Abstract
The performance of 24 density functionals, Hartree-Fock, and MP2 is assessed with respect to the CCSD(T)/CBS* energetics of 49 sulfate-water clusters with three to six water molecules. Included among the density functionals are GGA, meta-GGA, hybrid GGA, hybrid meta-GGA, and double hybrid density functionals, as well as the LDA. Three types of dispersion corrections (VV10, XDM, and -D) are tested in conjunction with these functionals. The 26 methods are compared using the relative and binding energies of the sulfate-water clusters as the main criteria. It was discovered that a majority of the tested density functionals are unable to simultaneously capture the physics necessary to describe both the relative and binding energies of the anionic solvation clusters. The three density functionals with the best overall performance are XYG3, omega 1397X-2, and XYGJ-OS. The only other density functional that performs comparably to these three double hybrids is M11. A majority of the density functionals that contain a fraction of exact exchange tend to perform well only for the relative energies, while functionals lacking exact exchange generally perform poorly with respect to both criteria. However, the meta-GGA functional, M11-L, stands out due to its superior performance for the relative energies. While dispersion correction functionals cannot replace the accuracy provided by MP2 correlation, it is shown that the proper combination of a hybrid GGA functional (LC-omega PBE) with a dispersion correction functional (VV10) can lead to drastic improvements in the binding energies of the parent functional, while preserving its performance with respect to the relative energies. Ultimately, however, MP2 has the best overall performance out of the 26 benchmarked methods.