Evaluation of MP2, DFT, and DFT-D Methods for the Prediction of Infrared Spectra of Peptides

The prediction accuracy of second-order Moller−Plesset theory MP2 and density functional theory DFT-(D) with and without empirical dispersion correction within the resolution of identity approximation (ri) have been investigated for the assignment of infrared spectra of gas-phase peptides. A training set of 17 conformers of phenylalanine containing capped peptides have been used to establish mode-specific local scaling factors. Inclusion of dispersion terms at the DFT level turns to significantly improve the accuracy of predicted IR spectra. At the DFT-D level, the nonhybrid generalized gradient approximation functional B97-D (TZVP basis set) provides even better results than the popular hybrid functional B3LYP (6-31+G* basis set) while reducing the computational cost by almost 1 order of magnitude. Besides, MP2 (SVP basis set) outperforms all other tested methods in terms of reliability and transferability to larger molecular systems with typical prediction errors of about 5 cm−1.