A formal partially dynamical approach to ergodic sampling, hybrid Mont
e Carlo, has been adapted for the first time from its proven applicati
on in quantum chromodynamics to realistic molecular systems. A series
of simulations of pancreatic trypsin inhibitor were run using temperat
ure-rescaled molecular dynamics and hybrid Monte Carlo. It was found t
hat simulations run using hybrid Monte Carlo equilibrated an order of
magnitude faster than those run using temperature-rescaled molecular d
ynamics. Certain aspects of improved performance obtained using hybrid
Monte Carlo are probably due to the increased efficiency with which t
his algorithm explores phase space. To discuss this we introduce the n
otion of ''trajectory stiffness.'' (C) 1993 John Wiley & Sons, Inc.