The time has come for a major update to (soon to be renamed) OBeLiX. The descriptor calculator has been reworked, mistakes were made (and fixed). We are now able to calculate descriptors for XYZ files, CREST output and DFT log files. Morfeus and cclib are used for descriptor calculation/extraction. For CREST, these descriptors are boltzmann weighted and averaged.
The donors are labeled based on their GFN2-xTB calculated charge, where less negative charge is called the 'max donor' and more negative charged is called the 'min donor'.
In our dataset we have different types of structures: metal with just the bidentate ligand coordinated ('pristine'), metal with the bidetnate ligand coordinated on one side and norbornadiene on the other side ('nbd') and metal with the bidentate ligand on one side and two acetonitrile molecules on the other side ('acetonitrile').
For the nbd structures, additional descriptors such as H-C-M-donor dihedral angles or C=C bond distances can be calculated since the nbd is always at the same position (bottom) in our xyz files. This needs to be extended towards a substructure search, but getting a proper openbabel or RDKit mol object from a TM complex has proven to be a challenge on its own.
For now, enjoy the many descriptors we can calculate already. Which include:
DFT descriptors: thermodynamic descriptors (electronic energy, enthalpy etc.), (anti-)bonding orbital occupations (donor – metal, donor – any other element), dipole moment, lone pair occupancy of the donors, dispersion energy, NBO and Mulliken charge on metal and donors, HOMO-LUMO gap and derivations (hardness, softness, electronegativity etc.)
Hi all,
The time has come for a major update to (soon to be renamed) OBeLiX. The descriptor calculator has been reworked, mistakes were made (and fixed). We are now able to calculate descriptors for XYZ files, CREST output and DFT log files. Morfeus and cclib are used for descriptor calculation/extraction. For CREST, these descriptors are boltzmann weighted and averaged.
The donors are labeled based on their GFN2-xTB calculated charge, where less negative charge is called the 'max donor' and more negative charged is called the 'min donor'.
In our dataset we have different types of structures: metal with just the bidentate ligand coordinated ('pristine'), metal with the bidetnate ligand coordinated on one side and norbornadiene on the other side ('nbd') and metal with the bidentate ligand on one side and two acetonitrile molecules on the other side ('acetonitrile').
For the nbd structures, additional descriptors such as H-C-M-donor dihedral angles or C=C bond distances can be calculated since the nbd is always at the same position (bottom) in our xyz files. This needs to be extended towards a substructure search, but getting a proper openbabel or RDKit mol object from a TM complex has proven to be a challenge on its own.
For now, enjoy the many descriptors we can calculate already. Which include: