Too Small Band Gap of Silicon (~0.65 eV) Calculated by VASP

[wenpengzhu]

Hello VASP users,

I am a beginner in VASP. I tried to redo the Si example. BUT the band gap is ~0.65 eV, which is much smaller than the real value (1.17 eV). I am wondering what I am doing wrong.

The below are the steps I have taken and corresponding inscripts. The pseudo-potential I used is from the folder /paw_pbe/Si. I also used the ultrasoft potential from the folder \pot\Si. But the

band gap is the same. The results are similar to the one that presents on the website (https://wiki.alcf.anl.gov/wparker/index.php/Silicon)

Could you please check my flow chart and see if I did something wrong? How can I get accurate energy calculated by VASP?

Besides, I have another two questions for VASP usage.
1. I can get different Fermi energies from setp #2, #3.1 and #3.2. WHICH one is correct? The one from the SCF calculation? Should I use this value for the next steps (DOS and Band

structure)?
2. What files should I copy to the folder of the next steps for continuation? Just CHGCAR and CHG, or CHGCAR, CHG and WAVECAR, or all the output files? If the ISTART = 0, does it

mean I don't need to use any files from the previous step?

===========================
Step #1: Structure Optimization.
I initially set the lattice constants as 5.5 Angstrom (POSCAR). Then I did several ion steps to get the optimized structure (CONTCAR).

INCAR file:
System = Diamond Si
ISTART = 0
ISMEAR = 1
SIGMA = 0.05
ENCUT = 300.0
NELM = 200
GGA = PE
EDIFF = 1E-06
EDIFFG = -0.02
NPAR = 8
NSW = 50
IBRION = 2
ISIF = 3
ISYM = 1

KPOINTS file:
K-Points
0
Monkhorst Pack
21 21 21
0 0 0

Output: CONTCAR, CHG, CHGCAR (for the next step #2)

=====================================================
Step #2: SCF Calculation (Run self-consistently to get the charge density)
I changed the CONTCAR file to POSCAR file for input structure. The k-point file is the same as above.

INCAR file:
System = Diamond Si
ISTART = 0
ISMEAR = 0
SIGMA = 0.05
ENCUT = 300.0
NELM = 200
GGA = PE
EDIFF = 1E-06
EDIFFG = -0.02
NPAR = 8
ISYM = 1

Output: CHG, CHGCAR, WAVECAR (for the next steps #3.1-DOS and #3.2-Band Structure) (Is it right? What files should I copy to the next step for continuation?)
Fermi energy: 5.83091414 eV (read by the file DOSCAR. It wil be used for the next steps.) (We also can get different Fermi energies from step #3.1 and #3.2. Which one is the correct Fermi

energy?)

=================================================
Step #3.1: Non-SCF Calculation (DOS)
I used the accurate charge density files from the previous step #2 for DOS.The k-point file is the same as above.

INCAR file:
System = Diamond Si
ISTART = 1
ICHARG = 11
ISMEAR = -5
SIGMA = 0.05
ENCUT = 300.0
NELM = 200
GGA = PE
EDIFF = 1E-06
EDIFFG = -0.02
NPAR = 8
ISYM = 1

Output: DOSCAR file --> DOS curve

=================================================
Step #3.2: Non-SCF Calculation (Band Structure)
I used the accurate charge density files from the previous step #2. The k-point file is different with a line mode.

INCAR file:
System = Diamond Si
ISTART = 1
ICHARG = 11
ISMEAR = 0
SIGMA = 0.05
ENCUT = 300.0
NELM = 200
GGA = PE
EDIFF = 1E-06
EDIFFG = -0.02
NPAR = 8
ISYM = 1

KPOINTS file:
K-Points for FCC Bandstructure G-X-W-K-G-L-U-W-L-K|U-X
10
line
reciprocal
0.000 0.000 0.000 ! G
0.000 0.500 0.500 ! X

0.000 0.500 0.500 ! X
0.250 0.750 0.500 ! W

0.250 0.750 0.500 ! W
0.375 0.750 0.375 ! K

0.375 0.750 0.375 ! K
0.000 0.000 0.000 ! G

0.000 0.000 0.000 ! G
0.500 0.500 0.500 ! L

0.500 0.500 0.500 ! L
0.250 0.625 0.625 ! U

0.250 0.625 0.625 ! U
0.250 0.750 0.500 ! W

0.250 0.750 0.500 ! W
0.500 0.500 0.500 ! L

0.500 0.500 0.500 ! L
0.375 0.750 0.375 ! K

0.250 0.625 0.625 ! U
0.000 0.500 0.500 ! X

Output: EIGENVAL file --> Band Structure

================================

Thank you for your attention. If you have any suggestions or corrections, I will very appreciate that.

Best,
Wenpeng Zhu

[kzhuo3]

0.
Don't worry, it is normal for DFT/VASP to underestimate band gaps. 0.65 eV gap in Si seems reasonable for a VASP calculation. If you try Germanium, you will even get no band gap. You can get good experimental agreement only if you try some advanced hybrid functional, HSE or +U methods.

1.
I believe the Fermi energy only makes sense only after you have defined your energy zero. I like to use the vacuum potential as the energy zero. This might help ( http://cms.mpi.univie.ac.at/vasp-forum/ ... php?4.1609 )

2.
For non-scf DOS and band structure calculations, I keep my CONTCAR and CHGCAR only from step 2. CONTCAR is of course renamed to POSCAR.