Improving replica exchange using driven scaling

Replica exchange is a powerful simulation method in which simulations are run at a series of temperatures, with the highest temperature chosen so phase space can be sampled efficiently. In order for swaps to be accepted, the energy distributions of adjacent replicas must have some overlap. This can create the need for many replicas for large systems. In this paper, we present a new method in which the potential energy is scaled by a parameter, which has an explicit time dependence. Scaling the potential energy broadens the distribution of energy and reduces the number of replicas necessary to span a given temperature range. We demonstrate that if the system is driven by the time-dependent potential sufficiently slowly, then equilibrium is maintained and energetic and structural properties are identical to those of conventional replica exchange. The method is tested using two systems, the alanine dipeptide and the trpzip2 polypeptide, both in water