Efficient Parallel Implementation of the CCSD External Exchange Operator and the Perturbative Triples (T) Energy...

Area of Science:

• Computational Chemistry
• Quantum Chemistry
• High-Performance Computing

Background:

• Coupled cluster singles and doubles (CCSD) is a cornerstone of accurate quantum chemical calculations.
• The external exchange operator (EEO) represents a significant computational bottleneck in CCSD energy calculations.
• Previous algorithms for EEO were limited by input/output operations and suboptimal utilization of integral screening and sparsity.

Purpose of the Study:

• To develop and present a novel, efficient parallel algorithm for the external exchange operator (EEO) in CCSD calculations.
• To improve computational efficiency by reducing input/output requirements and enhancing integral screening.
• To enable the simultaneous application of dense matrix multiplication and sparsity utilization for accelerated CCSD procedures.

Main Methods:

• Formulation of the EEO calculation as a series of matrix multiplications.
• Decomposition of atomic orbital integrals and CC coefficients into smaller blocks to reduce memory footprint.
• Presorting of integrals to regularize sparsity patterns, facilitating efficient dense matrix routines and sparsity exploitation.
• Parallel implementation of the perturbative triples correction to CCSD using the Array Files tool for distributed file systems.

Main Results:

• The new parallel algorithm demonstrates significantly reduced input/output compared to previous methods.
• The implementation effectively combines dense matrix multiplication and sparsity utilization for substantial speedup.
• Representative timings confirm efficiency for calculations involving a wide range of atomic orbitals and correlated electrons across various symmetries.

Conclusions:

• The presented parallel algorithm offers a more efficient approach to a critical step in CCSD energy calculations.
• The method's design allows for effective parallelization without compromising computational performance.
• This work provides a valuable tool for accelerating quantum chemical simulations, particularly for large systems.