STRANGE OCCUPATION BANDS FOR GW0 CALCULATIONS

Message

giacomo giorgi · #1 Post by **giacomo giorgi** » Thu Oct 29, 2009 6:39 am

Dear Vasp Users and Administrator

I am performing some calculations on Ge-doped GaAs alloys at GW0 level

My INCAR

ISTART=1
ISMEAR=0
SIGMA=0.05
NSW=0
ENCUT=287.700
ENAUG=531.356
NBANDS=150
LOPTICS=.TRUE.
ENCUTGW=90
ALGO=GW0
NOMEGA=48
NELM=4

kpoint sampling -> Gamma centered 6x6x6

I get at the fourth iteration the following band occupation on Gamma

k-point 1 : 0.0000 0.0000 0.0000
band No. old QP-enery QP-energies sigma(DFT) T+V_ion+V_H V^pw_x(r,r') Z occupation

37 0.8090 0.8099 -10.8544 11.6646 -13.0712 0.7608 2.0000
38 0.8090 0.8099 -10.8544 11.6646 -13.0712 0.7608 2.0000
39 3.8832 3.8864 -10.8135 14.7006 -9.8633 0.8068 0.5997
40 4.0285 4.0333 -10.9570 14.9916 -9.7787 0.7918 0.0000
41 3.6519 3.6552 -11.3874 15.0435 -12.4925 0.7866 2.0000
42 3.6519 3.6552 -11.3874 15.0435 -12.4925 0.7866 2.0000
43 4.6651 4.6725 -8.3847 13.0591 -5.2681 0.7979 0.0000
44 4.6651 4.6725 -8.3847 13.0591 -5.2681 0.7979 0.0000
45 5.1422 5.1500 -8.9985 14.1506 -5.9325 0.7949 0.0000
46 5.4410 5.4490 -8.7814 14.2325 -5.6324 0.7957 0.0000

My system has NELECT=82.

This occupation seems very strange. But maybe I am wrong . On gamma I have 80.5997 electrons and on L (0.5,0.5,0.5) I have 83.4 electrons. Indeed on L, I have:

41 2.9373 2.9400 -11.2163 14.1570 -12.6719 0.7805 2.0000
42 3.8461 3.8539 -9.8640 13.7197 -8.2733 0.8098 1.4003
43 5.0311 5.0387 -9.6372 14.6779 -6.8278 0.7941 0.0000

Why this strange occupation?

How to explain this behavior? I know from literature that in general LDA initial guess can not be ideal for GW0 calculations, especially for systems including Germanium (band disentanglement).
Please keep in mind that at LDA level the system has a gap ~ 0 eV, even if those systems are experimentally narrow gap semiconductors.

My question is :
Do you think it is useless to calculate at GW0 level a system that at LDA is almost metallic?

In case how to improve the results? Only reducing SIGMA -> 0? Or should I set ISYM=0? Starting from a Hartree Fock initial guess? Is it possible to start from an LDA+U initial guess for subsequent GW0 calculations?

Any comment is extremely appreciated!!!!
Please help me.

Very BEST,
G

irene · #2 Post by **irene** » Fri Oct 30, 2009 9:27 am

Have you tried to increase NELM to 10, for example?
I had exactly the same error. It was related to what I posted on the forum on Spetember the 1st:

http://cms.mpi.univie.ac.at/vasp-forum/ ... php?4.5883

Maybe you can try to increase NELM and check if it is the same problem I got. It is a nonsense, from my point of view, that NELM has that effect, but apparently nobody from the VASP team wants to answer GW questions...

irene · #3 Post by **irene** » Fri Oct 30, 2009 9:43 am

By the way, I don't think it is related to the fact that your system is almost metallic in LDA.

In my case, it was a semiconductor, with LDA band gap 0.9 eV and I was using NELM 4. At the 4th iteration, my occupations were absurd, I had almost one electron more for spin up!

I changed NELM to 10, and at the 4th iteration, the system was perfectly converged.

giacomo giorgi · #4 Post by **giacomo giorgi** » Thu Nov 05, 2009 6:30 am

Thanks Irene!
I will try for sure according to your suggestion!

By the way, same impression about what you say...

I will keep you posted!

Very Best,
G