Wednesday, February 28, 2007

Great talk today

Alain Aspect gave the departmental colloquium today, and his talk was fantastic. He let the audience choose whether to hear about his more recent work on the Hanbury Brown-Twiss experiment with cold atoms, or his very famous work on Bell's Inequalities. By show of hands the packed audience picked the latter, and Aspect gave an extremely clear talk about why local hidden variable theories like the kind desired by Einstein just aren't compatible with quantum mechanics. I know that the talk has been fine-tuned and updated over the years, so the fact that it's polished shouldn't be surprising. Still, it was an impressively well structured colloquium: a good, generally accessible set-up and statement of the problem, a discussion of the experiment and what it means, and conclusions updated to include modern experiments about entanglement and quantum cryptography.

Sunday, February 25, 2007

Physics, smarts, and perspective

There's a great post on Cosmic Variance about the "cult of genius" in physics - the myth in our discipline that if you're not supermegabrilliant (Feynman/Einstein/Hawking, as Julianne puts it), you're basically a pedestrian loser. Hand in hand with this is the still-persistent attitude out there that if you get a physics PhD but don't end up a full professor at Harvard, you're a plodder. Read the post and the comments. It's great stuff. It also makes me remember my first real intellectual wake-up call, realizing that I was surrounded by really smart folks and would have to get used to it. First semester, freshman year, taking this class from this fellow, and getting 6 out of 30 on the first exam. The mean was a 9. One real advantage to getting an undergrad degree at a top-tier place is the character-building early realization that there are many people smarter than you. Better to come to that conclusion at 18 than at 22 or 25....

This week in cond-mat

One theory paper, and two experimental papers this time.
cond-mat/0702446 - Poggio et al., Feedback cooling of a cantilever's fundamental mode below 5 mK
Suppose you had a mechanical resonator (mass on a spring). At moderate temperatures you know from the good, old equipartition theorem that the average kinetic energy and average potential energy in the resonator would each equal 1/2 k_{B}T. (Note to self: get LaTeX working in blogger....) At low enough temperatures (k_{B}T < \hbar \omega), you should instead think about the number of vibrational quanta in the resonator. Suppose you could actively damp the resonator - if it's moving toward you, you push back to slow it down. It is possible to effectively cool the resonator this way (though in a Maxwell's demon sense, there's no such thing as a free lunch). How far you can go depends on the noise in your measurement system used for the feedback. In this paper by Dan Rugar's group, they demonstrate that they can cool a Si cantilever from a base temperature of around 4.2 K all the way down to 5 mK, limited by the noise in their feedback system. This is impressive, and of obvious interest to those who want to examine the fundamental quantum properties of mechanical systems (including detector back-action).

cond-mat/0702472 - Kalb et al., Organic small-molecule field-effect transistors with Cytop(tm) gate dielectric: eliminating gate bias stress effects
A persistent problem with organic FETs is that their performance degrades if the gate is biased for long periods. There can be many reasons for this, but one major issue involves the interaction between the semiconductor and the gate dielectric. It is widely believed that in many OFETs charge leaking through the gate dielectric introduces defects and trap states right at the channel interface in the organic semiconductor. Here, Batlogg's group at ETH seems to have found, with collaborators, a fluoropolymer dielectric that doesn't seem to have these problems, and has impressive breakdown strength as well. I'll have to look into getting some.

cond-mat/0702505 - Khodas et al., One-dimensional Fermi-Luttinger liquid
Fermi liquid theory is the standard model of electrons in metals (as well as normal-state liquid 3He). The upshot of FLT is that the quasiparticles of the interacting electron gas look very much like weakly interacting electrons, and have well defined quantum numbers (spin 1/2, charge -e, k-vectors and band indices). In 1d, though, FLT doesn't do well. Luttinger, by assuming that the dispersion E(k) of the carriers around the Fermi points is linear, came up with an exact solution to the 1d problem now called the Luttinger liquid (LL). The LL has some very interesting properties, including separate spin and charge excitations. In this paper, Glazman, Pustilnik, Khamanev, and Khodas consider what happens when the dispersion a the Fermi points is more realistic: linear with a little bit of quadratic correction. This breaks particle-hole symmetry around the Fermi points, and has some profound effects on the structure of the density of states. This is a long paper, and while I think I get the main point, I haven't had a chance to look at it thoroughly. It seems important, though, since the slight nonlinear correction considered here seems very physically reasonable for many systems.

Saturday, February 17, 2007

This week in cond-mat

Several papers caught my eye this week; I'll be brief, particularly since I haven't had time to read them in detail. Now that our paper is in and our search is nearing the end, I'll have more time soon. Maybe I'll even get time to work on my book. Anyway....

cond-mat/0702246 - Capelle et al., Energy gaps and interaction blockade in confined quantum systems
The authors consider the general problem of interacting quantum particles confined in a harmonic potential. This could apply to electrons in a small quantum dot, or cold atoms in a magneto-optic trap. They then come up with expressions for the addition energies (how much energy is needed to add one more particle to the confined, interacting system) based on single-particle properties plus the interactions. They predict phenomena analogous to Coulomb blockade for other interacting systems, including some kind of Van der Waals blockade for trapped atoms.

cond-mat/0702259 - Kornyushin, An introduction to the polaron and bipolaron theoretical concepts
This looks like a nice pedagogical derivation of polarons and bipolarons. Should be good for students.

cond-mat/0702332 - Wu et al., Shot noise with interaction effects in single walled carbon nanotubes
This is a typically nice piece of experimental work from the Helsinki group. They've measured shot noise in carbon nanotube devices, and while they have seen interesting quantum coherence effects (Fabry-Perot electronic resonances as have been observed in dc conduction in these systems), they do not see any clear signs of Luttinger liquid physics.

cond-mat/0702348 - Phillips, Mottness
This is a longer article by Phil Phillips on his ideas about the properties and excitations of Mott insulators - materials that are insulating not because their bands are all full, but because strong electron-electron interactions lock the carriers in place. Interesting ideas explained in a compelling way, though theorists have been arguing about this stuff (in particular, the role or lack thereof of Mott physics in, e.g., the normal state of the high Tc compounds) for some time. Prof. Phillips is also the best dressed scientist I've ever met, bar none.

Tuesday, February 13, 2007

Quantum computing: are we there yet?

(Updated and corrected) As others in the blogging world have pointed out, today is the big day for D-Wave, a privately held, VC-financed Canadian company that plans a public demonstration of a 16 qubit quantum computer. One of the main ideas behind quantum computation is that, because of the way quantum mechanics works, performing a linear number of operations, N, allows you to build up quantum states that can be written as superpositions containing an exponentially large (e.g. 2^N) number of terms. If one can do this and not have decoherence (due to environmental interactions) mess up the superposition states, it is possible to use this property of quantum mechanics to do certain computations much faster than classical computers. Another way to view the power of this quantum parallelism: suppose you want to solve a math problem, and the input is an N-bit binary number. With a generic quantum computer, you can imagine preparing an initial state built out of N qubits that is actually a superposition of all 2^N possible inputs. Your quantum computer could then solve the problem, producing a superposition of all solutions corresponding to those inputs. Readout is the tricky bit, of course, since simple-minded measurement of the final state will only pick out one of those solutions.

There have been many ideas proposed for physical implementations of quantum computers. The requirement that decoherence be small is extremely restrictive. With so-called "fault-tolerant" quantum computation, one can beat down that requirement a bit by using additional qubits to do error correction. In the last few years, there has been great progress in using small superconducting systems as quantum mechanical bits (qubits), either thinking about the charge on small "Cooper pair box" metal islands, or persistent currents in superconducting loops with Josephson junctions. One can do a form of quantum computation using NMR, though the number of effective qubits is strongly limited in molecules. There have been proposals to use tunable hyperfine interactions in phosphorous doped Si to get around that restriction. Some people want to do quantum computation using photons, or through optical manipulations of excitons in semiconductor dots, or directly using individual electron spins in semiconductor nanostructures. The current record (6 qubits) for producing superpositions like the ones I described above, or other related superpositions (8 qubits) has been set using trapped ions.

The D-wave demo is an attempt to do adiabatic quantum computation. The idea is to formulate a problem such that one can start out with the initial data being represented by the ground state (lowest energy state) of a system of interacting qubits. Then one very gently changes the Hamiltonian of the system such that the system never leaves its instantaneous ground state (that's the adiabatic part), but arranges matters so that the solution to the problem is represented by the ground state of the final Hamiltonian. The main proponent of this approach has been Seth Lloyd. Empirically, the D-wave folks are going to use 16 qubits made out of Nb loops and Josephson junctions (as explained here), and they cool this whole mess (128 filtered leads) down to 5 mK in a dilution refrigerator.

There seem to be three big questions here: (1) Is this really quantum computation? It's difficult for me to assess this, as I'm no expert. There seem to be arguments about which problems can really be solved in the adiabatic formulation that's implemented here, and about whether one can actually get significant improvements relative to classical algorithms. (2) Will the demo be fair? The high tech world is no stranger to rigged demos, and in this is a particular black-box affair. One has to trust that the stuff displayed on the screen of the PC controlling the electronics is really being determined by the chip at the bottom of the fridge, and not by some clever software. I'm willing to give them the benefit of the doubt, provided that they let some independent experts play with the system. (3) Why haven't they published everything in the open literature and let outsiders come in to verify that it's all legit? Well, I can't say I really blame them. The paper I linked to up there for their implementation never got into PRL, as far as I can see. I don't see Intel hurrying up to get outside approval for their new gate metallization. If these folks think they can actually get this to work and make a buck at it, more power to them. The truth will out.

Saturday, February 10, 2007

A scientific direction that I think is promising

I haven't written too much about my own research on this blog, mostly because I figure that people who really care about it can read my group homepage or my papers. However, there is one area out there that I think has real promise, and I'd like to get other folks thinking about it, at least in general terms.

Electronic transport measurements in nanoscale systems can be considered a kind of spectroscopy. In particular, when a chunk of conducting material is sufficiently small and relatively weakly coupled to leads (call them a "source" and a "drain", after transistor terminology), conduction can be dominated by one or a few specific quantum states of that material. There has been great work done by many groups over the past 15 years or so, looking at these individual electronic states in a bunch of systems, including metal nanoparticles, patches of doped semiconductor, and semiconductor nanowires and nanocrystals. As neat as these systems are, they're all comparatively simple from the electron-electron interaction point of view. With a few exceptions (like Kondo-based physics), you can pretty much work in a single-particle picture. That is, adding one more electron to these systems doesn't drastically change the spectrum of electronic states - the spectrum itself is mostly unchanged except for the population of the states, one of which has increased by 1.

Many interesting materials exist where strong electronic correlations are more important. For example, the high-Tc superconductors in their normal state are often "bad metals" that are not well described by a picture of weakly interacting electrons. There are similar phases in the heavy fermion compounds. Even magnetite (Fe3O4), a comparatively simple compound, has strong correlation effects: it's not really a metal or a semiconductor; it has a room temperature resistivity in the milliOhm-cm range (say 1000 times higher than Cu or Au), and that resistivity increases with decreasing temperature, but not in a simple way as in a semiconductor.

I think it would be very revealing for transport spectroscopy experiments to be performed on nanostructures made from these strongly correlated materials. This won't be easy for many practical reasons (e.g., stoichiometry can be tough to control in nanomaterials; noone knows how to make many of these systems in nanostructured forms yet), but I'm convinced that there is much to learn in such experiments.

Another claim to fame

See this comic? See how, down at the bottom, it says "This comic courtesy of Jeff from Rice U."? That's my grad student, Jeff, who has been having problems with me walking by and having his devices die mysteriously.

Monday, February 05, 2007

My touch with fame

I can't resist posting a link to this article (NY Times, reg. req.), about two friends of mine from college. If you ever see a joke on The Daily Show that involves pretty serious math or science, there's a good chance it was written by Rob. He had a great one a year or two ago involving Venn diagrams....

Friday, February 02, 2007

This week in PRL (last year in cond-mat)

Real life continues to limit my blogging time. I'll hopefully be posting more often again soon. In the mean time, here's a neat paper that just came out in Phys. Rev. Lett. today, and was actually on the arxiv last year:

cond-mat/0603079 - Matthey et al., Electric field modulation of
transition temperature, mobile carrier density and in-plane penetration depth in NdBa2Cu3O(7-delta) thin films

In this
work the authors grow (by sputtering) underdoped high-Tc superconducting films on top of a SrTiO3 gate dielectric with an underlying gate electrode. A number of people (e.g. Allen Goldman's group at Minnesota) have played with SrTiO3 as a high-k gate dielectric to do experiments involving large gated charge densities. It's almost a ferroelectric, so it is possible to get an extremely large electric polarization in that material. The reason to do this is that in principle it allows you to tune the carrier density in an overlying material via the field effect: in a properly designed field-effect transistor, applying a potential difference between the gate electrode and the source/drain electrodes capacitively accumulates or depletes charge at the interface between the overlying material and the dielectric. Of course, there's no guarantee that the interface is nice, and that all the gated charge is actually mobile, even at "clean" interfaces between simple materials (Si, SiO2). However, if you can get it to work, you can tune charge density (at least in a thin layer of material) without accompanying changes in disorder that result from chemical doping. Anyway, the authors of this paper have managed to get this approach to work surprisingly well in this essentially 2d (the sample is only 3-4 unit cells thick) high-Tc material, and can electrostatically tune the transition temperature by about a factor of two within a given sample. This has allowed them to do detailed studies of the superconductor-insulator transition that happens as a function of carrier density, without having to worry about variable disorder. (This kind of phase transition, driven by a control parameter rather than temperature, can occur at T=0 and is called a quantum phase transition.) Very nice stuff. They've been working on this for several years, and it's nice to see them succeed. I met Jean-Marc Triscone, the PI, when we were working on this.