Ion Diffusion through 2D material convergence issues

[brobinson10296]

Hi all,

I am using this tool to study single ion diffusion through graphene. I am able to get the forces below 1 eV/A pretty easily, but after that, each image's forces will oscillate. I have tried using IOPT=3 and 7, along with varying the TIMESTEP tag, yet the system's forces are still now converging towards 15 meV/A. Do you have any suggests on what I could do to fix this? Below is the INCAR that I am using.

PREC=Normal
LREAL=Auto
ALGO=Fast
ISMEAR=0 ; SIGMA=0.05
ENCUT=520
NSW=500
NPAR=5
ICHARG=2
LMAXMIX=4

EDIFF=1.E-6
EDIFFG=-15E-4

IBRION=3
ISIF=2
POTIM=0
NSW=500

IOPT=3

LWAVE=.TRUE.
LCHARG=.FALSE.

ICHAIN=0
LCLIMB=.TRUE.
IMAGES=9
SPRING=-5
NELMIN=5


Thanks!

[graeme]

If you post the calculation, I can take a look at it.

[brobinson10296]

Unfortunately, I am unable to share the exact calculation from the system that I was talking about in the original post, however, I can share a similar calculation. Everything is behaving the same, except some images are still converging to below 1 eV/A. Hopefully this will help with figuring out why I am getting the force oscillation issues. Thank you for your help!

[graeme]

Ok, well this graphene Ag NEB looks just fine. The values of the barrier and the distances that the atoms move (the distance along the path) are both very high. I assumed this indicated a problem, but now that I see the pathway, it makes sense.

In systems like this with both soft and zero modes, convergence can be tedious. You can try lowering EDIFF to make sure that your forces are significantly accurate. You will have very high forces along the band as you pull the graphene apart. But also note that you have a barrier of over 14 eV - do you really need to know that precisely? I mean, this is just not going to happen as a diffusive process. Even at 1000K this process will take well over the age of the universe to occur diffusively.