Optimizing Computational Parameters for Nuclear Electronic Orbital Density Functional Theory: A Benchmark Study on Proton Affinities.

This study benchmarks the nuclear electronic orbital density functional theory (NEO-DFT) method for a set of molecules that is larger than in previous studies. The focus is on proton affinity predictions to assess the influences of computational parameters. NEO-DFT incorporates nuclear quantum effects for protons involved in protonation processes. Using a test set of 72 molecules with experimental proton affinities as reference, we evaluated various exchange-correlation functionals, finding that B3LYP-based functionals deliver the most accurate results. Among the tested functionals, CAM-B3LYP performs the best with an MAD value of 6.2 kJ/mol with respect to experimental data. In NEO-DFT, electron-proton correlation (epc) functionals were assessed, with LDA-type epc17-2 yielding comparable results to the GGA-type epc19 functional. Compared to traditional DFT (MAD value of 31.6 kJ/mol), which treats nuclei classically, NEO-DFT provides enhanced accuracy for proton affinities when electron-proton correlation is included. Regarding basis sets, the def2-QZVP electronic basis set achieved the highest accuracy with an MAD value of 5.0 kJ/mol, though at a higher computational cost compared to def2-TZVP and def2-SVP, while nuclear basis sets showed minimal impact on proton affinity accuracy and no consistent trend. Overall, this study demonstrates NEO-DFT's efficacy in addressing nuclear quantum effects for proton affinity predictions, providing guidance on optimal parameter selection for future NEO-DFT applications.