Scalable implementation of analytic gradients for second-order Z-averaged perturbation theory using the distributed...

Area of Science:

• Computational Chemistry
• Quantum Mechanics
• Theoretical Chemistry

Background:

• Second-order Z-averaged perturbation theory (ZAPT2) is crucial for calculating molecular properties.
• Efficient parallel implementation of analytic gradients is needed for large-scale computations.
• Existing methods may face challenges with communication costs and convergence.

Purpose of the Study:

• To revise the analytic gradient expression for second-order Z-averaged perturbation theory.
• To describe a detailed parallel implementation of this revised gradient.
• To improve the computational efficiency and scalability of ZAPT2 calculations.

Main Methods:

• The study revises the analytic gradient expression for ZAPT2.
• A parallel implementation is developed using a distributed data interface for molecular-orbital integral arrays.
• The algorithm prioritizes local data access and minimizes communication overhead.
• Iterative solutions and a preconditioner are employed for coupled-perturbed Hartree-Fock (CPHF) equations.

Main Results:

• The revised analytic gradient expression is presented.
• A detailed description of the parallel implementation is provided.
• The algorithm demonstrates efficient use of distributed memory and reduced communication costs.
• Illustrative timing examples showcase the performance of the implementation.

Conclusions:

• The revised analytic gradient and its parallel implementation offer an efficient approach for ZAPT2 calculations.
• The developed algorithm effectively manages distributed data and minimizes communication, leading to improved performance.
• This work contributes to the advancement of computational methods in quantum chemistry.