The performance of the currently used generalized gradient approximation density functionals is analyzed using several simple, yet critical requirements. We analyze the effects of the self-interaction error, the inclusion of the exact exchange, and the parameter settings used in the popular three-parameter hybrid density functionals. The results show that the elimination of the self-interaction error from the current density functionals lead to very poor results for H2. The inclusion of the exact exchange does not significantly influence the self-interaction corrected results. The variation of the A, B, and C parameters of a hybrid DFT method influences the H - H equilibrium bond length through a very simple linear equation, and it is possible to reproduce the experimental H - H distance with appropriate selection of these parameters, although an infinite number of solutions exists. Similar results were obtained for the total energy and the electron density along the internuclear axis. The analysis of the exact KS potential at the bond critical point of the dissociating H2 molecule shows that, for this property, the second order Moller-Plesset perturbation theory yields a better potential than the density functionals studied in this article. © 1997 John Wiley & Sons, Inc.
CITATION STYLE
Csonka, G. I., Nguyen, N. A., & Kolossvary, I. (1997). simple tests for density functional methods. Journal of Computational Chemistry, 18(12), 1534–1545. https://doi.org/10.1002/(SICI)1096-987X(199709)18:12<1534::AID-JCC10>3.0.CO;2-K
Mendeley helps you to discover research relevant for your work.