Morrone Joseph A, Zhou Ruhong, Berne B J
Department of Chemistry, Columbia University, 3000 Broadway, MC 3103, New York, New York 10027, and IBM Thomas J. Watson Research Center, 1101 Kitchawan Road, Yorktown Heights, New York 10598.
J Chem Theory Comput. 2010 Jun 8;6(6):1798-804. doi: 10.1021/ct100054k. Epub 2010 May 19.
Multiple time scale methodologies have gained widespread use in molecular dynamics simulations and are implemented in a variety of ways across numerous packages. However, performance of the algorithms depends upon the details of the implementation. This is particularly important in the way in which the nonbonded interactions are partitioned. In this work, we show why some previous implementations give rise to energy drifts, and how this can be corrected. We also provide a recipe for using multiple time step methods to generate stable trajectories in large scale biomolecular simulations, where long trajectories are needed.
多时间尺度方法在分子动力学模拟中已得到广泛应用，并在众多软件包中以多种方式实现。然而，算法的性能取决于实现的细节。这在非键相互作用的划分方式中尤为重要。在这项工作中，我们展示了为什么一些先前的实现会导致能量漂移，以及如何纠正这一问题。我们还提供了一种使用多时间步方法在大规模生物分子模拟中生成稳定轨迹的方法，在这种模拟中需要长轨迹。
