Massively parallel linear-scaling Hartree-Fock exchange and hybrid exchange-correlation functionals with plane wave...

Area of Science:

• Computational Chemistry
• Materials Science
• Quantum Mechanics

Background:

• Accurate calculation of Hartree-Fock exchange energy is crucial for electronic structure methods.
• Existing methods struggle with computational cost for large systems.
• Linear-scaling methods are needed to overcome these limitations.

Purpose of the Study:

• To extend a linear-scaling approach for Hartree-Fock exchange energy calculation.
• To leverage massive parallelism for computational efficiency.
• To implement and validate the approach within the ONETEP framework.

Main Methods:

• Utilized localized in situ optimized orbitals and non-orthogonal generalized Wannier functions (NGWFs).
• Employed a resolution-of-identity approach with truncated spherical waves (SWs) for exchange calculations.
• Combined distributed (MPI) and shared (OpenMP) memory parallelism for efficient CPU core utilization.

Main Results:

• Achieved linear scaling of computational cost with system size.
• Demonstrated superlinear strong parallel scaling in many cases due to time-memory trade-offs.
• Maintained linear scaling and graceful performance degradation even with limited memory.

Conclusions:

• The new implementation efficiently calculates Hartree-Fock exchange energy for large systems.
• The approach is practical for production calculations, as shown with imogolite nanotubes (>1400 atoms).
• The study analyzed the impact of approximations (exchange cutoff, SW basis) on accuracy and performance.