Question about spin flips

[jkshenton]

Hi! Thanks for developing this code and including all the input and output files from the test systems -- that's super helpful!

I have a question about including the cost of a spin flip. In the Fe2O3 example, you look at a polaron hopping in the basal plane (in which all Fe belong to the same spin sublattice). How would one go about calculating the hopping barriers in the perpendicular direction (i.e. between different spin sublattices)? Would that be captured in this implementation?

Similarly, the 1D chain example seems to be ferromagnetic (?). If we wanted to extract the electron-transfer barrier for an antiferromagnetic 1D chain, would we be able to capture that correctly using this implementation?

Thanks again for making this public and I look forward to learning a lot by using it.

Cheers,
Kane

[pavankum]

Hi Kane,

Definitely, as long as you are able to create localized states you can do so. Creating localized states is independent of VAB calculation and it is more of an issue of parameter optimization with CP2K. Plane wave codes delocalize the excess electron unless you provide a good initial guess. With CP2K you can make use of the broken symmetry section to start with a good initial wavefunction to localize your electron on a specific atom, elongating bonds around the site by 0.15 Ang or so, and also a higher +U value if you are using DFT+U. Sometimes basis sets also affect the ease of localization. I uploaded the input files for basal plane hop in hematite (under ET_CP2K/5.Hematite) for your reference. You can see the mulliken charges showing a high spin configuration up and down, layer by layer. You can experiment with a different site of your choice and see how it goes.

Hope this helps.

Best,
Pavan.

[pavankum]

Hi Kane,

Looking back at your question I am not quite sure whether spin flip is accounted within my implementation. I found an earlier study (http://dx.doi.org/10.1063/1.1869492) where they discussed the nature of such a transfer in spin-bilayer Fe2O3. They used GMH for electronic coupling instead of a quasidiabatic approach (like here). It looks more involved than I previously thought. I will let you know if I find anything new.

Thank you.

Best,
Pavan.

[jkshenton]

Hi Pavan,

Thanks for looking into this and apologies for not getting back straight away. I wanted to understand a little more about the method and try out your code first.

I saw the 2005 Iordanova paper and was excited to try to reproduce their findings with your method :) They say:

The good agreement between the different models validates both methods, the UHF-based quasidiabatic method and the GMH method...

Does this mean that we should expect your quasidiabatic method to work, at least in principle?

Having the extension to account for spin-flips in the solid state would be immensely useful, not only to understand anisotropic conductivity, but also how defects affect transport!

Cheers,
Kane

[pavankum]

Hi Kane,

I apologize for the delay. Yeah, in principle it should work. I tried to do the c-dir transfer in hematite but found that cp2k doesn't allow to have more beta orbitals than alpha, which was problematic in generating the localized configuration in one of the states. We have spin-up and spin-down layers, and I was able to generate a localized configuration with an excess alpha placed in spin-down layer but not the other way around. I couldn't find a solution to it. One tricky way to go around this might be if we place a hole on the same atom far away from the excess electron in both the states. I haven't tested it but it seems to be a solution for this practical issue.

Once again sorry for the delay.

Best,
Pavan.