This paper was converted on www.awesomepapers.org from LaTeX by an anonymous user.
Want to know more? Visit the Converter page.

Asymptotic behavior of the first Dirichlet eigenvalue of AHE manifolds

Xiaoshang Jin
Abstract

In this article, we investigate the rate at which the first Dirichlet eigenvalue of geodesic balls decreases as the radius approaches infinity. We prove that if the conformal infinity of an asymptotically hyperbolic Einstein manifold is of nonnegative Yamabe type, then the two-term asymptotic of the eigenvalues is the same as that in hyperbolic space.

1 Introduction

Suppose that n+1\mathbb{H}^{n+1} is an n+1n+1- dimensional hyperbolic space, then it is well-known that the spectrum of the Laplacian σ(Δ)=[n24,+).\sigma(-\Delta_{\mathbb{H}})=[\frac{n^{2}}{4},+\infty). Later the result is extended by R. Mazzeo in [9] and [10]. He showed that if (X,g)(X,g) is an n+1n+1- dimensional asymptotically hyperbolic manifold, then the spectrum of the Laplacian

σ(Δ)=σp(Δ)[n24,+).\sigma(-\Delta)=\sigma_{p}(-\Delta)\cup[\frac{n^{2}}{4},+\infty).

Here Δ=Δg=1Gxi(gijGxj)\Delta=\Delta_{g}=\frac{1}{\sqrt{G}}\frac{\partial}{\partial x^{i}}(g^{ij}\sqrt{G}\frac{\partial}{\partial x^{j}}) stands for the Laplace-Beltrami operator and σp(Δ)(0,n24)\sigma_{p}(-\Delta)\subseteq(0,\frac{n^{2}}{4}) is a finite set of point spectrums (L2L^{2} eigenvalues). Lee [6] discovered a connection between its spectrum and conformal infinity when gg is also Einstein. In other words, σp(Δ)\sigma_{p}(-\Delta) is empty when (X,g)(X,g) is an asymptotically hyperbolic Einstein (AHE) manifold with conformal infinity of nonnegative Yamabe type. One can also see [14] for another proof.

We rewrite Lee’s result: Y(X,[g^])0λ1(X)=n24.Y(\partial X,[\hat{g}])\geq 0\Rightarrow\lambda_{1}(X)=\frac{n^{2}}{4}. It is clear from the property of the first Dirichlet eigenvalue on the noncompact manifold that

limR+λ1(B(p,R))=n24\lim\limits_{R\rightarrow+\infty}\lambda_{1}(B(p,R))=\frac{n^{2}}{4}

holds for any pX.p\in X. Here B(p,R)B(p,R) is the geodesic ball centered at pp of radius R.R. In this paper, we will present the rate of how λ1(B(p,R))\lambda_{1}(B(p,R)) tends to n24.\frac{n^{2}}{4}.

Before stating our primary theorem, let’s first present some basic conceptions. Suppose that X¯\overline{X} is a compact manifold with smooth boundary X\partial X and gg is a complete metric in its interior X.X. We say that (X,g)(X,g) is conformally compact if there exists a defining function ρ\rho such that g¯=ρ2g\bar{g}=\rho^{2}g extends continuously to X¯.\overline{X}. Here

ρ>0inX,ρ=0onX,dρ0onX.\rho>0\ in\ X,\ \ \ \rho=0\ on\ \partial X,\ \ \ d\rho\neq 0\ on\ \partial X.

(X,g)(X,g) is called Cm,αC^{m,\alpha} (smoothly) conformally compact if g¯=ρ2g\bar{g}=\rho^{2}g is Cm,αC^{m,\alpha} (smooth) on X¯.\overline{X}. For any defining function ρ,\rho, we call g^=ρ2g|TX\hat{g}=\rho^{2}g|_{T\partial X} the boundary metric. Hence the conformal class (X,[g^])(\partial X,[\hat{g}]) is uniquely determined by gg and we call it the conformal infinity of g.g.

Let g¯=ρ2g\bar{g}=\rho^{2}g be a C2C^{2} conformal compactification of (X,g),(X,g), then a simple calculation, such as that in [9] indicates that the sectional curvature of gg tends to |dρ|g¯2|X-|d\rho|^{2}_{\bar{g}}|_{\partial X} as ρ0.\rho\rightarrow 0. Therefore no matter what the topology and geometry of gg look like in X,X, the boundary behavior would always remind us of hyperbolic space. We call (X,g)(X,g) an asymptotically hyperbolic (AH for short) manifold if it is conformally compact and |dρ|g¯2=1|d\rho|^{2}_{\bar{g}}=1 on the boundary X.\partial X.

Let (X,g)(X,g) be a C2C^{2} conformally compact manifold of dimension n+1,n+1, if gg is also Einstein, i.e. Ric[g]=ng,Ric[g]=-ng, then |dρ|ρ2g2=1|d\rho|^{2}_{\rho^{2}g}=1 on X\partial X for any smooth defining function ρ.\rho. In this case, we say that (X,g)(X,g) is an asymptotically hyperbolic Einstein (AHE for short) manifold.

Here is the main result of this paper:

Theorem 1.1.

Let (X,g)(X,g) be an n+1n+1- dimensional AHE manifold with conformal infinity (X,[g^]).(\partial X,[\hat{g}]). If the Yamabe constant Y(X,[g^])0,Y(\partial X,[\hat{g}])\geq 0, then for any pX,p\in X,

λ1(B(p,R))=n24+π2R2+O(R3),R+.\lambda_{1}(B(p,R))=\frac{n^{2}}{4}+\frac{\pi^{2}}{R^{2}}+O(R^{-3}),\ \ R\rightarrow+\infty. (1.1)

Here λ1\lambda_{1} denotes the first Dirichlet eigenvalue.

Theorem 1.1 makes clear that the rate at which the first Dirichlet eigenvalue of geodesic balls tends to n24\frac{n^{2}}{4} is the same as that in hyperbolic space, at least in the second term of the expansion.

On the other hand, we believe that the rate at which the first Dirichlet eigenvalue decreases is related to the geometry structure of manifolds. It is connected to the number of ends. Let’s recall the work of Witten-Yau [15]. They showed that the boundary X\partial X of an AHE manifold (X,g)(X,g) is connected if Y(X,[g^])>0.Y(\partial X,[\hat{g}])>0. Later the work was extended by Cai and Galloway in [2] where they relaxed the assumption that X\partial X has nonnegative Yamabe constant. In [13], Wang proved that if the first eigenvalue of an AHE manifold λ1(X)n1,\lambda_{1}(X)\geq n-1, then it either has only one end or it must be a warped product (×N,dt2+cosh2tgN).(\mathbb{R}\times N,dt^{2}+\cosh^{2}tg_{N}). It would provide a new proof for Cai-Galloway’s result if combined with Lee’s work in [6]. Let’s summarize their work: for an AHE manifold (X,g),(X,g),

Y(X,[g^])0λ1(X)=n24Xisconnected(Xhasoneend).Y(\partial X,[\hat{g}])\geq 0\Longrightarrow\lambda_{1}(X)=\frac{n^{2}}{4}\Longrightarrow\partial X\ is\ connected\ (X\ has\ one\ end).

Later, Li and Wang expanded the results in [7] and [8] where they didn’t require XX to be conformally compact. In this case, XX either has one end or is a warped product. Now we could rule out the case of warped product by a direct calculation.

In fact, as an application of theorem 0.5 and 0.6 in [8], we could obtain the following property:

Proposition 1.2.

Let (X,g)(X,g) be a complete n+1n+1-dimensional manifold with n2n\geq 2 and Ric[g]ng.Ric[g]\geq-ng. If

λ1(B(p,R))=n24+π2R2+O(R3),R+\lambda_{1}(B(p,R))=\frac{n^{2}}{4}+\frac{\pi^{2}}{R^{2}}+O(R^{-3}),\ \ R\rightarrow+\infty (1.2)

for some pX,p\in X, then XX has only one end with infinite volume.

This paper is organized as follows. In section 2, we first provide some background information on the Dirichlet eigenvalue. Then in sections 3 and 4, we prove theorem 1.1. In order to get the upper bound of the first Dirichlet eigenvalue of geodesic balls, we use the eigenvalue comparison theory and the eigenvalue formula in hyperbolic space. To estimate the lower bound, we somewhat enhance Lee’s work. To be more precise, we create a new test function un2sin(alnεu)u^{-\frac{n}{2}}\cdot\sin(a\ln\varepsilon u) on a bounded domain. Here uu is the eigenfunction solution to Δu=(n+1)u\Delta u=(n+1)u and was first used by Lee in [6]. In the end, we prove proposition 1.2 in section 5.

2 The first Dirichlet eigenvalue of manifolds

Let’s introduce some materials about Dirichlet eigenvalue in this section. Suppose that (M,g)(M,g) is a complete manifold and ΩM\Omega\subseteq M is a bounded domain of MM with piecewise smooth boundary. The Dirichlet eigenfunctions are defined by solving the following problem for u0u\neq 0 and eigenvalue λ.\lambda.

{Δu=λuinΩ,u=0onΩ\left\{\begin{array}[]{l}\Delta u=-\lambda u\ \ \ \ in\ \Omega,\\ u=0\ \ \ \ \ \ \ \ \ \ \ \ \ on\ \partial\Omega\end{array}\right. (2.1)

where Δ=1Gxi(gijGxj).\Delta=\frac{1}{\sqrt{G}}\frac{\partial}{\partial x^{i}}(g^{ij}\sqrt{G}\frac{\partial}{\partial x^{j}}). The smallest eigenvalue is denoted by λ1=λ1(Ω)>0.\lambda_{1}=\lambda_{1}(\Omega)>0. Recall the Sobolev space H1(Ω)=W1,2(Ω)H^{1}(\Omega)=W^{1,2}(\Omega) and H01(Ω)H1(Ω)H^{1}_{0}(\Omega)\subseteq H^{1}(\Omega) is defined to be the closure of the infinitely differentiable functions compactly supported in Ω.\Omega. Then by the max-min principle,

λ1(Ω)=inffH01(Ω){0}Ω|f|2𝑑VgΩf2𝑑Vg\lambda_{1}(\Omega)=\inf\limits_{f\in H^{1}_{0}(\Omega)\setminus\{0\}}\frac{\int_{\Omega}|\nabla f|^{2}dV_{g}}{\int_{\Omega}f^{2}dV_{g}} (2.2)

It’s easy to see that the eigenvalue has domain monotonicity: if Ω1Ω2M,\Omega_{1}\subseteq\Omega_{2}\Subset M, then λ1(Ω1)λ1(Ω2).\lambda_{1}(\Omega_{1})\geq\lambda_{1}(\Omega_{2}).

Now we suppose that (M,g)(M,g) is a noncompact manifold, and denote the greatest lower bound for the L2L^{2}-spectrum of the Laplacian by

λ1(M):=infspec(Δ)=inffH01(M){0}M|f|2𝑑VgMf2𝑑Vg.\lambda_{1}(M):=\inf spec(-\Delta)=\inf\limits_{f\in H^{1}_{0}(M)\setminus\{0\}}\frac{\int_{M}|\nabla f|^{2}dV_{g}}{\int_{M}f^{2}dV_{g}}. (2.3)

Notice that λ1(M)\lambda_{1}(M) does not necessarily be an L2L^{2} eigenvalue of Δ,-\Delta, but is motivated by the characterization by

λ1(M)=limkλ1(Ωk)\lambda_{1}(M)=\lim\limits_{k\rightarrow\infty}\lambda_{1}(\Omega_{k}) (2.4)

for any smoothly compact exhaustion {Ωk}\{\Omega_{k}\} of M.M.

For example, for the hyperbolic space M=n+1,M=\mathbb{H}^{n+1}, we know that spec(Δ)=[n24,+),spec(-\Delta)=[\frac{n^{2}}{4},+\infty), see [11]. Then for any pn+1,p\in\mathbb{H}^{n+1},

limR+λ1(B(p,R))=n24.\lim\limits_{R\rightarrow+\infty}\lambda_{1}(B(p,R))=\frac{n^{2}}{4}. (2.5)

It is an interesting problem what the formula of λ1(B(p,R)\lambda_{1}(B(p,R) looks like. Or how λ1(B(p,R))\lambda_{1}(B(p,R)) tends to n24?\frac{n^{2}}{4}? It is shown in [12] and [1] that

λ1(B(p,R))=n24+π2R2+O(R3),R+.\lambda_{1}(B(p,R))=\frac{n^{2}}{4}+\frac{\pi^{2}}{R^{2}}+O(R^{-3}),\ \ R\rightarrow+\infty. (2.6)

In this paper, we prove that (2.6) still holds for AHE manifolds with conformal infinity of nonnegative Yamabe type.

3 The upper bound of eigenvalues

Let’s recall the classic eigenvalue comparison theorem of Cheng [3]. If (X,g)(X,g) is an n+1n+1 dimensional complete manifold satisfying that Ric[g]ng,Ric[g]\geq-ng, then for any pXp\in X and R>0,R>0, λ1(B(p,R))λ1(B(R)).\lambda_{1}(B(p,R))\leq\lambda_{1}(B^{\mathbb{H}}(R)). Here B(R)B^{\mathbb{H}}(R) is a geodesic ball of radius RR in hyperbolic space. He also showed that λ1(B(R))n24+CR2\lambda_{1}(B^{\mathbb{H}}(R))\leq\frac{n^{2}}{4}+\frac{C}{R^{2}} for some positive constant C.C. Later the upper bound estimate was extended by Gage, see theorem 5.2 in [4]. In the following, we provide a weak version of the estimate for the upper bound and the proof is also simpler.

Theorem 3.1.

Let n+1\mathbb{H}^{n+1} be the hyperbolic space of n+1n+1 dimension, then for any pn+1,p\in\mathbb{H}^{n+1},

λ1(B(p,R))n24+π2R2+O(R3)R+.\lambda_{1}(B(p,R))\leq\frac{n^{2}}{4}+\frac{\pi^{2}}{R^{2}}+O(R^{-3})\ \ \ R\rightarrow+\infty. (3.1)
Proof.

Consider the rotationally symmetric model:

(n+1,g=dt2+sinh2tg𝕊)(\mathbb{R}^{n+1},g_{\mathbb{H}}=dt^{2}+\sinh^{2}tg_{\mathbb{S}})

and let pp be the center point. For any R>0,R>0, we define the function

f=en2tsin(πRt)H01((B(p,R))f=e^{-\frac{n}{2}t}\cdot\sin(\frac{\pi}{R}t)\in H^{1}_{0}((B(p,R)) (3.2)

Then

λ1(B(p,R))\displaystyle\lambda_{1}(B(p,R)) B(p,R)|f|2𝑑V[g]B(p,R)f2𝑑V[g]\displaystyle\leq\frac{\int_{B(p,R)}|\nabla f|^{2}dV[g^{\mathbb{H}}]}{\int_{B(p,R)}f^{2}dV[g^{\mathbb{H}}]} (3.3)
=0Rent(n2sin(πRt)+πRcos(πRt))2ωnsinhntdt0Rentsin2(πRt)ωnsinhntdt\displaystyle=\frac{\int_{0}^{R}e^{-nt}(-\frac{n}{2}\sin(\frac{\pi}{R}t)+\frac{\pi}{R}\cos(\frac{\pi}{R}t))^{2}\cdot\omega_{n}\sinh^{n}tdt}{\int_{0}^{R}e^{-nt}\sin^{2}(\frac{\pi}{R}t)\cdot\omega_{n}\sinh^{n}tdt}
=0R(1e2t)n(n2sin(πRt)+πRcos(πRt))2𝑑t0R(1e2t)nsin2(πRt)𝑑t\displaystyle=\frac{\int_{0}^{R}(1-e^{-2t})^{n}\cdot(-\frac{n}{2}\sin(\frac{\pi}{R}t)+\frac{\pi}{R}\cos(\frac{\pi}{R}t))^{2}dt}{\int_{0}^{R}(1-e^{-2t})^{n}\cdot\sin^{2}(\frac{\pi}{R}t)dt}
=0π(1e2Rθπ)n(n2sinθ+πRcosθ)2𝑑θ0π(1e2Rθπ)nsin2θdθ\displaystyle=\frac{\int_{0}^{\pi}(1-e^{-\frac{2R\theta}{\pi}})^{n}\cdot(-\frac{n}{2}\sin\theta+\frac{\pi}{R}\cos\theta)^{2}d\theta}{\int_{0}^{\pi}(1-e^{-\frac{2R\theta}{\pi}})^{n}\cdot\sin^{2}\theta d\theta}
=F(R)G(R)\displaystyle=\frac{F(R)}{G(R)}

where

F(R)0π(n2sinθ+πRcosθ)2𝑑θ=π2(n24+π2R2)F(R)\leq\int_{0}^{\pi}(-\frac{n}{2}\sin\theta+\frac{\pi}{R}\cos\theta)^{2}d\theta=\frac{\pi}{2}\cdot(\frac{n^{2}}{4}+\frac{\pi^{2}}{R^{2}})

For the term G(R),G(R), a direct calculation indicates that

0πerθsin2θdθ=2r(r2+4)(1eπr)=2r3+O(r4),r+\int_{0}^{\pi}e^{-r\theta}\cdot\sin^{2}\theta d\theta=\frac{2}{r(r^{2}+4)}(1-e^{-\pi r})=\frac{2}{r^{3}}+O(r^{-4}),\ r\rightarrow+\infty (3.4)

Hence we could get that

G(R)\displaystyle G(R) =0π[1+k=1nCnk(e2Rθπ)k]sin2θdθ\displaystyle=\int_{0}^{\pi}[1+\sum\limits_{k=1}^{n}C_{n}^{k}(-e^{-\frac{2R\theta}{\pi}})^{k}]\sin^{2}\theta d\theta (3.5)
=π2π34[k=1nCnk(1)k+1k3]1R3+O(R4)\displaystyle=\frac{\pi}{2}-\frac{\pi^{3}}{4}[\sum\limits_{k=1}^{n}C_{n}^{k}\frac{(-1)^{k+1}}{k^{3}}]\frac{1}{R^{3}}+O(R^{-4})

In the end, we deduce that

λ1(B(p,R))F(R)G(R)n24+π2R2+n2π28[k=1nCnk(1)k+1k3]1R3+O(R4).\lambda_{1}(B(p,R))\leq\frac{F(R)}{G(R)}\leq\frac{n^{2}}{4}+\frac{\pi^{2}}{R^{2}}+\frac{n^{2}\pi^{2}}{8}[\sum\limits_{k=1}^{n}C_{n}^{k}\frac{(-1)^{k+1}}{k^{3}}]\frac{1}{R^{3}}+O(R^{-4}). (3.6)

4 The lower bound of eigenvalues

Suppose that (X,g)(X,g) satisfies the conditions in theorem 1.1. Lee proved that λ1(X)=n24\lambda_{1}(X)=\frac{n^{2}}{4} in [6]. The key step is to construct a proper test function φ=un2.\varphi=u^{-\frac{n}{2}}. Here uu is an important positive eigenfunction with prescribed growth at infinity. In order to make our proof more clear, we would give a short quick introduction to Lee’s proof.

4.1 A quick review of Lee’s work

Lemma 4.1.

[6] Let (X,g)(X,g) be an n+1n+1- dimensional AHE manifold with boundary metric g^\hat{g} and let xx be the associated geodesic defining function. Then there is a unique positive eigenfunction uu on XX such that

Δu=(n+1)u.\Delta u=(n+1)u.

and uu has the following form of expansion at the boundary

u=1x+R^4n(n1)x+O(x2).u=\frac{1}{x}+\frac{\hat{R}}{4n(n-1)}x+O(x^{2}).

Let φ=un2,\varphi=u^{-\frac{n}{2}}, then

Δφφ=n24+n(n+2)4(1|du|g2u2).-\frac{\Delta\varphi}{\varphi}=\frac{n^{2}}{4}+\frac{n(n+2)}{4}(1-\frac{|du|^{2}_{g}}{u^{2}}).

One can estimate near the boundary:

u2|du|g2=R^n(n1)+o(1).u^{2}-|du|^{2}_{g}=\frac{\hat{R}}{n(n-1)}+o(1).

On the other hand, the Bochner formula implies that

Δ(u2|du|g2)=2|Δun+1g+2u|20.-\Delta(u^{2}-|du|^{2}_{g})=2|\frac{\Delta u}{n+1}g+\nabla^{2}u|^{2}\geq 0.

When Y(X,[g^])0Y(\partial X,[\hat{g}])\geq 0 we can choose a representative g^\hat{g} whose scalar curvature R^0,\hat{R}\geq 0, then the maximum principle implies that u2|du|g20u^{2}-|du|^{2}_{g}\geq 0 in X.X. So Δφφn24-\frac{\Delta\varphi}{\varphi}\geq\frac{n^{2}}{4} in X.X. According to the eigenvalue comparison theorem of Cheng-Yau [3], λ1(X)n24.\lambda_{1}(X)\geq\frac{n^{2}}{4}.

Now we turn to research the first Dirichlet eigenvalue of geodesic balls. For sufficiently small ε>0,\varepsilon>0, let Xε=X×(0,ε).X_{\varepsilon}=\partial X\times(0,\varepsilon). We study the first Dirichlet eigenvalue of XXε.X\setminus X_{\varepsilon}. If R^>0,\hat{R}>0, we get that

1|du|g2u2R^n(n1)1u2+o(1u2)=R^n(n1)x2+o(x2).1-\frac{|du|^{2}_{g}}{u^{2}}\geq\frac{\hat{R}}{n(n-1)}\frac{1}{u^{2}}+o(\frac{1}{u^{2}})=\frac{\hat{R}}{n(n-1)}x^{2}+o(x^{2}).

Then

Δφφn24+cε2,onXXε-\frac{\Delta\varphi}{\varphi}\geq\frac{n^{2}}{4}+c\varepsilon^{2},\ \ \ on\ X\setminus X_{\varepsilon}

for some positive constant cc and hence λ1(XXε)n24+cε2.\lambda_{1}(X\setminus X_{\varepsilon})\geq\frac{n^{2}}{4}+c\varepsilon^{2}. As a consequence,

λ1(B(p,R))n24+Ce2R\lambda_{1}(B(p,R))\geq\frac{n^{2}}{4}+\frac{C}{e^{2R}} (4.1)

for some C>0C>0 provided RR is large enough. If R^=0,\hat{R}=0, we know that 1|du|g2u21-\frac{|du|^{2}_{g}}{u^{2}} is still positive in X,X, see [5]. Then a similar estimate of (4.1) could be obtained. The lower bound Ce2R\frac{C}{e^{2R}} is too ”small” compared to π2R2.\frac{\pi^{2}}{R^{2}}. We need to find a better test function to get a sharper lower bound of λ1(B(p,R)).\lambda_{1}(B(p,R)).

4.2 A new test function

Let uu be the eigenfunction that is defined in lemma 4.1 and φ=un2.\varphi=u^{-\frac{n}{2}}. In the following, for sufficiently small ε>0,\varepsilon>0, we consider a new test function

ψ=φh=un2sin(alnεu)\psi=\varphi\cdot h=u^{-\frac{n}{2}}\cdot\sin(a\ln\varepsilon u) (4.2)

on the bounded domain

Fε={pX:u(p)<1ε}F_{\varepsilon}=\{p\in X:u(p)<\frac{1}{\varepsilon}\} (4.3)

where a=a(n,ε)<0a=a(n,\varepsilon)<0 is a constant to be determined. A simple calculation indicates that

h=acos(alnεu)u,h′′=a2sin(alnεu)acos(alnεu)u2=a2hu2hu.h^{\prime}=\frac{a\cos(a\ln\varepsilon u)}{u},\ \ h^{\prime\prime}=\frac{-a^{2}\sin(a\ln\varepsilon u)-a\cos(a\ln\varepsilon u)}{u^{2}}=-a^{2}\frac{h}{u^{2}}-\frac{h^{\prime}}{u}. (4.4)

Hence

Δh=hΔu+h′′|du|2=(n+1)uh(a2h+uh)|du|2u2\Delta h=h^{\prime}\Delta u+h^{\prime\prime}|du|^{2}=(n+1)uh^{\prime}-(a^{2}h+uh^{\prime})\frac{|du|^{2}}{u^{2}} (4.5)

and

2g(dlnφ,dlnh)=2g(n2duu,hhdu)=nuhh|du|2u2.2g(d\ln\varphi,d\ln h)=2g(-\frac{n}{2}\frac{du}{u},\frac{h^{\prime}}{h}du)=-n\frac{uh^{\prime}}{h}\frac{|du|^{2}}{u^{2}}. (4.6)

As a consequence,

Δψψ\displaystyle-\frac{\Delta\psi}{\psi} =Δφφ(Δhh+2g(dlnφ,dlnh)\displaystyle=-\frac{\Delta\varphi}{\varphi}-(\frac{\Delta h}{h}+2g(d\ln\varphi,d\ln h) (4.7)
=n24+n(n+2)4(1|du|2u2)[(n+1)uhh(a2+uhh)|du|2u2nuhh|du|2u2]\displaystyle=\frac{n^{2}}{4}+\frac{n(n+2)}{4}(1-\frac{|du|^{2}}{u^{2}})-[(n+1)\frac{uh^{\prime}}{h}-(a^{2}+\frac{uh^{\prime}}{h})\frac{|du|^{2}}{u^{2}}-n\frac{uh^{\prime}}{h}\frac{|du|^{2}}{u^{2}}]
=n24+a2+(1|du|2u2)[n(n+2)4(n+1)uhha2]\displaystyle=\frac{n^{2}}{4}+a^{2}+(1-\frac{|du|^{2}}{u^{2}})[\frac{n(n+2)}{4}-(n+1)\frac{uh^{\prime}}{h}-a^{2}]
=n24+a2+(1|du|2u2)[n(n+2)4(n+1)acot(alnεu)a2]\displaystyle=\frac{n^{2}}{4}+a^{2}+(1-\frac{|du|^{2}}{u^{2}})[\frac{n(n+2)}{4}-(n+1)a\cdot\cot(a\ln\varepsilon u)-a^{2}]

We could assume that u1u\geq 1 on X,X, or else we use kuku instead where kk is a constant large enough. Now set

a=πlnε+cnln2εa=\frac{\pi}{\ln\varepsilon}+\frac{c_{n}}{\ln^{2}\varepsilon} (4.8)

for some constant cn>0,c_{n}>0, then a<0.a<0. Hence on Fε,F_{\varepsilon}, we have that

alnεu(0,π+cnlnε](0,π).a\ln\varepsilon u\in(0,\pi+\frac{c_{n}}{\ln\varepsilon}]\subseteq(0,\pi).

As a result, hh is smooth and positive on FεF_{\varepsilon} and so is ψ.\psi. Furthermore,

acot(alnεu)\displaystyle-a\cdot\cot(a\ln\varepsilon u) acot(π+cnlnε)=acos(π+cnlnε)sin(π+cnlnε)\displaystyle\geq-a\cdot\cot(\pi+\frac{c_{n}}{\ln\varepsilon})=-a\frac{\cos(\pi+\frac{c_{n}}{\ln\varepsilon})}{\sin(\pi+\frac{c_{n}}{\ln\varepsilon})} (4.9)
>asin(π+cnlnε)=πlnε+cnln2εsin(cnlnε)πcn.\displaystyle>\frac{a}{\sin(\pi+\frac{c_{n}}{\ln\varepsilon})}=\frac{\frac{\pi}{\ln\varepsilon}+\frac{c_{n}}{\ln^{2}\varepsilon}}{\sin(-\frac{c_{n}}{\ln\varepsilon})}\rightarrow-\frac{\pi}{c_{n}}.

Therefore

lim infε0[n(n+2)4(n+1)acot(alnεu)a2]n(n+2)4(n+1)πcn.\liminf\limits_{\varepsilon\rightarrow 0}[\frac{n(n+2)}{4}-(n+1)a\cdot\cot(a\ln\varepsilon u)-a^{2}]\geq\frac{n(n+2)}{4}-(n+1)\frac{\pi}{c_{n}}. (4.10)

If we choose cn4π(n+1)n(n+2),c_{n}\geq\frac{4\pi(n+1)}{n(n+2)}, then the formula (4.10)is nonnegative and finally we can get that

Δψψn24+a2-\frac{\Delta\psi}{\psi}\geq\frac{n^{2}}{4}+a^{2} (4.11)

on FεF_{\varepsilon} provided ε\varepsilon is sufficiently small. Then

λ1(Fε)n24+a2=n24+π2ln2ε+O(1ln3ε).\lambda_{1}(F_{\varepsilon})\geq\frac{n^{2}}{4}+a^{2}=\frac{n^{2}}{4}+\frac{\pi^{2}}{\ln^{2}\varepsilon}+O(\frac{1}{\ln^{3}\varepsilon}). (4.12)

For any pXp\in X and large R>0,R>0, let’s consider the first Dirichlet eigenvalue of B(p,R).B(p,R). Since

|u1x|C1,|lnx()dg(p,)|C2|u-\frac{1}{x}|\leq C_{1},\ \ \ |-\ln x(\cdot)-d_{g}(p,\cdot)|\leq C_{2} (4.13)

where C1C_{1} and C2C_{2} are positive constants, we have that

edg(p,)eC2xeC2(uC1)e^{d_{g}(p,\cdot)}\geq\frac{e^{-C_{2}}}{x}\geq e^{-C_{2}}(u-C_{1}) (4.14)

Then

B(p,R){qX:u(q)<eR+C2+C1}FeRC3B(p,R)\subseteq\{q\in X:u(q)<e^{R+C_{2}}+C_{1}\}\subseteq F_{e^{-R-C_{3}}}

for some constant C3>0C_{3}>0 when RR is large enough. Hence

λ1(B(p,R)\displaystyle\lambda_{1}(B(p,R) λ1(FeRC3)\displaystyle\geq\lambda_{1}(F_{e^{-R-C_{3}}}) (4.15)
n24+π2(R+C3)2+O(1(R+C3)3)\displaystyle\geq\frac{n^{2}}{4}+\frac{\pi^{2}}{(R+C_{3})^{2}}+O(\frac{1}{(R+C_{3})^{3}})
=n24+π2R2+O(1R3).\displaystyle=\frac{n^{2}}{4}+\frac{\pi^{2}}{R^{2}}+O(\frac{1}{R^{3}}).

Proof of theorem 1.1: theorem 3.1 and the eigenvalue comparison theorem in [3] provide the upper bound of the first Dirichlet eigenvalue of balls while (4.15) provides the lower bound. Then we finish the proof for theorem 1.1.

5 The geometric property of the asymptotical behavior

To prove proposition 1.2, we introduce an important result of Li and Wang:

Theorem 5.1.

[8] Let (M,g)(M,g) be an n+1n+1 dimensional complete manifold with n2.n\geq 2. Suppose that Ric[g]nRic[g]\geq-n and λ1(M)=n24.\lambda_{1}(M)=\frac{n^{2}}{4}. Then

(1) MM has only one end with infinite volume; or

(2) (M,g)=(×N,dt2+e2tgN)(M,g)=(\mathbb{R}\times N,dt^{2}+e^{2t}g_{N}) where (N,gN)(N,g_{N}) is an nn-dimensional compact manifold satisfying that Ric[gN]0;Ric[g_{N}]\geq 0; or

(3) n=2n=2 and (M,g)=(×N,dt2+cosh2tgN)(M,g)=(\mathbb{R}\times N,dt^{2}+\cosh^{2}tg_{N}) where NN is a compact surface satisfying that the sectional curvature KN1.K_{N}\geq-1.

In the following, we will show that the rate of eigenvalues in case (2) and (3) of theorem 5.1 does not match the formula (1.2).

Example 5.2.

Let (M,g)=(×N,dt2+e2tgN)(M,g)=(\mathbb{R}\times N,dt^{2}+e^{2t}g_{N}) be an n+1n+1-dimensional manifold (n2)(n\geq 2) where (N,gN)(N,g_{N}) is an nn-dimensional compact manifold satisfying that Ric[gN]0.Ric[g_{N}]\geq 0. Then Ric[g]nRic[g]\geq-n and λ1(M)=n24.\lambda_{1}(M)=\frac{n^{2}}{4}.

Now we are going to study the formula of λ1(B(p,R)).\lambda_{1}(B(p,R)). We assume that dN=diam(N,gN)>0d_{N}=diam(N,g_{N})>0 and for any pMp\in M and large R>0,R>0, let dp=distg(p,N).d_{p}=dist_{g}(p,N). Then

B(p,Rdp)B(N,R)B(p,R+dN+dp).B(p,R-d_{p})\subseteq B(N,R)\subseteq B(p,R+d_{N}+d_{p}). (5.1)

Here B(N,R)=(R,R)×N.B(N,R)=(-R,R)\times N. Let f=en2tcos(π2Rt),f=e^{-\frac{n}{2}t}\cos(\frac{\pi}{2R}t), then

f>0inB(N,R),f=0onB(N,R).f>0\ \ in\ B(N,R),\ \ \ \ f=0\ \ on\ \partial B(N,R).

On the other hand,

Δf\displaystyle\Delta f =f(t)Δt+f′′(t)|t|2=nf(t)+f′′(t)\displaystyle=f^{\prime}(t)\Delta t+f^{\prime\prime}(t)|\nabla t|^{2}=nf^{\prime}(t)+f^{\prime\prime}(t) (5.2)
=nen2t[n2cos(π2Rt)π2Rsin(π2Rt)]+en2t[n24cos(π2Rt)\displaystyle=ne^{-\frac{n}{2}t}[-\frac{n}{2}\cos(\frac{\pi}{2R}t)-\frac{\pi}{2R}\sin(\frac{\pi}{2R}t)]+e^{-\frac{n}{2}t}[\frac{n^{2}}{4}\cos(\frac{\pi}{2R}t)
+nπ4Rsin(π2Rt)+nπ4Rsin(π2Rt)π24R2cos(π2Rt)]\displaystyle\ \ \ \ +\frac{n\pi}{4R}\sin(\frac{\pi}{2R}t)+\frac{n\pi}{4R}\sin(\frac{\pi}{2R}t)-\frac{\pi^{2}}{4R^{2}}\cos(\frac{\pi}{2R}t)]
=en2t[n24cos(π2Rt)π24R2cos(π2Rt)]\displaystyle=e^{-\frac{n}{2}t}[-\frac{n^{2}}{4}\cos(\frac{\pi}{2R}t)-\frac{\pi^{2}}{4R^{2}}\cos(\frac{\pi}{2R}t)]
=(n24+π24R2)f\displaystyle=-(\frac{n^{2}}{4}+\frac{\pi^{2}}{4R^{2}})f

which means that uu is an eigenfunction and λ1(B(N,R))=n24+π24R2.\lambda_{1}(B(N,R))=\frac{n^{2}}{4}+\frac{\pi^{2}}{4R^{2}}. Then the monotonicity of the first Dirichlet eigenvalue and (5.1) would imply that

λ1(B(p,R))=n24+π24R2+O(R3),R+.\lambda_{1}(B(p,R))=\frac{n^{2}}{4}+\frac{\pi^{2}}{4R^{2}}+O(R^{-3}),\ \ R\rightarrow+\infty. (5.3)
Example 5.3.

Let (M,g)=(×N,dt2+cosh2(t)gN)(M,g)=(\mathbb{R}\times N,dt^{2}+\cosh^{2}(t)g_{N}) be a 33-dimensional manifold where (N,gN)(N,g_{N}) is a compact surface with Gaussian curvature bounded from below by 1.-1. Then Ric[g]2Ric[g]\geq-2 and λ1(M)=1.\lambda_{1}(M)=1.

As discussed above, we only need to calculate the first Dirichlet eigenvalue of B(N,R)=(R,R)×N.B(N,R)=(-R,R)\times N. Let f=cos(π2Rt)cosh(t),f=\frac{\cos(\frac{\pi}{2R}t)}{\cosh(t)}, then

f>0inB(N,R),f=0onB(N,R).f>0\ \ in\ B(N,R),\ \ \ \ f=0\ \ on\ \partial B(N,R).

Furthermore,

f(t)=π2Rsin(π2Rt)cosh(t)tanh(t)ff^{\prime}(t)=-\frac{\pi}{2R}\frac{\sin(\frac{\pi}{2R}t)}{\cosh(t)}-\tanh(t)\cdot f (5.4)

and

f′′(t)=(π24R4+tanh2(t)1cosh2(t))f+πRsinh(t)cosh2(t)sin(π2Rt)f^{\prime\prime}(t)=(-\frac{\pi^{2}}{4R^{4}}+\tanh^{2}(t)-\frac{1}{\cosh^{2}(t)})f+\frac{\pi}{R}\frac{\sinh(t)}{\cosh^{2}(t)}\sin(\frac{\pi}{2R}t) (5.5)

and hence

Δf\displaystyle\Delta f =f(t)Δt+f′′(t)|t|2=f(t)2tanh(t)+f′′(t)\displaystyle=f^{\prime}(t)\Delta t+f^{\prime\prime}(t)|\nabla t|^{2}=f^{\prime}(t)\cdot 2\tanh(t)+f^{\prime\prime}(t) (5.6)
=πRsinh(t)cosh2(t)sin(π2Rt)2tanh2(t)f\displaystyle=-\frac{\pi}{R}\frac{\sinh(t)}{\cosh^{2}(t)}\sin(\frac{\pi}{2R}t)-2\tanh^{2}(t)\cdot f
+(π24R4+tanh2(t)1cosh2(t))f+πRsinh(t)cosh2(t)sin(π2Rt)\displaystyle\ \ \ \ +(-\frac{\pi^{2}}{4R^{4}}+\tanh^{2}(t)-\frac{1}{\cosh^{2}(t)})f+\frac{\pi}{R}\frac{\sinh(t)}{\cosh^{2}(t)}\sin(\frac{\pi}{2R}t)
=(π24R4tanh2(t)1cosh2(t))f\displaystyle=(-\frac{\pi^{2}}{4R^{4}}-\tanh^{2}(t)-\frac{1}{\cosh^{2}(t)})f
=(1+π24R2)f\displaystyle=-(1+\frac{\pi^{2}}{4R^{2}})f

We obtain that λ1(B(N,R))=1+π24R2\lambda_{1}(B(N,R))=1+\frac{\pi^{2}}{4R^{2}} and hence for any pM,p\in M,

λ1(B(p,R))=1+π24R2+O(R3),R+.\lambda_{1}(B(p,R))=1+\frac{\pi^{2}}{4R^{2}}+O(R^{-3}),\ \ R\rightarrow+\infty. (5.7)

Theorem 5.1 and (5.3), (5.7) all lead naturally to Proposition 1.2.

References

  • [1] Sergei Artamoshin. Lower bounds for the first dirichlet eigenvalue of the laplacian for domains in hyperbolic space. In Mathematical Proceedings of the Cambridge Philosophical Society, volume 160, pages 191–208. Cambridge University Press, 2016.
  • [2] Mingliang Cai and Gregory J Galloway. Boundaries of zero scalar curvature in the ads/cft correspondence. arXiv preprint hep-th/0003046, 2000.
  • [3] Shiu-Yuen Cheng. Eigenvalue comparison theorems and its geometric applications. Mathematische Zeitschrift, 143:289–297, 1975.
  • [4] Michael E Gage. Upper bounds for the first eigenvalue of the laplace-beltrami operator. Indiana University Mathematics Journal, 29(6):897–912, 1980.
  • [5] Colin Guillarmou and Jie Qing. Spectral characterization of poincaré–einstein manifolds with infinity of positive yamabe type. International Mathematics Research Notices, 2010(9):1720–1740, 2010.
  • [6] John M Lee. The spectrum of an asymptotically hyperbolic einstein manifold. Communications in Analysis and Geometry, 3(2):253–271, 1995.
  • [7] Peter Li and Jiaping Wang. Complete manifolds with positive spectrum. Journal of Differential Geometry, 58(3):501–534, 2001.
  • [8] Peter Li and Jiaping Wang. Complete manifolds with positive spectrum, ii. Journal of Differential Geometry, 62(1):143–162, 2002.
  • [9] Rafe Mazzeo. The hodge cohomology of a conformally compact metric. Journal of differential geometry, 28(2):309–339, 1988.
  • [10] Rafe Mazzeo. Unique continuation at infinity and embedded eigenvalues for asymptotically hyperbolic manifolds. American Journal of Mathematics, 113(1):25–45, 1991.
  • [11] Henry P McKean. An upper bound to the spectrum of \triangle on a manifold of negative curvature. Journal of Differential Geometry, 4(3):359–366, 1970.
  • [12] Alessandro Savo. On the lowest eigenvalue of the hodge laplacian on compact, negatively curved domains. Annals of Global Analysis and Geometry, 35:39–62, 2009.
  • [13] Xiaodong Wang. On conformally compact einstein manifolds. Mathematical Research Letters, 8(5):671–688, 2001.
  • [14] Xiaodong Wang. A new proof of lee’s theorem on the spectrum of conformally compact einstein manifolds. Communications in Analysis and Geometry, 10(3):647–651, 2002.
  • [15] Edward Witten and ST Yau. Connectedness of the boundary in the ads/cft correspondence. Advances in Theoretical and Mathematical Physics, 3(6):1–19, 1999.

Xiaoshang Jin
School of mathematics and statistics, Huazhong University of science and technology, Wuhan, P.R. China. 430074
Email address: jinxs@hust.edu.cn