Q
stringlengths
4
3.96k
A
stringlengths
1
3k
Result
stringclasses
4 values
Proposition 4.5.7 (Hahn Decomposition Theorem). Let \( \nu \) be a signed measure on \( \left( {X,\mathcal{B}}\right) \) . Then there exists a positive set \( A \) and a negative set \( B \) with \( X = A \cup B \) and \( A \cap B = \varnothing \) .
Proof. By definition, \( \nu \) can take on at most one of the values \( + \infty \) , \( - \infty \) . Without loss of generality, assume \( \nu \) never takes on the value \( + \infty \) . Let\n\n\[ \lambda = \sup \{ \nu \left( E\right) \mid E \in \mathcal{B}\text{ and }E\text{ is positive }\} .\](In the case that \(...
Yes
Let \( C \) be the Cantor set. Again, the Cantor set is a Borel set and so is Hausdorff measurable. To find the Hausdorff dimension of the Cantor set, recall that at the \( k \) th stage of construction, \( {C}_{k} \) consisted of \( {2}^{k} \) intervals of length \( \frac{1}{{3}^{k}} \) . Hence, a candidate for \( {H}...
\[ \mathop{\lim }\limits_{{k \rightarrow \infty }}\frac{{2}^{k}}{{3}^{k\alpha }} = \mathop{\lim }\limits_{{k \rightarrow \infty }}{\left( \frac{2}{{3}^{\alpha }}\right) }^{k}. \] If \( \frac{2}{{3}^{\alpha }} < 1 \), this limit will be 0 . If \( \frac{2}{{3}^{\alpha }} > 1 \), this limit will be \( + \infty \) . As pre...
No
1.16 Proposition Let \( X \subseteq \mathbb{R} \) and \( f : X \rightarrow E \) be right and left differentiable at \( a \in X \) with \( {\partial }_{ + }f\left( a\right) = {\partial }_{ - }f\left( a\right) \) . Then \( f \) is differentiable at \( a \) and \( \partial f\left( a\right) = {\partial }_{ + }f\left( a\rig...
Proof By hypothesis and Proposition 1.1(iii), there are functions\n\n\[ \n{r}_{ + } : X \cap \lbrack a,\infty ) \rightarrow E\;\text{ and }\;{r}_{ - } : ( - \infty, a\rbrack \cap X \rightarrow E \n\] \n\nwhich are continuous at \( a \) and satisfy \( {r}_{ + }\left( a\right) = {r}_{ - }\left( a\right) = 0 \) and \n\n\[...
Yes
Let \( - \infty < a < b < \infty \) and \( f \in {C}^{2}\left( {\left\lbrack {a, b}\right\rbrack ,\mathbb{R}}\right) \) be such that \( {f}^{\prime }\left( x\right) \neq 0 \) for all \( x \in \left\lbrack {a, b}\right\rbrack \) . We suppose further that there is some \( \xi \in \left( {a, b}\right) \) such that \( f\le...
Proof (i) By the extreme value theorem (Corollary III.3.8), there are constants \( {M}_{1},{M}_{2}, m > 0 \) such that \( m \leq \left| {{f}^{\prime }\left( x\right) }\right| \leq {M}_{1},\;\left| {{f}^{\prime \prime }\left( x\right) }\right| \leq {M}_{2},\;x \in \left\lbrack {a, b}\right\rbrack \). For the function \(...
Yes
2.1 Theorem If \( \left( {f}_{n}\right) \) converges uniformly to \( f \) and almost all \( {f}_{n} \) are continuous at \( a \in X \), then \( f \) is also continuous at \( a \) .
Proof Let \( \varepsilon > 0 \) . Because \( {f}_{n} \) converges uniformly to \( f \), there is, by Remark 1.3(e), some \( N \in \mathbb{N} \) such that \( {\begin{Vmatrix}{f}_{n} - f\end{Vmatrix}}_{\infty } < \varepsilon /3 \) for all \( n \geq N \) . Since almost all \( {f}_{n} \) are continuous at \( a \), we can s...
Yes
6.9 Theorem (Fubini) For \( f \in {\mathcal{L}}_{1}\left( {\mathbb{R}}^{m + n}\right) \), (i) \( f\left( {x, \cdot }\right) \in {\mathcal{L}}_{1}\left( {\mathbb{R}}^{n}\right) \) for \( {\lambda }_{m} \) -almost every \( x \in {\mathbb{R}}^{m} \), \( f\left( {\cdot, y}\right) \in {\mathcal{L}}_{1}\left( {\mathbb{R}}^{m...
Proof (a) For \( f \in {\mathcal{L}}_{1}\left( {{\mathbb{R}}^{m + n},{\mathbb{R}}^{ + }}\right) \), the claim follows from Tonelli’s theorem and Remark 3.3(e). (b) Given the representation \( f = {f}_{1} - {f}_{2} + i\left( {{f}_{3} - {f}_{4}}\right) \), with \( {f}_{j} \in {\mathcal{L}}_{1}\left( {{\mathbb{R}}^{m + n}...
Yes
Theorem 1. Let \( I \) be a bounded interval. (i) If the bounded functions \( f \) and \( g \) are integrable on \( I \), then so likewise is \( {\alpha f} + {\beta g} \) for any constants \( \alpha \) and \( \beta \) , and
(2.3) \[ m\left( {{\alpha f} + {\beta g}}\right) = {\alpha m}\left( f\right) + {\beta m}\left( g\right) . \]
No
Theorem 2. If the real functions \( f \) and \( g \) are integrable on \( I \), so are the functions \( \sup \left( {f, g}\right) \) and \( \inf \left( {f, g}\right) \) .
By Theorem 1 and the formula above it is enough to show that if \( f \) is integrable then so is \( {f}^{ + } \) . This follows immediately from the definition,(1) or (1’), and from the inequality \( \left| {{f}^{ + } - {\varphi }^{ + }}\right| \leq \left| {f - \varphi }\right| \) .
No
Theorem 3. Let \( f \) and \( g \) be two bounded integrable functions on a compact interval \( I \) . Then the function \( f\bar{g} \) is integrable and (Cauchy-Schwarz inequal- \( {it}{y}^{4} \) )
(2.5) \[ {\left| m\left( f\bar{g}\right) \right| }^{2} \leq m\left( {\left| f\right| }^{2}\right) m\left( {\left| g\right| }^{2}\right) . \] In checking that \( f\bar{g} \) is integrable we may assume \( f \) and \( g \) real, and even positive, since every integrable real function \( f \) is the difference of the inte...
Yes
Theorem 4. Let \( \left( {f}_{n}\right) \) be a uniformly convergent sequence of integrable functions on a bounded interval \( I \) . Then the function \( f\left( x\right) = \lim {f}_{n}\left( x\right) \) is integrable and\n\n(4.2)\n\n\[ m\left( f\right) = {\int }_{I}f\left( x\right) {dx} = \lim {\int }_{I}{f}_{n}\left...
For \( r > 0 \) given, and for every \( n \), let us choose a step function \( {\varphi }_{n} \) such that \( m\left( \left| {{f}_{n} - {\varphi }_{n}}\right| \right) < r \), and let \( N \) be an integer such that\n\n\[ n > N \Rightarrow {\begin{Vmatrix}f - {f}_{n}\end{Vmatrix}}_{I} < r \]\n\nfrom the definition of un...
Yes
Consider a power series\n\n\[ f\left( z\right) = \sum {c}_{n}{z}^{n}/n! = \sum {c}_{n}{z}^{\left\lbrack n\right\rbrack } \]\n\nwhich converges on a disc \( \left| z\right| < R \) of nonzero radius, and let us calculate the integral of \( f\left( x\right) \) over an interval \( \left\lbrack {a, b}\right\rbrack \) with \...
We know that the series converges normally on every disc \( \left| z\right| \leq r < R \), so on the interval considered: we can therefore integrate term-by-term. But we also know (Chap. II, \( {\mathrm{n}}^{ \circ }{11} \) ) that\n\n\[ {\int }_{a}^{b}{x}^{n}{dx} = \frac{{b}^{n + 1}}{n + 1} - \frac{{a}^{n + 1}}{n + 1} ...
Yes
Theorem 5 (Borel-Lebesgue). Let \( K \) be a compact subset of \( \mathbb{R} \) (resp. \( \mathbb{C} \) ) and \( {\left( {U}_{i}\right) }_{i \in I} \) a family of open sets in \( \mathbb{R} \) (resp. \( \mathbb{C} \) ). Suppose that \( K \) is contained in the union of the \( {U}_{i} \) . Then there is a finite subset ...
First we show that if \( K \) is bounded one can, for every \( r > 0 \), find a finite number of points \( {x}_{k} \) of \( K \) such that \( K \) is contained in the union of the open balls \( B\left( {{x}_{k}, r}\right) \) . Since \( K \) is certainly contained in a compact interval or square, it is clear that one ca...
Yes
Let \( {\left( {K}_{i}\right) }_{i \in I} \) be a family of nonempty compact sets in \( \mathbb{R} \) or \( \mathbb{C} \) . Suppose that the intersection of the \( {K}_{i} \) is empty. Then there is a finite subset \( F \) of \( I \) such that the intersection of the \( {K}_{i}, i \in F \), is empty.
We choose any index \( j \) and replace each \( {K}_{i} \) by \( {K}_{i} \cap {K}_{j} \) . If one of these intersections is empty, the corollary is proved. So assume they are nonempty. This is equivalent to assuming that all the \( {K}_{i} \) are contained in the same compact set \( K \), namely \( {K}_{j} \) . Let \( ...
Yes
Theorem 6. Let \( f \) be a scalar function defined on an interval \( I \) of \( \mathbb{R} \) . The two following properties are equivalent: (i) \( f \) has left and right limits at every point of \( I \) ; (ii) there is a sequence of step functions on \( I \) which converges to \( f \) uniformly on every compact subs...
The implication (ii) \( \Rightarrow \) (i) was established from Cauchy’s criterion in Chap. III, \( {\mathrm{n}}^{ \circ } 12 (Corollary of Theorem 16). The implication (i) \( \Rightarrow \) (ii) is obtained, when \( I \) is compact, by observing, as at the beginning of the preceding \( {\mathrm{n}}^{ \circ } \), that ...
Yes
Theorem 7. The integral of a regulated (resp. continuous) positive function \( f \) is zero if and only if the set \( D = \{ f\left( x\right) \neq 0\} \) is countable (resp. empty).
The condition is sufficient. For consider a step function \( \varphi \leq f \) . One can have \( \varphi \left( x\right) > 0 \) only if \( x \in D \) . Since the set of points of a nonsingleton interval is uncountable (Chap. I), the function \( \varphi \) is necessarily negative on all the intervals of nonzero length w...
Yes
Theorem 8 (Heine \( {}^{12} \) ). Every scalar function defined and continuous on a compact set \( K \subset \mathbb{C} \) is uniformly continuous on \( K \) .
Given \( r > 0 \) let us choose for each \( x \in K \) an open ball \( B\left( x\right) \) with centre \( x \) such that \( f \) is constant to within \( r \) in \( B\left( x\right) \cap K \) . Let \( {B}^{\prime }\left( x\right) \) be the open ball with centre \( x \) and of radius half that of \( B\left( x\right) \) ...
Yes
Corollary 1. Let \( f \) be a scalar function defined and continuous on \( \mathbb{R} \) (resp. \( \mathbb{C} \) ) and zero for \( \left| x\right| \) large. Then \( f \) is uniformly continuous on \( \mathbb{R} \) (resp. \( \mathbb{C} \) ).
We need only treat the case of \( \mathbb{C} \) . Let \( K \) be a compact set outside which \( f = 0 \), and \( H \) the set of \( x \in \mathbb{C} \) such that \( d\left( {x, K}\right) \leq 1 \) . Since \( d\left( {x, K}\right) \) is a continuous function of \( x \) (Chap. III, \( {\mathrm{n}}^{ \circ }{10} \) ), the...
Yes
Corollary 2. Let \( f \) be a function defined and continuous on a bounded set \( X \subset \mathbb{C} \) . The following two properties are equivalent: (i) \( f \) is uniformly continuous on \( X \) ; (ii) \( f \) is the restriction to \( X \) of a continuous function on the compact set \( \overline{X} \) .
We have just seen that (i) implies (ii). The converse implication follows from Theorem 8 since \( \bar{X} \) is compact.
Yes
Corollary 3. Let \( f \) be a scalar function defined and continuous on a compact interval \( I \) . For every \( r > 0 \) there exists an \( {r}^{\prime } > 0 \) such that\n\n\[ \left| {{\int }_{I}f\left( x\right) {dx}-\sum f\left( {\xi }_{k}\right) \left( {{x}_{k + 1} - {x}_{k}}\right) }\right| < r \] \n\nfor any poi...
For example one can decompose \( I \) into \( n \) equal intervals \( {I}_{1},\ldots ,{I}_{n} \) and choose a \( {\xi }_{k} \in {I}_{k} \) at random for each \( k \) . The corresponding Riemann sum is just\n\n\[ m\left( I\right) \frac{f\left( {\xi }_{1}\right) + \ldots + f\left( {\xi }_{n}\right) }{n}. \] \n\nIt tends ...
No
Theorem 10. Let \( K \) and \( H \) be two compact intervals in \( \mathbb{R} \) and \( f \) a continuous function on \( K \times H \) . Then\n\n\[{\int }_{H}\nu \left( y\right) {dy}{\int }_{K}f\left( {x, y}\right) \mu \left( x\right) {dx} = {\int }_{K}\mu \left( x\right) {dx}{\int }_{H}f\left( {x, y}\right) \nu \left(...
To prove the equality of the two sides of (7), note that, by (3), there exist finite partitions of \( K \) and \( H \) into intervals \( {K}_{p} \) and \( {H}_{q} \) such that \( f \) is constant to within \( r \) on each rectangle \( {K}_{p} \times {H}_{q} \) . Then\n\n\[{\int }_{H}f\left( {x, y}\right) \nu \left( y\r...
No
Corollary 1. Let \( K \) be a compact interval and \( \left( {f}_{n}\right) ,\left( {g}_{n}\right) \) two everywhere increasing or two everywhere decreasing sequences of real continuous functions on \( K \) . Assume that \( \lim {f}_{n}\left( x\right) = \lim {g}_{n}\left( x\right) \) for every \( x \in K \) . Then\n\n\...
Consider for example the case of increasing sequences, put\n\n\[ \varphi \left( x\right) = \sup {f}_{n}\left( x\right) \leq + \infty \]\n\nand consider the set \( {C}_{\text{inf }}\left( \varphi \right) \) of all real functions \( h \) defined and continuous on \( K \) such that \( h\left( x\right) \leq \varphi \left( ...
Yes
Corollary 2. Let \( \sum {u}_{n}\left( x\right) \) be a series of continuous functions on a compact interval \( K \) . Assume that the series converges simply to a continuous function \( s\left( x\right) \) and that\n\n(10.3)\n\[ \sum {\int }_{K}\left| {{u}_{n}\left( x\right) }\right| {dx} < + \infty \]\n\nThen\n\n(10....
To prove (4), one may assume \( s = 0 \) by replacing \( {u}_{1} \) by \( {u}_{1} - s \), which is again continuous. One may also assume the \( {u}_{n} \) real and then use the decomposition \( {u}_{n} = {u}_{n}^{ + } - {u}_{n}^{ - } \) of \( {\mathrm{n}}^{ \circ }2 \) . These positive functions again satisfy the hypot...
Yes
Theorem 11. Let \( f \) be a regulated function on an interval \( I \) in \( \mathbb{R} \). Then the function \( F \) defined by the relation (3) is continuous and has right and left derivatives equal to \( f\left( {x + }\right) \) and \( f\left( {x - }\right) \) at each point \( x \in I \).
This result, for functions as then understood, is already in Newton in 1665-66 with essentially the same proof, phrased in his language of fluentes and fluxions (Chap. III, \( {\mathrm{n}}^{ \circ }{14} \) ): if \( y \) is the fluent which defines the curve \( \left\lbrack {y = f\left( x\right) \text{in the language of...
No
For \( x > 0 \) and \( s \in \mathbb{C} \), the function\n\n\[ \n{x}^{s} = \exp \left( {s \cdot \log x}\right)\n\]\n\nhas derivative \( s{x}^{s - 1} \) [Chap. IV, formula (10.10)]. The function \( {x}^{s + 1}/\left( {s + 1}\right) \) is therefore, for \( s \neq - 1 \), a primitive of \( {x}^{s} \) .
Whence the formula\n\n(12.7)\n\n\[ \n{\int }_{a}^{b}{x}^{s}{dx} = \frac{{b}^{s + 1} - {a}^{s + 1}}{s + 1}\;\left( {0 < a, b;s \in \mathbb{C}, s \neq - 1}\right)\n\]\n\nalready obtained for \( s \in \mathbb{N} \) by a direct calculation of the integral (Chap. II, \( {\mathrm{n}}^{ \circ }{11} \) ), but now valid for any...
Yes
For \( x > 0 \) the function \( \log x \) has derivative \( 1/x \)
\[ {\int }_{a}^{b}{dx}/x = \log b - \log a\;\left( {0 < a, b}\right) \]
No
The derivative of the function \( \arctan x \) is \( 1/\left( {1 + {x}^{2}}\right) \) ; whence\n\n\[{\int }_{a}^{b}\frac{dx}{1 + {x}^{2}} = \arctan b - \arctan a\]
we must pay attention to the fact that, in this calculation, we take the \
No
For \( c \in \mathbb{C}, c \neq 0 \), the derivative of \( {e}^{cx}/c \) is \( {e}^{cx} \)
we again find the formula\n\[ \n{\int }_{a}^{b}{e}^{cx}{dx} = \left( {{e}^{cb} - {e}^{ca}}\right) /c.\n\]
No
Theorem 13. Let \( I \) and \( J \) be two intervals, \( f \) a continuous function on \( I \times J \) and let \( \varphi ,\psi : I \rightarrow J \) be two differentiable functions. Suppose that \( f \) has a continuous derivative \( {D}_{1}f \) on \( I \times J \) . Then the function\n\n\[ g\left( x\right) = {\int }_...
By subtracting, one need only prove this in the case where \( \varphi \left( x\right) = b \) is constant. Put\n\n\[ F\left( {x, y}\right) = {\int }_{b}^{y}f\left( {x, t}\right) {dt} \]\n\nwhence \( g\left( x\right) = F\left\lbrack {x,\psi \left( x\right) }\right\rbrack \) . Since \( f \) is continuous the FT shows that...
Yes
Theorem 14. Let \( f\left( {x, y}\right) \) be a function defined and continuous on \( I \times J \) where \( I \) and \( J \) are two intervals in \( \mathbb{R} \) . Assume that \( f \) has continuous second derivatives \( {D}_{1}{D}_{2}f \) and \( {D}_{2}{D}_{1}f \) on \( I \times J \) . Then they are equal.
Since it is enough to verify the statement on a neighbourhood of an arbitrary point of \( I \times J \) one may reduce to the case where \( I = \left\lbrack {a, b}\right\rbrack \) and \( J = \left\lbrack {c, d}\right\rbrack \) are compact. The FT applied to the functions \( y \mapsto {D}_{2}{D}_{1}f\left( {x, y}\right)...
Yes
A differentiable function \( f \) is convex on an open interval if and only if its derivative is increasing. A twice differentiable function is convex if and only if \( {f}^{\prime \prime }\left( x\right) \geq 0 \) for every \( x \) .
For if \( {f}^{\prime }\left( x\right) \) exists everywhere, and is increasing, so regulated, then \( f \) is a primitive of \( {f}^{\prime } \) and is therefore convex. If \( {f}^{\prime } \) is differentiable it is increasing if and only if \( {f}^{\prime \prime }\left( x\right) \geq 0 \) everywhere, by the mean valu...
Yes
If \( f \) and \( g \) are regulated functions on a bounded interval \( I \), then
\[ \left| {{\int }_{I}f\left( x\right) g\left( x\right) {dx}}\right| \leq {N}_{p}\left( f\right) {N}_{q}\left( f\right) ,\;{N}_{p}\left( {f + g}\right) \leq {N}_{p}\left( f\right) + {N}_{p}\left( g\right) \]\n\nwhere one puts\n\n\[ {N}_{p}\left( f\right) = {\left( {\int }_{I}{\left| f\left( x\right) \right| }^{p}dx\rig...
No
Take for \( X \) a finite set and \( {\mu }^{ * }\left( f\right) = \sum \left| {f\left( x\right) }\right| \) . One obtains, in more traditional notation, the original versions of the inequalities:
\[ \left| {\sum {x}_{k}{y}_{k}}\right| \leq {\left( \sum {\left| {x}_{k}\right| }^{p}\right) }^{1/p}{\left( \sum {\left| {y}_{k}\right| }^{q}\right) }^{1/q}, \] \[ {\left( \sum {\left| {x}_{k} + {y}_{k}\right| }^{p}\right) }^{1/p} \leq {\left( \sum {\left| {x}_{k}\right| }^{p}\right) }^{1/p} + {\left( \sum {\left| {y}_...
Yes
Like the preceding, but with an infinite set \( X \) and, again, \( {\mu }^{ * }\left( f\right) = \) \( \sum \left| {f\left( x\right) }\right| \leq + \infty \) for every function \( f \) with positive values. If the series \( \sum {\left| {x}_{n}\right| }^{p} \) and \( \sum {\left| {y}_{n}\right| }^{q} \) converge then...
All this assumes \( p, q > 1 \) and \( 1/p + 1/q = 1 \) . The case \( p = q = 2 \) is the Cauchy-Schwarz inequality for series, which may be proved much more easily by passing to the limit starting from the case of a finite sum.
No
Let us calculate the primitive\n\n\[ \int \log \left( x\right) {dx} \]
\[\n\int \log \left( x\right) {dx} = \int \log \left( x\right) {.1} \cdot {dx} = \int \log \left( x\right) \cdot {\left( x\right) }^{\prime } \cdot {dx} = \n\]\n\n\[ \n= \log \left( x\right) x - \int {\log }^{\prime }\left( x\right) {xdx} = x\log x - \int {1dx} \n\]\n\nwhence\n\n(15.3)\n\n\[ \n\int \log \left( x\right)...
Yes
\[ \int {x}^{5}{e}^{x}{dx} = \]
\[ = \int {x}^{5}{\left( {e}^{x}\right) }^{\prime }{dx} = {x}^{5}{e}^{x} - \int {\left( {x}^{5}\right) }^{\prime }{e}^{x}{dx} = {x}^{5}{e}^{x} - 5\int {x}^{4}{e}^{x}{dx} = \]\n\[ = {x}^{5}{e}^{x} - 5{x}^{4}{e}^{x} + {5.4}\int {x}^{3}{e}^{x}{dx} = \]\n\[ = {x}^{5}{e}^{x} - 5{x}^{4}{e}^{x} + {5.4}{x}^{3}{e}^{x} - {5.4.3}...
Yes
[{I}_{n} = \int {x}^{n}\cos x \cdot {dx},\;{J}_{n} = \int {x}^{n}\sin x \cdot {dx}.]
[{I}_{n} = \int {x}^{n}{\sin }^{\prime }x \cdot {dx} = {x}^{n}\sin x - n\int {x}^{n - 1}\sin x \cdot {dx}] and continue. It is more economical to observe that [{I}_{n} + i{J}_{n} = \int {x}^{n}{e}^{ix}{dx}] and to calculate as in Example 2 or, for good measure, to calculate \( \int {x}^{n}{e}^{tx}{dx} \) for every \( t...
Yes
Example 4. Put \( {\log }^{2}x = {\left( \log x\right) }^{2} \) and calculate\n\n\[ \int x.{\log }^{2}x.{dx} \]
\n\[ \int x.{\log }^{2}x.{dx} = \int {\left( \frac{1}{2}{x}^{2}\right) }^{\prime }{\log }^{2}x.{dx} = \frac{1}{2}{x}^{2}{\log }^{2}x - \frac{1}{2}\int {x}^{2}{\left( {\log }^{2}x\right) }^{\prime }{dx} = \]\n\n\[ = \frac{1}{2}{x}^{2}{\log }^{2}x - \frac{1}{2}\int {x}^{2}2\log x\left( {{\log }^{\prime }x}\right) {dx} = ...
Yes
Theorem 16. Let \( f \) be a function defined and of class \( {C}^{n + 1} \) on an interval \( I \) of \( \mathbb{R} \) . Then, for any \( a, x \in I \) ,\n\n\[ f\left( x\right) = f\left( a\right) + {f}^{\prime }\left( a\right) \left( {x - a}\right) + {f}^{\prime \prime }\left( a\right) {\left( x - a\right) }^{\left\lb...
where\n\n\[ {r}_{n}\left( x\right) = {\int }_{a}^{x}{f}^{\left( n + 1\right) }\left( t\right) {\left( x - t\right) }^{\left\lbrack n\right\rbrack }{dt}. \]
Yes
Take \( a = 0 \) and \( f\left( x\right) = \sin x \) . For \( n = {2p} \) one finds \( \left| {{r}_{n}\left( x\right) }\right| \leq \) \( {\left| x\right| }^{2p}/\left( {2p}\right) ! since the successive derivatives are everywhere less than 1 in modulus. On passing to the limit one thus recovers the formula
\[ \sin x = \lim \left\lbrack {x - {x}^{3}/3! + \ldots + {\left( -1\right) }^{p - 1}{x}^{{2p} - 1}/\left( {{2p} - 1}\right) !}\right\rbrack . \]
Yes
Take \( f\left( x\right) = {\left( 1 + x\right) }^{s} \) with \( s \in \mathbb{C} \) and \( - 1 < x \)
Here, by (8),\n\n\[ {r}_{n}\left( x\right) = s\left( {s - 1}\right) \ldots \left( {s - n}\right) \frac{{x}^{n + 1}}{n!}{\int }_{0}^{1}{\left( 1 + ux\right) }^{s - n - 1}{\left( 1 - u\right) }^{n}{dt} \]\n\nor\n\n\[ {r}_{n}\left( x\right) = \frac{s\left( {s - 1}\right) \ldots \left( {s - n}\right) }{n!}{x}^{n + 1}{\int ...
Yes
Theorem 17. Let \( u \) be a real function defined and of class \( {C}^{1} \) on a compact interval \( I = \left\lbrack {a, b}\right\rbrack \), and \( f \) a function defined and continuous on the interval \( J = u\left( I\right) \) . Then\n\n\[ {\int }_{u\left( a\right) }^{u\left( b\right) }f\left( y\right) {dy} = {\i...
For, let \( F \) be a primitive of \( f \) on \( J \), so that the left hand side is equal to \( F\left\lbrack {u\left( b\right) }\right\rbrack - F\left\lbrack {u\left( a\right) }\right\rbrack \) . The function \( G\left( x\right) = F\left\lbrack {u\left( x\right) }\right\rbrack \) is differentiable on \( I \) and \( {...
Yes
Calculate the indefinite integral \( \int {\left( {x}^{2} + 1\right) }^{3}{xdx} \)
Putting \( u\left( x\right) = \) \( {x}^{2} + 1 \) we have \( {u}^{\prime }\left( x\right) {dx} = {2xdx} \), so we need to calculate \( \frac{1}{2}\int u{\left( x\right) }^{3}{u}^{\prime }\left( x\right) {dx} \) ; this is situation (2) with \( f\left( y\right) = {y}^{3} \) . Thus\n\n\[ \int {\left( {x}^{2} + 1\right) }...
Yes
Let \( f \) be a real function of class \( {C}^{1} \) on an interval \( I \), not vanishing on \( I \) . To calculate \( \int {f}^{\prime }\left( x\right) /f\left( x\right) .{dx} \)
one performs the change of variable \( y = f\left( x\right) \), whence \( {dy} = {f}^{\prime }\left( x\right) {dx} \) and\n\n\[\n\int \frac{{f}^{\prime }\left( x\right) }{f\left( x\right) }{dx} = \int {dy}/y\n\]\n\nIt remains to find a primitive of the function \( 1/y \) on the interval \( J = f\left( I\right) \) . Sin...
Yes
To calculate \( \int {dx}/\sin x \) on an interval where the sine function does not vanish, for example on \( \rbrack 0,\pi \lbrack \) .
If one is inspired, or if one has read all the books, one observes that\n\n\[ 1/\sin x = 1/2\sin \left( {x/2}\right) \cos \left( {x/2}\right) = 1/2\tan \left( {x/2}\right) {\cos }^{2}\left( {x/2}\right) = {f}^{\prime }\left( x\right) /f\left( x\right) \]\n\nwhere \( f\left( x\right) = \tan x/2 \) and \( {f}^{\prime }\l...
Yes
To calculate\n\n\\[ \int \frac{{x}^{4} + 1}{\sqrt{\left( {x + 1}\right) \left( {x - 5}\right) }}{dx} \\]
we have to work in the interval \( x < - 1 \), or in the interval \( x > 5 \) to obtain a real result. We have \( \left( {x + 1}\right) \left( {x - 5}\right) = {\left( x - 3\right) }^{2} - 4 \), which suggests the change of variable \( x = 3 + {2y} \), whence \( {dx} = {2dy} \) and reduction to\n\n\\[ \int \frac{{\left...
Yes
Theorem 18. (i) Let \( f \) be a positive regulated function defined on an interval \( X = \\left( {a, b}\\right) \) . Then \( f \) is integrable on \( X \) if and only if the integrals over the compact subsets \( K \subset X \) are bounded above; and then\n\n(22.1)\n\n\[ \n{\\int }_{X}f\\left( x\\right) {dx} = \\matho...
To prove point (i) we observe that, \( f \) being positive, \( s\\left( K\\right) \) is an increasing function of \( K \) :\n\n\[ \nK \subset {K}^{\\prime } \Rightarrow s\\left( K\\right) \leq s\\left( {K}^{\\prime }\\right) \n\]\n\nThe arguments of Chap. II, \( {\\mathrm{n}}^{ \u2060 }9 \) on increasing sequences tran...
No
Let \( f \) be a bounded regulated function on an interval \( X \), and \( \mu \) an absolutely integrable regulated function on \( X \) . Then the function \( f\left( x\right) \mu \left( x\right) \) is absolutely integrable on \( X \) and \[ \int \left| {f\left( x\right) \mu \left( x\right) {dx}}\right| \leq \parallel...
Obvious. As we shall do on various occasions in the rest of this \( § \), we have used the \( \int \) sign to denote integrals extended over \( X \) .
No
Corollary 2. Let \( f \) and \( g \) be two regulated square integrable functions on an interval \( X \) ; then the function \( f\left( x\right) \overline{g\left( x\right) } \) is absolutely integrable on \( X \) and\n\n\[{\left| \int f\left( x\right) \overline{g\left( x\right) }dx\right| }^{2} \leq \int {\left| f\left...
One replaces \( f \) and \( g \) by \( \left| f\right| \) and \( \left| g\right| \), writes the Cauchy-Schwarz inequality for every compact interval \( K \subset X \) and notes that the left hand side is, for any \( K \), majorised by the right hand side of the inequality to be established, whence the result on passage...
No
Consider Euler's ubiquitous Gamma function\n\n\[ \Gamma \left( s\right) = {\int }_{0}^{+\infty }{e}^{-x}{x}^{s - 1}{dx}. \]
Absolute convergence at infinity is automatic, but, at 0, requires \( \operatorname{Re}\left( s\right) > 0 \) . An integration by parts \( {}^{43} \) then shows that\n\n\[ \Gamma \left( {s + 1}\right) = {\int }_{0}^{+\infty }{e}^{-x}{x}^{s}{dx} = - {\left. {e}^{-x}{x}^{s}\right| }_{0}^{+\infty } + s{\int }_{0}^{+\infty...
Yes
\[ B\left( {x, y}\right) = {\int }_{0}^{1}{t}^{x - 1}{\left( 1 - t\right) }^{y - 1}{dt} \] where \( x \) and \( y \) are a priori complex (and rational for him). Absolute convergence on a neighbourhood of 0 requires \( \operatorname{Re}\left( x\right) > 0 \) and, on a neighbourhood of \( 1,\operatorname{Re}\left( y\rig...
\[ B\left( {x, y}\right) = B\left( {y, x}\right) \] (change of variable \( t \mapsto 1 - t \) ). The change of variable \( t \mapsto {\sin }^{2}t \) shows that \[ B\left( {x, y}\right) = 2{\int }_{0}^{\pi /2}{\sin }^{{2x} - 1}t \cdot {\cos }^{{2y} - 1}t \cdot {dt} \] We shall see later \( \left( {\mathrm{n}}^{ \circ }\...
No
Theorem 19 (Poor man’s dominated convergence). Let \( \left( {f}_{n}\right) \) be a sequence of regulated functions, absolutely integrable on an interval \( X \subset R \) . Assume that\n\n(i) the \( {f}_{n} \) converge to a limit \( f \) uniformly on every compact \( K \subset X \) ,\n\n(ii) there exists a positive fu...
First of all, it is clear that \( f \), being regulated like the \( {f}_{n} \), is absolutely integrable on \( X \) since \( \left| {f\left( x\right) }\right| \leq p\left( x\right) \) for every \( x \) . Since \( p \) is positive and integrable, for every \( r > 0 \) there exists a compact interval \( K \subset X \) su...
Yes
Consider the function\n\n\\[ \Gamma \\left( s\\right) = {\\int }_{0}^{+\\infty }{e}^{-x}{x}^{s - 1}{dx},\\;\\operatorname{Re}\\left( s\\right) > 0, \\]\n\nagain, and observe that\n\n\\[ {e}^{-x}{x}^{s - 1} = \\lim {\\left( 1 - x/n\\right) }^{n}{x}^{s - 1}. \\]
We cannot just bluntly apply Theorem 19 since the functions on the right hand side are not integrable between 0 and \\( + \\infty \\) : convergence at 0 presupposes \\( \\operatorname{Re}\\left( s\\right) > 0 \\) and convergence at infinity \\( \\operatorname{Re}\\left( s\\right) < - n \\) . For \\( x < n \\) we always...
Yes
Theorem 20. Let \( \sum {u}_{n}\left( x\right) \) be a series of absolutely integrable regulated functions on an interval \( X \) . Assume that (i) the series converges uniformly on every compact \( K \subset X \) ; (ii) there exists a positive function \( p\left( x\right) \), integrable on \( X \), such that \( \sum \...
The hypothesis (i) shows that the partial sums \( {s}_{n}\left( x\right) \) converge uniformly on every compact \( K \subset X \) ; since (ii) shows that \( \left| {{s}_{n}\left( x\right) }\right| \leq p\left( x\right) \), one need only apply the preceding theorem to the \( {s}_{n} \).
Yes
Theorem 21. Let \( X \) be an interval and \( {u}_{n}\left( x\right) \) a series of regulated functions which converges normally on every compact \( K \subset X \) . Assume that\n\n\[ \sum \int \left| {{u}_{n}\left( x\right) }\right| {dx} < + \infty \]\n\nThen the function \( s\left( x\right) = \sum {u}_{n}\left( x\rig...
Let us put,\n\n\[ {m}_{I}\left( f\right) = {\int }_{I}f\left( x\right) {dx} \]\n\nand consider a compact interval \( K \subset X \) . Since the given series converges normally on \( K \) (\
No
Theorem 22. Let \( X \) be an interval, \( H \) a compact subset of \( \mathbb{C}, f \) a function defined and continuous on \( X \times H \) and \( \mu \) a function defined and regulated in \( X \) . Assume that there is a positive function \( p \) on \( X \) such that \( \left| {f\left( {x, y}\right) }\right| \leq p...
Hypotheses (i) and (iii) above are clearly satisfied by \( f\left( {x, y}\right) \mu \left( x\right) \) when \( y \) tends to a \( b \in H \) . If \( K \) is a compact subset of \( X \), then the function \( f \) is uniformly continuous on the compact \( K \times H \) ; consequently, the hypothesis (ii) is satisfied al...
No
Theorem 23. Let \( f \) be a positive regulated function, defined for \( x \geq a > - \infty \) , decreasing, and tending to 0 at infinity. Then the integral\n\n\[ \varphi \left( y\right) = {\int }_{a}^{+\infty }f\left( x\right) \sin \left( {2\pi xy}\right) {dx} \]\n\nconverges for any \( y \neq 0 \), and is a continuo...
To see this, assume \( y > 0 \) and perform the change of variable \( {2xy} = u \), \n\nwhence\n\[ {2y\varphi }\left( y\right) = {\int }_{2ay}^{+\infty }f\left( {u/{2y}}\right) \sin \left( {\pi u}\right) {du}. \]\n\nConvergence is clear, and (3) can now be written\n\n(24.4)\n\[ \left| {{2y\varphi }\left( y\right) - {\i...
Yes
Theorem 24. Let \( X \) and \( J \) be two intervals in \( \mathbb{R} \), let \( \mu \) be a regulated function on \( X \) and \( f \) a function defined and continuous on \( X \times J \). Assume that\n\n(i) the integral\n\n\[ g\left( y\right) = {\int }_{X}f\left( {x, y}\right) \mu \left( x\right) {dx} \]\n\nconverges...
Example 1. If \( X = Y
No
If \( X = Y = \mathbb{R} \), if \( \mu \) is an absolutely integrable regulated function on \( \mathbb{R} \) and if \( f\left( {x, y}\right) = {e}^{-{2\pi ixy}} \), then the function \( g\left( y\right) \) is just the Fourier transform \( \widehat{\mu } \) of \( \mu \) .
Here\n\n\[ \n{D}_{2}f\left( {x, y}\right) = - {2\pi ix}{e}^{-{2\pi ixy}} \n\]\n\nand so \( \left| {{D}_{2}f\left( {x, y}\right) }\right| = {2\pi }\left| x\right| = p\left( x\right) \), and this is clearly the smallest positive function which dominates \( x \mapsto {D}_{2}f\left( {x, y}\right) \) for a (or for all) \( y...
Yes
In particular choose \( \mu \left( x\right) = \exp \left( {-\pi {x}^{2}}\right) \), an integrable function on \( \mathbb{R} \) since it decreases at infinity more rapidly than \( {\left| x\right| }^{-n} \) for any \( n > 0 \) . We have \( - {2\pi ix\mu }\left( x\right) = i{\mu }^{\prime }\left( x\right) \), whence, int...
\[ {\widehat{\mu }}^{\prime }\left( y\right) = i{\int }_{-\infty }^{+\infty }{\mu }^{\prime }\left( x\right) \exp \left( {-{2\pi ixy}}\right) {dx} = - {2\pi y}{\int }_{-\infty }^{+\infty }\mu \left( x\right) \exp \left( {-{2\pi ixy}}\right) {dx} \] since the integrated-out part is zero because of the decrease of \( \mu...
Yes
If \( \mu \) is a regulated function on the closed interval \( \lbrack 0, + \infty \lbrack \) and is \( O\left( {t}^{N}\right) \) at infinity for some \( N \), its Laplace transform or complex Fourier transform\n\n\[ \n{L}_{\mu }\left( z\right) = {\int }_{0}^{+\infty }{e}^{2\pi itz}\mu \left( t\right) {dt} \n\]\n\nis d...
Here \( f\left( {t, z}\right) = {e}^{2\pi itz} \), whence \( \left| {{f}^{\prime }\left( {t, z}\right) }\right| = {2\pi t}{e}^{-{2\pi ty}} \) . Since every compact subset \( H \) of \( U \) is contained in a half plane \( \operatorname{Im}\left( z\right) \geq \sigma > 0 \), we have, in \( H \) , that \( \left| {{f}^{\p...
Yes
The function \( \Gamma \left( s\right) = \int {e}^{-x}{x}^{s - 1}{dx} \) is holomorphic in the half plane \( \operatorname{Re}\left( s\right) > 0 \) where it is defined.
It is clear that\n\n(i) the function \( s \mapsto {e}^{-x}{x}^{s - 1} = {e}^{-x}\exp \left\lbrack {\left( {s - 1}\right) \log x}\right\rbrack \) is holomorphic for every \( x > 0 \) since it is the composite of two holomorphic functions;\n\n(ii) its complex derivative \( {}^{48}{e}^{-x}{x}^{s - 1}\log x \) is continuou...
Yes
\[ \Gamma \left( s\right) = {\int }_{0}^{1}{e}^{-x}{x}^{s - 1}{dx} + {\int }_{1}^{+\infty }{e}^{-x}{x}^{s - 1}{dx}. \]
The second integral converges for any \( s \in \mathbb{C} \) . So, as in the preceding example, is a holomorphic function of \( s \) in all of \( \mathbb{C} \) . In the first integral, term-by-term integration of the exponential series gives, for \( \operatorname{Re}\left( s\right) > 0 \) ,\n\n\[ {\int }_{0}^{1}{e}^{-x...
Yes
Theorem 25 (Poor man’s Lebesgue-Fubini). Let \( X \) and \( Y \) be two intervals and \( f \) a continuous function on \( X \times Y \) . Suppose that the following conditions are satisfied:\n\n(i) for every compact \( K \subset X \) there exists a positive integrable function \( {q}_{K}\left( y\right) \) on \( Y \) su...
In what follows we shall put\n\n(26.3)\n\n\[{g}_{J}\left( x\right) = {\int }_{J}f\left( {x, y}\right) {dy},\;{h}_{I}\left( y\right) = {\int }_{I}f\left( {x, y}\right) {dx}\]\nfor any intervals \( I \subset X \) and \( J \subset Y \) . We shall also employ the notation \( {m}_{I} \) to denote an integral over \( I \) .\...
Yes
First note that if the continuous functions \( f\left( x\right) \) and \( g\left( y\right) \) are defined and absolutely integrable on the intervals \( X \) and \( Y \) then the function \( f\left( x\right) g\left( y\right) \) is absolutely integrable on \( X \times Y \), and clearly\n\n\[ \n{\iint }_{X \times Y}f\left...
If, for each \( x \), one effects the change of variable \( y = \left( {{u}^{-1} - 1}\right) x \) in the \( y \) -integration, one finds\n\n(26.15)\n\[ \n\Gamma \left( a\right) \Gamma \left( b\right) = {\int }_{0}^{+\infty }{dx}{\int }_{0}^{1}{e}^{-u/x}{x}^{a + b - 1}{\left( 1 - u\right) }^{b - 1}{u}^{-b - 1}{du} = \n\...
Yes
Consider on \( \mathbb{R} \) a function \( u\left( x\right) \) which is regulated, positive, with total integral 1, and put \( {u}_{n}\left( x\right) = {nu}\left( {nx}\right) \). The condition (D 1) is satisfied, also (D2) (change of variable \( {nx} = y \) in the integral) and condition (D 3) is satisfied because \[ {...
If \( u \) is of compact support the function \( {u}_{n}\left( x\right) = {nu}\left( {nx}\right) \) is zero for \( \left| x\right| \geq A/n \), i.e. outside an ever-shrinking interval with centre 0 ; the factor \( n \) in its definition shows that, on the other hand, it takes very large values on a neighbourhood of 0 ,...
Yes
Theorem 26. Let \( \left( {u}_{n}\right) \) be a Dirac sequence. For every function \( f \) defined and continuous on \( \mathbb{R} \) one has\n\n\[ f\left( x\right) = \lim \int f\left( {x - y}\right) {u}_{n}\left( y\right) {dy} \]\n\nuniformly on every compact subset of \( \mathbb{R} \) if \( f \) is bounded or just i...
To establish (8) for \( f \) bounded it is enough to apply the lemma to the function \( y \mapsto f\left( {x - y}\right) \) . The little calculation in Dirac’s lemma shows moreover that\n\n\[ \left| {f\left( x\right) \int {u}_{n}\left( y\right) {dy}-\int f\left( {x - y}\right) {u}_{n}\left( y\right) {dy}}\right| \leq \...
Yes
Theorem 27. For every function \( f \) defined and continuous on \( \mathbb{R} \) there exists a sequence \( {f}_{n} \) of \( {C}^{\infty } \) functions which converges to \( f \) uniformly on every compact subset of \( \mathbb{R} \) . If \( f \) is of compact support one may assume that the \( {f}_{n} \) are zero outs...
Obvious: one applies Theorem 26 to a Dirac sequence formed by functions of \( \mathcal{D} \), bearing in mind what we want to establish. If \( f \) is zero for \( \left| x\right| > A \) and if one assumes, for example, that the \( {\varphi }_{n} \) vanish for \( \left| x\right| \geq 1/n \), it is clear that the \( {f}_...
No
Theorem 28 (Weierstrass, 1885). Let \( f \) be a real function defined and continuous on a compact interval \( K \subset \mathbb{R} \). Then there exists a sequence of polynomials which converges to \( f \) uniformly on \( K \).
It is enough to observe that \( f \) can be extended to a continuous function on all \( \mathbb{R} \), zero for \( \left| x\right| \) large: complete the graph of \( f \) by linear functions.
No
Theorem 29 (Emile Borel,1895). For every sequence \( \left( {a}_{n}\right) \) of complex numbers there exists an indefinitely differentiable function \( f \) of compact support on \( \mathbb{R} \) such that \( {f}^{\left( n\right) }\left( 0\right) = {a}_{n} \) for every \( n \in \mathbb{N} \) .
Our first move, faced by this theorem, is to put\n\n(29.1)\n\n\[ f\left( x\right) = \sum {a}_{n}{x}^{n}/n! \]\n\nin accordance with Maclaurin's formula. Bad idea: the series has every chance of diverging for \( x \neq 0 \) .\n\nAll the same, (1) contains the germ of an idea for a proof. The function \( {a}_{n}{x}^{n}/n...
Yes
Choose a function \( \mu \left( x\right) \), integrable (in the usual sense) on an interval \( K \subset \mathbb{R} \), and put\n\n\[ \mu \left( f\right) = \int f\left( x\right) \mu \left( x\right) {dx} \]\n\nfor every \( f \in L\left( K\right) \) .
Linearity is obvious and continuity follows from the inequalities\n\n\[ \left| {\mu \left( f\right) }\right| \leq \int \left| {f\left( x\right) }\right| \cdot \left| {\mu \left( x\right) }\right| {dx} \leq \parallel f{\parallel }_{K} \cdot \int \left| {\mu \left( x\right) }\right| {dx}. \]\n\nHere \( \parallel \mu \par...
Yes
Choose a countable set \( D \) of points of \( K \) and, for every \( \xi \in D \), a number \( c\left( \xi \right) \in \mathbb{C} \) ; assuming \( \sum \left| {c\left( \xi \right) }\right| < + \infty \) one may define
\[ \mu \left( f\right) = \sum c\left( \xi \right) f\left( \xi \right) \] for every continuous function \( f \) on \( K \), the series being taken over \( D \) . No hypothesis on the compact set \( K \) is necessary here.
Yes
Example 3. Take \( K = A \times B \) where \( A \) and \( B \) are compact intervals in \( \mathbb{R} \) and put\n\n\[ m\left( f\right) = {\iint }_{A \times B}f\left( {x, y}\right) {dxdy} \]
for every continuous function \( f \) on \( K\left( {\mathrm{n}}^{ \circ }\right. \) 9, Theorem 10).
No
Take for \( X \) the open interval \( \rbrack 0, + \infty \lbrack \) and\n\n\[ \mu \left( f\right) = {\int }_{0}^{+\infty }f\left( x\right) {dx}/x \]
There is no problem with convergence for \( f \in L\left( X\right) \) since \( f\left( x\right) \) is zero on a neighbourhood of 0 and for \( x \) large. If \( f \) is zero outside \( K = \left\lbrack {u, v}\right\rbrack \) then\n\n\[ \left| {\mu \left( f\right) }\right| \leq \left( {\log v - \log u}\right) \parallel f...
Yes
If one replaces the open interval \( \rbrack 0, + \infty \lbrack \) by the closed interval \( \lbrack 0, + \infty \lbrack \) the formula (6) is no longer meaningful, since, in this case, a function \( f \in L\left( X\right) \) is required to be zero for \( x \) large but not on a neighbourhood of 0, so allowing every c...
But one can replace \( 1/x \) by a function that poses no problem at the origin, and, for example, put\n\n(31.7)\n\n\[ \mu \left( f\right) = {\int }_{0}^{+\infty }f\left( x\right) {x}^{s}{dx}\;\text{ with }\operatorname{Re}\left( s\right) > - 1. \]\n\nIf \( f \) is zero outside \( K = \left\lbrack {0, v}\right\rbrack \...
Yes
Choose a compact interval \( K \subset X \), a measure \( \mu \) on \( K \) and consider the linear form \( f \mapsto \int f\left( x\right) {d\mu }\left( x\right) \), where one integrates over \( K \), so involving only the values of \( f \) on this fixed compact set.
This example shows that a measure on \( K \) may also be considered as a measure on \( X \) : all the mass is supported by \( K \) .
No
Example 4. For \( X = \mathbb{R} \) put\n\n(31.8)\n\n\[ \mu \left( f\right) = \sum f\left( n\right) \]\n\nsumming over \( \mathbb{Z} \). If \( f \) is zero outside a compact \( K \), only the \( n \in K \) count, whence \( \left| {\mu \left( f\right) }\right| \leq {M}_{K}\left( \mu \right) \parallel f{\parallel }_{X} \...
\[ \left| {\mu \left( f\right) }\right| \leq {M}_{K}\left( \mu \right) \parallel f{\parallel }_{X}\;\text{ where }\;{M}_{K}\left( \mu \right) = \mathop{\sum }\limits_{{\xi \in K}}\left| {c\left( \xi \right) }\right| \]\n\nsince the \( \xi \notin K \) do not appear in (9). If the total series \( \sum \left| {c\left( \xi...
Yes
Theorem 31. Let \( X \) and \( Y \) be two intervals, \( \mu \) and \( \nu \) positive measures on \( X \) and \( Y \), and \( \lambda \) the product measure on \( X \times Y \) . Let \( \varphi \) be an lsc (resp. usc) function on \( X \times Y \) which is \( {}^{71} \) the upper (resp. lower) envelope of the \( f \in...
We shall examine the case of an usc function, the other being trivial to deduce from it: multiply the function by -1 . As always, the two crucial points will be that (a) a lower envelope of continuous functions is usc; (b) one may calculate the integral of an usc function from any decreasing philtre of continuous funct...
Yes
Every absolutely integrable function \( f \) on every compact interval of \( \mathbb{R} \) (for example \( \log \left| x\right| \) despite its singularity at the origin) defines a distribution \( {}^{76} \) which is in fact a measure
\[ {T}_{f}\left( \varphi \right) = \int \varphi \left( x\right) f\left( x\right) {dx} \]
No
Choose an \( a \in \mathbb{R} \), an integer \( k \in \mathbb{N} \) and put\n\n\[ T\left( \varphi \right) = {\varphi }^{\left( k\right) }\left( a\right) \]
For \( k = 0 \) one obtains the Dirac measure at the point \( a \), denoted by \( {\delta }_{a} \) or \( {\varepsilon }_{a} \) :\n\n\[ {\delta }_{a}\left( \varphi \right) = \varphi \left( a\right) \]
No
Take for \( T \) the Dirac measure\n\n\[ T\left( \varphi \right) = \varphi \left( 0\right) \]\n\nWe find \( {T}^{\prime }\left( \varphi \right) = - {\varphi }^{\prime }\left( 0\right) \) in accordance with Dirac’s baroque formula to which we alluded at the beginning of \( {\mathrm{n}}^{ \circ }{27} \) .
One might, like Dirac himself, continue:\n\n\[ {T}^{\prime \prime }\left( \varphi \right) = + {\varphi }^{\prime \prime }\left( 0\right) ,\;{T}^{\prime \prime \prime }\left( \varphi \right) = - {\varphi }^{\prime \prime \prime }\left( 0\right) ,\text{ etc. } \]
No
Take for \( f \) the function equal to 1 for \( x > 0 \) and to 0 for \( x < 0 \). Then \[ {T}_{f}\left( \varphi \right) = {\int }_{0}^{+\infty }\varphi \left( x\right) {dx} \]
\[ {T}_{f}^{\prime }\left( \varphi \right) = - {T}_{f}\left( {\varphi }^{\prime }\right) = - {\int }_{0}^{+\infty }{\varphi }^{\prime }\left( x\right) {dx} = \varphi \left( 0\right) \] since the primitive \( \varphi \) of \( {\varphi }^{\prime } \) is zero for \( x \) large. In other words, the derivative of the distri...
Yes
Let us multiply term-by-term the relations\n\n\\[ \n{e}^{x} = 1 + x + {x}^{2}/2 + O\\left( {x}^{3}\\right) ,\\;\\sin x = x - {x}^{3}/6 + O\\left( {x}^{5}\\right) \n\\]\n\nvalid for \\( x \\rightarrow 0 \\) ; calculating \\( \\grave{a}{la} \\) Newton one finds
\n\n\\[ \n{e}^{x}\sin x = \\left( {1 + x + {x}^{2}/2}\\right) \\left( {x - {x}^{3}/6}\\right) + \\left( {1 + x + {x}^{2}/2}\\right) O\\left( {x}^{5}\\right) + \n\\]\n\n\\[ \n+ \\left( {x - {x}^{3}/6}\\right) O\\left( {x}^{3}\\right) + O\\left( {x}^{3}\\right) O\\left( {x}^{5}\\right) = \n\\]\n\n\\[ \n= x + {x}^{2} + {x...
Yes
When \( x \rightarrow 0 \) , what is the expansion of \( {\left( {x}^{4} + {x}^{2}\right) }^{1/3} \)?
\[ {\left( {x}^{4} + {x}^{2}\right) }^{1/3} = {x}^{2/3}{\left( 1 + {x}^{2}\right) }^{1/3} = {x}^{2/3}\left\lbrack {1 + {x}^{2}/3 - {x}^{4}/9 + O\left( {x}^{6}\right) }\right\rbrack \] by the binomial series, whence \[ {\left( {x}^{4} + {x}^{2}\right) }^{1/3} = {x}^{2/3} + {x}^{8/3}/3 - {x}^{{14}/3}/9 + O\left( {x}^{{20...
Yes
Consider the ratio\n\n\[ \frac{{x}^{2} - x + \log x}{{x}^{2} - {\left( \log x\right) }^{2}} \]\n\n as \( x \) tends to \( + \infty \) .
In the numerator, \( x \) and \( \log x \) are \( o\left( {x}^{2}\right) \), so it is \( \sim {x}^{2} \). In the denominator, \( \log x \) is \( o\left( x\right) \), so \( {\left( \log x\right) }^{2} \) is \( o\left( {x}^{2}\right) \), so that the denominator also is \( \sim {x}^{2} \). The fraction we are considering ...
Yes
We seek a truncated expansion of order 1 at \( x = 0 \) of the function \( h\left( x\right) = {e}^{x}/{x}^{2}\sin x \).
Here \( h\left( x\right) \sim {x}^{-3} \), so that a relation of the form \( h\left( x\right) = \) \( p\left( x\right) + o\left( x\right) \) can be written as \( {x}^{3}h\left( x\right) = q\left( x\right) + o\left( {x}^{4}\right) \) . We have to find a truncated expansion of order 4 for\n\n\[ {x}^{3}h\left( x\right) = ...
Yes
Theorem 2. If \( f \) is a continuous function on \( \mathbb{T} \) such that \( \sum \left| {\widehat{f}\left( n\right) }\right| < + \infty \) , then \[ f\left( u\right) = \sum \widehat{f}\left( n\right) {u}^{n}\;\text{ for every }u \in \mathbb{T}. \]
Let us denote the right hand side by \( g\left( u\right) \) . This is the sum of an absolutely convergent Fourier series, whence (Chap. V, \( {\mathrm{n}}^{ \circ }5 \) ) \( \widehat{g}\left( n\right) = \widehat{f}\left( n\right) \) for every \( n \) . Putting \( f = g + h \), we see that all the Fourier coefficients o...
No
Theorem 3 (Parseval-Bessel \( {}^{14} \) ). Let \( f \) be a regulated periodic function. Then the series \( \sum {\left| \widehat{f}\left( n\right) \right| }^{2} \) is convergent and\n\n\[ \sum {\left| \widehat{f}\left( n\right) \right| }^{2} = \parallel f{\parallel }_{2}^{2} = \int {\left| f\left( u\right) \right| }^...
\[ \mathop{\lim }\limits_{{N \rightarrow \infty }}{\int }_{0}^{1}{\left| f\left( t\right) - \mathop{\sum }\limits_{{\left| n\right| \leq N}}\widehat{f}\left( n\right) {\mathbf{e}}_{n}\left( t\right) \right| }^{2}{dt} = 0. \]
Yes
Consider the periodic function equal to \( t \) for \( \left| t\right| < \frac{1}{2} \) ; the values at the end-points are immaterial. Integrating by parts, we have, for \( n \neq 0 \) ,
\[ \widehat{f}\left( n\right) = {\int }_{-\frac{1}{2}}^{\frac{1}{2}}t\overline{{\mathbf{e}}_{n}\left( t\right) }{dt} = {\left. \frac{t\overline{{\mathbf{e}}_{n}\left( t\right) }}{-{2\pi in}}\right| }_{-\frac{1}{2}}^{\frac{1}{2}} + \frac{1}{2\pi in}{\int }_{-\frac{1}{2}}^{\frac{1}{2}}\overline{{\mathbf{e}}_{n}\left( t\r...
Yes
Consider the function of period 1 such that\n\n\[ f\left( t\right) = {e}^{2\pi izt}\;\text{ for }\left| t\right| < \frac{1}{2}, \]\n\nwhere \( z \) is a complex number, not an integer, since otherwise the interest of the problem evaporates.
We have\n\n\[ \widehat{f}\left( n\right) = {\int }_{-\frac{1}{2}}^{\frac{1}{2}}{e}^{{2\pi i}\left( {z - n}\right) t}{dt} = {\left. \frac{{e}^{{2\pi i}\left( {z - n}\right) t}}{{2\pi i}\left( {z - n}\right) }\right| }_{-\frac{1}{2}}^{\frac{1}{2}} \]\n\nsince, for every \( \lambda \in \mathbb{C} \), the derivative of \( ...
Yes
Expansion of \( \cot z \) as a series of rational fractions.
Consider the function of period 1 on \( \mathbb{R} \) given by\n\n\[ f\left( t\right) = \cos {2\pi zt}\;\text{ for }\left| t\right| < \frac{1}{2}, \]\n\nwhere \( z \in \mathbb{C} \) is not a rational integer, for otherwise there would be no problem. Since \( f\left( {-\frac{1}{2}}\right) = f\left( \frac{1}{2}\right) \)...
Yes
The Bernoulli polynomials are defined by the recurrence relations\n\n\[ \n{B}_{0}\left( x\right) = 1,\;{B}_{k}^{\prime }\left( x\right) = k{B}_{k - 1}\left( x\right) \n\]\n\nand by the condition\n\n\[ \n{B}_{k}\left( 0\right) = {B}_{k}\left( 1\right) \;\text{ for }k \geq 2. \n\]
The inventor was not acquainted with Fourier series, but condition (20) is exactly what one needs to transform the \( {B}_{k} \), for \( k \geq 2 \), into continuous periodic functions \( {B}_{k}^{ * } \), by putting\n\n\[ \n{B}_{k}^{ * }\left( t\right) = {B}_{k}\left( t\right) \;\text{ for }0 \leq t \leq 1 \n\]\n\nas ...
Yes
For every regulated periodic function \( f \) the arithmetic means of the partial sums of the Fourier series of \( f \) converge to \( \frac{1}{2}\lbrack f\left( {t + }\right) + f\left( {t - }\right) \rbrack \) for any \( t \) .
To establish the first one writes, as in (11.13), (12.7) \[ f \star {F}_{N}\left( t\right) - \frac{1}{2}\left\lbrack {f\left( {t + }\right) + f\left( {t - }\right) }\right\rbrack = \] \[ = \int \left\lbrack {f\left( {t + s}\right) - f\left( {t + }\right) }\right\rbrack {F}_{N}\left( s\right) {ds} + \int \left\lbrack {f...
Yes
Theorem 10 (Cauchy, 1831). Let \( f \) be a holomorphic function in an open set \( U \) in \( \mathbb{C} \) . Then \( f \) is analytic in \( U \) and, for every \( a \in U \), the Taylor series of \( f \) at a converges and represents \( f \) in the largest open disc with centre a contained in \( U \) .
It is enough, in the preceding arguments, to replace the disc \( \left| z\right| < R \) by the largest disc \( \left| {z - a}\right| < R \) in question or, if one prefers, to consider the function \( f\left( {a + z}\right) \) . Now the only power series that can possibly represent \( f \) on a neighbourhood of \( a \) ...
No
Theorem 11. Let \( f \) be a holomorphic function in a connected open set \( U \) . Then \( f \) is constant if at a point of \( U \) it has either a local maximum or a non zero local minimum.
The case of a local minimum reduces to the preceding case on considering the function \( 1/f \) : this is defined and holomorphic on a neighbourhood of a local minimum of \( f \) and has a local maximum there; \( 1/f \) (and so \( f \) ) is thus constant on a disc, so \( f \) is constant on \( U \) .
Yes
Corollary 1. Let \( G \) be a bounded domain in \( \mathbb{C}, K \) its closure, \( F = K - G \) its frontier and \( f \) a function defined and continuous in \( K \) and holomorphic in \( G \) . Then\n\n(15.4)\n\n\[ \parallel f{\parallel }_{G} = \parallel f{\parallel }_{K} = \parallel f{\parallel }_{F} \]
Since \( G \) is bounded, \( K \) is bounded and closed, hence compact. The continuous function \( \left| {f\left( z\right) }\right| \) therefore attains its maximum at a point \( a \in K \) . If \( a \in G \), Theorem 5 shows that \( f \) is constant in \( G \), hence in \( K \), and the corollary is obvious. If \( f ...
Yes
Corollary 2. Let \( G \) be a bounded domain and \( \left( {f}_{n}\right) \) a sequence of functions defined and continuous on the closure \( K \) of \( G \) and holomorphic in \( G \) . Assume that the \( {f}_{n} \) converge uniformly on the boundary \( F \) of \( G \) to a limit function. Then the \( {f}_{n} \) conve...
Consider the functions \( {f}_{pq} = {f}_{p} - {f}_{q} \) . Cauchy’s criterion for uniform convergence shows that, for every \( r > 0 \), one has \( {\begin{Vmatrix}{f}_{pq}\end{Vmatrix}}_{F} \leq r \) for \( p \) and \( q \) large, and thus (Corollary 1) \( {\begin{Vmatrix}{f}_{pq}\end{Vmatrix}}_{K} \leq r \) . The \(...
No
Corollary 3 ((H. A.) Schwarz’ lemma). Let \( f \) be a function holomorphic and bounded on a disc \( \left| z\right| < R \) and having a zero of order \( p \) at the origin. Then\n\n\[ \left| {f\left( z\right) }\right| \leq M{\left| z/R\right| }^{p}\;\text{ where }M = \sup \left| {f\left( z\right) }\right| .
The assumption about \( f \) implies that \( f\left( z\right) = {z}^{p}g\left( z\right) \) where \( g \) is, like \( f \) , the sum of a power series in \( \left| z\right| < R \) . The relation \( \left| {{z}^{p}g\left( z\right) }\right| \leq M \) shows that \( \left| {g\left( z\right) }\right| \leq M/{r}^{p} \) for \(...
Yes
Theorem 13 (Laurent). Let \( f \) be a holomorphic function in an annulus \( C : {R}_{1} < \left| z\right| < {R}_{2} \). Then we have a series expansion\n\n(16.3)\n\n\[ f\left( z\right) = \mathop{\sum }\limits_{{n \in \mathbb{Z}}}{a}_{n}{z}^{n}\;\text{ with }\;{a}_{n} = {r}^{-n}{\int }_{\mathbb{T}}f\left( {ru}\right) {...
An expansion of this type is called a Laurent series; it is the sum of a power series in \( z \) and of a power series in \( 1/z \). The first converges at least for \( \left| z\right| < {R}_{2} \) and the second for \( \left| z\right| > {R}_{1} \) since a power series necessarily converges on a disc. This allows us to...
Yes
Theorem 14 (Liouville). Let \( f \) be a holomorphic function of period 1 in an open strip \( B : a < \\operatorname{Im}\\left( z\\right) < b \) . We then have a series expansion\n\n\[ f\\left( z\\right) = \\sum {a}_{n}{e}^{2\\pi inz} \]\n\nwhich converges normally on every closed strip \( {B}^{\\prime } \\subset B \) ...
Conversely, if a complex Fourier series, i.e. of the form (5), converges absolutely in a strip \( a < \\operatorname{Im}\\left( z\\right) < b \), the preceding argument shows that the series converges normally on every closed strip (and so on every compact set) contained in the given open strip. Theorem 17 below will s...
No
Theorem 15 (Liouville). Let \( f \) be an entire function such that\n\n(18.1)\n\n\[ f\left( z\right) = O\left( {z}^{p}\right) \;\text{ when }\;\left| z\right| \rightarrow + \infty ,\]\n\nwhere \( p \) is an integer \( \geq 0 \) . Then \( f \) is a polynomial of degree \( \leq p \) . In particular, a bounded entire func...
By Theorem 10 one has an expansion \( f\left( z\right) = \sum {a}_{n}{z}^{n} \) valid for any \( z \) . The relation (14.10) then shows that, for every \( n \) ,\n\n(18.2)\n\n\[ \left| {a}_{n}\right| {r}^{n} \leq {M}_{f}\left( r\right) \]\n\nwhere \( {M}_{f}\left( r\right) \) is the upper bound of \( \left| {f\left( z\...
Yes