New implementation of high-level correlated methods using a general block tensor library for high-performance...

Area of Science:

• Computational Chemistry
• Scientific Computing

Background:

• Post-Hartree–Fock methods require efficient tensor algebra capabilities.
• Existing libraries may lack support for arbitrary tensor dimensions, sizes, and symmetries.

Purpose of the Study:

• To present an open-source, object-oriented C++ library for performing tensor algebra.
• To facilitate advanced electronic structure calculations and other scientific computations.

Main Methods:

• Implementation of data structures and algorithms for handling large tensors by block splitting.
• Utilizing both core memory and disk storage for large tensor operations.
• Application of divide-and-conquer parallel algorithms for tensor algebra.

Main Results:

• Support for tensors of arbitrary order, size, and symmetry.
• Inclusion of general tensor symmetry algorithms and specific electronic structure symmetries (permutational, spin, molecular point group).
• Successful integration into Q-Chem software for driving advanced electronic structure methods.

Conclusions:

• The library provides a versatile and efficient tool for tensor algebra in computational science.
• Its design enables scalability and applicability across various scientific domains.
• It enhances the capabilities of electronic structure software like Q-Chem.