No abstract

A method of "natural population analysis" has been developed to calculate atomic charges and orbital populations of molecular wave functions in general atomic orbital basis sets. The natural analysis is an alternative to conventional Mulliken population analysis, and seems to exhibit improved numerical stability and to better describe the electron distribution in compounds of high ionic character, such as those containing metal atoms. We calculated ab initio SCF-MO wave functions for compounds of type CH 3 X and LiX (X = F, OH, NH 2 , CH 3 , BH 2 , BeH, Li, H) in a variety of basis sets to illustrate the generality of the method, and to compare the natural populations with results of Mulliken analysis, density integration, and empirical measures of ionic character. Natural populations are found to give a satisfactory description of these molecules, providing a unified treatment of covalent and extreme ionic limits at modest computational cost.

We have carried out a natural bond orbital analysis of hydrogen bonding in the water dimer for the near-Hartree–Fock wave function of Popkie, Kistenmacher, and Clementi, extending previous studies based on smaller basis sets and less realistic geometry. We find that interactions which may properly be described as ‘‘charge transfer’’ (particularly the n-σ*OH interaction along the H-bond axis) play a critical role in the formation of the hydrogen bond, and without these interactions the water dimer would be 3–5 kcal/mol repulsive at the observed equilibrium distance. We discuss this result in relationship to Klemperer’s general picture of the bonding in van der Waals molecules, and to previous theoretical analyses of hydrogen bonding by the method of Kitaura and Morokuma.

The method of natural localized molecular orbitals (NLMOs) is presented as a novel and efficient technique for obtaining LMOs for SCF and CI wave functions. It is an extension of the previously developed natural atomic orbital (NAO) and natural bond orbital (NBO) methods, and uses only the information contained in the one-particle density matrix. Results are presented for methane and cytosine to indicate that NLMOs closely resemble LMOs obtained by the Boys and Edmiston–Ruedenberg methods, with the exception that the NLMO procedure automatically preserves the MO σ–π separation in planar molecules. The computation time is modest, generally only a small fraction of the SCF computation time. In addition, the derivation of NLMOs from NBOs gives direct insight into the nature of the LMO ‘‘delocalization tails,’’ thus enhancing the role of LMOs as a bridge between chemical intuition and molecular wave functions.

No abstract