A "fast growth" method of computing free energy differences

Abstract

Let Delta F be the free energy difference between two equilibrium states of a system. An established method of numerically computing Delta F involves a single, long ``switching simulation'', during which the system is driven reversibly from one state to the other (slow growth, or adiabatic switching). Here we study a method of obtaining the same result from numerous independent, irreversible simulations of much shorter duration (fast growth). We illustrate the fast growth method, computing the excess chemical potential of a Lennard-Jones fluid as a test case, and we examine the performance of fast growth as a practical computational tool.