Magnetite

Now that it's been published online, I can talk about our new paper about magnetite. Back in August I wrote a post about the different types of papers that I've been involved with. This one fits the third category that I'd mentioned, the (Well-Motivated) Surprise, and it's been a fun one.

Background: Magnetite is Fe3O4, also known as lodestone. This material is a ferrimagnet, meaning that it has two interpenetrating lattices of magnetic ions with oppositely directed polarizations of different magnitudes. Since one polarization wins, the material acts in many ways like a ferromagnet, which is how it was first used in technology: to make primitive compasses. The magnetic ordering temperature for magnetite is about 860 K. Anyway, at room temperature magnetite has a crystal structure called inverse spinel, with two kinds of lattice sites for iron atoms. The A sites are each in the center of a tetrahedron with oxygen atoms at the corners, and are occupied by Fe(3+). The B sites (there are twice as many as A sites) are each in the center of an oxygen octahedron, and are occupied by a 50/50 mix of Fe(3+) and Fe(2+), according to chemical formal charges.

It's been known for nearly 70 years that the simple single-electron band theory of solids (so good at describing Si, for example) does a lousy job at describing magnetite. Fe3O4 is a classic example of a strongly correlated material, meaning that electron-electron interactions aren't negligible. At room temperature it's moderately conducting, with a resistivity of a few milli-Ohm-cm. That's 1000 times worse than Cu, but still not too bad. When cooled, the resistivity goes weakly up with decreasing temperature (not a standard metal or semiconductor!), and at about 120 K the material goes through the Verwey transition, below which it becomes much more insulating. Verwey first noticed this in 1939, and suggested that conduction at high temperatures was through shifting valence of the B-site irons, while below the transition the B-site irons formed a charge ordered state. People have been arguing about this ever since, sometimes with amusing juxtapositions (hint: look at the titles and publication dates on those links).

Motivation: I'd been interested for a while about trying to do some nanoscale transport measurements in strongly correlated systems. The problem is, most relevant materials are very difficult to work with - not air stable, difficult to prepare, etc. Magnetite is at least a well-defined compound, and the Verwey transition acts as something of a gauge of material quality, at least in bulk. Screw up the oxygen content by a couple of percent, and the transition temperature falls through the floor.

What did we find: In two different kinds of magnetite nanostructures, we found that the I-V characteristics become dramatically hysteretic once the sample is cooled below the Verwey transition. This was completely unexpected! Basically it looks like you can take the system, which wants to be a decent insulator in equilibrium at low temperatures, and kick it back into a conducting state by applying a large enough electric field. Reduce the field back down, and the system remains in the conducting state until you pass a lower threshold, and then the magnetite snaps back into being an insulator. We worked very hard to check that this was not just some weird self-heating problem, and that's all described in the paper. I should point out that other strongly correlated insulators (vanadium oxides; some perovskite oxides) seem to be capable of qualitatively similar transitions. Hopefully we'll be able to use this transition as a way of getting a better handle on the nature of the Verwey transition itself - in particular, the role of structural degrees of freedom as well as electronic correlations.