Understanding the rates of hydrogen-atom abstraction reactions: empirical, semi-empirical and ab initio approaches

Abstract

The performance of the empirical algorithm proposed by Roberts and Steel (J. Chem. Soc., Perkin Trans. 2, 1994, 2155), which relates the activation energy for the H-atom transfer reaction (A) to properties of the A˙+ HB → AH + B˙(A) four molecules involved, is assessed. This approach has been strongly criticised by Zavitsas in two recent papers (J. Chem. Soc., Perkin Trans. 2, 1996, 391 and J. Am. Chem. Soc., 1995, 117, 10 645), mainly on the grounds that it performs badly in comparison with his semi-empirical method for estimating E_a. These criticisms are examined and, in the main, are found to be without foundation. The identity reaction H₃Si˙+ H₄Si is viewed as a test case and it is shown that our empirical approach predicts an activation energy in accord with that estimated here by ab initio methods and with published experimental data on related reactions. The value of E_a predicted by Zavitsas' published method (71.5 kJ mol^–1) is very much higher than the value deduced from our empirical analysis (37.6 kJ mol^–1) or that obtained by the ab initio procedure (46.2 kJ mol^–1, which itself is probably too high). It is argued that polar effects are important in H-atom abstraction from thiols by alkyl radicals and that hydrogen-bonding interactions between A˙ and HB can be important, in particular when the leading atoms of A and B are oxygen or nitrogen. It is concluded that empirical models have a role to play in aiding the understanding of radical reactivity and that the approach of Roberts and Steel is relatively successful, given the limitations implicit in the model.