Local Second Order Mo̷ller-Plesset Theory with a Single Threshold Using Orthogonal Virtual Orbitals: A Distributed...

Area of Science:

• Computational chemistry
• Quantum chemistry
• High-performance computing

Background:

• Superlinear computational scaling in traditional electron correlation methods poses challenges for large molecular systems.
• Local correlation methods offer a solution by approximating small contributions as zero, maintaining accuracy.
• Parallel computing enhances the efficiency and scalability of these methods.

Purpose of the Study:

• To implement a distributed memory parallel local correlation method for second-order Moller-Plesset (MP2) theory.
• To achieve efficient and scalable computation of electron correlation energies for large molecules.
• To validate the method's performance through numerical experiments.

Main Methods:

• Developed a distributed memory parallel implementation using the Message Passing Interface (MPI).
• Employed a single threshold for term dropping and accuracy control across computing kernels.
• Utilized a fixed sparsity pattern matrix multiplication and a distributed conjugate gradient solver.
• Implemented a process partitioning strategy tailored for distributed systems.

Main Results:

• The parallel MP2 implementation exhibits near-linear scaling in both strong and weak analyses.
• Demonstrated high accuracy (0.003%) with a computationally feasible runtime for complex molecules like vancomycin.
• Successfully calculated correlation energy for vancomycin in def2-TZVP basis within 30 minutes using 32 MPI ranks.
• Showcased significant performance gains over sequential or shared-memory implementations.

Conclusions:

• The developed distributed parallel local MP2 method overcomes computational limitations of traditional approaches.
• This technique offers a scalable and accurate solution for electronic structure calculations of large molecular systems.
• The implementation provides a practical tool for researchers in computational and quantum chemistry.