Problem calculating dielectric function of ZnO doped with Al

[samar_fawzy]

Hello Everyone,

I am trying to replicate a paper: Transparent Conductive Oxides as Near-IR Plasmonic Materials: The Case of Al-Doped ZnO Derivatives by Calzolari et al.

I got the same real and imaginary parts of the dielectric function as the paper incase of undoped ZnO.
However, after doping with Al, I am observing no change in the dielectric function.
The resulting structure after doping should be metallic, i.e. the real permittivity is negative between (0-2) eV, but there is no overlap between Density of states in the pure metallic sense.

I am using the following steps in both doped and undoped ZnO, the procedure works well for undoped ZnO
1- Relaxing the structure to get CONTCAR
2- Using CONTCAR as POSCAR and Doing SCF calculations to get WAVECAR and CHGCAR, and NBANDS from EIGENVAL file
3- Use NBANDS, WAVECAR and CHGCAR to calculate the frequency dependent dielectric function using:

EDIFF=1E-8
NBANDS= 192
LEPSILON=.TRUE.
#IBRION = 8 # perturbation theory effects, doesn't work with spin polarized calculation
LOPTICS=.TRUE.
ALGO=EXACT
#LRPA=.TRUE.
CSHIFT=0.1

When I got wrong results for ZnO-Al, I tried the below:

1- Changed smearing and sigma values, but the issue persisted
ISMEAR=0 ; SIGMA=0.04
ISMEAR=0 ; SIGMA=0.5
ISMEAR=1 ; SIGMA=0.2
ISMEAR=2 ; SIGMA=0.2
2- Tried with RPA and no RPA, the issue persists
3- Tried to do SCF calculation to get maximum number of planewaves (5940) and then start with ISTART=0,ICHARG=2, and NBANDS=5940, but got the error: Error EDDDAV: Call to ZHEGV failed. Returncode = 1 2 48


The complete INCAR file is as below:

ISTART=1
ICHARG=11

ENCUT=400
ISMEAR=2 ; SIGMA=0.2
EDIFF=1E-8

# no Ionic relaxation
LORBIT=10
NEDOS=8000
PREC=Accurate

# Hubbard
ISPIN=2
LDAU=.TRUE.
LDAUTYPE=4
LDAUL = 2 -1 1
LDAUU = 10 0 7 # 10 for Zn , 0 for Al, 7 for O
LDAUJ = 0 0 0
LMAXMIX=4


NBANDS= 192
LEPSILON=.TRUE.
#IBRION = 8
LOPTICS=.TRUE.
ALGO=EXACT
#LRPA=.TRUE.
CSHIFT=0.1

[andreas.singraber]

Hello!

Welcome to the VASP forum! Please provide all relevant input and output files for the steps you have described according to the forum posting guidelines.

Thank you!

Best,
Andreas Singraber

[samar_fawzy]

I attached the files as requested

[samar_fawzy]

Sorry this is the correct folder

[andreas.singraber]

Hello!

I had a look at the paper and your calculations but unfortunately I cannot yet offer a simple solution to the reproducibility problem. I suggest to take a step back and try to reproduce first the band structure shown in Figure 1 of the paper. My guess is that this shows the ZnO band structure for the primitive cell. If the results match I would further suggest to reproduce the colored lines, i.e., Fermi levels, for AZO (maybe one of them is enough). Only if all of that matches we should look further into the dielectric function.

I noticed that you did not use exactly the same setup described in the publications, is there a specific reason for that? You used a smaller supercell with 32 instead of 64/128 atoms and also a different k-point grid. Also, the LDA+U parameters were 12.0 eV (Zn) and 6.5 eV (O) whereas you set them to 10.0 eV (Zn) and 7 eV (O).

Best,

Andreas Singraber

[samar_fawzy]

Thank you for your reply
I used a smaller supercell size for faster convergence, which also gives metallic dielectric function as described.
Regarding the Hubbard parameters, I can reproduce correct ZnO density of states using them, so I just used them.
I also used them to reproduce the results using CASTEP.

I will try to reproduce the Band structure figure.