A note on the asymptotic variance of drift accelerated diffusions

The asymptotic variance is a natural indicator of the efficiency for Markov Chain Monte Carlo algorithm. In this note, we prove that drift accelerated diffusion converges to zero uniformly if and only there are no non-trivial first order Sobolev functions in kernel generating operator. Its proof based on spectral analysis space mean functions.