J. Chem. Phys. Abstract

Abstract


In this paper (paper I) and a companion paper (paper II), novel new algorithms and applications of the isokinetic ensemble as generated by Gauss' principle of least constraint, pioneered for use with molecular dynamics twenty years ago, are presented for biophysical, path integral, and Car-Parrinello based ab initio molecular dynamics. In paper I, a new ``extended system'' version of the isokinetic equations of motion that overcomes the ergodicity problems inherent in the standard approach, is developed using a new theory of non-Hamiltonian phase space analysis [M. E. Tuckerman, et al., Europhys. Lett. 45, 149 (1999); M. E. Tuckerman, et al., J. Chem. Phys. 115, (2001)]. Reversible multiple time step integrations schemes for the isokinetic methods, first presented by Zhang [J. Chem. Phys. 106, 6102 (1997)] are reviewed. Next, holonomic constraints are incorporated into the isokinetic methodology for use in fast efficient biomolecular simulation studies. Model and realistic examples are presented in order to evaluate, critically, the performance of the new isokinetic molecular dynamics schemes. Comparisons are made to, the now standard, canonical dynamics method, Nosé-Hoover chain dynamics [G. J. Martyna, et al., J. Chem. Phys. 97, 2635 (1992)]. The new isokinetic techniques are found to yield more efficient sampling than the Nosé-Hoover chain method in both path integral molecular dynamics and biophysical molecular dynamics calculations. In paper II, the use of isokinetic methods in Car-Parrinello based ab initio molecular dynamics calculations is presented.