new nudged elastic band questions

Queries about input and output files, running specific calculations, etc.


Moderators: Global Moderator, Moderator

Post Reply
Message
Author
mhyman
Newbie
Newbie
Posts: 19
Joined: Wed Jul 20, 2005 10:37 pm

new nudged elastic band questions

#1 Post by mhyman » Sun Oct 02, 2005 8:06 pm

Hello,

I thank those that have been so helpful in the past in answering my questions. I've become a bit more savvy about setting up neb calculations. I'm using a fairly complex system, so I don't expect easy convergence, but I think I've finally identified the correct end points. I first tried using IBRION = 3, but it moved my intermediate images around and the MEP didn't make sense. I didn't try to wait for it to converge, since it didn't look like it was going to.

I reran the calculations with IBRION = 1. This seemed to be going well. After 100 cycles of POTIM = 0.2, one image seemed to have essentially converged and the other two were oscillating only a little bit. I then set POTIM = 0.1, changed EDIFF to 1e-5, but kept EDIFFG = 1e-3. I figured that the forces would now be more accurate and the time step is smaller, so it should converge quickly. This didn't happen at all. In fact, after 40 cycles, two of the images started osciallating wildly over a range of about 0.1 eV.

I suppose that this could be a result of the sytem that I'm looking at, but is there something else that I can try?

Thanks,
Matt
Last edited by mhyman on Sun Oct 02, 2005 8:06 pm, edited 1 time in total.

graeme
Jr. Member
Jr. Member
Posts: 54
Joined: Sat Jul 02, 2005 6:11 am
Location: Austin, TX
Contact:

new nudged elastic band questions

#2 Post by graeme » Mon Oct 03, 2005 2:34 pm

If your initial band is pretty crazy, I recommend using IBRION=3, POTIM=0.01 until the forces get as low as 1 eV/Ang. Then look to see if the pathway makes physical sense. If it does, continue converging using a larger POTIM of 0.1, or switch to IBRION=1.
Last edited by graeme on Mon Oct 03, 2005 2:34 pm, edited 1 time in total.

Post Reply