A GENERALIZED GUIDED MONTE-CARLO ALGORITHM

A hamiltonian-guided Monte Carlo algorithm for simulations of lattice field theories is presented. It is a generalization of the Hybrid Monte Carlo algorithm allowing the trajectory length to be shrunk to the step-size without losing on the speed of configuration decorrelation. Without the Monte Carlo step, this limit of the generalized algorithm reduces to the second-order Langevin equation. The latter is shown in a gaussian model and in compact QED to be faster than the Hybrid algorithm. The performance of the two algorithms with the Monte Carlo step, HMC and L2MC, turned out to be comparable in compact QED. It is pointed out that L2MC is safer from rounding errors.