convergence test

Message

apple · #1 Post by **apple** » Wed Apr 27, 2011 3:15 pm

hi,
can someone comment on how to establish the right number of layers for a surface calculation given that the slab is asymmetric and there is fixed vacuum. Also, I am aware of the fact that the bottom layer needs to be passivated with H, but do we first perform convergence tests for the number of layers? I would appreciate your help.

#2 Post by **admin** » Wed May 11, 2011 6:18 pm

There is no general rule of thumb to that question. I would proceed as follows: keep 1-2 bottom layers + the passivation layer of your slab fixed (frozen) in geometry. then check the forces on the atoms of the lowermost layer which you relax. The slab thickness should be such that presence of the surface (top layer) does not lead to too large forces on the lowemost relaxed layer (i.e., the conditions for that layer are more or less 'bulk-like').

apple · #3 Post by **apple** » Fri May 13, 2011 7:27 pm

Hello,
thank you for replying to my post.
However, I am still not clear about the procedure: If I allow the atoms of the lowermost layer to relax (other than the very bottom 1-2 layers), then the forces on these atoms will be small as set with the EDIFFG flag. So, what am I checking for? I would appreciate you commenting again.

tlchan · #4 Post by **tlchan** » Sun May 22, 2011 9:29 am

One of the intentions behind passivating the bottom layer is to remove the surface states related to the bottom surface such that the bottom part of the slab is bulk-like. The admin suggested you to check the forces of the movable atoms for the bottom part of the slab during atomic relaxation (probably not after the relaxation is finished. As you have noted, the forces after relaxation are small). If the forces of those atoms are small throughout the whole relaxation run, then the atoms will be at bulk-like positions, which implies that the top surface does not significantly interact with the bottom part of the slab. You can then infer that the slab is thick enough.

In general, you can calculate the physical quantity that you are studying (for example, surface energy, surface diffusion) as a function of slab thickness to check the convergence.

apple · #5 Post by **apple** » Mon May 23, 2011 1:41 am

Thank you. It is clear now.