The previously developed valence bond configuration interaction (VBCI) method (Wu, W.; Song, L.; Cao, Z.; Zhang, Q.; Shaik, S., J. Phys. Chem. A, 2002, 105, 2721) that borrows the general CI philosophy of the MO theory, is further extended in this article, and its methodological features are improved, resulting in three accurate and cost-effective procedures: (a) the effect of quadruplet excitation is incorporated using the Davidson correction, such that the new procedure reduces size consistency roblems, with due improvement in the quality of the computational results. (b) A cost-effective procedure, named VBCI(D, S), is introduced. It includes doubly excited structures for active electrons and singly excited structures for inactive pairs. The computational results of VBCI(D, S) match those of VBCISD with much less computational effort than VBCISD. (c) Finally, a second-order perturbation theory is utilized as a means of configuration selection, and lead to considerable reduction of the computational cost, with little or no loss in accuracy. Applications of the new procedures to bond energies and barriers of chemical reactions are presented and discussed.

One of the landmark achievements of quantum chemistry, specifically of MO-based methods that include electron correlation, was the precise calculation of the barrier for the hydrogenexchange reaction (B. Liu, J. Chem. Phys. 1973, 58, 1925; P. Siegbahn, B. Liu, J. Chem. Phys. 1978, 68, 2457). This paper reports an accurate calculation of this barrier by two recently developed VB methods that use only the eight classical VB structures. To our knowledge, the present work is the first accurate ab initio VB barrier that matches an experimental value. Along with the accurate barrier, the VB method provides accurate bond energies and diabatic quantities that enable the barrier heightto be analyzed by the VB state correlation diagram approach, VBSCD (S. Shaik, A. Shurki, Angew. Chem. 1999, 111, 616; Angew. Chem. Int. Ed. Engl. 1999, 38, 586). This is a proof of principal that VB theory with appropriate account of dynamic electron correlation can achieve quantitative accuracy of reaction barriers, and still retain a compact and interpretable wave function. A sample of SN2 barriers and dihalogen bonding energies, which are close to CCSD (T) and G2(+) values, show that the H3 problem is not an isolated case, and while it is premature to conclude that VB theory has come of age, the occurrence of this event is clearly within sight.

Valence bond SCF calculations have been carried out on identity proton transfers of the type X-H + X-→X- + H-X and X-H+ + X→X + X-H+ for systems where X=F-, Cl-, Br-, OH-, SH-, NH2-, PH2-,CH3-, SiH3-, OH2, NH3, and PH3. A hybrid consisting of three contributing structuressreactantlike, productlike, and ionic (X- H+ X- or X H+ X)-gives reasonable results, though additional structures afford a slight further decrease in energy. Energies somewhat below Hartree-Fock and reasonable weights are obtained. Calculated barriers agree well with those obtained by high-level ab initio calculations with correlation. Insights into the factors determining barrier height are obtained.

d was tested by calculating the bond energies of H2, LiH, HF, HCl, F2, and Cl2 as well as the barriers of identity hydrogen abstraction reactions,X·+X'H→XH+X'·(X, X'=CH3, SiH3, GeH3, SnH3, or PbH3). It is shown that VBCIS gives results that are at par with breathing orbital valence bond method. The VBCISD method is better and its results match those of the molecular orbital based coupled cluster CCSD method. Future potential directions of the development of the VBCI approach are outlined.