G3-RAD and G3X-RAD: Modified Gaussian-3 (G3) and Gaussian-3X (G3X) procedures for radical thermochemistry

Abstract

The G3-RAD, G3X-RAD, G3(MP2)-RAD, and G3X(MP2)-RAD, procedures, designed particularly for the prediction of reliable thermochemistry for free radicals, are formulated and their performance assessed using the G2/97 test set. The principal features of the RAD procedures include (a) the use of B3-LYP geometries and vibrational frequencies (in place of UHF and UMP2), including the scaling of vibrational frequencies so as to reproduce ZPVEs, (b) the use of URCCSD(T) [in place of UQCISD(T)] as the highest-level correlation procedure, and (c) the use of RMP (in place of UMP) to approximate basis-set-extension effects. G3-RAD and G3X-RAD are found to perform well overall with mean absolute deviations (MADs) from experiment of 3.96 and 3.65 kJ mol⁻¹, respectively, compared with 4.26 and 4.02 kJ mol⁻¹ for standard G3 and G3X. G3-RAD and G3X-RAD successfully predict heats of formation with MADs of 3.68 and 3.11 kJ mol⁻¹, respectively (compared with 3.93 and 3.60 kJ mol⁻¹ for standard G3 and G3X), and perform particularly well for radicals with MADs of 2.59 and 2.50 kJ mol⁻¹, respectively (compared with 3.51 and 3.18 kJ mol⁻¹ for standard G3 and G3X). The G3(MP2)-RAD and G3X(MP2)-RAD procedures give acceptable overall performance with mean absolute deviations from experiment of 5.17 and 4.92 kJ mol⁻¹, respectively, compared with 5.44 and 5.23 kJ mol⁻¹ for standard G3(MP2) and G3X(MP2). G3(MP2)-RAD and G3X(MP2)-RAD give improved performance over their standard counterparts for heats of formation (MADs=4.73 and 4.44 kJ mol⁻¹, respectively, versus 4.94 and 4.64 kJ mol⁻¹). G3(MP2)-RAD shows similar performance to G3(MP2) for radical heats of formation (MAD=5.10 versus 5.15 kJ mol⁻¹) while G3X(MP2)-RAD performs significantly better than G3X(MP2) (MAD=4.67 versus 5.19 kJ mol⁻¹).