[CP2K-user] [CP2K:15054] The choices of training molecules in the optimization of basis set

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

Hi

without more information on what "quite large" is, it is difficult
to give advice. My experience is, that if one molecule has a considerably
larger difference, that points to an error in the reference wavefunction
or the reference input file. This usually also causes an increase of
the difference during optimization.

Anyway, here my output for PBE0:
The order of reference molecules is: VO, VH5, VOF3
The order of basis sets is: TZV2P, TZVP, DZVP

 BASOPT| Information at iteration number:     100
 BASOPT| Training set | Combination | Rho difference | Condition num. | Time
 BASOPT| -----------------------------------------------------------------------
 BASOPT|        1     |       1     | 0.49136022E-03 | 0.36941292E+03 |    1.940
 BASOPT|        1     |       2     | 0.66719832E-03 | 0.29529497E+03 |    1.285
 BASOPT|        1     |       3     | 0.20109649E-02 | 0.14175422E+03 |    1.112
 BASOPT|        2     |       1     | 0.56818688E-03 | 0.55825387E+03 |   13.851
 BASOPT|        2     |       2     | 0.72028902E-03 | 0.47884816E+03 |    8.550
 BASOPT|        2     |       3     | 0.16966073E-02 | 0.45474553E+02 |    5.470
 BASOPT|        3     |       1     | 0.66582112E-03 | 0.81643655E+03 |    8.729
 BASOPT|        3     |       2     | 0.10039605E-02 | 0.52118329E+03 |    5.611
 BASOPT|        3     |       3     | 0.43020517E-02 | 0.20016463E+03 |    6.313
 BASOPT| -----------------------------------------------------------------------
 BASOPT| Total residuum value: -.22424284E+04

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: 03/31/2021 08:50PM
Subject: [CP2K:15054] The choices of training molecules in the optimization of basis set


Dear all,

I am performing the optimization of new basis set for the functional B97M_V. I chose some small molecules provided in the "https://github.com/juerghutter/BASIS/tree/master/MOLSET". For example, I used the training molecules of VH5, VO and VOF3 for the optimization of basis set for element V. However, the rho difference (electron density difference) of VO between the reference basis set and optimized basis set is quite large, where the other two are relatively small. Should I remove VO and run a new optimization calculation? And should I include more training molecules?
I will try to compare the results with and without VO, but any suggestion regarding this is highly welcome. thanks a lot in advance.

Best,
Kai
  
  -- 
 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/2526f7b8-15d8-4533-a158-c88a6c60c3cbn%40googlegroups.com.