Exact Banach–Mazur distances of certain ℓ p \ell_{p} -sums and cones
Florian Grundbacher and Tomasz Kobos · 2026
Read Exact Banach–Mazur distances of certain ℓ p \ell_{p} -sums and cones by Florian Grundbacher and Tomasz Kobos (2026) 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-2603-18268.txt.
Introduction
Grundbacher and Kobos determine exact Banach-Mazur distances for certain l_p sums and cones. A recent result on optimal position against the Euclidean ball drives the work. They give a closed formula for the distance of an l_p direct sum to Euclidean space. For p equal to 1, sums with l_1 keep the distance of the base space when it is at most 3. Cone versions swap in the simplex, with equalities for symmetric bases in dimension 3.
Catalog record
| CREATOR | Florian Grundbacher and Tomasz Kobos |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 72021 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2603.18268
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Exact Banach–Mazur distances of certain ℓ p \ell_{p} -sums and cones Florian Grundbacher and Tomasz Kobos Abstract We determine certain Banach–Mazur distances involving ℓ p \ell_{p} -direct sums of finite-dimensional real normed spaces and related cone constructions of convex bodies. Using a recent characterization of the optimal Banach–Mazur position with respect to the Euclidean ball, we derive a closed formula for the distance from X 1 ⊕ p ⋯ ⊕ p X k X_{1}\oplus_{p}\cdots\oplus_{p}X_{k} to Euclidean space in terms of the distances of the spaces X i X_{i} to Euclidean space. For p = 1 p=1 we show that if d B M ( X , ℓ 1 n ) ≤ 3 d_{BM}(X,\ell_{1}^{n})\leq 3 , then d B M ( X ⊕ 1 ℓ 1 m , ℓ 1 n + m ) = d B M ( X , ℓ 1 n ) d_{BM}(X\oplus_{1}\ell_{1}^{m},\ell_{1}^{n+m})=d_{BM}(X,\ell_{1}^{n}) . Interpreting ℓ 1 \ell_{1} -sums geometrically as double cones motivates a study of single cones over arbitrary convex bases, for which we establish an analogous result with the simplex replacing ℓ 1 \ell_{1} . We further show that in dimension 3 3 the distance between single cones with symmetric bases equals the distance between the bases, and that the same equality holds for double cones over planar symmetric bases in arbitrary dimension, under an additional assumption on the distance of the bases to ℓ 1 2 \ell_{1}^{2} . As consequences, we obtain an explicit isometric embedding of the 2 2 -dimensional symmetric Banach–Mazur compactum into the 3 3 -dimensional (non-symmetric) compactum and lift a recent construction of arbitrarily large equilateral sets in the 2 2 -dimensional symmetric compactum to all higher dimensions. 0 0 footnotetext: 2020 Mathematics Subject Classification. Primary 52A21; Secondary 46B04, 46B20, 52A20. 0 0 footnotetext: Key words and phrases. Banach–Mazur distance, p-sum, convex cone, isometric embedding. 1 Introduction and Results The purpose of this paper is to determine exact Banach–Mazur distances between several \ell_{p} -sum constructions for finite-dimensional real normed spaces, as well as related cone constructions for general convex bodies in \mathbb{R}^{n} . Before we state the main results, we introduce the required notation and recall some basic facts. For any integer n\geq 1 and scalar p\in[1,\infty] , we denote by \ell_{p}^{n} the space \mathbb{R}^{n} equipped with the norm \|\cdot\|_{p} defined for x=(x_{1},\ldots,x_{n})\in\mathbb{R}^{n} by \|x\|_{p}=(|x_{1}|^{p}+\ldots+|x_{n}|^{p})^{\frac{1}{p}}, where for p=\infty we understand \|x\|_{\infty}=\max\{|x_{1}|,\ldots,|x_{n}|\} . In particular, \|\cdot\|_{1} , \|\cdot\|_{2} , and \|\cdot\|_{\infty} denote the \ell_{1} , Euclidean, and maximum norms, respectively. The unit ball of \ell_{p}^{n} is given by \mathbb{B}_{p}^{n} . More generally, given p\in[1,\infty] and finite-dimensional normed spaces X_{i}=(\mathbb{R}^{n_{i}},\|\cdot\|_{X_{i}}) for i=1,\ldots,k , the \ell_{p} -direct sum X_{1}\oplus_{p}\ldots\oplus_{p}X_{k} is the space \mathbb{R}^{n_{1}+\ldots+n_{k}} equipped with the norm \|x\|=\left(\|x^{1}\|^{p}_{X_{1}}+\ldots+\|x^{k}\|^{p}_{X_{k}}\right)^{\frac{1}{p}}, where x=(x^{1},\ldots,x^{k}) with x^{i}\in\mathbb{R}^{n_{i}} for each i=1,\ldots,k . For p=\infty , we again understand \|x\|=\max\{\|x^{1}\|_{X_{1}},\ldots,\|x^{k}\|_{X_{k}}\} . For a set A\subseteq\mathbb{R}^{n} , its v -translation by a vector v\in\mathbb{R}^{n} is given by A+v=\{a+v:a\in A\} . Its \rho -dilation for \rho\in\mathbb{R} is \rho A=\{\rho a:a\in A\} . We say that A is 0 -symmetric if (-1)A=A , and (centrally) symmetric if A+v is 0 -symmetric for some v\in\mathbb{R}^{n} . The Banach–Mazur distance between two real normed spaces X,Y of the same dimension is defined as d_{BM}(X,Y)=\inf\|T\|\,\|T^{-1}\|, where the infimum (in fact a minimum in the finite-dimensional setting) is taken over all invertible linear operators T:X\to Y and the standard operator norms are used. For convex bodies K,L\subseteq\mathbb{R}^{n} , i.e., compact convex sets with non-empty interior, their Banach–Mazur distance is defined as d_{BM}(K,L)=\inf\{\rho\geq 0:K+u\subseteq T(L+v)\subseteq\rho(K+u)\}, with the infimum (again a minimum) taken over all invertible linear operators T:\mathbb{R}^{n}\to\mathbb{R}^{n} and vectors u,v\in\mathbb{R}^{n} . More generally, if K and L are d -dimensional convex bodies, meaning their affine spans have dimension d , then d_{BM}(K,L) is understood like above after embedding K and L in \mathbb{R}^{d} . When K and L are 0 -symmetric, the translation vectors u and v may be omitted. Under the well-known correspondence of norms and 0 -symmetric convex bodies, the equality d_{BM}(X,Y)=d_{BM}(B_{x},B_{y}) holds for any two normed spaces X=(\mathbb{R}^{n},\|\cdot\|_{X}) , Y=(\mathbb{R}^{n},\|\cdot\|_{Y}) and their respective unit balls B_{X},B_{Y}\subseteq\mathbb{R}^{n} . The n -dimensional Banach–Mazur compactum \mathcal{BM}^{n} is the set of affine equivalence classes of n -dimensional convex bodies equipped with the (multiplicative) metric d_{BM} . By \mathcal{BM}_{s}^{n} we denote the subset of \mathcal{BM}^{n} corresponding to only 0 -symmetric convex bodies, i.e., the n -dimensional symmetric Banach–Mazur compactum. The \ell_{p} -direct sum of normed spaces is one of the fundamental constructions in the geometry of Banach spaces, and many of their aspects have been studied in various different contexts. In particular, estimates of the Banach–Mazur distance when \ell_{p} -sums are involved often play an important role, see, e.g., [ 4 , 5 , 11 , 12 , 13 , 14 , 18 , 19 , 20 , 22 ] . In the vast majority of these works, the Banach–Mazur distance to \ell_{p} -sums is only estimated rather than determined exactly. This reflects a broader fact about Banach–Mazur compacta: most of their properties have been established only in the asymptotic regime, i.e., as the dimension tends to infinity. In part, this is due to exact computations of distances requiring solutions to optimization problems that can be challenging even in very specific situations. One notable example is given by the classical spaces \ell_{1}^{n} and \ell_{\infty}^{n} . Their distance is known to be of order \sqrt{n} , but besides the trivial cases n=1,2 , the exact value has been determined only for n=3 as \frac{9}{5} and n=4 as 2 (see [ 17 ] ). The distance \frac{9}{5} appears to be the largest Banach–Mazur distance computed for 3 -dimensional real normed spaces to date. The main goal of this paper is to investigate some situations in which the Banach–Mazur distance to an \ell_{p} -sum can be determined exactly, which leads us to explicit computations of the Banach–Mazur distance between many different pairs of normed spaces or convex bodies. In concrete cases, obtaining an upper bound on the Banach–Mazur distance is usually straightforward, whereas establishing the matching lower bound is the main challenge, requiring appropriate tools and often restrictions. One of the few instances in which the Banach–Mazur distance to an \ell_{p} -sum has been computed precisely is [ 22 , Proposition 2 ] by Tomczak-Jaegermann. For any integer n\geq 1 and any 1<p\leq 2 , it is proved that d_{BM}(\ell_{p}^{n}\oplus_{p^{*}}\ldots\oplus_{p^{*}}\ell_{p}^{n},\ell_{2}^{n^{2}})=n^{\frac{2-p}{p}}, (1) where p^{*}=\frac{p}{p-1}\in[2,\infty) is the conjugate exponent of p and there are exactly n identical summands. The proof method heavily relies on the large set of symmetries present in this particular sum and thus has limited applicability. We begin with a broad generalization of ( 1 ), computing the Banach–Mazur distance between the Euclidean space and an arbitrary \ell_{p} -sum of finite-dimensional real normed spaces in terms of the distances of the individual summands. The result reads as follows. Theorem 1.1 . Let n_{1},n_{2},\ldots,n_{k}\geq 1 be integers and let X_{i} be an n_{i} -dimensional real normed space for i=1,\ldots,k . Then, for any p\in[1,\infty] , d_{BM}(X_{1}\oplus_{p}\ldots\oplus_{p}X_{k},\ell^{n_{1}+\ldots+n_{k}}_{2})=\|(d_{BM}(X_{1},\ell_{2}^{n_{1}}),\ldots,d_{BM}(X_{k},\ell_{2}^{n_{k}})\|_{r}, where r=\frac{2p}{|p-2|}\in[2,\infty] (with the conventions r=\infty when p=2 and r=2 when p=\infty ). The proof is presented in Section 2 and relies on a recent characterization [ 8 ] of the optimal Banach–Mazur position with respect to the Euclidean ball by an