DL_FFLUX: A Parallel, Quantum Chemical Topology Force Field.

Michael K Bane,Benjamin C B Symons,Paul L A Popelier

doi:10.1021/acs.jctc.1c00595

Abstract

DL_FFLUX is a force field based on quantum chemical topology that can perform molecular dynamics for flexible molecules endowed with polarizable atomic multipole moments (up to hexadecapole). Using the machine learning method kriging (aka Gaussian process regression), DL_FFLUX has access to atomic properties (energy, charge, dipole moment, etc.) with quantum mechanical accuracy. Newly optimized and parallelized using domain decomposition Message Passing Interface (MPI), DL_FFLUX is now able to deliver this rigorous methodology at scale while still in reasonable time frames. DL_FFLUX is delivered as an add-on to the widely distributed molecular dynamics code DL_POLY 4.08. For the systems studied here (103–105 atoms), DL_FFLUX is shown to add minimal computational cost to the standard DL_POLY package. In fact, the optimization of the electrostatics in DL_FFLUX means that, when high-rank multipole moments are enabled, DL_FFLUX is up to 1.25× faster than standard DL_POLY. The parallel DL_FFLUX preserves the quality of the scaling of MPI implementation in standard DL_POLY. For the first time, it is feasible to use the full capability of DL_FFLUX to study systems that are large enough to be of real-world interest. For example, a fully flexible, high-rank polarized (up to and including quadrupole moments) 1 ns simulation of a system of 10 125 atoms (3375 water molecules) takes 30 h (wall time) on 18 cores.

Highlights

A computational chemistry code needs to compete along two axes: accuracy and speed
An Message Passing Interface (MPI) process always binds to a single processor core
The rigorous methodology of FFLUX based on quantum chemical topology has already been shown to be accurate for small systems

Summary

INTRODUCTION

A computational chemistry code needs to compete along two axes: accuracy and speed. Generally, an improvement in one comes at the expense of the other. Molecular flexibility and polarization are achieved without recourse to the usual means of harmonic potentials,[33] Drude oscillators,[34] or other methods.[35,36] in DL_FFLUX, predicted (intramolecular) potential energy surfaces and atomic multipole moments handle molecular flexibility, polarization, and intramolecular charge transfer. All this information is captured by the kriging models trained on an effective paucity of computationally expensive QCT data. The advance presented here is that DL_FFLUX is currently optimized and parallelized such that QCT data are available during an MD simulation at a relatively low extra cost

THEORETICAL BACKGROUND

RESULTS

CONCLUSIONS

■ REFERENCES