Spin Orbit Coupling

[Juan]

Dear All,

         I'm studying a semiconductor (InSb) and need to do calculations on spin orbit coupling. In the tutorial "Noncollinear calculations for metallic nanowires" you give some parameter on calculations for Nickel and pre-calculations for the convergence of values. Are those parameters more or less good approximation to start a calculation on SOC for a semiconductor? I've been looking for information but don't find just a few papers. If you know a good reference or references for this, can you please let me know? Thank you.

Sincerely
Juan

[Anders Blom]

No, for soc in a semiconductor you don't need any special parameters, just select SpinOrbit in the options and run. The tutorial you refer to deals with noncollinar magnetization, so it's a bit different.

Some better tutorials for you are these:
http://docs.quantumwise.com/tutorials/spin_orbit_bandstructures.html
http://docs.quantumwise.com/tutorials/gold_spin-orbit.html

[Juan]

I thank you very much for the quick reply and information! 

Juan

[Juan]

Hello!

        I did calculations for the SOC for InSb(semiconductor). I got great bandstructure results with MGGA but once I changed the pseudopotential to the OMX, I get a closed bandstructure. I did all recommendations on the cut-off parameter and the K points. Also tried some other parameters to see if it changed. Is there another pseudopotential that can be used for the SOC calculation besides that one? Also, any other recommendation would be very helpful.

Thank you!

Juan

[Anders Blom]

What is a closed band structure?

You must use OMX, since the FHI pseudopotentials don't have spin-orbit projectors included. In 2016 there will be SG15 pps which also can be used

[Juan]

Doing calculations with FHI pseudopotential and MGGA method, I get the correct result. Once I change the OMX potential for the SOC calculation, the conduction and valence band get together. Zooming a lot, there is a small gap but it measures 0.0002 eV.

[Anders Blom]

You seem to be comparing apples and pears... One calculation is with spin-orbit (which splits the valence band, primarily), the other with MGGA (which fixes the band gap issue).

You need OMX + MGGA + spin-orbit, if you want spin-orbit splitting with a correct gap.

[Anders Blom]

You can try setting the c-parameter in MGGA by hand, and vary it until the band gap become large enough. You could also try to give it an initial spin different from zero (default), in case you are falling into a local minimum.

[Juan]

Thank you very much for the recommendations, Dr. Blom!