PySAGES: Enhanced Sampling for Ab Initio Dynamics and Machine Learning Potentials

PySAGES is a python library for performing enhanced sampling in molecular dynamics simulations. It provides a friendly interface and allows the user to leverage and write complex collective variables and enhanced sampling methods. Here we show how PySAGES can be coupled to Ab Initio integrators using the ASE interface for enhanced sampling in systems where frist principles accuracy is necessary. We also present how PySAGES can be used as a tool for studying the robustness of machine learned force fields. In particular we evaluate a set of different examples of DeePMD, GAP and Graph Neural Network potentials and look at their reliability for preserving the behavior of a system in term of a selected collection of relevant molecular descriptors.

Overall, we hope to convey how PySAGES can be used to accelerate the study and analysis of molecular dynamics simulation as well as the design of molecules and materials in a wide range of areas.