Exponential convergence in probability for empirical means of Brownian motion and of random walks (Q1303905): Difference between revisions

A bounded function \(f:\mathbb{Z}^d\to\mathbb{R}\) is said to have uniform mean \(M(f)\in\mathbb{R}\) if, uniformly in \(x\in\mathbb{Z}^d\), we have \(\lim_{n\to\infty}(2n)^{-d}\sum_{|y|\leq n}f(x+y)=M(f)\). Let \((X_n)_{n\in\mathbb{N}_0}\) be a random walk on \(\mathbb{Z}^d\) such that the support of the distribution of the steps generates \(\mathbb{Z}^d\). It is proved that, for any bounded function \(f:\mathbb{Z}^d\to\mathbb{R}\), the exponential convergence of \(L_n(f)=\frac 1n\sum_{k=1}^n f(X_k)\) (i.e., the assertion that \(\limsup_{n\to\infty}\frac 1n\log P(|L_n(f) - z|> \delta)<0\) for any \(\delta>0\)) is equivalent to the fact that \(f\) has a uniform mean \(z=M(f)\). Also the analogous assertion for the continuous case is derived, i.e., for Brownian motion instead of a random walk, with a similar notion of a uniform mean, but under the additional hypothesis that \(f:\mathbb{R}^d\to\mathbb{R}\) belongs to the Kato class. A corollary about the spectrum of a self-adjoint Schrödinger operator with small potential is added. The main tools in the proofs are exponential martingale inequalities and large-deviation results for the occupation time measures for Markov chains resp.\ Brownian motion.