A pairing method which improves convergence in Monte-Carlo estimation of quantum mechanical expectation values (Q1067743)

This paper introduces a pairing method which reduces the variance in Monte-Carlo estimates of integrals. The method involves several levels of estimation where at each level the maximum and minimum points of the integrand f(x) are saved. These two vectors are then used in the next level so that the samples are taken equally from regions where f(x) is large and small thus reducing the variance. The method is also applied to quantum mechanical expectation values which may be written as the ratio of two integrals. Calculations are given for \(H^+_ 2\) and \(H_ 2\) and the calculations converge at rates proportional to \(N^{-0.72}\) and \(N^{-0.62}\) respectively compared with \(N^{-0.5}\) for Monte Carlo estimates. Here N is the number of function evaluations.