spin-flip excitons

[milesj]

Hello,

In my (antiferromagnetic) material there is excitonic photoluminescence exhibiting a clear zeeman splitting with a magnetic field applied along the ground state spin axis. This indicates an excitonic state strongly related to a spin-flip transition of some sort.

Given the clear spin-flipping character, I would not expect a spin-polarized computation with spin-orbit coupling to be able to simulate this exciton. However, since it is visible in the optical spectra (both photoluminescence and absorption), I would expect the addition of spin-orbit coupling to be able to simulate it, and I would expect a clear peak to appear in the optical spectra that did not exist in the spin-polarized computation.

Instead, when I do a full BSE computation with spin-orbit, there is essentially no change in the optical spectra. Analyzing the excitonic states through slepc, numerous additional states appear with the addition of spin-orbit, but they all have extremely low Strengths as output by ypp. Even with BDmRange set to 0.01, these additional states are only visible as very small bumps in the optical spectra of height ~0.01 (in the calculation without spin-orbit, the more optically active weakly bound excitonic states show up as peaks with height ~10-20 at this damping, and the band-band peak reaches 100). See the attached pictures.

This makes me think that perhaps the spin-flip exciton is not being simulated in these calculations. Is there something I'm missing? Could it actually be one of those tiny, seemingly optically inactive states? Or is there some parameter that I need to change or turn on so that YAMBO considers spin-flip excitons more seriously? Or is it somehow one of the spin-conserving transitions that seem unaffected by spin-orbit? I've also attached my input files for BSE.

Any advice is appreciated!

Thanks,
Miles

[Davide Sangalli]

Dear Miles,

Without info on the material, what I can tell you is that spin-orbit coupling (SOC) is in general a weak perturbation, and this is what you see, e.g. the spectrum does not change much.
The calculation without SOC will consider only spin conserving transitions, while the calculation with SOC will consider in general all transition, e.g. the BSE will have twice the number of poles.
All the extra poles would be dark (due to spin selection rules) in the case without SOC. They may acquire some intensity in the case with SOC, since SOC breaks the spin selection rules, but I expect it to be little.

I suggest to check these papers for info about spin-conserving and spin-flip transitions:
- https://journals.aps.org/prb/abstract/1 ... 103.155152
- https://arxiv.org/abs/2502.06598 (the code did not enter yet in the main yambo repo)

In presence of a magnetic field, there will be a lifting of degeneracy between the spin down and the spin-up channel, which are usually degenerate in AFM materials.
This might be the source of the Zeeman splitting you mention. I cannot go beyond that. This is forum which offers support about the code, not about physical problems.

Best,
D.