Spin transport

[NW]

Hello,

I have a question related to the following tutorial:

https://docs.quantumatk.com/tutorials/fe_mgo_fe/fe_mgo_fe.html

In the calculation part, the Hubbard parameter (U) was not used for Fe. Is there any reason for that?

Thanks

[Petr Khomyakov]

What would be the reason to use it for Fe in this electron transport study?

[NW]

For electronic properties such as bandstructure and DOS, I am sure that we need Hubbard model. However, in case of electron transport I am not sure. Could you please give me a reference for more study about that?

Thanks

[Petr Khomyakov]

Fe|MgO|Fe tunnel junction is one of the most studied TMJ system in Spintronics. You may take a look at the original study by Butler et al cited in the tutorial, as well as the papers on this TMJ that cite the Butler's work. You may find them, e.g., through Google Scholar or Web of Science, or directly on the Publisher website.

[NW]

Dear Petr,

Thanks for the references. Does QuantumATK support Linear response theory to estimate U parameter? if not, how can I determine correct value for U parameter for a new structure that has not been studied before?

[Petr Khomyakov]

In QuantumATK, U parameter is a free parameter. U-parameter(s) is set for a given element and its orbital(s), so I think you may use U parameter value for a known material containing the corresponding element.

[NW]

I also have a general question about zero dimension magnetism (e.g. isolated Fe atoms). Is it possible to study this property by DFT implemented in QuantumATK?

[Petr Khomyakov]

I am not sure I understand what zero dimensional magnetism is, as magnetic order occurs due to exchange interaction among atoms. If you mean whether one can calculate a single Fe atom in QuantumATK, then the answer is yes, just place an Fe atom in a large-enough unit cell. One may use periodic boundary conditions for that, or Dirichlet ones to avoid spurious interaction with periodic images of the Fe atoms.

[NW]

Dear Petr,

Is it a good idea to compare the band structure of a material based on spin-polarized calculation with Hybrid functional (HSE06) in PW basis sets and then find the best U in LCAO-spin-polarized method to match the obtained bandstructure (to predict the U value)?

[Petr Khomyakov]

That can be done I suppose, and some people do that. Just a note that HSE06 formally is kind of correction to the exchange term, whereas U is a correction to the correlation term in the exchange-correlation functional. Of course, the latter can be seen as one term in DFT functional describing one Coulomb interaction in a many-electron system.