Abstract
A series of semiempirical PM6* and PM7 methods has been tested in reproducing relative conformational energies of 27 realistic-size complexes of 16 different transition metals (TMs). An analysis of relative energies derived from single-point energy evaluations on density functional theory (DFT) optimized conformers revealed pronounced deviations between semiempirical and DFT methods, indicating a fundamental difference in potential energy surfaces (PES). To identify the origin of the deviation, we compared fully optimized PM7 and respective DFT conformers. For many complexes, differences in PM7 and DFT conformational energies have been confirmed often manifesting themselves in false coordination of some atoms (H, O) to TMs and chemical transformations/distortion of coordination center geometry in PM7 structures. Despite geometry optimization with fixed coordination center geometry leading to some improvements in conformational energies, the resulting accuracy is still too low to recommend explored semiempirical methods for out-of-the-box conformational search/sampling: careful testing is always needed.