On the Xα Local Spin Density Approximation in the Study of Organic Molecules

Abstract

While the familiar wave function methods meet with considerable success in numerous applications to chemistry, they involve formidable computational demands in any evaluation of thermochemical data at a level approaching chemical accuracy. By far the most useful non-empirical alternatives to configuration interaction calculations are the methods rooted in density functional theory (DFT). Building on their success in applications to inorganic chemistry, we examine here the applicability of DFT to the “simpler” organic systems, with the hope of retrieving good-quality thermochemical information (atomization energies, enthalpies, and the like). Namely, we examine here the local spin density (LSD) approximation and treat exchange and correlation in the Xα approach where α is a variable parameter. The tentative hypothesis is that the Xα(LSD) method basically meets the required demands of accuracy, i.e., that the energy should be right, if α is properly selected. This is where the problem lies: ever since the appearance of the Xα method, the way of selecting α has been a fundamental problem. On the basis of regularities which are observed for exchange-correlation energies, we develop a recipe permitting valid estimates of α. The at least approximate validity of this procedure is illustrated by the accuracy of calculated atomization energies, thus paving the way to accurate evaluations of dissociation energies. Examples include alkanes, alkyl radicals, amines and chloroalkanes, as well as weaker interactions (hydrogen bonds), showing that the theoretical results obtained in this approach are well within experimental accuracy. Bond dissociation energies in compounds N_XH_y were also investigated, with a direct application to the explosive decomposition of HN₃.