Making Puzzle Pieces Fit or Reshaping MiMiC for Multiscale Simulations with CP2K and More

TL;DR

This paper enhances the MiMiC framework for multiscale chemical simulations by integrating CP2K, demonstrating that modular design maintains performance and enables flexible QM/MM MD simulations with multiple software tools.

Contribution

The paper introduces a major refactoring of MiMiC to include CP2K, expanding its flexibility and demonstrating seamless integration with other software like OpenMM without performance loss.

Findings

Adding CP2K does not impact performance

Modular design enables flexible multiscale simulations

Successful QM/MM MD simulation with multiple programs

Abstract

MiMiC is a framework for modeling large-scale chemical processes that require treatment at multiple resolutions. It does not aim to implement single-handedly all methods required to treat individual subsystems, but instead, it relegates this task to specialized computational chemistry software while it serves as an intermediary between these external programs, and computes the interactions between the subsystems. MiMiC minimizes issues typically associated with molecular dynamics performed with multiple programs, by adopting a multiple-program multiple-data paradigm combined with a loose-coupling model. In this article, we present the addition of a new client program, CP2K, to the MiMiC ecosystem, which required a major refactoring of the entire framework and in the end allowed us to unlock its full flexibility. By thorough timing analysis, we verify that the introduced changes do not affect the performance of MiMiC or CP2K, and neither are they a source of significant computational overheads that would be detrimental to simulation efficiency. Moreover, we demonstrate the benefits of the framework's modular design, by performing a QM/MM MD simulation combining CP2K with previously interfaced OpenMM, with no additional implementation effort required.