A characterization of \(L_{2}\) mixing and hypercontractivity via hitting times and maximal inequalities (Q2413244)

Consider a continuous-time reversible Markov chain \((X_t)_{t \geq 0}\) with state space \(\Omega\). Denote by \(P_x^t\) and \(\pi\) the distribution of \(X_t\) given that \(X_0=x\) and a stationary distribution, respectively. Given that \(X_0=x\), the \(L_p\)-distance at time \(t\) is defined as \[ d_{p,x}(t)=\left(\sum_{y \in \Omega} |P_x^t(y)/\pi(y)-1|^p \pi(y)\right)^{1/p} \] for \(1 \leq p < \infty\). The \(L_p\)-mixing time is \(\tau_p=\max_{x \in \Omega} \min\{t:d_{p,x}(t)\leq 1/2\}\). Although there are numerous characterizations of \(\tau_1\) in terms of stopping times or hitting times, similar characterizations of \(\tau_2\) were not known yet. In the article, \(\tau_2\) is characterized up to a constant factor using a hitting time distribution. For example, the following result is obtained. Let \(T_A=\inf\{t:X_t \in A\}\) denote the hitting time of a set \(A \subset \Omega\), \(P_x\) denote the distribution of \((X_t)_{t \geq 0}\) given that \(X_0=x\) and \(A^c=\Omega \setminus A\). Define \(\rho=\max_{x \in \Omega}\rho_x\), where \(\rho_x=\min t\) is calculated under the condition \(P_x\{T_{A^c}>t\} \leq \pi(A)+0.5 (\pi(A) \pi(A^c))^{1/2}\) for all \(A\) such that \(\pi(A)\leq 0.5\). Then there exists some absolute constant \(C_1\) that \[ \rho \leq \tau_2 \leq \rho+C_1/c_{LC}. \] Here \(c_{LC}\) is a Log-Sobolev constant for which a new extremal characterization is derived as a weighted version of the spectral gap. Some results concerning the robustness of \(\tau_2\) are also presented.