Optimization, Slabs, Surfaces

Optimization, Slabs, Surfaces

smpat 08 Jul 2013 10:58

From the tutorial "Surface Calculations", it appears that structural optimization with respect to volume, lattice parameters, etc, which will eventually be used in a supercell/slab, or surface can be found using, for example, a conventional unit cell. In other words, if my ultimate goal is to create a surface or supercell, is it acceptable to optimize my structure using the conventional unit cell (small number of atoms) and use these optimized parameters in my many-atom structure? Will the optimized parameters hold true for the larger structure (i.e., are the optimized lattice parameters for a conventional cell the same as those used in a larger cell)?

Re: Optimization, Slabs, Surfaces

giuliobiddau 08 Jul 2013 11:29

Hi smpat,

your questions is made of several.
Just to clarify, the tutorial you mention is just a tutorial. The size of your slab (etc) depends by your problem. You can get a simple idea out of simplified calculations.
Example 1: if your aim is to understand the Au(111) reconstruction, you can get some information on required vacuum from 1x1xn unit cell, but the appropriate description of the Au(111) surface will never come out. Similar arguments hold with adsorbates.
Example 2: consider the PDOS of a surface atom. It will change more or less substantially depending on the slab size.

Message to take home: test the smallest cell possible to have a "rough" idea of parameters like vacuum and number of layers. These will be a good start for much larger system calculations.

PS the lattice parameter comes for bulk calculations

Best regards
Giulio

Re: Optimization, Slabs, Surfaces

smpat 08 Jul 2013 11:45

Thank you for the quick reply, Giulio. I am learning and new to DFT calculations, so I suppose my question is somewhat general. To give you a more specific example, my first project is to design a 48 atom cell and perfom bandstructure/density of states calculations (not surface, just "bulk"). For my material, the conventional unit cell contains 12 atoms. My question specific to this case is, therefore, are the optimized parameters from my conventional cell calcuations valid for my (2x2x1) 48 atom cell? My question is motivated, as you might expect, by the fact that ground state calculations for a 48 atom cell will require much more computation time than for the 12 atom cell.

Re: Optimization, Slabs, Surfaces

giuliobiddau 08 Jul 2013 12:24

Hi again smpat

the unit cell is defined as follow "The smallest group of atoms that has the overall symmetry of a crystal, and from which the entire lattice can be built up by repetition". Thus your 12 atoms cell is sufficient to get the initial parameters, on which you can later build your 48 atoms cell.
With suggestion to optimise appropriately the parameters of your bulk (you can refer to the appropriate tutorial). You can later compare total energy per atom and other quantities obtained for your minimum energy configuration for the 12 atoms bulk, with the same for the 48 atoms cell.
PS be careful of the k point mesh when you move from 12 to 18 atoms cell.

Re: Optimization, Slabs, Surfaces

smpat 08 Jul 2013 13:26

This makes sense and was my logic as well. I should expect that if I double my cell in two directions, I would reduce the kpoints by half in these directions. In other words, go from 4 4 2 for the 1x1x1 cell to 2 2 2 for the 2x2x1 cell. Do you have more suggestions for taking care in this procedure?

Re: Optimization, Slabs, Surfaces

giuliobiddau 08 Jul 2013 14:40

Cross fingers and let me know…

Re: Optimization, Slabs, Surfaces

smpat 13 Jul 2013 14:41

I've been running the optimization process outlined in "General Lattice Optimization" (alternating between c/a and volume) for over a week (as I haven't been able to use this feature on the university HPC, it's running extraordinarily slowly on a MacBook Air). I've performed about seven alternating optimization runs for my specific structure, with no sign of convergence in sight.

My primary concern is that on every c/a run, the c parameter is increased and the a parameter is decreased, to the point where I'm about to have an entirely new structure. I'm using GGA (PBE '96), and accordingly started with parameters similar to those found in literature from other LAPW researches on the same structure with the same functional.

Does this sound familiar to anyone, or is this suggestive of an obvious operator error that I'm missing?

Last edited on 15 Jul 2013 11:04 by smpat Show more