvn waves can be estimated to be @xmath57 where @xmath58 this @xmath19 is a maximum value , determined by the condition that the dissipation length be sufficient so as to explain the plasma heating along the magnetic field from the truncated region to the star s surface . assuming a conversion of 5@xmath4 of the total accreting plasma energy flux , @xmath59 , into the alfvn wave flux , @xmath60
, we obtain a wave dissipation rate of @xmath61 erg @xmath33 s@xmath37 .
this dissipation rate produces a wave momentum transfer force @xmath62 ( see elfimov @xmath0 1996 ) . balancing the wave momentum transfer force with the electron - ion friction force @xmath63
, we obtain the density of the driven current , @xmath64 using a reduced friction , @xmath65 , for the electron tail with effective velocity @xmath66 and @xmath67 s@xmath37 ( @xmath68 cm s@xmath69 ) , we obtain the required current density , @xmath70 stata @xmath71 , in the filament . in this case , the electron tail that transports the current along the filament can be supported by the gaw via the electron landau damping , while the current in the filament can be driven by the gaw . here , we also present the results of calculations for the gaw in the filament . in order to calculate the wave field and dissipation , we use the eikonal model to analyze the wave propagation along the filament ,
exp@xmath72 $ ] , where @xmath73 is the coordinate along the filament axis and @xmath74 is the poloidal angle . in this case , in order to simplify the gaw calculations using numerical codes @xcite , instead of a dipole magnetic field , we assume a cylindrical model for the filament plasma with the current flowing along a magnetic field that has a helical magnetic field line configuration . to calculate the wave characteristics along the filament , we take the major radius @xmath75 cm comparable to the star s radius and the minor radius @xmath76 cm .
the calculations are carried out with the electron density profile of the form @xmath77 with @xmath78 @xmath33 and @xmath79 @xmath33 .
the temperatures are taken as @xmath80 ev and @xmath81 ev . the ion density is chosen so as to satisfy the requirement of charge neutrality , @xmath82 , where @xmath83 and @xmath84 are the densities of hydrogen and the other ions , respectively . in accordance with the spitzer conductivity @xmath85 , the current density profile has the form @xmath86 .
the value of the safety factor , @xmath13 , is assumed to be unity at the filament axis and to monotonically increase up to four at the plasma boundary . in fig .
[ fig2 ] , we show the characteristic profile of the magnetic field as well as the dissipated power of the gaw for the filament parameters discussed above . in fig .
[ fig2 ] , we observe the peaking of the dissipation at the filament center , which predicts filament center peaking of the electron temperature distribution , as well .
the gaw heats the core of the filament to higher temperature , while reducing the plasma temperature at the filament border due to heat diffusion .
the axial wavenumber of the gaw , @xmath87 , varies slowly along the magnetic field lines since it depends on changes in the local magnetic field and density .
our analysis of gaws in the magnetospheres of cttss shows that : + a ) the energy of the shock ( and/or accretion disk region ) can be transported from the surface of the star ( and/or accretion disk region ) to the magnetosphere region by gaws via current carrying filaments ; + b ) the dissipation length of gaws is on the order of the truncation radius ; + c ) the conversion of only a few percent of the shock energy into gaws is required in order to heat the filaments up to the observed temperature of the star s magnetosphere ; + d ) the current in the filament may be induced by the electric field , created by the @xmath1-magnetodynamo effect in the truncated region of the disk .
however , the major part of the current in the filaments is driven by gaws . | model for classical t tauri stars class ii objects @xcite central young star surrounded by thin accretion disk disk disrupted by magnetic field star smaller radii accretion flow follows magnetic field lines until stellar surface ghosh lamb 1979 knigl 1991 ) model explains signatures cttss excess optical ultraviolet continuum flux redshift absorption features emission line profiles muzerolle 1998 hartmann hewett calvet 1994 plasma disk flows along dipole magnetic field lines towards magnetic poles star inhomogeneous in perpendicular to magnetic field lines homogeneous field lines cavities plasma condensations plasmoids aligned by magnetic field present plasmoids distributed randomly perpendicular to plasma flow waves excited by plasma interact with plasmoids creating current current - carrying filaments electric current driven by waves or @xmath1-mhd effect in truncated region disk filaments observed in laboratory fadeev 1965 rosenberg 2001 occurrence filaments predicted in simulations rotating magnetized disks machida 2000 )hartmann hewett calvet 1994 muzerrole 1998 ) magnetosphere accretion models solving radiative transfer equations sobolev approximation theoretical balmer line profiles with observations major uncertainty magnetosphere temperature profile along tube martin 1996 ) calculated temperature profile from energy balance gas heating by adiabatic compression magnetic field balmer photoionization ambipolar diffusion values too low explain observed spectra additional heating mechanisms inside tube needed line - averaged electron temperature in magnetosphere about @xmath2 ev balancing heating rate with black body emission at temperature @xmath3 ev 5@xmath4 flux energy sufficient to heat downstream flux ev main consumption on plasma ionization data energy for plasma ionization rate about @xmath5 in filament for electron temperatures @xmath6 ev theoretical temperature profile varying from @xmath7 k to @xmath8 k reproduce optical line features evidence for turbulence necessity for additional heating mechanism vasconcelos jatenco - pereira opher 2002 ) suggested alfvn waves important in heating magnetic flux tubesstudied waves at star s surface due to shock produced by accreting matter suggested by scheurwater & kuijpers ( 1988 ) damping mechanisms for alfvn waves suggested alfvn resonant heating of solar loops hollweg 1990 elfimov 1996 wave damping by phase mixing cyclotron heating alfvn wave travels down magnetic field gradient until frequency matches cyclotron resonance of helium or other plasma specie due decreasing ion - cyclotron resonance magnetic conventional collisional viscous - resistive alfvn wave damping osterbrock 1961 ) vasconcelos @xmath0 2002 ) on nonlinear turbulent damping opher 2000 ) nonlinear turbulent damping studied by jatenco - pereira opher 1989a in solar wind protostellar winds late - type giant stars wolf - rayet stars quasar clouds extragalactic jets not previously analyzed possible mechanism for energy transport from shock truncation regions to magnetosphere of t tauri starspredicted ross 1982 cylindrical model magnetized plasmas axial current observed in tokamak experiments gaws excited in plasmas helical magnetic field poloidal component axial current @xmath9 @xmath10 axial magnetic field @xmath11 plasma radius length tokamak safety factor @xmath13 unity cold plasmas temperatures less than 0.1 ev gaws dissipate via electron - ion collisions hot plasmas alfvn velocity electron thermal velocity gaws dissipate via electron landau damping effective waves interact with electrons velocities equal to wave phase velocity @xmath18 frequency @xmath19 wavenumber parallel to magnetic field lines electromagnetic field components gaws proportional to @xmath20 poloidal @xmath21 axial @xmath22 wave numbers requirement waves propagate @xmath23 @xmath24 current field gaw large quality factor ratio frequency to dissipation decrement long dissipation length wave length good explain plasma heating far from source solar loopsgaw current calculated balancing wave momentum transfer with electron - ion friction force ( elfimov 1996 study two - fluid magnetohydrodynamic model of young star accreting plasma fig standard parameters cttss vasconcelos 2002 star radius cm truncated radius cm depth cm magnetic field star s surface kg kg truncated region electron density @xmath32 @xmath33 length magnetic field from to truncated region @xmath35 velocity shock cm star s surface cm truncated region ionization rate parameter for @xmath40 ev @xmath41 @xmath42 ( delcroix 1965 ) accretion energy density defined by ratio energy flux density to path length balancing accretion power density with ionization power density plasma average density along magnetic field @xmath45 @xmath33 magnetic field connecting truncated disk star s surface consists of filaments gaw propagates filament current driven by wave ponderomotive force magnetodynamo in truncated regioncurrent @xmath46}=5ab_\theta ef$ density @xmath47 @xmath48 in filament radius @xmath11 produced by electric field @xmath49@xmath37 poloidal magnetic field @xmath51 for gaw excitation electric field generated by perpendicular electric field @xmath52@xmath37 in truncated region disk due to magnetodynamo @xmath1 effect fig difference of potential filament radius @xmath53 statv potential higher than necessary filament current hot electron tail velocity higher than alfvn velocity @xmath54 electrons tail remain hot filament interaction gaw phase velocity of gaw @xmath55 is @xmath56 cm s@xmath37 at center filament plasma resistance dissipation length of alfvn waves estimated @xmath57 @xmath58 @xmath19 maximum value dissipation plasma heating along magnetic field from truncated region to star s surface conversion of 5@xmath4 of accreting plasma energy flux @xmath59 into alfvn wave flux @xmath60 wave dissipation rate of @xmath61 erg @xmath33 s@xmath37 dissipation rate produces wave momentum transfer force @xmath62 elfimov @xmath0 1996 balancing wave with electron - ion friction force @xmath63 obtain density driven current @xmath64 reduced friction @xmath65 electron tail velocity @xmath66 @xmath67 @xmath68@xmath69 obtain required current density @xmath70 @xmath71 in filament electron tail current supported by gaw electron landau damping current filament driven by gaw results calculations for gaw in filament calculate wave field dissipation eikonal model analyze wave propagation filament@xmath72 @xmath73 coordinate filament axis @xmath74 poloidal angle calculations assume cylindrical model filament plasma current flowing magnetic field helical magnetic field line configuration calculate wave characteristics filament major radius @xmath75 cm star s radius minor radius @xmath76 cm calculations with electron density profile @xmath77 @xmath78 @xmath33 @xmath79 temperatures taken as @xmath80 ev @xmath81 evion density chosen charge neutrality @xmath82 @xmath83 @xmath84 densities of hydrogen other ions spitzer conductivity @xmath85 current density profile form @xmath86 safety factor @xmath13 assumed unity at filament axis increase to four at plasma boundary fig . fig2 ] profile magnetic field dissipated power of gaw for filament parameters peaking dissipation at filament center predicts peaking electron temperature distribution gaw heats core filament to higher temperature plasma temperature at filament border heat diffusion axial wavenumber of gaw @xmath87 varies along magnetic field lines depends on changes local magnetic field density analysis of gaws in magnetospheres cttss shows energy shock disk transported from surface star to magnetosphere by gaws via current carrying filaments dissipation length of gaws on order of truncation radius conversion of few percent shock energy into gaws required to heat filaments to temperature star s magnetosphere current in filament induced by electric field @xmath1-magnetodynamo effect truncated region major part current driven by gaws | 0.505504 | 3,334 | 1,682 | , Saving $0.1 in GPT-4. |
21,008 | a fast trigger system is being designed as a potential upgrade to veritas , or as the basis for a future array of imaging atmospheric - cherenkov telescopes such as agis . the scientific goal is a reduction of the energy threshold by a factor of 2 over the current threshold of veritas of around 130 gev .
the trigger is being designed to suppress both accidentals from the night - sky background and cosmic rays
. the trigger uses field - programmable gate arrays ( fpgas ) so that it is adaptable to different observing modes and special physics triggers , e.g. pulsars .
the trigger consists of three levels : the level 1 ( l1.5 ) trigger operating on each telescope camera samples the discriminated pixels at a rate of 400 mhz and searches for nearest - neighbor coincidences . in l1.5 , the received discriminated signals are delay - compensated with an accuracy of 0.078 ns , facilitating a short coincidence time - window between any nearest neighbor of 5 ns .
the hit pixels are then sent to a second trigger level ( l2 ) that parameterizes the image shape and transmits this information along with a gps time stamp to the array - level trigger ( l3 ) at a rate of 10 mhz via a fiber optic link .
the fpga - based event analysis on l3 searches for coincident time - stamps from multiple telescopes and carries out a comparison of the image parameters against a look - up table at a rate of 10 khz .
a test of the single - telescope trigger was carried out in spring 2009 on one veritas telescope .
trigger , real - time processing , imaging atmospheric cherenkov technique | we are designing a fast topological trigger system @xcite with the goal of reducing the energy threshold for detection of gamma rays by a factor of two below the threshold of current imaging atmospheric - cherenkov telescopes ( iacts ) around 130 gev .
this trigger will analyze in real time the trigger - maps across all telescopes in the array and thereby improve the rejection of background over the current generation of trigger systems .
aspects that ultimately determine how low the energy threshold can be , include not only the speed and sophistication of the trigger , but also the memory depth in the front - end electronics and the speed of the data acquisition system
. this topological trigger may be part of a veritas upgrade proposal @xcite or it may interface with a highly integrated camera on next generation iacts , such as the advanced gamma - ray imaging system ( agis ) .
the current generation of imaging atmospheric cherenkov telescopes ( iacts ) typically achieve gamma - ray collection areas of @xmath0 m@xmath1 above 1 tev .
the gamma - ray collection area of veritas is shown in fig .
[ fig : collection_area ] at nominal trigger conditions ( _ solid line _ ) . the photomultiplier tube ( pmt )
thresholds are set at 5 photoelectrons ( pe ) , resulting in an energy threshold at the trigger level , e@xmath2 , of e@xmath3112 gev .
a lower energy threshold could be achieved by decreasing the pmt thresholds .
[ fig : collection_area ] shows two scenarios of lowered pmt thresholds that result in enhanced collection area below @xmath4100 gev .
however , these thresholds are not easily achievable because lowering the pmt threshold results in steeply rising data rates from background events : cosmic rays and night - sky background ( nsb ) light .
rejection of background events with veritas is achieved through an efficient three - stage trigger @xcite . at the primary level ( l1 ) ,
pmts generate a trigger if at least @xmath45 pe are detected within about 3 ns . following the threshold requirement , a pattern - sensitive logic ( l2 ) searches for 3 nearest - neighbor pmts that trigger within a time span of about 7 ns .
this camera trigger then sends a timing edge over optical fiber to a central station ( l3 ) .
a trigger is issued by l3 only if at least 2 telescopes produced a trigger within 100 ns of each other .
all parameters of the trigger setup are adjustable and were optimized for stable operation with less than 10% dead - time under a variety of observing conditions , including moon light .
the trigger rates of one of the telescopes and of l3 are shown in fig .
[ fig : array_bias ] as a function of the pmt threshold .
it can be seen that cosmic rays are the main source of background at the current operating threshold of 50 mv ( equivalent to 5 pe ) . however , if the pmt threshold is reduced to 3 pe , then a trigger would have to more efficiently reject both , nsb and cosmic rays .
[ fig : array_bias ] indicates that at a threshold of 3 pe , the efficiency for rejection of nsb must be improved by @xmath5 and that of cosmic rays by 70% .
such a background reduction is possible if the stereo images ( trigger maps ) could be analyzed in real time at the array level .
we are developing a trigger that implements a geometrical algorithm at the array level to reject more efficiently both , nsb and cosmic rays . in turn
, the reduced background rate would allow the energy threshold to be lowered without increasing the telescope dead - time .
30 mv ( about 3 pe ) without increasing the dead - time .
, width=288 ]
the topological trigger consists of high - speed camera - level triggers that transmit image parameters to the array trigger for real - time event classification .
the design is implemented with field - programmable gate arrays ( fpgas ) that allow the trigger to be adaptable to different observing modes ( e.g. a ring - shaped acceptance region for low - energy pulsar studies ) or concurrent special physics triggers ( e.g. searching for short bursts of gamma rays ) . while the design is flexible in the specific deployment , the hardware specification were chosen so as to be compatible with veritas . in order to maintain a responsive system with a dead - time of less than 10 the first level of the topological trigger ( l1.5 )
samples the discriminated camera signals at a rate of 400 mhz and searches for hits of three nearest neighbor pixels .
there are three l1.5 cards per camera .
the time window to detect coincidences is programmable to optimize the signal - to - noise ratio for detection of gamma rays .
a fast l1.5 is essential since gamma - ray events as imaged by the telescope are of @xmath45 ns duration .
currently , veritas is operating with a 7 ns coincidence window on each telescope .
the increased speed alone makes it possible to lower the energy threshold without increasing the single - telescope trigger rate . to achieve the required level of timing accuracy across the camera , programmable delay lines
are implemented on l1.5 .
this is necessary since nonuniform delays between pixels are introduced by cable lengths and variations in the high - voltage .
the processing speed of l1.5 is well matched to the 2 ns samples recorded by the fadcs and facilitate an appropriate reduction of the energy threshold in the off - line analysis .
the hit pixels are then sent to a second trigger level ( l2 ) that parameterizes the image shape and transmits this information along with the gps time stamp to the array - level trigger ( l3 ) at a rate of 10 mhz via a fiber optic link .
the fpga - based event analysis on l3 searches for coincident time - stamps from multiple telescopes and carries out a comparison of the image parameters against a look - up table at a rate of up to 10 khz .
we have identified an algorithm @xcite , described below , that is relatively simple to implement on l3 and is suitable for gamma / hadron separation below 100 gev .
the algorithm we intend to implement on l3 is based on the parallactic of shower - images seen from different points on the ground . as described in @xcite ,
only a first moment - parameterization of each image is required and is practical to implement within the limited time of @xmath410 @xmath6s available for data transmission and image processing .
the method is illustrated in fig .
[ fig : parallactic ] for a gamma - ray and a cosmic - ray primary arriving on - axis .
gamma rays produce narrow and well - defined images in each telescope , while cosmic rays produce sub - showers and result in erratic distribution of the cherenkov light between telescopes .
the parallactic displacement results from the different view points of telescopes on the ground and displaces the image away from the camera center . by drawing lines through the image centroid and the camera center
, pairs of telescopes produce an intersection . by projecting the intersecting lines on the ground ,
the impact location is reconstructed .
the spread of the pair - wise intersecting points is smaller for gamma rays than for cosmic rays .
the figure of merit , called _ parallaxwidth _ , is defined by the coordinates of the intersecting points , @xmath7 , for each telescopes @xmath8 , and the average over all such intersections , @xmath9 : @xmath10 the efficiency of the _ parallaxwidth _ parameter in separating gamma rays from cosmic rays is shown in fig .
[ fig : parallax_dist ] for an array of 50 telescopes .
gamma rays were simulated at 100 gev , while cosmic rays were drawn from a power - law distribution with index -2.7 between 50 gev and 10 tev .
the arrival direction for both , gamma and cosmic rays is on - axis , with telescopes pointed at zenith . since
cosmic rays are distributed isotropically , we are presenting a worst - case scenario for suppression of cosmic rays since off - axis cosmic rays will have even larger spread in the parallaxwidth parameter .
the distribution on fig .
[ fig : parallax_dist ] results from a minimum of 3 telescopes participating with a pmt threshold of 3.5 pe .
the nsb suppression expected with this topological trigger configuration is about @xmath11 when compared to the veritas standard `` 2 out 4 '' array trigger . 5 ) and the pair - wise intersection angle ( @xmath1230@xmath13 ) . , width=288 ]
a proof - of - principle trigger system is currently being developed by iowa state university and argonne national lab .
the hardware necessary for a complete single - telescope trigger has been produced and verified in the lab .
the trigger resides in one vme crate containing three l1.5 cards , one l2 card , and thirteen input / output ( i / o ) cards .
the i / o cards are necessary to test the topological trigger on a veritas telescope without disturbing night - time observations . for l3 ,
we have purchased a pci - based fpga card from faster technology that features four fiber optical connections to the telescopes .
the signal routing from the i / o cards to the l1.5 cards occurs over a custom backplane with gigabit connectors .
the operation of the l1.5 cards has been verified at 400 mhz through direct pattern injection into the i / o cards .
testing of the trigger was conducted in parallel with the standard trigger during late march 2009 on a single veritas telescope .
the fpga architecture includes adjustable front - end input delays to equalize the trigger arrival - times across all channels .
variations arise due to cable lengths and high - voltage settings .
delays up to 5 ns can be accommodated in the fpga and are programmable with a precision of 0.078 ns .
the measured timing distribution before and after compensating for the delay - variations is shown in fig .
[ fig : delay ] . following the delay - line calibration ,
simultaneous measurements were taken of the trigger rate versus pmt threshold , shown in fig .
[ fig : bias ] .
we observe a reduction in the night - sky accidentals by 50% and an increase of the cosmic - ray rate by 20% .
the reduction in the night - sky background ( nsb ) rate is expected due to the faster coincidence logic ( 6 ns vs. 7 ns window ) . the increased cosmic - ray rate may be due to a larger trigger area at the edge of the field of view ( 10% ) , and the fact that delay corrections are not applied to the veritas trigger inputs .
the single - telescope trigger test was thus successful and we anticipate further reducing the nsb rate by shortening the coincidence window to 3 ns through firmware development .
we would like to thank the veritas collaboration for use of the telescope and support during the trigger test .
this research is supported by iowa state university , argonne national lab through the ldrd program , and the us department of energy through the advanced detector research program ( defg0207er41497 ) . | designing fast topological trigger system @xcite reducing energy threshold for detection gamma rays by factor two below current telescopes around 130 gev trigger trigger - maps telescopes rejection background over current energy threshold include speed sophistication trigger memory depth front - end electronics speed data acquisition system trigger may part of veritas upgrade proposal or interface with integrated camera on next generation iacts advanced gamma - ray imaging system current generation telescopes achieve gamma - ray collection areas above 1 tev gamma - ray collection area veritas shown in fig . at nominal trigger conditions photomultiplier tube pmt thresholds set at 5 photoelectrons energy threshold at trigger level@xmath3112 gev lower energy threshold achieved by decreasing pmt thresholds scenarios lowered pmt thresholds enhanced collection area below @xmath4100 gev thresholds not easily achievable lowering pmt results in rising data rates from background events rejection of background events with veritas achieved through efficient three - stage trigger @xciteprimary level l1 pmts generate trigger if @xmath45 pe detected within 3 ns pattern - sensitive logic l2 ) searches for 3 pmts trigger within 7 ns camera trigger sends timing edge over to central station l3 trigger issued l3 if 2 telescopes trigger within 100 ns parameters trigger setup adjustable optimized for stable operation less than 10% dead - time under conditions including moon light trigger rates of l3 shown in fig function of pmt threshold cosmic rays main source background at threshold 50 mv 5 pe ) if pmt threshold reduced to 3 pe trigger nsb and cosmic rays 3 pe efficiency for rejection of nsb by @xmath5 cosmic rays by 70% background reduction possible if stereo images analyzed in real time at array level developing trigger geometrical algorithm to reject nsb and cosmic rays reduced background rate energy threshold without increasing telescope dead - time mv 3 pe topological trigger high - speed camera - level triggers transmit image parameters to array trigger for real - time event classificationdesign implemented with field - programmable gate arrays to observing modes or special physics triggers design flexible hardware specification compatible with veritas responsive system dead - time less than 10 first level topological trigger l1.5 samples camera signals 400 mhz searches for three nearest pixels three l1.5 cards per camera time window to coincidences programmable signal - to - noise ratio for detection gamma rays fast l1.5 essential gamma - ray events ns duration veritas with 7 ns coincidence window on each telescope increased speed energy threshold without increasing single - telescope trigger rate timing accuracy programmable delay lines implemented on l1.5 necessary nonuniform delays by cable lengths variations high - voltage processing speed l1.5 matched to 2 ns samples reduction energy threshold hit pixels sent to second trigger level ( l2 ) parameterizes image shape transmits information gps to array - level trigger ( l3 ) 10 mhz via fiber optic link fpga - based event analysis on l3 searches for coincident time - stamps from multiple telescopes image parameters against look - up table up to 10 khzidentified algorithm @xcite simple implement on l3 suitable for gamma / hadron separation below 100 gev algorithm l3 based on parallactic shower - images from points ground first moment - parameterization image required limited time of @xmath410 for data transmission image processing method illustrated in fig . for gamma - ray cosmic - ray arriving on - axis gamma rays produce narrow images cosmic rays produce sub - showers erratic distribution parallactic displacement from view points displaces image away from camera center drawing lines through image centroid camera center produce intersection projecting intersecting lines ground impact location reconstructed spread intersecting points smaller for gamma rays than cosmic rays figure merit _ parallaxwidth defined by coordinates intersecting points average over efficiency parallaxwidth separating gamma from cosmic rays shown in fig . for 50 telescopes gamma rays simulated at 100 gev cosmic rays from power - law distribution index -2.7 between 50 gev and 10 tev arrival direction for gamma cosmic rays on - axis telescopes pointed at zenithcosmic rays distributed isotropically worst - case scenario suppression off - axis rays larger spread parallaxwidth distribution fig . 3 telescopes pmt threshold 3.5 pe nsb suppression trigger @xmath11 compared veritas standard 2 out 4 array trigger pair - wise intersection angle @xmath1230@xmath13 proof - of - principle trigger system developed by iowa state university argonne national lab hardware single - telescope trigger produced verified trigger vme crate three l1.5 cards one l2 card thirteen input / output i / o cards test trigger night observations purchased pci - based fpga card from faster technology four fiber optical connections signal routing i to l1.5 cards over custom backplane gigabit connectors operation l1.5 verified at 400 mhz pattern injection testing standard late march 2009 single veritas telescope fpga architecture includes adjustable front - end input delays trigger arrival - times variations due to cable lengths high - voltage settings delays up to 5 ns programmable precision 0.078 ns measured timing distribution delay - variations shown in fig .fig delay line calibration measurements trigger rate versus pmt threshold fig . bias reduction night - sky accidentals 50% increase cosmic - ray rate 20% reduction night - sky background rate due faster coincidence logic 6 ns 7 increased cosmic - ray rate due larger trigger area edge field of view 10% delay corrections not applied veritas trigger inputs single - telescope trigger test successful anticipate reducing nsb rate shortening coincidence window 3 ns firmware development veritas collaboration telescope support trigger test research supported by iowa state university argonne national lab us department of energy advanced detector research program | 0.357607 | 2,410 | 1,157 | , Saving $0.1 in GPT-4. |
21,011 | we find a consistency between two different approaches of hard diffraction , namely the qcd dipole model and the soft colour interaction approach . a theoretical interpretation in terms of s - matrix and perturbative qcd properties in the small @xmath0 regime is proposed .
* 1 . * in the present paper , we focus on two different theoretical approaches to hard diffraction , which are shown to be compatible and complementary .
the first one is based on an extension @xcite to hard diffractive processes of the `` qcd dipole '' approach in the small @xmath0 regime of perturbative qcd . in this picture , the hard photon is supposed to probe the parton structure of the pomeron considered as an hadronic particle . in the original ingelman - schlein formulation ,
this hard probe was formulated in the parton model , later complemented by qcd evolution equations @xcite . in the dipole model , this hard interaction is described via the balitsky fadin kuraev lipatov ( bfkl ) resummation @xcite . a major uncertainty for this model is the relative normalization of diffractive over non - diffractive cross - sections which stays beyond the perturbative framework . a second approach @xcite is the soft colour interaction ( sci ) model , where hard diffraction is described as the superposition of two processes .
at short time , the hard probe initiates a typical deep - inelastic interaction with colour quantum numbers exchange .
then , at large times / distances , a `` soft '' colour interaction is assumed to rearrange the colour quantum numbers and gives rise to singlet exhanges -and thus diffraction- with a probability of order @xmath1 where @xmath2 is the number of colours . while in this approach , the relative normalization is fixed , the exact nature of the interplay between soft and hard components is not known . in the present paper
, we propose a way to relate these two approaches which allows one to consider them as compatible and complementary .
this provides a definite prediction for both the normalization and the analytic form of the amplitude .
our main results are the following : * i ) * the interplay between hard and soft components of hard diffraction is expressed via `` effective '' parameters of a pomeron interaction determined from leading log perturbative qcd resummation .
it is found to depend not only on @xmath3 but also on the ratio @xmath4 where y ( resp .
y ) are the total ( resp .
gap ) rapidity interval .
we obtain @xmath5 with @xmath6 is the bfkl evolution kernel @xcite ( together with its gaussian approximation near the minimum at @xmath7 ) and @xmath8 a non - perturbative scale associated with the proton .
we have @xmath9 note that one may write @xmath10 which is the known `` triple pomeron '' formula @xcite where @xmath11 defines the effective interaction cross - section of a virtual photon with a bfkl pomeron @xmath12 derived from the qcd dipole formalism . by analogy with bfkl @xcite , @xmath13 and @xmath14
can be defined , respectively , as the anomalous dimension , intercept and diffusion parameter of an `` effective '' bfkl @xmath15 cross - section .
the normalization , which remains unknown in the qcd dipole model description , is determined as the product of the factor @xmath16 and the normalization factor @xmath17 of the non - diffractive structure function , according to the sci prescription , see below .
* ii ) * in sci models , the following relation between the total structure function and the overall contribution of hard diffraction at fixed value reads of @xmath18 @xmath19 where @xmath20 is the minimal rapidity gap .
if we insert the qcd dipole prediction for @xmath21 in formula ( [ biczy4 ] ) , we find @xmath22 which is equivalent to a canonical bfkl expression for non - diffractive structure functions , apart the substitution of the bfkl effective anomalous dimension @xmath23 by @xmath24 namely @xmath25 where the `` universal '' value @xmath26 is solution ( for @xmath27 ) of the implicit equation @xmath28 . here , it is interesting to note that the shift ( [ value ] ) may be useful to avoid the objection to sci models @xcite based on `` low s theorem '' following which soft colour radiation can not be emmitted from inside a partonic process .
the differences we find with the original model means that , in our dipole formulation , the soft colour interaction indeed seems to modify the initial parton kinematics .
* iii ) * using s - matrix properties of triple - regge contributions , a relation is found between discontinuities of a @xmath29 amplitude and the two approaches to hard diffraction we consider . following old results of s - matrix theory in the regge domain @xcite , and as sketched in fig.1 , one may consider three types of discontinuities of a @xmath29 amplitude representing hard diffraction . a single discontinuity over the diffractive @xmath30 describing the hard pomeron interaction , a double discontinuity taking into account the analytic discontinuity in the subenergy variable of one of the incident pomeron exchanges ( and its complex conjugate ) for the sci model and the full triple discontinuity including those of the two pomeron exchanges , which is characteristic of the qcd dipole model description @xcite .
an interesting new feature is thus the s - matrix interpretation of the sci approach as a specific double discontinuity of the @xmath29 forward amplitude , which formulates the model in terms of simultaneous exchanges of a soft and a hard pomeron .
* 2 . *
let us sketch the derivation of our results .
our starting point is a triple - regge formula for the diffractive structure function for longitudinal and transverse photon in the qcd dipole formalism : f_t , l^diff(q^2,y , y)~ _ c - i^c+i ( 1-_1-_2- ) ( qq_0)^2 \{y ( ( _ 1)+(_2 ) ) + ( y - y ) ( ) } , biczy where @xmath31 is the bfkl evolution kernel ( [ chi ] ) and @xmath32 is a normalization containing both qcd perturbative and non - perturbative factors @xcite . strictly speaking @xcite
the @xmath33-function is the unique contribution in the differential diffractive structure function at momentum transfer @xmath34 however , it can be shown that this is the dominant perturbative contribution even at non zero transfer due to specific properties @xcite of the analytic qcd triple - pomeron couplings @xcite .
the first step of the computation of formula ( [ biczy ] ) is to use the saddle - point approximation @xcite at large @xmath35 to integrate over the difference @xmath36 one easily gets f_t , l^diff= n^diff1x_p _ c - i^c+i \{2y ( 2 ) + ( y - y ) ( ) + 2 qq_0 } .
biczy1 using the gaussian approximation ( [ chi ] ) for the bfkl kernels @xmath31 and @xmath37 in the relevant interval @xmath38 and again a saddle - point approximation at large rapidity gap @xmath35 , one obtains , up to a normalization factor , formula ( [ biczy10 ] ) , with @xmath13 and @xmath14 defined as in ( [ epsilon ] ) .
the normalization @xmath32 is not yet specified at this stage .
the derivation of the normalization is coming from the comparison with the sci approach . inserting ( [ biczy ] ) in the integral ( [ biczy4 ] ) ,
one is led to perform a two - dimensional saddle - point approximation in the @xmath39 complex plane .
the saddle - point equations read : -y ^ ( 2 ) + ( y - y ) ^ ( ) + 2 qq_0=0 2()-()=0 , cols whose solution @xmath40 is y^ * & = & ( y+ ) ( 1+)^-1 ( ^*)&=&2 ( ) solution resulting in a value of @xmath41 which is `` universal '' , i.e. independent of the kinematics of the reaction .
after computation of the prefactors to the saddle - point approximation , one finds : f_t , l^diff / tot = ^diff 1 ^(2)+^(^ * ) ( qq_0)^2^ * . biczy7 note that it is the linearity in @xmath35 of the saddle - point equation ( [ cols ] ) which allows one to get such an elegant form for the integrals
using ( [ biczy7 ] ) and by comparison with the canonical bfkl formula we identify the `` hard '' component of the sci model by the substitution @xmath42 see ( [ value ] ) . then using the sci ansatz ( [ biczy4 ] ) , we obtain the relation ^diff ^(2)+^(^ * ) norma which fixes the relative normalization of the diffractive vs. non diffractive structure functions .
this ends the derivation of formulae ( [ biczy10]-[value ] ) .
* 3 . *
let us finally come to the s - matrix interpretation of our approach .
for every fixed but arbitrary value of the parameters @xmath43 the triple - regge formula ( [ biczy ] ) can be obtained from the canonical formulalism _
elastic _ amplitudes in both the s - matrix framework @xcite and perturbative qcd at leading log level @xcite . ] corresponding to the vertex of three regge pole singularities in the complex plane of angular momentum @xcite .
as such , one can make use of the important s - matrix mueller - regge relation @xcite , valid in kinematical regions including the triple - regge limit , between semi - inclusive amplitudes and specific discontinuity contributions of forward elastic @xmath29 amplitudes .
it naturally applies to hard diffraction initiated by a virtual photon , as sketched in fig.1 , namely ^ * + pp + x disc_1\{^ * |p p ^ * |p p } .
[ 3to3 ] quite interestingly , the existence of regge phase factors allows one to relate other discontinuities of @xmath44 to @xmath45 as sketched in fig.1 , one may also consider a double discontinuity @xmath46 taking into account also the analytic discontinuity in the subenergy of one of the incident pomeron exchanges and a triple discontinuity @xmath47 including the discontinuity over the two pomeron exchanges .
the expression of the discontinuities , through generalized unitarity relations , is obtained through the imaginary part of the relevant regge phase factors @xcite .
moreover , one finds an equality relation @xmath48 which is due to the fact that the discontinuity taken over the mass variable ( corresponding to diffractively produced states ) is common to all three cases in fig.1 and factorizes the same @xmath49 vertex in @xmath50 ( cf . the classical derivation in the last paper of ref.@xcite ) .
let us now take advantage of the hard probe in the process , allowing one to introduce in the game the ( resummed ) perturbative qcd expansion at high energy ( small @xmath0 ) . in a generic s - matrix approach , the analytic discontinuities of scattering amplitudes
are related to a summation over a complete set of asymptotic _ hadronic _ final states . if however , the underlying microscopic field theory is at work with small renormalized coupling constant due to the hard probe , it is possible in some cases to approximate the same discontinuity using a complete set of _ partonic _ states .
in particular , at high energy and within the approximation of leading logs ( and also large @xmath2 ) , qcd dipoles can be identified as providing such a basis .
the discontinuity @xmath51 appears naturally in the dipole formulation of hard diffraction @xcite .
indeed , hard diffraction calculations make use of the probability distribution for finding two dipoles in the wave - function of one initial dipole in the virtual photon through the exchange over a bfkl pomeron exchange described by its discontinuity in the @xmath52 rapidity range
. then , each of these dipoles interact with the target through two perturbative ( bfkl ) pomeron exchanges described by their own discontinuities over the @xmath35 range .
thus the calculation of scattering amplitudes implies the full discontinuity over three pomerons described by intermediate dipole interactions , which is nothing but @xmath53 the appearance of @xmath54 is natural in the regge formalism @xcite .
however a physical interpretation involving perturbative qcd contributions has not been previously noticed . in the framework of our study
, it corresponds to the superposition of a hard perturbative cut pomeron interaction with a soft one described by a non - cut pomeron ( with complex conjugate contribution , see fig.1 ) .
this is similar to the superposition of hard and soft interactions which characterizes the sci approach .
we are thus led to propose @xmath54 as a way to get quantitative predictions , in particular the relative normalization of diffractive over non diffractive cross - sections .
indeed , it appears as a `` hard '' partonic interaction very similar to the one describing ordinary deep - inelastic processes , in parallel with a `` soft '' correction evolving during a long time , corresponding to the uncut pomeron singularity in the middle graphs ( including complex conjugate ) of fig.1 .
the equality between the different discontinuities allows the connection between the different models , leading to the results given in section * 1*. * figure * epsf = 8.truecm = 8.cm * figure 1 * _ s - matrix interpretation of the three approaches to hard diffraction . _
upper graph : description of @xmath55 ; middle graph : description of @xmath46 and its complex conjugate ( candidates for the sci approach ) ; lower graph : description of @xmath47 ( qcd dipole approach ) . | we thank a.bialas and w.buchmuller for valuable remarks .
99 a. h. mueller , b. patel , _ nucl . phys . _ * b425 * ( 1994 ) 471 ; + a.bialas , h.navelet , r.peschanski , _ phys .
lett . _ * b427 * ( 1998 ) 147 .
+ n. n. nikolaev , b. g. zakharov , _ z. phys .
_ * c49 * ( 1991 ) 607 ; _ z. phys . _ * c53 * ( 1992 ) 331 .
+ a. bialas , r. peschanski , _ phys .
_ * b378 * ( 1996 ) 302 ; * b387 * ( 1996 ) 405 .
l.n.lipatov , _ sov.j . nucl.phys._
* 23 * ( 1976 ) 642 ; v.s.fadin , e.a.kuraev and l.n.lipatov , _ phys .
* b60 * ( 1975 ) 50 ; e.a .
kuraev , l.n .
lipatov and v.s .
fadin , _ sov.phys .
* 44 * ( 1976 ) 45 , * 45 * ( 1977 ) 199 ; i.i.balitsky and l.n.lipatov , _ sov.j .
nucl.phys._ * 28 * ( 1978 ) 822 ; l.n .
zh.eksp.teor.fiz . _
* 90 * ( 1986 ) 1536 ( eng .
* 63 * ( 1986 ) 904 ) .
j. bartels , m. wusthoff , _ zeit .
_ * c66 * ( 1995 ) 157 ; j. bartels , h. lotter , m. wusthoff , _ zeit .
_ * c68 * ( 1995 ) 121 ; a.bialas , w.czyz , hep - ph/9808437 , _ acta phys .
polon . _ * b29 * ( 1998 ) 2095 . | thank a.bialas w.buchmuller remarks mueller patel nucl phys . b425 1994 ) 471 a.bialas h.navelet r.peschanski phys b427 1998 ) 147 nikolaev zakharov z c49 1991 ) c53 1992 ) 331 a bialas peschanski b378 1996 ) 302 b387 1996 ) 405 l.n.lipatov sov.j nucl.phys. 23 1976 ) 642 v.s.fadin.kuraev l.n.lipatov phys b60 1975 ) 50 kuraev lipatov fadin sov.phys. 44 1976 ) 45 45 1977 ) 199.balitsky l.n.lipatov sov.j nucl.phys. 28 1978 ) 822.eksp.teor.fiz 90 1986 ) 1536 63 ( 1986 ) 904 j. bartels m wusthoff zeit c66 1995 ) 157 bartels h. lotter wusthoff c68 ( 1995 ) 121 a.bialas w.czyz hep - ph/9808437 acta phys . b29 ( 1998 ) 2095 | 0.07392 | 490 | 287 | , Saving $0.0 in GPT-4. |
21,012 | the pool project is the common persistency framework for the lhc experiments to store petabytes of experiment data and metadata in a distributed and grid enabled way .
pool is a hybrid event store consisting of a data streaming layer and a relational layer .
this paper describes the design of file catalog , collection and metadata components which are not part of the data streaming layer of pool and outlines how pool aims to provide transparent and efficient data access for a wide range of environments and use cases - ranging from a large production site down to single disconnected laptops . the file catalog is the central pool component translating logical data references to physical data files in a grid environment .
pool collections with their associated metadata provide an abstract way of accessing experiment data via their logical grouping into sets of related data objects . | the pool project@xcite is the common persistency framework for the lhc experiments to store petabytes of experiment data and metadata in a distributed and grid enabled way .
the pool is a hybrid event store combining c++ object streaming technology such as root i / o @xcite for the bulk data with a transactionally safe relational database store such as mysql for file catalog , collection and metadata .
this paper describes the design of pool components which are not part of the data streaming layer@xcite and outlines how pool aims to provide transparent and efficient data access for a wide range of environments and use cases - ranging from a large production site down to single disconnected laptops .
the pool file catalog is the central pool component translating logical data references to physical data files in a grid environment .
pool collections with their associated metadata add another more abstract way of accessing experiment data via their logical grouping into sets of related data objects .
ad hoc queries on the collections provide physicist with an efficient way to extract useful data .
in a file based persistency mechanism such as pool , the storage components need to operate on the contents of the file .
one should be able to navigate from one object to the other even if they are not stored in the same file .
using an external file catalog to keep track of the physical location of the file is more flexible than hard - coding the location information in the file itself .
it allows the file to be moved or replicated . in the grid environment ,
high level applications view only logical files .
a logical file does not exist physically . a physical file and all
its copies can be viewed as the same logical file . in another word ,
physical files are representations of the logical file .
to operate on the contents of a logical file , a service is needed to map the logical file to one of its physical representations and this physical file can then be opened .
the physical location of a file can be represented by its physical file name(pfn ) while the logical file can be represented by a unique logical identifier . in pool , this identifier is called the file i d and is generated via the guid ( global unique identifier ) mechanism .
the file catalog component in pool is responsible for maintaining a list of physical locations of accessible files together with their unique and immutable ids and translating the logical file reference into its physical representation .
it is a basic component for object navigation and grid integration .
the basic content of the catalog is the one - to - many mapping of the file i d and the pfns .
this mapping is sufficient for object lookup and navigation .
optionally , the file catalog can store logical file names(lfns ) .
in contrast to the file i d , which is not easy for the user to read and to remember , the lfns may contain human readable and memorisable strings .
lfn can be seen as the alias of the file i d .
one file i d may have many lfns .
in addition to pfn and lfn , a set of user defined file metadata can also be associated with a file i d .
metadata - fileid mapping is one - to - one which means one file can be associated with one set of well defined metadata .
the purpose of the file metadata support is to help selecting a fragment of the catalog when necessary .
figure [ fig : fcschema ] shows the logical view of the content of the catalog : the pfn - fileid mapping , which is essential for the object lookup and optionally , the lfn - fileid and the metadata - fileid mapping .
-90 file registration and lookup are two fundamental functionalities of the component . to register a file
is to insert a file i d and pfn pair in the catalog while to lookup a file is to resolve a file i d into a pfn or vice versa .
similar functions are provided also for the file i d and lfn mapping . besides
, one can extract a subset of files from one concrete catalog and cross populate them into another catalog .
specific metadata can be associated to the file and queries on the metadata can be used to select the catalog fragments . in the cross catalog operations
the source and the destination catalogs need not to have the same backend implementation .
the component provides both c++ api and command - line tools .
the c++ api is used by the storage components in pool and the experiment framework to register and lookup a file inside the application process while command - line tools can be used outside the application process for catalog management operations . three concrete catalog : xml , mysql and edg - rls based implementations
are provided under single abstract interface .
concrete catalogs are loaded dynamically at run time . to take advantage of the relational database technology ,
transactions are supported by the component . in the transaction
the user can commit or rollback the changes to the catalog .
the three implementations are described in more detail as follows : * xml catalog + the xml catalog is useful when the user wants to run the application disconnected from the network and the number of entries in the catalog is not too large . running applications connected to a local xml catalog do not rely on any central service .
the xml catalog can also be used in data migration .
for example , when one migrates the data in a mysql backend into a , say , oracle backend , the xml catalog can be used as the intermediate format .
* mysql catalog + the mysql catalog is connected to the mysql database server .
the database based catalog can handle concurrent accesses and larger data volume than the xml catalog .
the mysql catalog can be used in larger scale applications such as in a production farm .
* edg catalog + a grid based catalog can be used by the entire virtual organisation ( vo ) .
the edg project @xcite will provide the replica management service , which controls files that belong to a vo .
in particular , the replica location service(rls ) component@xcite maintains information about the physical location of files , while the replica metadata catalog(rmc ) component@xcite provides the information on the logical file names and metadata .
the file catalog component in pool provides an interface to the edg - rls and edg - rmc . in another word ,
pool applications connected to the edg catalog are on the grid .
the file catalog component provides a graphic user interface for browsing the content of the catalog .
a first prototype of the browser has been developed in python scripting language .
* read / write files in pool applications + when the pool persistency manager writes in a file , it requests the i d of the file of given pfn from the file catalog component .
the pfn may or may not already be registered in the catalog . if the file is already registered , the file catalog component returns the file i d to the persistency manager , otherwise the file catalog component generates a new file i d , registers it in the catalog and returns it to the persistency manager .
+ when the persistency manager opens a file for reading , it requests the pfn of the file of a given file i d from the file catalog component .
the file catalog component looks up the pfn in the catalog and returns it to the persistency manager .
if more than one pfns are found , in the case of a grid based catalog , the optimal pfn will be returned , in an off - the - grid case , the pfn of the master copy of the file will be returned . *
manage a production in a local farm + the user extracts a catalog fragment needed by the job for reading from a central catalog(edg or mysql based ) into a local xml catalog .
user runs the job disconnected from the network . during the running of the job ,
the local input xml catalog is used when retrieving objects ; the output files are registered into another local xml catalog . after n job runs , n output xml catalogs are produced .
user cleans the entries in the catalogs produced by unsuccessful jobs and publishes the local catalog fragments to the central catalog making the produced data files available to a group of users or the entire production site .
+ in this setup , when jobs are running they do not depend on any central services .
performance tests have been set up to compare the behavior of different catalog implementations . for the xml catalog
, tests were run on a single pentium iii-1.2ghz computer with about 200 mb free memory .
the maximum number of entries have been tested in a single xml catalog is 50,000 . the major time spent for
the xml catalog is on the xml dom parser initialization .
it takes about 10ms to initialize a new catalog and about 6s to start a catalog with 20k entries .
the pfn registration operation takes about 0.3 ms / entry on average . for the mysql catalog
, tests were run on 10 client nodes and a mysql server node on the lan .
the maximum number of entries have been tested in a single mysql catalog is 1 million . in the case of 300 concurrent client processes from the 10 client nodes querying the mysql server and committing every 100 entries
, the pfn registration operation takes about 1.5 ms / entry on average .
similar tests have been performed against the edg - rls server which is web service on top of relational database .
the pfn registration time is in the range from about 30ms / entry to 6ms / entry for multithreaded applications depending on the combination and the setup of the web service and the backend database . in all tests ,
a single pfn entry is about 200 characters long and the file i d size is 36 characters .
this is the first performance test of the pool file catalog , no systematic performance test and tuning have been attempted so far .
the purpose of the pool collection component is to provide the tools needed to manage potentially large ensembles of objects stored by means of pool s persistence services .
the collection component provides the infrastructure to support definition , creation , population , and use of such ensembles , including query , selection , and iteration services , and higher- level utilities .
collections provide an entry point for physicists , a locus for recording the navigational information necessary to allow later iteration through a list of data objects that may be physically dispersed across a very large persistent store .
a collection of events that are of interest to a particular physics working group because they share certain physics characteristics is a motivating example . to support selection of specific subsets of interest from very large collections , it is useful to provide a means to query attributes of those objects without retrieving the objects themselves .
the metadata component in pool provides the machinery to define and manage object- and collection - level metadata . named collections populated by explicit calls to insertion operators are called `` explicit collections '' in pool .
pool also supports `` implicit collections''collections of objects grouped by physical containment .
the set of all objects of type t in a given container or file is an example of an implicit collection .
an object persistified by writing its state into a file is implicitly an element of the collection of all such objects in the file , but the same object may also be registered in one or more explicit collections .
explicit and implicit collections are read via the same abstract interface , so client code does not in any way depend upon whether an input collection is explicit or implicit .
the first implementation of explicit collections is in mysql .
the intent is to allow the possibility of server - side selection of objects of interest from queryable collections by means of database - provided sql services , so that the list of qualifying objects , and , by extension , the list of data files needed , may be determined at a stage prior to job scheduling and grid resource brokering , without the need to first retrieve the data files and navigate into application - specific data . pool also supports hierarchical collections , tree - like structures that accept other ( sub)collections as their elements .
the design allows complex selection based upon both collection- and object - level metadata .
* create / write explicit collections + user first defines the attribute lists of collections will be produced by the process .
then she processes the event and gathers attribute data . for each event
, the attribute data and a reference , which contains enough information to retrieve the persistent object , is added to the corresponding collection . at the end of process ,
several collections with different names are created and persisted . *
read explicit collections + during analysis , the user specifies a number of collections and provides a query to select only the objects of interest .
then she iterates through the returned persistent object references , retrieve the objects and process them in the analysis .
file catalog , collection and metadata components are part of the so - called relational layer of the hybrid event store pool .
such a relational layer takes advantage of the underlying relational database technology , e.g. the transaction , concurrency and queries , to achieve efficient management of files and object metadata .
the file catalog is one of the basic components which provide transparent object navigation in the event store .
the design of the component can accommodate wide ranges of use cases .
it is grid - aware but also preserves grid - decoupled modes .
9 d. dllmann _ et al .
_ , these proceedings mokt007 , and http://pool.cern.ch r. brun and f. rademakers , nucl .
meth . * a 389 * ( 1997 ) 81 , and http://root.cern.ch d. dllmann _ et al .
_ , these proceedings mokt008 , and physics/0306084 .
http://edg-wp2.web.cern.ch/edg-wp2/ http://proj-grid-data-build.web.cern.ch/ + proj - grid - data - build / edg - rls - server/ + user - guide / html/ http://proj-grid-data-build.web.cern.ch/ + proj - grid - data - build / edg - metadata - catalog/ + user - guide / html/ | pool project@xcite is persistency framework for lhc experiments petabytes experiment data metadata distributed grid enabled pool is hybrid event store combining c++ object streaming technology safe relational database store mysql for file catalog collection metadata paper describes design of pool components not data streaming layer@xcite transparent efficient data access for environments use cases large production site disconnected laptops pool file catalog central component translating logical data references to physical data files in grid environment pool collections metadata add accessing experiment data via logical grouping into related data objects ad hoc queries on collections provide efficient extract data in file based persistency mechanism pool storage components operate on contents file navigate from object to other even not file external file catalog physical location more flexible than hard - coding location file allows file moved replicated in grid environment applications view logical files logical file exist physically physical file copies viewed as same logical file physical files are representations of logical file operate service map logical file to physical physical file opened physical location file represented by physical file name logical file by unique logical identifierin pool identifier file i d generated via guid global unique identifier mechanism file catalog physical locations of files unique immutable ids translating logical file reference into physical representation basic for object navigation grid integration content catalog is one - to - many mapping of file i d pfns sufficient for object lookup navigation file catalog store logical file names(lfns contrast to file i d lfns contain human readable memorisable strings lfn alias of file i d one file i d may have many lfns pfn lfn user defined file metadata associated with file i d metadata - fileid mapping is one - to - one one file associated with one metadata file metadata support fragment catalog figure shows view catalog pfn - fileid mapping essential for object lookup lfn - fileid metadata - fileid mapping file registration lookup fundamental functionalities register file file i d pfn in catalog lookup file resolve file i d into pfn similar functions for file i d lfn mapping extract files from one catalog cross populate into another catalog specific metadata associated to file queries metadata to select catalog fragmentscross catalog operations source destination catalogs same backend implementation component provides c++ api command - line tools c++ api used by storage pool experiment framework to register lookup file inside application process command - line tools outside process for catalog management three concrete catalog xml mysql edg - rls based implementations provided under single abstract interface catalogs loaded dynamically at run time relational database technology transactions supported by component user can commit or rollback changes to catalog three implementations xml catalog useful when application disconnected from network entries not large applications connected to local xml catalog rely on central service used in data migration intermediate format mysql catalog + connected to mysql database server concurrent accesses larger data volume used in larger scale applications production farm edg catalog + grid based catalog used by entire virtual organisation vo ) edg project @xcite replica management service controls files vo replica location service(rls ) component@xcite maintains physical location files replica metadata catalog(rmc ) provides information logical file names metadata file catalog component pool interface to edg - rls edg - rmc pool applications connected to edg catalog on gridfile catalog provides graphic user interface for browsing content catalog first prototype browser developed in python read / write files in pool applications pool persistency manager writes in file requests i d of file given pfn from catalog pfn may or not be registered in if file registered returns i d to persistency manager otherwise generates new file i d registers returns to persistency manager opens file for reading requests pfn of from catalog looks pfn returns if more than one pfns found optimal pfn returned off - - grid pfn of master copy file returned manage production in local farm user extracts catalog fragment from central catalog into local xml catalog runs job disconnected from network local input xml catalog used retrieving objects output files registered into another local xml catalog after n job runs n output xml catalogs produced user cleans entries catalogs publishes local catalog fragments to central catalog making produced data files available to group users or entire production site jobs depend on central services performance tests set up to compare behavior catalog implementations tests run on single pentium iii-1.2ghz computer with 200 mb free memorymaximum entries tested in single xml catalog is 50,000 major time on xml dom parser initialization 10ms to initialize new catalog 6s to start catalog with 20k entries pfn registration takes 0.3 ms / entry mysql catalog tests run on 10 client nodes mysql server node on lan maximum entries tested in catalog is 1 million 300 concurrent client processes from every 100 entries pfn registration takes 1.5 ms / entry similar tests performed against edg - rls server web service relational database pfn registration time 30ms to 6ms / entry for multithreaded applications depending on web service backend database single pfn entry 200 characters long file i d size 36 characters first performance test of pool file catalog no systematic performance test tuning attempted pool collection tools manage large ensembles of objects pool persistence services provides infrastructure definition creation population use of ensembles query selection iteration services higher utilities collections provide entry point for physicists locus for recording navigational information later iteration data collection of events interest to physics working group motivating examplesupport selection subsets from large collections means query attributes without retrieving objects metadata component in pool provides manage object- collection - level metadata named collections populated by explicit calls insertion operators called explicit collections pool supports implicit collections''collections grouped by physical containment set objects type t in container file example implicit collection object persistified by writing state into file element of collection same object registered in explicit collections explicit implicit collections read via same interface client code depend upon input collection explicit or implicit first implementation of explicit collections in mysql intent server - side selection of objects from queryable collections by database - provided sql services list qualifying objects data files needed determined prior job scheduling grid resource brokering without need retrieve data files application - specific data pool supports hierarchical collections sub design allows complex selection based collection- and object - level metadata create / write explicit collections user defines attribute lists collections process processes event gathers attribute data each event attribute data reference persistent object added to collection end process several collections with different names created persistedread explicit collections analysis user specifies collections provides query select objects interest iterates through returned object references objects analysis file catalog collection metadata relational layer hybrid event store pool layer relational database technology transaction concurrency queries efficient management of files object metadata file catalog basic transparent object navigation event store design use cases grid - aware preserves grid - decoupled modes dllmann et al . proceedings mokt007 http://pool.cern.ch r. brun f. rademakers nucl . meth a 389 * ( 1997 ) 81 http://root.cern.ch dllmann et al proceedings mokt008 physics/0306084 http://edg-wp2.web.cern.ch http://proj-grid-data-build.web.cern.ch data build edg - rls - server user - guide edg - metadata - catalog user - guide | 0.439956 | 2,885 | 1,391 | , Saving $0.1 in GPT-4. |
21,018 | it is shown that the chemical freeze - out parameters obtained at cern / sps , bnl / ags and gsi / sis energies all correspond to a unique value of 1 gev for the average energy per hadron . | information about the chemical freeze - out parameters , the temperature , @xmath0 , and baryon chemical potential @xmath1 can be obtained from ratios of integrated particle yields @xcite .
this is because various effects like transverse flow or particle production from a superposition of fireballs cancel out in such ratios provided the freeze - out parameters are unique .
we will discuss in succession particle ratios at gsi / sis , bnl / ags and sps / cern .
we will then discuss common properties , in particular , the observation that all of them correspond to an average energy of 1 gev per hadron independent of the beam energy @xcite .
a systematic study of the particle ratios measured at gsi / sis energies has been done recently @xcite . in figure
[ fig : gsi ] we show the results obtained in @xcite for ni - ni at 1.0 agev .
the particle ratios are consistent with the interpretation that the hadronic composition of the final state is fixed at a unique temperature and baryon chemical potential .
since the temperature is low , the number of particles created is small and it is therefore necessary to take into account the exact conservation of strangeness because strange particles are always produced in pairs and it is more difficult to create two particles in a small cold system than it is to create one particle .
there is no necessity to introduce any other parameters and a good fit can be achieved with full chemical equilibrium including strange particles , i.e. with @xmath2 . in view of the fact that kaons are produced below threshold at sis
this is remarkable .
particle ratios at bnl / ags have been reanalyzed recently in references @xcite and @xcite .
the results are consistent with those obtained previously , e.g. ref .
@xcite obtains @xmath3 mev and @xmath4 mev .
the ratios together with error bands are shown in figure [ fig : ags ] @xcite .
this temperature is about double the one extracted from sis results , the baryon chemical potential is clearly much smaller . because of the higher temperature , it is no longer necessary to treat strangeness in a special way and corrections from the more exact canonical treatment are negligible .
there is agreement that a good fit can be achieved with @xmath2 and the data are again consistent with chemical equilibrium .
several papers have appeared recently analyzing the particle ratios measured in pb - pb collisions at cern @xcite . in reference
@xcite it is found that a strangeness suppression factor @xmath5 is needed to describe the data while in reference @xcite it is argued that a good description is possible using @xmath2 . despite this serious difference
both papers arrive at very similar results : @xmath6 mev in @xcite vs @xmath7 mev in @xcite , i.e. the difference is less than 3 percent .
the disagreement over the value of @xmath8 will be settled when more precise data become available , indications from this conference are that a value below 1 is necessary but a full analysis is still outstanding .
the analysis of reference @xcite introduces new parameters and is not directly comparable to the one discussed here .
the results from gsi , bnl and cern show a striking systematic behavior ( see figure 3 ) : the gsi / sis results have the lowest temperature and the highest baryon chemical potential , as the beam energy is increased a clear shift towards higher temperatures and lower baryon chemical potentials occurs .
the points have in common that the average energy per hadron is approximately 1 gev @xcite .
chemical freeze - out is thus reached when the energy per particle drops below 1 gev per hadron . when this value is reached inelastic collisions are no longer important and the abundances of the various hadronic species are fixed .
the consequences of this are discussed in detail in reference @xcite .
recent results from e895 collaboration presented at this conference @xcite are compatible with the above result .
bringing together results obtained at very different beam energies shows that the hadronic abundances seen in the final state of relativistic heavy ion collisions are fixed once the average energy per hadron drops below 1 gev .
we expect the results from the sps beams at 40 and 80 gev to follow this observation . it will be interesting to see how the results from rhic will relate to this observation . 9 u. heinz , j. phys .
g * 25 * , 263 ( 1999 ) .
j. cleymans , h. oeschler and k. redlich , j. phys .
g * 25 * 281 ( 1999 ) .
j. cleymans and k. redlich , phys .
lett . * 81 * 5284 ( 1998 ) .
j. cleymans , d. elliott , a. kernen and e. suhonen , phys .
rev . c*57 * 319 ( 1998 ) .
j. cleymans , h. oeschler and k. redlich , phys .
rev . c*59 * 3319 ( 1999 ) .
f. becattini , j. phys .
g * 25 * 287 ( 1999 ) .
j. stachel , private communication and i. heppe , diplomarbeit , university of heidelberg , 1998 .
j. cleymans , d. elliott , r.l .
thews and h. satz , z. phys . * 74 * 319 ( 1997 ) .
p. braun - munzinger , i. heppe and j. stachel , nucl - th/9904058 .
g.d . yen and m. gorenstein , phys .
rev . c*59 * 2788 ( 1999 ) .
j. rafelski and j. letessier , nucl - th/9903018 .
j. cleymans and k. redlich , nucl - th/9903063 .
g. rai , contribution to this conference . | chemical freeze - out parameters temperature @xmath0 baryon chemical potential @xmath1 from integrated particle yields @xcite effects like transverse flow particle production from superposition fireballs cancel out in ratios freeze - out parameters unique discuss particle ratios at gsi / sis bnl / ags sps / cern common properties correspond to average energy 1 gev per hadron independent of beam energy @xcite systematic study of particle ratios at gsi / sis energies done @xcite figure results for ni - ni at 1.0 agev ratios consistent with hadronic composition final fixed at unique temperature baryon chemical potential temperature low number particles small necessary conservation of strangeness produced in pairs difficult to create two particles small cold system one no necessity to introduce other parameters good fit with full chemical equilibrium including strange particles @xmath2 kaons produced below threshold at sis particle ratios at bnl / ags reanalyzed @xcite results consistent with @xmath3 @xmath4 mev ratios error bands shown in figure ags temperature double sis results baryon chemical potential smallerhigher temperature no longer necessary treat strangeness corrections from canonical treatment negligible agreement good fit with @xmath2 data consistent with chemical equilibrium papers appeared analyzing particle ratios in pb - pb collisions at cern @xcite strangeness suppression factor @xmath5 needed good description possible using @xmath2 papers similar results @xmath6 mev @xcite vs @xmath7 mev difference less than 3 percent disagreement over value @xmath8 settled when precise data available value below 1 necessary full analysis outstanding analysis of reference @xcite introduces new parameters not comparable results from gsi bnl cern show systematic behavior gsi / sis results have lowest temperature highest baryon chemical potential beam energy shift towards higher temperatures lower baryon chemical potentials average energy per hadron approximately 1 gev @xcite chemical freeze - out reached when energy per particle drops below 1 gev per hadron inelastic collisions no important abundances hadronic species fixed consequences discussed @xcite recent results from e895 collaboration compatible with result hadronic abundances in ion collisions fixed once average energy per hadron drops below 1 gevexpect results sps beams 40 80 gev follow observation interesting see results rhic heinz j phys 25 263 1999 ) cleymans h oeschler k redlich 25 281 1999 cleymans redlich 81 5284 1998 ) cleymans d. elliott. kernen suhonen*57 319 1998 cleymans oeschler redlich*59 3319 1999 f becattini phys g 25 * 287 1999 ) j stachel i. heppe diplomarbeit university heidelberg 1998 cleymans d. elliott r thews h. satz. 74 * 319 1997 ). braun - munzinger i. heppe j. stachel nucl th/9904058 yen m. gorenstein phys rev c*59 2788 1999 ) j. rafelski j letessier nucl th/9903018 j cleymans k. redlich nucl th/9903063 g. rai contribution conference | 0.186218 | 1,298 | 674 | , Saving $0.0 in GPT-4. |
21,027 | we present the results of searches for anisotropy in the right ascension ( @xmath0 ) distribution of arrival directions of cosmic rays ( crs ) detected with the yakutsk array during the 19742008 observational period in the energy range above @xmath1 ev .
two methods of analysis are applied to two sub - samples of the data .
particularly , estimations of the first and second harmonic amplitudes are given , as well as the first harmonic phase in adjacent energy intervals .
analysis of variance demonstrates a significant contraction of the minimal width of the @xmath0 distribution in the energy bin @xmath2 ev with respect to the isotropic distribution , which may be attributed to a possible source of crs within the interval @xmath3 . cosmic ray , extensive air shower , harmonic analysis , analysis of variance 95.85.ry , 96.50.sd , 98.70.sa | a conventional approach to shed light on the origin of cosmic rays ( crs ) is to search for anisotropy in the arrival directions distribution .
a number of attempts have been made to find excess fluxes of crs correlated with large - scale structures in the nearby universe , resulting in several indications of possible anisotropic effects , but none of these effects have been confirmed independently ( some examples can be found in refs .
@xcite ) . by restricting the data under analysis exclusively to ground - based arrays , an essentially uniform exposure in the right ascension can be obtained . under this restriction , in the present paper we examine the data obtained from scintillation counters surface detectors of the yakutsk array .
the results of previous analysis of the @xmath0 distribution of extensive air shower ( eas ) primaries detected within the time period from january 1974 to may 2000 were published in refs .
we have now analyzed the extended dataset up to june 2008 .
additional data include , specifically , 7598 eas events above the threshold energy 1 eev ( = @xmath1 ev ) @xcite .
the main aim of this analysis was to extend the time series of the data by using an observational time that was as long as possible .
these efforts were aimed at testing the previous indications of possible anisotropy in the arrival directions of crs in the extended dataset using different methods .
the yakutsk array site is located near oktyomtsy , the satellite village of yakutsk , at geographical coordinates @xmath4 and at a mean altitude of 105 m above sea level . at present
, it consists of 58 ground - based and 4 underground scintillation counters to measure charged particles ( electrons and muons ) , and 48 detectors of the air cherenkov light . during its 40-year lifetime
, the array has been reconfigured several times . before 1990 ,
the total area covered by detectors was at its maximum ( @xmath5 km@xmath6 ) ; now , it is @xmath7 km@xmath6 .
eas events were selected from the background using a two - level trigger of detector signals : the first level is a coincidence of signals from two scintillation counters in a station within 2 @xmath8s ; the second level is a coincidence of signals from at least three nearby stations ( not lined up ) within 40 @xmath8s .
stations spaced @xmath9 m and @xmath10 m form the so - called trigger-500 and trigger-1000 , respectively . over the entire observation period ,
more than @xmath11 showers of primary energy above @xmath12 ev were selected .
[ fig : map ] shows the positions of detectors of the yakutsk array and the core positions of eass with @xmath13 eev .
only scintillation counters of the array with spacing 500 and 1000 m are shown .
other types of detectors are described , for example , in refs .
@xcite .
the shower core coordinates were located fitting the lateral distribution of particle densities by the greisen - type function .
core location errors are @xmath14 m for trigger-500 and @xmath15 m for trigger-1000 events .
arrival angles of the eas primary particles were calculated in the plane shower front approximation using detection times at the stations . a clock pulse transmitter at the center of the array provided pulse timing to 100 ns accuracy .
errors in arrival angles depend on the primary energy decreasing from @xmath16 at @xmath17 eev to @xmath18 above @xmath19 eev .
more detailed information can be found in refs .
@xcite . in this work , a sample of the analyzed dataset consisted of eas events detected in the period january 1974 june 2008 within the array area , with energy above @xmath20 eev , at zenith angles @xmath21 .
the energy estimation method was based on the total flux measurement of the air cherenkov light and the number of electrons and muons at observation level @xcite . to unify the energies of showers detected in a variety of years
, we used the same @xmath22-to - energy relationship and attenuation length as in ref .
the number of eas events selected was 43710 at @xmath23 eev .
the energy estimation error was approximately @xmath24 for the showers with axes within the array area @xcite .
the systematic error in the energy estimation procedure of the yakutsk array , as in the case of other giant arrays , could be corrected by a specific factor @xcite .
[ fig : eqfig ] shows the distribution of arrival directions of selected showers in the equatorial system compared with expected isotropic distributions .
we used the rayleigh formalism in harmonic analysis of the @xmath0 distribution of cr arrival directions @xcite .
if the large - scale anisotropy is present in the arrival directions of crs , it can be expressed using the anisotropy coefficient for the intensity variation : @xmath25 where @xmath26 and @xmath27 are the maximum and minimum of cr intensity , respectively , as a function of the arrival angle , @xmath28 . in the simplest cases with one source at @xmath29 ( for example , @xmath26 from the galactic center , @xmath27 from the anticenter )
, the distribution can be described by a cosine wave @xmath30 , or with the two opposing sources as @xmath31 ( for example , @xmath26 from the galactic arm `` in '' and `` out '' directions , while @xmath27 is from perpendicular directions ) . in the first case ,
the coefficient is denoted @xmath32 , which is the amplitude of the first harmonic , and in the second case , the coefficient is denoted @xmath33 , which is the amplitude of the second harmonic . in the general case of harmonic expansion of an arbitrary distribution : @xmath34 if the observed distribution of arrival angles , @xmath35 , is given by a sum of delta functions @xmath36 , then @xmath37 where @xmath38 .
average values of the amplitudes are not zero even in the case of isotropic distribution , and converge to zero with @xmath39 ( see appendix a ) : @xmath40 the probability that isotropic amplitudes will be larger than the given @xmath41 is @xmath42 earth s rotation enables celestial regions to be represented as @xmath0 distributions .
scintillation counters of the yakutsk array have a 24-hour duty cycle , resulting in almost uniform directional exposure .
small deviations from the uniformity are caused by maintenance of the detectors ( mainly during working days ) and summertime shutdowns of the array due to thunderstorms .
diurnal and seasonal variations of the array exposure and atmospheric conditions result in spurious amplitude @xmath43 @xcite . with datasets of the size described in section 2
, we were able to determine only those amplitudes that were well above @xmath44 ( eq .
3 ) . therefore , in the energy region
@xmath23 eev , we neglected the effect of diurnal and seasonal variations of the array exposure in the analyzed dataset .
.yakutsk array observed amplitudes in energy bins , @xmath41 , and the probability that expected isotropic amplitudes will exceed observed amplitudes , @xmath45 .
@xmath46 . [ cols= " <
, > , > , > , > " , ] our alternative hypothesis , @xmath47 , has two parameters to fit the experimental data : the source position in @xmath0 , and the fraction of the total flux produced by the source luminosity .
we estimated the most probable @xmath47 parameters fitting the data from table 2 in the energy range @xmath48 eev .
the results showed that the ss position is @xmath49 , and the ratio of cr flux from ss to the total flux is @xmath50 .
we used two methods to examine the @xmath0 distribution of cr arrival directions measured with the yakutsk array : harmonic analysis and analysis of variance .
resultant first and second harmonic amplitudes increased with energy but were consistent with expected amplitudes of the isotropic distribution .
in the energy range @xmath51 eev , our data observed in the period 19742000 exhibited an excess flux with the first harmonic amplitude of @xmath52 , a chance probability of @xmath53 , but the second part of the data observed in 20002008 had the amplitude @xmath54 with a chance probability of @xmath55 . therefore , we found no significant deviation of the first and second harmonic amplitudes from those in the isotropic distribution , providing a hint of possible anisotropy above @xmath56 eev .
analysis of variance demonstrated the prominent excess in the same energy bin , namely , a significant contraction of the minimal width of the @xmath0 distribution of cr arrival directions with respect to the isotropic distribution . downsizing in width was found in both independent parts of the yakutsk array data , i.e. , data observed before and after 2000 .
the null hypothesis was rejected at the significance level @xmath57 . an alternative hypothesis with a single source in the uniform background flux of crs
was fitted to the observational data in the energy range @xmath48 eev with the source position @xmath3 and the ratio of cr flux from ss to the total flux @xmath50 .
the first harmonic phase did not appear to be randomly distributed in the interval @xmath58 , as would be expected in the isotropic case , but exhibited a gradual increase with energy in @xmath0 , at least in the energy interval above @xmath59 eev .
this behavior is inherent in both yakutsk observation periods , i.e. , before and after 2000 , and is in agreement with the possible regularity in the phase of the dipole anisotropy observed by the pao and ta collaborations .
we are grateful to the yakutsk array staff for the data acquisition and analysis .
the work is supported in part by the russian academy of sciences ( program 10.2 ) and rfbr ( grants 11 - 02 - 00158 , 13 - 02 - 12036 ) .
here , we apply the rayleigh formalism , founded in @xcite , to illustrate a calculation of the first harmonic amplitude , @xmath32 , for @xmath60 isotropic points in the interval @xmath61 . treating the amplitude as a sum of @xmath60 vectors of length @xmath62 with angle @xmath35 , we have the total length , @xmath63 , which accumulates @xmath60 equal random steps .
the inductive argument about a relationship between @xmath64 and @xmath65 in a triangle of vectors @xmath66 consists in @xmath67 due to @xmath68 .
consequently , @xmath69 .
asymptotically , the distribution of the amplitude is circular gaussian according to the central limit theorem : @xmath70 , where @xmath71 . the mean amplitude is @xmath72 while the mean square of amplitude is @xmath73 a comparison with the vector length gives the following : @xmath74 , the mean amplitude @xmath75 , and the variance @xmath76 .
the probability of obtaining an amplitude greater than or equal to @xmath77 is @xmath78
the fortran-90 program below illustrates the minimal width calculation for a distribution of @xmath60 random points in the right ascension ( ra ) circle . the function var computes the minimum variance with the trial mean scanning a circle @xmath79 .
the mean value and standard deviation of the distribution minimal width are calculated with a sample of size @xmath80 in the main program w. | conventional approach origin cosmic rays for anisotropy in arrival directions distribution attempts to find excess fluxes crs with large - scale structures universe possible anisotropic effects none confirmed independently examples in restricting data to ground - based arrays uniform exposure in right ascension obtained present paper data from scintillation counters surface detectors yakutsk array results of previous analysis extensive air shower january 1974 to may 2000 published in now analyzed extended dataset up to june 2008 additional data include 7598 eas events above threshold energy 1 eev ( aim to extend time series data observational time long testing indications of possible anisotropy in arrival directions crs in extended dataset methods yakutsk array site near oktyomtsy at coordinates @xmath4 altitude 105 m above sea level consists of 58 ground - based and 4 underground scintillation counters 48 detectors of air cherenkov light 40-year array reconfigured several times before 1990 total area covered by detectors was maximum ( @xmath5 km@xmath6 ) now @xmath7 km@xmath6 eas events selected two - level trigger first level two scintillation counters within 2 @xmath8s second level three stations within 40 @xmath8s stations spaced @xmath9 m m form trigger-500 and trigger-1000 @xmath11 showers above @xmath12 ev selected fig : map shows positions detectors yakutsk array core positions eass @xmath13 eev scintillation counters spacing 500 1000 m shown other types detectors described in refs. @xcite shower core coordinates located lateral distribution particle densities - function core location errors @xmath14 m for trigger-500 @xmath15 m for trigger-1000 arrival angles particles calculated detection times clock pulse transmitter pulse timing 100 ns accuracy errors depend primary energy decreasing from to above eev detailed information in refs . @xcite analyzed dataset eas events detected january 1974 june 2008 energy above @xmath20 eev zenith angles @xmath21 energy estimation based on total flux measurement air cherenkov light number electrons muons at observation levelunify energies showers years used same @xmath22-to - energy relationship attenuation length as ref . number eas events selected 43710 at @xmath23 eev energy estimation error approximately @xmath24 for showers axes within array area @xcite systematic error energy estimation yakutsk array corrected by specific factor @xcite fig ] shows distribution arrival directions showers equatorial system compared with expected isotropic distributions used rayleigh formalism harmonic analysis @xmath0 distribution cr arrival directions large - scale anisotropy present in arrival directions expressed using anisotropy coefficient for intensity variation : @xmath25 @xmath26 @xmath27 maximum minimum cr intensity function arrival angle @xmath28 simplest cases one source at @xmath29 distribution described by cosine wave @xmath30 or two opposing sources as @xmath31 galactic perpendicular first case coefficient denoted @xmath32 amplitude first harmonic second case @xmath33 amplitude second harmonicharmonic expansion arbitrary distribution @xmath34 observed distribution arrival angles @xmath35 given sum delta functions @xmath36 @xmath37 @xmath38 average values amplitudes not zero isotropic distribution converge to zero with @xmath39 appendix @xmath40 probability isotropic amplitudes larger @xmath41 @xmath42 earth s rotation enables celestial regions @xmath0 distributions scintillation counters yakutsk array 24-hour duty cycle uniform directional exposure small deviations caused maintenance summertime shutdowns thunderstorms diurnal seasonal variations array exposure atmospheric conditions result spurious amplitude @xmath43 datasets 2 amplitudes above @xmath44 3 energy region @xmath23 neglected diurnal seasonal variations array exposure.yakutsk array observed amplitudes energy bins @xmath41 probability expected isotropic amplitudes exceed observed amplitudes @xmath45 @xmath46 alternative hypothesis @xmath47 two parameters fit experimental data source position @xmath0 fraction total flux source luminosity estimated probable @xmath47 parameters fitting data table 2 energy range @xmath48 eevresults showed ss position @xmath49 ratio cr flux to total @xmath50 used two methods distribution cr arrival directions yakutsk array harmonic analysis variance first second harmonic amplitudes increased with energy consistent with isotropic distribution energy range @xmath51 eev data 19742000 excess flux first amplitude @xmath52 chance probability @xmath53 20002008 amplitude @xmath54 chance probability @xmath55 no significant deviation first second amplitudes from isotropic distribution hint possible anisotropy above @xmath56 eev analysis variance demonstrated excess energy bin significant contraction minimal width @xmath0 distribution isotropic downsizing in width found in yakutsk array data before 2000 null hypothesis rejected at significance level @xmath57 alternative hypothesis single source uniform background flux fitted data energy range @xmath48 eev source position @xmath3 ratio cr flux ss to total flux @xmath50 first harmonic phase randomly distributed interval @xmath58 gradual increase with energy @xmath0 above @xmath59 eev behavior inherent in both yakutsk observation periodsbefore after 2000 agreement with regularity dipole anisotropy pao ta collaborations grateful yakutsk array staff for data acquisition analysis supported by russian academy sciences program 10.2 rfbr grants 11 - 02 - 00158 13 - 02 - 12036 rayleigh formalism calculation first harmonic amplitude @xmath32 for @xmath60 isotropic points interval @xmath61 amplitude sum vectors length angle @xmath35 total length @xmath63 accumulates equal random steps relationship between @xmath64 @xmath65 triangle vectors @xmath66 @xmath67 @xmath68 distribution amplitude circular gaussian central limit theorem mean amplitude @xmath72 square @xmath73 comparison with vector length @xmath74 mean amplitude @xmath75 variance @xmath76 probability amplitude greater than equal to @xmath77 is @xmath78 fortran-90 program minimal width calculation distribution @xmath60 random points right ascension circle function var computes minimum variance with trial mean scanning circle @xmath79 mean value standard deviation calculated with sample @xmath80 | 0.354165 | 2,533 | 1,319 | , Saving $0.1 in GPT-4. |
21,038 | the temperature dependence of the resistivity of epitaxial ba(fe@xmath0co@xmath1)@xmath2as@xmath2 thin films ( with nominal doping @xmath3 and @xmath4 ) has been analyzed and compared with analogous measurements on single crystals taken from literature .
the @xmath5 of thin films looks different from that of single crystals , even when the cobalt content is the same .
all @xmath5 curves can be fitted by considering an effective two - band model ( with holes and electrons bands ) in which the electrons are more strongly coupled with the bosons ( spin fluctuations ) than holes , while the effect of impurities is mainly concentrated in the hole band . within this model
the mediating boson has the same characteristic energy in single crystals and thin films , but the shape of the transport spectral function at low energy has to be very different , leading to a `` hardening '' of the electron - boson spectral function in thin films , associated with the strain induced by the substrate . | in recent years a huge effort has been devoted to the investigation of all the physical aspects of iron pnictides .
these materials show very interesting properties such as the high superconducting transition temperature @xcite , the proximity of the magnetic and the superconducting states and the presence of complex band structures resulting in multi - band fermi surfaces @xcite . the electronic structure in general consists of hole pockets centered at the @xmath6 point and electron pockets at the corner of the brillouin zone .
in particular , it has been proved that for the so - called 122 " compounds of general formula aefe@xmath2as@xmath2 [ ae : alkali earth elements ] the size and even the dimension of these pockets drastically change with doping @xcite . due to
the variety and complexity of these compounds many questions about the symmetry of the order parameters , the coupling mechanism and the effects of the impurities are still open @xcite .
the situation becomes even more complex when a thin film is considered . in this case many differences with respect to the single crystals can appear also in the fundamental properties of the materials .
therefore a deeper investigation in this direction is indispensable @xcite .
nowadays very high - quality thin films are available and this allows using them for fundamental studies as well as for more technological applications @xcite .
in particular , co - doped bafe@xmath2as@xmath2 films of high quality can be readily fabricated by pulsed laser deposition ( pld ) and are thus one of the most suitable materials for these purposes .
the investigation of the effects of the substrate in thin films is thus a hot topic and a particular attention has been paid to the strain - dependent critical temperature @xcite . in this article
we present the study of the resistivity measured on ba(fe@xmath0co@xmath1)@xmath2as@xmath2 thin films grown on caf@xmath2 substrates and with different nominal cobalt content ( namely 8% , 10% and 15% ) in comparison with analogous measurements on single crystals taken from literature @xcite .
we use a model containing two different kinds of carriers ( holes and electrons ) as suggested by the presence of several sheets in the fermi surface and by the tendency of the @xmath5 curves to saturate at high temperature , as already observed in other iron - based compounds @xcite . on the basis of experimental evidences , we assume that the temperature dependence of the resistivity is dominated by the coupling between electrons and spin fluctuations ( sf ) .
this simple model allows us to fit very well the resistivity of single crystals by adjusting the free parameters ( in particular the characteristic boson frequency @xmath7 ) in agreement with the results obtained by inelastic neutron scattering and optical measurements @xcite . in the case of thin films ,
the @xmath5 curves can be fitted with a set of parameters that agree with the experimental data available in literature only if the electron - sf spectrum is depleted at low energy , resulting in a transfer of spectral weight to higher energies .
thanks to the interaction between phonons and spin fluctuations , this effect can be ascribed to the phonon hardening induced by the reduction of the unit - cell volume @xcite caused by the presence of the caf@xmath2 substrate @xcite .
the ba(fe@xmath0co@xmath1)@xmath2as@xmath2 ( @xmath8 ) epitaxial thin films with a thickness of the order of 50 nm were deposited on ( 001)caf@xmath2 substrates by pulsed laser deposition ( pld ) @xcite using a polycrystalline target with high phase purity @xcite .
the surface smoothness was confirmed by in - situ reflection high energy electron diffraction ( rheed ) during the deposition ; only streaky pattern were observed for all films , indicative of a smooth surface .
the details of the structural characterization and of the microstructure of these high - quality epitaxial thin films can be found in ref .
standard four - probe resistance measurements in van der pauw configuration were performed in a @xmath9he cryostat to determine the @xmath5 curve as well as the transport critical temperature and the width of the superconducting transition @xcite ( both reported in table [ tab : exp ] ) .
.critical temperatures of thin films ( tf ) @xcite and single crystals ( sc ) @xcite determined from electric transport measurements . @xmath10 and @xmath11 are the temperatures at which the resistance ( the resistivity ) is respectively 10% and 90% of the normal - state value immediately before the transition .
@xmath12 is defined here as @xmath13 . [
cols="^,^,^,^",options="header " , ]
it is already known that , in thin films of iron - based compounds , the transport critical temperature is affected by the strain due to the substrate @xcite .
here we have shown that this effect may also explain the different shape of the @xmath5 curve in crystals and thin films of co - doped ba-122 , even when the co content is the same . by using a simple effective two - band model for the transport
, we have shown that the @xmath5 curves of thin films can be fitted with the same characteristic boson energy @xmath7 used for single crystals only if some spectral weight is transferred from below @xmath7 to above @xmath7 in the transport electron - boson spectral function @xmath14 .
in other words , @xmath14 behaves like @xmath15 in a wider energy range than in single crystals .
this effect is possibly induced by the indirect coupling between spin fluctuations and phonons @xcite , whose frequencies are increased by the compressive strain @xcite . finally , we have shown that the decrease ( increase ) in the electron - boson coupling constant ( characteristic boson energy ) between the superconducting and the normal state ( @xmath16 , @xmath17 ) observed in crystals of various fe - based compounds also holds in thin films of co - doped ba-122 .
this behavior is also common to htsc @xcite and may thus be a unifying principle at the root of superconductivity in iron - based materials and cuprates .
the authors would like to thank f. galanti for his cooperation in resistivity measurements .
this work was done under the collaborative eu - japan project `` iron - sea '' ( nmp3-sl-2011 - 283141 ) . | recent years effort devoted to investigation physical aspects of iron pnictides materials show properties high superconducting transition temperature proximity of magnetic superconducting states complex band structures multi - band fermi surfaces electronic structure consists of hole pockets at @xmath6 point electron pockets at brillouin zone proved for 122 " compounds of formula size dimension of pockets change with doping variety complexity questions about symmetry order parameters coupling mechanism effects of impurities open situation more complex when thin film considered differences single crystals appear in fundamental properties materials deeper investigation indispensable high - quality thin films available for fundamental studies technological applications co - doped bafe@xmath2as@xmath2 films fabricated by pulsed laser deposition suitable for purposes investigation effects substrate in thin films hot topic attention paid to strain - dependent critical temperature article present study of resistivity measured on ba(fe@xmath0co@xmath2as@xmath2 thin films grown on caf@xmath2 substrates with different nominal cobalt content 8% 10% 15% ) comparison with analogous measurements on single crystalsuse model containing two carriers ( holes electrons suggested by sheets in fermi surface tendency @xmath5 curves to saturate at high temperature observed in iron - based compounds temperature dependence resistivity dominated by coupling between electrons spin fluctuations model resistivity of single crystals adjusting free parameters boson frequency @xmath7 with inelastic neutron scattering optical measurements thin films @xmath5 curves fitted with parameters data if electron - sf spectrum depleted at low energy transfer spectral weight to higher energies interaction between spin fluctuations effect ascribed to phonon hardening reduction unit - cell volume caf@xmath2 substrate epitaxial thin films thickness 50 nm deposited on)caf@xmath2 substrates by pulsed laser deposition using polycrystalline target high phase purity surface smoothness confirmed by in - situ reflection high energy electron diffraction during deposition streaky pattern observed for all films smooth surface details structural characterization microstructure of high - quality epitaxial thin films in ref .four - probe resistance measurements van der pauw configuration performed @xmath9he cryostat @xmath5 curve transport critical temperature width superconducting transition @xcite reported in table temperatures of thin films single crystals determined from electric transport measurements @xmath10 @xmath11 temperatures resistance 10% 90% of normal - state value before transition @xmath12 defined as @xmath13 in thin films iron - based compounds transport critical temperature affected by strain substrate @xcite different shape @xmath5 curve in crystals thin films of co - doped ba-122 - model @xmath5 curves films same characteristic boson energy @xmath7 if spectral weight transferred from below to above @xmath7 transport electron - boson spectral function @xmath14 @xmath14 behaves like @xmath15 wider energy range single crystals induced by indirect coupling between spin fluctuations phonons @xcite frequencies increased by compressive strain decrease in electron - boson coupling constant between superconducting normal state in crystals fe - based compounds holds in thin films co - doped ba-122behavior common htsc @xcite unifying principle superconductivity iron - based materials cuprates authors thank f. galanti cooperation resistivity measurements work eu - japan project iron - sea - 283141 | 0.258148 | 1,402 | 671 | , Saving $0.0 in GPT-4. |
21,053 | many planet candidates have been detected by radial velocity variations of the primary star ; they are planet _ candidates _ , because of the unknown orbit inclination .
detection of the wobble in the two other dimensions , to be measured by astrometry , would yield the inclination and , hence , true mass of the companions .
we aim to show that planets can be confirmed or discovered in a close visual stellar binary system by measuring the astrometric wobble of the exoplanet host star as periodic variation of the separation , even from the ground .
we test the feasibility with hd 19994 , a visual binary with one radial velocity planet candidate .
we use the adaptive optics camera naco at the vlt with its smallest pixel scale ( @xmath0 mas ) for high precision astrometric measurements .
the separations measured in single 120 images taken within one night are shown to follow white noise , so that the standard deviation can be devided by the square root of the number of images to obtain the precision . in this paper
we present the first results and investigate the achievable precision in relative astrometry with adaptive optics . with a careful data reduction it is possible to achieve a relative astrometric precision as low as 50 @xmath1as for a 0.6 arc sec , binary with vlt / naco observations in one hour , the best relative astrometric precision ever achieved with a single telescope from the ground .
the relative astrometric precision demonstrated here with ao at an 8-m mirror is sufficient to detect the astrometric signal of the planet hd 19994 ab as periodic variation of the separation between hd 19994 a and b. | since the radial velocity technique can yield only lower mass limits , all planets ( or planet candidates ) found by this technique have to be confirmed by other methods .
of @xmath2 radial velocity planet candidates found so far , 14 have been confirmed by transit and two by astrometry , see e.g. exoplanet.eu .
the two planets gj 876 b and 55 cancri d have been confirmed by astrometry using the hubble space telescope fine guidance sensor ( benedict et al .
2002 , mcarthur et al .
2004 ) , with a precision down to 0.04 mas ( milli arc sec ) sufficient to detect the wobble of the host star in the plane of the sky .
the astrometric displacement is given by @xmath3 = 0.960 \times \frac{a}{[5\,au ] } \times \frac{[10\,pc]}{d } \times \frac{m_{pl}}{[m_{jup } ] } \times \frac{[m_{\odot}]}{m_{\star}}\ ] ] with planet mass @xmath4 in jupiter masses at a semi - major axis @xmath5 ( in units of 5 au ) in a circular orbit around a host star with mass @xmath6 in solar masses at a distance @xmath7 ( in units of 10 pc ) . here , we present both a feasibility study to determine the astrometric precision of the adaptive optics ( ao ) camera naos - conica ( naco ) of the eso very large telescope ( vlt ) as well as first measurements in order to determine the mass of hd 19994 ab , a rv planet candidate in orbit around hd 19994 : the rv data suggest a lower mass limit of the planet candidate of @xmath8 in an orbit with semi - major axis @xmath9 au with an eccentricity of @xmath10 , i.e. a @xmath11 day orbital period ( mayor et al . 2004 ) .
the full ( peak - to - peak ) astrometric wobble , from kepler s 3rd law , is at least 0.155 mas ( for a circular orbit ) and 0.131 mas by taking the observed eccentricity into account this is when assuming the minimum mass @xmath12 as true mass .
we observed hd 19994 and a calibration binary ( hd 19063 ) with the s13 camera ( 13.26 mas / pixel pixel scale , 14@xmath13 @xmath14 14@xmath13 fov ) in a nb_2.17 narrow band filter .
short exposures of both target binaries where taken for one hour each , equally splitted up into two time slots before and after meridian passage .
we obtained a total of 120 frames for hd 19994 and 117 frames for hd 19063 .
data reduction followed the standard technique of dark subtraction , division by a flat field and the application of a bad pixel mask with the _ eclipse _ software . in each single exposure
( 0.347 sec ) , we measured the separation between a and b of both binaries .
we have then made sure that the separations follow white noise by a k - s test ( see a future paper by neuhuser et al . for details ) .
hence , we can devide the standard deviation of the mean by the square root of the number of measurements to obtain the precision of the separation measurement , the _ precision _ , not the _ accuracy_. for the latter , we would need to know the true pixel scale .
see figure 1 and table 1 .
error for individual measurements .
the standard deviation of the complete dataset is shown as dashed lines and grey - shaded areas for 1 , 2 , and 3@xmath15 , respectively .
the standard deviation of the mean , thus the precision of the measured binary separation , is @xmath16.,width=464 ] .astrometric results from the measurements of the binaries hd 19994 and hd 19063 .
k - s test values higher than 0.6 are interpreted as indicating white noise .
[ cols="^,^,^,^,^ , > " , ] differential chromatic refraction ( drc ) or allan noise ( see pravdo & shaklan 1996 ) do not matter here , because we observe in the near ir with a narrow - band filter ( hence no dcr ) and with ao ( hence no allan noise ) .
we have observed two bright visual binaries ( 0.6 to 2.4 arc sec separation ) with vlt / nacos smallest pixel scale ( 13.26 mas / pixel ) by taking @xmath17 short ( 0.347 s ) exposures per binary within a few hours , separated in four bins of @xmath18 exposures each ( 15 december 2004 , eso program 075.c-0288.a ) .
we could confirm that the separations measured follow white noise ( in three out of four bins ) , so that we are allowed to divide the mean of the separations by the square root of the number of measurements to obtain the precision of the separation measurement . by doing so , we measured the separation between hd 19994 a and b to be @xmath19 mas ( milli arc seconds ) with the relative precision of @xmath20 ( micro arc seconds ) . for hd 19063
, we obtained @xmath21 mas with a precision being as low as @xmath22 .
these are the most precise measurements in relative astrometry obtained with single - aperture telescopes from the ground , sufficient to detect the astrometric wobble due to a planet , to be seen as periodic change in the binary separation .
such a high precision as shown here can be applied not only to measurements of masses of previously detected rv planets ( or candidates ) , but also to search for new planets in wide binaries , or even single stars with a bright ( back- or foreground ) star within the isoplanatic angle .
other possible applications are orbit determinations in stellar multiples , ground - based parallaxe and proper motion measurements , observations of expanding , contracting , or rotating star clusters , etc .
the main limitation of this technique is the need for a bright star near the target , namely within the isoplanatic angle . and
, as long as no pixel scale calibrator is available with a precision down to 1/100000 , the measurement can achieve this high precision only for relative astrometry , but not in absolute terms . | radial velocity technique lower mass limits all planets methods @xmath2 radial velocity planet candidates 14 confirmed by transit two by astrometry exoplanet.eu two planets gj 876 b and 55 cancri d confirmed by astrometry using hubble space telescope sensor benedict 2002 mcarthur 2004 precision to 0.04 mas detect wobble host star astrometric displacement by @xmath3 = 0.960 planet mass @xmath4 in jupiter masses semi - major axis @xmath5 5 au circular orbit around host star mass @xmath6 in solar masses distance @xmath7 ( units of 10 pc ) feasibility study astrometric precision of adaptive optics camera - eso very large telescope first measurements mass of hd 19994 ab rv planet candidate orbit around hd 19994 data suggest lower mass limit of planet candidate of @xmath8 orbit with semi - major axis @xmath9 au eccentricity of @xmath10 @xmath11 day orbital period ( mayor 2004full peak astrometric wobble kepler s 3rd law 0.155 mas circular orbit 0.131 mas observed eccentricity assuming minimum mass @xmath12 true mass observed hd 19994 calibration binary hd 19063 s13 camera 13.26 mas / pixel scale 14@xmath13 nb_2.17 narrow band filter short exposures binaries one hour each splitted two time slots before after meridian passage 120 frames hd 19994 117 frames hd 19063 data reduction dark subtraction division flat field bad pixel mask eclipse software exposure 0.347 sec measured separation between a b binaries separations follow white noise by k - s test devide standard deviation mean by square root number measurements precision separation measurement not accuracy true pixel scale figure 1 table 1 error individual measurements standard deviation dataset shown as dashed lines grey - shaded areas for 1 2 3@xmath15 standard deviation mean precision measured binary separation is @xmath16.,width=464.astrometric results from measurements binaries hd 19994 hd 19063 k - s test values higher than 0.6 white noisedifferential chromatic refraction allan noise pravdo shaklan 1996 matter observe near narrow - band filter no dcr ao no allan noise observed two bright visual binaries 0.6 to 2.4 arc sec separation vlt / nacos smallest pixel scale 13.26 mas / pixel @xmath17 0.347 s exposures per binary four bins @xmath18 exposures 15 december 2004 eso program 075.c-0288.a separations follow white noise three out of four bins divide mean separations by square root number measurements precision separation measured separation between hd 19994 a and b @xmath19 mas precision @xmath20 micro arc seconds hd 19063 obtained @xmath21 mas precision low as @xmath22 most precise measurements relative astrometry single - aperture telescopes detect astrometric wobble due planet periodic change binary separation high precision applied measurements masses rv planets search for new planets wide binaries single stars bright star isoplanatic angle applications orbit determinations stellar multiples ground - parallaxe proper motion measurements observations expanding contracting rotating star clusterslimitation technique need bright star near target isoplanatic angle no pixel scale calibrator available precision down to 1/100000 measurement achieve high precision for relative astrometry not absolute | 0.256192 | 1,392 | 714 | , Saving $0.0 in GPT-4. |
21,058 | we update briefly our understanding of hadron production in relativistic nucleus - nucleus collisions in terms of statistical models with emphasis on the relation of the data to the qcd phase boundary and on a puzzle in the beam energy dependence . | hadron yields or more specifically yield ratios observed in central nuclear collisions at ags , sps , and rhic energies can be described with high precision within a hadro - chemical equilibrium approach @xcite , governed by a chemical freeze - out temperature t@xmath0 and baryo - chemical potential @xmath1 .
a recent review , found in @xcite , provides a wealth of information on the subject .
the data at sps and rhic energy comprise multi - strange hadrons including the @xmath2 , @xmath3 , @xmath4 and @xmath5 baryons .
their yields ratios ( to pions , e.g. ) agree particularly well with the chemical equilibrium calculation and are enhanced by more than an order of magnitude as compared to observations in pp collisions .
we present as an example the results for central au - au collisions at rhic energy in fig .
[ yields ] .
a remarkable exception is observed for strongly decaying , wide resonances . in particular , the @xmath6 ratio is overpredicted while the @xmath7 ratio , both determined in ( rather peripheral , though ) au - au collisions by the star collaboration @xcite , is underpredicted by the model @xcite . for an interesting and provocative speculation as to where this might come from see @xcite .
clearly precision data for central collisions and also at other beam energies are needed to settle this issue .
the chemical parameters t@xmath0 and @xmath1 determined from fits to data at all available energies are plotted in the qcd phase diagram shown in fig .
[ phase ] , taken ( slightly modified ) from @xcite .
the phase transition lines in this figure are obtained from recent analyses within the framework of lattice qcd @xcite and include recent estimates @xcite of the position of the tri - critical point . also included in this figure
are lines of constant total baryon density and lines of constant energy density , computed within the framework of the hadronic gas model of @xcite .
the line of constant energy density @xmath8 mev/@xmath9 should in our view be a reasonable phenomenological description of the phase boundary .
interestingly , it is , at @xmath10 mev , much closer to the chemical freeze - out points than to the phase lines from lattice qcd .
apparently , the phase boundary lines obtained from state - of - the - art calculations within the framework of lattice qcd are not at all lines of constant energy density .
rather , the energy density grows by more than a factor of 2 when going from 0 to 500 mev in @xmath1 .
this rather surprizing result deserves further attention .
an important observation , already made in@xcite , concerning the results of the thermal model calculations is that , for top sps energy and above , the chemical parameters determined from the measured hadron yields coincide within the uncertainties of about @xmath11 mev with the phase boundary as determined from lattice qcd calculations .
a natural question arises : is this coincidence accidental and , if not , what enforces equilibration at the phase boundary ? considerations about collisional rates and timescales of the hadronic fireball expansion @xcite imply that at sps and rhic energies the equilibrium can not be established in a purely hadronic medium and that is is the phase transition which drives the particles densities and ensures chemical equilibrium .
in @xcite it is further shown that many body collisions become important close to t@xmath12 and provide a mechanism for rapid equilibration .
because of the strong density increase near the qcd phase transition such multi - particle collisions provide a natural explanation for the observation of chemical equilibration at rhic energies and lead to t@xmath0=t@xmath12 within an accuracy of a few mev . any scenario with t@xmath0 substantially smaller than t@xmath12 would require that either multi - particle interactions dominate even much below t@xmath12 or that the two - particle cross sections are larger than in the vacuum by a high factor .
both of the latter hypotheses seem unlikely in view of the rapid density decrease in the hadronic phase of the expanding fireball .
the critical temperature determined from rhic for t@xmath13 coincides well with recent lattice estimates @xcite for @xmath14 .
the same arguments as discussed here for rhic energy also hold for sps energies : it is likely that also there the phase transition drives the particle densities and insures chemical equilibration .
it was alternatively proposed @xcite that the observed hadron abundances arise from a direct production of strange ( and non - strange ) particles by hadronization .
how this happens microscopically is unclear . to escape the above argument that t@xmath0=t@xmath12 one would have to argue that the particle yields are established without hadronic rescattering
this is unlikely since the abundances are determined by hadronic properties ( masses ) with high precision .
second , one may question if the `` chemical temperature '' extracted from the abundances is a universal temperature which also governs the local kinetic aspects and can be associated with the critical temperature of a phase transition in equilibrium .
indeed , in a prethermalization process , different equilibrium properties are realized at different time scales .
nevertheless , all experience shows that kinetic equilibration occurs before chemical equilibration .
it seems hard to imagine that chemical equilibrium abundances are realized at a time when the kinematic distributions are not yet close to their equilibrium values .
the chemical equilibrium curve and the qcd phase boundary as obtained from lattice qcd calculations begin to differ for baryo - chemical potential values of @xmath15 400 mev . despite
that multi - strange baryons are produced with yields very close to what is obtained in a full chemical equilibration scenario ( see below ) .
this could be an indication for new , as yet unidentified equilibration processes , different from those advocated in @xcite and also incompatible with the `` phase space filling '' scenario through hadronization put forward by @xcite , in a dense , baryon - rich medium .
considering the discussion above , a rather radical but appealing possibility is that the chemical freeze - out curve defines the phase boundary also at large @xmath1 values up to @xmath16 mev .
we note in this context that the chemical freeze - out points determined from measurements at sis energy should still be considered with caution as no multi - strange baryons have been measured there .
the relatively smooth energy dependence of the above discussed thermal parameters is to be contrasted with recent observation by the na49 collaboration of an anomaly in the k@xmath17 ratio near @xmath18 gev @xcite .
these results look rather striking , and indeed , the narrowness of the observed structure is missed by the thermal model predictions , as is visible in fig .
[ k2pi ] .
we note , however , that the anomaly seems to be confined to ratios involving light mesons . for strange baryons ,
the situation is quite different , as is demonstrated in fig .
[ xiom ] . here , the observed energy dependence is very well reproduced by the calculations and no anomaly is visible .
the situation is even more puzzling considering that the energy averaging inherent in ultra - relativistic nucleus - nucleus collisions leads to a characteristic smearing of all possible intrinsic structures .
the effect can be estimated @xcite using simple geometrical arguments and is of the order 10 % of the cm .
energy per nucleon at sps energy .
the corresponding width is very close to the width of the structure observed by the na49 collaboration and would imply that the strength of the intrinsic structure is even much larger than visible in fig .
further research is necessary to shed light on this puzzle .
99 p. braun - munzinger , j. stachel , j.p .
wessels and n. xu , phys . lett . *
b344 * ( 1995 ) 43 and * b365 * ( 1996 ) 1 .
j. cleymans , d. elliott , h. satz , and r.l .
thews , z. phys . *
c74 * ( 1997 ) 319 .
braun - munzinger , i. heppe and j. stachel , phys . lett .
* b465 * ( 1999 ) 15 .
j. cleymans and k. redlich , phys . rev .
* c60 * ( 1999 ) 054908 .
f. becattini , j. cleymans , a. keranen , e. suhonen , and k. redlich , phys . rev .
* c64 * ( 2001 ) 024901 .
p. braun - munzinger , d. magestro , k. redlich , and j. stachel , phys . lett .
* b518 * ( 2001 ) 41 .
n. xu and m. kaneta , nucl .
* a698 * ( 2002 ) 306c .
f. becattini , j. phys .
* g28 * ( 2002 ) 1553 .
r. rapp and e. shuryak , phys .
* 86 * ( 2001 ) 2980 .
p. braun - munzinger , k. redlich , and j. stachel , nucl - th/0304013 , invited review in quark gluon plasma 3 , eds .
r.c . hwa and x.n .
wang , ( world scientific publishing , 2004 ) .
p. fachini et al . , star collaboration , nucl - ex/0403026 , j. phys .
* g30 * ( 2004 ) s735 .
j. adams et al . ,
star collaboration , phys .
* 92 * ( 2004 ) 092301 .
brown , chang - hwan lee , and m. rho , nucl - th/0507073 .
a. andronic and p. braun - munzinger , hep - ph/0402291 .
s. ejiri et al . , hep - lat/0312006 .
z. fodor and s. katz , jhep * 0404 * ( 2004 ) 050 .
o. philipsen , proc .
lattice05 conf . ,
dublin , ireland , july 2005 , to be published .
see also http://www.maths.tcd.ie/lat05/plenary-talks/philipsen.pdf .
p. braun - munzinger , j. stachel , c. wetterich , phys
* b596 * ( 2004 ) 61 .
f. karsch , e. laermann , a. peikert , nucl . phys .
* a605 * ( 2001 ) 579 .
r. stock , phys .
b465 * ( 1999 ) 277 .
u. heinz , nucl . phys .
* a638 * ( 1998 ) 357 .
u. heinz , nucl .
* a685 * ( 2001 ) 414 .
j. adams et al . ,
star coll .
* 92 * ( 2004 ) 182301 , nucl - ex/0307024 .
f. becattini , z. physik * c69 * ( 1996 ) 485 , f. becattini and u. heinz , z. physik * c76 * ( 1997 ) 269 .
afanasiev et al . , na49 collaboration , phys . rev .
* c66 * ( 2002 ) 054902 . c. blume et al .
, na49 collaboration , j. phys .
* g31 * ( 2005 ) s685 .
a. andronic , p. braun - munzinger , j. stachel , manuscript in preparation .
p. braun - munzinger , d. miskowiec , j. stachel , manuscript in preparation . | hadron yields ratios in central nuclear collisions at ags sps rhic energies described hadro - chemical equilibrium approach @xcite governed by chemical freeze - out temperature t@xmath0 baryo - chemical potential @xmath1 recent review in @xcite provides information data at sps rhic energy multi - strange hadrons including @xmath2 @xmath5 baryons yields ratios agree with chemical equilibrium calculation enhanced compared to pp collisions results for central au - au collisions at rhic energy in fig . [ exception for strongly decaying wide resonances @xmath6 ratio overpredicted @xmath7 ratio underpredicted by model @xcite speculation see @xcite precision data for central collisions other beam energies needed chemical parameters t@xmath0 and @xmath1 from energies plotted in qcd phase diagram in fig . [ phase ] @xcite phase transition lines from recent analyses lattice qcd @xcite include estimates of position tri - critical point constant total baryon density constant energy density hadronic gas model of @xciteline constant energy density @xmath8/@xmath9 description phase boundary at @xmath10 mev closer to chemical freeze - out points than phase lines from lattice qcd phase boundary lines from not lines constant energy density energy density grows more than factor 2 from 0 to 500 mev in @xmath1 surprizing result deserves attention observation for top sps energy above chemical parameters hadron yields coincide @xmath11 mev with phase boundary lattice qcd calculations coincidence accidental what enforces equilibration at phase boundary ? collisional rates timescales hadronic fireball expansion imply at sps and rhic energies equilibrium in purely hadronic medium phase transition drives particles densities ensures chemical equilibrium body collisions important close to t@xmath12 for rapid equilibration strong density increase near qcd phase transition multi - particle collisions provide explanation for chemical equilibration at rhic energies lead to t@xmath0=t@xmath12 accuracy few mevscenario with t@xmath0 smaller than@xmath12 multi - particle interactions below or two - particle cross sections larger than vacuum hypotheses unlikely rapid density decrease in hadronic phase expanding fireball critical temperature for t@xmath13 coincides with lattice estimates for @xmath14 arguments for rhic energy hold for sps energies phase transition drives particle densities insures chemical equilibration hadron abundances arise from production particles by hadronization unclear t@xmath0=t@xmath12 particle yields established without hadronic rescattering unlikely abundances determined by hadronic properties with precision question if chemical temperature from abundances is universal temperature local kinetic aspects associated with critical temperature phase transition prethermalization different equilibrium properties realized at different time scales shows kinetic equilibration occurs before chemical equilibration hard to imagine chemical equilibrium abundances realized kinematic distributions not close to equilibrium values chemical equilibrium curve qcd phase boundary differ for baryo - chemical potential values of @xmath15 400 mev multi - strange baryons produced with yields close to full chemical equilibration scenarioindication for new equilibration processes different from @xcite incompatible with phase space filling scenario through hadronization in dense baryon - rich medium possibility chemical freeze - out curve defines phase boundary at @xmath1 values up to @xmath16 mev chemical freeze - out points from at sis energy caution no multi - strange baryons measured smooth energy dependence of parameters contrasted with anomaly in k@xmath17 ratio near @xmath18 gev @xcite results striking narrowness of structure missed by thermal model predictions visible in fig . [ k2pi anomaly confined to ratios light mesons for strange baryons different in fig . [ xiom ] observed energy dependence reproduced no anomaly visible energy averaging in ultra - relativistic nucleus - nucleus collisions leads to smearing of intrinsic structures effect estimated 10 % of cm . energy per nucleon at sps energy width close to structure observed by na49 collaboration strength of intrinsic structure larger than fig . further research necessary 99. braun - munzinger , j. stachel , j.wessels n xu phys b344 1995 ) 43 b365 1996 ) cleymans elliott satz thews phys c74 1997 ) 319 braun - munzinger heppe j stachel phys b465 1999 ) 15 j cleymans k redlich c60 1999 ) 054908 becattini cleymans keranen suhonen k redlich c64 2001 ) 024901 braun - munzinger d magestro k redlich j stachel phys b518 2001 ) 41 xu m kaneta nucl a698 2002 ) 306c becattini g28 2002 ) 1553 rapp e shuryak phys 86 2001 ) 2980 braun - munzinger redlich j stachel nucl th/0304013 invited review quark gluon plasma 3 hwa wang world scientific publishing 2004 fachini star collaboration nucl ex/0403026 phys g30 2004 s735 adams star collaboration phys 92 2004 ) 092301 brown chang - hwan lee m rho nucl - th/0507073andronic braun - munzinger hep - ph/0402291 ejiri hep - lat/0312006 fodor katz 0404 2004 ) 050 philipsen lattice05 conf dublin ireland july 2005 published.tcd.ie/lat05-talks/philipsen braun - munzinger stachel wetterich b596 2004 ) 61 karsch laermann peikert nucl a605 2001 ) 579 stock phys b465 1999 ) 277 heinz nucl phys a638 1998 ) 357 heinz nucl a685 2001 ) 414 adams star coll 92 2004 ) 182301 nucl - ex/0307024 becattini c69 1996 ) 485 becattini heinz physik c76 1997 ) 269 afanasiev na49 collaboration . c66 2002 ) 054902 blume al na49 collaboration phys. g31 2005 ) s685 andronic braun - munzinger stachel manuscript preparation braun - munzinger miskowiec stachel manuscript preparation | 0.354028 | 2,587 | 1,341 | , Saving $0.1 in GPT-4. |
21,065 | a predictive control scheme for a permanent - magnet synchronous machine ( pmsm ) is presented .
it is based on a suboptimal method for computationally efficient trajectory generation based on continuous parameterization and linear programming .
the torque controller optimizes a quadratic cost consisting of control error and machine losses in real - time respecting voltage and current limitations .
the multivariable controller decouples the two current components and exploits cross - coupling effects in the long - range constrained predictive control strategy .
the optimization results in fast and smooth torque dynamics while inherently using field - weakening to improve the power efficiency and the current dynamics in high speed operation .
the performance of the scheme is demonstrated by experimental results . | the efforts of implementing predictive controllers in electrical drives aim at replacing the classical cascaded field - oriented control structure with pi controllers .
the machine can be better exploited by improved control behavior , the system variables are optimized . in this contribution , the conventional torque and current control structure of two separate controllers is changed to multi - input multi - output ( mimo ) control . by transformation into the field - oriented frame ,
torque generation is decoupled from flux variation , however , the current controllers are still strongly coupled , therefore a mimo controller is advantageous .
the obtained improvements are better decoupling , better current and voltage constraint handling by exploiting cross - coupling between the orthogonal components , and better power efficiency and dynamics by optimally adjusting the currents in both dynamic and steady - state operation .
the major obstacle in implementing predictive control schemes is the limited computational power , inherited by the high sampling rates .
existing implementations suffer from this restriction and can not exploit the full advantages of model predictive control ( mpc ) .
for instance , generalized predictive control ( gpc ) has a high optimization horizon but is unconstrained , whereas predictive torque control ( ptc ) is constrained but so far only reaches @xmath0 steps of prediction @xcite . for good performance , both , inclusion of constraints and a high optimization horizon are required . using control in the field - oriented frame
, the analytical problem description enables using mathematical optimization algorithms .
such schemes are computationally efficient and maximize the obtained information for a given computational power .
the online solution of the linearly constrained linear - quadratic problem , typical for mpc , requires quadratic programming ( qp ) algorithms , which are , however , computationally too expensive for drive systems .
a recent development is the use of explicit mpc , where an offline solution is computed and stored as look - up table in the real - time controller @xcite @xcite .
the scheme reaches @xmath1 prediction steps with constraints .
while several fast online algorithms were recently proposed @xcite @xcite , most of them based on gradient search algorithms , the interest in online solutions is growing .
advantages of online optimization are manipulation and adaptation of parameters , resulting in a more flexible choice of machines for the control system .
this implementation is based on a suboptimal trajectory generation algorithm presented in @xcite , embedded in a flatness - based predictive control scheme @xcite .
it is based on a continuous approach , the variables are not discretized but represented as a polynomial with undetermined coefficients . with this method
, higher optimization horizons can be reached with comparably few parameters , considerably reducing the computational burden .
constraints are handled by linearization of the cost functional and the use of a linear programming ( lp ) solver , which is amongst the smallest and fastest numerical optimizers .
as result , a ( suboptimal ) prediction of @xmath0 ms with current and voltage constraints is obtained at @xmath2 khz sampling rate .
as linear - quadratic optimization problems with linear constraints are simpler to solve in real - time , the machine model is linearized . assuming that the rotor speed does not change too much over the optimization horizon @xmath3 , @xmath4 , \end{aligned}\ ] ] the pmsm model and the voltage equations become linear .
the electrical subsystem of the machine , consisting of the quadrature and direct currents @xmath5 and @xmath6 ( peak values ) , is given as @xmath7 the nomenclature is shown in table ii in the appendix .
it is noted that the reluctance torque @xmath8 is neglected , as this term is very small compared to the electromagnetic torque in surface - mounted pmsms or in machines with small saliency .
furthermore , it would render the model nonlinear , requiring nonlinear optimization methods @xcite .
the formulation of a suitable cost functional is a key point in predictive control , as it is the only tuning possibility of the control scheme .
the optimization is aiming at minimizing the control error for good dynamical performance as well as machine losses for better efficiency .
both goals are included in the cost functional . by choosing the cost functional and weights well , it is possible to find a good trade - off between both goals during transients , or eventually to fulfill both goals in steady - state .
the cost functional for the predictive torque controller is @xmath9 which trades off the squared control error @xmath10 with machine losses @xmath11 the first term in @xmath12 represents copper losses , and the second term represents the iron losses consisting of hysteresis losses .
eddy current losses are negligible on the tested machines , however , they could be included using the model presented in @xcite .
iron losses can be reduced by field - weakening , where a trade - off between copper and iron losses is found @xcite . imposing a negative direct current @xmath6 , the flux magnitude in the stator
is reduced while the copper losses increase .
as the iron loss constant @xmath13 is not part of standard motor parameters , it has to be determined experimentally @xcite .
the last term in @xmath14 is the end - weight of the control error and aims at reducing the steady - state control error .
the weight @xmath15 was set @xmath16 , the value was determined heuristically
. the optimization horizon is set @xmath17 ms such that the cost functional includes the complete setpoint change .
it is important that the optimization horizon is high enough , otherwise the open - loop and closed - loop trajectories differ and the behavior is strongly suboptimal .
this is illustrated in fig .
[ fig : horizon ] .
if the horizon is too small , due to the end - weight of the control error , a significant difference between the optimized open - loop trajectories , and the closed - loop trajectories resulting from regeneration at every sampling step , appears .
then , the closed - loop trajectories simply do nt fit the cost functional anymore . for a horizon higher than required for the setpoint change , the difference between open- and closed - loop trajectories becomes smaller .
, width=491 ] the most important nonlinearities of a pmsm , in terms of control , are the voltage and current limitations .
the current constraints prevent overheating of the machine , and the voltage is limited by the maximum dc - link voltage of the inverter .
the voltage constraints limit rotor speed as well as current dynamics in high - speed operation .
these constraints are linearized , in order to be computationally efficiently treated .
the current range for the direct current @xmath6 is limited to @xmath18 .
only negative values of @xmath6 are desirable , as they improve power efficiency and reduce the induced voltage by weakening the flux magnitude in the stator @xcite @xcite .
the lowest value @xmath19 is the optimum value at rated speed ( @xmath20 ) and is given as @xmath21 which is independent of quatrature current @xmath5 as the reluctance torque was neglected .
the value is doubled to enable further field - weakening to improve dynamics in high speed , an effect described in the experimental results section . for the quadrature current @xmath5 ,
the largest possible range of values should be available .
the resulting linear constraints , shown on fig .
[ fig : constraints ] , almost completely fill the current region of interest .
a linearization is thus acceptable .
the voltage linearization is a little bit more difficult .
the @xmath22-axis should not be restricted , as the induced voltage is aligned to it and is the largest value that will appear .
a steady - state analysis of the system equations ( 2 ) , ( 3 ) shows that a rectangular voltage area results @xmath23 this rectangle is expanded such that the outer circle of the voltage limitation is hit ( light grey on fig .
[ fig : constraints ] ) . during dynamical transients ,
the voltage vector points to one of the outer corners , subsequently touching the outer limiting circle .
therefore , linearizing the voltage limits as a rectangle by the presented method , as shown on fig .
[ fig : constraints ] , does not limit the operational range and only marginally affects dynamics .
a less restrictive method is presented in @xcite , where a time - varying linearization in form of a hexagon in stator frame is presented .
while such a linearization is possible with the underlying predictive control algorithm , the chosen method in the ( @xmath24)-frame is chosen for simplicity .
, width=321 ]
to study real - time applicability of the presented scheme , first , the highest possible amount of optimization parameters is determined .
the fastest optimizer with constraints is the widely known linear programming ( lp ) method .
table i shows some worst - case computational results of lp ( simplex method from @xcite ) as function of the number of free parameters ( cpu : @xmath25 ghz industrial pc ) .
more parameters lead to a higher number of iterations which are also more complex ; the worst - case number of iterations is the number of parameters plus the number of constraints . as in the underlying application , the constraints are decoupled , however , this worst - case is not to be expected .
the maximum runtime is given by the sampling rate minus latency of input / output , therefore at @xmath2 khz sampling rate , it must be less than about @xmath26 @xmath27s .
thus , at best , @xmath28 parameters can be optimized .
runtime of the predictive controller is further discussed at the end of the section .
s. bolognani , s. bolognani , l. peretti and m. zigliotto , combined speed and current model predictive control with inherent field - weakening features for pmsm drives , _ proc . of the ieee med .
electrotechnical conf .
_ , pp . 472478 , 2008 . c. cavallaro , a.o .
di tommaso , r. micelli , a. raciti , g.r .
galluzzo and m. trapanese , efficiency enhancement of permanent - magnet synchronous motor drives by online loss minimization approaches , _ ieee trans .
4 , pp . 11531160 , 2005 .
n. urasaki , t. senjyu and k. uezato , a novel calculation method for iron loss resistance suitable in modeling permanent - magnet synchronous motors , _ ieee trans .
energy convers .
1 , pp . 4147 , 2003 . | implementing predictive controllers in electrical drives cascaded field - oriented control structure with pi controllers machine exploited by improved control behavior system variables optimized conventional torque and current control structure of two changed to multi - input multi - output control torque generation decoupled from flux variation current controllers strongly coupled mimo controller advantageous improvements are better decoupling current voltage constraint handling better power efficiency dynamics currents in obstacle predictive control limited computational power high sampling rates existing implementations suffer from exploit advantages of model predictive control generalized predictive control high optimization horizon unconstrained predictive torque control constrained only reaches steps of prediction for performance inclusion constraints high optimization horizon required control in field - oriented frame problem enables mathematical optimization algorithms schemes computationally efficient maximize information for computational power online solution of constrained linear - quadratic problem requires quadratic programming algorithms computationally too expensive for drive systems recent development explicit mpc offline solution computed stored as in real - time controller @xcitescheme reaches @xmath1 prediction steps with constraints fast online algorithms proposed @xcite gradient search interest online solutions growing advantages online optimization manipulation adaptation parameters flexible choice machines control implementation based on suboptimal trajectory generation algorithm @xcite flatness - based predictive control scheme continuous approach variables not discretized represented as polynomial undetermined coefficients higher optimization horizons reached with few parameters reducing computational burden constraints handled by linearization cost functional linear programming solver smallest fastest numerical optimizers suboptimal prediction of @xmath0 ms with current voltage constraints obtained at @xmath2 khz sampling rate linear - quadratic optimization problems simpler machine model linearized rotor speed change optimization horizon pmsm model voltage equations become linear electrical subsystem machine quadrature direct currents @xmath5 @xmath6 given as @xmath7 nomenclature shown in table ii reluctance torque @xmath8 neglected small compared to electromagnetic torque in surface - mounted pmsmsmodel nonlinear requiring nonlinear optimization @xcite suitable cost functional key in predictive control only tuning possibility optimization control error dynamical performance machine losses efficiency goals in cost functional choosing cost functional trade - off goals during transients or fulfill goals in steady - state cost functional predictive torque controller @xmath9 trades off control error @xmath10 with machine losses @xmath11 first term @xmath12 represents copper losses second iron losses eddy current losses negligible on tested machines included iron losses reduced by field - weakening trade - off iron losses negative direct current @xmath6 flux magnitude stator reduced copper losses increase iron loss constant @xmath13 not standard motor parameters determined experimentally last term @xmath14 end - weight control error reducing steady - state control error weight @xmath15 set @xmath16 value determined heuristically optimization horizon set @xmath17 cost functional includes complete setpoint change important optimization horizon high open closed - loop trajectories differ behavior suboptimal illustrated in fig .horizon small due to control error significant difference between optimized open - loop closed - loop trajectories regeneration appears closed - loop trajectories do nt fit cost functional horizon higher for setpoint change difference between open closed - loop trajectories smaller important nonlinearities of pmsm voltage current limitations current constraints prevent overheating voltage limited by maximum dc - link voltage inverter voltage constraints limit rotor speed current dynamics in high - speed constraints linearized computationally current range for direct current @xmath6 limited to @xmath18 negative values of @xmath6 desirable improve power efficiency reduce induced voltage flux magnitude stator lowest value @xmath19 optimum value at rated speed given as @xmath21 independent of quatrature current @xmath5 reluctance torque neglected value doubled field - weakening dynamics in high speed for quadrature current @xmath5 largest range of values available linear constraints on fig . fill current region interest linearization acceptable voltage linearization more difficult @xmath22-axis not restricted induced voltage aligned largest valuesteady - state analysis system equations 3 shows rectangular voltage area results @xmath23 rectangle expanded outer circle voltage limitation hit light grey fig . during dynamical transients voltage vector points to outer outer limiting circle linearizing voltage limits as rectangle method limit operational range marginally affects dynamics less restrictive method in @xcite time - varying linearization hexagon in stator frame linearization possible with predictive control algorithm chosen method @xmath24)-frame chosen for simplicity study real - time applicability scheme highest optimization parameters determined fastest optimizer with constraints linear programming ( lp ) method table i shows worst - case computational results lp function number of free parameters more parameters lead higher iterations more complex worst - case number iterations number of parameters plus number constraints constraints decoupled worst - case not expected maximum runtime given by sampling rate minus latency input / output at @xmath2 khz sampling rate must less than @xmath26 @xmath27s best @xmath28 parameters can be optimizedruntime predictive controller discussed end section s bolognani peretti zigliotto speed current model predictive control field - weakening features pmsm drives electrotechnical conf. pp 472478 2008 c. cavallaro tommaso micelli raciti galluzzo trapanese efficiency enhancement permanent - magnet synchronous motor drives online loss minimization. 4 11531160 2005 urasaki senjyu k uezato novel calculation method iron loss resistance modeling permanent magnet synchronous motors energy convers . pp 4147 2003 | 0.352955 | 2,197 | 1,063 | , Saving $0.1 in GPT-4. |
21,073 | we report on multi - epoch x - ray observations of the type iin ( narrow emission line ) supernova with the _ rosat _ and _ asca_satellites .
the january 1998 _ asca _ x - ray spectrum is well fitted by a thermal bremsstrahlung ( @xmath0 kev , @xmath1 ) or power - law ( @xmath2 , @xmath3 ) model .
the x - ray light curve shows evidence for significant flux evolution between august 1996 and january 1998 : the count rate from the source decreased by 30% between our august 1996 and august 1997 _ rosat _ observations , and the x - ray luminosity most likely increased by a factor of @xmath42 between our august 1997 _
rosat _ and january 1998 _ asca _
observations , although evolution of the spectral shape over this interval is not ruled out .
the high x - ray luminosity , @xmath5 ergs s@xmath6 , places in a small group of type iin supernovae with strong circumstellar interaction , and the evolving x - ray luminosity suggests that the circumstellar medium is distributed inhomogeneously . [ firstpage ] supernovae : individual : x rays : stars stars : circumstellar matter | a small number of supernovae ( sne ) have been detected in x - rays in the near aftermath ( @xmath4years ) of their explosions .
sn 1978k , sn 1980k , sn 1986j , sn 1987a , and sn 1993j were found before 1995 ( see schlegel 1995 for a review ) ; more recent detections include sn 1979c , sn 1988z , and sn 1994i ( immler , pietsch , & aschenbach 1998a , b ) and sn 1999em ( fox & lewin 1999 ) . here
we report on observations of ( pollas 1995 ) .
supernova x - ray emission depends on the density structure of the progenitor s stellar wind as well as the structure of the sn ejecta , through which the initial shock passes .
after the shock wave emerges from the star , the characteristic velocity is @xmath410@xmath7 km s@xmath6 and the density distribution in the outer parts of the star can be approximated by a power - law in radius , @xmath8 , where the value of @xmath9 ranges from @xmath47 to @xmath420 ( models for sn1987a have @xmath10 ; matzner & mckee 1999 ) . the shock then propagates into the circumstellar material , typically a slow - moving wind with a density that decreases with the inverse square of the radius , @xmath11 , where @xmath12 is the stellar wind mass - loss rate and @xmath13 is the wind velocity ( typically @xmath410 km s@xmath6 ) .
the collision between the stellar ejecta and the circumstellar material produces a second ` reverse ' shock at @xmath41000 km s@xmath6 in the stellar ejecta ( which is expanding at @xmath410@xmath7 km s@xmath6 ) .
the forward shock produces a very hot shell ( @xmath410@xmath14 k ) , while the reverse shock produces a denser , cooler shell ( @xmath410@xmath15 k ) with much higher emission measure , from which the observable x - ray emission arises ( chevalier 1982 ) .
this picture is modified if the stellar wind is clumpy ; then the forward shock front in the clumps can yield cooler , stronger emission than the shock front in the diffuse wind ( chugai 1993 ) .
the exact time of the explosion of in the host galaxy is not known , but it may have occurred as much as ten months before its optical discovery in may 1995 ( benetti , bouchet & schwarz 1995 ) , making the supernova @xmath42 years old at the epoch of first detection ( lewin , zimmermann , & aschenbach 1996 ) .
was only the fourth supernova to be detected in x - rays so soon after its explosion .
regular observations of the x - ray temperature , luminosity , and line - of - sight hydrogen column density will help determine the structure of the sn ejecta , the structure of the circumstellar medium established by the wind of the progenitor , and the details of its evolution prior to the explosion .
these constraints can lead to an estimate of the progenitor stellar type and the type and length of its ( pre - explosion ) evolutionary stages .
we first observed in july 1996 with _ rosat _ when the optical ( benetti et al .
1995 ; garnavich & challis 1995 ) and radio ( van dyk et al .
1996 ) properties of this supernova made it a promising candidate for strong x - ray emission .
x - ray detection of with this pointing ( lewin et al .
1996 ) prompted our longer follow - up observation that august .
in addition , we were awarded time for another _ rosat _ observation in august 1997 , and applied simultaneously for _ asca _ spectral observations .
execution of our _ asca _ observation in january 1998 completed the program for this paper .
the log of our x - ray observations of is presented in table [ tbl : obs ] .
.log of x - ray observations [ cols=">,>,<,^ , > , > , > " , ] is the peak observed x - ray luminosity ( 0.110 kev ) .
references are : ( 1 ) petre et al . 1994 ;
( 2 ) schlegel et al .
1996 ; ( 3 ) houck et al.1998 ; ( 4 ) fabian & terlevich 1996 ; ( 5 ) this paper .
the interpretation of the x - ray emission from these sources is not clear .
chevalier & fransson ( 1994 ) presented models in which smooth supernova ejecta interact with a smooth circumstellar wind . in this picture , the reverse shock emission is characterized by @xmath16 kev and the forward shock emission by @xmath17 kev . however , the radio emission ( see weiler et al .
1990 for sn 1986j ) and narrow optical line emission give evidence that the circumstellar medium is clumpy and shock waves are being driven into the clumps ( see chugai et al .
1995 on the case of sn 1978k ) .
supernovae like sn 1978k and sn 1988z show very narrow line emission , which can be attributed to dense circumstellar clumps that are photo - ionized by the x - ray emission , and line components with a width of @xmath18 km s@xmath6that can be from cooling shocks moving into the clumps
. also shows evidence for these line components ( a.v .
filippenko , priv . comm . ) .
shock waves in clumps can give cooler x - ray emission than the forward shock front in the diffuse circumstellar medium .
chugai ( 1993 ) found that this picture may be preferred for the x - ray emission from sn 1986j .
a prediction of this model is that the emitting region should have a lower characteristic velocity than in the case of the reverse shock wave .
our series of flux measurements show that the x - ray luminosity of has probably dimmed by 30% and then brightened by as much as a factor of two over the period of observation , which corresponds to an age range of 2.0 to 3.5 years for a proposed explosion date in july 1994 ( benetti et al .
the x - ray light curve of sn 1978k shows approximately constant luminosity over the age range of 12 to 16 years ( schlegel , petre , & colbert 1996 ) . for sn 1986j , a decline in luminosity has been observed over the age range 8.6 to 13 years ( houck et al .
1998 ) ; the evolution is consistent with a @xmath19 time dependence . in the models of chevalier & fransson ( 1994 ) ,
a decline in flux is expected if the evolution is non - radiative , whereas a radiative reverse shock front leads to a roughly constant luminosity . in the clump model ( chugai 1993 ) , the evolution depends on the variation of clump properties with radius .
the light curve that we observe may be due to such variation .
characterization of the radio observations of is not yet complete .
however , five 8.4 ghz radio observations made between july 1995 and june 1998 are all consistent with a constant flux of 3.8 mjy ( van dyk et al .
2000 ) , indicating only mild evolution of the radio properties of the shocked region over this interval . in table
[ tbl : iin ] , we give the estimated temperature characterizing the x - ray emission , although it is only in the case of sn 1986j that an x - ray line has been detected ( houck et al .
1998 ) , and hence , the presence of thermal emission confirmed .
possible non - thermal sources of x - ray emission are inverse compton and synchrotron emission , but these are typically less efficient than thermal emission ( chevalier 1982 ) .
the temperature that we find is higher than expected for the reverse shock wave , but somewhat lower than that expected for the forward shock wave .
if the emission is from shocked clumps , the temperature depends on the density of the clumps , so that intermediate temperatures are possible .
the lack of detectable line emission does not allow us to directly test for the presence of clumps by examining the widths of lines .
the gas column density that we measure to the source is compatible ( within our roughly 50% uncertainty ) with the expected column density in our galaxy .
table [ tbl : iin ] shows that a higher column density is present to sn 1986j , but that can be attributed to absorption within the parent galaxy ( houck et al .
these supernovae do not show evidence for absorption by the dense shell that might be formed by a cooling reverse shock wave .
since the cool shell formed by a reverse shock will probably not be ionized , two years or more after the sn explosion , this low absorption may indicate inhomogeneities in the shell that allow the x - ray flux to emerge without absorption .
alternatively , low absorption may be seen as supporting the clump model ( chugai 1993 ) , since the x - ray emission from forward shocks penetrating high - density clumps will not be significantly absorbed .
our observations show that belongs to the class of x - ray luminous type iin supernovae .
the variations in x - ray flux over a 1.5 yr period suggest that the emission mechanism is more complicated than simple radiative emission from a reverse shock wave generated by interaction with a dense circumstellar medium , and that inhomogeneities in the circumstellar medium play a role .
the x - ray emission is expected to power the optical / ultraviolet emission from the supernova ; thus , optical spectroscopy will provide information on the gas motions so that a model for the supernova and the circumstellar structure can be determined .
whgl and rc gratefully acknowledge support from the national aeronautics and space administration .
kww thanks the office of naval research for the 6.1 support of this research .
this research has made use of the digitized sky survey , produced at the space telescope science institute under u.s .
government grant nag w-2166 . | small supernovae detected x - rays explosions sn 1978k 1980k 1986j 1987a 1993j found before 1995 schlegel 1995 recent detections include sn 1979c 1988z 1994i 1999em observations pollas 1995 supernova x - ray emission depends density progenitor s stellar wind structure sn ejecta initial shock passes shock wave characteristic velocity @xmath410@xmath7 km s@xmath6 density distribution outer parts star approximated by power - law radius @xmath8 @xmath9 ranges @xmath47 to @xmath420 models sn1987a @xmath10 shock propagates into circumstellar material slow - moving wind density decreases inverse square radius @xmath11 @xmath12 stellar wind mass - loss rate @xmath13 wind velocity @xmath410 km s@xmath6 collision between stellar ejecta circumstellar material produces second reverse shock at @xmath41000 km s@xmath6 ejecta expanding at @xmath410@xmath7 km s@xmath6forward shock produces hot shell ( @xmath410@xmath14 k ) reverse shock denser cooler shell ( @xmath410@xmath15 k ) higher emission measure observable x - ray emission arises ( chevalier 1982 ) picture modified if stellar wind clumpy forward shock front in clumps cooler stronger emission than diffuse wind ( chugai 1993 ) exact time explosion not known occurred ten months before optical discovery may 1995 supernova @xmath42 years old first detection fourth supernova detected in x - rays after explosion regular observations of x - ray temperature luminosity line sight hydrogen column density determine structure sn ejecta circumstellar medium details evolution prior explosion lead to estimate progenitor stellar type length pre - explosion evolutionary stages first observed july 1996 with _ rosat _ optical radio properties promising candidate for strong x - ray emission x - ray detection prompted longer follow - up observation awarded time for another _ rosat _ observation august 1997 applied simultaneously for _ asca _ spectral observationsasca observation january 1998 completed program paper log x - ray observations presented table peak observed x - ray luminosity 0.110 kev ) references petre et al . 1994 schlegel et al 1996 houck et al.1998 fabian & terlevich 1996 paper interpretation x - ray emission not clear chevalier fransson 1994 models smooth supernova ejecta interact smooth circumstellar wind reverse shock emission @xmath16 kev forward shock @xmath17 kev radio emission weiler et al . 1990 sn 1986j narrow optical line emission circumstellar medium clumpy shock waves driven into clumps chugai et al . 1995 sn 1978k supernovae sn 1978k sn 1988z show narrow line emission dense circumstellar clumps photo - ionized x - ray emission line components @xmath18 km s cooling shocks clumps shock waves clumps cooler x - ray emission forward shock front diffuse circumstellar medium chugai 1993 found picture preferred x - ray emission sn 1986jprediction model emitting region lower velocity than reverse shock wave flux measurements show x - ray luminosity dimmed by 30% brightened factor two over observation age range 2.0 to 3.5 years for proposed explosion july 1994 x - ray light curve of sn 1978k shows constant luminosity over 12 to 16 years sn 1986j decline in luminosity observed over 8.6 to 13 years evolution consistent with @xmath19 time dependence models chevalier & fransson 1994 decline in flux expected if evolution non - radiative radiative reverse shock front leads constant luminosity clump model evolution depends on variation clump properties with radius light curve may due to variation characterization radio observations not complete five 8.4 ghz radio observations between july 1995 and june 1998 consistent with constant flux of 3.8 mjy mild evolution of radio properties shocked region table estimated temperature x - ray emission only sn 1986j x - ray line detected presence thermal emission confirmed possible non - thermal sources emission are inverse compton synchrotron emission less efficient than thermal ( chevalier 1982 )temperature higher for reverse shock lower forward shock wave if emission from shocked clumps temperature depends on density intermediate temperatures possible lack of detectable line emission test for presence clumps gas column density compatible 50% uncertainty with expected column density in galaxy table higher column density to sn 1986j attributed to absorption parent galaxy supernovae show evidence for absorption by dense shell by cooling reverse shock wave cool shell not ionized two years after sn explosion low absorption may indicate inhomogeneities shell x - ray flux without absorption low absorption clump model x - ray emission from forward shocks penetrating high - density clumps not significantly absorbed observations show x - ray luminous type iin supernovae variations in x - ray flux over 1.5 yr period suggest emission mechanism more complicated than emission reverse shock wave inhomogeneities in circumstellar medium role x - ray emission power optical / ultraviolet emission supernova optical spectroscopy information on gas motions model for supernova circumstellar structurewhgl rc acknowledge support national aeronautics space administration kww thanks office naval research support research digitized sky survey space telescope science institute u government grant w-2166 | 0.345456 | 2,312 | 1,139 | , Saving $0.1 in GPT-4. |
21,098 | an algebraic model based on lie - algebraic techniques is applied to the analysis of thermodynamic vibrational properties of diatomic molecules .
the local anharmonic effects are described by a morse - like potential and corresponding anharmonic bosons associated with the @xmath0 algebra .
a vibrational high temperature partition function and the related thermodynamic functions are derived in terms of the parameters of the model .
the thermal expansion operator is defined and obtained using the renormalized frequency of the model . | algebraic models have been used very successfully in nuclear physics and have led to new insights into the nature of complex many body systems [ 1 ] .
similar methods have been proposed to describe molecular excitations [ 2,3 ] .
the methods combine lie algebraic techniques , describing the interatomic interactions , with discrete symmetry techniques associated with the molecules . in the framework of the algebraic model [ 2 ] ,
the anharmonic effects of the local interactions are described by substituting the local harmonic potentials by morse - like potential , associated with the @xmath0 algebra .
the one - dimensional morse hamiltonian is written in terms of the generators @xmath1of @xmath0 , @xmath2 where @xmath3 is a constant .
the eigenstates , @xmath4,v\rangle $ ] , correspond to the @xmath5 symmetry - adapted basis,@xmath6where @xmath7is the total number of bosons fixed by the potential shape , and @xmath8 is the number of quanta in the oscillator , @xmath9 $ ] .
the anharmonic effects are described by anharmonic boson operators [ 2 ] , @xmath10 where @xmath11is the morse phonon operator with an eigenvalue @xmath8 .
the operators satisfy the commutation relations , @xmath12 = \hat{b},\;\;\;\;\;\;\;\left [ \hat{b}^{^{\dagger } } , \hat{v}\right ] = -\hat{b}^{^{\dagger } } , \;\;\;\;\left [ \hat{b},\hat{b}^{^{\dagger } } \right ] = 1-\frac{2\hat{v}}{n}\ ] ] the harmonic limit is obtained when @xmath13 , in which case @xmath14 \rightarrow 1 $ ] giving the usual boson commutation relations .
the morse hamiltonian can be written in terms of the operators @xmath15 and @xmath16 , @xmath17 which corresponds to vibrational energies @xmath18 where @xmath19 is the harmonic oscillator frequency .
the morse potential spectrum thus leads to a deformation of the harmonic oscillator algebra . a more detailed relationship between
the morse coordinates and momenta and the @xmath20 generators can be derived through a comparison of their matrix elements [ 5 ] .
note that for an infinite potential depth , @xmath21 , the morse potential can not be distinguished from the harmonic potential .
the anharmonic model has been developed to analyze molecular vibrational spectra [ 2 - 8 ] .
it provides a systematic procedure for studying vibrational excitations in a simple form by describing the stretching and bending modes in a unified scheme based on @xmath0 algebras which incorporate the anharmonicity at the local level .
the aim of this paper is to apply the algebraic approach to the vibrational high - temperature thermodynamics of a diatomic molecule and obtain the basic thermodynamic functions in terms of the parameters of the anharmonic model .
these results must be combined with the translational and rotational thermodynamic functions in order to compare with experiment , in a fashion similar to that of reference [ 9 ] . here , we concentrate our attention on the derivation of analytic results which will be applied in a subsequent paper [ 10 ] .
the vibrational partition function of a diatomic anharmonic molecule is @xmath22}e^{-\beta \varepsilon _ { v}}\ ] ] where @xmath23 , the vibrational energies @xmath24 are given in equation ( 5 ) and @xmath25 is the fixed total number of anharmonic bosons .
introducing new parameters,@xmath26 $ ] and @xmath27-v$ ] , the vibrational partition function can be written as , @xmath28 at high temperatures @xmath29 , for @xmath30 large and @xmath31 small , the sum can be replaced by the integral , @xmath32 where @xmath33 .
this integral can be evaluated exactly in terms of the error function , @xmath34 ( see [ 11 ] ) , @xmath35 equation ( 9 ) represents the high - temperature value of the vibrational partition function in the morse - like spectrum .
when @xmath36 , the harmonic limit of the model is obtained , @xmath37 which coincides with the harmonic vibrational partition function of a diatomic molecule at high temperatures .
having the partition function @xmath38 in terms of the parameters of the algebraic model , we are now in position to derive the basic thermodynamic functions
. an algebraic approach has been used in [ 12 ] to study the thermodynamic properties of molecules . however
, the partition function in [ 12 ] uses an approximation of the classical density of states , while we have derived an explicit function in terms of the parameters of the algebraic model which can now be used to compute the high - temperature thermodynamic functions in analytic form .
the mean vibrational energy is given by @xmath39 taking into account that @xmath40 we obtain the following expression for the mean vibrational energy in terms of the partition function @xmath38 , @xmath41 substituting @xmath38 by expression ( 9 ) gives , @xmath42 the harmonic limit is obtained from equation ( 13 ) , when @xmath30 @xmath43 and @xmath38 is given by ( 10 ) , @xmath44 this is the classical mean energy of a diatomic molecule at very high temperatures .
the vibrational part of the specific heat is , @xmath45 substituting @xmath46 with equation ( 13 ) and using ( 12 ) , we obtain @xmath47 this represents the vibrational specific heat in the anharmonic model in terms of the partition function @xmath38 .
all anharmonic contributions depend on the temperature .
the explicit value is obtained from equation ( 17 ) by substituting @xmath38 by expression ( 9 ) .
when @xmath30 @xmath43 , the harmonic limit of the model gives the vibrational specific heat of a diatomic molecule at very high temperatures , @xmath48 the mean vibrational energy in the anharmonic model can be written in terms of mean number @xmath49 of anharmonic quanta , each with energy @xmath50 , @xmath51 substituting @xmath46 by equation ( 13 ) , we obtain @xmath49 in terms of the partition function @xmath52 @xmath53 using expression ( 9 ) in equation ( 20 ) , we obtain the high - temperature value , @xmath54 the harmonic limit is obtained from equation ( 20 ) when @xmath55 and @xmath38 is given by expression ( 10 ) , @xmath56 the free vibrational energy in terms of the partition function @xmath38 is given by @xmath57 substituting @xmath38 with equation ( 9 ) gives the free vibrational energy in the algebraic model at high temperatures , @xmath58 using expression ( 10 ) in equation ( 23 ) , we obtain the classical harmonic result for the free vibrational energy at very high temperatures , @xmath59
it is well - known that the thermal expansion in a molecule is a result of the local anharmonic effects and can not be explained by the harmonic models .
it depends on the positions of the atoms and their displacement from the equilibrium . as an approximation , we can write the coordinate operator @xmath60 in terms of the anharmonic operators @xmath15 and @xmath16 ( see also [ 5 ] ) as @xmath61 here ,
@xmath62 depends on the number of quanta in the oscillator [ 4 - 7 ] , @xmath63 now , we define the thermal expansion operator , @xmath64 where @xmath65 represents the displacement from the equilibrium position .
the harmonic limit is obtained when @xmath21 , in which case @xmath66 and @xmath67 .
evaluating the mean value of the operator , @xmath68 , between the states @xmath69v\rangle $ ] and taking into account ( 27 ) , we obtain the mean thermal expansion @xmath70 in terms of the parameters of the anharmonic model , @xmath71 the anharmonic effects are essential at high temperatures @xmath29 .
when @xmath72 , equation ( 29 ) becomes , @xmath73 taking into account the high - temperature value of the mean number of quanta ( 22 ) , @xmath74 the thermal expansion of anharmonic diatomic molecule has been obtained in [ 13 ] , using a morse - type potential and classical methods .
the comparison of equations ( 29 ) and ( 30)-(31 ) with the results in [ 13 ] at high temperatures shows very good agreement .
it also shows that @xmath75 is proportional to the mean separation between the atoms at @xmath76 and that @xmath77 is proportional to the depth of the potential well , the latter being in agreement with the algebraic model [ 4 - 7 ] .
we have applied the algebraic model to those thermodynamic properties of diatomic molecule which at high temperature strongly depend on its anharmonicity .
we have derived the vibrational partition function and the related thermodynamic functions , such as mean vibrational energy and specific heat , in terms of the parameters of the model .
the next step is to consider the rotational excitations of the molecule .
the thermal expansion , which is a typical anharmonic effect , has been derived in terms of the parameters of the model and its high - temperature value is in a good agreement with classical anharmonic results .
the model can be further applied to the thermodynamics of polyatomic molecules .
this and other applications will be discussed elsewhere .
this work was supported in part by project ( 32397-e ) conacyt , mexico . | algebraic models used in nuclear physics led to insights complex systems similar methods proposed molecular excitations methods combine algebraic techniques with symmetry techniques model anharmonic effects local interactions described by substituting local harmonic potentials by morse - like potential @xmath0 algebra one - morse hamiltonian written in generators @xmath1of @xmath0 @xmath2 @xmath3 constant eigenstates @xmath4,v $ correspond to @xmath5 symmetry - adapted basis @xmath7is total number of bosons fixed by potential shape @xmath8 number of quanta in oscillator @xmath9 anharmonic effects described by anharmonic boson operators @xmath10 @xmath11is morse phonon operator with eigenvalue @xmath8 operators satisfy commutation relations @xmath12 = \hat{b} 1-{2\hat{v}}{n harmonic limit obtained when @xmath13 @xmath14 \rightarrow 1 $ usual boson commutation relationsmorse hamiltonian written operators @xmath15 @xmath16 vibrational energies @xmath18 @xmath19 harmonic oscillator frequency morse potential spectrum leads deformation harmonic oscillator algebra detailed relationship between morse coordinates momenta @xmath20 generators derived comparison matrix elements infinite potential depth @xmath21 morse potential distinguished from harmonic potential anharmonic model developed molecular vibrational spectra systematic procedure studying vibrational excitations stretching bending modes unified scheme @xmath0 algebras anharmonicity aim paper apply algebraic approach to vibrational high - temperature thermodynamics diatomic molecule obtain basic thermodynamic functions parameters anharmonic model results combined with translational rotational thermodynamic functions compare with experiment derivation of analytic results subsequent paper vibrational partition function of diatomic anharmonic molecule is @xmath22}e^{-\beta \varepsilon _ @xmath23 vibrational energies @xmath24 equation @xmath25 fixed total number of anharmonic bosonsnew parameters@xmath26 vibrational partition function written @xmath28 high temperatures @xmath29 @xmath30 large small sum replaced by integral @xmath32 @xmath33 integral evaluated error function @xmath34 @xmath35 equation 9 represents high - temperature value vibrational partition function morse - spectrum @xmath36 harmonic limit @xmath37 coincides with vibrational partition function diatomic molecule high temperatures partition function @xmath38 algebraic model derive basic thermodynamic functions algebraic approach used 12 ] thermodynamic properties molecules partition function 12 ] uses approximation classical density states derived explicit function algebraic model compute high - temperature thermodynamic functions mean vibrational energy given by @xmath39 @xmath40 expression partition function @xmath38 @xmath41 substituting @xmath38 expression ( 9 ) gives @xmath42 harmonic limit from equation ( 13 ) @xmath30 @xmath43 @xmath38 10 @xmath44 classical mean energy diatomic molecule at high temperaturesvibrational part heat @xmath45 substituting equation 13 ) 12 ) @xmath47 vibrational heat anharmonic model partition function @xmath38 anharmonic contributions depend temperature explicit value equation 17 ) substituting @xmath38 expression 9 ) @xmath30 @xmath43 harmonic limit vibrational heat diatomic molecule high temperatures @xmath48 mean vibrational energy mean number @xmath49 anharmonic quanta energy @xmath50 @xmath51 substituting @xmath46 equation 13 ) obtain @xmath49 partition function @xmath52 @xmath53 expression 9 ) equation 20 ) high - temperature value @xmath54 harmonic limit equation 20 ) @xmath55 @xmath38 expression 10 ) @xmath56 free vibrational energy partition function substituting equation 9 ) energy high temperatures @xmath58 expression 10 ) equation 23 ) classical harmonic result free vibrational energy high temperatures @xmath59 thermal expansion molecule local anharmonic effects harmonic models depends positions atoms displacement equilibriumwrite coordinate operator @xmath60 anharmonic operators @xmath15 @xmath16 @xmath61 @xmath62 depends on number quanta oscillator define thermal expansion operator @xmath64 @xmath65 displacement from equilibrium harmonic limit obtained @xmath21 @xmath66 @xmath67 evaluating mean value operator @xmath68 states mean thermal expansion @xmath70 parameters anharmonic model @xmath71 anharmonic effects essential at high temperatures @xmath29 @xmath72 equation 29 ) becomes @xmath73 high - temperature value mean number quanta 22 @xmath74 thermal expansion anharmonic diatomic molecule obtained in [ 13 ] morse - type potential classical methods comparison equations ( 29 ) 30)-(31 ) with results 13 high temperatures shows agreement @xmath75 proportional to mean separation between atoms @xmath76 @xmath77 proportional depth potential well algebraic model applied algebraic model thermodynamic properties diatomic molecule depend on anharmonicity derived vibrational partition function related thermodynamic functions mean vibrational energy specific heat parameters model next consider rotational excitations moleculethermal expansion typical anharmonic effect derived parameters model high - temperature value with classical anharmonic results model applied to thermodynamics polyatomic molecules applications discussed elsewhere supported by project ( 32397-e ) conacyt mexico | 0.342521 | 2,196 | 1,116 | , Saving $0.1 in GPT-4. |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.