The Modern Temperature-Accelerated Dynamics Approach

摘要

Accelerated molecular dynamics (AMD) is a class of MD-based methods used to simulate atomistic systems in which the metastable state-to-state evolution is slow compared with thermal vibrations. Temperature-accelerated dynamics (TAD) is a particularly efficient AMD procedure in which the predicted evolution is hastened by elevating the temperature of the system and then recovering the correct state-to-state dynamics at the temperature of interest. TAD has been used to study various materials applications, often revealing surprising behavior beyond the reach of direct MD. This success has inspired several algorithmic performance enhancements, as well as the analysis of its mathematical framework. Recently, these enhancements have leveraged parallel programming techniques to enhance both the spatial and temporal scaling of the traditional approach. We review the ongoing evolution of the modern TAD method and introduce the latest development: speculatively parallel TAD.