DFT Functionals Benchmarking via Hellmann-Feynman Forces Evaluation

Computational methods based on density functional theory (DFT) have become an indispensable tool in many fields of chemistry due to their excellent cost-to-accuracy ratio. One of the main challenges in the development of DFT methods is to find a good approximation for the exchange-correlation functional. Currently, diverse approximations demonstrate optimal performance in replicating distinct characteristics, making it challenging to choose the appropriate functional for routine calculations. The promising class of highly parameterized functionals can deliver excellent results in reproducing specific properties but may exhibit non-physical behavior that is difficult to capture. Here, we provide a vivid example of these issues: we show how some highly parameterized functionals exploit basis set incompleteness to describe London dispersion forces. The potential for functionals to exhibit such behavior is unintuitive, making it difficult to trust other functionals without explicit verification. Although most highly parameterized functionals did not use this opportunity explicitly, they had significant basis set incompleteness errors (BSIE), so they could potentially exploit this in a way that is not immediately apparent. It seems that the approach of functionals to the non-physical application of basis set incompleteness and other artifacts requires closer attention.