Hi Antti,
Sorry, we don't have good radii for this metals.
There might be a couple of good choices dependent on the
problem you are modeling.
For naked metal cations it might be a good idea to include an implicit solvens shell for instance - at least for strongly coordinating solvents.
The normal cosmo radii are roughly 1.17*bondii radius. But as far as I know there are no bondii radii available for the metals you are interested in. Dependent on your molecule
there are two cases:
1) The metal has no (or nearly no) solvens accesable surface. In other words: The ligands coordinate the metal in a way that the metal surface is perfectly screened. In this case any metal radius that defines a sphere smaller that the ligand sphere is OK.
2) If the metal surface is accessible, like in square planar complexes, you need to find a useful radius.
The optimized cosmo radii have been fitted to thermodynamic data. But this is not a simple straight forward procedure. Furthermore, exp. data is not available for many systems.
I would try some reasonable values like metal radius*1.17 and do some test calculations in order to get an impression of the sensitivity.
hope this helps a bit
michael
