Efficient Markovian couplings: Examples and counterexamples. (Q1884821)

It is well-known that coupling techniques can be applied to obtain various estimates in probability and analysis, both in theoretical research and in applications. The authors study the notion of an efficient coupling of Markov processes. A diffusion coupling is called ``efficient'' if it can be used to obtain a sharp estimate for the spectral gap of the operator which is the generator of the diffusion in question. The main results of the paper state that among well-known coupling we can find both efficient and inefficient couplings. Two examples of Markov processes are given for which there is no ``efficient'' Markovian coupling. The first one is continuous-time Markov chains with finite state space and the second one is concerned with reflected Brownian motion in planar domains.