Random Phase Approximation Correlation Energy Using Real-Space Density Functional Perturbation Theory.

Area of Science:

• Computational Physics
• Quantum Chemistry
• Materials Science

Background:

• Kohn-Sham density functional theory (KS-DFT) is a cornerstone for electronic structure calculations.
• Calculating the random phase approximation (RPA) correlation energy is computationally demanding, especially for large systems.
• Existing methods often struggle with scalability and efficiency.

Purpose of the Study:

• To develop a novel real-space method for computing RPA correlation energy.
• To leverage the low-rank properties of the density response operator for computational efficiency.
• To create a scalable and accurate computational framework for electronic structure calculations.

Main Methods:

• A cubic-scaling formalism based on density functional perturbation theory.
• Circumventing direct calculation of the response function matrix via Sternheimer linear systems.
• Utilizing subspace iteration, spectral quadrature, Kronecker product methods, and conjugate orthogonal conjugate gradient solvers.
• A large-scale parallel implementation for high-performance computing.

Main Results:

• Demonstrated convergence with respect to key computational parameters.
• Verified accuracy by comparison with established plane-wave methods.
• Achieved excellent strong scaling on thousands of processors.
• Reduced computation time for a 128-electron system to approximately 150 seconds on 4608 processors.

Conclusions:

• The developed real-space method provides an efficient and accurate approach for RPA correlation energy calculations.
• The parallel implementation demonstrates significant scalability for large electronic systems.
• This framework offers a promising tool for advanced materials and molecular simulations within KS-DFT.