Enabling Large-Scale Condensed-Phase Hybrid Density Functional Theory Based Ab Initio Molecular Dynamics. 1. Theory,...

Area of Science:

• Computational Chemistry and Materials Science
• Quantum Mechanics and Electronic Structure Theory

Background:

• Hybrid functionals in Density Functional Theory (DFT) improve accuracy by including exact exchange (EXX), reducing self-interaction error.
• High computational cost of EXX calculations limits hybrid DFT's application to large molecules and condensed-phase systems.
• Existing methods struggle with the scalability needed for complex, large-scale simulations.

Purpose of the Study:

• To develop and present a linear-scaling approach for hybrid DFT calculations.
• To enable accurate ab initio molecular dynamics (AIMD) simulations of large condensed-phase systems.
• To overcome the computational bottlenecks associated with exact exchange evaluation.

Main Methods:

• Utilized a local representation of occupied orbitals, specifically maximally localized Wannier functions (MLWFs).
• Exploited sparsity in real-space evaluation of quantum mechanical exchange interaction for finite-gap systems.
• Integrated the MLWF-based approach into the Car-Parrinello AIMD framework, using MLWF-product potentials for EXX energy and forces.
• Implemented an efficient algorithm in the Quantum ESPRESSO program with hybrid MPI/OpenMP parallelization for High-Performance Computing (HPC).

Main Results:

• Demonstrated the feasibility of MLWF-based AIMD simulations for large condensed-phase systems (500-1000 atoms) at the hybrid DFT level.
• Achieved wall time costs comparable to semilocal DFT for simulations of liquid water ((H2O)256).
• Showcased excellent strong and weak scaling performance on modern HPC architectures.

Conclusions:

• The developed linear-scaling MLWF approach significantly reduces the computational cost of hybrid DFT calculations.
• This breakthrough enables routine AIMD simulations of large and complex condensed-phase systems at a high level of theory.
• The method brings routine, long-timescale simulations of complex materials closer to reality.