Help me for how to choose the k-point grid

[hong_tang1]

Dear Sir or Madam:

I am running a vasp calculation to get the band structure for a system. I met a problem need your help.

1. my system is a kind of modified single layer ribbon system. Because the nature of the system, I have to make a about 50x36x17 (angstroms) supercell. Only the a (or x) direction is parodical. Both y and z are not periodical. i.e. the ribbon direction is along x. the system is then relaxed before band structure calculations.

2. I use 2 steps to calculate band. stept1, kpoint: Gamma 5x1x1, ENCUT 650eV, to get CHGCAR. I used ISPIN=2.
step2, line_mode calculation, with CHGCAR from step1. and ICHARG = 11 and ISPIN=2
kpoints:
System
25 ! 20 grids
Line-mode
reciprocal
0.0000 0.0000 0.0000 ! \Gamma
0.5000 0.0000 0.0000 ! X


however, when I check the calculated results, the band structure along Gamma-X direction is almost flat, no dispersion. It looks not right.

Please help me what is the problem. is my kpoint grid in step 1 correct? How to choose a correct k grid for this system?

I appreciate your helps very much.

Best,

Hong

[henrique_miranda]

In principle, I don't see anything wrong with your setup. I cannot check the INCAR or POSCAR because you did not attach it.
What do you mean when you say that the band-structure is almost flat? Can you share an image?
Is this a large system? If so did you try this calculation on a smaller system before?

[hong_tang1]

Dear Henrique :

Thank you so much for helping.
1. "The band structure is flat" means that for whatever k value along Gamma-X direction, the energy is almost the same for each band. So the band structure looks like many flat parallel lines, not curved like usual band structure.

2. My system in the supercell is actually a 1D system, periodical and infinite along x (or a) direction. Both y and z (or (b and c) directions are inserted by 13 angstroms of vacuum layer. The supercell size is 50x36x17. However, the periodical unit along x direction is long (about 50 Angstroms). The system has about 320 atoms. I used PBE to calculate.

3. My question is that usually for k-grid choice, like for a system, supercell is 2x5x10 (angstroms or A0), then reciprocal lattice vectors are about 0.5 x 0.2 x 0.1 (unit:1/A0), so k grid is usually chosen to be 25x10x5, in order to make the distance between two k points is about ~ 0.02 (1/A0). Which also means in the above numbers, 2 times 25 is about 50, 5 times 10 is about 50, 10 times 5 is about 50 too.

4. However, for my system, the super cell is 50x 36x17, and only the x direction is periodical. I chose k grid as 5x1x1, which makes 50x5 not equal 36x1 not equal 17x1. Please let me know whether my choice is correct or not. Is there any problem here for vasp to calculate the energy for the system, if I choose k grid this way?

5. If I want to abide the rule in 3 to choose k grid. then I have to increase the size of my supercell in y and z directions to about 250 A0, to about 50x250x250. This will be a very huge supercell. It probably can not compute.

Please please help me.

Best,
Hong

[hong_tang1]

Please INCAR for step1
SYSTEM = XXX
ENCUT =650.00 eV
ISTART = 0; ICHARG = 2
ISMEAR = 0
SIGMA = 0.01

ISPIN = 2


NCORE = 10


GGA = PE
LASPH = .TRUE.
ADDGRID = .TRUE.
AMIN =0.01
!LMAXMIX = 4

PREC= Accurate
!NSW=300
IBRION=-1
!ISIF=3
!EDIFFG=-0.01
NELM=400

!NGX = 34
!NGY = 34
!NGZ = 34

LREAL = Auto !.False.
IALGO = 38
EDIFF = 3E-07


LWAVE = .True.
LCHARG = .True.
LORBIT = 11






INCAR for step 2
SYSTEM = XXX
ENCUT =650.00 eV

ICHARG = 11
ISMEAR = 0
SIGMA = 0.01

ISPIN = 2


NCORE = 10

GGA = PE
LASPH = .TRUE.
ADDGRID = .TRUE.
AMIN =0.01
!LMAXMIX = 4

PREC= Accurate
!NSW=300
IBRION=-1
!ISIF=3
!EDIFFG=-0.01
NELM=400

!NGX = 34
!NGY = 34
!NGZ = 34

LREAL = Auto !.False.
IALGO = 38
EDIFF = 3E-07


LWAVE = .False.
LCHARG = .False.
LORBIT = 11

[henrique_miranda]

I was hoping you could share a plot of the band-structure.
I think your choice for the k-point grid is correct, the problem might be somewhere else.
Can you share your POSCAR and KPOINTS files?
Did you try to run this calculation on a system with fewer atoms?

[henrique_miranda]

(User sent POSCAR and KPOINTS file in a private communication)

I've looked at your POSCAR and KPOINTS file and everything looks correct to me.
As I mentioned before, I really suggest that you try to do the same calculation on a small system that you can run in a couple of minutes.
For example, you can try the graphene nanoribbon example that I am attaching and check for yourself if you get what you would expect. You can also check what happens when you choose a path in a different direction.