Parallel implementation of efficient charge-charge interaction evaluation scheme in periodic divide-and-conquer...

Area of Science:

• Computational Chemistry
• Materials Science
• Condensed Matter Physics

Background:

• Density-Functional Tight-Binding (DFTB) is a computationally efficient method for electronic structure calculations.
• Scaling to large systems remains a challenge due to the computational cost of calculating interactions between atomic partial charges.
• Existing methods like the Ewald-based technique can be computationally intensive for large-scale simulations.

Purpose of the Study:

• To develop a low-computational-cost algorithm and its parallel implementation for periodic divide-and-conquer DFTB (DC-DFTB) calculations.
• To accelerate the computation of long- and short-range interactions between atomic partial charges.
• To enable efficient and accurate simulations of large-scale material systems.

Main Methods:

• A novel algorithm combining multipole- and interpolation-based approaches for efficient calculation of atomic partial charge interactions.
• Parallel implementation of the DC-DFTB algorithm.
• Assessment of numerical errors (energy and forces) controlled by multipole expansion order, unit cell replication, and interpolation grid size.
• Evaluation of parallel performance on the K computer using various schemes.

Main Results:

• The developed algorithm significantly reduces computational cost for DC-DFTB calculations, particularly for large systems.
• Numerical errors are controllable and comparable to conventional methods.
• Parallel implementation achieved high performance, enabling calculations on systems with over 3 million atoms in minutes.
• The method demonstrated high performance on both supercomputers (K computer) and single-node workstations.
• Feasibility was confirmed for various systems, including liquid water and solid materials like carbon, copper, and sodium chloride.

Conclusions:

• The presented low-computational-cost algorithm and its parallel implementation offer a highly efficient approach for DC-DFTB calculations.
• This method enables rapid and accurate simulations of large-scale atomic and material systems.
• The developed technique is suitable for diverse applications in computational chemistry and materials science.