Hi,

first, you could try use the very last geometry printed in H.out as starting geometry for the geometry optimisation. The second step would be to increase the number of iterations using the %geom-block ( %geom maxiter 1000 end) in your input.

Hope this helps and good luck

Conrad

Conrad Hübler Thank you so much for this info. Will try this steps as soon as possible. Thank you again.

Firstly, HF is not DFT. Secondly, STO-3G is not a suitable basis set for studying compounds of d-elements. For this reason, I would recommend switching to the GGA DFT method, such as PBE, and using the 3-21G* basis set or, better, def2-SVP. Perhaps this will already solve the problem, because the structure that is obtained as a result of optimization is very different from the original one, which should be very alarming.

It is true what Alexey N. Masliy said about the method. In general, the basis set is to small for production run.

Using PBE with def2-SVP is a good choice, for a final single point you should go with an even bigger basis set (def2-TZVPP). - Of course, the equivalent LAN basis sets would do it as well.

As you are using orca, just try the composite methods PBEh.3c, B97-3c and/or r2SCAN-3c. They have well balanced basis sets and slightly re-parametrised functionals. I assume, that Ti doesn't make as much difficulties as other d elements with respect to self-interaction error.

Alexey N. Masliy and Conrad Hübler thank you for answering my question I really appreciate it . I'd like to ask if is it okay if I use B3LYP and LANL2DZ basis set as my computational method instead of using PBE and some basis sets you have mentioned for the geometry optimization of the clusters?

I think that Conrad's suggestion to use composite methods (any of the ones suggested) is a good idea, since the method and basis set are well matched in them. But if there are reasons why you want to use B3LYP/LanL2DZ, then I can’t give any serious objections. Another question - what is the purpose of your optimization? Are you going to explore nanoparticles or crystal?

Well, as for your original question - Orca saves the latest optimized geometry in a file with a name that matches the name of the original file, but with the .xyz extension (H.xyz in your case). It is also a good idea to read the finished MO guess from the .gbw file (H.gbw in your case). These are options:

%moinp "full_path_to_file.gbw"

%scf

Guess MORead

GuessMode CMATRIX

end

Alexey N. Masliy thank you for this recommendation I already tried this and my cluster have already converge thank you so much.