A general functional version of Grünbaum’s inequality

David Alonso-Gutiérrez, Francisco Marín Sola, Javier Martín Goñi and Jesús Yepes Nicolás

arXiv: 2404.08319v1 · PDF · 12 April 2024
Categories: math.MG, math.FA
Licence: CC BY SA 4.0 The site shows the text.

Abstract

A classical inequality by Grünbaum provides a sharp lower bound for the ratio \mathrm{vol}(K^{-})/\mathrm{vol}(K), where K^{-} denotes the intersection of a convex body with non-empty interior K\subset\mathbb{R}^n with a halfspace bounded by a hyperplane H passing through the centroid \mathrm{g}(K) of K. In this paper we extend this result to the case in which the hyperplane H passes by any of the points lying in a whole uniparametric family of r-powered centroids associated to K (depending on a real parameter r\geq0), by proving a more general functional result on concave functions. The latter result further connects (and allows one to recover) various inequalities involving the centroid, such as a classical inequality (due to Minkowski and Radon) that relates the distance of \mathrm{g}(K) to a supporting hyperplane of K, or a result for volume sections of convex bodies proven independently by Makai Jr. & Martini and Fradelizi.

Items 13

Theorem

Theorem A

M6AA

Let K\subset\mathbb{R}^n be a convex body with non-empty interior and let H be a hyperplane. If K has centroid at the origin then

\frac{\mathrm{vol}_1\bigl(K^-|H^\perp\bigr)}{\mathrm{vol}_1\bigl(K|H^\perp\bigr)}\geq \frac{1}{n+1}.

Theorem

Theorem B

JPTE

Let K\subset\mathbb{R}^n be a convex body with non-empty interior, let H be a hyperplane and let f:[a,b]\longrightarrow \mathbb{R}_{\geq 0} be the function given by f(t) = \mathrm{vol}_{n-1}\bigl({K \cap (tu + H)}\bigr). If K has centroid at the origin then

\frac{f(0)}{\|f\|_{\infty}}\geq \left(\frac{n}{n+1}\right)^{n-1}.

Theorem

Theorem C

6BTY

Let K\subset\mathbb{R}^n be a compact set with non-empty interior and with centroid at the origin. Let H be a hyperplane such that the function f:H^{\bot}\longrightarrow\mathbb{R}_{\geq0} given by f(x)=\mathrm{vol}_{n-1}\bigl(K\cap(x+H)\bigr) is p-concave, for some p\in(0,\infty). Then

\frac{\mathrm{vol}(K^{-})}{\mathrm{vol}(K)}\geq\left(\frac{p+1}{2p+1}\right)^{(p+1)/p}.

The inequality is sharp.

Theorem

Theorem 1.1

MAJS

Let r\in[0,\infty) and let K\subset\mathbb{R}^n be a compact set with non-empty interior having the point \mathrm{g}_{r}\cdot u, with respect to some direction u\in\mathbb{S}^{n-1}, at the origin. Let H=\{x \in \mathbb{R}^{n}: \left\langle x,u\right\rangle=0\} be the hyperplane with normal vector u and assume that the function f:H^{\bot}\longrightarrow\mathbb{R}_{\geq0} given by f(x)=\mathrm{vol}_{n-1}\bigl(K\cap(x+H)\bigr) is p-concave, for some p\in(0,\infty). If r\geq1 then

\frac{\mathrm{vol}(K^{-})}{\mathrm{vol}(K)}\geq\left(\frac{p+1}{2p+r}\right)^{(p+1)/p},

whereas if 0\leq r\leq1 then

\frac{\mathrm{vol}(K^{-})}{\mathrm{vol}(K)}\geq\left(\frac{p+r}{2p+r}\right)^{(p+1)/p}.

Used by Remark 1.1

Theorem

Theorem 1.2

KPBZ

Let h:[a,b]\longrightarrow[0,\infty) be a non-negative concave function, and let \alpha,\beta>0. If \beta\leq\alpha then

\frac{\int_{\mathrm{g}_{\alpha}(h)}^bh^\beta(t)\,\mathrm{d} t} {\int_{a}^bh^\beta(t)\,\mathrm{d} t} \geq\left(\frac{\beta+1}{\alpha+2}\right)^{\beta+1},

whereas if \alpha\leq\beta then

\frac{\int_{\mathrm{g}_{\alpha}(h)}^bh^\beta(t)\,\mathrm{d} t} {\int_{a}^bh^\beta(t)\,\mathrm{d} t} \geq\left(\frac{\alpha+1}{\alpha+2}\right)^{\beta+1}.

Remark

Remark 1.1

Q5R6

We observe that Theorem 1.1 is directly obtained from the previous result by just taking h=f^{p}, \beta=1/p and \alpha=r\beta (where the case r=0 is derived when doing \alpha\to0^+). Moreover, Theorem 1.2 can be shown from Theorem 1.1 by just considering the set of revolution K associated to the radius function r=(1/\kappa_{n-1})f^{1/(n-1)}, where \kappa_{n-1} is the (n-1)-dimensional volume of the Euclidean unit ball in \mathbb{R}^{n-1}, f=h^\beta, p=1/\beta and r=\alpha/\beta. Therefore, in fact, both results (Theorems 1.1 and 1.2) are equivalent.

Uses Theorem 1.1

Proposition

Proposition 2.1

R25S

Let K\subset\mathbb{R}^n be a compact set with non-empty interior and with midpoint, with respect to some direction u\in\mathbb{S}^{n-1}, at the origin. Let H=\{x \in \mathbb{R}^{n}: \left\langle x,u\right\rangle=0\} be the hyperplane with normal vector u and assume that the function f:H^{\bot}\longrightarrow\mathbb{R}_{\geq0} given by f(x)=\mathrm{vol}_{n-1}\bigl(K\cap(x+H)\bigr) is p-concave, for some p\in(0,\infty). Then

\frac{\mathrm{vol}(K^{-})}{\mathrm{vol}(K)}\geq\left(\frac{1}{2}\right)^{(p+1)/p}.

Proof of Proposition 2.1

In the proof of [MSYN, Theorem 1.1] it is shown that there exists a p-affine function g_p:[-\gamma,\delta]\longrightarrow\mathbb{R}_{\geq0} given by g_p(t)=c(t+\gamma)^{1/p}, for some \gamma,\delta,c>0, such that g_p(0)=f(0),

\int_{-\gamma}^0 g_p(t)\,\mathrm{d} t=\int_a^0 f(t)\,\mathrm{d} t \quad \text{and} \quad \int_0^\delta g_p(t)\,\mathrm{d} t=\int_0^b f(t)\,\mathrm{d} t,

and further that -\gamma\leq a<0<\delta\leq b.

Here, since K has its midpoint (w.r.t. u) at the origin, we have that a=-b and then we get -\gamma+\delta\leq0. Thus

\frac{\mathrm{vol}(K^-)}{\mathrm{vol}(K)}=\frac{\int_{-\gamma}^0 g_p(t)\,\mathrm{d} t}{\int_{-\gamma}^\delta g_p(t)\,\mathrm{d} t} \geq\frac{\int_{-\gamma}^{(-\gamma+\delta)/2} g_p(t)\,\mathrm{d} t}{\int_{-\gamma}^\delta g_p(t)\,\mathrm{d} t} =\left(\frac{1}{2}\right)^{(p+1)/p},

as desired.

Theorem

Theorem D (The Borell-Brascamp-Lieb inequality)

DVTT

Let \lambda\in(0,1). Let -1/n\leq p\leq\infty and let f,g,h:\mathbb{R}^n\longrightarrow\mathbb{R}_{\geq0} be measurable functions with positive integrals such that

h((1-\lambda)x + \lambda y)\geq \bigl((1-\lambda)f(x)^p+\lambda g(y)^p\bigr)^{1/p}

for all x,y\in\mathbb{R}^n with f(x)g(y)>0. Then

\int_{\mathbb{R}^n}h(x)\,\mathrm{d} x\geq \left[(1-\lambda)\left(\int_{\mathbb{R}^n}f(x)\,\mathrm{d} x\right)^q+\lambda\left(\int_{\mathbb{R}^n}g(x)\,\mathrm{d} x\right)^q\right]^{1/q},

where q=p/(np+1).

Used by Proof, Proof

Corollary

Corollary 2.1

FMH7

Let r\in(0,\infty) and let K\subset\mathbb{R}^n be a compact set with non-empty interior having the point \mathrm{g}_{r}\cdot u, with respect to some direction u\in\mathbb{S}^{n-1}, at the origin. Let H=\{x \in \mathbb{R}^{n}: \left\langle x,u\right\rangle=0\} be the hyperplane with normal vector u and assume that the function f:\mathbb{R}\longrightarrow\mathbb{R}_{\geq0} given by f(t)=\mathrm{vol}_{n-1}\bigl(K\cap(tu+H)\bigr) is p-concave, for some p\in(0,\infty). Then

\frac{\mathrm{vol}(K^{-})}{\mathrm{vol}(K)} \geq \left(\frac{p}{2p + r}\right)^{(p+1)/p}.

Proof of Corollary 2.1

Denoting by [a,b] the support of f, let \bar{f},g:[a,b]\longrightarrow\mathbb{R}_{\geq0} be the functions given by

g(t)=\mathrm{vol}\bigl(K \cap (tu + H^{-})\bigr)^{p/(p+1)}=\left(\int_a^t f(s)\,\mathrm{d} s\right)^{p/(p+1)}

and

\bar{f}(t)=\mathrm{vol}_{n-1}\bigl(K\cap(tu+H)\bigr)^r=f(t)^r.

By hypothesis, it is clear that \bar{f} is (p/r)-concave, whereas from the Borell-Brascamp-Lieb inequality (D) we have that g is concave on [a,b]. So, from Proposition 2.2 applied to the functions \bar{f} and g, and taking into account that \mathrm{g}_{r}=0, we get that

\begin{split} \mathrm{vol}(K^{-}) = g\left(\frac{\int_a^b t \bar{f}(t) \,\mathrm{d} t}{\int_a^b \bar{f}(t)\,\mathrm{d} t}\right)^{(p+1)/p} &\geq \left(\frac{p}{2p + r}\right)^{(p+1)/p}\|g\|_{\infty}^{(p+1)/p}\\ &= \left(\frac{p}{2p + r}\right)^{(p+1)/p}\mathrm{vol}(K), \end{split}

as desired.

Uses Theorem D, Proposition 2.2

Proposition

Proposition 2.2

PCCE

Let K \subset \mathbb{R}^n be a convex body. Let g : K \longrightarrow \mathbb{R}_{\geq 0} be a concave function and let f : K \longrightarrow \mathbb{R}_{\geq 0} be a p-concave function, with p>0. Then

g\left(\frac{\int_{K} x f(x) \,\mathrm{d} x}{\int_{K} f(x) \,\mathrm{d} x}\right) \geq \left(\frac{p}{(n + 1)p + 1}\right)\|g\|_{\infty}.

Used by Proof

Proof of Proposition 2.2

Let \mu be the probability measure whose density function is given by

\mathrm{d}\mu(x) = \frac{f(x)}{\int_K f(x) \,\mathrm{d} x}\mathrm{d} x.

Since g is concave, using Jensen’s inequality we get that

\begin{split} g\left(\frac{\int_{K} x f(x) \,\mathrm{d} x}{\int_{K} f(x) \,\mathrm{d} x}\right) = g\left(\int_K x \, \mathrm{d}\mu(x)\right) &\geq \int_K g(x) \,\mathrm{d}\mu(x)\\ &= \int_0^{\|g\|_{\infty}}\mu\bigl(\{x \in K : g(x) \geq t\}\bigr)\,\mathrm{d} t, \end{split}

where in the last identity we have used Fubini’s theorem. Now, since the density of \mu, with respect to the Lebesgue measure, is p-concave, from the Borell-Brascamp-Lieb inequality (D) we have that the function \varphi(t) := \mu\bigl(\{x \in K : g(x) \geq t\}\bigr) is \bigl(p/(np + 1)\bigr)-concave. Indeed, it is enough to apply Theorem D with the functions

\frac{f\cdot\chi_{_{\{x\in K\,:\, g(x)\geq t_1\}}}}{\int_K f(x) \,\mathrm{d} x}, \quad \frac{f\cdot\chi_{_{\{x\in K\,:\, g(x)\geq t_2\}}}}{\int_K f(x) \,\mathrm{d} x} \quad \text{ and } \quad \frac{f\cdot\chi_{_{\{x\in K\,:\, g(x)\geq (1-\lambda)t_1+\lambda t_2\}}}}{\int_K f(x) \,\mathrm{d} x}

for any t_1,t_2\in\mathbb{R}, where \chi_{_M} denotes the characteristic function of the set M, one easily obtains the desired concavity of \varphi. Hence, and taking into account that \varphi(0)=1 and \varphi\bigl(\|g\|_{\infty}\bigr)\geq0, we may assure that \varphi(t)^{p/(np + 1)} \geq (1 - t/\|g\|_{\infty}) for all t \in [0,\|g\|_{\infty}]. So, by integrating we obtain that

\begin{split} \int_0^{\|g\|_{\infty}}\mu\bigl(\{x \in K : g(x) \geq t\}\bigr)\,\mathrm{d} t &\geq \int_0^{\|g\|_{\infty}}(1 - t/\|g\|_{\infty})^{(np + 1)/p}\,\mathrm{d} t\\ &= \left(\frac{p}{(n + 1)p + 1}\right)\|g\|_{\infty}, \end{split}

from where the result immediately follows.

Uses Theorem D

Source and licence

The text on this page is from arXiv:2404.08319v1 by David Alonso-Gutiérrez, Francisco Marín Sola, Javier Martín Goñi and Jesús Yepes Nicolás, under the licence CC BY SA 4.0. The corpus made these changes: it read the items out of the LaTeX source, it expanded the macros of the paper, it put each proof after its statement, and it converted the text to HTML. The formulas are the TeX of the paper, and KaTeX renders them. A formula or a reference can look different from the PDF. The page is published under CC BY SA 4.0.

The record of each item is in analysis/objects/2404/2404.08319.jsonl.