Study On The Theoretical Procedures Of Accurate Model Chemistry Calculations

A series of modifications on geometry optimization, core electronic correlation and the basis sets non-extrapolation to the well-known Gaussian-n theories of accurate model chemistry calculations proposed by J. A. Pople (the 1998 Nobel Prize-Chemistry Winner) and co-workers are carried out via the following process.Firstly, the basic approximations in the present fundamental theoretical procedures in Quantum Chemistry are examined or criticized. Each of the effects of approximate treatment to the accurate eigenvalue problem is discussed.Secondly, two families of theoretical model chemistry procedures based on the ab initio molecular orbital theory proposed, respectively, by Pople and co-workers (Gaussian-n series or Gn) and by G. A. Petersson and co-workers (Complete Basis Sets or CBS) are compared and investigated in detail. It is found that the treatment on the residual errors in CBS family is less empirical. However, the empirical treatment of the residual errors, namely HLC (high level correction), in the Gn theories is more simple and more complete. Therefore, our efforts have been mainly focused on the Pople's Gn family of theories.Finally, we carry out a series of comprehensive study on the applications of G2 theory and of those of the theoretical levels higher than G2. From these theoretical calculations, we find and understand that there are three main factors affecting the accuracy of the high accurate model chemistry calculations by geometry optimization, frozen-core (FC) approximation and basis sets extrapolation. These discoveries lay a solid foundation in our further investigations in modifying the G2, G2(QCI) and G3 theories.In addition, we have also obtained the following results from our comprehensive study, which are greatly helpful for the development of the high accurate model chemistry theoretical procedures. (1). The single-reference approximation in the energy calculations of a molecule at its equilibrium geometry will not significantly affect the accuracy of high accurate reaction energy calculations. (2). The energy from a single reference M(ller-Plesset perturbation theory is not reliable sometimes if the molecule is not at its equilibrium geometry and the Hartree-Fock wavefunction is not dominated. (3). In case the energy from a single reference M(ller-Plesset perturbation theory is incorrect, however the energy from the quadratic configuration interaction QCISD(T) is almost correct. (4). Contrast to the larger errors in the G2 frame for the high spin (triplets and quartets) species and all of the carbonyl structure C=O (carbon-oxygen ( bonds), these errors are not significant in the G3(QCI) theoretical frame. Based on our practice and the analysis of previous work in literature, we have further modified the G2, G2(QCI) and the G3 theories in the following procedures in this work. 1. A systematic comparative study of geometry optimization on all of the inorganic molecules from the 1996-1997 CRC Handbook of Chemistry and Physics (77th edition) is carried out at MP2, B3LYP, B3PW91 and partially at BPW91, QCISD and QCISD(T) levels with basis sets at 6-31G(d), 6-31G(d,p), 6-311G(d,p) and 6-311G(2d,p). Results show that the MP2(full)/6-311G(d,p), B3PW91 and B3LYP are superior to the other levels. Therefore, better geometry optimizations are suggested in the Gn single-point energy calculations. As an application, the B3PW91/6-311G(d,p) optimization has been used in our G3(QCI/fun)// B3PW91 (where n=1, 2, see below) frame.2. Core-electron correlation energy correction via equation: (E(full)= E[Level(full)/basis_sets] - E[Level(FC)/basis_sets](where Level denotes a post-SCF level) is suggested to be directly added into the total energy of the original Gn theories in facilitating the flaw of frozen-core approximations. From this scheme, by using the G2-1 test sets we have examined Level=MP2, basis_sets= 6-31G(d) and 6-311G(d,p) on G2, G2(QCI) and G3(QCI) theories and proposed some new theoretical procedures of G2(fu1), G2(fu2), G2(QCI/fu1), G2(QCI/...