Correcting density functionals for dispersion interactions using pseudopotentials

Noncovalent Interactions in Density Functional Theory

“…However, keeping in mind that only one DCACP function was used for each atom and that the fitting data used to generate the functions were very small, the performance is reasonable. The recent work of Karalti et al 253 convincingly demonstrates that using two functions per atom can offer much improved performance in the treatment of noncovalent interactions, with the MAE reduced by almost a factor of 2 over that obtained with the single function DCACPs. Table VI summarizes the performance of various dispersion-corrected DFT methods on the larger S66 set.…”

“…99 DCACP: DFT-DCACP2/plane wave. 253 DCP: B3LYP-DCP/6-31+G(2d,2p) 234 and LC-ωPBE/6-31+G(2d,2p). 243 MN: Minnesota/def2-QZVP.…”

“…252 Recently, Jordan's group found that DCACPs containing two functions offer much better performance than the single function approach. 253 The advantages of DCAPCs are analogous to those of DCPs. For example, DCACPs can be used with plane wave computational packages without the need for reprogramming and are able to take advantage of the full machinery of such packages, including ab initio molecular dynamics.…”

Dispersion-correcting potentials can significantly improve the bond dissociation enthalpies and noncovalent binding energies predicted by density-functional theory

“…The approach was motivated by our earlier work on quantum capping potentials, 24 which are atomcentered potentials that were designed to bridge the quantum and classical regions in QM/MM simulations. DCPs are similar in spirit to a plane wave-based technique introduced by von Lilienfeld and co-workers, [25][26][27] which were recently extended by Karalti et al 28 DCPs are generated by optimizing the exponents and coefficients of a set of Gaussian functions to minimize the error in the calculated binding energies (BEs) associated with the potential energy surfaces (PESs) of fitting sets composed of noncovalently interacting dimers. We demonstrated that the DCP approach can be combined with the B3LYP 29, 30 functional and 6-31+G(2d,2p) basis sets to allow for the very accurate prediction structures and BEs for a wide variety of non-covalently interacting dimer systems containing the H, C, N, and O atoms.…”

Ab initio-enabled phase transition prediction of solid carbon dioxide at ultra-high temperatures

“…In recent decades, great advances have been made in electronic structure methods for predicting molecular structure and crystal energy. 19,20 Density functional theory (DFT) [21][22][23][24] has been the standard method for handling solids and liquids due to its efficiency. However, its accuracy depends on the choice of density functionals, and there is currently no general procedure to systematically improve existing density functionals.…”