Hydrogen bonding described using dispersion-corrected density functional theory

In recent works, dispersion-corrected atom-centered potentials (DCACPs) were developed as a method to account for long-range dispersion forces between molecules in density functional theory calculations within the generalized gradient approximation (GGA). Here, we test the ability of DCACPs to improve the GGA treatment of hydrogen-bonded systems. We assessed both BLYP and dispersion-corrected BLYP with respect to benchmark calculations for the hydrogen bond lengths and binding energies of 20 complexes containing the elements C, H, N, O, and S. Benchmark data included geometries calculated using MP2 and CCSD(T) and binding energies using W2, W1, CBS-QB3, and other CCSD(T) extrapolation schemes. With respect to benchmark methods, dispersion-corrected BLYP exhibited a mean signed error of 0.010 Å in the hydrogen bond length and a mean relative error of 5.1% in the hydrogen bond binding energy. By comparison, uncorrected BLYP exhibited error statistics of 0.036 Å and 15.9%, respectively. We conclude that DCACPs robustly improve the BLYP description of hydrogen-bonded systems at small additional computational cost. New benchmark geometries (MP2/aug-cc-pVTZ) and new benchmark binding energies (W1) are presented for seven complexes, and the remaining benchmark data were taken from previous literature.