M-F. Reyniers

Thermochemical and kinetic data were calculated at four cost-effective levels of theory for a set consisting of five hydrogen abstraction reactions between hydrocarbons for which experimental data are available. The selection of a reliable, yet cost-effective method to study this type of reactions for a broad range of applications was done on the basis of comparison with experimental data or with results obtained from computationally demanding high level of theory calculations. For this benchmark study two composite methods (CBS-QB3 and G3B3) and two density functional theory (DFT) methods, MPW1PW91/6-311G(2d,d,p) and BMK/6-311G(2d,d,p), were selected. All four methods succeeded well in describing the thermochemical properties of the five studied hydrogen abstraction reactions. High-level Weizmann-1 (W1) calculations indicated that CBS-QB3 succeeds in predicting the most accurate reaction barrier for the hydrogen abstraction of methane by methyl but tends to underestimate the reaction barriers for reactions where spin contamination is observed in the transition state. Experimental rate coefficients were most accurately predicted with CBS-QB3. Therefore, CBS-QB3 was selected to investigate the influence of both the 1D hindered internal rotor treatment about the forming bond (1D-HR) and tunneling on the rate coefficients for a set of 21 hydrogen abstraction reactions. Three zero curvature tunneling (ZCT) methods were evaluated (Wigner, Skodje & Truhlar, Eckart). As the computationally more demanding centrifugal dominant small curvature semiclassical (CD-SCS) tunneling method did not yield significantly better agreement with experiment compared to the ZCT methods, CD-SCS tunneling contributions were only assessed for the hydrogen abstractions by methyl from methane and ethane. The best agreement with experimental rate coefficients was found when Eckart tunneling and 1D-HR corrections were applied. A mean deviation of a factor 6 on the rate coefficients is found for the complete set of 21 reactions at temperatures ranging from 298 to 1000 K. Tunneling corrections play a critical role in obtaining accurate rate coefficients, especially at lower temperatures, whereas the hindered rotor treatment only improves the agreement with experiment in the high-temperature range.

The group contribution method for activation energies is applied to hydrogen abstraction reactions. To this end an ab initio database was constructed, which consisted of activation energies calculated with the ab initio CBS-QB3 method for a limited set of well-chosen homologous reactions. CBS-QB3 is shown to predict reaction rate coefficients within a factor of 2–4 and Arrhenius activation energies within 3–5 kJ mol−1of experimental data. Activation energies in the set of homologous reactions vary over 156 kJ mol−1with the structure of the abstracting radical and over 94 kJ mol−1with the structure of the abstracted hydrocarbon. The parameters required for the group contribution method, the so-called standard activation group additivity values, were determined from this database. To test the accuracy of the group contribution method, a large set of 88 additional activation energies were calculated from first principles and compared with the predictions from the group contribution method. It was found that the group contribution method yields accurate activation energies for hydrogen-transfer reactions between hydrogen molecules, alkylic hydrocarbons, and vinylic hydrocarbons, with the largest deviations being less than 6 kJ mol−1. For reactions between allylic and propargylic hydrocarbons, the transition state is believed to be stabilized by resonance effects, thus requiring the introduction of an appropriate correction term to obtain a reliable prediction of the activation energy for this subclass of hydrogen abstraction reactions.

A complete and consistent set of 95 Benson group additive values (GAV) for the standard enthalpy of formation of hydrocarbons and hydrocarbon radicals at 298 K and 1 bar is derived from an extensive and accurate database of 233 ab initio standard enthalpies of formation, calculated at the CBS-QB3 level of theory. The accuracy of the database was further improved by adding newly determined bond additive corrections (BAC) to the CBS-QB3 enthalpies. The mean absolute deviation (MAD) for a training set of 51 hydrocarbons is better than 2 kJ mol-1. GAVs for 16 hydrocarbon groups, i.e., C(Cd)3(C), C−(Cd)4, C−(Ct)(Cd)(C)2, C−(Ct)(Cd)2(C), C−(Ct)(Cd)3, C−(Ct)2(C)2, C−(Ct)2(Cd)(C), C−(Ct)2(Cd)2, C−(Ct)3(C), C−(Ct)3(Cd), C−(Ct)4, C−(Cb)(Cd)(C)(H), C−(Cb)(Ct)(H)2, C−(Cb)(Ct)(C)(H), C−(Cb)(Ct)(C)2, Cd−(Cb)(Ct), for 25 hydrocarbon radical groups, and several ring strain corrections (RSC) are determined for the first time. The new parameters significantly extend the applicability of Benson's group additivity method. The extensive database allowed an evaluation of previously proposed methods to account for non-next-nearest neighbor interactions (NNI). Here, a novel consistent scheme is proposed to account for NNIs in radicals. In addition, hydrogen bond increments (HBI) are determined for the calculation of radical standard enthalpies of formation. In particular for resonance stabilized radicals, the HBI method provides an improvement over Benson's group additivity method.

A quantum chemical investigation is presented for the determination of accurate kinetic and thermodynamic parameters for hydrocarbon radical reactions. First, standard enthalpies of formation are calculated at different levels of theory for a training set of 58 hydrocarbon molecules, ranging from C1 to C10, for which experimental data are available. It is found that the CBS-QB3 method succeeds in predicting standard enthalpies of formation with a mean absolute deviation of 2.5 kJ/mol, after a systematic correction of −1.29 kJ/mol per carbon atom and −0.28 kJ/mol per hydrogen atom. Even after a systematic correction, B3LYP density functional theory calculations are not able to reach this accuracy, with mean absolute deviations of 9.2 (B3LYP/6-31G(d)) and 12.9 kJ/mol (B3LYP/6-311G(d,p)), and with increasing deviations for larger hydrocarbons. Second, high-level transition state geometries are determined for 9 carbon-centered radical additions and 6 hydrogen additions to alkenes and alkynes and 10 hydrogen abstraction reactions using the IRCMax(CBS-QB3//B3LYP/6-311G(d,p)) method. For carbon-centered radical addition reactions, B3LYP/6-311G(d,p) slightly overestimates the length of the forming C−C bond as compared to the IRCMax data. A correlation to improve the agreement is proposed. For hydrogen addition reactions, MPW1K density functional theory (MPW1K/6-31G(d)) is able to locate transition states. However, the lengths of the forming C−H bonds are systematically longer than reference IRCMax data. Here, too, a correlation is proposed to improve the agreement. Transition state geometries for hydrogen abstraction reactions obtained with B3LYP/6-311G(d,p) show good agreement with the IRCMax reference data. Third, the improved transition state geometries are used to calculate activation energies at the CBS-QB3 level. Comparison between both CBS-QB3 and B3LYP density functional theory predictions shows deviations up to 25 kJ/mol. Although main trends are captured by B3LYP DFT, secondary trends due to radical nucleophilic effects are not reproduced accurately.