DL_FFLUX: A Parallel, Quantum Chemical Topology Force Field

Area of Science:

• Computational Chemistry
• Materials Science
• Molecular Dynamics

Background:

• Molecular dynamics (MD) simulations are crucial for understanding molecular behavior.
• Accurate representation of molecular properties, especially polarizability, is essential for reliable simulations.
• Existing MD force fields often struggle with capturing the nuances of flexible molecules and polarizable systems.

Purpose of the Study:

• To introduce DL_FFLUX, a new force field for molecular dynamics simulations.
• To enable accurate calculations of atomic properties for flexible molecules using quantum chemical topology.
• To demonstrate the scalability and efficiency of DL_FFLUX for large-scale simulations.

Main Methods:

• Utilizing quantum chemical topology to derive atomic properties.
• Employing kriging (Gaussian process regression) for machine learning-based property prediction.
• Implementing domain decomposition Message Passing Interface (MPI) for parallelization.
• Integrating DL_FFLUX as an add-on to the DL_POLY 4.08 molecular dynamics code.

Main Results:

• DL_FFLUX achieves quantum mechanical accuracy for atomic properties like energy and charge.
• The parallelized DL_FFLUX shows minimal computational cost increase compared to standard DL_POLY.
• DL_FFLUX can be up to 1.25× faster than standard DL_POLY when high-rank multipole moments are enabled.
• Scalability of MPI implementation is preserved in parallel DL_FFLUX.

Conclusions:

• DL_FFLUX provides a computationally efficient and accurate method for molecular dynamics simulations of flexible, polarizable molecules.
• The developed force field makes large-scale, high-fidelity molecular simulations feasible for the first time.
• This advancement opens possibilities for studying complex, real-world molecular systems with unprecedented detail.