My Community

Hi everyone,

This is my first time using the DFT-D3 method with GGA functional for a magnetic heterostructure. The calculation works well if I don't consider the vdW correction (standard DFT) but crashes if DFT-D3 is included. I tried with IDW=11 and IVDW=12 but it got crashed. Then I found that I should also mention VDW_A1, VDW_A2, and VDW_S8 in the INCAR file. But there is no clear explanation of how to set them on vasp wiki. I found their values in one of the posts here and I used them but still, the calculation didn't converge. I tried with ALGO=Normal and ALGO=All but none of them worked. Attached is my INCAR file, I would appreciate it if anyone could help me find the right values of DFT-D3 parameters. Or is there anything missing in my INCAR file for DFT-D3 calculations?

ISTART = 0
ICHARG = 2
GGA = PE
IVDW = 12

VDW_A1 = 0.4289
VDW_A2 = 4.4407
VDW_S8 = 0.7875

LDAU = .TRUE.
LDAUTYPE = 2
LDAUL = 2 2 -1 -1 -1
LDAUU = 4.0 3.0 0.0 0.0 0.0
LDAUJ = 0.0 0.0 0.0 0.0 0.0
LMAXMIX = 4

NPAR = 4
KPAR = 4

PREC = Normal
ALGO = All
ENCUT = 400
EDIFF = 1.0E-7
LORBIT = 11
LREAL = F
GGA_COMPAT = .FALSE.
ADDGRID = .FALSE.

NSW = 500
EDIFFG = -1.0E-2
ISIF = 4
IBRION = 2
POTIM = 0.015
SYMPREC = 1.0E-6
ISYM = 0


AMIX = 0.1
BMIX = 0.00001
AMIX_MAG = 0.4
BMIX_MAG = 0.00001
MAXMIX = 40

ISMEAR = 0
SIGMA = 0.03
ISPIN = 2
MAGMOM = 3*4.5 2*-3.6 2*0.5 6*0.5 12*0.5 6*0.5

Thank you in advance.
-Roma

Hi Roma,

It is difficult to analyze the problem without further information. Please, provide all the relevant input and output files in accordance with the forum posting guidelines.

Hi Manuel,

Thank you for your reply. The attached zip file contains the input files as well as vasp.out file.
Thank you.

As far as I can tell, there is nothing wrong with your INCAR file. The DFT-D3 parameters are explained on this site on the VASP wiki. As you can see, default values are already provided for the standard PBE functional. They are identical to the values you specified.

The error in the standard output and the number of iterations indicates that the calculation might already be converged. Please check the forces of your last ionic step in the OUTCAR file to confirm that this is the case. If the calculation is well converged already, you may proceed by copying CONTCAR to POSCAR and restarting the calculation with IBRION=1. According to this forum post, you can try to set ADDGRID=true and specify a larger ENCUT in this final step.

Hi Manuel,

Thank you for the suggestion. However, I have already tried this and found that this method doesn't work. Every time I get the same energy pattern and the same error.
In fact, if you look at the energy pattern, it first tries to converge but after several ionic steps, it diverges and gives the random energy. The attached file contains the energies in each ionic as well as electronic steps. I also checked the XDATCAR file. It shows that after some ionic steps, the vdW distance between two monolayers starts decreasing and they come closer and closer, and eventually, they overlap with each other. From this point, the VASP starts throwing random energy values. Does this mean the heterostructure is not stable?
There is no issue with the standard DFT method. This is weird. Do you think EDIFF = 1E-6 and ENCUT = 350 are tight enough to have this issue? Does it make sense if I further lower both parameters?
I also have attached the XDATCAR file for reference.
Thank you.

My best guess at this point is that the calculation is not accurate enough.

Here are some parameters you can play around with to increase accuracy:

• Increase ENCUT (350 eV is rather low)
• Decrease EDIFF and/or EDIFFG
• Set PREC=accurate
• POTIM (try lowering to maybe 0.01)

I was able to run the calculations successfully by changing the number of cores or running the calculations on a separate cluster. The problem seems to be related to the parallelization. It is interesting that the same input setting works for DFT, but not for DFT-D3. I still have no idea how this happened.
Anyway, thank you so much for the help.

Hello there, I have the same problem, VASP crashes for large systems whenenver IVDW = 12 is applied.
Have you found any soulutions? Many thanks!

Hi,

I'm happy to look into the matter. If your issue is directly related to the original post, then we can remain in this topic. Otherwise, I would encourage you to open a new separate topic in the forum. In any case, I would ask you to provide more information by uploading the input and output files according to our forum posting guidelines.