Sunday, August 13, 2006

This week in cond-mat (mini-version + digression)

One particular paper caught my eye this week:
cond-mat/0608243
- Nakamura et al., Low-temperature metallic state induced by electrostatic carrier doping in SrTiO3.
The authors of this paper have managed to solve, at least well enough to do the experiment, the surface processing and ohmic contact challenges to make a field-effect transistor on the surface of a n undoped strontium titanate single crystal. At high enough gate voltages, they can accumulate enough carriers in the channel to drop the sheet resistance of the 2d charge layer well below the resistance quantum (~ h/2e^2 ~ 13 kOhms), and see metallic temperature dependence of the channel conductance (that is, the conductance improves with decreasing temperature). Anytime someone does this sort of thing with a new material system it's interesting, and SrTiO3 is particularly noteworthy because it's a perovskite (crystal structure not that different from high Tc materials), it's an incipient ferroelectric (very large dielectric constant as T decreases), and when doped at moderate levels, it's been known to superconduct. Field-effect "doping" is a very nice tool for studying this sort of physics, because the carrier density can be changed without introducing the disorder that comes with chemical doping. I'm actually a co-author on a forthcoming Reviews of Modern Physics paper about this general topic.

Now that you've glanced at that preprint, take a look at this PRL. Those folks have been looking at conduction in a semiconducting polymer, poly(3-hexylthiophene), and claim to observe a metal-insulator transition. The data are very pretty, but I just don't see how the interpretation matches the data well. These folks argue that, because the temperature dependence of the (highly nonlinear) conduction that they measure (at large source-drain voltage) gets weaker with increasing gated charge, and approaches temperature-independence, they are seeing a metal-insulator transition. It seems that the picture is: for high quality polymer films, the potential minima from disorder are relatively shallow, and when the potential is sufficiently tilted (by source-drain), and the deeper minima are filled (by large gated charge), then one can get tunneling (rather than thermal activation) out of the minima, and temperature-indep. conduction. This may well be right, but I really object to calling this a metal-insulator transition. There is no true transition here, and never does conduction improve with decreasing T, as in a metal. Again, the data are good, but the title and language are, to me, an example of wordsmithing. (Full disclosure: one reason this rubs me the wrong way is that in our own work we saw similar weakening of T-dep. several years ago. I would never have thought of calling this a transition to a metallic phase.)

4 comments:

  1. Hey Doug. I have similar reservations about such papers that claim to see a MIT in funny contexts. The terms 'metal' and 'insulator' mean specific things and can only be defined rigorously if you stick to narrow definitions. Otherwise you end up with gobbledygook. I'm not sure which standard this paper meets in that regard.

    As of course you know, there are other prototypes for dynamic phase transitions (current induced CDW depinning, current induced vortex lattice melting) in CM systems. Could this be seen in such a context?

    I think the burden of proof then is on the authors here to show that the idea of dynamic MIT even makes sense. We can only define metals and insulators on asymptotically low energy scales and thereby using large bias invalidates these assumptions. If they want to claim such a phenomenon I think they should have started off with a simpler material. I have little doubt that if you took Si:P on the insulating side of its MIT and used large currents you would see enhanced dI/dV at large V. Would this be a 'dynamic' MIT? I don't know, and certainly more work would have to be done before one could so. I do grant that the power law behavior in the conductivity over a few decades is interesting though.

    One other comment: It strikes me that their Efros Shklovskii T's are unphysically large. Are such #'s typical for polymers like this? They are certainly way way off from #'s in disordered semiconductors in regimes where you could see hopping.

    ReplyDelete
  2. Hey Peter - I agree with what you said, and I should take a closer look at their hopping parameters. These materials are known to be hopping conductors, though the situation is complicated because the density of (localized) states is strongly energy dependent. The best analogy I can make here is to consider a hot filament at some dc bias relative to a nearby ground plane. As the temperature is reduced, the conductance drops because thermionic emission is suppresed. However, at high enough bias, there is Fowler-Nordheim emission that is temperature-independent. I would never consider this the onset of "metallic" conduction. Depinning would probably be the best way to think about the case at hand.

    ReplyDelete
  3. Anonymous9:10 AM

    Hi Doug, thank you for featuring our paper. It has been finally published in APL89, 133504 (2006).
    As one of the co-authors of the RMP, I would like to mention a similar work has been also done by another co-author which is seen in a recent issue of Science. The work is independent of our present work, but is rather more similar to my previous work which is included in the RMP. The guys are clever enough to fairly extend my incomplete work...
    However, we are proud that we have also done it in more straightforward fashion, as JHS did five years ago.
    Isao.

    ReplyDelete
  4. Anonymous9:12 AM

    Hi Doug, thank you for featuring our paper. It has been finally published in APL89, 133504 (2006).
    As one of the co-authors of the RMP, I would like to mention a similar work has been also done by another co-author which is seen in a recent issue of Science. The work is independent of our present work, but is rather more similar to my previous work which is included in the RMP. The guys are clever enough to fairly extend my incomplete work...
    However, we are proud that we have also done it in more straightforward fashion, as JHS did five years ago.
    Isao.

    ReplyDelete