Abstract
Over the last couple of years, it has been shown that Time Dependent Density Functional Theory (TD-DFT) is able to predict accurately and efficiently the polarizability of molecules, when using appropriate exchange-correlation potentials and (large) basis sets. In a previous paper, we compared the accuracy of the predicted mean polarizabilities of 15 organic molecules with experiment, and with two other computational methods: the Restricted Hartree-Fock (RHF) method and the Direct Reaction Field (DRF) approach, the first of which is ignored in this paper. The (empirical) DRF approach however was shown to give comparable accuracies to TD-DFT with the values computed in just a few seconds. In this paper, we use TD-DFT to compute molecular polarizabilities of the twenty amino acid residues, and compare them with the results obtained with the DRF approach. Although the mean absolute deviation of the DRF values from the TD-DFT values is reasonable (7%), it is more than two times the accuracy normally found with the DRF approach. Therefore we decided to optimize the atomic parameters for these systems, and found after optimization, a good agreement with the TD-DFT values (mean absolute deviation 1.0%). As the TD-DFT calculations were necessarily obtained with two additional hydrogens to saturate the backbone bonds, the molecular value of the polarizability of the amino acid residues is overestimated by the TD-DFT calculations. Therefore, the DRF approach (with the newly optimized atomic parameters) has been used to get the actual polarizabilities of the amino acid residues.