Reducing Functional- and Density-Driven Errors with r²SCANX@r²SCANY

The self-consistent density obtained from a density functional approximation (DFA) is often reasonably accurate, with the total DFA error mainly arising from the functional-driven component. However, in systems where the DFA density is poor, the density-driven error can grow large compared to the functional-driven error. In such cases, evaluating a DFA on a Hartree–Fock (HF) density (DFA@HF) can substantially reduce the total error, suggesting that the HF density is more accurate than the self-consistent DFA density. Interestingly, recent studies^2,3 have shown that DFA@HF can yield improved results even when the functional-driven error dominates, due to fortuitous cancellation between functional- and density-driven errors. Recently, a hybrid approach⁴—r²SCANX@r²SCANY—based on the regularized-restored strongly constrained and appropriately normed (r²SCAN) functional, has been proposed to simultaneously reduce both sources of error. In this work, we apply this method to a diverse set of molecular properties and demonstrate improvements in accuracy relative to r²SCAN.

 

1. 1. Min-Cheol Kim, Eunji Sim and Kieron Burke, Phys. Rev. Lett. 111, 073003 (2013)
   
   2. B. Kanungo, A. D. Kaplan, C. Shahi, V. Gavini, AND J. P. Perdew, J. Phys. Chem. Lett. 15, 323 (2024)
   
   3.  A. D. Kaplan, C. Shahi, R. K. Sah, P. Bhetwal, B. Kanungo, V. Gavini, and J. P. Perdew, J. Chem. Theory Comput. 20, 5517 532 (2024)
   
   4. H. R. Gopidi, R. Zhang, Y. Wang, A. Patra, J. Sun, A. Ruzsinszky, J. P. Perdew, and P. Canepa, arXiv preprint arXiv: 2506.20635 (2025)