D3 Dispersion corrections to gradients

[175116]

Hi,

I came across another D3 dispersion correction issue. In case I want to carry out geometrical optimization in which dispersion correction gradients are taken into account,  do I have to add -grad parameter next to the $disp3 -bj keyword? (-bj form BJ-damping). If I run optimization with "$disp3 -bj -grad", the gradients are little bit different than in case of "$disp3 -bj" (which is logical). However, when I look at the rdgrad output in case of "$disp3 -bj" (i.e. without -grad parameter) it seems rdgrad also calls disp3 procedure. Therefore it raises a question about double counting the dispersion correction gradients in case of "$disp3 -bj -grad".

Thank you very much indeed.

[uwe]

Hi,

good point. The options can be added to the $disp3 line in the control file the same way as in the stand-alone version of the original program of the Grimme group. I have added a post in the FAQ section about that:

http://www.turbo-forum.com/index.php?topic=625.0

There is no double counting of the DFT-D3 contribution to the gradients if you add -grad to the $disp3 keyword, but -grad is not needed if a gradient calculation is being performed.

Regards,

Uwe

[175116]

Hi, Uwe.

Thanky you very much for clarification! However, one issue is still unclear to me. On the assumption that $disp3 -grad is supposed to be identical with $disp (i.e. both ways result in supplementing dispersion gradients to pure DF gradients), why the gradient files differ in the two identical runs (see my first post in this thread). Is there any way how to explain this discrepancy?

Thanks in advance

[uwe]

Hi,

when I run the same job with $disp3 -bj and with $disp3 -bj -grad, I get identical gradients. If you get something else, please contact the Turbomole Support (turbomole at cosmologic.de).

Regards,

Uwe