Reduced-scaling double hybrid DFT with rapid basis set convergence through localized pair natural orbital F12

TL;DR

This paper demonstrates that using PNO-L approximation in double hybrid DFT with F12 explicitly correlated methods significantly reduces computational scaling while maintaining accuracy, enabling efficient calculations on large, diverse chemical datasets.

Contribution

It introduces a reduced-scaling approach for double hybrid DFT with F12 correlation using PNO-L, improving efficiency without sacrificing accuracy.

Findings

Linearized CPU time scaling with PNO-L

Negligible accuracy loss compared to traditional methods

Effective on large, diverse benchmark suite

Abstract

Following earlier work [Mehta, N.; Martin, J. M. L.; J. Chem. Theory Comput. 2022, 18, acs.jctc.2c00426] that showed how the slow basis set convergence of double hybrid density functional theory can be obviated by the use of F12 explicit correlation in the GLPT2 step (second order G\"orling-Levy perturbation theory), we demonstrate here, for the very large and chemically diverse GMTKN55 benchmark suite, and using the B2GP-PLYP-D3BJ functional as a proof of principle, that the CPU time scaling of this step can be reduced (asymptotically linearized) using the PNO-L (pair natural orbitals, localized) approximation, at negligible cost in accuracy.