A small remark on small-dimensional normed barrelled spaces

Damian Sobota · 2025

Read A small remark on small-dimensional normed barrelled spaces by Damian Sobota (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2511-13355.txt.

Introduction

Damian Sobota proves a lower bound on the size of normed barrelled spaces. A barrelled space is a topological vector space where a natural completeness-like property holds for families of continuous maps. The question is how small the algebraic dimension of such a space can be.

Blending methods of Brian and Stuart with the classical Dvoretzky theorem, Sobota shows no endless-dimensional Banach space holds a barrelled subspace below a stated cardinal tied to the Lebesgue null ideal. Hence every endless-dimensional normed barrelled space reaches at least that size. One corollary is that standard set theory can consistently rule out a barrelled space whose dimension equals the bounding number.

Catalog record

CREATORDamian Sobota
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE12 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2511-13355

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

A small remark on small-dimensional normed barrelled spaces Thanks: The author was supported by the Austrian Science Fund (FWF), ESP 108-N Damian Sobota Abstract . Combining the methods of Brian and Stuart with the classical Dvoretzky theorem, we show that no infinite-dimensional Banach space contains a barrelled subspace of (algebraic) dimension < cov ​ ( 𝒩 ) <\mbox{cov}(\mathcal{N}) , the covering number of the Lebesgue null ideal 𝒩 \mathcal{N} . Consequently, every infinite-dimensional normed barrelled space has dimension ≥ cov ​ ( 𝒩 ) \geq\mbox{cov}(\mathcal{N}) and it is consistent with ZFC that no Banach space contains a barrelled subspace of dimension equal to the bounding number 𝔟 \mathfrak{b} . Recall that a topological vector space E is barrelled if every absolutely convex closed absorbing subset ( barrel ) of E is a neighborhood of 0 . Barrelled spaces constitute one of the most important classes of topological vector spaces because strong forms of such classical theorems as the Uniform Boundedness Principle or the Open Mapping Theorem hold in them, see [ 10 ] . The minimal (algebraic) dimension of infinite-dimensional metrizable barrelled spaces was studied by Sánchez Ruiz and Saxon in [ 5 , 6 , 7 ] , where it was proved that it is equal to the cardinal characteristic of the continuum called the bounding number \mathfrak{b} 1 1 1 See [ 2 ] for the definitions and basic facts concerning the cardinal characteristics \mathfrak{b} , \non(\mathcal{M}) , and \cov(\mathcal{N}) , used in this note. . In [ 5 ] and [ 6 ] they also defined the following cardinal number: \mathfrak{y}=\min\big\{\kappa\geq\aleph_{0}\colon\ \text{there is a normed barrelled space of dimension }\kappa\big\}, and asked whether it is equal to any standard cardinal characteristic of the continuum. Of course, we have \mathfrak{y}\geq\mathfrak{b} . The author proved in [ 8 ] that consistently \mathfrak{y} may be strictly smaller than the continuum \mathfrak{c} , i.e. the cardinality of the real line \mathbb{R} , which implies that in ZFC , the standard system of axioms of set theory, one cannot prove that \mathfrak{y}=\mathfrak{c} (see also [ 9 ] ). Brian and Stuart [ 3 ] recently showed that every infinite-dimensional separable Banach space contains a barrelled subspace of dimension \non(\mathcal{M}) , the minimal cardinality of a non-meager subset of \mathbb{R} . It follows in particular that \mathfrak{y}\leq\non(\mathcal{M}) . Moreover, they also proved that if the dual space E^{*} of a Banach space E contains an isomorphic copy of a Banach space F\in\{c_{0}\}\cup\{\ell_{p}\colon 1\leq p<\infty\} , then no barrelled subspace of E has dimension strictly less than \cov(\mathcal{N}) , the minimal cardinality of a family of the Lebesgue null sets whose union is \mathbb{R} ( [ 3 , Theorem 4.12] ). The question whether the conclusion of the latter statement holds for all Banach spaces E was left open. Question 1 ( [ 3 , Question 4.14] ) . Does every barrelled subspace of an infinite-dimensional Banach space have dimension \geq\cov(\mathcal{N}) ? In particular, \mathfrak{y}\geq\cov(\mathcal{N}) ? The aim of this note is to answer Question 1 in affirmative and hence to complete Brian and Stuart’s work—see Theorem 3 and Corollary 5 . Moreover, in Corollary 6 we get that the inequality \mathfrak{b}<\mathfrak{y} is consistent with ZFC , which sheds new light on the question of Sánchez Ruiz and Saxon. Let us briefly discuss the argument. The proof of aforementioned [ 3 , Theorem 4.12] is divided into two cases: separately for F=\ell_{1} and F\in\{c_{0}\}\cup\{\ell_{p}\colon 1<p<\infty\} . To establish the second case, the authors of [ 3 ] introduce a property of Banach spaces denoted (\dagger) , prove that c_{0} and \ell_{p} for 1<p<\infty satisfy (\dagger) , and finally show that (\dagger) implies the required conclusion concerning the dimension of barrelled subspaces. However, an analysis of their proof demonstrates that what is in fact used to obtain the last result is the following property, which we call (\dagger^{\prime}) and which is formally weaker than (\dagger) (see Remark 4 ): • there are a sequence (e_{k}^{*})_{k\in\mathbb{N}} of unit functionals in E^{*} and a partition of \mathbb{N}=\bigcup_{m\in\mathbb{N}}I_{m} into finite intervals such that \sum_{m\in\mathbb{N}}\frac{\big|\big\{i\in I_{m}\colon\ |e_{i}^{*}(e)|\geq 1/m\big\}\big|}{|I_{m}|}<\infty for every unit vector e\in E . (Here, as usual, \mathbb{N} denotes the set of positive integers, i.e. \mathbb{N}=\{1,2,3,...\} .) It turns out that the property (\dagger^{\prime}) is actually satisfied by every infinite-dimensional Banach space, including \ell_{1} , which follows from the classical Dvoretzky theorem asserting that the space \ell_{2} is finitely representable in every infinite-dimensional Banach space E , that is, for each finite-dimensional subspace F of \ell_{2} and each \varepsilon>0 there are a finite-dimensional subspace G of E and a surjective isomorphism T\colon F\to G with \|T\|=1 and \|T^{-1}\|<1+\varepsilon (cf. [ 1 , Theorem 12.3.13 and Remark 12.1.3.(a)] ), see Proposition 2 . Consequently, by using the arguments presented in the proofs of [ 3 , Theorems 4.4 and 4.10] , in Theorem 3 we get that every barrelled subspace of an infinite-dimensional Banach space must have dimension at least \cov(\mathcal{N}) , which basically answers Question 1 . We are ready to present the proofs. Proposition 2 . Every infinite-dimensional Banach space satisfies (\dagger^{\prime}) . Proof. Let E be an infinite-dimensional Banach space. Let (\tilde{e}_{n})_{n\in\mathbb{N}} be the standard basis of \ell_{2} , i.e. for j\neq n\in\mathbb{N} we have \tilde{e}_{n}(j)=0 and \tilde{e}_{n}(n)=1 . For every m\in\mathbb{N} set E_{m}=\spn\big\{\tilde{e}_{1},\ldots,\tilde{e}_{m^{4}}\big\} , so E_{m} is an m^{4} -dimensional subspace of \ell_{2} , and let T_{m}\colon E_{m}\to F_{m} be an isomorphism onto a subspace F_{m} in the dual space E^{*} given by the Dvoretzky theorem with \|T_{m}\|=1 and \big\|T_{m}^{-1}\big\|<2 . For each m\in\mathbb{N} and 1\leq i\leq m^{4} , let \alpha_{m,i}=\big\|T_{m}(\tilde{e}_{i})\big\|^{-1}, so \alpha_{m,i}<2 , and f_{m,i}^{*}=\alpha_{m,i}T_{m}(\tilde{e}_{i}), hence \big\|f_{m,i}^{*}\big\|=1 . Let \mathbb{N}=\bigcup_{m\in\mathbb{N}}I_{m} be a partition of \mathbb{N} into consecutive intervals such that |I_{m}|=m^{4} . For each k\in\mathbb{N} , let m_{k}\in\mathbb{N} and 1\leq i_{k}\leq m_{k}^{4} be such that k=i_{k}+\sum_{j=1}^{m_{k}-1}j^{4} , and set e_{k}^{*}=f_{m_{k},i_{k}}^{*}. We claim that the sequences (e_{k}^{*})_{k\in\mathbb{N}} and (I_{m})_{m\in\mathbb{N}} witness that E satisfies (\dagger^{\prime}) . To this end, fix a unit vector e\in E . Let m\in\mathbb{N} . Put J_{m}^{+}=\big\{k\in I_{m}\colon e_{k}^{*}(e)\geq 1/m\big\},\quad J_{m}^{-}=\big\{k\in I_{m}\colon-e_{k}^{*}(e)\geq 1/m\big\},\quad J_{m}=J_{m}^{+}\cup J_{m}^{-}. We have \bigg\|\sum_{k\in J_{m}^{\pm}}e_{k}^{*}\bigg\|=\bigg\|\sum_{k\in J_{m}^{\pm}}f_{m,i_{k}}^{*}\bigg\|\leq\|T_{m}\|\cdot\bigg\|\sum_{k\in J_{m}^{\pm}}\alpha_{m,i_{k}}\tilde{e}_{i_{k}}\bigg\|_{\ell_{2}}=\bigg(\sum_{k\in J_{m}^{\pm}}\alpha_{m,i_{k}}^{2}\bigg)^{1/2}<2\sqrt{\big|J_{m}^{\pm}\big|}. On the other hand, \bigg\|\sum_{k\in J_{m}^{+}}e_{k}^{*}\bigg\|\geq\bigg|\sum_{k\in J_{m}^{+}}e_{k}^{*}(e)\bigg|=\sum_{k\in J_{m}^{+}}e_{k}^{*}(e)\geq\big|J_{m}^{+}\big|/m, and similarly \bigg\|\sum_{k\in J_{m}^{-}}e_{k}^{*}\bigg\|\geq\bigg|\sum_{k\in J_{m}^{-}}e_{k}^{*}(e)\bigg|=-\sum_{k\in J_{m}^{-}}e_{k}^{*}(e)\geq\big|J_{m}^{-}\big|/m. Combining the above inequalities, we get \big|J_{m}^{\pm}\big|/m\leq\bigg\|\sum_{k\in J_{m}^{\pm}}e_{k}^{*}\bigg\|<2\sqrt{\big|J_{m}^{\pm}\big|}, hence, rearranging, \big|J_{m}^{\pm}\big|<4m^{2}, and finally |J_{m}|=\big|J_{m}^{+}\big|+\big|J_{m}^{-}\big|<8m^{2}. Using the latter inequality for every m\in\mathbb{N} , we get \sum_{m\in\mathbb{N}}\frac{\big|\big\{k\in I_{m}\colon\ |e_{k}^{*}(e)|\geq 1/m\big\}\big|}{|I_{m}|}=\sum_{m\in\mathbb{N}}\frac{|J_{m}|}{|I_{m}|}<\sum_{m\in\mathbb{N}}\frac{8m^{2}}{m^{4}}=8\sum_{m\in N}1/m^{2}<\infty, which finishes the proof of the proposition. ∎ Theorem 3 . Assume that A is a subset of an infinite-dimensional Banach space E such that the space \spn A is barrelled. Then, |A|\geq\cov(\mathcal{N}) . Proof. The proof is the same as of [ 3 , Theorem 4.10] , so we only sketch it very briefly for the sake of completeness. Let (e_{k}^{*})_{k\in\mathbb{N}} and (I_{m})_{m\in\mathbb{N}} be sequences given by (\dagger^{\prime}) for E (Proposition 2 ). Endow each set I_{m} with the discrete topology, and let P be the standard probability product measure on the compact space \Omega=\prod_{m\in\mathbb{N}}I_{m} , i.e. P\big(\{f\in\Omega\colon f(m)\in F\}\big)=|F|/|I_{m}| for every m\in\mathbb{N} and F\subseteq I_{m} . For every f\in\Omega and m\in\mathbb{N} set y_{f,m}^{*}=me_{f(m)}^{*}, so that \big\|y_{f,m}^{*}\big\|=m , and hence \sup_{m\in\mathbb{N}}\big\|y_{f,m}^{*}\big\|=\infty . We claim that for every x\in A it holds P\Big(\Big\{f\in\Omega\colon\sup_{m\in\mathbb{N}}\big|y_{f,m}^{*}(x)\big|=\infty\Big\}\Big)=0. To see this, let x\in A be non-zero. For all M\in\mathbb{N} we have P\Big(\Big\{f\in\Omega\colon\sup_{m\in\mathbb{N}}\big|y_{f,m}^{*}(x)\big|=\infty\Big\}\Big)=P\Big(\Big\{f\in\Omega\colon\sup_{m>M}\big|y_{f,m}^{*}(x)\big|=\infty\Big\}\Big)\leq

Read the complete text