ENERGIES OF SIGMA-STAR ORBITALS

In order to find the energies of sigma* orbitals we have parametrized a semiempirical method (CNDO) to give orbital energies for occupied and unoccupied orbitals, which can be interpreted as negative ionization energies and electron affinities, respectively. Experimental ionization energies, electron affinities, and UV excitations have been used for the parametrization. During the parametrization procedure, an intermediate version, based on a small number of electron affinities with reliable interpretations, was used to calculate UV spectra. It appeared that a number of previously unassigned UV bands could be understood as valence transitions. In this way the basis for the parametrization was extended to include a number of newly assigned states. With the final optimized parameters we obtained reasonable agreement between calculated and experimental energies of several kinds of electronic states for many molecules. It is therefore probable that the energies of sigma* orbitals given in this paper are reasonably accurate and useful for continued work. In a small number of cases there are discrepancies which we cannot explain. Our results for electron affinities are partly unexpected. In hydrocarbons (e.g., ethane) the sigma* orbitals are very low, and it is unexpected that they are not observed in electron transmission spectroscopy. In other saturated hydrocarbons (e.g., cyclopropane) broad bands are seen in such studies. Tentative explanations are given based on the present calculations. In order to further investigate the utility of the method it has also been used for calculation of core excitation, where valence type final states dominate. Reasonable agreement is obtained for a large number of experimental bands in these spectra, although the calculated energies for the higher energy sigma* orbitals are consistently underestimated. Previously, the importance of sigma* orbitals was not realized, and early interpretations of core spectra predominantly invoked Rydberg orbitals. Spectra for which reassignments have been made based on the present calculations are discussed in some detail. A general result of our work is that many processes which previously have been explained by use of Rydberg orbitals are now suggested to involve unoccupied valence orbitals. Another result is that the configuration interaction matrix elements may be very strongly dependent upon correlation. The findings will probably be important in the theory of chemical reactions. They will also be useful in a future development of a better semiempirical MO theory.