Sunday, October 29, 2006

The h parameter....

Well, since several folks are commenting on the h parameter, I might as well put in my two cents. The h number is defined here. In brief, if you've published h papers (and no more) that each have h or more citations, then your h number is, well, h. In principle, your h number is not supposed to count self-citations (though once h is above 10 or so, that becomes pretty irrelevant anyway). In some fields (e.g. CS) where people tend to publish on public electronic archives rather than in journals, citations of those preprints are counted. The idea is that the H number is a metric of scientific performance and impact, and is more robust than mere citation counting. Steady output that people actually cite is rewarded more than being one co-author on a Science paper that happened to get 750 citations. There are variations, too. You can calculate the h number divided by a person's years of "professional experience", or actually figure out dh/dt. For a fair comparison between people, one should normalize h numbers by subfield. In condensed matter physics, a typical person near tenure time has an h of around 10. In mid-career, an h of around 20-30 is about the average, and exceptional people like National Academy members tend to have h values higher than 50. The h number can be skewed in certain cases. Some people publish little, but their work can have enormous impact. Others, such as materials growers, can have enormous h numbers because they supply materials used by dozens of experimental groups.

Obviously trying to quantify a person's scientific impact and productivity in one number is a crude and rough thing to do, just as the subject GREs and qualifying exams are often crude indicators of actual aptitude. Just as I think the physics GRE is only really good at identifying outliers (the best 2.5% do very well on it; the worst 2.5% do very poorly; the middle 95% get scores that don't seem to correlate with their actual talent or ability), the h number is similar. I would never dream of assigning too much importance to it in tenure cases. As in grad school or postdoc or faculty applications, detailed letters of recommendation are far more useful, and in my experience correlate much better with actual performance. However, if someone has an h number far outside the expected norm in either direction, I'd like to know that. For example, I heard recently of an externally appointed dean at a research university where the faculty were rather shocked to find that the dean's h number was about 4. Unsurprisingly, people who have had vastly larger scientific impacts don't really like being told what to do or have their decisions scrutinized by someone who has essentially been a professional administrator.

Anyway, I wouldn't lose too much sleep over h numbers. They just get a lot of attention because they're a relatively new idea, and they do seem superior to the previous crude metric, citation counting.

Saturday, October 28, 2006

This week in cond-mat

Four papers this time, though brief descriptions. Eventually the semester will ease up a bit and I'll have more time to write.

cond-mat/0610572 - Gabelli et al., Violation of Kirchoff's Laws for a coherent RC circuit
Kirchoff's laws are the basic rules you learn in introductory circuits, and may be suitably generalized to think about high frequency systems. One of the basics is that impedances in series add. In this paper (also published in Science), the authors do some very nice work using gated two-dimensional electron gas to make an effective RC circuit, where part of the R is a quantum point contact. They find that when the whole system is quantum coherent, the basic idea of adding impedances goes out the window. This is neat, and it is a beautifully done experiment, but I don't find the conceptual point to be very surprising at all. Think about this simple case just in the dc limit: a single tunneling barrier has some effective tunneling resistance. A second, identical tunneling barrier has the same resistance. What is the resistance of the series combination of the tunneling barriers? Well, in the incoherent limit, the resistances just add. In the fully coherent limit, you have to worry about interference effects between the barriers, and can even arrive at perfect transmission for the series combination, even though each barrier individually is not very transmissive. This paper's analysis is more general than this, but I can't help but think that it's really the same basic physics at work.

cond-mat/0610634 - Neder et al., Controlled dephasing of electrons by non-Gaussian shot noise
This is another great experiment by the folks at the Weizmann Institute, studying the basic physics of quantum decoherence using an interferometer and a tunable detector at one arm of the interferometer, all made from GaAs 2d electron gas. In earlier work, they've shown that the interference of the electrons in the which-path interferometer can be suppressed in a controlled and continuous way, depending on how "on" the detector is, and how strongly the detector is coupled to the interferometer arm. Here, they work in the quantum Hall limit, and study directly the relationship between the back-action of the detector (via its noise) and the effect on the interference.

cond-mat/0610710 - Scalapino, Numerical studies of the 2d Hubbard model
The 2d Hubbard model is one of the favorite toy models suggested for high Tc. It's a square lattice, with some nearest neighbor hopping amplitude t and an on-site repulsion U so strong that each site can only hold 1 electron. Scalapino has written a review chapter summarizing numerical treatments of this model, and arguing that it has all the essential features of high-Tc. Numerical work in models like this is notoriously difficult computationally, in part because of the requirements that the whole many-body state be antisymmetric under exchange of any two electrons.

cond-mat/0610721- Potok et al., Observation of the two-channel Kondo effect
I want to write more about this later. In brief, David Goldhaber-Gordon and Yuval Oreg had proposed an experimental set-up to implement a tunable version of the long-sought two-channel Kondo model, in which a single localized spin is coupled via tunneling to two independent electronic baths. The 2CK model is of interest because its ground state is not a Fermi liquid (as opposed to the conventional Kondo model and ordinary metals). David's students Ron Potok and Illeana Rau have done the experiment, and the results look very interesting. Using the scaling of the conductance, it looks very much like they have succeeded in getting (at least) very close to the two-channel Kondo state. A cool experiment, and very technically demanding, in part because the temperature scales needed to see the physics are so low.

Wednesday, October 18, 2006

This week in cond-mat

Three papers this time out. The semester is very busy, so not much commentary for a while.

cond-mat/0610352 - Wu et al., Optical metamaterials at near and mid-IR range fabricated by nanoimprint lithography
There's been a lot of hubbub about making meso- and nanostructured materials that have negative permeability and permittivity over some limited frequency range. These materials can have very weird optical properties (obey a left-hand rule; refract in the opposite direction than conventional materials; can be used to try and beat the diffraction limit for imaging; can be hyped into Harry Potter-style invisibility cloaks). Here is the first example I've seen of someone making large-area 2d structures with these properties in an interesting frequency range (near-IR, close to the 1.5 micron telecommunications band).

610413 - Evers and Burke, Pride, prejudice, and penury of ab initio transport calculations for single molecules
I really like this paper, both for what it says and how it says it. The authors go into detail about different calculational approaches used to predict or retrodict electronic transport properties of single molecules. Very often people in this field crank out results using quantum chemistry techniques (density functional theory) and approximate methods without ever pointing out what those methods generally can't handle (strong correlation effects like Kondo; significant interaction corrections; Coulomb blockade). This paper really gets at what works, what doesn't work, why, and what can be done. Similar in topic is a recent preprint from Datta's group, where they look at Coulomb blockade in small molecules.

quant-ph/0610117 - Dyakonov, Is fault-tolerant quantum computation really possible?
I haven't read this one yet, but the abstract is attention-getting. It argues that the math upon which error correction schemes for quantum computers are based is unrealistic in terms of its relationship with real world systems. Therefore, it may be impossible in principle to scale up to large quantum computing systems. Anyone take a look at this and have an opinion?
Update: After reading Dave Bacon's comment, I actually looked at this preprint. Wow. The tone is very colloquial (it's based on a talk), and is hardly subtle, nor is it very convincing as reasoned technical argument. Is this the same Dyakonov as in the Dyakonov-Perel mechanism of spin relaxation? The initials are the same. Not that having something named after you necessarily means that you're right about everything; Brian Josephson's rather unorthodox views on telekinetics and levitation are the classic case in point.

Friday, October 06, 2006

CM Experimental position at Rice

Presumably any serious job-seekers out there would read about this on the AIP website or in Physics Today, but what the heck - it can't hurt to reproduce the ad here:

Faculty Position in Experimental Condensed Matter Physics
Rice University

The Department of Physics and Astronomy at Rice University invites applications for a tenure-track Assistant Professor position in experimental condensed matter physics, in the general area of quantum materials, including strongly correlated electronic systems and quantum nanostructures. This position will complement our existing strengths in condensed matter and materials physics and quantum degenerate gases. Applicants should send a dossier that includes a curriculum vitae, a statement of research and teaching interests, a list of publications, and two or three selected reprints, and arrange for at least three letters of recommendation to be sent to the Chair of the Condensed Matter Search Committee, Dept. of Physics and Astronomy, MS 61, Rice University, 6100 Main Street, Houston, TX 77005. Review of applications will begin in December, and the appointment is expected to be available July 2007. Rice University is an affirmative action/equal opportunity employer; women and underrepresented minorities are strongly encouraged to apply.


I hope we get some good candidates! I agree firmly with what a competitor of mine from Cornell once said to me about faculty searches: "I aspire someday to be the dumbest person in my department."

Thursday, October 05, 2006

Two fun science links

I haven't had a chance to watch these yet, but the Vega Trust in the UK has, on line, four full length lectures on quantum electrodynamics by Richard Feynman from 1979. Someday I'll have the four or five hours available to watch these.

Much shorter, and much more viscerally fun, check out this video to see that alkali metal chemistry really can be fun. (Thanks for the link, Pat!)

This week in cond-mat

Two papers this time around....
cond-mat/0610107 - Butenko et al., Electric field effect analysis of thin PbTe films on high-\epsilon SrTiO3 substrate
This paper is a nice example of using the three-terminal field-effect geometry as a way to probe the states of a material while keeping the disorder fixed. The authors use strontium titanate as the dielectric layer. Since SrTiO3 is almost a ferroelectric, it has an extremely high gateable polarization (gated charge density) at breakdown field. This means that the authors are able to shift the Fermi level over a very broad range, spanning the entire (relatively narrow compared to things like Si or GaAs) energy gap of the PbTe disordered film, and gate in either electrons or holes. They can see the effects of interface states, and the broadening of the conduction and valence bands due to disorder. Their main observation is that the mobility gap in the disordered case is actually larger than the standard band gap in PbTe. Pretty interesting, and written in a reasonably pedagogical style.

- Liu et al., Experimental observation of the inverse spin Hall effect at room temperature
The spin Hall effect is a neat concept that my friend Jairo Sinova at Texas A&M has been involved with heavily, as has Soucheng Zhang, who taught me many-body physics back in grad school. The basic idea is that, under the right conditions, it is possible for a dc longitudinal current to establish an unequal spin population on the transverse edges of a material (e.g. a GaAs heterostructure). That is, along the two edges of the sample that parallel the current flow, there will be an excess spin population (with no excess electronic population!), with one edge having an excess of spin-up, and the other edge having an excess of spin-down. Here, up and down are relative to the direction normal to the plane of the current flow. This spin population difference is analogous to the voltage difference that develops transverse to the current in the presence of a perpendicular magnetic field in the ordinary Hall effect. Anyway, the bottom line is that one can produce separated spin populations without actually injecting spins from a ferromagnet or something similarly difficult. The spin Hall effect can be intrinsic (due to spin-orbit coupling and a built-in electric field or lack of inversion symmetry in the material) or extrinsic (due to spin-dependent scattering off of disorder in the material). One of the first (the first?) observation of spin Hall was made by Awschalom's group at UCSB, using spatially resolved magneto-optic Kerr to map the spin density.

Anyway, in this paper the authors claim to observe the inverse spin Hall effect. That is, they establish an unequal spin population between edges of a sample using a spatially varying intensity of circularly polarized light to generate polarized carriers. Then, they observe a dc current transverse to the spin density gradient. The data look pretty convincing, though I'm no expert in photophysics of III-V materials.