A new method for the accurate a priori calculation of atomic and molecular energies is proposed. The The description of the electronic structure of atoms and molecules in terms of one electron orbitals provides a simple and therefore useful starting point for any more detailed theory. The energy obtained from calculations based upon orbitals is generally in error by about 0.015 au (10 kcal/ mole) * for each electron. Although these errors comprise a small fraction of the total energy, they are quite large by chemical standards. It is, therefore, necessary to go beyond an orbital picture to obtain a chemically adequate description of electronic structure.The errors inherent in an orbital model can be accounted for in terms of the correlated motion of the electron pairs (1). The novel feature of the new method is that the correlation correction is determined in an a priori fashion directly from the orbitals. This allows us to treat much larger systems than we could by calculating correlation energies using an ab initio variational approach, but maintains our ability to predict these correlation energies as functions of molecular geometry. The latter is absolutely necessary if we are to examine potential energy surfaces for chemical reactions. It is with an eye towards the study of transition states and reaction pathways that the new method has been developed.
GENERALIZED VALENCE BOND PAIR ENERGIESIf we introduce electron correlation into the wavefunction explicitly (2), our calculation becomes much more difficult.If we use configuration interaction (3), the wavefunction is no longer easy to interpret. We shall, therefore, follow Sinanoglu (1), and express the total energy as the sum of the energy resulting from an orbital calculation, plus a sum of pair energies. Since we want the orbital pair energies to include only the effects of real dynamic electron correlation, (4) In order to picture the general behavior of GVB electron pair correlation energies, let us consider two hydrogen atoms, A and B, at large separation (Fig. la). The electron repulsion is RB and is not altered -by electron correlation (if we ignore the small Van Der Waals attraction). As we bring the two atoms closer together, the charge distributions begin to overlap (Fig. lb). The electron repulsion is now RAB plus an additional term arising from the overlapping of the two charge distributions where r12 is very small. It is this additional term that is removed by electron correlation. In the correlated wavefunction, only one electron will occupy this region at a time. Hence, the correlation energy is determined by the overlapping of the two charge distributions, rather than by the total electron repulsion. This is-true not only for the large bond lengths we have considered above, but also for the