Hi all,
I am currently running simulations of hydrogenated graphene. I have made several (hundreds) previous calculations without failure. However, today VASP defeated me.
I am calculated a NEB for hydrogen adsorption in a strained graphene sheet using the Henkelman transition-state-code addition. I use 8 images, of which 2 do not converge using standard electronic relaxation setup: I only specify to use "IALGO = F". After 60 steps the difference was still oscillating between e-1 and e-2, and all others were at e-6. The two images that did not converge were then put under the microscope. The structures looked fine, so I started tweaking the mixing settings as recommended in one of the online workshops (https://www.vasp.at/vasp-workshop/slide ... ectron.pdf).
1) First I changed to ICHARG = 12, and noticed a quick relaxation.
2) Then I changed to linear mixing (amix=0.1 and bmix=0.001), but not convergence.
3) Then I started changing bmix and amin, but no convergence.
I am completely lost as to how I can fix the convergence. As stated, I have run extensive convergence checks regarding kpoints, etc.
My used settings are:
ENCUT = 500
PREC = Normal
EDIFF = 1e-6
ALGO = Fast
ISPIN = 2
ISMEAR = 0
SIGMA = 0.1
NELMIN = 4
#
ICHARG = 2
AMIX = 0.1
BMIX = 3.0
AMIN = 0.01
Can anyone help?
Best regards,
Jesper
Convergence error in graphene+hydrogen system
Moderators: Global Moderator, Moderator
-
- Newbie
- Posts: 2
- Joined: Thu May 30, 2013 1:45 pm
-
- Newbie
- Posts: 2
- Joined: Thu May 30, 2013 1:45 pm
Re: Convergence error in graphene+hydrogen system
Just wanted to give an update on my own problem.
After a long discussion with a colleague, I ended up changing the initial magnetic moments of the system.
A quick "grep" of the magnetization of the system for the failing images showed unrealistic values.
I also changed the mixing for the spin density (amix_mag and bmix_mag).
Now all images converge.
After a long discussion with a colleague, I ended up changing the initial magnetic moments of the system.
A quick "grep" of the magnetization of the system for the failing images showed unrealistic values.
I also changed the mixing for the spin density (amix_mag and bmix_mag).
Now all images converge.