A note on time regularity for discrete time heat kernels (Q2502261)

Let \(\left( \Omega ,\mu \right)\) be a \(\sigma\)-finite measure space and let \(L^{p}=L^{p}\left(\Omega ,d\mu\right)\), \(1\leq p\leq\infty \). Let \(T\) denote a linear operator acting in some suitable space \(\mathcal{F}\) of complex-valued functions on \(\Omega\). If \(T^{n}\) has a measurable integral kernel with respect to \(\mu\), let denote it by \(K_{n}:\Omega ^{2}\rightarrow \mathbb{C}\), so that \[ \left( T^{n}f\right) \left( x\right) =\int_{\Omega }K_{n}\left( x,y\right) f\left(y\right) \mu\left( dy\right),\;n\geq 1,\;x\in \Omega ,\;f\in \mathcal{F}. \] A simple method is given for deriving regularity estimates of the difference \(K_{n}-K_{n+1}\) for the cases \(\mathcal{F}=L^{2}\) or \(\mathcal{F}=L^{p},\;1\leq p\leq\infty\). This method applies, e.g., to suitable Markov chains and to random walks on weighted graphs. For related work, in which a more complicated approach is used, see \textit{S.~Blunck} [Colloq.\ Math.\ 86, 189--201 (2000; Zbl 0961.47005)].