[CP2K-user] [CP2K:10993] Re: WARNINGS: The Kohn Sham matrix is not 100% occupied. & The cutoff radius is larger than half the minimal cell dimension

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

Hi

HFX calculations (periodic) are dominated by the time for calculating
the 4 center ERIs (integrals). CP2K stores these integrals in memory,
that is why after the first SCF step you see the enormous speed up.
The ERI calculations depends critically on the smallest exponents
of the primitive Gaussian in the basis sets. The original MOLOPT
basis sets are therefore not suited for this kind of calculation.
Even the SR version will be very slow, but already considerably faster.
This is why we have developed the ADMM method. It allows for a fast
HFX calculation at a somewhat reduced accuracy.

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: "katarína stančiaková" 
Sent by: cp... at googlegroups.com
Date: 11/29/2018 12:39PM
Subject: [CP2K:10993] Re: WARNINGS: The Kohn Sham matrix is not 100% occupied. &  The cutoff radius is larger than half the minimal cell dimension

Dear Vladimir and Frederick,
thank you for your help. I changed interaction potential to TRUNCATED and also omega to 0.15. In 4 days (on 48 cores) I was able to get at least the first  SCF optimization. When I look at the output file, I see that the first step of the SCF cycle takes almost all time and the other steps takes about 1 second each. Is this normal? I don't have much experience with these functionals so I don't know what kind of timings I should expect, but 4 days on 48 cores for a system of approx 20 atoms and 1 SCF cycle seems quite a lot to me....

Thank you
Kind regards,
Katarina

Dňa utorok, 20. novembra 2018 13:42:46 UTC+1 Frederick Stein napísal(-a):
Dear Katarina,

if you really need a shortrange potential, you either use a larger value for omega or a larger supercell. Maybe you should look for some variants of HSE with larger omega.

The cutoff radius of the potential is not related to the cutoff radius that cp2k assumes here. The cutoff is chosen such that the potential at that radius equals eps_schwarz (see function erfc_cutoff in file hfx_types.F of the code, if interested). The cutoff (for your system around 32.8 A) should be printed somewhere ("HFX_INFO| Cutoff Radius") but I do not know why it is not printed, here.

I hope I could help you.

Frederick 

Am Montag, 19. November 2018 22:30:50 UTC+1 schrieb katarína stančiaková:
Dear CP2K users,
I am trying to run HSE06 energy calculation, however, I always got the following warning and I've never been able to get even one step of SCF cycle within 48 hours (a periodic system of 22 atoms running on 48 cores).  

*** WARNING in hfx_types.F:1343 :: Periodic Hartree Fock calculation      ***
 *** requested with use of a truncated or shortrange potential. The cutoff ***
 *** radius is larger than half the minimal cell dimension. This may lead  ***
 *** to unphysical total energies. Reduce the cutoff radius in order to    ***
 *** avoid possible problems.                                              ***

To speed up the calculation I am using restart WF from PBE functional, but the same issues appear when I don't use it.  I've tested input files for various CUTOFF_RADIUS values (even very small ones), but without any result. Does anyone know how to solve this issue? 

Input files are attached
Thank you for any help!

Katarina

Dňa pondelok, 30. januára 2017 12:50:28 UTC+1 donaldducksdaughter napísal(-a):


  *** WARNING in hfx_types.F:1342 :: Periodic Hartree Fock calculation      ***
  *** requested with use of a truncated or shortrange potential. The cutoff ***
  *** radius is larger than half the minimal cell dimension. This may lead  ***
  *** to unphysical total energies. Reduce the cutoff radius in order to    ***
  *** avoid possible problems.    




 *** WARNING in hfx_energy_potential.F:600 :: The Kohn Sham matrix is not  ***
  *** 100% occupied. This may result in incorrect Hartree-Fock results. Try ***
  *** to decrease EPS_PGF_ORB and EPS_FILTER_MATRIX in the QS section. For  ***
  *** more information see FAQ: https://www.cp2k.org/faq:hfx_eps_warning    ***




Dear all,


I'm desperately trying to implement B3LYP on my system and to run a geometry optimisation. The calculations take horribly long (over two weeks) and still haven't finished. My error file is empty, however I seem to get the above two warnings, no matter what I do.


With respect to the first truncated potential warning, I'm using this truncated potential, because I have received an earlier warning telling me to use it and read about it on the cp2k/groups website. I'm using a cutoff of 15.0 angstrom with a box size of 30.0, which should be exactly half and fine. I'm uncertain whether I should just ignore it, because of the responses given in 

https://groups.google.com/forum/#!searchin/cp2k/The$20cutoff$20***$20$20***$20radius$20is$20larger$20than$20half$20the$20minimal$20cell$20dimension.$20This$20may$20lead$20$20***$20$20***$20to$20unphysical$20total$20energies.$20Reduce$20the$20cutoff$20radius$20in$20order$20to$20$20$20$20***$20$20***$20avoid$20possible$20problems.$20$20$20$20$20$20$20$20$20$20$20$20$20$20$20$20$20$20$20$20$20$20$20$20$20$20$20$20$20$20$20$20$20$20$20%7Csort:relevance/cp2k/g1sFck3SYF8/pVS-MPF-GQAJ. If I can ignore it, I reckon I should be using smaller auxiliary basis sets, as in cFIT types, rather than DZVP ones. Would that give me the same result? Additionally, I have already changed all my admms to shorter range basis sets to reduce the BSSE.


Regarding the second warning, I have already lowered the EPS Default from 1.0E-10 to 1.0E-12, but I'm still getting the same error. Also changing the poisson solver to WAVELET for my cubic 30x30x30 box would still give me the KS matrix as not 100 % occupied. https://groups.google.com/forum/#!searchin/cp2k/hfx_energy_potential%7Csort:relevance/cp2k/Xog7jpV5eRk/o254NK-8NGwJ


Any experience or help will be appreciated.


Many thanks in advance


Julia











  
  -- 
 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 "energy_HSE06.out" removed by Jürg Hutter/at/UZH]