Possible bug: ScfConvergenceWarning

Total energy, geometry optimization, DFT+U, spin....

Moderator: bguster

Locked
sheng
Posts: 64
Joined: Fri Apr 11, 2014 3:44 pm

Possible bug: ScfConvergenceWarning

Post by sheng » Mon Mar 21, 2016 3:09 am

I discovered a weird behaviour during structural relaxation of a slab geometry. The convergence in an iteration simply stopped even though the toldff requirement is not yet met. The toldff is my case is set to 5.0d-6.

The following is from the log file (norm-conserving potentials are used):

Code: Select all

 ITER STEP NUMBER    15
 vtorho : nnsclo_now=  1, note that nnsclo,dbl_nnsclo,istep=  0 0 15
 Total charge density [el/Bohr^3]
,     Maximum=    1.3567E+00  at reduced coord.    0.4222    0.5000    0.2950
,     Minimum=    7.3609E-06  at reduced coord.    0.9778    0.0000    0.0000
,  Integrated=    1.3600E+02
 ETOT 15  -458.70158659597     5.786E-10 2.329E-07 1.636E-05 5.623E-06 2.183E-03
 scprqt: <Vxc>= -3.2009586E-01 hartree

 Pulay update with  7 previous iterations:
 mixing of old trial potential : alpha(m:m-4)=  0.819      0.925     -0.398     -0.494      0.145

 ITER STEP NUMBER    16
 vtorho : nnsclo_now=  1, note that nnsclo,dbl_nnsclo,istep=  0 0 16
 Total charge density [el/Bohr^3]
,     Maximum=    1.3567E+00  at reduced coord.    0.4222    0.5000    0.2950
,     Minimum=    7.3610E-06  at reduced coord.    0.9778    0.0000    0.0000
,  Integrated=    1.3600E+02
 ETOT 16  -458.70158659519     7.804E-10 4.196E-07 5.951E-06 1.171E-05 2.184E-03
 scprqt: <Vxc>= -3.2009586E-01 hartree

 Cartesian components of stress tensor (hartree/bohr^3)
  sigma(1 1)=  6.72003171E-05  sigma(3 2)=  1.85397380E-07
  sigma(2 2)=  3.17760047E-05  sigma(3 1)= -1.49007920E-09
  sigma(3 3)=  5.66625408E-05  sigma(2 1)=  6.75298050E-08


--- !ScfConvergenceWarning
message: |
    nstep 1000 was not enough SCF cycles to converge.
...

 scprqt:  WARNING -
  nstep= 1000 was not enough SCF cycles to converge;
  maximum force difference=  1.171E-05 exceeds toldff=  5.000E-06

   1
   2
   3
   4
   5
   6
   7
   8
   9
  10
  11
  12
  13
  14
  15
  16
  17
 ioarr: writing density data
ioarr: file name is BaTiO3o_TIM25_DEN
ioarr: data written to disk file BaTiO3o_TIM25_DEN
 ihist @ write_md_hist          25
 mxhist @ write_md_hist        1000
 OPEN NETCDF FILE


The relaxation then proceeds to another iteration using the unconverged result, leading to to an huge deviation to the tolmxf value, essentially causing the whole structure to be relaxed from the start.

Below is the output from the commad "grep ETOT log":

Code: Select all

 ETOT  1  -458.70149552702    -4.587E+02 1.031E-06 6.161E+00 2.415E-02 2.373E-02
 ETOT  2  -458.70158165706    -8.613E-05 8.520E-07 2.767E-01 2.801E-02 4.284E-03
 ETOT  3  -458.70158779408    -6.137E-06 6.937E-07 2.274E-02 5.998E-03 2.568E-03
 ETOT  4  -458.70158908919    -1.295E-06 4.179E-07 4.555E-03 2.349E-03 7.276E-04
 ETOT  5  -458.70158964835    -5.592E-07 5.684E-07 6.488E-04 4.616E-04 1.092E-03
 ETOT  6  -458.70158967604    -2.768E-08 3.918E-07 4.288E-05 3.289E-04 1.055E-03
 ETOT  7  -458.70158968403    -7.990E-09 5.133E-07 2.070E-05 9.211E-05 1.038E-03
 ETOT  8  -458.70158968433    -3.040E-10 3.764E-07 1.704E-05 2.534E-05 1.063E-03
 ETOT  9  -458.70158968471    -3.755E-10 4.728E-07 2.579E-07 3.376E-05 1.052E-03
 ETOT 10  -458.70158968468     2.416E-11 3.631E-07 4.542E-07 4.100E-06 1.054E-03
 ETOT 11  -458.70158968463     4.968E-11 4.348E-07 8.096E-08 2.558E-06 1.056E-03
 ETOT  1  -458.70153059105    -4.587E+02 2.700E-06 3.978E+00 1.588E-02 1.615E-02
 ETOT  2  -458.70158305948    -5.247E-05 1.013E-06 3.824E-02 1.542E-02 3.542E-03
 ETOT  3  -458.70158618880    -3.129E-06 3.478E-07 1.254E-02 3.900E-03 6.145E-04
 ETOT  4  -458.70158640010    -2.113E-07 5.887E-07 3.341E-03 1.812E-03 1.970E-03
 ETOT  5  -458.70158661256    -2.125E-07 2.904E-07 3.941E-04 4.686E-04 1.927E-03
 ETOT  6  -458.70158664298    -3.041E-08 5.321E-07 1.515E-04 1.060E-04 1.981E-03
 ETOT  7  -458.70158661404     2.894E-08 2.661E-07 9.428E-05 1.542E-04 2.135E-03
 ETOT  8  -458.70158661240     1.642E-09 5.115E-07 5.295E-05 3.102E-05 2.144E-03
 ETOT  9  -458.70158660910     3.300E-09 2.510E-07 2.241E-05 2.572E-05 2.154E-03
 ETOT 10  -458.70158660695     2.145E-09 4.909E-07 2.645E-06 2.490E-05 2.158E-03
 ETOT 11  -458.70158661018    -3.230E-09 2.421E-07 1.577E-05 3.369E-05 2.148E-03
 ETOT 12  -458.70158659345     1.674E-08 4.609E-07 1.278E-04 8.531E-05 2.190E-03
 ETOT 13  -458.70158660024    -6.792E-09 2.360E-07 7.775E-06 4.833E-05 2.173E-03
 ETOT 14  -458.70158659655     3.688E-09 4.386E-07 2.291E-05 8.910E-06 2.182E-03
 ETOT 15  -458.70158659597     5.786E-10 2.329E-07 1.636E-05 5.623E-06 2.183E-03
 ETOT 16  -458.70158659519     7.804E-10 4.196E-07 5.951E-06 1.171E-05 2.184E-03
 ETOT  1  -458.70147352986    -4.587E+02 3.343E-06 3.209E+00 8.044E-02 7.825E-02
 ETOT  2  -458.70149145204    -1.792E-05 2.878E-06 5.480E-01 1.657E-02 7.814E-02
 ETOT  3  -458.70146480407     2.665E-05 2.793E-06 4.166E-01 3.020E-03 8.116E-02
 ETOT  4  -458.70146315380     1.650E-06 2.185E-06 3.278E-01 9.645E-04 8.124E-02
 ETOT  5  -458.70145440049     8.753E-06 3.014E-06 6.542E-02 2.570E-03 8.190E-02
 ETOT  6  -458.70145647509    -2.075E-06 2.123E-06 1.940E-02 2.825E-03 8.143E-02
 ETOT  7  -458.70145645575     1.934E-08 2.610E-06 1.200E-04 9.274E-04 8.152E-02
 ETOT  8  -458.70145625346     2.023E-07 1.922E-06 1.781E-04 4.723E-05 8.153E-02
 ETOT  9  -458.70145560527     6.482E-07 2.430E-06 1.072E-03 1.057E-04 8.161E-02

which shows the force differences increases to an order of -2 from -3, due to the ScfConvergenceWarning error.

Strangely I can avoid the problem by restarting the relaxation from coordinates of previously converged iteration, but the error can appear now and then.

On the other hand, using the PAW I have also encountered this problem, but this time the ScfConvergenceWarning did not explicit appear. The toldff convergence in an iteration is not reached, but the code simply proceed to another iteration as though nothing unusual happen.

Code: Select all

 ETOT  1  -1374.6908884409    -1.375E+03 1.734E-07 2.158E-08 9.336E-04 8.229E-04
 ETOT  2  -1374.6908625437     2.590E-05 1.381E-07 1.245E-08 2.186E-04 6.043E-04
 ETOT  3  -1374.6908498869     1.266E-05 2.058E-07 2.461E-07 9.075E-05 5.135E-04
 ETOT  4  -1374.6908472461     2.641E-06 1.103E-07 3.943E-09 4.963E-05 5.631E-04
 ETOT  5  -1374.6908239901     2.326E-05 1.647E-07 2.729E-09 1.282E-04 4.349E-04
 ETOT  6  -1374.6908047443     1.925E-05 8.821E-08 1.575E-08 1.004E-05 4.450E-04
 ETOT  7  -1374.6907927434     1.200E-05 1.319E-07 2.438E-09 1.089E-04 3.360E-04
 ETOT  8  -1374.6907598415     3.290E-05 7.052E-08 2.688E-09 1.075E-04 2.285E-04
 ETOT  9  -1374.6907554376     4.404E-06 1.056E-07 1.327E-09 4.615E-05 1.824E-04
 ETOT 10  -1374.6907467425     8.695E-06 5.639E-08 1.627E-09 1.024E-04 1.204E-04
 ETOT 11  -1374.6907442919     2.451E-06 8.454E-08 4.458E-10 4.208E-05 1.201E-04
 ETOT 12  -1374.6907395508     4.741E-06 4.511E-08 1.189E-09 3.426E-05 1.215E-04
 ETOT 13  -1374.6907353028     4.248E-06 6.771E-08 2.011E-10 5.666E-05 1.218E-04
 ETOT 14  -1374.6907308127     4.490E-06 3.610E-08 3.324E-10 2.004E-05 1.226E-04
 ETOT 15  -1374.6907269428     3.870E-06 5.422E-08 3.146E-10 1.086E-06 1.224E-04
 ETOT 16  -1374.6907238048     3.138E-06 3.033E-08 3.647E-10 8.627E-06 1.229E-04
 ETOT 17  -1374.6907201591     3.646E-06 4.267E-08 5.130E-10 1.119E-05 1.225E-04
 ETOT 18  -1374.6907249081    -4.749E-06 2.893E-08 4.889E-10 2.414E-05 1.217E-04
 ETOT 19  -1374.6907261053    -1.197E-06 3.281E-08 3.524E-10 8.945E-06 1.218E-04
 ETOT 20  -1374.6907274970    -1.392E-06 2.759E-08 3.048E-10 2.928E-06 1.220E-04
 ETOT  1  -1374.6905318140    -1.375E+03 2.627E-08 4.490E-08 8.276E-04 8.868E-04
 ETOT  2  -1374.6893065979     1.225E-03 2.000E-04 8.404E-04 7.527E-02 7.616E-02
 ETOT  3  -1374.6905221834    -1.216E-03 8.066E-05 6.579E-06 7.514E-02 1.016E-03
 ETOT  4  -1374.6905355955    -1.341E-05 1.379E-05 2.411E-07 1.106E-04 1.127E-03
 ETOT  5  -1374.6905785950    -4.300E-05 1.081E-05 1.108E-07 2.637E-04 8.629E-04
 ETOT  6  -1374.6906113346    -3.274E-05 1.873E-06 2.570E-08 2.359E-05 8.393E-04
 ETOT  7  -1374.6906886135    -7.728E-05 1.481E-06 9.734E-08 3.756E-05 8.017E-04
 ETOT  8  -1374.6906653930     2.322E-05 3.761E-07 3.424E-08 4.263E-05 7.591E-04
 ETOT  9  -1374.6906458968     1.950E-05 2.392E-07 3.018E-08 6.899E-06 7.660E-04
 ETOT 10  -1374.6906068161     3.908E-05 1.392E-07 2.359E-08 2.109E-05 7.449E-04


Code: Select all

 ETOT 19  -1374.6907261053    -1.197E-06 3.281E-08 3.524E-10 8.945E-06 1.218E-04
 scprqt: <Vxc>= -2.8344701E-01 hartree
 
 Pulay update with  7 previous iterations:
 mixing of old trial potential : alpha(m:m-4)=  0.886     -0.945     -0.358     -0.483      0.837
 
--- !WARNING
message: |
    Density went too small (lower than xc_denpos) at206735 points
    and was set to xc_denpos=  1.00E-14. Lowest was  -0.13E-04.
    Likely due to too low boxcut or too low ecut for pseudopotential core charge.
src_file: Subroutine Unknown
src_line: 0
...
 
 average electrostatic smooth potential [Ha] , [eV]      0.000000        0.000000
 
 ****** Psp strength Dij in Ha (atom      1) *****
   0.65870  -0.09920   0.00000  -0.00215  -0.00029   0.00000  -0.00135  -0.00018   0.00000   0.00000   0.00032  -0.00002 ...
  -0.09920   0.01593   0.00000   0.00016   0.00002   0.00000  -0.00016  -0.00002   0.00000   0.00000  -0.00002   0.00000 ...
   0.00000   0.00000  -0.08877   0.00000   0.00000   0.00645   0.00000   0.00000  -0.00008  -0.00058   0.00000   0.00000 ...
  -0.00215   0.00016   0.00000  -0.08802  -0.00003   0.00000   0.00700  -0.00002   0.00000   0.00000  -0.00066  -0.00008 ...
  -0.00029   0.00002   0.00000  -0.00003  -0.08885   0.00000  -0.00002   0.00639   0.00000   0.00000   0.00004  -0.00057 ...
   0.00000   0.00000   0.00645   0.00000   0.00000  -0.97409   0.00000   0.00000  -0.00009  -0.00064   0.00000   0.00000 ...
  -0.00135  -0.00016   0.00000   0.00700  -0.00002   0.00000  -0.97285  -0.00004   0.00000   0.00000  -0.00074  -0.00009 ...
  -0.00018  -0.00002   0.00000  -0.00002   0.00639   0.00000  -0.00004  -0.97423   0.00000   0.00000   0.00005  -0.00064 ...
   0.00000   0.00000  -0.00008   0.00000   0.00000  -0.00009   0.00000   0.00000  -0.02311   0.00000   0.00000   0.00000 ...
   0.00000   0.00000  -0.00058   0.00000   0.00000  -0.00064   0.00000   0.00000   0.00000  -0.02297   0.00000   0.00000 ...
   0.00032  -0.00002   0.00000  -0.00066   0.00004   0.00000  -0.00074   0.00005   0.00000   0.00000  -0.02294   0.00000 ...
  -0.00002   0.00000   0.00000  -0.00008  -0.00057   0.00000  -0.00009  -0.00064   0.00000   0.00000   0.00000  -0.02298 ...
   ...  only 12  components have been written...
 
 
 ITER STEP NUMBER    20
 vtorho : nnsclo_now=  1, note that nnsclo,dbl_nnsclo,istep=  0 0 20
 
 *********** RHOIJ (atom   1) **********
   2.02793  -0.02439   0.00000   0.04437   0.00571   0.00000  -0.00510  -0.00071   0.00000   0.00000   0.02103  -0.00272 ...
  -0.02439   0.10647   0.00000  -0.09058  -0.01260   0.00000   0.00441   0.00076   0.00000   0.00000  -0.01480   0.00282 ...
   0.00000   0.00000   2.05449   0.00000   0.00000  -0.00722   0.00000   0.00000  -0.01549  -0.07903   0.00000   0.00000 ...
   0.04437  -0.09058   0.00000   1.99413   0.00077   0.00000  -0.00047  -0.00017   0.00000   0.00000  -0.07558  -0.00988 ...
   0.00571  -0.01260   0.00000   0.00077   2.05872   0.00000  -0.00017  -0.00762   0.00000   0.00000   0.00647  -0.07464 ...
   0.00000   0.00000  -0.00722   0.00000   0.00000   0.00104   0.00000   0.00000   0.00092   0.00380   0.00000   0.00000 ...
  -0.00510   0.00441   0.00000  -0.00047  -0.00017   0.00000   0.00066   0.00001   0.00000   0.00000   0.00320   0.00049 ...
  -0.00071   0.00076   0.00000  -0.00017  -0.00762   0.00000   0.00001   0.00103   0.00000   0.00000  -0.00037   0.00352 ...
   0.00000   0.00000  -0.01549   0.00000   0.00000   0.00092   0.00000   0.00000   0.09585  -0.00207   0.00000   0.00000 ...
   0.00000   0.00000  -0.07903   0.00000   0.00000   0.00380   0.00000   0.00000  -0.00207   0.07396   0.00000   0.00000 ...
   0.02103  -0.01480   0.00000  -0.07558   0.00647   0.00000   0.00320  -0.00037   0.00000   0.00000   0.05694   0.00463 ...
  -0.00272   0.00282   0.00000  -0.00988  -0.07464   0.00000   0.00049   0.00352   0.00000   0.00000   0.00463   0.07077 ...
   ...  only 12  components have been written...
 
 *********** RHOIJ (atom   6) **********
   2.02880  -0.01242   0.00000   0.00566   0.00576   0.00000  -0.00069  -0.00068   0.00000   0.00000   0.00416  -0.00041 ...
  -0.01242   0.09777   0.00000  -0.01208  -0.01200   0.00000   0.00067   0.00062   0.00000   0.00000  -0.00348   0.00041 ...
   0.00000   0.00000   2.05547   0.00000   0.00000  -0.00641   0.00000   0.00000  -0.01152  -0.01095   0.00000   0.00000 ...
   0.00566  -0.01208   0.00000   2.04923   0.00001   0.00000  -0.00562  -0.00002   0.00000   0.00000  -0.00880  -0.01128 ...
   0.00576  -0.01200   0.00000   0.00001   2.05835   0.00000  -0.00002  -0.00673   0.00000   0.00000   0.00358  -0.01127 ...
   0.00000   0.00000  -0.00641   0.00000   0.00000   0.00107   0.00000   0.00000   0.00057   0.00062   0.00000   0.00000 ...
  -0.00069   0.00067   0.00000  -0.00562  -0.00002   0.00000   0.00103   0.00000   0.00000   0.00000   0.00033   0.00051 ...
  -0.00068   0.00062   0.00000  -0.00002  -0.00673   0.00000   0.00000   0.00107   0.00000   0.00000  -0.00014   0.00065 ...
   0.00000   0.00000  -0.01152   0.00000   0.00000   0.00057   0.00000   0.00000   0.07478  -0.00065   0.00000   0.00000 ...
   0.00000   0.00000  -0.01095   0.00000   0.00000   0.00062   0.00000   0.00000  -0.00065   0.06440   0.00000   0.00000 ...
   0.00416  -0.00348   0.00000  -0.00880   0.00358   0.00000   0.00033  -0.00014   0.00000   0.00000   0.04968   0.00103 ...
  -0.00041   0.00041   0.00000  -0.01128  -0.01127   0.00000   0.00051   0.00065   0.00000   0.00000   0.00103   0.06425 ...
   ...  only 12  components have been written...
 
 Total charge density [el/Bohr^3]
,     Maximum=    1.3998E+00  at reduced coord.    0.5156    0.5000    0.2135
,     Minimum=    6.9363E-10  at reduced coord.    0.4844    0.0000    0.0260
,  Integrated=    4.1600E+02
 ETOT 20  -1374.6907274970    -1.392E-06 2.759E-08 3.048E-10 2.928E-06 1.220E-04
 scprqt: <Vxc>= -2.8344701E-01 hartree
 
 Cartesian components of stress tensor (hartree/bohr^3)
  sigma(1 1)= -5.29682400E-05  sigma(3 2)=  0.00000000E+00
  sigma(2 2)= -5.71511683E-05  sigma(3 1)=  0.00000000E+00
  sigma(3 3)= -7.45734300E-06  sigma(2 1)=  0.00000000E+00
 
   1  1 51 51
   2  2 47 47
   3  3 48 48
   4  4 52 52
   5  5 50 50
   6  6 46 46
   7  7 42 42
   8  8 43 43
   9  9 49 49
  10 10 45 45
  11 11 41 41
  12 12 37 37
  13 13 38 38
  14 14 44 44
  15 15 40 40
  16 16 36 36
  17 17 32 32
  18 18 33 33
  19 19 39 39
  20 20 35 35
  21 21 31 31
  22 22 27 27
  23 23 28 28
  24 24 34 34
  25 25 30 30
  26 26 26 26
  27 27 22 22
  28 28 23 23
  29 29 29 29
  30 30 25 25
  31 31 21 21
  32 32 17 17
  33 33 18 18
  34 34 24 24
  35 35 20 20
  36 36 16 16
  37 37 12 12
  38 38 13 13
  39 39 19 19
  40 40 15 15
  41 41 11 11
  42 42  7  7
  43 43  8  8
  44 44 14 14
  45 45 10 10
  46 46  6  6
  47 47  2  2
  48 48  3  3
  49 49  9  9
  50 50  5  5
  51 51  1  1
  52 52  4  4
 ioarr: writing density data
ioarr: file name is BaTiO3o_TIM17_DEN
ioarr: data written to disk file BaTiO3o_TIM17_DEN
 
 PAW TEST:
 ==== Compensation charge inside spheres ============
 The following values must be close to each other ...
 Compensation charge over spherical meshes =     39.042477960097763
 Compensation charge over fine fft grid    =     39.042522578943689
 
 ==== Results concerning PAW augmentation regions ====
 
 Total pseudopotential strength Dij (hartree):
 Atom #  1
   0.65870  -0.09920   0.00000  -0.00215  -0.00029   0.00000  -0.00135  -0.00018   0.00000   0.00000   0.00032  -0.00002 ...
  -0.09920   0.01593   0.00000   0.00016   0.00002   0.00000  -0.00016  -0.00002   0.00000   0.00000  -0.00002   0.00000 ...
   0.00000   0.00000  -0.08877   0.00000   0.00000   0.00645   0.00000   0.00000  -0.00008  -0.00058   0.00000   0.00000 ...
  -0.00215   0.00016   0.00000  -0.08802  -0.00003   0.00000   0.00700  -0.00002   0.00000   0.00000  -0.00066  -0.00008 ...
  -0.00029   0.00002   0.00000  -0.00003  -0.08885   0.00000  -0.00002   0.00639   0.00000   0.00000   0.00004  -0.00057 ...
   0.00000   0.00000   0.00645   0.00000   0.00000  -0.97409   0.00000   0.00000  -0.00009  -0.00064   0.00000   0.00000 ...
  -0.00135  -0.00016   0.00000   0.00700  -0.00002   0.00000  -0.97285  -0.00004   0.00000   0.00000  -0.00074  -0.00009 ...
  -0.00018  -0.00002   0.00000  -0.00002   0.00639   0.00000  -0.00004  -0.97423   0.00000   0.00000   0.00005  -0.00064 ...
   0.00000   0.00000  -0.00008   0.00000   0.00000  -0.00009   0.00000   0.00000  -0.02311   0.00000   0.00000   0.00000 ...
   0.00000   0.00000  -0.00058   0.00000   0.00000  -0.00064   0.00000   0.00000   0.00000  -0.02297   0.00000   0.00000 ...
   0.00032  -0.00002   0.00000  -0.00066   0.00004   0.00000  -0.00074   0.00005   0.00000   0.00000  -0.02294   0.00000 ...
  -0.00002   0.00000   0.00000  -0.00008  -0.00057   0.00000  -0.00009  -0.00064   0.00000   0.00000   0.00000  -0.02298 ...
   ...  only 12  components have been written...
 Atom # 52
   0.62649  -1.79536   0.00000   0.00078  -0.00002   0.00000  -0.00546   0.00013
  -1.79536   5.20990   0.00000  -0.00205   0.00005   0.00000   0.01457  -0.00037
   0.00000   0.00000  -0.31368   0.00000   0.00000   0.99545   0.00000   0.00000
   0.00078  -0.00205   0.00000  -0.31439  -0.00001   0.00000   0.99909   0.00005
  -0.00002   0.00005   0.00000  -0.00001  -0.31376   0.00000   0.00005   0.99584
   0.00000   0.00000   0.99545   0.00000   0.00000  -0.44117   0.00000   0.00000
  -0.00546   0.01457   0.00000   0.99909   0.00005   0.00000  -0.45919  -0.00028
   0.00013  -0.00037   0.00000   0.00005   0.99584   0.00000  -0.00028  -0.44312
 
 
 Augmentation waves occupancies Rhoij:
 Atom #  1
   2.02793  -0.02439   0.00000   0.04437   0.00571   0.00000  -0.00510  -0.00071   0.00000   0.00000   0.02103  -0.00272 ...
  -0.02439   0.10647   0.00000  -0.09058  -0.01260   0.00000   0.00441   0.00076   0.00000   0.00000  -0.01480   0.00282 ...
   0.00000   0.00000   2.05449   0.00000   0.00000  -0.00722   0.00000   0.00000  -0.01549  -0.07903   0.00000   0.00000 ...
   0.04437  -0.09058   0.00000   1.99413   0.00077   0.00000  -0.00047  -0.00017   0.00000   0.00000  -0.07558  -0.00988 ...
   0.00571  -0.01260   0.00000   0.00077   2.05872   0.00000  -0.00017  -0.00762   0.00000   0.00000   0.00647  -0.07464 ...
   0.00000   0.00000  -0.00722   0.00000   0.00000   0.00104   0.00000   0.00000   0.00092   0.00380   0.00000   0.00000 ...
  -0.00510   0.00441   0.00000  -0.00047  -0.00017   0.00000   0.00066   0.00001   0.00000   0.00000   0.00320   0.00049 ...
  -0.00071   0.00076   0.00000  -0.00017  -0.00762   0.00000   0.00001   0.00103   0.00000   0.00000  -0.00037   0.00352 ...
   0.00000   0.00000  -0.01549   0.00000   0.00000   0.00092   0.00000   0.00000   0.09585  -0.00207   0.00000   0.00000 ...
   0.00000   0.00000  -0.07903   0.00000   0.00000   0.00380   0.00000   0.00000  -0.00207   0.07396   0.00000   0.00000 ...
   0.02103  -0.01480   0.00000  -0.07558   0.00647   0.00000   0.00320  -0.00037   0.00000   0.00000   0.05694   0.00463 ...
  -0.00272   0.00282   0.00000  -0.00988  -0.07464   0.00000   0.00049   0.00352   0.00000   0.00000   0.00463   0.07077 ...
   ...  only 12  components have been written...
 Atom # 52
   1.97970   0.02642   0.00000  -0.01261  -0.00115   0.00000  -0.00392   0.00000
   0.02642   0.00068   0.00000  -0.00921   0.00018   0.00000  -0.00033   0.00000
   0.00000   0.00000   1.70896   0.00000   0.00000   0.02720   0.00000   0.00000
  -0.01261  -0.00921   0.00000   1.72822  -0.01181   0.00000   0.03141  -0.00042
  -0.00115   0.00018   0.00000  -0.01181   1.70497   0.00000  -0.00042   0.02749
   0.00000   0.00000   0.02720   0.00000   0.00000   0.00050   0.00000   0.00000
  -0.00392  -0.00033   0.00000   0.03141  -0.00042   0.00000   0.00068  -0.00002
   0.00000   0.00000   0.00000  -0.00042   0.02749   0.00000  -0.00002   0.00051
 
 
 ihist @ write_md_hist          17
 mxhist @ write_md_hist        1000
 OPEN NETCDF FILE


Unlike the norm conserving case, its effect on the maximum force only appears after a few iterations, leading to a hard convergence on toldff.

Jordan
Posts: 282
Joined: Tue May 07, 2013 9:47 am

Re: Possible bug: ScfConvergenceWarning

Post by Jordan » Tue Mar 22, 2016 8:31 am

Hi
I already encountered the problem but I could not resolve it.
The best thing to do is to restart form the previous geometry.

In you PAW example, I wonder if it doesn't come just from the default value of nstep which is 20 ? Do you set nstep to something higher than 20 in your input file ?

Cheers

Jordan

sheng
Posts: 64
Joined: Fri Apr 11, 2014 3:44 pm

Re: Possible bug: ScfConvergenceWarning

Post by sheng » Tue Mar 22, 2016 9:12 am

Hi,

In the PAW calculation nstep is set at 1000, the same as norm conserving calculation. If this is not the case, an error saying "SCF cycle is not enough to converge ....." or something like that will appear.

For the norm conserving I can still detect the problem by grepping the keyword "ScfConvergenceWarning" using terminal.

However the PAW case, nothing is shown when the problem occurs, not even the usual "For the second time, the differences in forces have ......" statement. The program just proceed to the netcdf related part and then go on to the next iteration. In this case I will have to look through the log file file manually, which is exhausting when the iterations reached over hundreds (of course you can estimate the breaking point by noticing the spike in convergence).

It is nice to know that the problem had already been noted. I guess from now on I have to be more cautious when doing my calculations.

Thanks!

Jordan
Posts: 282
Joined: Tue May 07, 2013 9:47 am

Re: Possible bug: ScfConvergenceWarning

Post by Jordan » Wed Mar 23, 2016 8:43 am

Sorry we cannot help more.
the problem is that this issue is very specific to one abinit executale. If you give us your starting point and input file and we run it, we probably won't be able to reproduce the bug.

Neverthelesse, if you work out to produce a minimal input file (taht can be run on a few cpus, one is better) with given pseudo to produce the bug on you machine and give us all the way you compile abinit we can give it a try.

But no promise.

Jordan

Locked