Angewandte Chemie International Edition
© WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Different Strokes for Different Folks
Hoffmann, Schleyer, and Schaefer present a delightful essay about subjects that all serious computational chemists should know well. The discussion of precision and accuracy is an important reminder to us all, and it is going to be very useful for the education of the many young computational chemists in my group and elsewhere.
I hope this essay will not dissuade chemists from using less accurate methods for some problems: Huckel theory and Extended Huckel theory are wonderful examples of theories of low accuracy and precision providing excellent explanations and guides to experiment.
The thermodynamic quantities of greatest interest to synthetic chemists, my frequent collaborators, are the sign of the free energy of reaction (does the reaction occur or not?), the relative energies of conformational or equilibrating isomers (what is the ratio in an equilibrium?), or the relative energies of diastereomeric transition states (what ratio of isomers will be formed in a reaction?). Quantum mechanical methods of relatively low accuracy for the computation of heats of formation often prosper well for the relative energies of isomers or of diastereomeric transition states. Demonstrations that various levels of quantum mechanics are of low accuracy for one problem do not disqualify that method for use in other problems.
There are ample examples of the accurate assessment of relative energies of stereoisomeric transition states, going back to the Anh-Eisenstein study of Cram's rule. Of course, the direct comparison of computational predictions with experiments on reactions of synthetic value in solution is much more complex than calculations for gas-phase properties, since free energies in solution are the quantities of interest. Errors in solvation, and the necessity of averaging over accessible structures or transition states, may introduce more inaccuracy than is inherent in the quantum mechanical method. Building credibility for the accuracy of the results in such cases requires calibration versus known systems. The best procedure is the exploration of complex systems with a robust DFT functional like B3LYP or newer functionals like MO6-2X, followed by higher accuracy correlated ab initio methods to validate the energetics.
K. N. Houk
University of California Los Angeles (USA)