text
stringlengths
128
2.05k
An easy consequence of 5.1 above is that the type of the co-induced action only depends on the type of the action. Corollary 5.3
Let [MATH] and [MATH] . If [MATH] then [MATH] Remark 5.4 If [MATH] and [MATH] , we may also view [MATH] as an element of [MATH] . In BGK17 the notion of a characteristic random subgroup is defined to be an IRS which moreover is invariant under the action of the full automorphism group. So, as any group [MATH] is contai...
Continuity of co-induction We will consider here continuity properties of the co-induction operation. First, on the level of invariant random subgroups we have the following result.
Proposition 6.1 Let [MATH] . The map [MATH] is continuous if and only if either [MATH] or [MATH] Proof. It is easily seen that if [MATH] , then [MATH] for any [MATH] . Thus in this case the co-induction operation is constant. If [MATH] , then the operation is continuous because the product in the definition is finite.
Conversely, assume [MATH] and [MATH] For each [MATH] let [MATH] Then [MATH] as [MATH] in [MATH] . But we have [EQUATION] for any [MATH] finite, while
[EQUATION] for all [MATH] finite. Since [MATH] is a homeomorphism between the space of weak equivalence classes of an amenable group and the space of invariant random subgroups on the group, we now have the following corollary.
Corollary 6.2 Let [MATH] and assume [MATH] amenable. Then we have that the map [MATH] is continuous if and only if either [MATH] or [MATH]
One implication holds in general. To prove this we will use a sequence of weak equivalence classes, which converges to the weak equivalence class of the trivial action in [MATH] . We denote the trivial action by [MATH]
Proposition 6.3 Let [MATH] and assume [MATH] and [MATH] . Then [MATH] is not continuous. Proof. Consider a sequence of actions [MATH] for which there exists [MATH] Borel satisfying that [MATH] [MATH] is free and [MATH] is trivial for all [MATH] . Then, since
[EQUATION] for all [MATH] and all Borel [MATH] , we have [MATH] as [MATH] in [MATH] . However, since [MATH] for all [MATH] , it follows as in the proof of 6.1 , that [MATH] , when [MATH] , in [MATH] . Thus [MATH] cannot be continuous.
By similar arguments as those above, we can also say something about the continuity of countable powers of an action. In general for an action [MATH] and [MATH] the action [MATH] is defined by
[EQUATION] for all [MATH] and [MATH] . We then have the following result. Proposition 6.4 The map [MATH] from [MATH] to [MATH] is not continuous if [MATH]
Proof. Let [MATH] and [MATH] be as in the proof of 6.3 and assume towards a contradiction that the map is continuous. Then we would have [MATH] as [MATH] in [MATH] . But,
[EQUATION] for all [MATH] , while [MATH] Remark 6.5 In B18 , Theorem 1.2] it is shown that for a class of groups, containing the non-abelian free groups, the operation [MATH] is not continuous, not even when restricted to the space of free weak equivalence classes. As a corollary, for any group [MATH] in this class, th...
[EQUATION] is not continuous, again, not even when restricted to the space of free weak equivalence classes. So, while co-induction on weak equivalence classes is continuous in the finite index case, when the big group is amenable, this is not the case in general.
Properties of the co-induced invariant random subgroups We will study here different properties of the co-induced invariant random subgroups such as mixing properties and non-atomicity. We also obtain a characterization of when the co-induced action is free.
First note that it is clear that if [MATH] is ergodic (resp., weakly mixing), then [MATH] is ergodic (resp., weakly mixing). The converse does not hold in general. For example, [MATH] can be a free non-ergodic action, but [MATH] is ergodic. However, for each [MATH] , if [MATH] is ergodic (resp., weakly mixing), the act...
By use of the analogous result for actions, we obtain the following result. Proposition 7.1 Let [MATH] with [MATH] . Then [MATH] is weakly mixing for any [MATH]
Proof. Let [MATH] and let [MATH] satisfy [MATH] . Then by 3.3 we have [MATH] is weakly mixing and hence, by 5.1 , so is [MATH] In general, weakly mixing is the strongest mixing property one can hope for a non-atomic IRS. Indeed, by the result of T-D15a , if an IRS [MATH] is totally ergodic (i.e., the restriction of the...
Let [MATH] be any group and [MATH] . We let [MATH] Note that [MATH] is a subgroup of [MATH] Proposition 7.2 Let [MATH] with [MATH] and [MATH] . If [MATH] is atomic, then [MATH]
Proof. Assume [MATH] satisfies [MATH] . Then the orbit of [MATH] must be finite and [MATH] restricted to this orbit is a uniform measure. The diagonal of the orbit is then a fixed positive measured subset of [MATH] , which is invariant under the diagonal action of [MATH] . Thus, as [MATH] is weakly mixing, the orbit mu...
[EQUATION] Thus [MATH] , as wanted. From the proposition above, it follows that in order to obtain a non-atomic IRS via the co-induction operation, we just need to ensure that we do not obtain a Dirac measure. Thus we have the following criterion.
Corollary 7.3 Let [MATH] with [MATH] and [MATH] . Moreover, let [MATH] be a transversal for the left cosets in [MATH] . Then [MATH] is non-atomic if and only if there is [MATH] such that
[EQUATION] and [MATH] for all [MATH] Proof. First note that [MATH] is weakly mixing by 7.1 . Hence it follows by 7.2 that [MATH] is non-atomic if and only if [MATH] , and the latter is equivalent to the statement in the corollary.
Note that if [MATH] for some [MATH] , then we have [MATH] . Thus [EQUATION] where [MATH] denotes the automorphism in [MATH] induced by [MATH]
It is clear that if [MATH] and [MATH] is normal with [MATH] , we have that [MATH] satisfies [MATH] . Thus all possible Dirac measures in [MATH] are contained in the image of the co-induction operation. In some cases these are the only ones.
Proposition 7.4 If [MATH] with [MATH] , then any [MATH] satisfies [MATH] Proof. Let [MATH] satisfy [MATH] . The for each [MATH] and [MATH] we have [MATH] . Thus, since [MATH] we have
[MATH] if [MATH] and [MATH] , if [MATH] , as wanted. In such cases, this means that the co-induced action has almost everywhere fixed stabilizers and the co-induced action of a faithful action is free. In general it is easily seen that if an action is free, then so are all its co-induced actions. By use of the descript...
Proposition 7.5 Let [MATH] [MATH] a transversal for the left cosets in [MATH] and [MATH] . Then [MATH] is not free if and only if for some [MATH] we have
[EQUATION] and [MATH] for all [MATH] Proof. It follows directly from 5.1 that [MATH] is not free if and only if for some [MATH] we have
[EQUATION] Since the latter is equivalent to the statement in the proposition, the conclusion follows. Note that for any [MATH] we have
[EQUATION] So for the co-induced action to be non-free, in the case [MATH] , the conjugates of some [MATH] under the transversal [MATH] must uniformly converge very fast to the identity in [MATH]
Remark 7.6 For any group [MATH] there exists a group [MATH] such that [MATH] and [MATH] . Thus for such groups any action [MATH] will satisfy that [MATH] is free.
New constructions of non-atomic, weakly mixing invariant random subgroups In this section we will apply the co-induction operation on invariant random subgroups to construct new examples of continuum size families consisting of non-atomic, weakly mixing invariant random subgroups on several classes of groups.
8.A A sufficient criterion We will provide in this subsection a sufficient criterion for an infinite index subgroup to generate continuum many non-atomic, weakly mixing co-induced invariant random subgroups on the bigger group.
In the following, for a group [MATH] and a subset [MATH] , we let [MATH] denote the subgroup [MATH] in [MATH] and [MATH] denote the normal subgroup [MATH] in [MATH]
Proposition 8.1 Let [MATH] with [MATH] . Consider the statements: (1) There exists a transversal [MATH] for the left cosets in [MATH] and [MATH] such that the chain of normal subgroups [MATH] , given by
[EQUATION] is not constant. (2) There exists a continuum size family [MATH] such that we have [MATH] are all non-atomic, weakly mixing and satisfy [MATH] for all [MATH] with [MATH]
(3) There exists [MATH] such that [MATH] is non-atomic. (4) There exists [MATH] such that [MATH] is not a Dirac measure. (5) For any transversal [MATH] for the left cosets in [MATH] there is [MATH] such that the chain of subgroups [MATH] , given by
[EQUATION] is not constant. Then [MATH] Proof. It is clear that [MATH] . Thus it suffices to prove [MATH] and [MATH] For the implication [MATH] , assume [MATH] holds for [MATH] and [MATH] We will first construct one such invariant random subgroup. Afterwards we will argue how to obtain uncountably many.
Let [MATH] . Then the non-constant assumption on the sequence [MATH] ensures that [MATH] , as for some [MATH] with [MATH] we have [MATH] . Moreover, it follows directly by the construction of [MATH] that [MATH] for all [MATH] and that
[EQUATION] So the assumptions of 7.3 are satisfied and thus the co-induced measure must be non-atomic. Since [MATH] , it will also be weakly mixing.
Now to construct uncountably many of these, let [MATH] be least such that [MATH] and let [MATH] . Next, fix [MATH] such that for [MATH] we have [MATH] if and only if [MATH] For each [MATH] put
[EQUATION] Then we have [EQUATION] for all [MATH] , while [EQUATION] for all [MATH] . Thus, by the description of the co-induction operation given in 5.1 , we obtain that
[EQUATION] So [MATH] is a continuum size family of non-atomic, weakly mixing invariant random subgroups of [MATH] , as wanted. For the implication [MATH] , assume that [MATH] holds. Let [MATH] be a transversal for [MATH] . We have that
[EQUATION] for all [MATH] finite. Since [MATH] is not a Dirac measure, there exists [MATH] such that [MATH] . Thus, if we let [MATH] satisfy that [MATH] , we have [MATH] for some [MATH] and
[EQUATION] By convergence of the series, it follows that for some [MATH] we have [EQUATION] Thus we must have [MATH] , as wanted.
Remark 8.2 In general, if [MATH] is a normal subgroup, then [MATH] is a transversal for the left cosets in [MATH] if and only if [MATH] is a transversal, as well. Thus in this case, the statement
(1’) There exists a transversal [MATH] for the left cosets in [MATH] and [MATH] such that the chain of normal subgroups [MATH] given by
[EQUATION] is not constant. is equivalent to condition [MATH] in 8.1 Remark 8.3 If [MATH] in 8.1 is abelian, then all the statements are equivalent. We also point out that the invariant random subgroups constructed in the proof of [MATH] above, are not weakly mixing when restricted to [MATH]
8.B Wreath products and HNN-extensions We will here apply the criterion in 8.1 to wreath products and HNN-extensions. (1) Let [MATH] be countable groups and consider the action [MATH] given by [MATH] . The wreath product of [MATH] by [MATH] is then the semidirect product [MATH] and is denoted by [MATH]
Construction of continuum many non-atomic, weakly mixing invariant random subgroups on [MATH] , for [MATH] countable groups such that [MATH] is infinite and [MATH] is not trivial.
Let [MATH] and [MATH] . Then [MATH] and [MATH] is a transversal for the left cosets [MATH] . Fix an enumeration [MATH] and let [MATH] . Define [MATH] by
[EQUATION] Then, as [MATH] is not constant, following 8.1 we construct continuum many non-atomic, weakly mixing invariant random subgroups on [MATH]
If [MATH] is a countable set and we have an action [MATH] , we may form a wreath product by letting [MATH] be given by [MATH] and then consider the semidirect product [MATH] . We denote such a wreath product by [MATH] . Arguments similar to those in the preceding paragraph work as well for [MATH] , if the action [MATH]...
(2) Next we will consider HNN-extensions over “small” subgroups. Construction of continuum many non-atomic, weakly mixing invariant normal subgroups for the HNN extension [MATH] , where [MATH] is a countable group, [MATH] and [MATH] is an embedding with
[MATH] Let [MATH] for each [MATH] be a copy of [MATH] and put [EQUATION] Then [MATH] , where [EQUATION] for all [MATH] and [MATH] (see B08 , Theorem 17.1] ). Now let [MATH] and consider the homomorphism [MATH] induced by the homomorphisms [MATH] for [MATH] given by [MATH] if [MATH] and [MATH] for all [MATH] For a fixed...
[EQUATION] and hence using 8.1 we construct continuum many non-atomic, weakly mixing invariant random subgroups in [MATH] Note that this covers the case where [MATH] is an automorphism of a non-trivial normal subgroup of [MATH] . We also have the following application.
Corollary 8.4 If [MATH] are not relatively prime, then there are continuum many non-atomic, weakly mixing invariant random subgroups on [MATH]
Proof. We have that [MATH] is the HNN-extension of [MATH] with respect to the isomorphism [MATH] and [MATH] 8.C Non-abelian free groups
We will now turn our attention towards the non-abelian free groups. It follows already from the results in BGK17 that these groups admit continuum many non-atomic, weakly mixing invariant random subgroups. In this subsection we show how the co-induction can be used to give alternative constructions of invariant random ...
(1) First we will use the co-induction operation from [MATH] to various semi-direct products of the form [MATH] , where [MATH] is induced by a permutation of the generators of [MATH] , to construct new invariant random subgroups on [MATH]
Construction of continuum many non-atomic, weakly mixing invariant random subgroups on [MATH] Fix [MATH] and let [MATH] be a Borel partition with [MATH] . For each [MATH] , put [MATH] and fix a free action [MATH] . Then define [MATH] by [MATH] if [MATH] and [MATH] if [MATH] for some [MATH] . Finally, let [MATH] act as ...
Next, let [MATH] be infinite and let [MATH] be a permutation which is transitive on [MATH] and fixes every element of [MATH] . We then define [MATH] by [MATH] and [MATH] for all [MATH] and [MATH] . Consider [MATH] and let
[EQUATION] Note that by 3.3 we have [MATH] is weakly mixing and for [MATH] we have [EQUATION] So for [MATH] it holds that [EQUATION]
and hence [MATH] is non-atomic. Moreover, this implies that whenever [MATH] are infinite with [MATH] we have [MATH] (2) Next consider another construction using co-induction, which allows us to construct continuum many non-atomic, weakly mixing invariant random subgroups on every non-abelian free group.
Construction of continuum many non-atomic, weakly mixing invariant random subgroups on [MATH] for [MATH] with [MATH] Fix [MATH] with [MATH] and some free generators [MATH] . Consider the surjective group homomorphism [MATH] given by [MATH] for [MATH] and [MATH] . Then let [MATH] and note that
[EQUATION] This set freely generates [MATH] as a copy of [MATH] inside [MATH] Moreover, the set [MATH] constitutes a transversal for [MATH]
Now for each [MATH] let [MATH] satisfy that the action induced by [MATH] is weakly mixing and for each [MATH] we have [EQUATION]
One way to choose [MATH] is to decompose [MATH] such that [MATH] and then let [MATH] be weakly mixing when restricted to [MATH] and trivial on [MATH] for [MATH] Next, define an action [MATH] by letting [MATH] and [MATH] for all [MATH] and [MATH] . Then put [MATH] . Note that all conditions of 7.3 are satisfied with res...
[EQUATION] Thus [MATH] for all [MATH] with [MATH] Remark 8.5 Using an action similar to the one in the construction above one can give a proof of the following algebraic fact: Let [MATH] satisfy [MATH] if [MATH] . Then the sequence
[EQUATION] does not extend to a basis of [MATH] . Indeed, assume towards a contradiction, that we may extend the sequence to a basis of [MATH] . Then we would have that [MATH] generates a copy, [MATH] , of [MATH] as a subgroup of [MATH] . So let [MATH] be an action such that [MATH] for all [MATH] and
[EQUATION] as [MATH] in [MATH] Now, since the sequence extends to a basis, we may extend this action to an action [MATH] . Therefore we would have
[EQUATION] which contradicts the convergence above. 8.D Free products with normal amalgamation Here we will use co-induction to give constructions of non-atomic, weakly mixing invariant random subgroups on certain free products of groups with normal amalgamation. Other constructions can be found in BGK17 but our proofs...
Construction of continuum many non-atomic, weakly mixing invariant random subgroups on [MATH] , where [MATH] [MATH] are non-trivial countable groups with [MATH] infinite, with support in
[EQUATION] Moreover, we can ensure that these invariant random subgroups are weakly mixing, when restricted to [MATH] Let [MATH] , consider the homomorphism [MATH] induced by the homomorphisms [MATH] , and put [MATH] . Then [MATH] is freely
[EQUATION] where [MATH] and [MATH] is a transversal for the left cosets [MATH] . Now fix [MATH] and [MATH] . For each [MATH] , let [MATH] be an action satisfying [MATH] and [MATH] for all [MATH] and [MATH] . Let [MATH] and note that
[EQUATION] We have [EQUATION] for all [MATH] [MATH] . Thus [MATH] or its inverse is in the reduced word over the alphabet of non-trivial commutators of
[EQUATION] if and only if [EQUATION] Therefore [MATH] and so [MATH] constitute a continuum size family of non-atomic, weakly mixing invariant random subgroups of [MATH]
To ensure weakly mixing when restricted to [MATH] , let [MATH] satisfy that [MATH] are distinct and that [MATH] [MATH] for all [MATH] . Then modify [MATH] such that the action of [MATH] is weakly mixing and each satisfies [MATH] for [MATH] . Note that the relation constrains on [MATH] ensure that at most one of [MATH] ...
[EQUATION] when [MATH] and [MATH] . Moreover, as with [MATH] , each will appear exactly four times. So we then have [EQUATION] Again, [MATH] constitute a continuum size family of non-atomic, weakly mixing invariant random subgroups of [MATH] . These will now also be weakly mixing, when restricted to [MATH] by 3.3
Remark 8.6 Gaboriau pointed out that in the paper D. Gaboriau and N. Bergeron, Asymptotique des nombres de Betti, invariants [MATH] et laminations, Comment. Math. Helv. 79(2) (2004), 362–395, 2004, the following result is proved: Let [MATH] and [MATH] be residually finite, infinite groups such that either
[MATH] or [MATH] for [MATH] Then the free product [MATH] admits continuum many IRS. These IRS are distinguished by their [MATH] -Betti numbers. (Here, [MATH] is the [MATH] -th Betti number of [MATH] while [MATH] is its [MATH] -th [MATH] -Betti number.)
Next, note the following well-known simple fact. Proposition 8.7 Let [MATH] be countable groups and [MATH] a surjective group homomorphism. Then there is an embedding [MATH] such that if [MATH] is ergodic, weakly mixing or non-atomic, so is [MATH]
Proof. Note that the map [MATH] given by [MATH] is a homeomorphism with image [EQUATION] Moreover, we have [EQUATION] for all [MATH] and [MATH] . So let [MATH] be given by [MATH] . It is then clear that [MATH] is ergodic, weakly mixing or non-atomic if [MATH] is. Since
[EQUATION] it follows by 4.2 that [MATH] is continuous. Now by use of the previous construction for free products and 8.7 , we can construct continuum many non-atomic, weakly mixing invariant random subgroups for the groups [MATH] , where [MATH] and [MATH] are countable groups satisfying that [MATH] with [MATH] non-tri...
The same applies to all the groups [MATH] , where [MATH] is a countable family of countable groups with [MATH] infinite and [MATH] non-trivial, by looking at the natural surjective group homomorphism [MATH]
8.E Automorphism invariant random subgroups of the free group of rank two In this part we will use the co-induction operation to construct non-atomic invariant random subgroups on [MATH] which are invariant under the action of the full automorphism group, as well. Moreover, these invariant random subgroups will be weak...
Fix a basis [MATH] . We think of an element of [MATH] as represented by the induced reduced word in the letters [MATH] . Consider the automorphisms [MATH] given by
[EQUATION] [EQUATION] Let [MATH] denote the set of automorphisms [MATH] and [MATH] (and not [MATH] ). Then [EQUATION] is a set of representatives for the left cosets in [MATH] , where [MATH] is identified with the subgroup of inner automorphisms (see CMZ81 , Section 3] ). Note that [MATH] denotes the identity map.
Consider the word [MATH] for some [MATH] . The first goal is to prove that the family [EQUATION] satisfies the [MATH] cancellation property. Recall that a subset of words [MATH] has the [MATH] cancellation property if the set [MATH] of all cyclically reduced cyclic conjugates of the words in [MATH] and their inverses s...
[EQUATION] Here [MATH] denotes the length of a word in [MATH] . In case [MATH] satisfies the [MATH] cancellation property and [MATH] is a cyclically reduced word, then there is [MATH] such that [MATH] For a proof of this see for example LS77 , Theorem 4.5 in Chapter V] We will use this fact to ensure that condition [MA...
Put [EQUATION] and let [MATH] for all [MATH] Below we will use the following terminology. For two words [MATH] cancellation of [MATH] and [MATH] is a string [MATH] which appears in the reduced cycles of both [MATH] We say that [MATH] is a bad cancellation of [MATH] and [MATH] if
[EQUATION] Here [MATH] denote the length of the induced cyclically reduced word. We call a cancellation for maximal if it cannot be extended. The goal is then to prove that there is no bad cancellation between any pair of words in the set
[EQUATION] Let [EQUATION] Then it suffices to prove that there is no bad cancellation among the words in [MATH] and then prove that there cannot be any bad cancellation between a word from [MATH] and a word from [MATH] . We will begin with the former. Most of our arguments for this are based on the following two lemmas...
For a word [MATH] , we let [MATH] denote the word obtained from [MATH] by switching every negative power of [MATH] and [MATH] to be positive.
Lemma 8.8 Let [MATH] [MATH] and let [MATH] be a cancellation of [MATH] . Assume [MATH] satisfies that for any cancellation [MATH] of [MATH] the total number of [MATH] ’s and the total number of [MATH] ’s in [MATH] are both less than [MATH] . Then
[EQUATION] Proof. First let [MATH] and [MATH] be such that [MATH] . Assume [MATH] is a maximal cancellation of [MATH] Then there is a maximal cancellation [MATH] of [MATH] such that [MATH] is equal to one of the following strings:
[EQUATION] This can be seen by considering the pre-images through [MATH] of all possible strings of the form [EQUATION] where [MATH] and [MATH] are such that [MATH] and [MATH] is a legal string in the reduced cycles induced by [MATH] and [MATH] , together with the cases where [MATH] [MATH] and [MATH] . The latter cases...
Next, for [MATH] , we let [MATH] and [MATH] be such that [MATH] and [MATH] , for [MATH] . Now let [MATH] be a maximal cancellation of [MATH] . Then, by repeating the argument above, there is a maximal cancellation [MATH] of [MATH] such that