Heat kernels on manifolds, graphs and fractals (Q4408045)

The author gives a brief survey of heat kernel long-time estimates on different spaces such as Riemannian manifolds, discrete graphs and abstract metric spaces including fractals. After defining the heat kernels on the different spaces the author begins with on-diagonal upper (resp.\ lower) estimates of the heat kernel on manifolds in relation with first eigenvalue lower (resp.\ upper) estimates of Faber-Krahn type. As an application, the author deduces lower estimates on the \(k\)-th Dirichlet eigenvalue of the Laplacian on precompact subsets of the manifold. Upper estimates of Gaussian type on \(p_t(x,y)\) on manifolds can be obtained from upper \textit{on-diagonal} estimates without further assumptions. A two-sided Gaussian estimate for \(0<t<t_0\) is equivalent to a volume growth condition and a Poincaré-type inequality, both valid on balls of radius \(0<r<\sqrt {t_0}\). Finally, the author reviews sub-Gaussian estimates, i.e., estimates on \(p_t(x,y)\) of the form \[ \frac 1 {t^{\alpha/\beta}} \exp \left( - \left( \frac {d^{\beta}(x,y)}{ct} \right)^{1/(\beta-1)} \right) \] where \(\alpha \geq 1\) is the (Hausdorff) dimension and \(\beta \geq 2\) is an additional parameter, a so-called ``walk dimension'' (on fractals, one has \(\beta>2\)). Equivalent characterisations for sub-Gaussian estimates like polynomial volume growth and Green function decay or a Harnack inequality are stated. This comprehensible review contains no proof but a long list of references. For the entire collection see [Zbl 0972.00031].