Orbital transformations for system with shallow defect states

[Nicholas Winner]

Hello all,

I am running some defect calculations for vacancies using CP2K, and some of my calculations induce states very close to the band edge. 

I believe this is causing the OT method to be very slow in its convergence, since OT requires (to my knowledge) a clear delineation between occupied/unoccupied states. I would like to continue using OT because of its good scaling behavior, so I'm wondering if there are any tricks to help converge systems with shallow states, or if I simply need to let the SCF proceed for a very long time.

Any suggestions would be appreciated, thanks.

-Nick

[Marcella Iannuzzi]

Dear Nick,

It depends on the system. 

If you expect localised defect states, it might be that DFT is not able to find the right state.

You can try to bias at the initial guess the localisation and hope that the SCF falls in the right minimum. 

Using OT you can help the SCF convergence by tuning the preconditioner.

If the system is really metallic, you better give up OT.

Regards

Marcella

[Matt W]

Couple of tricks I regulalry used are:

(i) include distortion around the defect likely to cause localisation

(ii) include a core correction (excess charge) on atoms to break symmetry / get localisation. Then restart from the wavefunction without the core correction (it is in the KIND section).