Q
stringlengths
4
3.96k
A
stringlengths
1
3k
Result
stringclasses
4 values
Proposition 12.1.8. The space \( {L}_{p} \) is finitely representable in \( {\ell }_{p} \) for \( 1 \leq p < \infty \) .
Proof. For each \( n \) let \( {E}_{n} = \left\lbrack {{\chi }_{\left( \left( k - 1\right) /{2}^{n}, k/{2}^{n}\right) } : 1 \leq k \leq {2}^{n}}\right\rbrack \) be the subspace spanned in \( {L}_{p} \) by the characteristic functions on dyadic intervals. Since \( {E}_{n} \) is isometrically isomorphic to a subspace of ...
Yes
Proposition 12.1.10 \( \left( {L}_{q}\right. \) -subspaces of \( \left. {L}_{p}\right) \) . (i) For \( 1 \leq p \leq 2,{L}_{q} \) embeds in \( {L}_{p} \) if and only if \( p \leq q \leq 2 \) . (ii) For \( 2 < p < \infty ,{L}_{q} \) embeds in \( {L}_{p} \) if and only if \( q = 2 \) or \( q = p \) . Moreover, if \( {L}_...
Proof. Let \( 1 \leq p, q < \infty \) and suppose that \( {L}_{q} \) embeds in \( {L}_{p} \) . Then, since \( {\ell }_{q} \) embeds in \( {L}_{q} \), it follows that \( {\ell }_{q} \) embeds in \( {L}_{p} \) . This implies, by Theorem 6.4.18, that either \( q = p \), or \( q = 2 \), or \( 1 \leq p < q < 2 \) . It remai...
Yes
Lemma 12.1.11. Let \( E \) be a finite-dimensional normed space and suppose \( {\left( {x}_{j}\right) }_{j = 1}^{N} \) is an \( \epsilon \) -net on the surface of the unit ball \( \{ e : \parallel e\parallel = 1\} \), where \( 0 < \epsilon < 1 \) . Suppose \( T : E \rightarrow X \) is a linear map such that \( 1 - \eps...
Proof. First suppose \( \parallel e\parallel = 1 \) . Pick \( j \) such that \( \begin{Vmatrix}{e - {x}_{j}}\end{Vmatrix} \leq \epsilon \) . Then\n\n\[ \parallel {Te}\parallel \leq \begin{Vmatrix}{{Te} - T{x}_{j}}\end{Vmatrix} + \left( {1 + \epsilon }\right) \]\n\nand so\n\n\[ \parallel T\parallel \leq \parallel T\para...
Yes
Proposition 12.1.12. Let \( X, Y \) be infinite-dimensional Banach spaces.\n\n(i) The ultraproduct \( {X}_{\mathcal{U}} \) is finitely representable in \( X \) .
Proof. (i) Let \( E \) be a finite-dimensional subspace of \( {X}_{\mathcal{U}} \) and suppose \( \epsilon > 0 \) . We can (by selecting representatives for a basis in \( E \) ) suppose \( E \subset {\ell }_{\infty }\left( X\right) \) and that \( \parallel \cdot {\parallel }_{\mathcal{U}} \) is a norm on \( E \) . Choo...
Yes
Theorem 12.1.15. Let \( X \) be a Banach space. Then\n\n(i) \( X \) fails to have type \( p > 1 \) if and only if \( {\ell }_{1} \) is finitely representable in \( X \) .\n\n(ii) \( X \) fails to have cotype \( q < \infty \) if and only if \( {\ell }_{\infty } \) is finitely representable in \( X \) .
Proof. We will use Proposition 7.2.5. For \( \left( i\right) \) it suffices to note that \( {\alpha }_{N}\left( X\right) = \sqrt{N} \) for every \( N \) . Thus for fixed \( N \) and all \( n \) we can find \( {\left( {x}_{nk}\right) }_{k = 1}^{N} \) such that\n\n\[{\left( \mathop{\sum }\limits_{{k = 1}}^{N}{\begin{Vmat...
No
Proposition 12.1.18. If \( 1 < p < \infty \), then the space \( {\ell }_{p} \) is superreflexive.
Proof. Fix \( 1 < p < \infty \) . Let \( X \) be a Banach space that is finitely representable in \( {\ell }_{p} \) and let \( Y \) be a separable subspace of \( X \) . By transitivity, \( Y \) is finitely representable in \( {\ell }_{p} \), so that by Theorem 12.1.9, the space \( Y \) is isomorphic to a subspace of \(...
Yes
Proposition 12.1.19. If a Banach space \( Y \) is crudely finitely representable in a superreflexive Banach space \( X \), then \( Y \) is superreflexive.
Proof. Let \( Z \) be a separable subspace of \( Y \) . By transitivity, \( Z \) is crudely finitely representable in \( X \) . Proposition 12.1.13 yields that \( Z \) is finitely representable in \( X \) equipped with an equivalent norm; hence \( Z \) is reflexive. Using again Lemma 12.1.17, we deduce that \( Y \) is ...
Yes
Corollary 12.1.21. The space \( {L}_{p} \) is superreflexive for \( 1 < p < \infty \) .
Proof. We know that \( {L}_{p} \) is finitely representable in \( {\ell }_{p} \) and that \( {\ell }_{p} \) is superreflexive. Now just apply Proposition 12.1.19.
Yes
Proposition 12.2.1. Let \( T : X \rightarrow Y \) be an operator with closed range. Suppose \( y \in \) \( Y \) is such that the equation \( {T}^{* * }{x}^{* * } = y \) has a solution \( {x}^{* * } \in {X}^{* * } \) with \( \begin{Vmatrix}{x}^{* * }\end{Vmatrix} < 1 \) . Then the equation \( {Tx} = y \) has a solution ...
Proof. This is almost immediate. We must show that \( y \in T\left( {U}_{X}\right) \), where \( {U}_{X} \) is the open unit ball of \( X \) . First suppose \( y \notin T\left( X\right) \) . In this case there exists \( {y}^{ * } \in {Y}^{ * } \) with \( {T}^{ * }\left( {y}^{ * }\right) = 0 \) but \( {y}^{ * }\left( y\...
Yes
Proposition 12.2.2. Let \( T : X \rightarrow Y \) be an operator with closed range and suppose \( K : X \rightarrow Y \) is a finite-rank operator. Then \( T + K \) also has closed range.
Proof. Suppose \( T + K \) does not have closed range. Then there is a bounded sequence \( {\left( {x}_{n}\right) }_{n = 1}^{\infty } \) with \( \mathop{\lim }\limits_{{n \rightarrow \infty }}\left( {T + K}\right) \left( {x}_{n}\right) = 0 \) but \( d\left( {{x}_{n},\ker \left( {T + K}\right) }\right) \geq 1 \) for all...
Yes
Theorem 12.2.3. Let \( X \) be a Banach space, \( A = {\left( {a}_{jk}\right) }_{j, k = 1}^{m, n} \) an \( m \times n \) real matrix, and \( B = {\left( {b}_{jk}\right) }_{j, k = 1}^{p, n} \) a \( p \times n \) real matrix. Let \( {y}_{1},\ldots ,{y}_{m} \in X,{y}_{1}^{ * },\ldots ,{y}_{p}^{ * } \in \) \( {X}^{ * } \),...
Proof. Consider the operator \( {T}_{0} : {\ell }_{\infty }^{n}\left( X\right) \rightarrow {\ell }_{\infty }^{m}\left( X\right) \) defined by\n\n\[ {T}_{0}\left( {{x}_{1},\ldots ,{x}_{n}}\right) = {\left( \mathop{\sum }\limits_{{k = 1}}^{n}{a}_{jk}{x}_{k}\right) }_{j = 1}^{m}. \]\n\nWe claim that \( {T}_{0} \) has clos...
Yes
Theorem 12.2.4 (The Principle of Local Reflexivity). Let \( X \) be a Banach space. Suppose that \( F \) is a finite-dimensional subspace of \( {X}^{* * } \) and \( G \) is a finite-dimensional subspace of \( {X}^{ * } \) . Then given \( \epsilon > 0 \) there exist a subspace \( E \) of \( X \) containing \( F \cap X \...
Proof. Given \( \epsilon > 0 \) let us take \( v > 0 \) such that \( \left( {1 + v}\right) {\left( 1 - 3v\right) }^{-1} < 1 + \epsilon \) and pick a \( v \) -net \( {\left( {x}_{j}^{* * }\right) }_{j = 1}^{N} \) in \( \left\{ {{x}^{* * } \in F : \begin{Vmatrix}{x}^{* * }\end{Vmatrix} = 1}\right\} \) . Let \( S : {\math...
Yes
Proposition 12.3.6. Suppose \( {\left( {x}_{n}\right) }_{n = 1}^{\infty } \) is a nonconstant spreading sequence in a Banach space X.\n\n(i) If \( {\left( {x}_{n}\right) }_{n = 1}^{\infty } \) fails to be weakly Cauchy, then \( {\left( {x}_{n}\right) }_{n = 1}^{\infty } \) is a basic sequence equivalent to the canonica...
Proof. (i) If \( {\left( {x}_{n}\right) }_{n = 1}^{\infty } \) is not weakly Cauchy, then no subsequence can be weakly Cauchy (by the spreading property), and so by Rosenthal's theorem (Theorem 11.2.1), some subsequence is equivalent to the canonical \( {\ell }_{1} \) -basis; but then again, this means that the entire ...
Yes
Lemma 12.3.10. Suppose \( \mathcal{X} \) is a spreading sequence space. Then there is a spreading sequence space \( \mathcal{Y} \) that is block finitely representable in \( \mathcal{X} \) such that the canonical basis of \( \mathcal{Y} \) is unconditional with unconditional basis constant \( {K}_{u} = 1 \) .
Proof. By the previous remarks we can assume that the canonical basis \( {\left( {e}_{n}\right) }_{n = 1}^{\infty } \) of \( \mathcal{X} \) is 2-unconditional, and that \( \mathcal{X} \) is not isomorphic to \( {c}_{0} \) . Thus, if we let \( {y}_{n} = \) \( \mathop{\sum }\limits_{{j = 1}}^{n}{\left( -1\right) }^{j}{e}...
Yes
Lemma 12.3.12. Suppose \( \mathcal{X} \) is a 1-unconditional spreading sequence space. Then\n\n(i) There exists a sequence \( {\left( {\xi }_{n}\right) }_{n = 1}^{\infty } \) in \( \mathcal{X}\left( {\mathbb{Q}}_{0}\right) \) with \( {\begin{Vmatrix}{\xi }_{n}\end{Vmatrix}}_{\mathcal{X}\left( {\mathbb{Q}}_{0}\right) }...
Proof. (i) Let us start by observing that\n\n\[ \mathop{\lim }\limits_{{n \rightarrow \infty }}{\begin{Vmatrix}{T}_{2}^{n}\end{Vmatrix}}^{\frac{1}{n}} = \mathop{\inf }\limits_{n}{\begin{Vmatrix}{T}_{2}^{n}\end{Vmatrix}}^{\frac{1}{n}} \]\n\n(12.3)\n\nand so\n\n\[ \begin{Vmatrix}{T}_{2}^{n}\end{Vmatrix} \geq {2}^{n\theta...
Yes
Theorem 12.3.13 (Dvoretzky’s Theorem). The space \( {\ell }_{2} \) is finitely representable in every infinite-dimensional Banach space.
Proof. An immediate conclusion from Krivine’s theorem is that some \( {\ell }_{p}(1 \leq \) \( p < \infty \) ) or \( {c}_{0} \) is finitely representable in every infinite-dimensional Banach space \( X \) . In the case of \( {c}_{0} \) this implies that \( {\ell }_{\infty } \) is finitely representable in \( X \), and ...
Yes
Theorem 13.1.1. Suppose \( \left( {X,\parallel \cdot {\parallel }_{X}}\right) \) is an n-dimensional normed space. Then there exists an n-dimensional ellipsoid of maximum volume contained in \( {B}_{X} \) .
Proof. Pick an ellipsoid \( {\mathcal{E}}_{n} \) contained in \( {B}_{X} \) . This ellipsoid corresponds to the image \( S\left( {B}_{{\ell }_{2}^{n}}\right) \) of some linear map \( S : {\ell }_{2}^{n} \rightarrow X \) . Let \( \parallel \cdot {\parallel }_{\mathrm{E}} \) be the Euclidean norm induced on \( X \) by \(...
Yes
Lemma 13.1.2. Suppose \( \left( {X,\parallel \cdot {\parallel }_{X}}\right) \) is an \( n \) -dimensional normed space. Let \( x \in X \) be such that \( \parallel x{\parallel }_{X} = \parallel x{\parallel }_{{\mathcal{E}}_{X}} = 1 \) . Then \( \parallel x{\parallel }_{{X}^{ * }} = 1 \) .
Proof. We must see that \( \langle x, y\rangle \leq 1 \) whenever \( \parallel y{\parallel }_{X} \leq 1 \) . For \( t > 0 \) we have\n\n\[ \parallel \left( {1 + t}\right) x - {ty}{\parallel }_{{\mathcal{E}}_{X}} \geq \parallel \left( {1 + t}\right) x - {ty}{\parallel }_{X} \geq \left( {1 + t}\right) \parallel x{\parall...
Yes
Lemma 13.1.3. Suppose that \( \left( {X,\parallel \cdot {\parallel }_{X}}\right) \) is an \( n \) -dimensional normed space and that \( T \in \mathcal{B}\left( X\right) \) . Then there is \( x \in X \) with \( \parallel x{\parallel }_{{\mathcal{E}}_{X}} = \parallel x{\parallel }_{X} = 1 \) such that\n\n\[ \left| {\oper...
Proof. For any \( t > 0 \) pick \( {x}_{t} \) such that \( {\begin{Vmatrix}{x}_{t}\end{Vmatrix}}_{{X}_{\mathcal{E}}} = 1 \) and\n\n\[ {\begin{Vmatrix}{x}_{t} + tT\left( {x}_{t}\right) \end{Vmatrix}}_{X} = \parallel I + {tT}{\parallel }_{{\mathcal{E}}_{X} \rightarrow X} \]\n\n(13.4)\n\nSince \( {\mathcal{E}}_{X} \) is a...
Yes
Theorem 13.1.4. Suppose \( \left( {X,\parallel \cdot {\parallel }_{X}}\right) \) is an n-dimensional normed space. Then,\n\n\[{\pi }_{2}\left( {I}_{X \rightarrow {\mathcal{E}}_{X}}\right) \leq \sqrt{n}\]
Proof. Suppose \( {x}_{1},\ldots ,{x}_{k} \in X \) . Consider the operator \( T : X \rightarrow X \) given by \( T = \) \( \mathop{\sum }\limits_{{i = 1}}^{k}{x}_{i} \otimes {x}_{i} \), that is,\n\n\[{Tu} = \mathop{\sum }\limits_{{i = 1}}^{k}\left\langle {{x}_{i}, u}\right\rangle {x}_{i}\]\n\nWe note that\n\n\[\operato...
Yes
Theorem 13.1.5 (John). If \( \left( {X,\parallel \cdot {\parallel }_{X}}\right) \) is an n-dimensional normed space, then \( d\left( {X,{\ell }_{2}^{n}}\right) \leq \sqrt{n} \)
Proof. We have \( \parallel I{\parallel }_{{\mathcal{E}}_{X} \rightarrow X} = 1 \) and \( \parallel I{\parallel }_{X \rightarrow {\mathcal{E}}_{X}} \leq {\pi }_{2}\left( {I}_{X \rightarrow {\mathcal{E}}_{X}}\right) \leq \sqrt{n} \) . Geometrically,\n\n\[ \n{\mathcal{E}}_{X} \subset {B}_{X} \subset \sqrt{n}{\mathcal{E}}...
Yes
Proposition 13.1.6. If \( X = {\ell }_{\infty }^{n} \) (or \( X = {\ell }_{1}^{n} \) ), then \( d\left( {X,{\ell }_{2}^{n}}\right) = \sqrt{n} \).
Proof. Let \( S : {\ell }_{\infty }^{n} \rightarrow {\ell }_{2}^{n} \) be a linear isomorphism. Let \( D = \parallel S{\parallel }_{{\ell }_{\infty }^{n} \rightarrow {\ell }_{2}^{n}} \) and \( C = \) \( {\begin{Vmatrix}{S}^{-1}\end{Vmatrix}}_{{\ell }_{2}^{n} \rightarrow {\ell }_{\infty }^{n}} \) . We have\n\n\[ \n{C}^{...
Yes
Theorem 13.1.7 (The Kadets-Snobar Theorem). Let \( X \) be a Banach space of dimension \( n \) . Then for every Banach space \( Y \) containing \( X \) as a subspace there is a projection \( P \) of \( Y \) onto \( X \) with \( \parallel P\parallel \leq \sqrt{n} \) .
Proof. According to Theorem 13.1.4, there is an operator \( S : X \rightarrow {\ell }_{2}^{n} \), where \( n = \) \( \dim X \), such that \( \begin{Vmatrix}{S}^{-1}\end{Vmatrix} = 1 \) and \( {\pi }_{2}\left( S\right) \leq \sqrt{n} \) . Using Theorem 8.2.13, we see that \( S \) extends to a bounded operator \( T : Y \r...
Yes
Theorem 13.2.2 (The Concentration of Measure Phenomenon). Let \( f \) be a Lipschitz function on \( {\mathcal{S}}^{n - 1} \) with Lipschitz constant 1. Then for \( t > 0 \) , \[ {\sigma }_{n}\left( {\left| {f - \bar{f}}\right| > t}\right) \leq 4{e}^{-n{t}^{2}/{72}{\pi }^{2}}, \] where \[ \bar{f} = {\int }_{{\mathcal{S}...
Proof. We shall assume that \( \bar{f} = 0 \), and so \( \left| {f\left( x\right) }\right| \leq 1 \) for all \( x \in {\mathcal{S}}^{n - 1} \) . Let us first extend \( f \) to \( {\mathbb{R}}^{n} \smallsetminus \{ 0\} \) by putting \[ f\left( x\right) = \parallel x{\parallel }_{2}f\left( {x/\parallel x{\parallel }_{2}}...
Yes
Lemma 13.3.1. Let \( \left( {{Y}_{0},\parallel \cdot \parallel }\right) \) be an m-dimensional Euclidean space. Suppose \( \epsilon > \) 0 . Then there is an \( \epsilon \) -net \( {\left\{ {x}_{j}\right\} }_{j = 1}^{N} \) for \( \left\{ {x \in {Y}_{0} : \parallel x\parallel = 1}\right\} \) with \( N \leq {\left( 1 + \...
Proof. Pick a maximal subset \( {\left\{ {x}_{j}\right\} }_{j = 1}^{N} \) of \( \left\{ {x \in {Y}_{0} : \parallel x\parallel = 1}\right\} \) with the property that \( \begin{Vmatrix}{{x}_{i} - {x}_{j}}\end{Vmatrix} \geq \epsilon \) whenever \( i \neq j \) . It is clear that this is an \( \epsilon \) -net. The open bal...
Yes
Theorem 13.3.2. Suppose \( \parallel \cdot {\parallel }_{\mathrm{F}} \) is a norm on \( {\mathbb{R}}^{n} \) with \( \parallel x{\parallel }_{\mathrm{F}} \leq \parallel x{\parallel }_{2} \) for all \( x \in {\mathbb{R}}^{n} \) . (a) Suppose \( 0 < \epsilon < \frac{1}{3} \) . Then there is a \( k \) -dimensional subspace...
Proof. (a) Let us fix some \( k \) -dimensional subspace \( {Y}_{0} \) of \( {\mathbb{R}}^{n} \) and use Lemma 13.3.1 to pick an \( \epsilon /3 \) -net \( {\left\{ {x}_{j}\right\} }_{j = 1}^{N} \) for \( \left\{ {x \in {Y}_{0} : \parallel x{\parallel }_{2} = 1}\right\} \) with \( N \leq {\left( 1 + 6/\epsilon \right) }...
Yes
Lemma 13.3.3. Let \( \parallel \cdot {\parallel }_{\mathrm{F}} \) be a norm on \( {\mathbb{R}}^{n} \) with \( \parallel x{\parallel }_{\mathrm{F}} \leq \parallel x{\parallel }_{2} \) for all \( x \) . Then:\n\n(a) \( {\theta }_{\mathrm{F}} \geq 1/\parallel I{\parallel }_{\mathrm{F} \rightarrow {\ell }_{2}^{n}} \) .\n\n...
Proof. Part (a) follows from \( 1 \leq \parallel I{\parallel }_{\mathrm{F} \rightarrow {\ell }_{2}^{n}}\parallel \xi {\parallel }_{\mathrm{F}} \) for all \( \xi \in {\mathcal{S}}^{n - 1} \) .\n\nSince \( \parallel G{\parallel }_{2} \) and \( G/\parallel G{\parallel }_{2} \) are independent random variables,\n\n\[ \math...
Yes
Theorem 13.3.4. Suppose \( 1 \leq p < \infty \) and \( n \in \mathbb{N} \) . Then for \( 0 < \epsilon < 1/3 \), the space \( {\ell }_{p}^{n} \) contains a subspace \( {X}_{0} \) with \( \dim {X}_{0} = k \) and \( {d}_{{X}_{0}} \leq 1 + \epsilon \), provided:\n\n- \( k \leq c{n}^{2/p}{\epsilon }^{2}{\left| \log \epsilon...
Proof. We consider \( {\mathbb{R}}^{n} \) equipped with the norm \( \parallel \cdot {\parallel }_{\mathrm{F}} = \parallel \cdot {\parallel }_{p}\left( {1 \leq p < \infty }\right) \) and denote the corresponding \( {\theta }_{\mathrm{F}} \) simply by \( {\theta }_{p} \) .\n\nIf \( p > 2 \), by Hölder’s inequality we hav...
Yes
Proposition 13.3.5 (The Dvoretzky-Rogers Lemma). Let \( \left( {X,\parallel \cdot {\parallel }_{X}}\right) \) be an \( n \) - dimensional normed space and suppose that \( \parallel \cdot {\parallel }_{{\mathcal{E}}_{X}} \) is the norm induced on \( X \) by the John ellipsoid. Then there is an orthonormal basis \( {\lef...
Proof. We must recall the definition of the John ellipsoid of \( X \) as the ellipsoid of maximal volume contained in \( {B}_{X} \) . We pick \( {\left( {e}_{j}\right) }_{j = 1}^{n} \) inductively such that \( {\begin{Vmatrix}{e}_{1}\end{Vmatrix}}_{{\mathcal{E}}_{X}} = \) \( {\begin{Vmatrix}{e}_{1}\end{Vmatrix}}_{X} = ...
Yes
Lemma 13.3.6. There is an absolute constant \( c > 0 \) such that if \( G \) is an m-dimensional Gaussian, then\n\n\[ \mathbb{E}\left( {\parallel G{\parallel }_{\infty }}\right) \geq c{\left( \log m\right) }^{1/2},\;m = 2,3,\ldots \]
Proof. If \( g \) is a scalar Gaussian and \( s > 0 \), then\n\n\[ \mathbb{P}\left( {\left| g\right| > s}\right) = \sqrt{\frac{2}{\pi }}{\int }_{s}^{\infty }{e}^{-\frac{1}{2}{\xi }^{2}}{d\xi } \geq \sqrt{\frac{2}{\pi }}s{e}^{-2{s}^{2}}. \]\n\nThus\n\n\[ \mathbb{P}\left( {\parallel G{\parallel }_{\infty } > s}\right) \g...
Yes
Theorem 13.3.7 (Dvoretzky’s Theorem). There is an absolute constant \( c > 0 \) with the following property: If \( \left( {X,\parallel \cdot {\parallel }_{X}}\right) \) is an n-dimensional normed space and \( 0 < \epsilon < 1/3 \), then \( X \) has a subspace \( {X}_{0} \) with \( \dim {X}_{0} = k \) and \( {d}_{{X}_{0...
Proof. Let \( \parallel \cdot {\parallel }_{{\mathcal{E}}_{X}} \) be the norm induced on \( X \) by the John ellipsoid. By the Dvoretzky-Rogers lemma, we can pass to a subspace \( Y \) of \( X \) with \( m = \dim Y \geq n/2 \), and with the property that \( \left( {Y,\parallel \cdot {\parallel }_{{\mathcal{E}}_{X}}}\ri...
Yes
Lemma 13.4.4. Suppose \( E \) is a finite-dimensional subspace of \( X \) and let \( v > 0 \) . Then there is a finite-codimensional subspace \( Y \) of \( E \) such that\n\n\[ \parallel e + y\parallel \geq \left( {1 - v}\right) \parallel e\parallel ,\;e \in E, y \in Y. \]
Proof. Let \( {\left\{ {x}_{i}\right\} }_{i = 1}^{N} \) be a \( v \) -net for \( S = \{ x \in E : \parallel x\parallel = 1\} \) . We use the Hahn-Banach theorem to construct \( {\left\{ {x}_{i}^{ * }\right\} }_{i = 1}^{N} \) in \( {X}^{ * } \) such that \( \begin{Vmatrix}{x}_{i}^{ * }\end{Vmatrix} = 1 \) and \( {x}_{i}...
Yes
Theorem 13.4.5 (Lindenstrauss and Tzafriri [200]). Suppose \( X \) is an infinite-dimensional Banach space in which every closed subspace is complemented. Then \( X \) is isomorphic to a Hilbert space.
Proof. By Theorem 13.4.3 we must prove that there is \( \lambda \geq 1 \) such that every finite-dimensional subspace of \( X \) is \( \lambda \) -complemented. Suppose that this is not the case and\n\nfor a finite-dimensional subspace \( E \) of \( X \), set,\n\n\[ \lambda \left( E\right) = \inf \{ \parallel P\paralle...
Yes
Problem 14.1.1. When are two separable Banach spaces homeomorphic as topological spaces?
This question was the subject of intense research in the years after the Second World War and was resolved beautifully by Kadets (announced in 1965) [145]:\n\nTheorem 14.1.2. All sep
No
Theorem 14.1.3 (Mazur and Ulam [217]). Let \( X \) and \( Y \) be real normed spaces and suppose that \( \Phi \) is an isometry mapping \( X \) onto \( Y \) that carries 0 into 0 . Then \( \Phi \) is a bounded linear operator. In particular, if two real Banach spaces \( X \) and \( Y \) are isometric as Banach spaces, ...
Proof. It suffices to show that \( \Phi \) preserves midpoints of line segments, i.e.,\n\n\[ \Phi \left( \frac{x + y}{2}\right) = \frac{\Phi \left( x\right) + \Phi \left( y\right) }{2} \]\n\n(14.2)\n\nfor all \( x, y \in X \) . Indeed, if this is the case, setting \( y = 0 \) and using the assumption that \( \Phi \left...
Yes
Lemma 14.1.7. Let \( f : X \rightarrow Y \) be a map between two metric spaces.\n\n(iii) If \( f \) is uniformly continuous and \( X \) is metrically convex, then \( {\omega }_{f}\left( s\right) < \infty \) for all \( s > 0 \) .
Proof. We do (iii) and leave the other statements as an exercise. We need to show that for \( s > 0 \) there is \( {C}_{s} > 0 \) such that \( d\left( {f\left( x\right), f\left( y\right) }\right) \leq {C}_{s} \) whenever \( d\left( {x, y}\right) \leq s \) . From the definition of uniform continuity there exists \( {\de...
No
Proposition 14.1.8 (Corson and Klee [53]). Let \( f : X \rightarrow Y \) be a uniformly continuous map. If \( X \) is metrically convex, then for every \( \theta > 0 \) there exists a constant \( {K}_{\theta } > 0 \) such that \( d\left( {f\left( x\right), f\left( y\right) }\right) \leq {K}_{\theta }d\left( {x, y}\righ...
Proof. Fix \( \theta > 0 \) . Given \( x, y \) in \( X \) with \( d\left( {x, y}\right) \geq \theta \), let \( m \) be the smallest integer such that \( d\left( {x, y}\right) /m < \theta \) . By the metric convexity of \( X \) we may choose points \( x = {x}_{0},{x}_{1},\ldots ,{x}_{m} = y \) in \( X \) with \( d\left(...
Yes
Lemma 14.1.14. Let \( X \) be an unbounded metrically convex space. Given \( x \in X \) and \( r > 0 \) there is \( y \in X \) such that \( d\left( {x, y}\right) = r \) .
Proof. Since \( X \) is unbounded, there is \( z \in X \) such that \( d\left( {x, z}\right) = R > r \) . Let \( t = r/R \in \left( {0,1}\right) \) . The metric convexity of \( X \) yields \( y \in X \) such that \( d\left( {x, y}\right) = {td}\left( {x, z}\right) = r. \)
Yes
Lemma 14.1.15. Let \( f : X \rightarrow Y \) be a map between two unbounded metric spaces.\n\n(i) Suppose that for some constants \( A \geq 1 \) and \( B \geq 0 \) the inequalities\n\n\[ \frac{1}{A}d\left( {x, y}\right) - B \leq d\left( {f\left( x\right), f\left( y\right) }\right) \leq {Ad}\left( {x, y}\right) + B \]\n...
Proof. (i) Suppose \( d\left( {x, y}\right) > \theta \) for a given \( \theta > 0 \) . Then,\n\n\[ d\left( {f\left( x\right), f\left( y\right) }\right) \leq {Ad}\left( {x, y}\right) + B \leq {Ad}\left( {x, y}\right) + \frac{B}{\theta }d\left( {x, y}\right) = \left( {A + \frac{B}{\theta }}\right) d\left( {x, y}\right) ....
Yes
Lemma 14.1.17. A metric space \( X \) contains an \( \alpha \) -separated \( \alpha \) -net for every \( \alpha > 0 \) .
Proof. For \( \alpha > 0 \), let \( \mathcal{I} = \{ \mathcal{N} \subseteq X : d\left( {{s}_{1},{s}_{2}}\right) \geq \alpha \) for all \( \left( {{s}_{1},{s}_{2}}\right) \in \mathcal{N} \times \mathcal{N} \) , \( \left. {{s}_{1} \neq {s}_{2}}\right\} \) . The set \( \mathcal{I} \) is nonempty, since it contains \( \var...
Yes
Proposition 14.1.21. Let \( X \) and \( Y \) be normed spaces and suppose \( f : X \rightarrow Y \) is a uniform homeomorphism. Then \( f \) is a (surjective) coarse Lipschitz embedding.
Proof. Since \( f \) is uniformly continuous, Proposition 14.1.8 yields for every \( \theta > 0 \) a constant \( {K}_{\theta } \) such that \( \parallel f\left( x\right) - f\left( y\right) \parallel \leq {K}_{\theta }\parallel x - y\parallel \) for every \( x, y \in X \) with \( \parallel x - y\parallel \geq \theta \) ...
Yes
Proposition 14.2.2. Let \( X \) be a finite-dimensional Banach space, and let \( f \) be a Lipschitz map from an open set in \( X \) to a (possibly infinite-dimensional) Banach space \( Y \) . Iff is Gâteaux differentiable at a point \( x \), then \( f \) is Fréchet differentiable at \( x \) .
Proof. Given \( \epsilon > 0 \), by compactness of the unit sphere of \( X \) we may choose an \( \epsilon \) -net \( {\left\{ {u}_{i}\right\} }_{i = 1}^{N} \) in \( {S}_{X} \) . Let \( \delta > 0 \) be such that \( \begin{Vmatrix}{f\left( {x + t{u}_{i}}\right) - f\left( x\right) - t{D}_{f}\left( x\right) \left( {u}_{i...
Yes
Lemma 14.2.4. Let \( f : X \rightarrow Y \) be a Lipschitz map between Banach spaces. Suppose \( E \) is a finite-dimensional subspace of \( X \) . Define \( g : X \rightarrow Y \) by\n\n\[ g\left( x\right) = {\int }_{E}f\left( {x - \xi }\right) \varphi \left( \xi \right) {d\lambda }\left( \xi \right) ,\;x \in X, \]\n\...
Proof. Using the properties of the Bochner integral, we have\n\n\[ \parallel g\left( y\right) - g\left( y\right) \parallel = \begin{Vmatrix}{{\int }_{E}\left( {f\left( {x - \xi }\right) - f\left( {y - \xi }\right) }\right) \varphi \left( \xi \right) {d\lambda }\left( \xi \right) }\end{Vmatrix} \]\n\n\[ \leq {\int }_{E}...
Yes
Lemma 14.2.5. Let \( f : X \rightarrow Y \) be a Lipschitz map between Banach spaces. Suppose \( E \) is a \( k \) -dimensional subspace of \( X \) . Define maps \( {\left( {g}_{n}\right) }_{n = 1}^{\infty } \) from \( X \) to \( Y \) by\n\n\[ \n{g}_{n}\left( x\right) = {2}^{nk}{\int }_{E}f\left( {x - \xi }\right) \var...
Proof. Write \( f\left( x\right) = {2}^{nk}{\int }_{E}f\left( x\right) \varphi \left( {{2}^{n}\xi }\right) {d\xi } \) . Then \n\n\[ \n\begin{Vmatrix}{{g}_{n}\left( x\right) - f\left( x\right) }\end{Vmatrix} = {2}^{nk}\begin{Vmatrix}{{\int }_{E}\left( {f\left( {x - \xi }\right) - f\left( x\right) }\right) \varphi \left(...
Yes
Lemma 14.2.9. Let \( h : X \rightarrow \left\lbrack {0,\infty }\right\rbrack \) be a measurable function. Suppose there exists a finite-dimensional subspace \( E \) of \( X \) such that \( h\left( {x + \xi }\right) = 0 \) a.e. \( \xi \in E \) for all \( x \in X \) , i.e., \( \lambda \left( {\{ \xi \in E : h\left( {x + ...
Proof. Pick a probability measure \( \mu \) on \( \left( {X,\mathcal{B}\left( X\right) }\right) \) such that \( \mu \left( {X \smallsetminus E}\right) = 0 \), and \( \mu \left( A\right) = 0 \) if and only if \( A \cap E \) has Lebesgue measure zero. Then we have \( h\left( {x + \xi }\right) = \) \( {0\mu } \) -a.e \( \...
Yes
Lemma 14.2.11. Let \( h \) be a Haar-null map on a separable infinite-dimensional Banach space \( X \) . For every \( \epsilon > 0 \) there exist a probability measure space \( \left( {\Omega ,\sum ,\mathbb{P}}\right) \) and a random variable \( \eta : \Omega \rightarrow X \) with \( \parallel \eta \parallel < \epsilon...
Proof. We know that there exist a probability space \( \left( {\Omega ,\sum ,\mathbb{P}}\right) \) and a random variable \( \eta : \Omega \rightarrow X \) such that \( \mathbb{E}\left( {h\left( {x + \eta }\right) }\right) = 0 \) for all \( x \in X \) . Since \( X \) is separable, we can decompose \( \Omega \) into coun...
Yes
Lemma 14.2.12. Let \( X \) be a separable infinite-dimensional Banach space.\n\n(i) If \( {\left( {h}_{n}\right) }_{n = 1}^{\infty } \) is a sequence of Haar-null maps on \( X \), then \( h = \mathop{\sum }\limits_{{n = 1}}^{\infty }{h}_{n} \) is Haar-null.\n\n(ii) If \( {\left( {A}_{n}\right) }_{n = 1}^{\infty } \) is...
Proof. For each \( n \) choose a probability measure space \( \left( {{\Omega }_{n},{\mathbb{P}}_{n}}\right) \) and a random variable \( {\eta }_{n} : {\Omega }_{n} \rightarrow X \) such that \( \mathbb{E}\left( {{h}_{n}\left( {x + {\eta }_{n}}\right) }\right) = 0 \) for all \( x \in X \) and (by Lemma 14.2.11) \( \beg...
Yes
Theorem 14.2.15 (Heinrich and Mankiewicz [124]). Let \( X \) and \( Y \) be Banach spaces and suppose \( f : X \rightarrow Y \) is a Lipschitz embedding. Assume \( X \) is separable and that \( Y \) has the Radon-Nikodym property. Then \( X \) is linearly isomorphic to a subspace of \( Y \) . More specifically, if \( {...
Proof. By Theorem 14.2.13 there exists at least one point \( {x}_{0} \in X \) where \( f \) is Gâteaux differentiable. Then, by the discussion leading to the estimate (14.11), \( {D}_{f}\left( {x}_{0}\right) \) is an isomorphic embedding of \( X \) into \( Y \) and \( \begin{Vmatrix}{{D}_{f}\left( {x}_{0}\right) }\end{...
Yes
Theorem 14.2.19 (Infinite-dimensional weak* Rademacher Theorem). Let \( X \) and \( Z \) be separable Banach spaces and let \( f : X \rightarrow {Z}^{ * } \) be a Lipschitz map. Then\n\n(a) \( f \) is weak* differentiable outside a Haar-null set.\n\n(b) \( \begin{Vmatrix}{{D}_{f}^{ * }\left( x\right) }\end{Vmatrix} \le...
Proof. To show part (c) let \( a \) be the inverse of \( \operatorname{Lip}\left( {f}^{-1}\right) \) . Choose an increasing sequence \( {\left( {E}_{n}\right) }_{n = 1}^{\infty } \) of finite-dimensional subspaces of \( X \) whose union is dense. For each \( n \), consider the set \( {D}_{n} \) of all points \( x \in X...
No
Proposition 14.2.21 (Existence of linear Hahn-Banach extension operators). Let \( Z \) be a Banach space, and let \( V \) be a separable subspace of \( Z \) . There exist a separable space \( W \) with \( V \subseteq W \subseteq Z \) and a linear map \( T : {W}^{ * } \rightarrow {Z}^{ * } \) such that \( \parallel T\pa...
Proof. Certainly, if \( Z \) is separable, then the result is clear, since it suffices to take \( W = Z \) . Having disposed of this case, we may assume that \( Z \) is nonseparable. Then, in particular, \( Z \) is infinite-dimensional and \( V \) can be assumed to be also infinite-dimensional. Pick a sequence \( {\lef...
Yes
Corollary 14.2.22. Let \( Y \) be a separable subspace of a dual space \( {Z}^{ * } \) . Then there exists a separable Banach space \( W \) such that \( Y \) embeds linearly isometrically into \( {W}^{ * } \) and \( {W}^{ * } \) embeds linearly isometrically into \( {Z}^{ * } \) .
Proof. Choose a separable subspace \( V \subseteq Z \) that is isometrically norming for \( Y \) . By Proposition 14.2.21, there exist \( V \subseteq W \subseteq Z \) and a linear isometric embedding \( T : {W}^{ * } \rightarrow {Z}^{ * } \) such that \( Q \circ T = {I}_{{W}^{ * }} \) . Since \( W \) is also isometrica...
Yes
Theorem 14.2.23. If a separable Banach space \( X \) Lipschitz embeds into the dual \( {Z}^{ * } \) of a Banach space \( Z \), then \( X \) is isomorphic to a subspace of \( {Z}^{ * } \) . Quantitatively, if \( f : X \rightarrow {Z}^{ * } \) is a Lipschitz embedding, then there is a linear embedding \( T : X \rightarro...
Proof. Assume \( f : X \rightarrow {Z}^{ * } \) is a Lipschitz embedding. Since \( X \) is separable, \( f\left( X\right) \) is contained in a separable subspace \( Y \) of \( {Z}^{ * } \) . By Corollary 14.2.22 we can find a separable Banach space \( W \) and isometric linear embeddings \( {S}_{0} : Y \rightarrow {W}^...
Yes
Theorem 14.2.27 (Ribe [268]). If there is a coarse Lipschitz embedding of a Banach space \( X \) into a Banach space \( Y \), then \( X \) is crudely finitely representable in \( Y \) .
Proof. Suppose \( X \) coarse Lipschitz embeds into \( Y \) . By Corollary 14.1.26 there is a Lipschitz embedding \( f : X \rightarrow {Y}_{\mathcal{U}} \) for some free ultrafilter \( \mathcal{U} \) on \( \mathbb{N} \) . Since \( {\left\lbrack {Y}_{\mathcal{U}}\right\rbrack }^{* * } \) is finitely representable in \( ...
Yes
Corollary 14.2.29. Let \( X \) and \( Y \) be Banach spaces such that there exists a coarse Lipschitz embedding from \( X \) into \( Y \) . If \( Y \) has type \( p \) [respectively, cotype \( q \) ], then \( X \) has type p [respectively, cotype q].
Proof. Use Theorem 14.2.27 and the definitions of type and cotype.
No
Corollary 14.2.30. If a Banach space \( X \) coarse Lipschitz embeds into a Hilbert space, then it is isomorphic to a Hilbert space.
Proof. This follows directly from Theorem 14.2.27 and Theorem 12.1.6.
Yes
Let \( q : {\ell }_{\infty } \rightarrow {\ell }_{\infty }/{c}_{0} \) be the canonical quotient map. We start from the existence of a continuum of infinite subsets \( {\left( {\mathbb{A}}_{i}\right) }_{i \in \mathcal{I}} \) of \( \mathbb{N} \) with the property that \( {\mathbb{A}}_{i} \cap {\mathbb{A}}_{j} \) is finit...
If \( x = \mathop{\sum }\limits_{{n = 1}}^{\infty }{a}_{n}q\left( {\xi }_{{i}_{n}}\right) \), where \( {a}_{1} \geq {a}_{2} \geq {a}_{3} \geq \cdots \geq 0 \), we let\n\n\[ f\left( x\right) \left( k\right) = \left\{ \begin{array}{ll} {a}_{1} & \text{ if }k \in {\mathbb{A}}_{{i}_{1}}, \\ {a}_{n} & \text{ if }k \in {\mat...
Yes
Proposition 14.3.5 (Heinrich and Mankiewicz [124]). Let \( f \) be a Lipschitz embedding of a Banach space \( X \) into a Banach space \( Y \) . Assume that \( f \) is Gâteaux differentiable at a point \( {x}_{0} \in X \) . If \( f\left( X\right) \) is a Lipschitz retract of \( Y \) and \( X \) is complemented in \( {X...
Proof. By a simple translation argument, without loss of generality we may assume that \( {x}_{0} = 0 \) and \( f\left( 0\right) = 0 \) . We will write \( {D}_{f} \) instead of \( {D}_{f}\left( {x}_{0}\right) \) . Let \( g : Y \rightarrow X \) be a Lipschitz map such that \( g \circ f = {I}_{X} \) . For all \( {y}_{1},...
Yes
Theorem 14.3.8. Let \( X \) and \( Y \) be separable Banach spaces such that \( Y \) has the Radon-Nikodym property and \( X \) is complemented in \( {X}^{* * } \) . If \( X \) is Lipschitz isomorphic to a Lipschitz retract of \( Y \), then \( X \) is linearly isomorphic to a complemented subspace of \( Y \) .
Proof. Let \( f : X \rightarrow Y \) and \( g : Y \rightarrow X \) be Lipschitz maps such that \( g \circ f = {I}_{X} \) . Theorem 14.2.13 ensures the existence of points in \( X \) where \( f \) is Gâteaux differentiable. Let \( {D}_{f} : X \rightarrow Y \) be the Gâteaux derivative of \( f \) at one of these points. ...
Yes
Theorem 14.3.12. If \( X \) is Lipschitz isomorphic to \( {\ell }_{1} \) and is a dual space, then \( X \) is linearly isomorphic to \( {\ell }_{1} \) .
Proof. The extra hypothesis on \( X \) guarantees that it is complemented in its bidual. By Theorem 14.3.8 we obtain that \( X \) is complemented in \( {\ell }_{1} \) . Since \( {\ell }_{1} \) is a prime space, \( X \) must be isomorphic to \( {\ell }_{1} \) .
Yes
Lemma 14.4.1. A convex function \( f : \mathbb{R} \rightarrow \mathbb{R} \) is differentiable at a point \( x \) if and only if
\[ \mathop{\lim }\limits_{{t \rightarrow {0}^{ + }}}\frac{f\left( {x + t}\right) + f\left( {x - t}\right) - {2f}\left( x\right) }{t} = 0. \]
No
Lemma 14.4.2. Let \( g : {\mathbb{R}}^{n} \rightarrow \mathbb{R} \) be a convex function. Then for all \( x \in {\mathbb{R}}^{n} \) and all \( t > 0 \)
Proof. Pick \( x = {\left( {x}_{i}\right) }_{i = 1}^{n} \in {\mathbb{R}}^{n} \) and \( t > 0 \) . Convexity shows that\n\n\[ \mathop{\sup }\limits_{{\parallel h{\parallel }_{1} \leq t}}g\left( {x + h}\right) - g\left( x\right) = \mathop{\max }\limits_{\substack{{1 \leq i \leq n} \\ {\epsilon = \pm 1} }}\left( {g\left( ...
Yes
A convex function \( f : {\mathbb{R}}^{n} \rightarrow \mathbb{R} \) is Fréchet differentiable at a point \( x \in {\mathbb{R}}^{n} \) if and only if the \( n \) partial derivatives of \( f \) at \( x \) exist and are finite. In particular, \( f \) is Fréchet differentiable at \( x \) if and only if it is Gâteaux differ...
Suppose the \( n \) partial derivatives of \( f \) exist at \( x \in {\mathbb{R}}^{n} \) . Define the linear map \( T : {\mathbb{R}}^{n} \rightarrow \mathbb{R} \) by\n\n\[ T\left( h\right) = \mathop{\sum }\limits_{{i = 1}}^{n}\frac{\partial f}{\partial {x}_{i}}\left( x\right) {h}_{i},\;h = \left( {{h}_{1},\ldots ,{h}_{...
Yes
Proposition 14.4.5. Suppose \( E \) is a finite-dimensional normed space. If \( f : E \rightarrow \mathbb{R} \) is a convex function, then it is continuous on \( E \) and Fréchet differentiable at every point of a dense subset of \( E \) .
Proof. Without loss of generality we may and do assume that \( E \) is \( {\mathbb{R}}^{n} \) equipped with the norm \( \parallel \cdot {\parallel }_{1} \) . The continuity of a convex function \( f : {\mathbb{R}}^{n} \rightarrow \mathbb{R} \) when \( n > 1 \) follows from Lemma 14.4.2 and the 1-dimensional case. For \...
Yes
Lemma 14.4.7. Suppose \( \left( {E,\parallel \cdot \parallel }\right) \) is a finite-dimensional normed space.\n\n(i) If \( x \in {\Omega }_{\parallel \cdot \parallel } \), then\n\n\[ \begin{Vmatrix}{{D}_{\parallel \cdot \parallel }\left( x\right) }\end{Vmatrix} = {D}_{\parallel \cdot \parallel }\left( x\right) \left( ...
Proof. (i) For \( u \in E \) and \( t > 0 \), by the reverse triangle inequality,\n\n\[ \left| \frac{\parallel x + {tu}\parallel - \parallel x\parallel }{t}\right| \leq \frac{\parallel x + {tu} - x\parallel }{\left| t\right| } = \parallel u\parallel .\n\nMaking \( t \rightarrow {0}^{ + } \) gives \( \left| {{D}_{\paral...
Yes
Lemma 14.4.9. Let \( \left( {E,\parallel \cdot \parallel }\right) \) be a finite-dimensional normed space. Pick \( x \in E \) with \( \parallel x\parallel = 1 \) a point of differentiability of the norm. Then \( {D}_{\parallel \cdot \parallel }\left( x\right) \) is the only 1-Lipschitz map \( \varphi : E \rightarrow \m...
Proof. By Lemma 14.4.7 (i), \( {D}_{\parallel \cdot \parallel }\left( x\right) : E \rightarrow R \) is 1-Lipschitz and \( {D}_{\parallel \cdot \parallel }\left( x\right) \left( {\alpha x}\right) = \alpha \) for all \( \alpha \in \mathbb{R} \) . Conversely, let \( \varphi : E \rightarrow \mathbb{R} \) be a 1-Lipschitz m...
Yes
Corollary 14.4.11. Every onto isometry \( \Phi : X \rightarrow Y \) between Banach spaces such that \( \Phi \left( 0\right) = 0 \) is linear.
Proof. Theorem 14.4.10 applied to \( \Phi \) shows that \( \Phi = {T}^{-1} \) is a linear isometry.
No
Theorem 14.4.12 (Godefroy and Kalton [105]). Let \( X \) be a separable Banach space. If there exists an isometry \( \Phi \) from \( X \) into a Banach space \( Y \), then \( Y \) contains a closed linear subspace that is linearly isometric to \( X \) .
Proof. We may and do assume that \( \Phi \left( 0\right) = 0 \) and that \( \left\lbrack {\Phi \left( X\right) }\right\rbrack = Y \) . By Theorem 14.4.10, there is a norm-one quotient map \( Q : Y \rightarrow X \) such that \( Q \circ \Phi = {I}_{X} \) . We can therefore apply Theorem 14.3.3 with \( g = \Phi \), and th...
Yes
Theorem 14.5.2 (Enflo-Lindenstrauss). Let \( 1 \leq p < \infty \) . If the spaces \( {L}_{p}\left( {\mu }_{1}\right) \) and \( {L}_{q}\left( {\mu }_{2}\right) \) are uniformly homeomorphic, then either they are of the same finite dimension or \( p = q \) .
Proof. Suppose \( 1 \leq p < q < \infty \) and that both \( {L}_{p}\left( {\mu }_{1}\right) \) and \( {L}_{q}\left( {\mu }_{2}\right) \) are of infinite dimension. If \( {L}_{p}\left( {\mu }_{1}\right) \) and \( {L}_{q}\left( {\mu }_{2}\right) \) are uniformly homeomorphic, by Proposition 14.1.21, \( {L}_{p}\left( {\mu...
Yes
Theorem 14.5.3 (Bourgain-Enflo-Gorelik). For \( 1 \leq p < \infty \) with \( p \neq 2 \) the spaces \( {\ell }_{p} \) and \( {L}_{p} \) are not uniformly homeomorphic.
Here the case \( p = 1 \) is due to Enflo in the 1970s (unpublished; see a proof in [23, Theorem 10.13]); the case \( 1 < p < 2 \) was established in 1987 by Bourgain [29], and the case \( 2 < p < \infty \) was not settled until 1994 by Gorelik [112].
No
Proposition 14.5.5. Let \( X \) and \( Y \) be Banach spaces, and let \( f : X \rightarrow Y \) be a coarse Lipschitz map. If \( {\operatorname{Lip}}_{\infty }\left( f\right) > 0 \), then for every \( \theta > 0,\epsilon > 0 \), and \( 0 < \delta < 1 \), there exist \( x, y \in X \) with \( \parallel x - y\parallel > \...
Proof. Suppose \( \theta ,\delta ,\epsilon \) are given. For \( \eta > 0 \) as small as we wish we pick \( {\theta }^{\prime } > \theta \) such that \( {\operatorname{Lip}}_{{\theta }^{\prime }}\left( f\right) < \left( {1 + \eta }\right) {\operatorname{Lip}}_{\infty }\left( f\right) \) . Then we choose \( x, y \in X \)...
Yes
Lemma 14.5.6. Suppose \( 1 \leq p < \infty \) . Let \( {\left( {e}_{i}\right) }_{i = 1}^{\infty } \) be the canonical basis of \( {\ell }_{p} \) , and for \( N \in \mathbb{N} \), denote by \( {E}_{N} \) the closed linear span of \( \left\{ {{e}_{i} : i > N}\right\} \) . Let \( x, y \in {\ell }_{p} \) , \( \delta \in \l...
Proof. Given \( 0 < \eta < 1 \), pick \( N \in \mathbb{N} \) such that \( \mathop{\sum }\limits_{{i = 1}}^{N}{\left| {v}_{i}\right| }^{p} \geq \left( {1 - {\eta }^{p}}\right) \parallel v{\parallel }_{p}^{p} \) .\n\n(i) We may clearly assume that \( p > 1 \) . Let now \( z \in {E}_{N} \) be that \( \parallel z{\parallel...
Yes
Proposition 14.5.7. Let \( 1 \leq p < q < \infty \) and suppose that \( f : {\ell }_{q} \rightarrow {\ell }_{p} \) is a coarse Lipschitz map. Then for every \( t > 0 \) and \( \epsilon > 0 \) there exist \( u \in {\ell }_{q},\tau > t, N \in \mathbb{N} \), and a compact subset \( K \) of \( {\ell }_{p} \) such that\n\n\...
Proof. We assume that \( {\operatorname{Lip}}_{\infty }\left( f\right) > 0 \), since if \( {\operatorname{Lip}}_{\infty }\left( f\right) = 0 \), the conclusion is clear. We choose a small \( \delta > 0 \) (to be specified later). Then we pick \( \theta \) large enough (also to be detailed later) that \( {\operatorname{...
Yes
Corollary 14.5.8. If \( 1 \leq p < q < \infty ,{\ell }_{q} \) does not coarse Lipschitz embed into \( {\ell }_{p} \) .
Proof. Let \( f : {\ell }_{q} \rightarrow {\ell }_{p} \) be a coarse Lipschitz map. With the notation of the previous proposition, we can find a sequence \( {\left( {u}_{n}\right) }_{n = 1}^{\infty } \) in \( u + \tau {B}_{{E}_{N}} \) such that \( {\begin{Vmatrix}{u}_{n} - {u}_{m}\end{Vmatrix}}_{q} \geq \tau \) for \( ...
Yes
Corollary 14.5.10. If \( 1 \leq q < p < \infty ,{\ell }_{q} \) does not coarse Lipschitz embed into \( {\ell }_{p} \) .
Proof. Suppose that \( {\ell }_{q} \) coarse Lipschitz embeds into \( {\ell }_{p} \) . Then, using homogeneity, there exist \( f : {\ell }_{q} \rightarrow {\ell }_{p} \) and \( B \geq 1 \) such that\n\n\[ \parallel x - y{\parallel }_{q} \leq \parallel f\left( x\right) - f\left( y\right) {\parallel }_{p} \leq B\parallel...
Yes
Example 14.6.2. (a) If \( X = {\ell }_{p} \) equipped with its natural norm, then\n\n\[{\bar{\rho }}_{{\ell }_{p}}\left( \tau \right) = {\left( 1 + {\tau }^{p}\right) }^{1/p} - 1,\;\tau > 0,\]
and so \( {\ell }_{p} \) is asymptotically uniformly smooth if \( 1 < p < \infty \) . The natural norm actually has an optimal (i.e., minimal) modulus of asymptotic uniform smoothness among all equivalent norms.
Yes
Lemma 14.6.4 (The Gorelik Principle). Let \( E \) and \( X \) be Banach spaces. Suppose \( \varphi : E \rightarrow X \) is a homeomorphism whose inverse \( {\varphi }^{-1} \) is Lipschitz. Let \( b \) and \( c \) be positive constants such that \( c > \operatorname{Lip}\left( {\varphi }^{-1}\right) \cdot b \), and let ...
Proof. Put \( a = \operatorname{Lip}\left( {\varphi }^{-1}\right) \cdot b \) . We first construct a compact subset \( \widetilde{K} \) of the ball \( c{B}_{E} \) such that if \( \psi : \widetilde{K} \rightarrow E \) is a continuous map with \( \parallel x - \psi \left( x\right) \parallel \leq a \) for every \( x \in \w...
Yes
Proposition 15.1.2. There is a norm-one operator \( T : \mathcal{X} \rightarrow X \) defined by \( T{e}_{n} = {x}_{n} \) for \( n \in \mathbb{N} \) . The operator \( T \) is a quotient map.
Proof. It is easy to see that \( \xi \in X \) implies that \( \mathop{\sum }\limits_{{j = 1}}^{\infty }\xi \left( j\right) {x}_{j} \) must converge and that\n\n\[\n\begin{Vmatrix}{\mathop{\sum }\limits_{{j = 1}}^{\infty }\xi \left( j\right) {x}_{j}}\end{Vmatrix} \leq \parallel \xi {\parallel }_{\mathcal{X}}\n\]\n\nThus...
Yes
Lemma 15.1.3. \( {T}^{ * }\left( {X}^{ * }\right) \cap \mathcal{Y} = \{ 0\} \), and \( {T}^{ * }{X}^{ * } + \mathcal{Y} \) is norm closed.
Proof. It is enough to note that if \( {x}^{ * } \in {X}^{ * } \) and \( {\xi }^{ * } \in \mathcal{Y} \), then\n\n\[ \n{\begin{Vmatrix}{T}^{ * }{x}^{ * }\end{Vmatrix}}_{\mathcal{X}} = \begin{Vmatrix}{x}^{ * }\end{Vmatrix} \leq {\begin{Vmatrix}{T}^{ * }{x}^{ * } + {\xi }^{ * }\end{Vmatrix}}_{{\mathcal{X}}^{ * }}\n\]\n\n...
No
Theorem 15.1.6. For every separable Banach space \( X \) there is a separable Banach space \( \mathcal{Z} \) such that \( {\mathcal{Z}}^{* * }/\mathcal{Z} \) is isomorphic to \( X \) . Furthermore, \( {\mathcal{Z}}^{ * } \) has a shrinking basis.
Proof. We take \( \mathcal{Z} = \ker T \) in the above construction. We show that \( \mathcal{X} \) can then be identified canonically with \( {\mathcal{Z}}^{* * } \) . More precisely, we show that under the pairing between \( \mathcal{X} \) and \( \mathcal{Y} \) we can identify \( \mathcal{Y} \) with \( {\mathcal{Z}}^...
Yes
Theorem 15.3.1 (Pelczyński's universal basis space). There is a unique separable Banach space \( U \) with a basis and with the property that every Banach space with a basis is isomorphic to a complemented subspace of \( U \) .
Proof. To prove the existence of \( U \) it suffices to construct a Banach space \( X \) with a basis \( {\left( {x}_{n}\right) }_{n = 1}^{\infty } \) such that every normalized basic sequence (in any Banach space) is equivalent to a complemented subsequence of \( {\left( {x}_{n}\right) }_{n = 1}^{\infty } \) . Then th...
No
Theorem 15.3.2. There is a unique Banach space \( {U}_{1} \) with an unconditional basis \( {\left( {u}_{n}\right) }_{n = 1}^{\infty } \) and with the property that every Banach space with an unconditional basis is isomorphic to a complemented subspace of \( {U}_{1} \) .
Proof. Suppose \( X \) is the space constructed in the preceding proof. Then we can define a norm on \( {c}_{00} \) by\n\n\[ \parallel \xi {\parallel }_{{U}_{1}} = \mathop{\sup }\limits_{{{\epsilon }_{j} = \pm 1}}{\begin{Vmatrix}\mathop{\sum }\limits_{{j = 1}}^{\infty }{\epsilon }_{j}\xi \left( j\right) {e}_{j}\end{Vma...
No
Proposition 15.4.2. Suppose a basis \( {\left( {x}_{n}\right) }_{n = 1}^{\infty } \) of a Banach space \( X \) satisfies a lower 2-estimate on blocks. Then,\n\n(i) The formula\n\n\[ \parallel \parallel x\parallel \parallel = \max \left\{ {\parallel x\parallel ,\sup {\left( \mathop{\sum }\limits_{{j = 1}}^{n}{\begin{Vma...
Proof. We leave the verification of \( \left( i\right) \) to the reader. To show (ii), suppose\n\n\[ \mathop{\sup }\limits_{n}\begin{Vmatrix}{\mathop{\sum }\limits_{{k = 1}}^{n}{a}_{k}{x}_{k}}\end{Vmatrix} < \infty \]\n\nbut the series \( \mathop{\sum }\limits_{{k = 1}}^{\infty }{a}_{k}{x}_{k} \) does not converge. The...
No
Proposition 15.4.4. The norm \( \parallel \cdot {\parallel }_{\mathcal{X}} \) has the following properties:\n\n(i) For every \( B \in \mathcal{B} \),\n\n\[ \parallel \xi {\parallel }_{B} = \parallel \xi {\parallel }_{\mathcal{X}},\;\xi \in {c}_{00}\left( B\right) .\n\]\n\n(ii) If \( {E}_{1},\ldots ,{E}_{n} \) are disjo...
Proof. (i) follows directly from (15.8).\n\n(ii) Given \( \epsilon > 0 \), pick disjoint segments \( {\left( {S}_{jk}\right) }_{k = 1}^{{m}_{n}} \) for \( j = 1,2,\ldots, n \) such that\n\n\[ \mathop{\sum }\limits_{{j = 1}}^{n}\mathop{\sum }\limits_{{k = 1}}^{{m}_{n}}{\begin{Vmatrix}{S}_{jk}{E}_{j}\xi \end{Vmatrix}}^{2...
Yes
Lemma 15.4.6. Suppose \( \xi \in {c}_{00} \) is supported on \( \left\lbrack {1, N}\right\rbrack \) and \( \eta \in {c}_{00} \) is supported on \( \lbrack N + 1,\infty ) \) . Then\n\n\[ \parallel \xi + \eta {\parallel }_{\mathcal{X}} \leq {\left( \parallel \xi {\parallel }_{\mathcal{X}}^{2} + \parallel \eta {\parallel ...
Proof. Let \( \delta = \mathop{\sup }\limits_{{m \geq N + 1}}{\begin{Vmatrix}{T}_{m}\eta \end{Vmatrix}}_{\mathcal{X}} \) . Suppose \( \epsilon > 0 \) and pick disjoint segments \( {\left( {S}_{j}\right) }_{j = 1}^{m} \) such that\n\n\[ \parallel \xi + \eta {\parallel }_{\mathcal{X}}^{2} < \mathop{\sum }\limits_{{j = 1}...
Yes
Theorem 15.4.8. There is a Banach space \( \mathcal{Y} \) such that \( {\mathcal{Y}}^{ * } \) is separable and \( {\mathcal{Y}}^{* * }/\mathcal{Y} \) is isometric to \( {\ell }_{2}\left( \mathcal{I}\right) \), where \( \mathcal{I} \) has the cardinality of the continuum.
Proof. We use the space \( \mathcal{J} \) but with the basis of Problem 3.11, which is a special case of the construction of Theorem 15.1.6. It is trivial to see that the basis \( {\left( {e}_{n}\right) }_{n = 1}^{\infty } \) of the space \( \mathcal{X} \) constructed in Section 15.1 has an exact lower 2-estimate on bl...
No
Theorem 15.4.9. The space \( {\mathcal{Y}}^{ * } = \mathcal{J}\mathcal{T} \) has nonseparable dual, but \( {\ell }_{1} \) does not embed into \( \mathcal{J}\mathcal{T} \) .
Proof. Obviously, \( \mathcal{J}{\mathcal{T}}^{ * } \) is nonseparable. Since \( \mathcal{J}\mathcal{T} \) is a dual space, it is complemented in its bidual, and so \( \mathcal{J}{\mathcal{T}}^{* * } = \mathcal{J}\mathcal{T} \oplus W \), where \( W \) can be identified as the dual of the space \( \mathcal{J}{\mathcal{T...
Yes
Theorem 15.4.10. Let \( X \) be any separable dual space. Then there is a Banach space \( Z \) such that \( {Z}^{* * }/Z \) is isomorphic to \( {\ell }_{2}{\left( X\right) }_{i \in \mathcal{I}} \) where \( \mathcal{I} \) has the cardinality of the continuum.
Proof. Let \( X = {Y}^{ * } \) and construct \( \mathcal{Z} \) as in Section 15.1 such that \( {\mathcal{Z}}^{* * }/\mathcal{Z} \approx Y \) . Using the canonical basis of \( \mathcal{Z} \) as in Theorem 15.4.8 will give us a space \( Z \) such that \( {Z}^{* * }/Z \) is isomorphic to \( {\ell }_{2}{\left( {Y}^{ * }\ri...
Yes
Lemma 1.4.1. Setting \( F\left( x\right) \mathrel{\text{:=}} {P}_{x}\left( {\{ \mathbf{0}\} }\right) \) and\n\n\[ k = {i}_{1} + {i}_{2}N + \cdots + {i}_{n}{N}^{n - 1}\;\left( { \in {\mathbb{N}}_{0}}\right) ,\]\n\nwe get the formula\n\n\[ {P}_{x}\left( {\{ k\} }\right) = F\left( {x + k}\right) \]\n\nwhere we have identi...
Proof. To see this, identify functions on \( \left\lbrack {0,1}\right\rbrack \) with 1-periodic functions on \( \mathbb{R} \), and note that the second formula in (1.4.1) yields \( {\tau }_{{i}_{n}}\cdots {\tau }_{{i}_{1}}\left( x\right) = \left( {x + k}\right) /{N}^{n} \) where \( k \) is given by (1.4.2). Hence, if \...
Yes
Lemma 1.5.1. Let \( N, W, F,{P}_{x} \), and \( h \) be as described above; see (1.5.2)-(1.5.3). Then \( h \) satisfies the following cocycle identity:\n\n\[ h\left( x\right) W\left( x\right) = {P}_{Nx}\left( {N\mathbb{Z}}\right) ,\;x \in \mathbb{R}. \]
Proof. We calculate the left-hand side in (1.5.6), using the earlier equations:\n\n\[ h\left( x\right) W\left( x\right) \underset{\begin{matrix} \text{ by (1.4.3) } \\ \text{ and (1.5.3) } \end{matrix}}{ = }W\left( x\right) \mathop{\sum }\limits_{{j \in \mathbb{Z}}}F\left( {x + j}\right) \]\n\n\[ \underset{\text{by per...
Yes
Theorem 1.5.2. Let \( N, W, F,{P}_{x} \), and \( h \) be as described above. Let \( x \in \mathbb{R} \), and suppose that \( {P}_{x}\left( {\{ \mathbf{0}\} }\right) > 0 \) . Then the following two conditions are equivalent.\n\n(a) The limit on the right-hand side below exists, and\n\n\[ F\left( x\right) = \mathop{\lim ...
Proof. (a) \( \Rightarrow \) (b). An iteration of the identity (1.5.6) in Lemma 1.5.1 above yields\n\n\[ {P}_{x}\left( {{N}^{k}\mathbb{Z}}\right) = \left( {\mathop{\prod }\limits_{{j = 1}}^{k}W\left( \frac{x}{{N}^{j}}\right) }\right) h\left( \frac{x}{{N}^{k}}\right) .\n\]\n\n(1.5.7)\n\nUsing (1.5.5), and working in \( ...
Yes
Lemma 2.5.1. (Kolmogorov) Let \( N \geq 2 \) be fixed, and let\n\n\[ \Omega = \{ 0,1,\ldots, N - 1{\} }^{\mathbb{N}} \]\n\nFor \( n = 1,2,\ldots \), let\n\n\[ {P}^{\left( n\right) } : {\mathfrak{A}}_{n} \rightarrow \mathbb{C} \]\n\nbe a sequence of linear functionals such that (i)-(iii) hold:\n\n(i) \( {P}^{\left( n\ri...
Proof of Lemma 2.5.1. The proof of Kolmogorov's extension result may be given several forms, but we note that the argument we used above (in a special case), based on an application of the Stone-Weierstraß theorem, also works in general.
No
Lemma 2.6.2. There is a unique mapping \( \rho : M\left( \Omega \right) \rightarrow M\left( X\right) \) which satisfies\n\n\[ \rho \left( {fg}\right) = \rho \left( f\right) \rho \left( g\right) \]\n\n(2.6.3)\n\nand\n\n\[ \rho \left( {\chi }_{A\left( {{i}_{1},\ldots ,{i}_{n}}\right) }\right) = {\chi }_{{\tau }_{{i}_{1}}...
Proof. Recalling (2.4.9), we note that\n\n\[ {\chi }_{A\left( {{i}_{1},\ldots ,{i}_{n}}\right) }\left( \omega \right) = {\delta }_{{i}_{1},{\omega }_{1}}\cdots {\delta }_{{i}_{n},{\omega }_{n}},\;\omega \in \Omega . \]\n\nAs a result,\n\n\[ {\chi }_{A\left( {{i}_{1},\ldots ,{i}_{n}}\right) }{\chi }_{A\left( {{j}_{1},\l...
Yes
Lemma 2.6.3. Let \( X, W \), and \( N \) be as described in the beginning of this chapter, and let \( \left\{ {{P}_{x} \mid x \in X}\right\} \) be the process obtained in the conclusion of Lemma 2.4.1. Then\n\n\[ \mathop{\sum }\limits_{{i = 0}}^{{N - 1}}W\left( {{\tau }_{i}x}\right) {P}_{{\tau }_{i}x}\left\lbrack {f\le...
Proof of Lemma 2.6.3. It follows from (2.4.6) and the arguments in the proof of Lemma 2.4.1 that it is enough to verify (2.6.6) for \( f \in {C}_{\text{fin }}\left( \Omega \right) \), or for \( f \in {\mathfrak{A}}_{n} \) . Let \( f \in {\mathfrak{A}}_{n} \) . Then\n\n\[ \mathop{\sum }\limits_{i}W\left( {{\tau }_{i}x}\...
Yes
Theorem 2.7.1. Let \( X, W \), and \( N \) be as described in the beginning of this chapter, and suppose in addition that (2.4.1) is satisfied. Let \( R = {R}_{W} \) be the Ruelle operator on \( {L}^{\infty }\left( X\right) \), and let \( \left\{ {{P}_{x} \mid x \in X}\right\} \) be the process on \( \Omega \) from Lem...
Proof. (2.7.1) \( \Rightarrow \) (2.7.2). This follows by Lemma 2.6.3 and the cocycle property. Let \( V \) be a cocycle, and define \( h = {h}_{V} \) by (2.7.1). Then for \( x \in X \),\n\n\[\left( {Rh}\right) \left( x\right) = \mathop{\sum }\limits_{i}W\left( {{\tau }_{i}x}\right) h\left( {{\tau }_{i}x}\right) = \mat...
Yes
Corollary 2.7.3. The \( 1 - 1 \) correspondence between the harmonic functions \( h \) in Theorem 2.7.1 and the cocycles \( V \) is an order-isomorphism when the ordering of both families of functions is defined in the pointwise sense.
Proof. We noted that when \( V \) is a given cocycle, then the corresponding function \( h \) is defined by (2.7.1); and conversely, \( V \) may be computed from \( h \) via the formula (2.7.6).\n\nSince each \( {P}_{x} \) is a positive measure,(2.7.1) yields the implication\n\n\[ \n{V}_{1} \leq {V}_{2} \Rightarrow {h}...
Yes
Corollary 2.7.4. Let \( v : \Omega \rightarrow \mathbb{C} \) be a bounded measurable function, and let \( X, W \) , and \( R = {R}_{W} \) be as stated in the theorem. Then\n\n\[ V\left( {x,\omega }\right) \mathrel{\text{:=}} v\left( \omega \right) ,\;\omega \in \Omega ,\]\n\n(2.7.8)\n\nis a cocycle if and only if \( v ...
Proof. Let \( v : \Omega \rightarrow \mathbb{C} \) be given, and suppose \( V \) in (2.7.8) is a cocycle. Then\n\n\[ v\left( \omega \right) = V\left( {x,\omega }\right) = V\left( {{\tau }_{{\omega }_{1}}x,\left( {{\omega }_{2},{\omega }_{3},\ldots }\right) }\right) = v\left( {{\omega }_{2},{\omega }_{3},\ldots }\right)...
Yes
Corollary 2.7.6. Let \( E \subset \Omega \), and let \( X,\sigma ,{\tau }_{i}, N \), and \( W \) be as described in the beginning of the chapter. Set\n\n\[ \n{h}_{E}\left( x\right) \mathrel{\text{:=}} {P}_{x}\left\lbrack {\chi }_{E}\right\rbrack \]\n\n(2.7.11)\n\nwhere \( {\chi }_{E} \) denotes the indicator function o...
Proof. This follows directly from Corollary 2.7.4 above, if we note that\n\n\[ \n{\chi }_{E} \circ {\sigma }^{\Omega } = {\chi }_{{\left( {\sigma }^{\Omega }\right) }^{-1}E}. \]\n
Yes
Proposition 2.8.2. Let \( v \) be \( R \) -invariant, and let \( h \) satisfy \( {Rh} = h \) . Then the measure\n\n\[ d{v}_{h} \mathrel{\text{:=}} {hdv} \]\n\n(2.8.3)\n\nis invariant.\n\nConversely, if \( {v}_{1} \) is a \( \mathcal{B} \) -measure on \( X \) which is assumed \( \sigma \) -invariant, and if \( {v}_{1} \...
Proof. (Part one!)\n\n\[ \int f \circ {\sigma d}{v}_{h} = \int f \circ {\sigma hdv} \]\n\n\[ = \int R\left( {f \circ {\sigma h}}\right) {dv} = \int {fRhdv} \]\n\n\[ = \int {fhdv} = \int {fd}{v}_{h} \]\n\nwhich implies that \( {v}_{h} \) is \( \sigma \) -invariant.\n\nTo prove (2.8.4) under the assumptions in the second...
Yes
Lemma 3.1.1. Let \( W \) and \( f \) be bounded 1-periodic functions on \( \mathbb{R} \). (If the two functions are measurable, they will be assumed only essentially bounded, and a.e. 1-periodic.)\n\nThen the function\n\n\[ \mathop{\sum }\limits_{{k = 0}}^{{N - 1}}W\left( \frac{x + k}{N}\right) f\left( \frac{x + k}{N}\...
Proof. Note that the expression in (3.1.12) is the right-hand side in (3.1.9), or equivalently in (3.1.10), and so it is the function \( {R}_{W}f \). We now give the explicit argument as to why it is 1-periodic. Note that the individual terms in the sum (3.1.12) have period \( N \) and not 1 . Let\n\n\[ g\left( x\right...
Yes
Lemma 3.2.1. Let \( W : \left\lbrack {0,1}\right\rbrack \rightarrow \left\lbrack {0,1}\right\rbrack \) be a given measurable function and extend \( W \) from \( \left\lbrack {0,1}\right\rbrack \) to \( \mathbb{R} \) by periodicity. Then the densities for the probabilities \( {P}_{x} \) in (1.2.4) are\n\n\[ W\left( {{\t...
Proof. A direct calculation, using (3.2.1) and the periodicity of \( W \), yields\n\n\[ W\left( {{\tau }_{{\omega }_{s}}{\tau }_{{\omega }_{s - 1}}\cdots {\tau }_{{\omega }_{1}}x}\right) = W\left( \frac{x + k}{{N}^{s}}\right) \;\text{ for }1 \leq s \leq n. \]\n\nIt is understood that the integer \( k \in {\mathbb{N}}_{...
Yes