[CP2K-user] [CP2K:14859] Pseudopoentials generation for density functional B97M_V

[hut... at chem.uzh.ch]

Hi

sometimes it might not be possible to reduce the function value
more than that. Check that the eigenvalues of valence states
and core states are well reproduced. You might have to live with
larger errors for the unoccupied states.
See the POTETIAL files on GITHUB for error values from the optimized
pseudos.
The final and only test of the pseudos is always in applications,
but you can try to test relative accuracy with several atomic states.

regards

Juerg Hutter
--------------------------------------------------------------
Juerg Hutter                         Phone : ++41 44 635 4491
Institut für Chemie C                FAX   : ++41 44 635 6838
Universität Zürich                   E-mail: hut... at chem.uzh.ch
Winterthurerstrasse 190
CH-8057 Zürich, Switzerland
---------------------------------------------------------------

-----cp... at googlegroups.com wrote: -----
To: "cp2k" <cp... at googlegroups.com>
From: "Kaixuan Chen" 
Sent by: cp... at googlegroups.com
Date: 03/03/2021 02:55AM
Subject: [CP2K:14859] Pseudopoentials generation for density functional B97M_V

Hi,
I am trying to use the cp2k module ATOM to generate pseudopotentials for 3d transition metals with the use of density functional B97M_V. The GTH pseudopotentials file from "https://github.com/juerghutter/GTH" are taken as initial guess.
As recommended, I restarted multiple runs with adjusted parameters (step_size, step_size_scaling, max_fun, max_init) until a small value of function (error) is obtained (less than 1.0). Unfortunately, the optimization of Cr seems to have some problem (input file attached), that is, no mater how I alter the parameters, a relatively high value of function appears (around 7.0).
 I would really appreciate your answers for the following questions:
1. Does any one have an idea on how to resolve this problem, that is, reduce the value of function for Cr optimization?
2. In addition to the value of function, is there any other way to benchmark the pseudopotentials that are generated?

Thanks,
Kaixuan Chen  
  -- 
 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 cp... at googlegroups.com.
 To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/23a985c7-df0c-4d0d-90d3-b165a899eaf2n%40googlegroups.com.
 

[attachment "cp2k.inp" removed by Jürg Hutter/at/UZH]