Paperlane

no. 16 · Tuesday 17 November 2026

The norm formula in a Banach space and in its complete theory

Pedro Poitevin

math.LO · v2

Abstract

Let $h$ be the Hilbert norm on the real space $\ell_2(\mathbb{N})\oplus_2\ell_2(\mathbb{N})$, and put $q(u,v)=\sup_{i<j}|u_i+v_j|$. For each of Bollobás's additive norms $h+\varepsilon q$, $\varepsilon>0$, and for the quadratic norm $\sqrt{h^2+q^2}$, the norm formula $\lVert x+y\rVert$ is unstable on the renormed space and has no independence property in its complete theory. We approximate $q(x+y)$ uniformly on bounded Hilbert balls by truncating small coordinates. For $n$ fixed test vectors and thresholds, the number of threshold patterns of $q(x+y)$ as $y$ varies is bounded by a polynomial in $n$ when the support sizes of the test vectors and of $y$ are bounded. These estimates give finite fat-shattering dimension for the translates of $q$ at every positive scale. The norm formula on a space can be stable, unstable without the independence property, or have the independence property. We give an example in each of these three cases where the formula has the independence property in the complete theory, and an example in each of the first two cases where it lacks that property in the complete theory. All five spaces satisfy a common midpoint estimate for uniform convexity, and each admits an isomorphism to Hilbert space of distortion at most $\sqrt3$.

comments
Revised abstract and introduction, with citations to Lupini
fields
Logic, Functional Analysis
pdf ↗ page arXiv:2611.15930 ↗