betaFeedback

WorksMathematics

Alexandrov's theorem in the anti-Gaussian space: A Reilly-formula approach

Abstract

We give an alternative proof of Alexandrov's theorem about constant weighted mean curvature hypersurfaces in the anti-Gaussian space, due to the authors and Johne-Silini independently, via a Reilly-type formula approach. Moreover, we establish a sharp Heintze-Karcher-type inequality for hypersurfaces in manifolds with a generalized substatic condition, together with its rigidity in Euclidean space with density.

Introduction

Alexandrov’s soap bubble theorem asserts that every closed, connected, embedded hypersurface of constant mean curvature in Euclidean space is a round sphere [1]. Alexandrov’s original proof uses his moving plane method. An alternative proof, which is based on the Heintze–Karcher inequality and the Minkowski formula, was later given by Reilly [20] and Ros [21]. The Heintze–Karcher inequality has been first established by Heintze-Karcher using the normal exponential map [10], while Reilly’s integral formula provides another approach through a boundary value problem [20, 21].

Heintze–Karcher’s inequality and Alexandrov’s theorem has been studied in other space forms [16, 13]. Moreover, Alexandrov’s theorem for a class of substatic warped product manifolds has been proved by Brendle [6], via establishment of Heintze–Karcher’s inequality by means of the normal exponential map with respect to a conformal metric. We recall the concept of sub-staticity. A Riemannian triple (M,g‾,V)(M,\overline{g},V) is called sub-static if V∈C∞(M)V \in C^\infty(M) such that V≥0V \ge0 in MM with V=0V=0 only on ∂M\partial M, and

Δ‾Vg‾−∇‾2V+VRic⁡‾≥0.(1)\overline{\Delta}V\overline{g}-\overline{\nabla}^{2}V+V\overline{\operatorname{Ric}}\ge0. \tag*{(1)}

Li-Xia [14] (see also Qiu-Xia [19]) has established a generalized Reilly-type formula which fits perfectly for substatic manifolds and used it to generalize Brendle’s Heintze–Karcher inequality to general substatic manifolds. Later, Borghini-Fogagnolo-Pinamonti [5] further proved the rigidity for such inequality.

Manifolds with density attracts many attention in recent decades. Let (Mm+1,g‾)(M^{m+1},\overline{g}) be a Riemannian manifold endowed with a smooth weighted measure dμϕ=e−ϕdμg‾d\mu_\phi=e^{-\phi}d\mu_{\overline{g}}, where ϕ∈C∞(M)\phi\in C^\infty(M) is called the potential function. If Ω⊂M\Omega\subset M is a domain with boundary Σ=∂Ω\Sigma=\partial\Omega, its weighted volume and the weighted area of Σ\Sigma are defined, respectively, by

Vol⁡ϕ(Ω)=∫Ωdμϕ,Area⁡ϕ(Σ)=∫Σdσϕ=∫Σe−ϕdσ.\operatorname{Vol}_{\phi}(\Omega)=\int_{\Omega}d\mu_{\phi},\qquad\operatorname{Area}_{\phi}(\Sigma)=\int_{\Sigma}d\sigma_{\phi}=\int_{\Sigma}e^{-\phi}d\sigma.

A particularly important model is Euclidean space equipped with a radial log-convex density. Manifolds with density and their associated isoperimetric problems have been extensively investigated; see, for example, [17, 22]. In particular, the Log-Convex Density Conjecture, commonly attributed to Wolfe [22], has been resolved in full by Chambers [7], which asserts that balls centered at the origin are the only isoperimetric regions in Rm+1\mathbb{R}^{m+1} with radial log-convex density.

2020 Mathematics Subject Classification. 53C42, 53C21, 35J25.

Key words and phrases. Heintze–Karcher’s inequality, Reilly’s formula, manifolds with density, anti-Gaussian space, Alexandrov’s theorem, self-expander.

This work is supported by NSFC (Grant No. 12271449, 12671069, 12526203, 12526102) and the Natural Science Foundation of Fujian Province of China (Grant No. 2024J011008).

The stationary hypersurfaces for the isoperimetric problems in such spaces are given by constant weighted mean curvature hypersurfaces

Hϕ=H−⟨∇ϕ,ν⟩=λ,H_{\phi}=H-\langle\nabla\phi,\nu\rangle=\lambda,

where HH is the sum of the principal curvatures, ν\nu is the outward unit normal, λ∈R\lambda\in\mathbb{R}. In an important special case ϕ=−∣x∣2/4\phi=-|x|^{2}/4, the constant weighted mean curvature hypersurfaces satisfy

Hϕ=H+12⟨x,ν⟩=λ,H_{\phi}=H+\frac{1}{2}\langle x,\nu\rangle=\lambda,

It is also called the λ\lambda-self-expander equation, compare to the λ\lambda-self-shrinkers as stationary hypersurfaces in the Gaussian space. We [3] and Johne-Silini [12] have independently proved the following Alexandrov-type theorem for λ\lambda-self-expanders.

Theorem 1.1 (Bai-Xia [3], Johne-Silini [12]). Every closed, connected, embedded λ\lambda-self-expander in Rm+1\mathbb{R}^{m+1} is a sphere centered at the origin.

There are non-compact and immersed non-spherical λ\lambda-self-expander, [2, 18], and also several rigidity results of λ\lambda-self-expanders, see for example [2, 4, 9, 15]. The proof of Theorem 1.1 is based on the following Minkowski formula for λ\lambda-self-expanders

∫Σ(m+∣x∣22)e∣x∣2/4 dσ=Hϕ∫Ω(m+1+∣x∣22)e∣x∣2/4 dx.(2)\int_{\Sigma}\left(m+\frac{|x|^{2}}{2}\right)e^{|x|^{2}/4}\,d\sigma=H_{\phi}\int_{\Omega}\left(m+1+\frac{|x|^{2}}{2}\right)e^{|x|^{2}/4}\,dx. \tag*{(2)}

and the following sharp Heintze-Karcher-type inequality.

Theorem 1.2 (Bai-Xia [3], Johne-Silini [12]). Let Ω⊂Rm+1\Omega\subset\mathbb{R}^{m+1} be a bounded domain with boundary Σ=∂Ω\Sigma=\partial\Omega. If H+⟨x,ν⟩/2>0H+\langle x,\nu\rangle/2>0 on Σ\Sigma, then

∫Σm+∣x∣2/2H+⟨x,ν⟩/2e∣x∣2/4 dσ≥∫Ω(m+1+∣x∣22)e∣x∣2/4 dx.(3)\int_{\Sigma}\frac{m+|x|^{2}/2}{H+\langle x,\nu\rangle/2}e^{|x|^{2}/4}\,d\sigma\geq\int_{\Omega}\left(m+1+\frac{|x|^{2}}{2}\right)e^{|x|^{2}/4}\,dx. \tag*{(3)}

Equality holds if and only if Ω\Omega is a ball centered at the origin.

The proof of Theorem 1.2 is based on an adaption of Brendle’s idea by using normal flow with respect to the conformal metric (m+∣x∣2/2)−2g‾(m+|x|^{2}/2)^{-2}\overline{g}. We remark that the approach also works in a large class of warped product manifolds with radial density satisfying a ϕ\phi-substatic assumption, see the details in [3].

In this paper, inspired by Li-Xia’s work [14], we find a Reilly-formula-approach to Theorem 1.2, and more generally, to Heintze-Karcher-type inequality in ϕ\phi-substatic manifolds (without warped product structure for metric and radial structure for density). To state our result, we introduce some notations. Let Δϕ=Δ−⟨∇‾ϕ,∇‾⟩\Delta_{\phi}=\Delta-\langle\overline{\nabla}\phi,\overline{\nabla}\rangle be the Witten-Laplacian and Ric⁡‾ϕ=Ric⁡‾+∇‾2ϕ\overline{\operatorname{Ric}}_{\phi}=\overline{\operatorname{Ric}}+\overline{\nabla}^{2}\phi be the Bakry-Emery-Ricci curvature. Denote

Ric⁡‾ϕV:=ΔϕVVg‾−∇‾2VV+Ric⁡‾ϕ,(4)\overline{\operatorname{Ric}}_{\phi}^{V}:=\frac{\Delta_{\phi}V}{V}\overline{g}-\frac{\overline{\nabla}^{2}V}{V}+\overline{\operatorname{Ric}}_{\phi}, \tag*{(4)}

and we call it ϕ\phi-static-Ricci tensor. Our main result is the following.

Theorem 1.3. Let (Mm+1,g‾,e−ϕdμg‾)(M^{m+1},\overline{g},e^{-\phi}d\mu_{\overline{g}}) be a Riemannian manifold with density, and Ω⊂M\Omega\subset M be a bounded domain with boundary Σ=∂Ω\Sigma=\partial\Omega such that Hϕ>0H_{\phi}>0. Let V∈C∞(Ω‾)V\in C^{\infty}(\overline{\Omega}) and N∈C∞(Ω‾)\mathcal{N}\in C^{\infty}(\overline{\Omega}) be such that V>0V>0, N≥m+1\mathcal{N}\geq m+1 and

(N−(m+1))Ric⁡‾ϕV≥∇‾ϕ⊗∇‾ϕ.(5)(\mathcal{N}-(m+1))\overline{\operatorname{Ric}}_{\phi}^{V}\geq\overline{\nabla}\phi\otimes\overline{\nabla}\phi. \tag*{(5)}

At points where N=m+1\mathcal{N}=m+1, assume in addition that ∇‾ϕ=0\overline{\nabla}\phi=0 and Ric⁡‾ϕV≥0\overline{\operatorname{Ric}}_{\phi}^{V}\geq0. Then it holds that

∫ΣVHϕ dσϕ≥∫ΩNN−1V dμϕ.(6)\int_{\Sigma}\frac{V}{H_{\phi}}\,d\sigma_{\phi}\geq\int_{\Omega}\frac{\mathcal{N}}{\mathcal{N}-1}V\,d\mu_{\phi}. \tag*{(6)}

Remark 1.4. When ϕ≡0\phi\equiv0, we can choose N≡m+1\mathcal{N} \equiv m+1 and Theorem 1.3 reduces to Li-Xia’s [14] Heintze-Karcher inequality for substatic manifolds. For general ϕ\phi and a constant N>m+1\mathcal{N} > m+1, Theorem 1.3 reduces to Huang-Ma-Zhu [11].

The proof of Theorem 1.3 uses Li-Xia’s Reilly-type formula [14], see also Huang-Ma-Zhu [11]. For simplicity, we state here a special case when f∣Σ=0f|_{\Sigma}=0.

Proposition 1.5 (Li-Xia [14], Huang-Ma-Zhu [11]). Let f∈C∞(Ω‾)f \in C^{\infty}(\overline{\Omega}) be such that f∣Σ=0f|_{\Sigma}=0. Then

∫ΣVHϕfν2 dσϕ=∫ΩV[(Δ‾ϕf−fVΔ‾ϕV)2−∣∇‾2f−fV∇‾2V∣2]dμϕ−∫ΩVRic⁡‾ϕV(∇‾f−fV∇‾V,∇‾f−fV∇‾V)dμϕ.(7)\int_{\Sigma} V H_{\phi} f_{\nu}^{2}\,d\sigma_{\phi}=\int_{\Omega} V\left[\left(\overline{\Delta}_{\phi}f-\frac{f}{V}\overline{\Delta}_{\phi}V\right)^{2}-\left|\overline{\nabla}^{2}f-\frac{f}{V}\overline{\nabla}^{2}V\right|^{2}\right]d\mu_{\phi} -\int_{\Omega} V\overline{\operatorname{Ric}}_{\phi}^{V}\left(\overline{\nabla}f-\frac{f}{V}\overline{\nabla}V,\overline{\nabla}f-\frac{f}{V}\overline{\nabla}V\right)d\mu_{\phi}. \tag*{(7)}

Next we consider the rigidity for the Heintze-Karcher inequality (6) in Theorem 1.3. For this purpose, we need a slightly stronger condition (9) than (5). For simplicity, we only consider the case M=Rm+1M=\mathbb{R}^{m+1} and choose

N=m+1−⟨x,∇‾ϕ⟩,V=N−1=m−⟨x,∇‾ϕ⟩.(8)\mathcal{N}=m+1-\langle x,\overline{\nabla}\phi\rangle,\qquad V=\mathcal{N}-1=m-\langle x,\overline{\nabla}\phi\rangle. \tag*{(8)}

To ensure N≥m+1\mathcal{N}\ge m+1, we assume ⟨x,∇‾ϕ⟩≤0\langle x,\overline{\nabla}\phi\rangle\le0.

Theorem 1.6. Let Ω⊂Rm+1\Omega\subset\mathbb{R}^{m+1} be a bounded connected smooth domain with density dμϕ=e−ϕdxd\mu_{\phi}=e^{-\phi}dx and boundary Σ=∂Ω\Sigma=\partial\Omega such that Hϕ>0H_{\phi}>0. Choose N\mathcal{N} and VV as in (8). Assume that ⟨x,∇‾ϕ⟩≤0\langle x,\overline{\nabla}\phi\rangle\le0, and

Ric⁡‾ϕV≥∣∇‾ϕ∣2N−(m+1)g‾,(9)\overline{\operatorname{Ric}}_{\phi}^{V}\ge\frac{|\overline{\nabla}\phi|^{2}}{\mathcal{N}-(m+1)}\overline{g}, \tag*{(9)}

where the quotient on the right hand side admits a smooth, strictly positive extension to Ω‾\overline{\Omega}. Then (6) holds. Moreover, equality in (6) holds if and only if Ω=BR(0)\Omega=B_{R}(0) for some R>0R>0 and ϕ\phi is a radial function about the origin.

Remark 1.7. Suppose that ϕ=ϕ(r)\phi=\phi(r), where r=∣x∣r=|x|. Since

∣∇‾ϕ∣2N−(m+1)=−ϕ′(r)r,\frac{|\overline{\nabla}\phi|^{2}}{\mathcal{N}-(m+1)}=-\frac{\phi'(r)}{r},

condition (9) reduces to

Δ‾ϕV‾ g‾−∇‾2V+VRic⁡‾ϕ+Vϕ′(r)rg‾≥0.\overline{\Delta}_{\phi}\overline{V}\,\overline{g}-\overline{\nabla}^{2}V+V\overline{\operatorname{Ric}}_{\phi}+V\frac{\phi'(r)}{r}\overline{g}\ge0.

This is precisely the weighted substatic condition used in [3] for the Euclidean case λ(r)=r\lambda(r)=r. In this case, together with the smooth positive extension assumption, this condition is stronger than radial log-convexity on a ball centered at the origin (cf. [12, Appendix B]).

We return to the case of the anti-Gaussian space, that is, ϕ=−∣x∣24\phi=-\frac{|x|^{2}}{4}. Then

V=m+∣x∣22,N=m+1+∣x∣22V=m+\frac{|x|^{2}}{2},\qquad\mathcal{N}=m+1+\frac{|x|^{2}}{2}

and

(N−(m+1))Ric⁡‾ϕV=∣x∣24g‾=∣∇‾ϕ∣2g‾≥∇‾ϕ⊗∇‾ϕ.(\mathcal{N}-(m+1))\overline{\operatorname{Ric}}_{\phi}^{V}=\frac{|x|^{2}}{4}\overline{g}=|\overline{\nabla}\phi|^{2}\overline{g}\ge\overline{\nabla}\phi\otimes\overline{\nabla}\phi.

Moreover,

Ric⁡‾ϕV=12g‾=∣∇‾ϕ∣2N−(m+1)g‾,\overline{\operatorname{Ric}}_{\phi}^{V}=\frac{1}{2}\overline{g}=\frac{|\overline{\nabla}\phi|^{2}}{\mathcal{N}-(m+1)}\overline{g},

where the quotient extends smoothly to the origin. Hence, Theorem 1.2 can be viewed as a special case of Theorem 1.3 and Theorem 1.6.

Finally, we mention a related work by de Lima [8], where an Alexandrov-type theorem in the Gaussian half-space (R+m+1,e−∣x∣24dx)(\mathbb{R}^{m+1}_{+},e^{-\frac{|x|^{2}}{4}}dx) has been proved via Qiu-Xia’s Reilly-type formula [19] with ∞\infty-Bakry-Emery-Ricci. Compare to de Lima [8], the Alexandrov-type theorem in the anti-Gaussian space proved here uses Li–Xia’s Reilly-type formula [14] with ϕ\phi-static-Ricci and a special function N\mathcal{N}.

The rest of the paper is organized as follows. In § 2 we give a proof of Proposition 1.5 for convenience of reader and then prove Theorem 1.3. In § 3, we prove Theorem 1.6.

Acknowledgments

Acknowledgments. Li–Xia’s Reilly formula used in this paper was obtained while CX was a postdoctoral fellow at McGill University under the mentorship of Professor Pengfei Guan. CX is deeply grateful to Professor Pengfei Guan for his continued support.

A weighted Reilly formula and a Heintze–Karcher inequality

Throughout this section we use the notation of the introduction. The unbarred operators ∇\nabla, Δ\Delta, and div⁡\operatorname{div} act on Σ\Sigma, and for ambient vector fields XX we write

div⁡‾ϕX=div⁡‾X−⟨∇‾ϕ,X⟩.\overline{\operatorname{div}}_{\phi}X=\overline{\operatorname{div}}X-\langle\overline{\nabla}\phi,X\rangle.

Our convention is II(X,Z)=⟨∇‾Xν,Z⟩\mathrm{II}(X,Z)=\langle\overline{\nabla}_{X}\nu,Z\rangle for X,Z∈TΣX,Z\in T\Sigma, with H=tr⁡ΣIIH=\operatorname{tr}_{\Sigma}\mathrm{II}.

Related Reilly-type integral formulas were established by Qiu–Xia [19] and Li–Xia [14], and extended to manifolds with density by Huang–Ma–Zhu [11]. For completeness, we give a brief proof of Proposition 1.5.

Proof of Proposition 1.5. Set

Y=∇‾f−fV∇‾V=V∇‾(fV).Y=\overline{\nabla}f-\frac{f}{V}\overline{\nabla}V=V\overline{\nabla}\left(\frac{f}{V}\right).

Direct differentiation gives

div⁡‾ϕ(VY)=VΔ‾ϕf−fΔ‾ϕV,∇‾XY=∇‾X∇‾f−fV∇‾X∇‾V−⟨X,Y⟩V∇‾V.(10)\begin{aligned} \overline{\operatorname{div}}_{\phi}(VY)&=V\overline{\Delta}_{\phi}f-f\overline{\Delta}_{\phi}V,\\ \overline{\nabla}_{X}Y&=\overline{\nabla}_{X}\overline{\nabla}f-\frac{f}{V}\overline{\nabla}_{X}\overline{\nabla}V-\frac{\langle X,Y\rangle}{V}\overline{\nabla}V. \tag*{(10)} \end{aligned}

It follows that

div⁡‾ϕ((VΔ‾ϕf−fΔ‾ϕV)Y)=V(Δ‾ϕf−fVΔ‾ϕV)2+V⟨∇‾(Δ‾ϕf−fVΔ‾ϕV),Y⟩.(11)\overline{\operatorname{div}}_{\phi}\left(\left(V\overline{\Delta}_{\phi}f-f\overline{\Delta}_{\phi}V\right)Y\right) =V\left(\overline{\Delta}_{\phi}f-\frac{f}{V}\overline{\Delta}_{\phi}V\right)^{2} +V\left\langle\overline{\nabla}\left(\overline{\Delta}_{\phi}f-\frac{f}{V}\overline{\Delta}_{\phi}V\right),Y\right\rangle. \tag*{(11)}

To calculate the second divergence, use the weighted Bochner commutation formula

div⁡‾ϕ(∇‾2h)(Z)=⟨∇‾Δ‾ϕh,Z⟩+Ric⁡‾ϕ(∇‾h,Z),\overline{\operatorname{div}}_{\phi}(\overline{\nabla}^{2}h)(Z) =\langle\overline{\nabla}\overline{\Delta}_{\phi}h,Z\rangle+\overline{\operatorname{Ric}}_{\phi}(\overline{\nabla}h,Z),

where the divergence of a Hessian is taken in its first index. Applied to ff and VV, this formula and (10) give

div⁡‾ϕ(V∇‾Y∇‾f−f∇‾Y∇‾V)=V⟨∇‾Δ‾ϕf,Y⟩−f⟨∇‾Δ‾ϕV,Y⟩+VRic⁡‾ϕ(Y,Y)+V∣∇‾2f−fV∇‾2V∣2−∇‾2V(Y,Y).\begin{aligned} \overline{\operatorname{div}}_{\phi}\left(V\overline{\nabla}_{Y}\overline{\nabla}f-f\overline{\nabla}_{Y}\overline{\nabla}V\right) ={}&V\langle\overline{\nabla}\overline{\Delta}_{\phi}f,Y\rangle-f\langle\overline{\nabla}\overline{\Delta}_{\phi}V,Y\rangle+V\overline{\operatorname{Ric}}_{\phi}(Y,Y)\\ &+V\left|\overline{\nabla}^{2}f-\frac{f}{V}\overline{\nabla}^{2}V\right|^{2}-\overline{\nabla}^{2}V(Y,Y). \end{aligned}

Since

V⟨∇‾Δ‾ϕf,Y⟩−f⟨∇‾Δ‾ϕV,Y⟩=V⟨∇‾(Δ‾ϕf−fVΔ‾ϕV),Y⟩+(Δ‾ϕV)∣Y∣2,\begin{aligned} V\langle\overline{\nabla}\overline{\Delta}_{\phi}f,Y\rangle-f\langle\overline{\nabla}\overline{\Delta}_{\phi}V,Y\rangle ={}&V\left\langle\overline{\nabla}\left(\overline{\Delta}_{\phi}f-\frac{f}{V}\overline{\Delta}_{\phi}V\right),Y\right\rangle+(\overline{\Delta}_{\phi}V)|Y|^{2}, \end{aligned}

the definition of Ric⁡‾ϕV\overline{\operatorname{Ric}}^{V}_{\phi} reduces this to

div⁡‾ϕ(V∇‾2f−f∇‾2V)=V⟨∇‾(Δ‾ϕf−fVΔ‾ϕV),Y⟩+V∣∇‾2f−fV∇‾2V∣2+VRic⁡‾ϕV(Y,Y).(12)\overline{\operatorname{div}}_{\phi}\left(V\overline{\nabla}^{2}f-f\overline{\nabla}^{2}V\right) =V\left\langle\overline{\nabla}\left(\overline{\Delta}_{\phi}f-\frac{f}{V}\overline{\Delta}_{\phi}V\right),Y\right\rangle+V\left|\overline{\nabla}^{2}f-\frac{f}{V}\overline{\nabla}^{2}V\right|^{2}+V\overline{\operatorname{Ric}}^{V}_{\phi}(Y,Y). \tag*{(12)}

Subtracting (2.3) from (2.2) and integrating gives the interior expression in (1.7). Its boundary term is

⟨(VΔ‾ϕf−fΔ‾ϕV)Y−V∇‾Y∇‾f+f∇‾Y∇‾V,ν⟩.\left\langle\left(V\overline{\Delta}_{\phi}f-f\overline{\Delta}_{\phi}V\right)Y-V\overline{\nabla}_{Y}\overline{\nabla}f+f\overline{\nabla}_{Y}\overline{\nabla}V,\nu\right\rangle.

Because f=0f=0 on Σ\Sigma, we have Y=∇‾f=fννY=\overline{\nabla}f=f_{\nu}\nu. For tangent vectors X,Z∈TΣX,Z\in T\Sigma, differentiating Z(f)=0Z(f)=0 gives

∇‾2f(X,Z)=−fν⟨ν,∇‾XZ⟩=fνII⁡(X,Z).\overline{\nabla}^{2}f(X,Z)=-f_{\nu}\langle\nu,\overline{\nabla}_{X}Z\rangle=f_{\nu}\operatorname{II}(X,Z).

Taking the tangential trace, and using the definition of Δ‾ϕ\overline{\Delta}_{\phi}, yields

Δ‾ϕf−∇‾2f(ν,ν)=(H−⟨∇‾ϕ,ν⟩)fν=Hϕfν.\overline{\Delta}_{\phi}f-\overline{\nabla}^{2}f(\nu,\nu)=\left(H-\langle\overline{\nabla}\phi,\nu\rangle\right)f_{\nu}=H_{\phi}f_{\nu}.

Thus the boundary integrand is VHϕfν2VH_{\phi}f_{\nu}^{2}, and the weighted divergence theorem proves (1.7). □

Proof of Theorem 1.3. Let ff be the solution to the Dirichlet problem:

{Δ‾ϕf−Δ‾ϕVVf=NN−1in Ω,f=0on Σ.(13)\begin{cases} \overline{\Delta}_{\phi}f-\dfrac{\overline{\Delta}_{\phi}V}{V}f=\dfrac{\mathcal{N}}{\mathcal{N}-1} & \text{in }\Omega,\\ f=0 & \text{on }\Sigma. \tag*{(13)} \end{cases}

Writing f=Vuf=Vu, we have

Δ‾ϕ(Vu)−uΔ‾ϕV=1Vdiv⁡‾ϕ(V2∇‾u),\overline{\Delta}_{\phi}(Vu)-u\overline{\Delta}_{\phi}V=\frac{1}{V}\overline{\operatorname{div}}_{\phi}(V^{2}\overline{\nabla}u),

so the equation can be written in divergence form:

div⁡‾ϕ(V2∇‾u)=NN−1V,u∣Σ=0.\overline{\operatorname{div}}_{\phi}(V^{2}\overline{\nabla}u)=\frac{\mathcal{N}}{\mathcal{N}-1}V,\qquad u|_{\Sigma}=0.

By the weighted divergence theorem, we have

∫ΣVfν dσϕ=∫ΩNN−1V dμϕ.(14)\int_{\Sigma}Vf_{\nu}\,d\sigma_{\phi}=\int_{\Omega}\frac{\mathcal{N}}{\mathcal{N}-1}V\,d\mu_{\phi}. \tag*{(14)}

The vector field from Proposition 1.5 is

Y=∇‾f−fV∇‾V.Y=\overline{\nabla}f-\frac{f}{V}\overline{\nabla}V.

The trace estimate for the Hessian difference is

∣∇‾2f−fV∇‾2V∣2≥1m+1(Δ‾f−fVΔ‾V)2.\left|\overline{\nabla}^{2}f-\frac{f}{V}\overline{\nabla}^{2}V\right|^{2}\geq\frac{1}{m+1}\left(\overline{\Delta}f-\frac{f}{V}\overline{\Delta}V\right)^{2}.

At points where N>m+1\mathcal{N}>m+1, the second tensor inequality in (1.5) gives

Ric⁡‾ϕV(Y,Y)≥1N−(m+1)⟨∇‾ϕ,Y⟩2.\overline{\operatorname{Ric}}^{V}_{\phi}(Y,Y)\geq\frac{1}{\mathcal{N}-(m+1)}\langle\overline{\nabla}\phi,Y\rangle^{2}.

Combining the two estimates with the scalar Cauchy–Schwarz inequality yields

∣∇‾2f−fV∇‾2V∣2+Ric⁡‾ϕV(Y,Y)≥1m+1(Δ‾f−fVΔ‾V)2+1N−(m+1)⟨∇‾ϕ,Y⟩2≥1N(Δ‾f−fVΔ‾V−⟨∇‾ϕ,Y⟩)2=1N(Δ‾ϕf−fVΔ‾ϕV)2.(15)\begin{aligned} \left|\overline{\nabla}^{2}f-\frac{f}{V}\overline{\nabla}^{2}V\right|^{2}+\overline{\operatorname{Ric}}_{\phi}^{V}(Y,Y) &\ge\frac{1}{m+1}\left(\overline{\Delta}f-\frac{f}{V}\overline{\Delta}V\right)^{2}+\frac{1}{\mathcal{N}-(m+1)}\langle\overline{\nabla}\phi,Y\rangle^{2}\\ &\ge\frac{1}{\mathcal{N}}\left(\overline{\Delta}f-\frac{f}{V}\overline{\Delta}V-\langle\overline{\nabla}\phi,Y\rangle\right)^{2}\\ &=\frac{1}{\mathcal{N}}\left(\overline{\Delta}_{\phi}f-\frac{f}{V}\overline{\Delta}_{\phi}V\right)^{2}. \tag*{(15)} \end{aligned}

Since ∇‾ϕ=0\overline{\nabla}\phi=0 and Ric⁡‾ϕV≥0\overline{\operatorname{Ric}}_{\phi}^{V}\ge0 at points where N=m+1\mathcal{N}=m+1, we also have (15). Inserting the pointwise bound into the Reilly identity of Proposition 1.5, and using (13), yields

∫ΣVHϕfν2 dσϕ≤∫ΩV(1−1N)(Δ‾ϕf−Δ‾ϕVVf)2 dμϕ=∫ΩNN−1V dμϕ.(16)\int_{\Sigma}VH_{\phi}f_{\nu}^{2}\,d\sigma_{\phi}\le\int_{\Omega}V\left(1-\frac{1}{\mathcal{N}}\right)\left(\overline{\Delta}_{\phi}f-\frac{\overline{\Delta}_{\phi}V}{V}f\right)^{2}\,d\mu_{\phi} =\int_{\Omega}\frac{\mathcal{N}}{\mathcal{N}-1}V\,d\mu_{\phi}. \tag*{(16)}

On Σ\Sigma, (14) and Cauchy–Schwarz give

(∫ΩNN−1V dμϕ)2=(∫ΣVfν dσϕ)2≤(∫ΣVHϕ dσϕ)(∫ΣVHϕfν2 dσϕ).(17)\left(\int_{\Omega}\frac{\mathcal{N}}{\mathcal{N}-1}V\,d\mu_{\phi}\right)^{2}=\left(\int_{\Sigma}Vf_{\nu}\,d\sigma_{\phi}\right)^{2} \le\left(\int_{\Sigma}\frac{V}{H_{\phi}}\,d\sigma_{\phi}\right)\left(\int_{\Sigma}VH_{\phi}f_{\nu}^{2}\,d\sigma_{\phi}\right). \tag*{(17)}

Therefore, (16) and (17) yields (6). □

Rigidity for Heintze–Karcher inequality

Proof of Theorem 1.6. Since N=V+1\mathcal{N}=V+1, the curvature condition (9) and the smooth positive extension of ∣∇‾ϕ∣2/(V−m)\lvert\overline{\nabla}\phi\rvert^{2}/(V-m) ensure that all the assumptions of Theorem 1.3 are satisfied, including at V=mV=m. Thus (6) follows.

Suppose equality holds, and take the solution ff of (13). As before, write

u=fV,Y=∇‾f−u∇‾V=V∇‾u.u=\frac{f}{V},\qquad Y=\overline{\nabla}f-u\overline{\nabla}V=V\overline{\nabla}u.

Since N=V+1\mathcal{N}=V+1, the Dirichlet equation gives

Δ‾ϕf−uΔ‾ϕV=V+1V.\overline{\Delta}_{\phi}f-u\overline{\Delta}_{\phi}V=\frac{V+1}{V}.

By the definition of the weighted Laplacian,

Δ‾f−uΔ‾V=V+1V+⟨∇‾ϕ,Y⟩.(18)\overline{\Delta}f-u\overline{\Delta}V=\frac{V+1}{V}+\langle\overline{\nabla}\phi,Y\rangle. \tag*{(18)}

At points where V>mV>m, the trace inequality and (9) yield

∣∇‾2f−u∇‾2V∣2+Ric‾ϕV(Y,Y)≥(Δ‾f−uΔ‾V)2m+1+∣∇‾ϕ∣2V−m∣Y∣2≥(Δ‾f−uΔ‾V)2m+1+⟨∇‾ϕ,Y⟩2V−m≥1V+1(Δ‾f−uΔ‾V−⟨∇‾ϕ,Y⟩)2=V+1V2.\left\lvert\overline{\nabla}^{2}f-u\overline{\nabla}^{2}V\right\rvert^{2}+\overline{\mathrm{Ric}}_{\phi}^{V}(Y,Y) \ge\frac{\left(\overline{\Delta}f-u\overline{\Delta}V\right)^{2}}{m+1}+\frac{\lvert\overline{\nabla}\phi\rvert^{2}}{V-m}\lvert Y\rvert^{2} \ge\frac{\left(\overline{\Delta}f-u\overline{\Delta}V\right)^{2}}{m+1}+\frac{\langle\overline{\nabla}\phi,Y\rangle^{2}}{V-m} \ge\frac{1}{V+1}\left(\overline{\Delta}f-u\overline{\Delta}V-\langle\overline{\nabla}\phi,Y\rangle\right)^{2} = \frac{V+1}{V^{2}}.

The second and third inequalities follow from the vector and scalar Cauchy–Schwarz inequalities, respectively. As in the proof of Theorem 1.3, equality in (6) forces equality in these pointwise estimates and in the boundary Cauchy–Schwarz inequality. Equality in the scalar Cauchy–Schwarz inequality gives

Δ‾f−uΔ‾Vm+1=−⟨∇‾ϕ,Y⟩V−m.\frac{\overline{\Delta}f-u\overline{\Delta}V}{m+1}=-\frac{\langle\overline{\nabla}\phi,Y\rangle}{V-m}.

Together with (18), this yields

V+1V=Δ‾f−uΔ‾V−⟨∇‾ϕ,Y⟩=(1+V−mm+1)(Δ‾f−uΔ‾V)=V+1m+1(Δ‾f−uΔ‾V).\frac{V+1}{V}=\overline{\Delta}f-u\overline{\Delta}V-\langle\overline{\nabla}\phi,Y\rangle =\left(1+\frac{V-m}{m+1}\right)\left(\overline{\Delta}f-u\overline{\Delta}V\right) =\frac{V+1}{m+1}\left(\overline{\Delta}f-u\overline{\Delta}V\right).

Consequently,

Δ‾f−uΔ‾V=m+1V,⟨∇‾ϕ,Y⟩=−V−mV.(19)\overline{\Delta}f-u\overline{\Delta}V=\frac{m+1}{V},\qquad\langle\overline{\nabla}\phi,Y\rangle=-\frac{V-m}{V}. \tag*{(19)}

Equality in the trace inequality and (19) give

∇‾2f−u∇‾2V=1Vg‾.\overline{\nabla}^{2}f-u\overline{\nabla}^{2}V=\frac{1}{V}\overline{g}.

Equality in the vector Cauchy–Schwarz inequality makes ∇‾ϕ\overline{\nabla}\phi and YY linearly dependent. Using the second identity in (19), we obtain

∇‾ϕ=−∣∇‾ϕ∣2V−mVY.\overline{\nabla}\phi=-\frac{\lvert\overline{\nabla}\phi\rvert^{2}}{V-m}VY.

At points where V=mV=m, the smooth positive extension of the quotient gives ∇‾ϕ=0\overline{\nabla}\phi=0, and (9) implies Ric‾ϕV>0\overline{\mathrm{Ric}}_{\phi}^{V}>0. In this case (18) gives Δ‾f−uΔ‾V=(m+1)/m\overline{\Delta}f-u\overline{\Delta}V=(m+1)/m, so

∣∇‾2f−u∇‾2V∣2+Ric‾ϕV(Y,Y)≥m+1m2+Ric‾ϕV(Y,Y)≥m+1m2.\left\lvert\overline{\nabla}^{2}f-u\overline{\nabla}^{2}V\right\rvert^{2}+\overline{\mathrm{Ric}}_{\phi}^{V}(Y,Y) \ge\frac{m+1}{m^{2}}+\overline{\mathrm{Ric}}_{\phi}^{V}(Y,Y)\ge\frac{m+1}{m^{2}}.

Equality forces Y=0Y=0 and equality in the trace inequality, hence the two interior identities remain valid at these points as well, with the quotient understood as its smooth extension.

Finally, equality in the boundary Cauchy–Schwarz inequality implies that HϕfνH_{\phi}f_{\nu} is constant on Σ\Sigma. By (17) and equality in (6), this constant is 1. We have therefore obtained

∇‾2f−u∇‾2V=1Vg‾,∇‾ϕ=−∣∇‾ϕ∣2V−mVYin Ω,fν=1Hϕon Σ.(20)\begin{aligned} \overline{\nabla}^{2}f-u\overline{\nabla}^{2}V&=\frac{1}{V}\overline{g},\\ \overline{\nabla}\phi&=-\frac{|\overline{\nabla}\phi|^{2}}{V-m}VY &&\text{in }\Omega,\\ f_{\nu}&=\frac{1}{H_{\phi}} &&\text{on }\Sigma. \tag*{(20)} \end{aligned}

We now prove that Ω\Omega is a ball centered at the origin. Differentiating Y=∇‾f−u∇‾VY=\overline{\nabla}f-u\overline{\nabla}V and using the first identity in (20), we get

∇‾XY=1VX−X(u)∇‾V.\overline{\nabla}_{X}Y=\frac{1}{V}X-X(u)\overline{\nabla}V.

for every vector field XX. Since Y=V∇‾uY=V\overline{\nabla}u and ∇‾Xx=X\overline{\nabla}_{X}x=X,

∇‾X(VY−x)=X(V)Y+V∇‾XY−X=⟨∇‾V,X⟩Y−⟨Y,X⟩∇‾V.\begin{aligned} \overline{\nabla}_{X}(VY-x)=X(V)Y+V\overline{\nabla}_{X}Y-X \\ &=\langle\overline{\nabla}V,X\rangle Y-\langle Y,X\rangle\overline{\nabla}V. \end{aligned}

For any vector fields X,ZX,Z, the preceding identity gives

⟨∇‾X(VY−x),Z⟩+⟨∇‾Z(VY−x),X⟩=0.\langle\overline{\nabla}_{X}(VY-x),Z\rangle+\langle\overline{\nabla}_{Z}(VY-x),X\rangle=0.

Thus VY−xVY-x is a Euclidean Killing vector field.

By (13) and N=V+1\mathcal{N}=V+1,

div⁡‾ϕ(V2∇‾u)=V+1>0,u∣Σ=0.\overline{\operatorname{div}}_{\phi}(V^{2}\overline{\nabla}u)=V+1>0,\qquad u|_{\Sigma}=0.

The maximum principle gives u<0u<0 in Ω\Omega. Thus uu attains an interior minimum at some x0∈Ωx_{0}\in\Omega, where Y(x0)=V∇‾u(x0)=0Y(x_{0})=V\overline{\nabla}u(x_{0})=0. The preceding formula then implies ∇‾X(VY−x)(x0)=0\overline{\nabla}_{X}(VY-x)(x_{0})=0 for every XX. Every Euclidean Killing vector field has the form Ax+bAx+b, where AA is a constant skew-symmetric matrix and bb is a constant vector. Since its derivative vanishes at x0x_{0}, we have A=0A=0. Evaluating at x0x_{0}, where Y(x0)=0Y(x_{0})=0, gives b=−x0b=-x_{0}. Consequently,

VY=x−x0in Ω.(21)VY=x-x_{0}\qquad\text{in }\Omega. \tag*{(21)}

Substituting (21) into the second identity in (20), we find

∇‾ϕ=−∣∇‾ϕ∣2V−m(x−x0).\overline{\nabla}\phi=-\frac{|\overline{\nabla}\phi|^{2}}{V-m}(x-x_{0}).

In particular, ∇‾ϕ(x0)=0\overline{\nabla}\phi(x_{0})=0, so (8) gives V(x0)=mV(x_{0})=m. As V≥mV\ge m, the function VV attains its minimum at x0x_{0} and hence ∇‾V(x0)=0\overline{\nabla}V(x_{0})=0. By (8),

V−m=∣∇‾ϕ∣2V−m⟨x,x−x0⟩.V-m=\frac{|\overline{\nabla}\phi|^{2}}{V-m}\langle x,x-x_{0}\rangle.

Since ∇‾⟨x,x−x0⟩=2x−x0\overline{\nabla}\langle x,x-x_{0}\rangle=2x-x_{0}, differentiating gives

∇‾V=⟨x,x−x0⟩∇‾(∣∇‾ϕ∣2V−m)+∣∇‾ϕ∣2V−m(2x−x0).\overline{\nabla}V=\langle x,x-x_{0}\rangle\overline{\nabla}\left(\frac{|\overline{\nabla}\phi|^{2}}{V-m}\right)+\frac{|\overline{\nabla}\phi|^{2}}{V-m}(2x-x_{0}).

At x=x0x=x_{0}, the first term vanishes, so

0=∇‾V(x0)=∣∇‾ϕ∣2V−m∣x0x0.0=\overline{\nabla}V(x_{0})=\left.\frac{|\overline{\nabla}\phi|^{2}}{V-m}\right|_{x_{0}}x_{0}.

The quotient has a strictly positive smooth extension, which forces x0=0x_{0}=0. Hence

VY=xin Ω.(22)VY=x\qquad\text{in }\Omega. \tag*{(22)}

On Σ\Sigma, we have u=0u=0, and the third identity in (20) gives Y=fνν=ν/HϕY=f_\nu\nu=\nu/H_\phi. Hence (22) yields

x=VHϕνon Σ.x=\frac{V}{H_\phi}\nu\quad\text{on }\Sigma.

For every tangent vector X∈TΣX\in T\Sigma,

X(∣x∣2)=2⟨x,X⟩=2VHϕ⟨ν,X⟩=0.X(|x|^2)=2\langle x,X\rangle=2\frac{V}{H_\phi}\langle\nu,X\rangle=0.

Thus ∣x∣|x| is constant on each connected boundary component, which is therefore a sphere centered at the origin. Since V/Hϕ>0V/H_\phi>0, the outward normal is ν=x/∣x∣\nu=x/|x|, excluding inner boundary components. As Ω\Omega is bounded and connected, we conclude that Ω=BR(0)\Omega=B_R(0) for some R>0R>0.

Furthermore, (22) and the second identity in (20) give

∇‾ϕ=−∣∇‾ϕ∣2V−mx.\overline{\nabla}\phi=-\frac{|\overline{\nabla}\phi|^2}{V-m}x.

Thus ϕ\phi is constant on every sphere centered at the origin, and therefore radial throughout Ω\Omega. By (8), VV is radial as well.

Conversely, suppose Ω=BR(0)\Omega=B_R(0). Since x=Rνx=R\nu on Σ\Sigma, (8) gives V=RHϕV=RH_\phi there. Moreover,

div⁡‾ϕx=m+1−⟨x,∇‾ϕ⟩=V+1=N,NN−1V=N.\overline{\operatorname{div}}_\phi x=m+1-\langle x,\overline{\nabla}\phi\rangle=V+1=\mathcal{N},\qquad\frac{\mathcal{N}}{\mathcal{N}-1}V=\mathcal{N}.

The weighted divergence theorem yields

∫ΣVHϕ dσϕ=R∫Σdσϕ=∫Σ⟨x,ν⟩ dσϕ=∫ΩN dμϕ=∫ΩNN−1V dμϕ.\begin{aligned} \int_\Sigma\frac{V}{H_\phi}\,d\sigma_\phi =R\int_\Sigma d\sigma_\phi =\int_\Sigma\langle x,\nu\rangle\,d\sigma_\phi\\ &=\int_\Omega\mathcal{N}\,d\mu_\phi =\int_\Omega\frac{\mathcal{N}}{\mathcal{N}-1}V\,d\mu_\phi. \end{aligned}

Thus equality holds in (6).

(J.B.) School of Mathematical Sciences, Xiamen University, 361005, Xiamen, P. R. China Email address: [email protected]

(C.X.) School of Mathematical Sciences, Xiamen University, 361005, Xiamen, P. R. China Email address: [email protected]

References

  1. [1]A. D. Aleksandrov, Uniqueness theorems for surfaces in the large. I, Vestnik Leningrad. Univ. 11 (1956), no. 19, 5–17.
  2. [2]Saul Ancari and Xu Cheng, Some rigidity properties for λ-self-expanders, Nonlinear Analysis 230 (2023), 113230.
  3. [3]Jinchuan Bai and Chao Xia, An Alexandrov-type theorem in warped product manifolds with radial density, 2026, arXiv:2608.08548.arxiv.org/abs/2608.08548
  4. [4]Jinchuan Bai and Chao Xia, Stable constant weighted mean curvature hypersurfaces in anti-gaussian space, 2026, arXiv:2608.22225.arxiv.org/abs/2608.22225
  5. [5]Stefano Borghini, Mattia Fognano, and Andrea Pinamonti, The equality case in the substatic heintze–karcher inequality, Archive for Rational Mechanics and Analysis 248 (2024), no. 6, 108.arxiv.org/abs/2307.04253
  6. [6]Simon Brendle, Constant mean curvature surfaces in warped product manifolds, Publ. Math. Inst. Hautes Études Sci. 117 (2013), 247–269.arxiv.org/abs/1105.4273
  7. [7]Gregory R. Chambers, Proof of the log-convex density conjecture, Journal of the European Mathematical Society 21 (2019), no. 8, 2301–2332.arxiv.org/abs/1311.4012
  8. [8]Levi Lopes de Lima, Critical domains for certain dirichlet integrals in weighted manifolds, 2024, arXiv:2409.03554v2.arxiv.org/abs/2409.03554
  9. [9]Liling Deng, Yunwen Yao, Yang Yu, and Hui Zhang, Some classification of 2-dimensional λ-self-expanders, Bulletin of the Malaysian Mathematical Sciences Society 47 (2024), 177.
  10. [10]Ernst Heintze and Hermann Karcher, A general comparison theorem with applications to volume estimates for submanifolds, Annales scientifiques de l’École Normale Supérieure 11 (1978), no. 4, 451–470.DOI
  11. [11]Guangyue Huang, Bingqing Ma, and Mingfang Zhu, A Reilly type integral formula and its applications, Differential Geom. Appl. 94 (2024), 102136.
  12. [12]Florian Johne and Lauro Silini, Alexandrov theorem for constant weighted mean curvature surfaces, 2026, arXiv:2608.05119.arxiv.org/abs/2608.05119
  13. [13]Nicholas J. Korevaar, Sphere theorems via alexandrov for constant weingarten curvature hypersurfaces: Appendix to a note of a. ros, Journal of Differential Geometry 27 (1988), no. 2, 221–223.
  14. [14]Junfang Li and Chao Xia, An integral formula and its applications on sub-static manifolds, J. Differential Geom. 113 (2019), no. 3, 493–518.arxiv.org/abs/1603.02201
  15. [15]Tongzhu Li and Mengdan Qi, Rigidity of closed λ-self-expanders, 2026, arXiv:2608.28307.
  16. [16]Sebastián Montiel and Antonio Ros, Compact hypersurfaces: The alexandrov theorem for higher order mean curvatures, Differential Geometry, Pitman Monographs and Surveys in Pure and Applied Mathematics, vol. 52, Longman Scientific & Technical, 1991, pp. 279–296.
  17. [17]Frank Morgan, Manifolds with density, Notices of the American Mathematical Society 52 (2005), no. 8, 853–858.
  18. [18]Mengdan Qi, Examples of topological sphere λ-self-expander, 2026, arXiv:2610.10634.arxiv.org/abs/2610.10634
  19. [19]Guohuan Qiu and Chao Xia, A generalization of Reilly’s formula and its applications to a new Heintze–Karcher type inequality, Int. Math. Res. Not. IMRN 2015 (2015), no. 17, 7608–7619.arxiv.org/abs/1405.4518
  20. [20]Robert C. Reilly, Applications of the Hessian operator in a Riemannian manifold, Indiana Univ. Math. J. 26 (1977), no. 3, 459–472.
  21. [21]Antonio Ros, Compact hypersurfaces with constant higher order mean curvatures, Revista Matemática Iberoamericana 3 (1987), no. 3–4, 447–453.
  22. [22]César Rosales, Antonio Cañete, Vincent Bayle, and Frank Morgan, On the isoperimetric problem in euclidean space with density, Calculus of Variations and Partial Differential Equations 31 (2008), no. 1, 27–46.DOI

Paper details

vlnc:W_er44qun2
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

Not recorded for this paper

View statement
Looking for related papers…
Contents