Warning: cannot automatically bond atom with strange atomic number

Hello everyone,

I’m testing psi4 with later actinides such as berkelium and californium. When I try to run a simple optimization calculation, it ends with the following message,

    Internal coordinates to be generated automatically.
    The INTCO_EXCEPTion handler:
    Warning: cannot automatically bond atom with strange atomic number
    Dynamic level is 0.
    exc.g_really_quit() is 0.

Here is the input

molecule bko2 {
0 8
O 1 1.764
O 1 1.764 2 114.0

basis mybas {
** assign sarc**
** assign O def2-tzvp**

set {
reference uhf
basis mybas
relativistic dkh
dkh_order 2
print 3
scf_type pk

optimize (‘scf’)

This is the first time I’m using Psi4. Apparently these atoms are not included in the optimization module, however I don’t know if is easy to overcome this issue. I would really appreciate any help.

Looks like the covalent_radii table ends just before your atoms of interest (https://github.com/psi4/psi4/blob/master/psi4/src/psi4/optking/cov_radii.h#L146). Are you working from compiled or conda?

Please just be generally aware that post-Krypton is not a region of the periodic table that has gotten much exercise with Psi4. And we seem to be less reliable than usual at correctly guessing occupancies when ECPs are involved, so please check up on those. (Not that you probably need such a warning if you’re accustomed to actinides :slight_smile: )

1 Like

Bad luck that the standard set ends at Cm! Annoyingly, I don’t see an easy
workaround short of changing the header file as @loriab suggests. In the
‘old days’, we would simply make sure the coordinates we wanted were
provided in a file, but with the process id file suffixes, this doesn’t
work anymore. In the ‘new days’, we should be able to simply provide the
coordinate set in the input, which would be easy with the new python
optimizer but not in the old one.

1 Like

Thanks @loriab for answering quickly.

I’m currently working with the conda version of Psi4 and I’m using an all-electron basis set so no ECP are needed. I’m just trying to compare results from Psi4 with those obtained by ORCA.

Unfortunately seems to be not as simple as I thought. Thanks anyway @Rollin_King !