Sunday, February 03, 2008

One PRL/arxiv paper

I'll write more in the next day or two about what I think is a very exciting new result of ours. For now, I wanted to write a little about this paper:
arxiv:0801.4021, Frolov et al., Electrical generation of pure spin currents in a two-dimensional electron gas
For quite some time there has been a strong interest in using the spin degree of freedom of electrons for information processing. In some sense this is old news (see this past year's Nobel in physics), but the real trick is to see whether one can generate currents of only spin, rather than pushing whole, spin-polarized electrons through a circuit. In principle pure spin currents can be moved without dissipation, so if they can be generated and detected in a "nice" way, it may be possible to reduce the power required for certain computations. Of course, unlike charge, spin polarization is not conserved - spins generally prefer to relax back to an unpolarized state in the absence of big magnetic fields. This paper reports a way of generating spin currents that is quite clever - use quantum point contacts + spin-orbit scattering to generate an excess spin population in a region of 2d electron gas, and then the excess spin population diffuses away (without a net flow of charge). This paper also demonstrates that reducing the dimensionality of the system leads to an enhanced spin lifetime. It's a neat result and a very pretty experiment.

2 comments:

Alison Chaiken said...

Wouldn't diffuse spins necessarily drag charges along just as charges drag entropy along in thermoelectric phenomena? It seems to me that there should be an entire set of analogous coupling coefficients. After all, the spins belong to electrons, and the electrons have a band structure.

Doug Natelson said...

Hi Alison - You're right that there are analogous coupling coefficients such as the spin diffusion length, etc. Still, the spin-Hall effect is a clear example of a situation where it is possible to build up a gradient of spin without a net gradient of charge. I'm still not sure that I am happy with the way I think about these phenomena. The main differences, to me, between spin transport and thermoelectric phenomena are (1) the fact that excess spin population can be both created and destroyed, unlike the one-way street of the second law; (2) spin-orbit couplings can lead to really nontrivial effects; and (3) in principle, with localized charges but exchange interactions you can still get spin flow.