Methods
All calculations were performed with ORCA suite of programs, version 4.2.1.78 The geometries were optimised from the closest available crystal structure using unrestricted Kohn–Sham DFT calculations with the TPSS density functional.79Ahlrich’s def2-TZVP basis set was used for all elements except for carbon and hydrogen, for which def2-SVP was used.80The split-RI-J approximation with the def2/J basis sets was used.81-83 Grimme’s dispersion correction with Becke–Johnson damping (D3BJ) was employed.84,85Solvation effects were included with the SMD model,86choosing water as the modelled solvent. The grid was set to 6 and integration accuracy was increased to 6.0 in ORCA nomenclature. The convergence criteria for the SCF and the geometry optimisation were set to “tight” in ORCA nomenclature; the only exception to this was complex 8 which was optimised with the “NormalOpt“ keyword due to its large size.
For the single point calculations, the hybrid functionals TPSSh,87 PBE088 and B3LYP89 90 were used in conjunction with the RIJCOSX82 approximation and a GridX choice of 8. Otherwise the same settings as for the geometry optimisations were used, except for a change in basis set: the CP(PPP)91basis set was used for Fe and def2-TZVP80 basis set for all other elements. Additionally, the integration grid for iron was increased to 7.0 in ORCA nomenclature. While all reported calculations use the SMD solvation model for water, we note that omission of the solvation model or its inclusion of with dielectric constants ranging from ε = 4–80 altered the Mössbauer parameters very little.
All calibrations and corresponding visualizations were performed with OriginPro 9.1,92 except for those related to bootstrapping, which were performed with Python 3.6.893 (including the packages Numpy 1.18.1,94 Scikit-learn 0.22.2,95Matplotlib 3.1.1,96 and Seaborn 0.9.097). Contact densities, fit parameters, and details on the bootstrapping analysis are reported in the Supporting Information. We also make available a free interactive notebook that allows researchers to request predictions of Mössbauer parameters, including virtual measurement uncertainties (error bars). The notebook can be accessed attinyurl.com/mbs-notebook.