Résumé:
Selecting a method to compute partial atomic charges is not trivial because different methods usually provide different charge values and there is no consensus on the most useful approach. In this work, Mulliken, MBS, Chelp, Chelpg, MK, Hirshfeld, NPA, DMA and AIM methods were selected to compute atomic charges and electric dipole moment vector of N-{N-(Pterin-7-yl)carbonylglycyl}-L-tyrosine molecule, a ricin inhibitor which has different types of bonds and chemical environments. While MBS and DMA methods provided the most chemically consistent charges according to atomic electronegativity and electron resonance effects criteria, Chelp, Chelpg and MK had the worst performances. Atomic charges and dipole moment calculated by the Hirshfeld method had the smallest magnitudes, a well-known behavior. Despite the differences among atomic charges predicted by all methods, the direction of the dipole moment vector was essentially the same. Further charge calculations using different basis sets and quantum methods indicated that the dependency on this aspect was the highest for Mulliken and Chelp and the lowest for MBS, Hirshfeld and DMA methods. Thus, results point to MBS and DMA as the most suitable methods for computing chemically consistent atomic charges and dipole moment vectors of similar systems for different applications; e.g., molecular dynamics.
