Abstract
The ability to use calculated OH frequencies to assign experimentally observed peaks in hydrogen bonded systems hinges on the accuracy of the calculation. Here we test the ability of several commonly employed model chemistries—HF, MP2, and several density functionals paired with the 6‐31+G(d) and 6‐311++G(d,p) basis sets—to calculate the interaction energy (D~e~) and shift in OH stretch fundamental frequency on dimerization (δ(ν)) for the H~2~O → H~2~O, CH~3~OH → H~2~O, and H~2~O → CH~3~OH dimers (where for X → Y, X is the hydrogen bond donor and Y the acceptor). We quantify the error in D~e~ and δ(ν) by comparison to experiment and high level calculation and, using a simple model, evaluate how error in D~e~ propagates to δ(ν). We find that B3LYP and MPWB1K perform best of the density functional methods studied, that their accuracy in calculating δ(ν) is ≈ 30–50 cm^−1^ and that correcting for error in D~e~ does little to heighten agreement between the calculated and experimental δ(ν). Accuracy of calculated δ(ν) is also shown to vary as a function of hydrogen bond donor: while the PBE and TPSS functionals perform best in the calculation of δ(ν) for the CH~3~OH → H~2~O dimer their performance is relatively poor in describing H~2~O → H~2~O and H~2~O → CH~3~OH. © 2009 Wiley Periodicals, Inc. J Comput Chem, 2010