On the conjectured capillary Blaschke-Santaló inequality

Carlos Cabezas-Moreno, Yingxiang Hu and Mohammad N. Ivaki

arXiv: 2509.20257v3 · PDF · 12 June 2026
Categories: math.DG, math.FA, math.MG
DOI: 10.1112/mtk.70109
Licence: CC BY 4.0 The site shows the text.

Abstract

We prove that the conjectured capillary Blaschke–Santaló inequality holds for any unconditional, strictly convex capillary hypersurface when \theta \in \left(0, \tfrac{\pi}{2}\right). Moreover, for \theta \in \left(\tfrac{\pi}{2}, \pi\right), we show that the capillary volume product has no finite upper bound.

Items 17

Remark

Remark Remark

HFL7

It is clear that \mathcal{C}_theta^* = \mathcal{C}_theta. In general, \Sigma^* is not a convex capillary hypersurface.

Conjecture

Conjecture Conjecture

VT94

[MWW25] Let theta in (0, \frac{pi}{2}) and \Sigma be an even, smooth, strictly convex capillary hypersurface. Then

\operatorname{vol}(\widehat{\Sigma}) \operatorname{vol}(\widehat{\Sigma^*}) \leq \operatorname{vol}(\widehat{\mathcal{C}_theta})^2.

Moreover, equality holds if and only if \Sigma = \lambda \mathcal{C}_theta for some \lambda > 0.

Theorem

Theorem 1.1

Y3A2

Let theta in \left(0, \frac{pi}{2}\right). For any unconditional convex body K (not necessarily smooth or capillary) with the standard support function h_K we have

\frac{\operatorname{vol}(K)}{2n} \int_{\mathbb{S}^{n-1}_theta} \frac{(1 - \costheta \, x_n)^{n+1}}{h_K^n(x)} \, d\sigma(x) \leq \operatorname{vol}(\widehat{\mathcal{C}_theta})^2.

In particular, when \Sigma is an unconditional, smooth, strictly convex capillary hypersurface, we have

\operatorname{vol}(\widehat{\Sigma}) \operatorname{vol}(\widehat{\Sigma^*}) \leq \operatorname{vol}(\widehat{\mathcal{C}_theta})^2.

Moreover, equality holds if and only if \Sigma = \lambda \mathcal{C}_theta for some \lambda > 0.

Theorem

Theorem 1.2

IBI0

Let theta in \left(0, \frac{pi}{2}\right) and let f in C^2(\mathcal{C}_theta) be an unconditional function satisfying \bar{\nabla}_\mu f = 0, where \mu denotes the unit normal vector to \partial \mathcal{C}_theta \subset \mathcal{C}_theta pointing towards the lower half-space. Then

\int_{\mathcal{C}_theta} f \left( \Delta_{\mathcal{C}_theta} f + 2n f \right) \, dV_{\mathcal{C}_theta} \leq 2n \frac{\left( \int_{\mathcal{C}_theta} f \, dV_{\mathcal{C}_theta} \right)^2}{\int_{\mathcal{C}_theta} dV_{\mathcal{C}_theta}}.

Lemma

Lemma 2.1

3PEC

The function x\mapsto V(\sqrt{x}) is concave in (0,\infty)^n.

Proof of Lemma 2.1

Let s=x_1+\cdots+x_{n-1} and x_n=t. Define

\begin{alignedat}{2} v(s,t)=V(\sqrt{x})=\frac{1}{2\sin^4theta}\left((1+\cos^2theta)t+(\sin^2theta)s +2\costheta \sqrt{t^2+s t\sin^2theta}\right). \end{alignedat}

Then the Hessian of v (a function of two variables s,t) is

\begin{alignedat}{2} D^2v=\begin{bmatrix} v_{ss} & v_{st}\\ v_{ts} & v_{tt} \end{bmatrix}=-\frac{\costheta}{4(t^2+s t\sin^2theta)^\frac{3}{2}}\begin{bmatrix} t^2 & -st \\ -st & s^2 \end{bmatrix}. \end{alignedat}

From this we obtain

D^2V(\sqrt{x}) = -\frac{\costheta}{4(t^2+s t\sin^2theta)^\frac{3}{2}} \begin{bmatrix} t^2 & \cdots & t^2 & -st \\ \vdots & \ddots & \vdots & \vdots \\ t^2 & \cdots & t^2 & -st \\ -st& \cdots & -st & s^2 \end{bmatrix}.

Recall thetain (0,\frac{pi}{2}). Then for any a=(a_1,\ldots, a_n)in \mathbb{R}^n, there holds

\begin{alignedat}{2} D^2V(\sqrt{x})(a,a)=-\frac{\costheta}{4(t^2+s t\sin^2theta)^{\frac{3}{2}}}\big(t\sum_{i=1}^{n-1}a_i-sa_n\big)^2 \leq 0. \end{alignedat}

Lemma

Lemma 2.2

EMGD

The function V^{\ast} is of class C^1, and for y=(y',y_n) satisfying |y_n| \leq |y| \cos theta, we have

DV^{\ast}(y) = \sin^2 theta \, (y', 0).

Moreover, we have

\begin{alignedat}{2} \det D^2V^{\ast}(y) = \begin{cases} 0 & \text{if } |y_n| < |y| \cos theta , \\ (1-\frac{|y_n|}{|y|}\costheta)^{n+1} & \text{if } |y_n| > |y| \cos theta. \end{cases} \end{alignedat}

In particular, the Monge-Ampère measure of V^{\ast} is

\begin{alignedat}{2} \mu_{V^{\ast}}(A)=\int_{A\cap\{y:\, |y_n|>|y|\costheta \}}\(1-\frac{|y_n|}{|y|}\costheta\)^{n+1}\, dy, \end{alignedat}

for any Borel set A.

Proof of Lemma 2.2

Since C is strictly convex, by [Schneider, Cor. 1.7.3], h_C is differentiable at y \neq 0. Since V^{\ast} is two-homogeneous, we have DV^{\ast}(0) = 0. Moreover, a convex differentiable function on \mathbb{R}^n is C^1; see [Schneider, Thm. 1.5.4].

The second claim follows from a well-known identity: a smooth, strictly convex hypersurface with support function h and Gauss curvature \mathcal{K} satisfies

\begin{alignedat}{2} \det D^2\frac{1}{2}h^2(x)=\frac{h^{n+1}}{\mathcal{K}}\left(\frac{x}{|x|}\right). \end{alignedat}

Here the Gauss curvature is regarded as a function on the unit sphere via the Gauss map. See, for instance, [MP14, Thm. 3.1].

Lemma

Lemma 2.3

A2X2

Let A=(0,\infty)^n and B=\{ y in (0,\infty)^n :\, y_n> |y|\costheta \}. Then D V(A) =B, DV: A\to B is a C^{\infty} diffeomorphism and

\begin{alignedat}{2} \det D^2V(x)\det D^2V^{\ast}(DV(x))=1\quad \forall xin A. \end{alignedat}

Proof of Lemma 2.3

Let r=\sqrt{x_n^2+|x'|^2\sin^2theta} and define \phiin (0,\frac{pi}{2}) through \cos \phi=\frac{x_n}{r} and \sin \phi=\frac{\sintheta |x'|}{r}. Let y=(y',y_n)=DV(x). Then

\begin{alignedat}{2} y_n&=\frac{r}{\sin^4theta}(1+\costheta \cos\phi)(\costheta+\cos\phi),\\ y'&=\frac{1+\costheta\cos\phi}{\sin^2theta}x',\quad |y'|=\frac{r}{\sin^3theta}(1+\costheta\cos\phi)\sin\phi. \end{alignedat}

Define f(\phi)=\frac{\sintheta \sin\phi}{\costheta+\cos\phi}=\frac{|y'|}{y_n} for \phiin (0,\frac{pi}{2}). Since f is strictly increasing and maps (0,\frac{pi}{2}) onto (0,\tantheta), we have 0<|y'|/y_n<\tantheta, which is equivalent to y_n>|y|\costheta. Hence DV(A)\subset B.

Let yin B. We choose \phiin (0,\frac{pi}{2}) such that f(\phi)=|y'|/y_n, and define

\begin{alignedat}{2} r=\frac{\sin^4theta\, y_n}{(1+\costheta\cos\phi)(\costheta+\cos\phi)}. \end{alignedat}

Then

\begin{alignedat}{2} x=r\left(\frac{\sin\phi}{\sintheta}\frac{y'}{|y'|},\cos\phi\right)in A \quad \text{and} \quad DV(x)=y. \end{alignedat}

Hence B\subset DV(A).

Note that V=\frac{1}{2}p_C^2 is C^\infty-smooth and strictly convex on the open set A, and thus the gradient map DV: A \to B is bijective. By the inverse function theorem, since D^2 V(x) is positive definite for x in A, the map DV is a diffeomorphism. Moreover, from the definition of the Legendre transform, for y in B, we have DV^{\ast}(y) = x, where x in A satisfies y = DV(x). Thus, DV^{\ast}(DV(x))=x and the relation between the determinants follows.

Theorem

Theorem Theorem (Prékopa–Leindler inequality, equality by Dubuc)

I2E7

[Pre71, Lei72, Bor75, Dub77] Let \lambda in (0, 1) and let f, g and h be non-negative integrable functions on \mathbb{R}^n. Assume that

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

for all x, y in \mathbb{R}^n. Then

\int_{\mathbb{R}^n} h \geq \left( \int_{\mathbb{R}^n} f \right)^{1-\lambda} \left( \int_{\mathbb{R}^n} g \right)^{\lambda}.

If f and g have positive integrals and equality holds, then there exist a > 0, w in \mathbb{R}^n, and a log-concave function \tilde{h}, such that h = \tilde{h}, f(x) = a^{\lambda} \tilde{h}(x +\lambda w), and g(x) = a^{ -(1-\lambda)} \tilde{h}(x -(1- \lambda) w) for almost every x.

Lemma

Lemma 2.4

TINT

Let A=\operatorname{diag}(a_1,\ldots,a_n) with a_i>0. If A\mathcal{C}_{theta} is theta-capillary, then for all i we have a_i=a_n.

Proof of Lemma 2.4

Let x in \mathbb{S}^{n-1}_{theta}. The outer unit normal to A \mathbb{S}^{n-1}_{theta} at Ax is given by

\begin{alignedat}{2} \nu(x) = \frac{A^{-1}x}{|A^{-1}x|}, \end{alignedat}

and we have

\begin{alignedat}{2} \langle \nu(x), E_n \rangle = \frac{x_n}{a_n\sqrt{\sum_{i=1}^n \tfrac{x_i^2}{a_i^2}}}. \end{alignedat}

In particular, for i=1,\ldots,n-1, the point e_i = \sintheta\, E_i + \costheta\, E_n in \partial \mathbb{S}^{n-1}_{theta} satisfies

\begin{alignedat}{2} \langle \nu(e_i), E_n \rangle = \frac{\costheta}{\sqrt{\cos^2theta + \tfrac{a_n^2}{a_i^2}\sin^2theta}}. \end{alignedat}

If A \mathcal{C}_{theta} is theta-capillary, then there holds

\begin{alignedat}{2} \cos^2theta + \frac{a_n^2}{a_i^2}\sin^2theta = 1, \quad i=1,\ldots,n-1. \end{alignedat}

Hence, a_i = a_n for all i.

Corollary

Corollary 2.5

DM6K

Let \alpha in (-n, 0). For an unconditional convex body K define

\begin{alignedat}{2} \mathcal{E}_\alpha(K) = \int_{\mathbb{S}_{theta}^{n-1}} \left( \frac{h_K}{h_C} \right)^\alpha h_C\, d\sigma. \end{alignedat}

If \operatorname{vol}(K) = \operatorname{vol}(C), then

\begin{alignedat}{2} \mathcal{E}_\alpha(K) \leq \mathcal{E}_\alpha(C) &= \int_{\mathbb{S}_{theta}^{n-1}} h_C\, d\sigma, \\ \int_{\mathbb{S}_{theta}^{n-1}} h_C \log h_K\, d\sigma &\geq \int_{\mathbb{S}_{theta}^{n-1}} h_C \log h_C\, d\sigma. \end{alignedat}

Moreover, equality holds only when K = C.

Proof of Corollary 2.5

The inequalities follow from Jensen’s inequality and (?). If equality holds in either inequality, then equality must also hold in Jensen’s inequality, and therefore h_K/h_C is constant on \mathbb{S}^{n-1}_theta and hence K=C.

Example

Example 1

8N7J

We construct a family of C^{1,1}-smooth convex bodies that are smooth near x_{n} = 0 and meet this hyperplane at the constant angle theta in (\frac{pi}{2}, pi). In this example, we write points in \mathbb{R}^n in the form (x,y) with xin \mathbb{R}^{n-1} and yin \mathbb{R}. For \lambda > 0, define

\begin{aligned} K_{\lambda} &\coloneqq \left\{ (x, y) in \frac{1}{\lambda} \widehat{\mathcal{C}_theta} : y \leq -\frac{\cos theta}{\lambda} \right\} \\ &\quad \cup \left\{ (x, y) in \mathbb{R}^n : |x| \leq \frac{1}{\lambda}, -\frac{\cos theta}{\lambda} \leq y \leq \lambda^{n-1} \right\} \cup \left( \frac{1}{\lambda} B^n + \lambda^{n-1} E_n \right). \end{aligned}

Example

Example 2

7N1U

We consider the intersection of an ellipse and the half-space y\geq 0 given by:

\begin{alignedat}{2} K= \left\{(x,y): \frac{x^2}{a^2} + \frac{(y - c)^2}{b^2} \leq 1, \quad y \geq 0\right\}, \end{alignedat}

where a > 0, b > c > 0 are constants to be determined later. Let us put \Sigma=\overline{\partial K\cap \mathbb{R}^2_+}. Suppose the contact angle of \Sigma with y=0 is thetain (\frac{pi}{2},pi).

Source and licence

The text on this page is from arXiv:2509.20257v3 by Carlos Cabezas-Moreno, Yingxiang Hu and Mohammad N. Ivaki, under the licence CC BY 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 4.0.

The record of each item is in analysis/objects/2509/2509.20257.jsonl.