There's an old quote from Isaac Asimov that is very true: "The most exciting phrase to hear in science, the one that heralds new discoveries, is not 'Eureka!' (I found it!) but 'That's funny...'." In my group, we recently had an experience that supports this, and it highlights what I think is some of the most fun you can have as an experimental scientist: trying to use the tools at your disposal to learn as much as you can about what's behind some unexpected and surprising phenomenon.
The story starts out several years ago. We'd been having some nice success making single-molecule electronic junctions and using them as model systems to study a particular piece of physics, the Kondo effect. Our theorist colleagues pointed out that these molecular devices, unlike typical semiconductor quantum dots, might give us an opportunity to study a particularly rich and interesting piece of physics called a quantum phase transition, because molecular devices are easier to attach to ferromagnetic electrodes. The idea, not directly germane to this story, is that an unpaired electron on the molecule is torn between two competing "baths" of excitations. On the one hand, the unpaired electron can undergo Kondo processes with the conduction electrons of the electrodes. On the other hand, spin waves in the ferromagnetic electrodes can also talk to the unpaired electron. By varying a gate voltage appropriately, the hope was to tune from one limit (the Kondo regime) into the other (expected to be a more exotic "non-Fermi liquid" state).
Anyway, for various reasons, it became clear that working with palladium electrodes might be a good place to start. Palladium is almost ferromagnetic. That means that it has long-lived spinwave-like excitations (paramagnons). At the same time, it's chemically friendlier (less prone to forming magnetically complicated oxides) than common ferromagnetic metals like iron, nickel, or cobalt. So, step zero of this project would be to make some bare palladium tunnel junctions (just two pointy palladium electrodes, without any molecule bridging them) and make sure that they're simple and boring, as expected. After all, we'd looked at literally thousands of gold tunnel junctions like this, and if properly made (so that you don't have extra metal nanoparticles around), they are dull as dirt: current-voltage (I-V) curves that are nearly linear and essentially temperature-independent.
Surprise! My postdoc, Gavin Scott, found that Pd tunnel junctions are very much not boring. While they look dull at, say, 10 K, if they are cooled down to lower temperatures, all kinds of sharp features appear in their differential conductance (dI/dV as a function of V). The features appear at voltages symmetric around V = 0, and they evolve with temperature in a very interesting way. In fact, if you look at the temperature dependence of those features, it looks very much like what you see for the temperature dependence of the order parameter in a "mean field" phase transition. We spent months trying various things, turning all the easily turned "knobs" like temperature, magnetic field, gate voltage, etc. One striking trend is the observation that, looking at all of our devices on one set of axes, the voltages where the conductance features appear extrapolate to zero when the conductance of the junction approaches e2/h. In other words, when the metal tips touch, the whole effect goes away.
It's been science-as-puzzle-solving, trying to figure out what could be going on here. We came up with many possible explanations, and tried to come up with ways to test the possibilities, eliminating the ones that didn't fit. For example, the data look (qualitatively) rather like superconductor tunnel junctions, but the quantitative values (specifically, the relationship between the voltage scale and the temperature scale) are far, far away from numbers that would make sense for a superconductor. In the end, we think that the most likely physics ingredient is the onset of magnetic order at the tips, though that's not a perfect explanation by any means. It is clear, though, that Pd is special - other metals (Au, Ni, Pt) just don't seem to show this. The paper is out here (email me if you want a copy). Hopefully others will get interested. Suggestions are always appreciated. In the meantime, it's a good example of how sometimes systems that you think are dull can surprise you, and how science is supposed to work.