SIESTA-PEXSI: massively parallel method for efficient and accurate ab initio materials simulation without matrix...

Area of Science:

• Computational Physics
• Materials Science
• Quantum Chemistry

Background:

• Electronic structure calculations are crucial for understanding material properties.
• Traditional methods like matrix diagonalization struggle with large-scale systems.
• Kohn-Sham Density Functional Theory (KSDFT) is a widely used framework.

Purpose of the Study:

• To develop an efficient and scalable method for large-scale electronic structure calculations.
• To combine the Pole Expansion and Selected Inversion (PEXSI) technique with the SIESTA method.
• To evaluate the performance and accuracy of the combined SIESTA-PEXSI method.

Main Methods:

• Integration of the PEXSI technique with the SIESTA method, which employs numerical atomic orbitals within KSDFT.
• Leveraging the sparsity of Hamiltonian and overlap matrices for computational efficiency.
• Utilizing massively parallel PEXSI for high-performance computing environments.

Main Results:

• The SIESTA-PEXSI method demonstrates significantly lower computational complexity for large systems compared to matrix diagonalization.
• The method accurately computes electron density, free energy, atomic forces, and density of states without eigenvalue/eigenvector computation.
• Achieved accuracy is comparable to traditional methods, even for metallic systems at low temperatures.
• Demonstrated high scalability, efficiently utilizing over 10,000 processors.

Conclusions:

• The SIESTA-PEXSI method provides an accurate, efficient, and highly scalable approach for large-scale electronic structure calculations.
• This method is suitable for diverse systems, including 1D, 2D, and bulk materials with various electronic characteristics.
• The technique offers a powerful alternative to conventional diagonalization for complex material simulations.