Right Answer for the Right Reason? Benchmarking Protocols and Pitfalls on a Ru-Metathesis Example.

A layered meta-benchmarking analysis was devised with the aim of illustrating how to produce an experimental/computational protocol for the method selection and estimation of Gibbs energies of catalytically prominent reactions. Our test subject involved the active-latent equilibrium through the cis-trans isomerism of two metathesis catalysts: mesitylene-Ru-SCF-Cl and diisopropylphenyl-Ru-SCF-Cl. The strategy was two-fold: first to perform a computational benchmark for the energies in the gas phase, followed by benchmarking the enthalpy and the Gibbs energy, including solvation and entropy, from experimental references. This "wedding cake" build-up of subsequent methods applied to our particular small test reaction indicates that: (1) DLPNO-CCSD(T)/CBS works well as a reference method for large systems. (2) Among several functionals ωB97XD and M06 were the most accurate. (3) Choosing between IEF-PCM and SMD solvation models turned out to be case dependent. (4) For the vibrational entropic component, low-frequency vibrations often produce humungous errors, which can be improved by Cramer and Truhlar's or Grimme's methods; however, their cut-off parameters had to be lowered from their standard values. (5) Solvation methods are important for enthalpies, but they are inadequate for entropies. (6) All of these components are equally important for the accuracy of organometallic complexes' reactions. The only way to find the right method for the right reasons is to be sure to match all of the Gibbs energy terms to benchmarked experimental and computational values.