Abstract
When using the adaptive biasing force (ABF) algorithm in free-energy calculations, unbiased sampling is performed across the transition coordinate to estimate the biasing force (F bias). In practice, however, setting the number of molecular dynamics steps (N samples) that precede the application of the time-dependent bias is arbitrary. In this paper, we find that N samples has a significant effect on the outcome of the free-energy calculation. When N samples is too small, marked non-equilibrium effects may occur, conducive to slow convergence, or possibly erroneous results. This phenomenon is observed in a variety of prototypical cases. Conversely, setting arbitrarily N samples to an exaggeratedly large value implies extra computational efforts. To avoid possible non-equilibrium artifacts while maintaining computational effectiveness, we propose a reliable scheme to make an appropriate choice of N samples in ABF-based free-energy calculations, resting on the evolution F bias as a function of N samples. Monitoring the latter, a value of N samples that minimises the computational cost while guaranteeing convergence of F bias can be determined. This method offers a safeguard for reliable free-energy calculations using ABF-based algorithms.
Funding Information
  • China Postdoctoral Science Foundation (bs6619012)
  • National Natural Science Foundation of China (21773125)
  • National Science Foundation of Tianjin, China (18JCYBJC20500)

This publication has 42 references indexed in Scilit: