PhysicsPreprintExperiment4 min read

ELECTRONS THAT CROSS IN THREES

Superconductivity rests on a pact between electrons. In a superconductor, they bind into Cooper pairs, and the current flows in units of two electron charges, written 2e. That number is not hidden deep in theory: it can be read off a voltmeter.

A ruler made of microwaves

Put two superconductors very close together, separated by a thin barrier, and a current can cross without any voltage: this is the Josephson effect. Apply a voltage and something else happens. The current starts to oscillate, at a frequency set by the voltage — about 1 gigahertz for every 2 microvolts, according to the paper.

Now shine microwaves of frequency f onto the junction. The two rhythms lock together, and the voltage no longer rises smoothly as the current increases: it climbs in flat steps, called Shapiro steps. Their height follows a simple formula, V = nhf/2e, where h is Planck’s constant and n a whole number. The 2e in the denominator is the signature of pairs. If the current were carried by groups of four electrons, the steps would shrink to hf/4e.

Steps of the wrong size

Yuying Zhu, Heng Wang, Qi-Kun Xue, Ding Zhang and colleagues, at the Beijing Academy of Quantum Information Sciences, Tsinghua University and the Southern University of Science and Technology in Shenzhen, built junctions from two thin flakes of a copper-oxide superconductor called BSCCO. They peeled off the flakes and stacked them at low temperature, the top one rotated by 45 degrees. Electron microscopy confirmed the twist.

In a control junction, the steps appeared where they should, at multiples of hf/2e. But in two other junctions, measured with the same equipment, they landed at hf/3e, 2hf/3e, hf/e, and so on, up to seven steps on each side of zero, for both directions of current. The unit was no longer that of a pair, but that of three electrons.

The authors checked that nothing was hiding. At 14.8 gigahertz, the first two steps were clearly separated, and a step at the usual hf/2e would have sat right between them, plainly visible. It was not there. As the microwave frequency was changed, the step voltages rose in proportion to it, at exactly the rate expected for a charge of 3e.

Ruling out the impostors

A mixture of ordinary effects could, in principle, mimic the pattern. The team tested several:

  • Fractional steps. Some junctions add small in-between steps to the normal ones. If that were the case here, the normal steps, at hf/e and 2hf/e, would be much wider than the others — a step’s width being the range of current over which the voltage stays flat on it. Instead, the widths shrink steadily: up to about 10 microamps for the first step, 7 for the second, 2 for the third.
  • Heat. In-between steps fade quickly when the sample warms up. These persisted from 1.6 to 70 kelvin, while the supercurrent itself vanishes near 80 kelvin.
  • Trapped magnetic flux. Magnetic field trapped in the sample can create unstable in-between steps. But the pattern returned after three separate cool-downs, with the sample warmed to 120 kelvin in between.
  • Several junctions in a row. A chain of ordinary junctions cannot produce a step smaller than hf/2e. Two junctions each carrying six electrons could, but they would need to be identical, and the device has only one twisted interface.

The main junction also behaved like its ordinary cousins in one respect. After a strong current was passed through it while cooling, it became a superconducting diode, letting current through more easily in one direction. Under microwaves, when the current was flipped back and forth in direction, the output voltage switched between zero in one direction and a step of exactly hf/3e or 2hf/3e in the other. The same treatment also created half-steps, at hf/6e: half of hf/3e. Half-steps sit halfway between whole steps, so finding them at half of hf/3e confirms that hf/3e was itself the basic, whole step.

A slightly starved crystal

What sets these two junctions apart? Their superconductivity vanishes about 10 kelvin lower than in the control: they hold slightly fewer charge carriers than the optimum, probably because oxygen escaped while they were being stacked. Twelve other junctions close to the optimum, besides eight studied before, showed only the standard steps.

The authors offer a hypothesis, which they present as speculation. In a two-layer system with strong interactions, an electron in one layer could bind to a missing electron — a hole — in the other. If such pairs coexist with superconductivity, a crossing could combine one Cooper pair and one single electron. That would give a junction with a strange property: its energy would not return to its starting value after one full turn of the superconducting phase — the quantity that, under a voltage, keeps turning and makes the current oscillate. Why only this process shows up, and not the more ordinary ones that should accompany it, remains, in the authors’ words, a question for further investigation.

Odd one out

Every superconducting process established so far moves an even number of charges. A current that crosses in threes does not fit, and the paper does not yet explain it. The team sees it as a sign that, in twisted crystal layers, superconductivity competes with other forms of order, and suggests looking for the same effect in other twisted materials where those forces can be tuned.