Semi-empirical convergence

[Виктор Овсянников]

Dear CP2K users and developers,

I am trying to minimize charged fullerene cage using AM1, but it stucks and cannot converge with default SCF parameters and BFGS minimizer, and if I try to use different mixing methods, CP2K gives me segmentation fault. Input, output, and xyz files are attached.

[Anton S. Lytvynenko]

Dear Victor,

It looks like your SCF is very far from convergence, please try to increase max_scf to say 300, and decrease alpha to 0.2 (if it doesn't help -- to 0.1).

You should try to find the parameters yielding your SCF converged on the first geometry optimization step (preferably) or at least on the 2nd-3rd.

It doesn't depend on the geometry optimization parameters -- if your SCF did not exhibit at least fair convergence at the first steps, you gradients, and, subsequently, the geometry, are completely irrelevant.

Hope that helps.

Yours,

Anton

27.05.20 17:55, Виктор Овсянников пише:

Dear CP2K users and developers,

I am trying to minimize charged fullerene cage using AM1, but it stucks and cannot converge with default SCF parameters and BFGS minimizer, and if I try to use different mixing methods, CP2K gives me segmentation fault. Input, output, and xyz files are attached.

--
You received this message because you are subscribed to the Google Groups "cp2k" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+uns...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/b3d696fc-dfda-442f-b2e7-63ff39ba4a78%40googlegroups.com.

[Виктор Овсянников]

Thank you for your help, alpha 0.1 worked just fine

ср, 27 мая 2020 г. в 18:13, Anton S. Lytvynenko <tenebrosu...@gmail.com>:

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/f6d9f62c-a62b-616f-27db-0c8e91ff52bd%40gmail.com.