OctoInformBeta

OpenAI preprint, September 2026 · §1. Introduction

Theorem 1.1: For 0≤λ≤10\le\lambda\le1, |W[hK1−λhLλ]|≥|K|1−λ|L|λ|W[h_K^{1-\lambda}h_L^{\lambda}]|\ge|K|^{1-\lambda}|L|^{\lambda}, and a variance inequality for soft slab laws implies it

Origin-symmetric convex bodies satisfy the logarithmic Brunn-Minkowski inequality.

Hover over a blue or green term in the text, or a node in the graph, to find it in the other. Select it to read its declaration both in Lean and in words.

The Setting

For convex bodies KK and LL in ℝn\mathbb{R}^n and 0≤λ≤10\le\lambda\le1, the Brunn-Minkowski inequality says that the Minkowski combination (1−λ)K+λL(1-\lambda)K+\lambda L, the body with support function (1−λ)hK+λhL(1-\lambda)h_K+\lambda h_L, has volume at least |K|1−λ|L|λ|K|^{1-\lambda}|L|^{\lambda}, where hKh_K is the support function and |K||K| the volume of KK. The logarithmic Brunn-Minkowski conjecture of Böröczky, Lutwak, Yang and Zhang replaces the arithmetic mean of the support functions by the geometric mean hK1−λhLλh_K^{1-\lambda}h_L^{\lambda}, whose Wulff body is contained in the Minkowski combination. It asks whether, for KK and LL symmetric about the origin, this smaller body still has volume at least |K|1−λ|L|λ|K|^{1-\lambda}|L|^{\lambda}.✦

The library states this inequality together with a reduction of it to a variance inequality for the soft slab laws: log-concave probability measures on ℝn\mathbb{R}^n whose density is a Gaussian times, for each of finitely many slabs symmetric about the origin, a smoothed indicator of the slab.

Convex Bodies and Wulff Bodies

Euclidean space ℝn\mathbb{R}^n carries the inner product ⟨x,u⟩\langle x,u\rangle and the norm |x||x|, and |K||K| is the Lebesgue volume of a set K⊂ℝnK\subset\mathbb{R}^n, a number in [0,∞][0,\infty]. A convex body in ℝn\mathbb{R}^n is a compact convex subset of ℝn\mathbb{R}^n with nonempty interior.✦ A set K⊂ℝnK\subset\mathbb{R}^n is origin-symmetric when K=−KK=-K.✦

The support function of a set K⊂ℝnK\subset\mathbb{R}^n is the function hKh_K on ℝn\mathbb{R}^n given by hK(u)=supx∈K⟨x,u⟩.h_K(u)=\sup_{x\in K}\langle x,u\rangle .(1)✦ For a real function ff on ℝn\mathbb{R}^n, the Wulff body of ff is the subset of ℝn\mathbb{R}^n W[f]={x∈ℝn:⟨x,u⟩≤f(u)for every unit vector u}.W[f]=\{x\in\mathbb{R}^n : \langle x,u\rangle\le f(u)\ \text{for every unit vector } u\}.(2)✦ For sets KK and LL in ℝn\mathbb{R}^n and a real number λ\lambda, the logarithmic combination of KK and LL with parameter λ\lambda is the Wulff body W[hK1−λhLλ]W[h_K^{1-\lambda}h_L^{\lambda}].✦

Soft Slab Laws

Take a finite nonempty index set II and, for each i∈Ii\in I, a vector pi∈ℝnp_i\in\mathbb{R}^n and a width hi>0h_i>0, together with an exponent q∈ℕq\in\mathbb{N} and a number ε>0\varepsilon>0. The slab polynomial of pip_i and hih_i is the qq-th power of the linear form x↦⟨pi,x⟩/hix\mapsto\langle p_i,x\rangle/h_i: ψi(x)=(⟨pi,x⟩hi)q.\psi_i(x)=\Bigl(\frac{\langle p_i,x\rangle}{h_i}\Bigr)^{q}.(3)✦

For real weights βi\beta_i and a parameter t∈ℝt\in\mathbb{R}, the regularized potential of the family (ψi)i∈I(\psi_i)_{i\in I} is the function Vt(x)=ε2|x|2+∑i∈Ieβitψi(x),V_t(x)=\frac{\varepsilon}{2}|x|^2+\sum_{i\in I}e^{\beta_i t}\psi_i(x),(4)✦ and for a natural number kk, the kk-th derivative polynomial of the family (ψi)i∈I(\psi_i)_{i\in I} is ∑i∈Iβikeβitψi\sum_{i\in I}\beta_i^k e^{\beta_i t}\psi_i✦, the kk-th derivative in tt of the sum in (4).

For each t∈ℝt\in\mathbb{R}, the Gibbs law μt\mu_t of the family (ψi)i∈I(\psi_i)_{i\in I} is the measure Zt−1e−Vt(x)dxZ_t^{-1}e^{-V_t(x)}\,dx on ℝn\mathbb{R}^n, where e−Vte^{-V_t} is the Gibbs density of μt\mu_t and Zt=∫ℝne−Vt(x)dxZ_t=\int_{\mathbb{R}^n}e^{-V_t(x)}\,dx is its partition function. Under μt\mu_t, a function FF on ℝn\mathbb{R}^n has mean 𝔼μtF=Zt−1∫ℝnF(x)e−Vt(x)dx\mathbb{E}_{\mu_t}F=Z_t^{-1}\int_{\mathbb{R}^n}F(x)e^{-V_t(x)}\,dx✦ and variance Var⁡μtF=𝔼μt(F−𝔼μtF)2\operatorname{Var}_{\mu_t}F=\mathbb{E}_{\mu_t}(F-\mathbb{E}_{\mu_t}F)^2.✦

The soft slab law μ\mu of the slab polynomials ψi\psi_i is the Gibbs law μ0\mu_0, the measure on ℝn\mathbb{R}^n with potential V(x)=ε2|x|2+∑i∈Iψi(x)V(x)=\tfrac{\varepsilon}{2}|x|^2+\sum_{i\in I}\psi_i(x) and partition function Z=Z0Z=Z_0.

The Claim

Theorem 1. Let n≥1n\ge1, let KK and LL be origin-symmetric convex bodies in ℝn\mathbb{R}^n, and let 0≤λ≤10\le\lambda\le1. Then |W[hK1−λhLλ]|≥|K|1−λ|L|λ.\bigl|W[h_K^{1-\lambda}h_L^{\lambda}]\bigr|\ \ge\ |K|^{1-\lambda}\,|L|^{\lambda}.(5)✦

Theorem 2. Suppose that for every dimension n≥2n\ge2, every finite nonempty family of nonzero vectors pi∈ℝnp_i\in\mathbb{R}^n with widths hi>0h_i>0, all real numbers bib_i, every even integer q≥2q\ge2 and every ε>0\varepsilon>0, the soft slab law μ\mu of the slab polynomials ψi\psi_i of (3) satisfies the inequality Var⁡μ(∑i∈Ibiψi)≤𝔼μ(∑i∈Ibi2ψi).\operatorname{Var}_{\mu}\Bigl(\sum_{i\in I}b_i\psi_i\Bigr)\ \le\ \mathbb{E}_{\mu}\Bigl(\sum_{i\in I}b_i^2\psi_i\Bigr).(6)✦ Then (5) holds for every n≥1n\ge1, all origin-symmetric convex bodies KK and LL in ℝn\mathbb{R}^n and every λ∈[0,1]\lambda\in[0,1].

What Is Checked

The inequality (5) is stated over convex bodies and origin-symmetric sets in Euclidean space ℝn\mathbb{R}^n, the support function (1), the Wulff body (2) and Lebesgue volume, and the variance inequality (6) over the soft slab law of the slab polynomials (3), with its regularized potential (4), partition function, mean and variance.