Perhaps you could try to replace the two Pt-(PMe3)2X side groups for a methyl group and optimize the organic core and do the same thing to the Pt-group, i.e., replace the organic core with a methyl group for one Pt unit. Then assemble the individual molecules you obtained and re-optimize. Also: what does the output file say exactly on what's going wrong?

Dear Dr Sigismund,

Thank you for your reply!

Very common the job ended with Berny optimization, error in internal coordination.

If cartesian command added, then it just simply fail to converge.

It is the first time i heard about the ideal or partially optimize the molecule. I did it before with large molecures, which i optimize the core first then added the attachments.

But your suggestion is opt two components seperately, then how to combine the two fragment together? Lets say we have two output files with optimized half structures.

Thank you!

First, you should read the output for a possible explanation. Next, try with somewhat lighter calculations. Lower the approximation level and remove freq. If this only works with opt=(loose, cartesian) try to continue from this optimization step. Finally add freq at the end.

you just go through the Gaussian manual, and check whether the basis set you have used is suitable for the molecule or not? and then try it again

I think your having problem is not just a basis set one.

I know that the mixed basis set you're using is not so-called the best.

However, at the same time, I thing it is not bad.

I reccomend you to check the accuracy of basis set, after you solved your having problem on internal coordinates.

From different point of view, I'd like to have a comment. Did you check your initial structure is well symmetrized? Do you have any strong reason for using nosymm option? If your calculation properly used point group symmetry, the number of internal coordinates are limited, and your optimization will be much easier. Could you please check this point, first?

Then, if you'll find that you're using symmetry, you have to check the size of step used in your geometry optimization. When we treat molecules that have very flat PES, sometimes geometry optimization fail with similar problem you have. If you're using Gaussian, you can use e.g. MaxStep=N to reduce step size. And in such a case, opt=calcFC or calc All option is often very helpful. Please check Gaussian IOP manual for details.

A) Is it possible to run the opt=tight int=grid=ultrafinegrid cphf=grid=fine nosym Calculation (which failed to converge) for more steps or was it cycling between two energy values?

B) Have you tried optimizing using opt=(loose, cartesian) and then optimizing the structure that comes out of that to a tighter level of convergence?

C) You could try changing the internal coordinates used for optimization: remove any dihedrals which are close to 0 or 180 degrees (in the modredundant section, add the line 'D N1 N2 N3 N4 R' or 'D N1 N2 N3 N4 K' where N1-4 are the numbers of the atoms in the dihedral) and replace them with linear bends using 'L N1 N2 N3 N4 B'

You probably have bond angles near 180 degree. Try using dummy atoms (x) to remove these kind of angles.

Thank you all for your time and advice. I will try to follow your advice see if it works for my case.

Thanks gain!

In The Name Of GOD

Dear Sir/Madam

In nature the real flat molecule does not exist. You should consider a small z-coordinate if molecule is in xy-plane. For example if you want to optimize a graphene cluster or flake and set all z-coordinates to zero your data will not converge. Of course after optimization software will set the z-coordinates equal to a very small value.

Best Regards

H.Simchi

HI,

I would use SDD basis for for platinum. LANL2DZ doesn't have F functions and the SDD basis wont cost you more. This is just a comment for when you get a good structure as it will most likely not help your optimization. You could also drop down to palladium to simplify the system at the start. Co-linearity problems are a pain and I would stay away from z-matrix coordinates.

Best,

Peter

Many of the above comments are useful and you should consider them. However in addition, given the symmetry of your structures you might also have the HOMO level degenerate with the LUMO, and during the scf iterations these 2 levels which have different symmetry might switch places and you 'll never reach convergence. So in practice you want to equally occupy them by using a Fermi-Dirac occupation factor, so that the level switching will not affect the resulting charge density. 

I am not very familiar with Gaussian, but there should be such an option to impose fractional occupations near the Fermi level as if the system was at a finite temperature.

This is also called "smeared" occupation. So either "smearing" or "finite temperature" might be the keywords you want to lookup in order to setup your runs.

Thank you all for your very useful suggestions.

The problem is resolved when i spent more time for each molecule. Closely monitor and keep track of the work. Extend the number of cycles allowed and run several opt jobs continuously appear to help to resolved the problem.

Thank you all and best wishes.

-Mer-