small selection of the published mean absolute deviations (MADs) between computed and experimental heats of formation. Of course, the corresponding values cannot be compared directly, since they are based on different sets of reference molecules and reference data, but they should provide some indication of the errors that can be expected in such calculations. The reader should consult the original literature for further information [1-31].

5.DISCUSSION

The statistical evaluations of the preceding section indicate that the semiempirical MO methods can predict heats of formation with useful accuracy and at very low computational costs. When comparing with

ab initio or DFT methods, the following points [32] should be kept in mind, however:

1.In general, errors tend to be more systematic at a given ab initio or DFT level and may therefore often be taken into account by suitable corrections. Errors in semiempirical calculations are normally less uniform and thus harder to correct.

2.The accuracy of the semiempirical results may be different for different classes of compounds, and there are elements that are more ”difficult” than others. Such variations in the accuracy are again less pronounced in high-level ab initio and DFT calculations.

3.Semiempirical methods can only be applied to molecules containing elements that have been parameterized, while ab initio and DFT methods are generally applicable.

4.Semiempirical parameterizations require reliable experimental or theoretical reference data and are impeded by the lack, of such data. Such problems do not occur in ab initio or DFT approaches.

In spite of these limitations, there are many areas where the established MNDO-type semiempirical methods can be applied successfully in calculations of thermochemical properties. This suggests that the underlying MNDO model includes the physically relevant interactions so that the parameterization can absorb the errors due to the MNDO approximations in an average sense. However, further improvements are clearly needed, and the inclusion of orthogonalization corrections that account for Pauli exchange repulsion indeed seems to enhance the accuracy of the calculated thermochemical properties (see the OM1 and OM2 results). This supports our belief [12] that a theoretically guided search for better models offers the most promising perspective for general-purpose semiempirical methods with better overall performance.

ACKNOWLEDGEMENTS

The author wishes to thank his coworkers for their contributions, particularly M. Kolb, A. Voityuk, and W. Weber.

30.B. Ahlswede and K. Jug, J. Comp. Chem. 20, 572 (1999).

31.K. Jug, G. Geudtner, and T. Homann, J. Comp. Chem. 21, 974 (2000).

32.W. Thiel, ACS Symp. Ser. 677, 142 (1997).

33.J. A. Pople, D. P. Santry, and G. A. Segal, J. Chem. Phys. 43, S 129 (1965).

34.W. J. Hehre, L. Radom, P. v. R. Schleyer, and J. A. Pople, Ab Initio Molecular Orbital Theory, Wiley, New York (1986).

35.L. A. Curtiss, K. Raghavachari, P. C. Redfern, and J. A. Pople, J. Chem. Phys. 106, 1063 (1997).

36.L. A. Curtiss, K. Raghavachari, P. C. Redfern, V. Rassolov, and J.A. Pople, J. Chem. Phys. 109, 7764 (1998).

37.W. Thiel, Program MNDO99, Mülheim (1999).

38.J. B. Pedley, R. D. Naylor, and S. P. Kirby, Thermochemical Data of Organic Compounds, 2nd ed., Chapman Hall, London (1986).

39.M. W. Chase, C. A. Davies, J. R. Downey, D. R. Frurip, R. A. McDonald, and

A.N. Syverud, JANAF Thermochemical Tables, 3rd ed., J. Phys. Chem. Ref. Data 14 (1985), Suppl. 1.

40.S. G. Lias, J. E. Bartmess, J. F. Liebman, J. L. Holmes, R. D. Levin, and W.

G.Mallard, Gas Phase Ion and Neutral Thermochemistry, J. Phys. Chem. Ref. Data 17 (1988), Suppl. 1.

41.M. J. S. Dewar and H. S. Rzepa, J. Am. Chem. Soc. 100, 58 (1978).

42.D. Higgins, C. Thomson, and W. Thiel, J. Comp. Chem. 9, 702 (1988).