Assessment of the suitability of using the composite G2, G3, and CBS-RAD methods for predicting activation energies

In this work, the accuracy of the G2, G3, and CBS-RAD methods for predicting activation barriers of ligand transfer reactions is investigated. We find that the zero point corrected G2 method has an RMS error of 3.82 kcal/mol for activation barriers. The G3 method has an RMS error of 4.16 kcal/mol. After adding thermal corrections to the G2 and zero point corrected results, the RMS error for the G2 method is 4.92 kcal/mol, while the error for the G3 method is 4.55 kcal/mol. In contrast, the CBS-RAD method has errors of 3.80 kcal/mol for zero point energy corrected activation energies, and 2.82 kcal/mol for thermally corrected results. The G3 method was found to require only 40% of the computational time required for the G2 method, making it an attractive alternative for predicting activation energies yielding errors of about 4 kcal/mol. The CBS-RAD method has a computational cost four times greater than that of the G2 method and gives an improvement of only about 1 kcal/mol.