Structure relaxation

Message

TAT · #1 Post by **TAT** » Thu Jun 04, 2009 1:49 am

Hi,
I want to relax only the ions in the unit cell such that only to change one of the lattice vector of the unit cell (say z). For the rest two directions (x&y) I want to keep the length of the lattice vectors fixed to keep sufficient vacuum space in the unit cell in the (x&Y direction). In this case what would it be sensible thing to use ISIF = 4 for relaxation using CG algorithm.
The problem with ISIF = 3 is that the full relaxation ends up in changing everthing resulting the vacuum space (~15A with Oxygen atoms in the system as the largest atom) to be reduced as a result of which the system becomes interacting due to periodic boundary conditions. Is this problem because may be the vacuum space is less and I should increase it?
Thanks

alex · #2 Post by **alex** » Thu Jun 04, 2009 7:26 am

Hi,

I think you'll have a hard time to get a slab cell converged. The main (physical) problem is that Coulomb interaction converges slowly with distance (1/r). I don't know, if your system may show such forces.
I'd assume that optimizing the structure with ISIF = 3 will result in a reasonable slab a and b. Just set c back and reoptimize without letting change the cellsize.
However, be aware of optimizing slab cellsizes should show dependence on the slab thickness (more bulk atoms ...). So the standard procedure is that you just take the optimized bulk cell, build your slab and just optimize the positions of the atoms in the slab and _not_ the cell.

Cheers

Alex

san · #3 Post by **san** » Fri Mar 18, 2011 3:10 pm

hi Alex,
could you explain what is meant by "Just set c back and reoptimize without letting change the cellsize"? It would be helpful for me as I want to relax the cell in two directions by keeping fixed in one direction.

Thanks!

alex · #4 Post by **alex** » Fri Mar 18, 2011 4:24 pm

Hi san,

in case you are optimising a slab with ISIF=3 you'll experience slight changes to the lattice parameter perpendicular to the surface. This is probably due to long range Coulomb interactions, which decay slowly with r.
If you really want optimised cell parameters for a slab structure, you might wish to have a rounded 'c'. So you set it back to some value (and adjust the fractional coordinates).

Cheers,

Alex

san · #5 Post by **san** » Wed Mar 23, 2011 11:20 am

hi,
Thank you!

Another question:
How to relax the cell in one direction alone, by keeping it fixed in the other two directions?
What is the ISIF value to be used for this?