Dear all,
For the NEB calculation, it is difficult to the convergence of small EDIFFG. What's the biggest value of EDIFFG I can use for publishing a paper ?
Thanks in advance.
Best regard,
Jiang
The biggest value of EDIFFG for NEB I can set
Moderator: moderators
-
wangjiang12
- Posts: 2
- Joined: Sat Jan 10, 2015 2:51 am
Re: The biggest value of EDIFFG for NEB I can set
If you are close to convergence, I suggest using the IOPT=1 (lbfgs NEB optimizer) for the most rapid convergence, and make sure to use the climbing image. If you primarily care about the barrier, just worry about the force on the climbing image for your convergence criterion. You can also using the neb2dim script to run a dimer calculation, which can converge more quickly because you only have 1 image, so you can typically use more processors for that image.
For publication purposes, I suggest starting with a higher force criterion, say ediffg=0.03 and then run a test at 0.01 to see how much the energy changes. This will give you a sense for how sensitive your reported energies are to the convergence parameter. Then, the real question is If you are comfortable with the uncertainty in the barriers, for example, due to the residual force.
For publication purposes, I suggest starting with a higher force criterion, say ediffg=0.03 and then run a test at 0.01 to see how much the energy changes. This will give you a sense for how sensitive your reported energies are to the convergence parameter. Then, the real question is If you are comfortable with the uncertainty in the barriers, for example, due to the residual force.