Factorized Quadruples and a Predictor of Higher-Level Correlation in Thermochemistry

Area of Science:

• Computational Chemistry
• Quantum Chemistry
• Theoretical Chemistry

Background:

• Coupled cluster theory, particularly CCSD(T), is crucial for accurate ab initio molecular property prediction.
• Higher-order excitations (quadruples) are computationally expensive, scaling as O(N^6) or worse.
• Predicting the necessity of post-CCSD(T) corrections for reaction energies is challenging.

Purpose of the Study:

• To investigate cost-effective methods for incorporating quadruple excitation corrections.
• To evaluate the accuracy of factorized coupled cluster methods and new estimation techniques for reaction energies.

Main Methods:

• Employed the factorized CCSD(TQf) method, a computationally cheaper alternative to CCSD(TQ).
• Assessed Goodson's continued fraction method for estimating CCSDT(Q)Λ contributions.
• Developed and tested a scaled perturbation estimator based on CCSD(TQf)/cc-pVDZ.

Main Results:

• The factorized CCSD(TQf) method showed minimal error compared to CCSD(TQ).
• The scaled perturbation estimator accurately predicted CCSDT(Q)Λ contributions with an average error of 0.07 kcal/mol for ~3000 reactions.
• This estimator achieves high accuracy at a computational cost no greater than CCSD(T).

Conclusions:

• Factorized coupled cluster methods offer a viable path to reduce the cost of high-accuracy calculations.
• The scaled perturbation estimator provides a reliable and computationally inexpensive way to estimate post-CCSD(T) effects on reaction energies.
• This approach enables efficient assessment of the importance of higher-order correlation effects in thermochemical predictions.