Properties of random walks on discrete groups: time regularity and off-diagonal estimates (Q932524): Difference between revisions

The author studies convolution powers \(K^{(n)}\) of a non-symmetric probability distrubution \(K\) on a discrete group \(G\). To describe his results, we introduce the concept of centeredness. Suppose that \(G/[G,G]=\mathbb{Z}^q\times A\), where \(A\) is a finite abelian group, and let \(\pi_0:G\rightarrow G/[G,G]\) be the canonical projection. Set \(G_1=\pi_0^{-1}(\{0\}\times A)\) and let \(\pi_1^{(j)}\) be the \(j\)th component of the homomorphism \(\pi_1:G\rightarrow G/G_1\cong \mathbb{Z}^q\). Then \(K\) is centered when \[ \int_GdgK(g)\pi_1^{(j)}(g)=0, \] for \(j=1,2,\dotsc,q\). The author proves that if \(K\) is centered then the associated Markov operator \(T\) is analytic in \(L^p\), \(1<p<\infty\): this means that there exists \(c>0\) such that \[ \|(I-T)T^n\|_{L_p}\leq cn^{-1} \] for all \(n\in\mathbb{N}\). He also proves Davies-Gaffney estimates in \(L^2\) for \(T^n\) and Gaussian upper bounds for \(K^{(n)}\). In the particular case of an amenable group \(G\), the author proves that analitycity of \(T\) or the Davies-Gaffney estimate hold if and only if \(K\) is centered, while for groups of polynomial growth he obtains new proofs of the Gaussian estimates with precise time decay.