Thursday, May 31, 2007

Hype. Again.

Remember this post, where I reported on interesting Shubnikov-deHaas oscillations in very pure high-Tc material? Well, that paper has now come out in Nature. Unsurprisingly, there has been an associated flurry of press, including this article. In case that link doesn't work, I'll spoil the punchline for you:
Canadian physicists have cracked a decades-old mystery surrounding metals that carry electricity without resistance, opening the door for everyday trains that levitate on magnetic fields, ultrapowerful quantum computers and big savings for utilities.
Taillefer predicted the discovery would lead to room-temperature superconductors within 10 years, triggering a technological revolution similar to the invention of the transistor.

One of the most promising applications for such superconducting metals is in magnetic levitation trains, which can theoretically run at speeds of up to 500 km/h.
Other possible superconducting applications include shrinking MRI machines to the size of laptops, eliminating the 10 to 20 per cent electricity lost from resistance inside power stations and building quantum computers, machines so powerful they would make today's supercomputers resemble mere pocket calculators.

Wow. They get from Shubnikov-deHaas oscillations to room temperature superconductors to maglev trains and quantum computers.
I had no idea that getting clean samples could do so much. I'm presuming that most of the fault for this lies in the journalism rather than the scientists, but let this be a cautionary tale.

Annoying conventions

What do you find to be the most annoying conventions in physics? The classic example is the choice (darn you, Ben Franklin) of sign for the charge of the electron. Franklin had a 50/50 chance, and we ended up with the often confusing situation that current flow and particle flow are oppositely directed, that "up" on energy level diagrams corresponds to more negative voltages, etc. [EDIT: I think my earlier statements here about UPS conventions stem from a particular paper that isn't representative; never mind.... ]. Do any of you out there have other examples of really bad/misleading conventions?

Thursday, May 24, 2007

This week in cond-mat

Only one quick blurb for now - there have been a number of neat looking papers on the arxiv lately, but I just haven't had time to read them. I am actually making some progress on my book, though.

arxiv:0705.2180 - Martin et al., Observation of electron-hole puddles in graphene using a scanning single electron transistor
A single-electron transistor (SET) consists of an "island" (in this case, a patch of aluminum film) weakly connected by tunnel barriers (in this case, aluminum oxide) to source and drain electrodes (also aluminum films here). Defining the total capacitance of the island to be C, the Coulomb energy cost of adding another electron to the island is E_c ~ e^2/C. If E_c >> kT, the thermal energy scale, and the tunneling resistances of the barriers are >~ h/e^2 (~ 26 kOhms), then the number of electrons on the island is fixed to be an integer. By varying the voltage on a nearby gate electrode coupled capacitively to the island, it is possible to change the average population of the island by one electron at a time. When the gate is set such that the island is just on the cusp of going from an electronic population of n to n+1, the source-island-drain conductance of the device has a peak and is very strongly dependent on that gate voltage. Instead of using a gate electrode, one could use the local electronic environment near the island to modulate the island potential (and hence the conductance). SETs are incredibly good electrometers, able to sense tiny fractions of an electronic charge nearby. Now consider sticking such an SET electrometer on the end of a scanned probe tip (in this case, fabricate it directly on the end of a tapered optical fiber). This is the scanning SET, a wonderful imaging tool developed and refined originally at Bell Labs by people like Harald Hess, Ted Fulton, Bob Willett, Mike Yoo, Amir Yacoby, and Nicolai Zhitenev.

In this paper Amir and colleagues (von Klitzing and company) use the scanning SET to look at graphene near the charge neutrality point as well as in the quantum Hall regime. They can see how the system breaks up into puddles of electron-rich and hole-rich regions with ~ 100 nm spatial resolution. This is a nice application of the S-SET technique, which can be extremely arduous - meeting the temperature requirement for good charge sensitivity requires working at very low temperatures (at least 3He fridge); the SET itself is very fragile and static sensitive; and the scanned probe setup is easy to crash into the sample surface. All in all, a tour de force tool that is unlikely to make its way into common usage any time soon.

Thursday, May 17, 2007


I got a very surprising email this morning from the NSF. Someone made a Freedom of Information Act request to get a copy of one of my NSF grant proposals. Now, I know that technically this is allowed - in principle, if someone wanted to, they could get (via FOIA) copies of their direct competitor's federal grants (with certain privacy information like social security numbers redacted). However, I've never actually heard of anyone doing this in practice - it's just not cricket, so to speak. The NSF gave me the name of the person, and I'm left to wonder: did they do this just to see an example of a funded proposal? Why didn't they contact me directly? Did they know that NSF was going to tell me about this? It's all perfectly legal, but I find it unsettling, and I can't pinpoint the precise reason. Has this ever happened to anyone else out there?

SCES '07 day 4

Back to the SCES conference this afternoon, after my campus commitments, for the second session on strong correlations in mesoscopic systems. Some highlights:

David Goldhaber-Gordon gave a nice talk about his group's work on using semiconductor nanostructures to engineer the two-channel Kondo effect. The work has been published here and is available on the arxiv here. In the single-channel Kondo problem, a free spin is coupled via tunneling to a single electronic bath. Antiferromagnetic exchange between the spin and the conduction electrons leads to the formation of a singlet at low temperatures - the spin is screened, and the ground state of the system is a Fermi liquid. In the two-channel Kondo problem, a single spin is coupled via tunneling to two independent electronic baths. Each bath tries to "screen" the spin via antiferromagnetic exchange, with the result that the spin is overscreened. The ground state of that system is supposed to be a non-Fermi-liquid, meaning that its low energy excitations don't look like weakly interacting quasiparticles. The hard part about testing this is actually making two truly independent electronic baths. The paper shows a clever implementation that effectively does this, at least over a limited temperature range.

Yong Chen, one of Randy Hulet's postdocs, gave a talk about using cold atoms to study Anderson localization. By sending a laser through frosted glass, they can use the resulting speckle pattern to provide a disordered potential for trapped cold bosonic atoms. They can dial around the strength of the disorder potential by changing that laser's intensity. Then they can play games with the trap potential to test how delocalized the Bose-condensed atoms are (kick the trap and look for resulting oscillations), and independently check for coherence by looking for interference fringes. The preliminary data are pretty exciting.

Ravin Bhatt talked about (theoretical) ways to try and produce ferromagnetism in doped semiconductors containing only nonmagnetic atoms, at very low carrier densities. The trick is to somehow get the system to have more electrons than there are donors. One can imagine doing this with clever modulation doping schemes. No one's pulled it off yet, but it sounds cool and the numerical results look suggestive.

Unfortunately more Rice commitments mean that I won't make it to the last day of the conference tomorrow. Ahh well. It was an interesting meeting.

Tuesday, May 15, 2007

SCES '07 day 2

Again I only was able to see the morning session today (and will be at Rice until Thursday pm). This means I'll miss the big "BCS@50" plenary session. However, here are a couple of talks that I did get to see....

First, T. Senthil started the day with a talk about spin liquids. This is a theoretically deep concept that I would love to understand better. The basic idea is that one can recast the interacting many-body problem in terms of new excitations of spinons (chargeless spin 1/2 excitations). The cost of doing this is that the spinons have "infinitely nonlocal" statistical correlations. However, these interactions can be made to look simple by introducing some effective gauge "charge" for the spinons and some effective gauge "magnetic field" - then the correlations look like the Aharonov-Bohm effect in this gauge language. If this sounds vague, it's partly because I don't really understand it. The upshot is that the spinons can be fermionic, and therefore have a Fermi surface, and this leads to nontrivial low temperature properties, particularly in systems where the whole weakly interacting quasiparticle picture falls apart. If anyone can point me to a good review article about this, I'd appreciate it.

There were a couple of other strong theory talks. Natan Andrei talked about a general approach to quantum impurities driven out of equilibrium (e.g., as in a quantum dot in the Kondo regime at large source-drain bias). Strong correlations + nonequilibrium is a tough nut to crack. Andrei argued that one can rewrite the problem in terms of scattering of initial states via simple phase shifts, provided that one picks the right (nasty, complicated) basis for the initial states that somehow wraps up the strong correlation effects. This choice of basis is apparently a form of the Bethe Ansatz, which I also need to understand better.

On the experimental side, besides my talk, Gleb Finkelstein from Duke gave a very nice talk about Kondo physics in carbon nanotube quantum dots. The really clever aspect of the work is that, through careful engineering of the contacts to the tube, the actual leads to the dot + the tunnel barriers + the dot itself are all formed out of the same nanotube. As a result the tunnel barriers preserve the special band structure symmetry (SO(4)) of the tube and the leads, leading to profoundly neat effects in transport.

Monday, May 14, 2007

SCES '07, Day 1

Today was the first day of the 2007 International Conference on Strongly Correlated Electron Systems (SCES), hosted this year in Houston jointly by Rice and UH. It's a pretty big meeting, typically with between 600 and 700 participants. Traditionally the meeting has had a very strong European and Asian participation, with a focus on heavy fermion compounds and high-Tc superconductivity. This year, there's an increased inclusion of strongly correlated physics in mesoscopic systems (quantum dots, nanotubes, graphene, single-molecule devices), as well as discussion of model correlated systems based on ultracold trapped atoms and molecules.

Because I'm on an internal search committee for a dean, my semester still hasn't really ended, which means that I'm going to miss a fair bit of the meeting. However, I'll still try to blog a couple of highlights daily. Here are two neat, new (and as yet unpublished) results that I saw this morning.

First, Louis Taillefer spoke about new measurements done in extremely high quality hole-doped YBCO at high magnetic fields (pulsed up to 60 Tesla). This work has been focused on trying to suppress superconductivity with a field in this, the grand-daddy of the high-Tc compounds, and to understand the ground state and possible quantum phase transitions (as a function of doping) of the normal phase. The exciting new result is that Taillefer and collaborators have been able to see Shubnikov-deHaas oscillations in this material for the first time. This is a big deal. First, it tells you that there are some sort of excitations in the normal state that can execute closed cyclotron orbits in the presence of a magnetic field. Since the validity of weakly interacting quasiparticles in the normal state is in significant doubt, this is interesting. Second, the frequency of the oscillations in 1/B reveals the area enclosed by those orbits in k-space - essentially it tells you how big the hole pockets are in the Brillouin zone, and therefore how many mobile holes there are per copper atom. Third, the temperature dependence of the S-deH oscillations lets you infer an effective mass (in this case, about 1.9 free electron masses) for whatever's doing the cyclotron motion. Very neat!

Second, Abhay Pasupathy from Ali Yazdani's group at Princeton showed some beautiful new STM data on BSCCO. The neat thing here is that their superfancy STM is absurdly stable over a big temperature range for days. That means that they can map out the tunneling density of states of the material on the atomic scale as a function of temperature, from deep within the superconducting state to well above the resistively detected Tc. They see that the gap in the density of states indicative of pair formation vanishes nonuniformly over the surface, with local bits persisting to well above the average Tc. They also show that the temperature dependence of the gap as a function of gap size is very different than that in low-Tc materials.

Sunday, May 13, 2007

Tenure - some advice.

This past week there was a flurry of science blogging regarding tenure, such as these posts at Cosmic Varience (here and here), Rob Knop's post about his situation, Chad Orzel's commentary, and the Incoherent Ponderer's take on the tenure process here. The IP's take on things summarizes my general thoughts on the tenure process pretty well. In terms of the job pipeline, the biggest cut in population happens when trying to get a faculty position, not at the tenure stage. In reasonable departments, no one is happy when a tenure promotion case fails. Good departments (and schools and universities) try very hard to filter at the hiring level and give their faculty the resources they need to succeed. I can only think of two or three places (in physics anyway) that historically have had a "sink or swim" attitude (that is, hiring a junior person in an area today means that seven years from now the university wants the best senior person in the world in that area - being in-house is not advantage), and I'm not sure that's even true anymore.

In the links above, people are mostly focused on the process and outcomes; I think it would be useful to give some suggestions about how to approach the tenure process from the position of the junior candidate. I am hardly in a position to give too much sage advice about tenure, and what follows below is largely common sense. Obviously the situation is different in various disciplines and at different universities, but here's some basic points that I think should be considered. I'm sure I'll leave things out - feel free to chide me in the comments.

Understand the process. Find out how the tenure process works at your institution. This should be written down in a faculty handbook. Talk to your department chair, your faculty mentor (if your department has such a thing) or senior colleagues. Understand the timeline. Get a sense of the weight that your institution places on the different components of the job (see below). Does the departmental vote carry a lot of weight (as it usually does at Rice, for example), or are the deans or the university promotions and tenure (P&T) committee commonly overriding departmental decisions?

The process probably goes something like this: the candidate is hired for a 4-year tenure-track appointment, with some kind of annual reviews and a more major renewal review in year 3 or 4. (This gives the university a chance to end the process early if there's a major problem with an assistant prof, and forces departments to give some concrete feedback to the assistant prof about how they stand.) In the summer before year 6 (at most places) the candidate is asked to put together a dossier (complete CV, reprints of papers, a summary of funding, a statement about university service, a statement about teaching, a summary of research accomplishments, etc.) and suggest names for external evaluators. The department comes up with additional names for external evaluation, and sends the full dossier to some mix of the external people. Eventually these external letters come back, and the department reads them, puts the whole package together, and there's a vote of the tenured faculty (in October or November) about whether to recommend the assistant prof for tenure. The departmental recommendation then goes to the cognizant dean, and from there to the university P&T committee (which generally would have people from all sorts of disciplines on there, from bio to French lit). Sometimes P&T committees or deans can request more external letters, and they get copies of teaching evaluations, etc., and may meet directly with department chairs. Eventually the P&T committee makes its decisions (in late spring) and the candidate finds out. That decision is finally signed off by the president of the university and the board of trustees.

The research component. To get tenure you need actually need to be getting science done. There's no sure-fire recipe for success here, but let me make a few suggestions:
  • Have a mix of projects that range from easier to high-risk/high-reward. Having only one major project can be very risky, particularly if it takes five years to get any results. One key element of getting tenure is that people in your community need to know who you are, what you've done, and what you've been doing that's really yours - new stuff from your professorial position, not rehash of your thesis or postdoc work.
  • Make sure that your colleagues know what you're doing. Your colleagues are going to need to understand your work at least on some level, and particularly for hard projects, they will need to have some idea why it may take four years before a paper comes out.
  • Have backup plans. High risk things may not succeed (no kidding.). Make sure, for your students' sake and yours, that you have thought out the projects well, so that even if you don't achieve the BIG goal, you are still learning useful things that are worth publishing.
  • Have a high attempt frequency for funding. If there's literally only one agency in the world that funds your work, that's risky and unfortunate. Make sure that you know what your options are for funding sources. Call up program officers. Ask to get a chance to serve on review panels - you'll learn a huge amount about writing proposals that way! Know if there are state funding opportunities. Think ahead about private foundations (e.g., Research Corporation).
  • Do some self-promotion but don't sell your soul. If your external evaluators don't know who you are, that's the kiss of death. Make sure you give talks at meetings. See what you can do about getting invited to give seminars at other schools. Yes, this is one issue where "well-connected" people really benefit, but if you go to meetings and get to know the people in your field, it's not that bad. Get involved in your own department's seminar series, and invite in people that you'd like to meet and talk to.
  • Publish good stuff. This is always the tricky bit, and people joke about the "least publishable unit". Still, holding back everything for the one big Nature paper that may not happen is not necessarily the best strategy, for you or your students.
The teaching component. Do a good job teaching. Most universities have resources available to help you - teaching centers that will videotape your lectures, offer suggestions for improved technique, etc. Good teaching can only help tenure in limited ways at a research university, but poor teaching can certainly hurt a borderline case.

The service component. Do a decent job in departmental and university service. Don't let it eat all your time, but get involved in things that matter to you. It's also a good way to get to know your administrators and people in other departments. I'm not suggesting currying favor - just be a solid citizen.

Common sense. People argue about whether blogging can hurt your tenure chances. Blogging is only one example of a public forum, though. Use some common sense. Publicly badmouthing your institution, colleagues, administrators, etc. is not a good idea. (I'm not talking about hushing up legitimate grievances - I'm saying don't antagonize people gratuitously.)

I'm sure I could write more, and will probably update this later. This is some food for thought for now, at least.

Thursday, May 10, 2007

The trouble with mercenaries

The trouble with mercenaries is that they can be bought. For example, the state of Texas bribed fair and square - errr, gave $40M and lots of tax incentives - to International Sematech, the big consortium of semiconductor manufacturers, in 2004 in exchange for them staying in Austin. That worked out really well for all concerned: today Sematech announced that they're picking up and moving to Albany because New York offered them more money. If I was Gov. Perry, I'd be pretty annoyed.
Update: Some Sematech people came to Rice yesterday and were grilled a bit about this. They say that the New York business is a parallel operation and won't affect their Texas activities; they also said that the reporting on this was pretty awful. Interesting. I guess time will tell about how much the focus of their work shifts to Albany. Given that the state of NY put over $3B into that setup, it seems like Texas will either have to do something similar, or face a possible eventual slide into secondary status.

Thursday, May 03, 2007

An article I'd missed

While I was traveling, the Wall Street Journal ran this article about Bob Laughlin and his tenure as president of KAIST, one of the premiere research universities in South Korea. The article is definitely worth a read. It has some classic understatements:
Dr. Laughlin, a Stanford physicist, is a talkative man quick to express his opinions.
KAIST hired Laughlin to come in and shake things up. When he did, he did so in characteristic Bob fashion, and they reacted negatively. Things went south from there:
In an attempt to assert his control, Dr. Laughlin in December 2005 set out to personally interview and evaluate every one of Kaist's 400 or so faculty members, focusing mainly on the quality of their research projects and academic work. For those professors who agreed to the interviews, he gave them a one-paragraph summary grading their work from "unimportant" to "very important."
Yeah, that may have rubbed people the wrong way.

To be fair to Bob, the leaders of KAIST were crazy to hire him - all issues of personality clashes aside, he'd never managed a group of more than a handful of people, let alone an enormous research institution with a complex bureaucracy, large staff, and huge budget. Surprise: a Nobel prize in physics doesn't automatically imply success in extremely sophisticated management problems. It's also entirely possible that his assigned task was essentially impossible by design. An interesting read, anyway.

Wednesday, May 02, 2007

NSF grantees conference post mortem

It was useful to network with fellow NSF grantees for the last two days in Reno - an interesting mix of people and projects. A few observations:
  • Rice's webmail client is so pathetically slow that it's nearly unusable.
  • Hotels in Nevada walk a fine line between needing to compete with other hotels on the one hand, and trying to make sure you'd rather sit in the casino than your room on the other.
  • "Nuggets" = out. "Highlights" = in.
  • "Disruptive technology" = out. "Transformative" = in.
  • 4 hours of 5-minute talks in one day is too many, at least for me.
  • Some people don't understand the meaning of "Your talk should be three slides."
  • Most of the NER projects (high-risk, high-reward one-year single investigator grants) from FY04 in the ECS part of the engineering directorate of NSF actually worked, and were pretty cool.
  • The free wifi in the Reno airport makes up for the beeping, blinking slot machines in the terminal.
UPDATE: Rice upgraded their webmail service today. How's that for timing?