Semiempirical molecular orbital theory current theories

The underlying theoretical approach is characterized by the type of the wavefunction and the choice of the basis set. Most current general-purpose semiempirical methods are based on molecular orbital theory and employ a minimal basis set for the valence electrons. Electron correlation is treated explicitly only if this is necessary for an appropriate zero-order description (e.g., in the case of electronically excited states or transition states in chemical reactions). Correlation effects are often included in an average sense by a suitable representation of the two-electron integrals and by the overall parametrization. 

Semiempirical methods, for example, MNDO. AMI, and PM3. are simplifications of ab initio molecular orbital theory and employ empirically determined parameters in essence, they only differ in the approximations being made. These methods involve adjustable parameters associated with molecular properties that are calibrated against experimental data. The chief advantage of semiempirical calculations over ab initio calculations is that they are several orders of magnitude faster. Thus, calculations for systems of up to c. 200 atoms are currently possible, whereby with ab initio methods, the limit is a moderately sized molecule (about 50 atoms at the time of writing), if rational results are to be obtained. Frequently, semieinpirical methods have proved to be the computational procedures of choice for studying relatively large molecules. 

We start with some biographical notes on Erich Huckel, in the context of which we also mention the merits of Otto Schmidt, the inventor of the free-electron model. The basic assumptions behind the HMO (Huckel Molecular Orbital) model are discussed, and those aspects of this model are reviewed that make it still a powerful tool in Theoretical Chemistry. We ask whether HMO should be regarded as semiempirical or parameter-free. We present closed solutions for special classes of molecules, review the important concept of alternant hydrocarbons and point out how useful perturbation theory within the HMO model is. We then come to bond alternation and the question whether the pi or the sigma bonds are responsible for bond delocalization in benzene and related molecules. Mobius hydrocarbons and diamagnetic ring currents are other topics. We come to optimistic conclusions as to the further role of the HMO model, not as an approximation for the solution of the Schrodinger equation, but as a way towards the understanding of some aspects of the Chemical Bond. 

Semiempirical approaches to quantum chemistry are thus characterized by the use of empirical parameters in a quantum mechanical framework. In this sense, many current methods contain semiempirical features. For example, some high-level at initio treatments of thermochemistry employ empirical corrections for high-order correlation effects, and several advanced density functionals include a substantial number of empirical parameters that are fitted against experimental data. We shall not cover such approaches here, but follow the conventional classification by considering only semiempirical methods that are based on molecular orbital (MO) theory and make use of integral approximations and parameters already at the MO level. 

A third approach, the one most important for the current discussions, is to treat the electric-field as a source of perturbation of the total molecular energy using real and virtual excited state transitions. This approach uses electronic wave functions either for all of the electrons of the molecule (ab initio calculations) or for only the valence electrons (so-called semiempirical theories). Semiempirical Hamiltonians may ignore electron interactions completely (Huckel theory). They may assume one jt-orbital per carbon and assume no overlap between adjacent electron orbitals (Pariser-Parr-Pople or PPP). Or, they may include both the a and... 

---