Surface geometry optimisation

[aschenk]

Hi,
I'm currently trying to look at the electronic properties of surface modified (100) diamond, and have started by initially trying to get the correct geometry of the bare surface and some simple, well known systems. I'm having trouble with the optimisation process; I start with the database diamond structure, cleave it along 100, replicate along x,y and z, and then set up the optimisation and send it to the job manager. The calculations all run, but the surface doesn't reconstruct-the atoms in the topmost layers do shift slightly, but not dramatically enough for it to really be a reconstruction. Is there something I am missing? Are there particular, vital settings that should be used for a surface geometry optimisation that I may have missed?

[Nordland]

What you are doing, sound very right to me.

In order to model surface reconstruction, it is important that the cell is big enough in x-y plane to allow of the reconstruction to form,
that said for some surfaces is nearly impossible to simulate a reconstruction because the required periodicity of the surface
could contain millions of atoms.

Personally, I would, slightly depending on the system, do what you would + I would repeat the system perhaps a little more in the z-axis, and then constraint the layers you want to represent the underlying bulk of the surface. I would then introduce a disorder in the surface in order to remove the configuration from any symmetry minimum that might end up breaking things.

[aschenk]

Thanks-too many late nights lately, I hadn't thought of nudging the surface atoms in case there is an energy minimum, most of your other suggestions I had already considered. I can't be sure because I'm tired, but I have a vague feeling that this may be what the issue is-something in the back of my mind says that there is an alternate reconstruction which is very similar to the bulk, which will appear under the right experimental conditions.

I'll give this a try next, and put up the (hopefully) success story when the calculation completes. Happy to hear any other advice