My Community

Good morning,

I am trying to perform a G0W0 calculation on a 13 atoms cell. To do that I followed the steps given here http://cms.mpi.univie.ac.at/wiki/index. ... le_shot_GW
I have two questions about this matter.
First of all, can anyone confirm that ‘ALGO = Exact’ is missing in the one step calculation for the WAVECAR and WAVEDER files ?! It is written in the second step of the two step calculations but not in the one step calculation.

I assumed that it was true and I applied this recipe to my cell. However my job stops and I got the following lines in my OUTCAR

################### OUTCAR ######################

----------------------------------------- Iteration 1( 1) ---------------------------------------


POTLOK: cpu time 0.02: real time 0.03
SETDIJ: cpu time 0.01: real time 0.01

total amount of memory used by VASP on root node 639291. kBytes
========================================================================

base : 30000. kBytes
nonl-proj : 4955. kBytes
fftplans : 2706. kBytes
grid : 4677. kBytes
one-center: 31. kBytes
wavefun : 596922. kBytes

################### OUTCAR ######################

I tried a few combinations of keywords and I found out that when I change ‘ALGO = exact’ to ‘ALGO = normal’, then it works fine.
Here is the INCAR and KPOINTS files I am using for my calculations

################### INCAR ######################
PREC = Accurate # Usually Normal for geom. optimisation and Accurate for Etot and DOS calc.
ENCUT = 600 # Cutoff energy of the plane wave basis set (eV)
LREAL = .FALSE. # Auto for geom. opt. and .FALSE. for Etot and DOS calc.

ALGO = Exact

IBRION = 2 # Algorithm for the geom. optim.
NSW = 0 # Max. number of iterations / 0 if no geom. optim.
ISIF = 0 # 2 = only atomic positions / 6 = only cell param. / 3 = Full geom. optim.
EDIFF = 1.0E-4
EDIFFG =-1E-02
LOPTICS = .TRUE.

ISMEAR = -4
NEDOS = 1601
NBANDS = 240
EMIN = -5
EMAX = 10
LORBIT = 11

################### INCAR ######################



################### POTCAR ######################
Automatic mesh
0
M
3 3 3
################### POTCAR ######################

Thank you in advance for any help you can provide.

Emmanuel

The word "step" is apparently making some confusion. Lets say it this way:
The GW0 calculation is made in two stages: 1/basic DFT, 2/GW0 calculation.
Basic DFT is performed in two steps:
1/simple DFT calculation (no LOPTICS)
and after step 1 is converged
2/ one electronic step (NELM=1) with LOPTICS=.TRUE., and ALGO=Exact is performed
This way a large amount of the computer time is saved, because ALGO=Exact is very time consuming.
Next is the stage 2, the GW0 calculation with LOPTICS=.TRUE., ALGO=GW0, and NOMEGA = XX

Thanks for the reply.
I tried the method you explained over the week-end however I still have a problem.
The first step of the basic DFT calculation is OK. However for the second step I have the following error

--------------------------------------------------------------------------------------------------------


Maximum index for augmentation-charges 1993 (set IRDMAX)


--------------------------------------------------------------------------------------------------------


First call to EWALD: gamma= 0.238
Maximum number of real-space cells 3x 3x 2
Maximum number of reciprocal cells 2x 2x 4

FEWALD executed in parallel
FEWALD: cpu time 0.00: real time 0.00


----------------------------------------- Iteration 1( 1) ---------------------------------------


POTLOK: cpu time 0.18: real time 0.18
SETDIJ: cpu time 0.02: real time 0.02

total amount of memory used by VASP on root node 1651548. kBytes
========================================================================

base : 30000. kBytes
nonl-proj : 13157. kBytes
fftplans : 7176. kBytes
grid : 12472. kBytes
one-center: 93. kBytes
wavefun : 1588650. kBytes

UPDATE: I figured out that my ENCUT was too high and hence the computation time was too big.
The problem now is that after the GW0 calculation, the energy of the electrons is off limit. For example, for Si, the Fermi level is at 146 eV !!

UPDATE 2: My bad, I am dumb. I forgot to copy the WAVECAR in the folder of my G0W0 calculation.
Sorry for the trouble.