molecule/metal(111)

Message

fabella · #1 Post by **fabella** » Mon Feb 13, 2006 3:17 am

Am am performing calculations on molecule/metal interface (interaction). I would like a suggestion or advice on processes of the calculation.

My question is in regard to INCAR file.

TRUE OR FALSE?

Is it negligible to omit spin polarization when performing geometry optimization on a molecule/ferromagnetic system (ie. ISPIN=1). Then when relaxed using ISPIN=2 and ICHARG=11 to calculate eigenvales (DOS).

Is the ISMEAR=1 a good account for relaxation in metals (the system at hand). (I am using the IBRION=2 method). (The ISMEAR=-5 is very good for total energy and bulk materials., but what about metals)

Thanks ahead of time

W

#2 Post by **admin** » Mon Feb 13, 2006 10:50 am

1) yes, you can calculate the DOS like that, but, one has to check (after relaxing the structure with ISPIN=1) whether turning on spin-polarization still leads to the same equilibrium geometry.

2) please read the online manual concerning the appropriate choice of ISMEAR for RELAXATION in metals (and metal surfaces): ISMEAR=1 should be used for geometry relaxation of all metallic systems.
For the fully converged geometry, I would do one more static run with ISMEAR=-5 to get the total energy as accurate as possible.

fabella · #3 Post by **fabella** » Tue Feb 14, 2006 4:14 am

Concerning this model. Using Selective dynamics in POSCAR.

I have fixed two layers (3 layers total) of the metal(111) surface and leaving the top layer + molecule to relax. The problem is after about 60 ionic steps the forces in the bottom layers are still significant ~0.4 eV/A (?) While the molecule and metal top layer is quite relaxed to about 0.05 eV ... any way "trick" to relax the bottom layers more efficiently? Does it matter, maybe I need more layers to effectively model this behavior?

Thanks

W

Veronika · #4 Post by **Veronika** » Tue Feb 14, 2006 8:42 am

3 layers for an asymmetric slab calculation is VERY FAR from describing a realistic system!

#5 Post by **admin** » Tue Feb 14, 2006 1:26 pm

I agree with you, Veronika, 3 layers are just a very crude model of a metal surface.
If some vacuum-adjacent layers are kept fixed at bulk layer distance, the forces of them will never vanish, because bulk-termination is not the equilibrium geometry of (almost) all surfaces. The magnitude of the remaining forces depends on the surface (material, orientation).

fabella · #6 Post by **fabella** » Tue Feb 14, 2006 2:28 pm

Thank You. This is part of a system we are still building upon. Probably five to seven layers would be more ideal ... inorder to relax the forces from interacting surfaces (maybe more?).

Thanks for your insight

W

#7 Post by **admin** » Wed Feb 15, 2006 3:10 pm

this really depends on te specific surface: usually
[1] for unsymmetric slabs (in the sense that one of the surfaces is kept at fixed bulk-spacing) one should take 2-3 'fixed layers' + N 'relaxed layers': the number of relaxed layers (N) should be large enough that surface effects are well-screened over that distance (i.e: no buckling, the fully relaxed interlayer spacing should be the bulk spacing...). How large N has to be varies strongly from 4-5 to >10
[2] for symmetric slabs (both surfaces are free to relax) the number of layers should be 2N+2 (the innermost two layers are kept fixed at bulk spacing)