Conformational sampling issue with Josiphos-related ligand

[binhmai]

I'm running conformational sampling using CREST with gfn2 for Pd-catalyzed reactions supported by Josiphos ligand (which has a ferrocene core). However, during the sampling, I found that the Pd and Fe atoms tend to come closer to each other (giving bimetallic Pd-Fe species). This situation does not happen with gfn1, but I only obtained 1 or 2 conformers! I'm also considering gfn2/gfnff, but I'm not sure whether it's good for organometallic systems. Is there any good way to overcome this issue?

Thanks,
Binh

[thfroitzheim]

While all GFN methods cover al of the transition metals, they are approximate and special-purpose methods. Especially, in complicated situations with two different metals, it is not clear which GFN method will give the most accurate results (the specific situation with Fe with Pd was likely not in the fit, so that there can be errors). If GFN1-xTB or GFN-FF work better it should be fine to use instead (you have to test for such a case).

If the formation of the bimetallic complex goes along with bond dissociation to your ligands, you might want to but a constraint on the bonds during the CREST run (see either here for constraints in xtb or here for constraints in CREST).