The choices of training molecules in the optimization of basis set

[Kaixuan Chen]

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

[hut...@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...@chem.uzh.ch
Winterthurerstrasse 190
CH-8057 Zürich, Switzerland
---------------------------------------------------------------

-----cp...@googlegroups.com wrote: -----
To: "cp2k" <cp...@googlegroups.com>
From: "Kaixuan Chen"
Sent by: cp...@googlegroups.com
Date: 03/31/2021 08:50PM
Subject: [CP2K:15054] The choices of training molecules in the optimization of basis set

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

[Kaixuan Chen]

Dear Juerg,

Thanks a lot for the sharing, I attach the output (for B97M_V) as below, same order of VO, VH5 and VOF3, and same order of basis TZV2P, TZVP, DZVP.

<<

 BASOPT| Information at iteration number:     490

 BASOPT| Training set | Combination | Rho difference | Condition num. | Time
 BASOPT| -----------------------------------------------------------------------

 BASOPT|        1     |       1     | 0.23283667E+01 | 0.49213932E+03 |   42.778
 BASOPT|        1     |       2     | 0.23290373E+01 | 0.43605921E+03 |   40.057
 BASOPT|        1     |       3     | 0.23355899E+01 | 0.88093685E+02 |   54.031
 BASOPT|        2     |       1     | 0.69033125E-04 | 0.32483764E+04 |   45.937
 BASOPT|        2     |       2     | 0.14196100E-03 | 0.30747941E+04 |   45.446
 BASOPT|        2     |       3     | 0.52714561E-03 | 0.23001768E+04 |   33.616
 BASOPT|        3     |       1     | 0.80156330E-04 | 0.73244742E+03 |   60.372
 BASOPT|        3     |       2     | 0.31742371E-03 | 0.63488031E+03 |   49.030
 BASOPT|        3     |       3     | 0.22712058E-02 | 0.18313406E+03 |   46.383
 BASOPT| -----------------------------------------------------------------------
 BASOPT| Total residuum value: -.21416233E+03

>>

The error between the reference and optimized basis is around 2~3, which I think is quite large. Thank you for the suggestion, I will check the reference input carefully.

Best,

Kai

[Kaixuan Chen]

Dear Juerg,

thank you very much for your suggestion, it is true that the reference wavefunction is not good enough. After I increase the convergence threshold, the problem disappears. Again, thanks a lot.

Best,

Kai