Comparison of different strategies for modelling hydrogen atoms in charge density analyses

The quality of various approximation methods for modelling anisotropic displacement parameters (ADPs) for hydrogen atoms was investigated in a comparative study. A multipole refinement was performed against high-resolution single crystal X-ray data of 9-diphenylthiophosphoranylanthracene (SPAnH) and 9,10-bis-diphenylthiophosphoranylanthracene·toluene (SPAnPS). Hydrogen-atom parameters and structural properties derived from our collected neutron data sets were compared with those obtained from the SHADE-server, the software APD-Toolkit based on the invariom database, the results from Hirshfeld atom refinement conducted in the OLEX2 GUI (HARt), and the results of anisotropic hydrogen refinement within XD2016. Additionally, a free refinement of H-atom positions against X-ray data was performed with fixed ADPs from various methods. The resulting C-H bond distances were compared with distances from neutron diffraction experiments and the HARt results. Surprisingly, the refinement of anisotropic hydrogen displacement parameters against the X-ray data yielded the smallest deviations from the neutron values. However, the refinement of bond-directed quadrupole parameters turned out to be vital for the quality of the resulting ADPs. In both model structures, SHADE and, to a lesser extent, APD-Toolkit showed problems in dealing with atoms bonded to carbon atoms with refined Gram-Charlier parameters for anharmonic motion. The HARt method yields the most accurate C-H bond distances compared to neutron data results. Unconstrained refinement of hydrogen atom positions using ADPs derived from all other used approximation methods showed that even with well approximated hydrogen ADPs, the resulting distances were still significantly underestimated.