A DENSITY-FUNCTIONAL STUDY OF THE GLYCINE MOLECULE - COMPARISON WITH POST-HARTREE-FOCK CALCULATIONS AND EXPERIMENT

The potential energy surface of un-ionized glycine has been explored w ith density functional. theory. The performance of several nonlocal fu nctionals has been evaluated and the results are presented in the cont ext of available experimental information and post-Hartree-Fock quantu m chemical results. The zero-point and thermal vibrational energies al ong with vibrational entropies play a very important role in determini ng the relative stability of glycine conformers; the realization of th is has led to some revision and reinterpretation of the experimental r esults. Uncertainties in the vibrational contributions to the energy d ifferences of several tenths of a kilocalorie/mole remain. The uncerta inty in the vibrational free energy is even larger, about 1 kcal/mol. In the final analysis, we suggest that the best estimate of the electr onic energy difference between the two lowest glycine conformers shoul d be revised downward from 1.4 to 1.0 kcal/mol. Thirteen stationary po ints on the potential energy surface have been localized. For the majo rity of these, there is close agreement among various nonlocal density functionals and the post-Hartree-Fock methods. However, the second co nformer (IIn), which has a strong hydrogen bond between the hydroxyl h ydrogen and the nitrogen of the amine group, presents a distinct chall enge. The relative energy of this conformer is extremely sensitive to the basis set, the level of correlation, or the functional used. The w idely used BP86, PP86, and BP91 nonlocal functionals overestimate the strength of the hydrogen bond and predict that this conformer is the l owest energy structure. This contradicts both experiment and high-leve l post-Hartree-Fock studies. The adiabatic connection method (ACM) and the BLYP functional yield the correct order. The ACM method, in parti cular, gives energies which are in reasonable agreement with MP2, alth ough these are somewhat low as compared with experiment. Based on this study, ACM should perform well for this type of bioorganic applicatio n, with typical. errors of a few tenths of a kilocalorie/mole and only rarely exceeding 0.5 kcal/mol. (C) 1997 John Wiley & Sons, Inc.