[CP2K:5567] Re: Geometry optimization by using LIBXC caused no SCF iteration with error NaN

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

Hi

so this is the GAPW bug for meta-functionals. As I have mentioned
before, there is an inconsistency in the way the kinetic energy density
is handled in the GAPW one-center terms.

regards

Juerg 
--------------------------------------------------------------
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: cp... at googlegroups.com
From: Geoffrey Wood 
Sent by: cp... at googlegroups.com
Date: 07/28/2014 10:43PM
Subject: Re: [CP2K:5567] Re: Geometry optimization  by using LIBXC caused no SCF iteration with error NaN

I ran some others these are the results:
 
               Key Words        %HF    CP2K    G09    diff (kJ/mol)        XC_GGA_XC_HCTH_407        0    -76.40895    -76.40918    -0.6        XC_HYB_GGA_XC_B97_1        0.21    -76.39399    -76.39374    0.7        XC_MGGA_C_M06_L        0    -67.68914                XC_MGGA_X_M06_L        0    -76.04416                XC_MGGA_C_M06_L    XC_MGGA_X_M06_L    0    -76.43029    -76.41007    53.1        XC_HYB_MGGA_XC_MPWB1K        0.44    SCF crash    -76.38142        
 
 
The MGGA seems to be causing the problems.
 
 
Thanks, Geoff.
 
 

On Monday, July 28, 2014 11:59:38 AM UTC-4, Geoffrey Wood wrote:
 
Hello
 
I am using GAPW and the cutoff is 1000 (not 10000, sorry for the error).  I used such a large cutoff because I was testing the convergence behavior of the energy wrt cutoff (I tried 200, 400, 800, 1000 and 4000) in the end I reported the 1000 numbers above as it seemed to give convergence to ~10^-5 Hartree, At this level the difference between the B3LYP and PBE0 energies between the two programs is very small.  Knowing that GAPW is a different approximation to using the Gaussian basis sets this difference is to be expected (as reported in the original paper). Using the same settings the difference between the M06 energies seems to be rather large, which is the concern I have. ll try the M06L functional and provide feedback.
 
Thanks, Geoff.
 
 
 
 
 
 

On Monday, July 28, 2014 10:44:49 AM UTC-4, jgh wrote:Hi 
 
Are you running GPW or GAPW? With that cutoff one never knows. 
 
M06L might be another good test. 
 
Juerg  
-------------------------------------------------------------- 
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: cp... at googlegroups.com 
From: Geoffrey Wood  
Sent by: cp... at googlegroups.com 
Date: 07/28/2014 03:27PM 
Subject: Re: [CP2K:5563] Re: Geometry optimization  by using LIBXC caused no SCF iteration with error NaN 
 
Hello- 
  
Thanks for everyone's comments. I discovered that part of the problem I was having was an error in my "arch" file. With that fixed I have compiled and linked CP2K (SVN 14141) libint (1.1.4) and libxc (2.1.0) and tested the XC libraries.  These are the results: 
  
system: 
CUTOFF=10000 
  
    O   0.000000    0.000000   -0.065587 
    H   0.000000   -0.757136    0.520545 
    H   0.000000    0.757136    0.520545 
    &KIND H 
      BASIS_SET 6-31Gxx 
      POTENTIAL ALL 
    &END KIND 
    &KIND O 
      BASIS_SET 6-31Gxx 
      POTENTIAL ALL 
    &END KIND 
 
  
XC_HYB_GGA_XC_PBEH (+25% HF x),  CP2k:  -76.33577878183063  another popular QM code:  -76.3357457 
XC_HYB_GGA_XC_B3LYP (+20% HF x), CP2k: -76.41801937267151  another popular QM code: -76.4180549 
XC_MGGA_C_M06 (+27% HF x), CP2K: -70.00722600780877 
XC_MGGA_X_M06 (+27% HF x) CP2K: -76.02248099520743 
XC_MGGA_X_M06 XC_MGGA_C_M06 (+27% HF x): CP2K:  -76.42254621238931  another popular QM code: -76.3842635 
XC_HYB_MGGA_XC_M06 (+27% HF x): CP2K (crashes on SCF) 
XC_HYB_MGGA_XC_M06 (no HF x included): CP2K (crashes on SCF) 
  
So there still seems to be some problems with M06. I.e. the hybrid-meta XC doesn't seem to work and trying to combine the X and C parts together gives the incorrect energy.  Thoughts? 
  
Geoff 
  
  
 
On Friday, July 25, 2014 8:22:08 AM UTC-4, Geoffrey Wood wrote: 
Hi Matthias and Michael.. 
  
Thanks for your comments. I have previously run cp2k 2.5.1 with libint 1.1.5 and tested it without any problems i.e. correct energies, gradients 2nd derivatives, BOMD using GGAs, hyrbrid GGAs, with and without libxc.  It is only with the current trunk that I'm running into problems.  I will re-link with libint 1.1.4 and provide feedback. 
  
Thanks again. 
  
Geoff. 
  
  
 
On Thursday, July 24, 2014 10:58:59 AM UTC-4, Michael wrote:On Thu, Jul 24, 2014 at 07:43:03AM -0700, Matthias Krack wrote:  
> you write that you compiled libint v1.1.5, but CP2K is only validated for   
> libint v1.1.4.  
  
As a data point, Debian/Ubuntu ships libint-1.1.4 and we have not seen  
any problems with that, including building cp2k-2.5.1 against it.    
  
Michael  
   
  --  
 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+... at googlegroups.com. 
 To post to this group, send email to cp... at googlegroups.com. 
 Visit this group at http://groups.google.com/group/cp2k. 
 For more options, visit https://groups.google.com/d/optout. 
  
  
  -- 
 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 http://groups.google.com/group/cp2k.
 For more options, visit https://groups.google.com/d/optout.