The Effective Fragment Molecular Orbital Method: Achieving High Scalability and Accuracy for Large Systems

Area of Science:

• Computational Chemistry
• Quantum Mechanics
• Materials Science

Background:

• The effective fragment molecular orbital (EFMO) method enables accurate energy predictions for large molecular systems.
• Accurate prediction relies on modeling intermolecular interactions, including many-body polarization and dispersion effects.
• Solving coupled perturbed Hartree-Fock (CPHF) and time-dependent Hartree-Fock (TDHF) equations is computationally intensive and a bottleneck in EFMO.

Purpose of the Study:

• To enhance the efficiency and scalability of the EFMO method.
• To improve the computational bottleneck associated with CPHF and TDHF equation solving.
• To demonstrate the accuracy and applicability of the improved EFMO method for large-scale systems.

Main Methods:

• Developed a new CPU memory-based implementation for CPHF and TDHF solvers.
• Parallelized solvers using Message Passing Interface (MPI) or hybrid MPI/OpenMP.
• Systematically examined the effects of basis sets and the cutoff parameter (R_cut) on accuracy.

Main Results:

• Significantly improved efficiency and scalability of the EFMO method.
• Achieved excellent accuracy (<1 kcal/mol error per fragment) with optimized basis sets and R_cut.
• Demonstrated near-ideal strong scaling for CPHF and TDHF calculations with increased MPI ranks.
• Successfully performed large-scale EFMO calculations on a hydrated nanoparticle (>15k atoms) and a nanoscale chemical reaction.

Conclusions:

• The new parallel implementation substantially reduces EFMO computational time.
• The improved EFMO method accurately models complex systems with explicit solvent effects.
• This advancement enables unprecedented large-scale quantum mechanical calculations for chemical and materials science problems.