[CP2K:10530] wB97X-V in CP2K

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

Thank you. I have updated the test.

best 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: "Yuezhi Mao" 
Sent by: cp... at googlegroups.com
Date: 07/14/2018 04:17AM
Subject: [CP2K:10530] wB97X-V in CP2K

Dear CP2K developers,

I was trying to figure out how to run wB97X-V calculations with CP2K, as AIMD with this functional using CP2K was reported earlier this year. By testing the binding energy for the water dimer in S22 set (the results are summarized in the attached .log file), I found the follows:
(i) The setup of this functional in tests/QS/regtest-hybrid-4/wB97X-V.inp seems incorrect. The line "FRACTION 0.167" should be deleted. As a mixed Coulomb operator is used and the 1/r and erf(wr)/r operators have been scaled separately, there is definitely no need to apply an additional scaling factor to scale down the exact exchange. From the output generated with the same setup as this regtest ("water_dimer_gapw_wB97X-V_regtest_def2-qzvp.out"), we can find that HF exchange energy (as well as the total energy) is particularly small.

(ii) Without the global scaling factor, the total energy given by CP2K (using the built-in parameterization) looks too low for this system (the corresponding output is "water_dimer_gapw_wB97X-V_regtest_def2-qzvp.out") compared to the q-chem reference. It was reported before that wB97X-V, as a parameterization of BECKE97, was not implemented correctly because it was not attenuated as an SR functional (see a previous discussion on this issue: https://groups.google.com/forum/#!msg/cp2k/JEvnnfAEgPY/CFwlbsroSOcJ). We suspected that this problem has not been fixed since then.

(iii) Replacing the built-in implementation of the XC functional with the counterpart in libxc, the results produced by CP2K (especially the GAPW/def2-QZVP calculation) are in great agreement with the Q-Chem numbers. Therefore, I believe that the setup in "water_dimer_gapw_wB97X-V_def2-qzvp.inp" is the correct way of running this functional.

I suggest that the regtest job should be modified, as it is often the first thing a user would refer to when one wants to try out some new features. The current one turns out to be misleading. Also, I think the built-in implementation of this functional should be either fixed or disabled, as it produces incorrect results.

Best regards,

Yuezhi Mao
Postdoctoral scholar
Department of Chemistry, 
Stanford University  
  -- 
 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... at googlegroups.com.
 To post to this group, send email to cp... at googlegroups.com.
 Visit this group at https://groups.google.com/group/cp2k.
 For more options, visit https://groups.google.com/d/optout.
 

[attachment "wB97X-V_water_dimer.log" removed by Jürg Hutter/at/UZH]
[attachment "tests.zip" removed by Jürg Hutter/at/UZH]