Page 1 of 1

Electronic convergence problems

Posted: Thu Jun 02, 2005 4:46 pm
by buber
I am trying to do a Vasp calculation, and I'm having problems with the system not converging the electronic wavefunction. I've tried both IALGO=38 and 48, with no success. Any suggestions would be greatly appreciated.

I'm doing 1x1x1 k-points. My INCAR is:

IBRION=2
NSW=100
PREC=med
LCHARG=F
LWAVE=F
EDIFFG=-0.01

An example output listing follows:

N E dE d eps ncg rms rms(c)
DAV: 1 0.159673026903E+05 0.15967E+05 -0.60831E+04 850 0.294E+03
DAV: 2 0.923387659527E+04 -0.67334E+04 -0.57703E+04 803 0.151E+03
DAV: 3 0.556498604359E+04 -0.36689E+04 -0.31910E+04 954 0.130E+03
DAV: 4 0.331061470126E+04 -0.22544E+04 -0.19414E+04 970 0.126E+03
DAV: 5 0.200959354767E+04 -0.13010E+04 -0.11520E+04 987 0.129E+03 0.154E+02
DAV: 6 0.241403013800E+04 0.40444E+03 -0.74231E+03 923 0.158E+03 0.614E+01
DAV: 7 0.535896872666E+03 -0.18781E+04 -0.18201E+04 963 0.184E+03 0.535E+01
DAV: 8 0.319986981357E+03 -0.21591E+03 -0.35492E+03 562 0.199E+03 0.466E+01
DAV: 9 -0.101825030215E+04 -0.13382E+04 -0.13550E+04 994 0.748E+02 0.124E+02
DAV: 10 -0.140271462518E+04 -0.38446E+03 -0.44150E+03 996 0.385E+02 0.152E+02
DAV: 11 -0.152241149621E+04 -0.11970E+03 -0.10365E+03 996 0.150E+02 0.168E+02
DAV: 12 -0.156922411900E+04 -0.46813E+02 -0.20275E+02 996 0.125E+02 0.175E+02
DAV: 13 -0.157302318700E+04 -0.37991E+01 -0.37082E+01 996 0.719E+01 0.177E+02
DAV: 14 -0.157916667474E+04 -0.61435E+01 -0.62497E+00 996 0.405E+01 0.177E+02
DAV: 15 -0.157057904564E+04 0.85876E+01 -0.83353E-01 994 0.153E+01 0.176E+02
DAV: 16 -0.170615586018E+04 -0.13558E+03 -0.15463E+01 964 0.196E+01 0.188E+02
DAV: 17 -0.159218815069E+04 0.11397E+03 -0.78862E+00 980 0.216E+01 0.181E+02
DAV: 18 -0.128293360530E+04 0.30925E+03 -0.11372E+02 916 0.466E+01 0.150E+02
DAV: 19 -0.773424785276E+03 0.50951E+03 -0.69972E+02 916 0.103E+02 0.757E+01
DAV: 20 -0.667849678649E+03 0.10558E+03 -0.33701E+02 980 0.775E+01 0.420E+01
DAV: 21 -0.666636751902E+03 0.12129E+01 -0.16997E+01 996 0.657E+01 0.417E+01
DAV: 22 -0.670482338037E+03 -0.38456E+01 -0.12301E+00 996 0.181E+01 0.428E+01
DAV: 23 -0.634265169401E+03 0.36217E+02 -0.41843E+01 980 0.288E+01 0.338E+01
DAV: 24 -0.625274752574E+03 0.89904E+01 -0.34184E+01 996 0.353E+01 0.344E+01
DAV: 25 -0.624998084522E+03 0.27667E+00 -0.15824E+01 996 0.278E+01 0.384E+01

Electronic convergence problems

Posted: Fri Jun 03, 2005 7:03 am
by Veronika
It does converge (look at dE and rms) but rather slow, give it more time!

greetings

Electronic convergence problems

Posted: Fri Jun 03, 2005 2:11 pm
by buber
I'll try that, but it seems that the dE isn't consistently negative, which I would have expected. So, the total energy keeps going up and down, not going monotonically to some value. I guess I shouldn't worry about that, but rather if the magnitude of dE is getting smaller?

PS You aren't the Veronika that was in Denmark when Hannes was, are you?

Electronic convergence problems

Posted: Mon Jun 06, 2005 6:47 am
by Veronika
If dE gets smaller during the relaxation, this is a good hint for convergence, but it does not mean that dE is necessarily always negative!
Obviously also the total energy should also converge to a certain value. ;-)
btw is it possible to increase the k-mesh slightly and not only using the Gamma Point?

@PS.: Sorry no, I've never been to Denmark ...
Greetings
<span class='smallblacktext'>[ Edited Mon Jun 06 2005, 11:59AM ]</span>

Electronic convergence problems

Posted: Fri Jun 10, 2005 3:05 pm
by buber
Thanks for the help, Veronika. I was just doing gamma point to do a quick minimization and then was going to do more k-points to get a better number.

In this system, which is Lu2SiO5, I added a Ce atom and now I get horrible convergence with Gamma point. The energy keeps going positive (the total energy). Would 2x2x2 help with this? Is there anything else I can try? I've never had such problems with convergence before.

Electronic convergence problems

Posted: Fri Jun 10, 2005 4:03 pm
by Veronika
I think, that you should try 2x2x2.
You can also try to do spin-polarized calculation, since Ce has one f electron it is magnetic (but I do not know the electronic configuration of the Ce for your case, it depends ...) - if your system is not too large (but from the total energies I guess it is ...)
Did you check the size of your (super)cell and the external pressure in the OUTCAR file? This should not be too large, too ...

hope, I could give some hints
greetings

Electronic convergence problems

Posted: Mon Jun 13, 2005 9:01 am
by admin
1) please have a look whether you need to set ISPIN =2 ( electronic
configuration of the valence states, listed in V_RHFIN in the directory
of the respective POTCAR)

to improve convergence:
1) if you use PAW potentials: set LMAXMIX = 6 in INCAR to ensure that the
high-l components are mixed properly
2) reduce the mixing parameters until a certain pre-convergence is reached:
I suggest to do a few steps with linear mixing first, keep WAVECAR and
then continue with the default mixing parameters:
a) set
AMIX = 0.2
BMIX = 0.0001
(AMIX_MAG = 0.8)
(BMIX_MAG = 0.0001)
LWAVE =.True.
for the first steps and,
b) comment the mixing - tags and set
ISTART = 1
in addition to continue from this run. Do not forget to copy WAVECAR to the
working directory in the continuation run.

Electronic convergence problems

Posted: Mon Jun 13, 2005 8:05 pm
by buber
Thanks. To know if I should set ISPIN=2 based on V_RHFIN, is that just a matter of looking if there is only one electron in the last shell? Or is there another criteria for this?

Electronic convergence problems

Posted: Tue Jun 14, 2005 11:37 am
by admin
Yes, you are right, you should have a look at the occupancies of the levels.
Of course it is also always helpful to have some information from experiment,
regarding the magnetic structure of a solid (ferromagnetic, antiferromagnetic,
non-collinear arrangement of the spins,...)