Semi-Empirical Evaluation of Hindered Internal Rotors for Accelerated Thermodynamics Predictions

Accurate determination of a molecule's thermodynamic properties using quantum chemistry methods is crucial for developing kinetic models. A vital step in the quantum chemical workflow is calculating torsional energy profiles to apply the one-dimensional hindered-rotor approximation. However, these profiles are typically obtained by performing rotational scans using density functional theory (DFT) methods, which adds high computational cost to the workflow. Here, we assess the performance of the semiempirical GFN2-xTB method for rapid generation of these torsional profiles by calculating the contributions of GFN2-xTB and B3LYP profiles to the enthalpy, entropy, heat capacity, and Gibbs free energy of molecules. Several correction methods to improve the results are proposed and evaluated. It is found that correcting the energy profiles based on their second derivative at the optimized-geometry point yields the most accurate results. This optimal correction method results in a mean absolute error on the Gibbs free energy at 1000 K of 0.43 kJ mol^-1 for a hydrocarbons data set and 0.93 kJ mol^-1 for a data set of nitrogen-containing compounds. In conclusion, the GFN2-xTB method, when combined with a suitable correction method, can accelerate the generation of hindered rotor profiles by a factor of 700, with only a minor loss of accuracy.