Correlation-Driven Spin-Component-Scaled Second-Order Møller–Plesset Perturbation Theory (CD-SCS-MP2)

Møller–Plesset second-order perturbation theory (MP2) is one of the most popular and successful methods in computational chemistry, but it is not without disadvantages. It fails to capture nondynamic correlation, overestimates dispersion interactions in strongly polarizable systems, and inaccurately describes delocalized molecules. Spin-component scaling techniques improve MP2 energies by compensating for the fact that, in general, opposite-spin correlation plays a significantly greater role than same-spin correlation. On average, SCS-MP2 improves the reaction energies of small organic molecules, vibrational frequencies, thermodynamic properties, and π-stacking interactions; however, the optimal scaling values are known to be system-dependent, resulting in multiple SCS-MP2 methods. In this work, we propose improving the accuracy of SCS-MP2 by scaling the opposite-spin correlation according to the amount of dynamic correlation as measured from recently developed correlation indices that depend on the natural orbital occupations. In this way, the method is correlation-driven and can effectively adapt to the system-specific nature of spin-scaling factors. The correlation-driven SCS-MP2 (CD-SCS-MP2) method adds a negligible cost to the MP2 calculation and provides results superior to those obtained from SCS-MP2.