# How do you optimize an organo-metalic compound containing cerium in in Gaussian?

I am trying to perform optimisation of the molecule attached. I need to get partial charges and frequencies. The issue is that it doesn't converge with following input:

%nprocshared=8
%mem=16GB
%chk=CTB.chk
#n opt b3lyp/gen pop=(ReadRadii, savechelpg) SCF=QC
CTB
0 2
Ce -6.41954 5.21611 0.61820
...
N -4.38490 3.87720 0.35349

H -5.52145 4.13584 -2.66372
Ce 0
SDD
****
H Si C N 0
6-31G
****
Ce 2.88


As well I get error message: Restarting incremental Fock formation. Gradient too large for Newton-Raphson or scaled steepest descent -- use steepest descent instead.

• Hello, it is necessary to add the keywords "pseudo=read" after "b3lyp/gen" because you use pseudopotential SDD. Moreover, I would suggest to use "scf=xqc". Apr 29 '20 at 13:55
• Baffled by the negative vote. Seems to me to be a perfectly valid question. Apr 29 '20 at 19:08
• It's hard to know your problem with convergence without seeing more of the output (e.g., how the SCF progressed). I agree that XQC is a good idea, but I'd actually suggest starting by minimizing your compound with xtb an approximate method from the Grimme group that generally provides good starting geometries and usually converges more easily. Apr 29 '20 at 23:44
• Other options to try and find a better starting geometry and save time: simplify the ligands to $\ce{N(SiH3)2}$ or even $\ce{NH2}$, use a GGA such as PBE for a pre-optimization. As a side-note, use a modern dispersion correction. Apr 30 '20 at 6:25
• With a little more detail, this question would likely be well text on the new Materials Modeling SE May 18 '20 at 0:46