Stochastic tensor contraction for quantum chemistry

Area of Science:

• Computational quantum chemistry
• High-performance computing
• Theoretical chemistry

Background:

• Ab initio quantum chemistry methods rely on high-order tensor contractions.
• The computational cost of these contractions limits the size of chemical systems that can be studied.
• Existing methods face a trade-off between accuracy and computational expense.

Purpose of the Study:

• To introduce stochastic tensor contraction as a method to reduce computational cost in quantum chemistry.
• To apply this method to coupled cluster theory, a gold-standard quantum chemistry technique.
• To challenge the established cost-to-accuracy landscape in computational chemistry.

Main Methods:

• Development and application of stochastic tensor contraction algorithms.
• Implementation within coupled cluster theory, including perturbative triples.
• Benchmarking against state-of-the-art local correlation approximations.

Main Results:

• Achieved significantly reduced computational scaling, approaching that of exact mean-field theory for stringent energy errors.
• Demonstrated performance close to practical mean-field absolute cost.
• Obtained an order-of-magnitude improvement in computation time and error compared to existing methods.
• Showed reduced sensitivity to system dimensionality and electron delocalization.

Conclusions:

• Stochastic tensor contraction is a powerful computational primitive for accelerating quantum chemistry calculations.
• This method offers a new paradigm for achieving high accuracy at reduced computational cost.
• It has broad applicability across various quantum chemistry methods and problems.