A comparison of the efficiency of Fourier- and discrete time-path integral Monte Carlo

We compare the efficiency of Fourier and discrete time path integral Monte Carlo (PIMC) methods on a cluster of 22 hydrogen molecules at 6 K. The discrete time PIMC with a pair density matrix approximation to the path action is shown to be the most efficient for evaluating all the observables studied here. The Fourier PIMC technique has a comparable efficiency for observables diagonal in the coordinates but is significantly worse for estimating the kinetic and total energies. The superior performance of the discrete time PIMC is shown to be due to the more accurate treatment of the path action using the pair density matrix approach; the discrete time PIMC simulation within the primitive approximation is much less efficient. Complete details of the implementation of all algorithms are given.