Definition 2.1
QZC2The Sobolev space W^{k,p}(M) is the completion of C^p_k(M) with respect to the norm
\left\lVertu\right\rVert_{W^{k,p}}=\sum_{j=0}^{k}\Big(\int_M|\nabla^ju|^pdV\Big)^{1/p}.
Integral Liouville theorem in a complete Riemannian manifold
- arXiv
- 1808.00180v1 · PDF · 1 August 2018
- Categories
- math.DG
- Journal
- Balkan Journal of Geometry and Its Applications, Vol. 23, No. 2 (2018), 56-66
- Licence
- CC BY 4.0 The site shows the text.
Abstract
If the Killing vector field in a Riemannian manifold is the gradient of a smooth real valued function, then it is called Killing potential. In this paper we have deduced a necessary condition for the existence of Killing potential in a complete Riemannian manifold. Yau proved the Liouville theorem of harmonic function in a Riemannian manifold using gradient estimation and after that many authors have generalized this concept and investigated various types of Liouville theorems of harmonic functions. In this article we have also proved a Liouville theorem in integral form of harmonic functions in a complete Riemannian manifold. Finally we have studied the behaviour of harmonic functions in a complete Riemannian manifold that satisfies some gradient Ricci solition and showed that harmonic function is a constant multiple of distance function along some geodesics.
Items 21
Definition 2.2
I8B6[YAU75] A C^2-function u:M\rightarrow\mathbb{R} is said to be harmonic if \Delta u=0. The function u is called subharmonic (resp. superharmonic) if \Delta \geq 0 (resp. \Delta u\leq 0).
Definition 2.3
748N[UDR94] A real valued function u on M is called convex if for every geodesic \gamma:[a,b]\rightarrow M, the following inequality holds
u\circ\gamma((1-t)a+tb)\leq (1-t)u\circ\gamma(a)+tu\circ\gamma(b)\quad \forall t\in [0,1],
or if u is differentiable, then
g(\nabla u,X)_x\leq u(exp_x\nabla u)-u(x), \ \forall X\in T_xM.
Theorem 1
LVFRThere does not exist any non-trivial Killing potential u\in C^2(M) in a complete Riemannian manifold M with Ric_M\geq cg for some constant c>0.
Proof of Theorem 1
Suppose u\in C^2(M) is a non-trivial Killing potential. Now Hessian of u vanishes [CM99]. Since u is a Killing potential so u is harmonic function [CM99]. Hence using Bochner’s formula we get
\frac{1}{2}\Delta|\nabla u|^2=|Hess(u)|^2+Ric(\nabla u,\nabla u)=Ric(\nabla u,\nabla u).
Again Ric_M\geq cg implies that
\Delta |\nabla u|^2\geq 2cg(\nabla u,\nabla u)=2c|\nabla u|^2.
Since u is Killing potential so \nabla u is covariant constant field [CM99], i.e., \nabla_ X(\nabla u)=0 \quad \forall X\in \chi(M), which implies that g(\nabla u,\nabla u)=C, for some constant C> 0. Hence \Delta |\nabla u|^2=0 and it leads to a contradiction. Hence u is constant.
Lemma 2
LMX9[BN08, Corollary 2] A complete Riemannian manifold M of negative definite Ricci curvature has no nontrivial Killing fields of constant length.
Corollary 2.1
YQAQIf a complete Riemannian manifold M admits a Killing potential u\in C^\infty(M), then Ric(\nabla u,\nabla u)_p=0\ \forall p\in M.
Theorem 3 (Integral Liouville Theorem)
IORYLet M be a complete Riemannian manifold and u:M\rightarrow\mathbb{R} be a harmonic function. If there exists a point p\in M, where M is rotationally symmetric and u(p)<0 with inj(p)>1, then
\int_{B_r}(K+u)dV=0\quad \forall r\in [0,inj(p)),
for some constant K>0.
Proof of Theorem 3
Let p\in M be a fixed point and B_r be an open geodesic ball with center at p and radius r such that 0\leq r\leq inj(p) and also inj(p)>1. Now for each r>0, we define
v(r)=\frac{1}{Vol(\partial B_r)}\int_{\partial B_r}udS_r,
where dS_r is the (n-1)-dimensional volume element of B_r. Since M is rotationally symmetric at p so using (?) and (Proof), we get
v(r)=\frac{1}{Vol(\partial B_1)}\int_{\partial B_1}u(r\eta)dS_1.
Now differentiating the above function with respect to r, we obtain
v'(r)=\frac{1}{Vol(\partial B_1)}\int_{\partial B_1}\frac{\partial u}{\partial r}(r\eta)dS_1=\frac{1}{Vol(\partial B_r)}\int_{\partial B_1}\partial_rudS_r.
Hence for r>0, by using divergence theorem, we get
v'(r)Vol(\partial B_r)=\int_{B_r}\Delta u dV.
In view of harmonicity of u, the above equation implies that v'(r)=0, i.e., v=C, a constant. Since v is continuous so v is constant in [0,inj(p)). Hence (Proof) implies that
\int_{\partial B_r}udS_r=CVol(\partial B_r).
Since u(p)<0, then there exists \epsilon>0, such that u<0 in \partial B_\epsilon. Hence
C=\frac{1}{Vol(\partial B_\epsilon)}\int_{\partial B_\epsilon}udS_\epsilon<0.
Suppose C=-K, where K>0. Hence
Vol(\partial B_r)=-\frac{1}{K}\int_{\partial B_r}udS_r.
Now from (Proof), we get
Vol(B_r)=\int_{0}^{r}Vol(\partial B_t)dt=-\frac{1}{K}\int_{0}^{r}\int_{\partial B_t}udS_tdt=-\frac{1}{K}\int_{B_r}udV.
The above equation implies that
\int_{B_r}dV+\frac{1}{K}\int_{B_r}udV=0,
i.e.,
\int_{B_r}(K+u)dV=0,\quad \forall r\in [0,inj(p)).
Corollary 3.1
CGQ3Suppose (M,o) is a Riemannian manifold with a pole o and is rotationally symmetric at o. If u:M\rightarrow\mathbb{R} is a harmonic function such that f(o)<0, then there exists a constant K>0 such that
\int_{B_r(o)}(K+u)dV=0,\quad \forall r\in [0,\infty).
Used by Theorem 4
Theorem 4
B38RAssuming the conditions of the above corollary with Ric_M\geq 0, there exists a constant C such that
\int_{B_r(o)}(C+u)dV=0,\quad \forall r\in [0,\infty).
Proof of Theorem 4
Let u\in C^2(M) be a harmonic function. Since a manifold with positive semi-definite Ricci curvature does not admit any non-constant positive harmonic function [SY94, p. 21], so there exists a point q\in M such that u(q)\leq 0. Without loss of generality we can take u(q)<0, otherwise we have to subtract some positive constant. If u(o)<0, then taking p=o the proof is complete. If u(o)\geq 0, then take f(x)=u(x)-u(o) for x\in M. Hence f is harmonic and f(o)<0. So using above corollary we get there is a constant K such that for all r>0
\begin{aligned}
0&= \int_{B_r(o)}(K+f)dV,\\
&=\int_{B_r(o)}(K+u(x)-u(o))dV.
\end{aligned}
Hence taking C=K+u(o), we get our result.
Corollary 4.1
95EJLet (M,o) be a Riemannian manifold with Ric_M\geq 0. Suppose M is rotationally symmetric at o. Then all non-constant convex functions u\in C^2(M) with |\nabla u|\in L^1(M) satisfy
\int_{B_r}(K+u)dV=0,\quad \forall r\in [0,\infty),
for some constant K.
Proof of Corollary 4.1
Since u is convex so u is subharmonic. Since v has integrable norm so u is harmonic [YAU76]. Hence the proof follows from the above Theorem.
Uses Theorem 4
Definition 4.1
7EJ4[CM97] A complete Riemannian manifold M is said to satisfy a uniform Neumann-Poincaré inequality if there exists C_N<\infty such that, for all p\in M, r>0 and u\in W^{1,2}_{loc}(M),
\int_{B_r}\Big(u-\frac{1}{Vol(B_r)}\int_{B_r}udV \Big)^2dV\leq C_Nr^2\int_{B_r}|\nabla u|^2dV.
Proposition 5
ACBKLet M be a n-dimensional complete Riemannian manifold with Ric_M>-K, for K>0 and u\in W^{1,2}_{loc}(M) be a harmonic function and also M satisfies a uniform Neumann-Poincaré inequality. If u is bounded below and there is a point p\in M where M is rotationally symmetric and such that u(p)<0 with inj(p)>1, then there exist C>0,C_N and c>0, depending only on n, such that, for 0\leq r<inj(p), the following inequality holds
\int_{B_r}\Big(u+C \Big)^2dV\leq (n-1)KC_Nr^2\int_{B_r}(u+c)^2dV.
Theorem 6
EGR0Let (M,g) be a complete non-compact Riemannian manifold with Ric_M>-K, where K>0. If there exists a harmonic function u\in C^2(M) which is bounded below and satisfies
\mathcal{L}_{\nabla u}g+Ric\geq cg,
then for some geodesic \gamma:[0,\infty)\rightarrow M, the function u takes the form
u(\gamma(t))=\frac{1}{\sqrt{(n-1)K}}(ct+G), \quad \forall 0\leq t<\infty,
for some constant G.
Used by Proof
Proof of Theorem 6
Since M is not compact, so by Ambrose’s compactness criteria [AM57] there exists a geodesic \gamma:[0,\infty)\rightarrow M, parametrized by arc length, such that
\int_{0}^{\infty}Ric(\gamma',\gamma')dt<\infty.
Then taking X=\nabla u, along \gamma we have
\mathcal{L}_Xg(\gamma',\gamma')=2g(\nabla_{\gamma'}X,\gamma')=2\frac{d}{dt}[g(X,\gamma')].
Hence using above equation and (6), we obtain
Ric(\gamma',\gamma')\geq cg(\gamma',\gamma')-2\frac{d}{dt}[g(X,\gamma')].
Now taking \gamma(0)=q and integrating and using the Cauchy-Schwarz inequality, we get
\begin{aligned}
\int_{0}^{T}Ric(\gamma',\gamma')dt &\geq & c\int_{0}^{T}g(\gamma',\gamma')dt-2\int_{0}^{T}\frac{d}{dt}[g(X,\gamma')]dt\\
&\geq & cT+2g(X_q,\gamma'(0))-2g(X_{\gamma(T)},\gamma'(T))\\
&\geq & cT+2g(X_q,\gamma'(0))-2\left\lVertX_{\gamma(T)}\right\rVert.
\end{aligned}
Now from (Proof), we get
\lim\limits_{T\rightarrow\infty}(cT+2g(X_q,\gamma'(0))-2\left\lVertX_{\gamma(T)}\right\rVert)<\infty.
Hence
\lim\limits_{T\rightarrow\infty}(cT-2\left\lVertX_{\gamma(T)}\right\rVert)<\infty.
Since Ric_M>-K and u is harmonic with lower bound, then gradient estimate Theorem (?) implies that
\lim\limits_{T\rightarrow\infty}(cT-2\sqrt{(n-1)K}(u-\inf_Mu))<\infty.
Consequently
\lim\limits_{T\rightarrow\infty}(cT-2u\sqrt{(n-1)K})<\infty.
Hence there exists a constant G such that
u(\gamma(t))=\frac{1}{\sqrt{(n-1)K}}(ct+G), \quad \forall 0\leq t<\infty,
Corollary 6.1
L6H1Suppose (M,g) is a complete non-compact Riemannian manifold with Ric_M>-K. If there is non-positive harmonic function u\in C^2(M) satisfying (6) for some positive constant c, then u is constant. Hence M satisfies Ric\geq cg.
Corollary 6.2
O1QTSource and licence
The text on this page is from arXiv:1808.00180v1 by Absos Ali Shaikh and Chandan Kumar Mondal, 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/1808/1808.00180.jsonl.