Page 1 of 1
no electronic convergence, even with ICHARG=12
Posted: Mon Jul 28, 2008 3:55 pm
by rgc
Hi,
I am trying to do the electronic optimisation of a system with d electrons, but I can't manage to make it converge.
It doesn't converge even when using ICHARG=12 (no charge update), so I guess it should not be a problem with the mixing.
I have used both ALGO=Normal and Fast, and the result (no convergence) was the same.
Increasing the number of bands did not work either.
I am using DFT+U, and I have already included LMAXMIX=4 in the INCAR file.
Any suggestion on what else I can try??? Many thanks!
no electronic convergence, even with ICHARG=12
Posted: Mon Jul 28, 2008 5:38 pm
by rgc
Mmmm...
An update: Having checked everything without success, I decided to change to a different machine (with a different vasp version)... and it converged!
So I guess my question is now different: what in my vasp installation or version could be affecting so strongly the convergence???
This is the output from the "good" computer, where the job converges nicely (vasp 4.6.13):
---------------------------------------------
running on 64 nodes
distr: one band on 8 nodes, 8 groups
vasp.4.6.13 15Sep03 complex
POSCAR found : 3 types and 48 ions
scaLAPACK will be used
LDA part: xc-table for Ceperly-Alder, Vosko type interpolation para-ferro
POSCAR, INCAR and KPOINTS ok, starting setup
WARNING: wrap around errors must be expected
WARNING: random wavefunctions but no delay for mixing, default for NELMDL
entering main loop
N E dE d eps ncg rms rms(c)
DAV: 1 .325711242569E+04 .32571E+04 -.12490E+05 800 .134E+03
DAV: 2 .197889655543E+03 -.30592E+04 -.28160E+04 848 .396E+02
DAV: 3 -.391267032028E+03 -.58916E+03 -.56306E+03 864 .187E+02
DAV: 4 -.433686354531E+03 -.42419E+02 -.41041E+02 960 .555E+01
DAV: 5 -.435753676621E+03 -.20673E+01 -.20414E+01 1072 .117E+01 .625E+01
DAV: 6 -.355914234770E+03 .79839E+02 -.58500E+02 1192 .105E+02 .449E+01
DAV: 7 -.348264886152E+03 .76493E+01 -.12578E+02 1176 .737E+01 .222E+01
DAV: 8 -.351667373849E+03 -.34025E+01 -.28908E+01 1256 .386E+01 .226E+01
DAV: 9 -.352784555835E+03 -.11172E+01 -.74016E+00 1240 .237E+01 .102E+01
DAV: 10 -.357625843686E+03 -.48413E+01 -.24460E+00 1296 .126E+01 .693E+00
DAV: 11 -.359551353773E+03 -.19255E+01 -.20642E+00 1112 .768E+00 .286E+00
DAV: 12 -.360152231486E+03 -.60088E+00 -.66315E-01 1088 .471E+00 .162E+00
DAV: 13 -.360367781154E+03 -.21555E+00 -.36850E-01 1016 .283E+00 .133E+00
DAV: 14 -.360464299182E+03 -.96518E-01 -.84648E-02 1104 .193E+00 .944E-01
DAV: 15 -.360563618922E+03 -.99320E-01 -.83649E-02 1176 .186E+00 .574E-01
DAV: 16 -.360602627378E+03 -.39008E-01 -.80696E-02 1104 .159E+00 .262E-01
DAV: 17 -.360612452617E+03 -.98252E-02 -.27998E-02 976 .654E-01 .191E-01
DAV: 18 -.360621993707E+03 -.95411E-02 -.81063E-03 1240 .509E-01 .111E-01
DAV: 19 -.360622120570E+03 -.12686E-03 -.58273E-03 976 .275E-01 .614E-02
DAV: 20 -.360622557393E+03 -.43682E-03 -.16600E-03 760 .165E-01 .360E-02
DAV: 21 -.360622808640E+03 -.25125E-03 -.17580E-04 648 .757E-02 .223E-02
DAV: 22 -.360622917166E+03 -.10853E-03 -.11313E-04 560 .436E-02 .144E-02
DAV: 23 -.360622919090E+03 -.19237E-05 -.46117E-05 480 .267E-02
1 F= -.36062292E+03 E0= -.36062292E+03 d E =-.734552E-10 mag= 24.0000
---------------------------------
and this is the output in the "bad" computer, with vasp 4.6.28, where the job doesn't converge:
-------------------------------
vasp.4.6.28 25Jul05 complex
POSCAR found : 3 types and 48 ions
LDA part: xc-table for Ceperly-Alder, Vosko type interpolation para-ferro
POSCAR, INCAR and KPOINTS ok, starting setup
WARNING: wrap around errors must be expected
FFT: planning ... 1
reading WAVECAR
WARNING: random wavefunctions but no delay for mixing, default for NELMDL
entering main loop
N E dE d eps ncg rms rms(c)
DAV: 1 0.321842594964E+04 0.32184E+04 -0.12435E+05 800 0.134E+03
DAV: 2 0.240262890969E+03 -0.29782E+04 -0.28589E+04 840 0.422E+02
DAV: 3 -0.386891065221E+03 -0.62715E+03 -0.61862E+03 916 0.215E+02
DAV: 4 -0.433378797593E+03 -0.46488E+02 -0.57850E+02 1172 0.754E+01
DAV: 5 -0.434803601633E+03 -0.14248E+01 -0.46069E+01 1208 0.288E+01 0.623E+01
DAV: 6 -0.375811999119E+03 0.58992E+02 -0.56710E+02 1180 0.116E+02 0.533E+01
DAV: 7 -0.345075413199E+03 0.30737E+02 -0.13149E+02 1188 0.901E+01 0.167E+01
DAV: 8 -0.352935799345E+03 -0.78604E+01 -0.31080E+01 1192 0.335E+01 0.133E+01
DAV: 9 -0.356142341745E+03 -0.32065E+01 -0.76530E+00 1240 0.292E+01 0.928E+00
DAV: 10 -0.358044455274E+03 -0.19021E+01 -0.37914E+00 1308 0.188E+01 0.501E+00
DAV: 11 -0.359468879888E+03 -0.14244E+01 -0.23002E+00 1124 0.170E+01 0.270E+00
DAV: 12 -0.359241530350E+03 0.22735E+00 -0.12851E+00 1120 0.112E+01 0.198E+00
DAV: 13 -0.359608359925E+03 -0.36683E+00 -0.65110E-01 988 0.984E+00 0.196E+00
DAV: 14 -0.360014360557E+03 -0.40600E+00 -0.24470E-01 1036 0.920E+00 0.160E+00
DAV: 15 -0.360248751357E+03 -0.23439E+00 -0.15698E-01 1092 0.881E+00 0.130E+00
DAV: 16 -0.359336145750E+03 0.91261E+00 -0.23544E-01 1108 0.792E+00 0.121E+00
DAV: 17 -0.359237438868E+03 0.98707E-01 -0.82987E-02 1164 0.641E+00 0.123E+00
DAV: 18 -0.359443331250E+03 -0.20589E+00 -0.65271E-02 1168 0.694E+00 0.101E+00
DAV: 19 -0.360338681156E+03 -0.89535E+00 -0.72727E-02 1224 0.678E+00 0.103E+00
DAV: 20 -0.359611138574E+03 0.72754E+00 -0.69807E-02 1148 0.819E+00 0.974E-01
DAV: 21 -0.359860099390E+03 -0.24896E+00 -0.16010E-01 1180 0.979E+00 0.101E+00
DAV: 22 -0.359571415471E+03 0.28868E+00 -0.36180E-01 1216 0.919E+00 0.116E+00
DAV: 23 -0.359865632051E+03 -0.29422E+00 -0.26144E-02 1244 0.923E+00 0.126E+00
DAV: 24 -0.359599661006E+03 0.26597E+00 -0.28946E-01 1300 0.851E+00 0.102E+00
DAV: 25 -0.358977567354E+03 0.62209E+00 -0.68942E-02 1300 0.833E+00 0.101E+00
DAV: 26 -0.359993355751E+03 -0.10158E+01 -0.10116E+00 1212 0.836E+00 0.929E-01
DAV: 27 -0.359405898450E+03 0.58746E+00 -0.15578E-02 1228 0.801E+00 0.103E+00
DAV: 28 -0.359055644736E+03 0.35025E+00 -0.99597E-03 1232 0.799E+00 0.820E-01
DAV: 29 -0.359669934741E+03 -0.61429E+00 -0.15730E-02 1244 0.723E+00 0.124E+00
DAV: 30 -0.359417276396E+03 0.25266E+00 -0.43916E-02 1256 0.695E+00 0.110E+00
DAV: 31 -0.359105723178E+03 0.31155E+00 -0.38973E-02 1188 0.718E+00 0.104E+00
DAV: 32 -0.359666563311E+03 -0.56084E+00 -0.38142E-01 1232 0.767E+00 0.139E+00
.
.
.
------------
and so on until it reaches 60 cycles with no convergence... Any hint on how I could solve this problem in the 'bad' computer? Many thanks!
no electronic convergence, even with ICHARG=12
Posted: Wed Aug 06, 2008 11:56 am
by admin
hmm, to me it seems as if there is some problem with the libraries (LAPACK, of BLAS) or optimization. please in any case
-- try to run with vasp compiled with lowest optimization, (on a single cpu)
-- only compare the outputs of the same vasp-version
-- check which contributions to the total energy cause the difference.