Hybrid Monte Carlo simulations theory and initial comparison with molecular dynamics

Abstract A formal partially dynamical approach to ergodic sampling, hybrid Monte Carlo, has been adapted for the first time from its proven application in quantum chromodynamics to realistic molecular systems. A series of simulations of pancreatic trypsin inhibitor were run using temperature‐rescaled molecular dynamics and hybrid Monte Carlo. It was found that simulations run using hybrid Monte Carlo equilibrated an order of magnitude faster than those run using temperature‐rescaled molecular dynamics. Certain aspects of improved performance obtained using hybrid Monte Carlo are probably due to the increased efficiency with which this algorithm explores phase space. To discuss this we introduce the notion of “trajectory stiffness”. © 1993 John Wiley & Sons, Inc.