Exchange energy density and exchange potential via a Hartree-Fock plus mp2 study of the electron liquid in the ground-state conformer of glycine

Very recent criticisms of existing exchange-correlation functionals by Wanko et al. applied to systems of biological interest have led us to reopen the question of the ground-state conformer of glycine: the simplest amino acid. We immediately show that the global minimum of the Hartree-Fock (HF) ground-state leads to a planar structure of the five non-hydrogenic nuclei, in the non-ionized form NH2-CH2-COOH. This is shown to lie lower in energy than the zwitterion structure NHB3+-CH2-COO-, as required by experiment. Refinement of the nuclear geometry using second-order Mller-Plesset perturbation theory (MP2) is also carried out, and bond lengths are found to accord satisfactorily with experimentally determined values. The ground-state electron density for the MP2 geometry is then redetermined by HF theory and equidensity contours are displayed. The HF first-order density matrix gamma(r, r ') is then used to obtain similar exchange-energy density (epsilon(x)(r)) contours for the lowest conformer of glycine. At first sight, their shape looks almost the same as for the density rho(r), which seems to vindicate the LDA proportional to rho(r)(3/4). However, by way of an analytically soluble model for an atomic ion, it is shown that this has to be corrected to obtain an accurate HF exchange energy E-x as the volume integral of epsilon(x)(r). Finally, recognizing that for larger amino acids, the use of HF plus MP2 perturbation corrections will become prohibitive, we have used the HF information for epsilon(x)(r) and rho(r) to plot the truly non-local exchange potential proposed by Slater, from the density matrix gamma(r, r'). This latter calculation should be practicable for large amino acids, but there adopting Becke's one-parameter form of epsilon(x)(r) correcting LDA exchange. Some future directions are suggested.