Heat kernel estimate for the Laplace-Beltrami operator under Bakry-Émery Ricci curvature condition and applications (Q6141321): Difference between revisions

This paper contains a proof of Gaussian bounds for the heat kernel, and some consequences, for manifolds whose Bakry-Émery Ricci curvature is bounded below. By Bakry-Émery Ricci curvature, we mean \(\mathrm{Ric}_f\left(g\right) := \mathrm{Ric}\left(g\right) + \nabla^2 f\), where \(f : M \rightarrow \mathbb{R}\) is a smooth function and \(\left(M,g\right)\) is a Riemannian manifold. When working with this curvature quantity, it is natural to work with respect to the weighted volume measure \(e^{-f}dV_g\). This then leads to considering a second order operator which is self-adjoint wtih respect to this measure: \(\Delta_f := \Delta - \left<\nabla f, \nabla\right>\). The main estimate in the paper is the following Gaussian bound, which (locally) holds under the lower bound \(\mathrm{Ric}_f\left(g\right) \geq -Kg\) for a constant \(K \geq 0\): \[ H\left(x,y,t\right) \leq \frac{C_1\left(n,\varepsilon\right)e^{C_2\left(n\right)\left(Kt + L\left(R\right)\right)}}{\mathrm{Vol}^{\frac{1}{2}}\left(B_{\sqrt{t}}\left(x\right)\right)\mathrm{Vol}^{\frac{1}{2}}\left(B_{\sqrt{t}}\left(y\right)\right)}e^{-\frac{d^2\left(x,y\right)}{\left(4+\varepsilon\right)t}}. \] Here \(\varepsilon, R, t > 0\) satisfying certain bounds while \(C_1\left(n,\varepsilon\right),C_2\left(n\right),L\left(R\right)\) are constants with the specified dependences. Importantly, the \(H\left(x,y,t\right)\) in the above estimate refers to the heat kernel which solves \(\left(\partial_t - \Delta\right) H\left(x,y,t\right) = 0\). Note that the drift Laplacian \(\Delta_f\), while natural to work with, is not present. In general, working with the drift operator comes with its own issues involving properties of \(f\). Also, \(\Delta\) and \(\Delta_f\) are not equivalent operators in general. Hence bounds on \(H\) do not necessarily imply bounds on \(H_f\) (the heat kernel with respect to \(\partial_t - \Delta_f\)) and vice-versa. The authors provide \(\mathbb{R}^n\) with the standard metric and \(f = x_1\) as a very explicit example. Key tools in the proof of the Gaussian bound are a volume comparison theorem for Bakry-Émery manifolds and local Poincaré and Sobolev inequalities. Some references for these are included in the paper and the reader may also want to refer to [\textit{G. Wei} and \textit{W. Wylie}, J. Differ. Geom. 83, No. 2, 377--405 (2009; Zbl 1189.53036)]. As for applications of the estimate: \begin{itemize} \item[1.] Under the additional assumption of a quadratic upper bound on \(f\), an \(L^1\)-Liouville theorem is proved for subharmonic functions. \item[2.] Specializing to the compact setting, lower bounds for the \(k\)\textsuperscript{th} eigenvalue of the Laplace-Beltrami operator are proved. \item[3.] For complete manifolds with Bakry-Émery curvature bounded below, the authors prove an upper bound on the lower spectrum of the Laplace-Beltrami operator. If \(f\) is taken to be a constant, then this recovers a result of Cheng and gives a sharp bound, with equality being achieved on hyperbolic space \(\mathbb{H}^n\). \end{itemize} As noted by the authors, a class of manifolds which satisfy the various assumptions in the result above are gradient Ricci solitons. These are Riemannian manifolds \(\left(M,g\right)\) with \(\mathrm{Ric}_f\left(g\right) = \lambda g\) for some \(\lambda \in \mathbb{R}\). These frequently arise when studying Ricci flow, either as singularity models or candidates to use in a desingularization procedure. These solitons, which contain Einstein manifolds (\(\mathrm{Ric} = \lambda g\)) as particular examples, can have behavior which seems contradictory to the choice of \(\lambda\), so adjustments to old methods or the development of new ones are usually needed when studying them. In particular, the potential function \(f\) typically plays a large role, as it does in the current article. When \(\lambda > 0\) there are non-compact examples, the most simple of which is the round cylinder \(S^{n-1} \times \mathbb{R}\). Furthermore, one can have solitons with \(\lambda > 0\), but which also have regions of negative Ricci curvature [\textit{M. Feldman}, \textit{T. Ilmanen}, and \textit{D. Knopf}, J. Differ. Geom. 65, No. 2, 169--209 (2003; Zbl 1069.53036)]. For \(\lambda = 0\), there are examples which are not Ricci flat, the standard one being the so-called Bryant soliton. When \(\lambda < 0\) one can have Ricci expanders with non-negative curvature operator asymptotic to a cone with positively curved link [\textit{A. Deruelle}, Geom. Funct. Anal. 26, No. 1, 188--249 (2016; Zbl 1343.53040)], [\textit{F. Schulze} and \textit{M. Simon}, Math. Z. 275, No. 1--2, 625--639 (2013; Zbl 1278.53072)].