Understanding the Success of Density-Corrected DFT for Dipole Moments

The framework of Density Corrected DFT introduced allows for rigorous decomposition of errors in DFT calculations into Density-Driven and functional-driven errors. In cases where the density driven error dominates, significant error reduction can be achieved by evaluating the energy functional on a non self-consistent density. In principle, the substitution of an improved density will reduce the error of the calculation. Hartree-Fock densities have become widely used in this context, and careful use of HF densities can lead to significantly improved performance for cases where the density-driven error is dominant. In recent literature, it has been argued that the success of DC-DFT for certain systems is due to an accidental cancellation of errors as an artifact of HF densities. Focusing on molecular dipole moments, we present numerical and analytical results to clarify that density corrections indeed systematically reduce errors. We discuss the difficulties in measuring the quality of approximate densities, and emphasize that the framework of DC-DFT is a general theory, and is not tied to the specific choice of HF densities.