High order actions proposed by Chin have been used for the first time in pathintegral Monte Carlo simulations. Contrarily to the Takahashi-Imada action,which is accurate to fourth order only for the trace, the Chin action is fullyfourth order, with the additional advantage that the leading fourth and sixthorder error coefficients are finely tunable. By optimizing two free parametersentering in the new action we show that the time step error dependence achievedis best fitted with a sixth order law. The computational effort per bead isincreased but the total number of beads is greatly reduced, and the efficiencyimprovement with respect to the primitive approximation is approximately afactor of ten. The Chin action is tested in a one-dimensional harmonicoscillator, a H$_2$ drop, and bulk liquid $^4$He. In all cases a sixth-orderlaw is obtained with values of the number of beads that compare well with thepair action approximation in the stringent test of superfluid $^4$He.
展开▼
