CELL_OPT convergence issues (energy unstable becomes positive)

[illidanus... at gmail.com]

Hi Everyone,

I'm having a problem with several structures I am looking at which GEO_OPT 
successfully, but fail almost immediately when I modify the .restart files 
for a CELL_OPT run. The problem occurs also occurs when I recreate the 
CELL_OPT input file from scratch. The problem occurs with a cp2k 4.1 
version installed as a debian package, and for cp2k 4.1 and 5.0 (trunk) 
versions I compiled this weekend with no regtest errors apart from two 
minor energy discrepancies at the E-12 precision level in both versions.

Please see attached for a zip file for two structures where the 1GEO_OPT 
directories contains the input and put of the geometry run, while the 
2CELL_OPT directories contain the input (starting from the GEO_OPT restart 
file) and output of the CELL_OPT run.

#######################################
The error looks as follows:
#######################################

For the CBMZPN16 structure at optimization step 1:

----------------------------------- OT 
---------------------------------------

  Step     Update method      Time    Convergence         Total energy    
Change
  
------------------------------------------------------------------------------
     1 OT DIIS     0.15E+00    8.5     0.00016276     -*1027.6184205332* 
-1.03E+03
     2 OT DIIS     0.15E+00   13.8     0.00005874     -1027.6204951253 
-2.07E-03
     3 OT DIIS     0.15E+00   13.7     0.00002844     -1027.6207762772 
-2.81E-04
     4 OT DIIS     0.15E+00   13.8     0.00001684     -1027.6208072065 
-3.09E-05
     5 OT DIIS     0.15E+00   13.9     0.00000333     -1027.6208151200 
-7.91E-06
     6 OT DIIS     0.15E+00   13.8     0.00000144     -1027.6208154844 
-3.64E-07
     7 OT DIIS     0.15E+00   13.8     0.00000041     -1027.6208155528 
-6.85E-08

  *** SCF run converged in     7 steps ***

For the CBMZPN16 structure at optimization step 69:

 ----------------------------------- OT 
---------------------------------------

  Step     Update method      Time    Convergence         Total energy    
Change
  
------------------------------------------------------------------------------
     1 OT DIIS     0.15E+00    8.9     0.00519175   *4150284.2584174965*  
4.15E+06
     2 OT DIIS     0.15E+00   14.2     0.00323145   4150282.5769866798 
-1.68E+00
     3 OT DIIS     0.15E+00   14.3     0.00287989   4150282.5971149658  
2.01E-02
     4 OT DIIS     0.15E+00   14.3     0.00109589   4150282.3514816891 
-2.46E-01
     5 OT DIIS     0.15E+00   14.3     0.00060207   4150282.3092384278 
-4.22E-02
     6 OT DIIS     0.15E+00   14.3     0.00032125   4150282.2993985782 
-9.84E-03
     7 OT DIIS     0.15E+00   14.3     0.00020179   4150282.2963638688 
-3.03E-03
     8 OT DIIS     0.15E+00   14.3     0.00007977   4150282.2952897698 
-1.07E-03
     9 OT DIIS     0.15E+00   14.2     0.00003647   4150282.2947119516 
-5.78E-04

#######################################

In most of the structures I'm looking at the problem occurs within 10 cell 
optimization steps (I have 8 different examples of this problem - two 
attached). There are no problems when geometry optimizing in all cases. 
Also, I have successfully optimized many other structure in the past using 
these settings and I retested one of them successfully yesterday.

Any advice will be much appreciated. Apologies if something like this has 
been posted before but I really searched for examples.

Kind regards,

Manuel


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20170604/e8040305/attachment.htm>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: CBMZPN16-full_molecules-cp2k-D2.out
Type: application/octet-stream
Size: 402693 bytes
Desc: not available
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20170604/e8040305/attachment.obj>