Saturday, November 06, 2021

The noise is the signal

 I am about to attend a gathering of some physics department chairs/heads from around the US, and I'll write some about that after the meeting, but I wanted to point out a really neat paper (arxiv version here) in a recent issue of Science.  A group at Leiden has outfitted their scanning tunneling microscope with the ability to measure not just the tunneling current, but the noise in the tunneling current, specifically the "shot noise" that results out of equilibrium because charge is transported by the tunneling of discrete carriers.  See here for a pretty extensive discussion about how charge shot noise is a way to determine experimentally whether electrons are tunneling one at a time independently, or whether they are, for example, being transported two at a time because of some kind of pairing.

Adapted from Fig. 1 of this paper.
The experiment is quite pretty, looking at disordered thin films of TiN, with a macroscopic superconducting transition temperature of \(T_{c} =\) 2.95 K.  With the shot noise measurement, the experimenters see enhanced noise at low applied voltages consistent with pairing (with pairs being transported presumably by the process of Andreev reflection).  The interesting point is that this enhanced noise persists up to temperatures as high as 2.7 times \(T_{c}\), despite the fact that the tunneling conductance \(dI/dV\) shows no sign of a gap or pseudogap up there.  This implies that superconductivity in this material dies as \(T\) exceeds \(T_{c}\) not because the pairing between electrons falls apart, but instead because of the loss of the global coherence needed for the superconducting state.  That's an exciting result.  

I'm a big fan of noise measurements and applying them to a broader class of condensed matter systems.  We'd seen enhanced noise in cuprate tunnel junctions above \(T_{c}\) and at large biases, as mentioned here, but in the cuprates such persistence of pairing is less surprising than in the comparatively "simple" TiN system.  Noise measurements on demand via STM should be quite the enabling capability!

6 comments:

  1. If the old adage goes "one person's trash is another person's treasure", perhaps the analogous phrase here would be "one scientist's noise is another scientist's signal"!

    ReplyDelete
  2. Yes, yes, yes!!! More noise measurements and lots of signal analysis, please. Look out for pareidolia, because we will find the weirdest things if we look, but some of those weird things will be systematic and useful and different from what we'll see if we look only for events and particles.

    The identification of events on a noisy signal line out of an apparatus is already a fraught —typically hardware-implemented— signal analysis: I want to know, at a ridiculous extreme, what we'll see if we record the signal level out of a device every picosecond, and analyze the Terabytes that would give us (though at what computational cost?!?), instead of accepting only the events that a single hardware-implemented digestion of the signal gives us. I'll settle for multiple hardware-implemented analyses of the same signal, as in this case, of the tunneling current and of the noise in the tunneling current, but that's only two out of a lot more. I hope this is coherent enough to make some sense.

    ReplyDelete
  3. Anonymous1:13 AM

    Would not the more simpler Nernst effect measurements tell us the same thing?

    ReplyDelete
  4. @PPP, as an early 21st century nanoscience person, I learned a long time ago that one person's packaging problem is another person's sensor.

    @Peter, the real challenge with a comprehensive approach (e.g. extracting the full counting statistics of electrons flowing vs time through a device - the first moment (average current), the second moment (the noise), and all higher moments) is that usually the timescales are extremely short. A one microamp current = 10^13 electrons per sec.

    @anon, Nernst is nice and has been applied to the high-Tcs by, e.g., Ong, but I think this really tells you something different. For example, you can get energy-resolved information thanks to the bias scale. If there are more sophisticated forms of Nernst measurements, I'd very much like to learn about them, as I haven't worked on that technique much.

    ReplyDelete
  5. Anonymous2:18 AM

    Is there a reason a normal (low-temp, UHV) STM can't due these measurements?

    ReplyDelete
  6. Anonymous1:39 AM

    This signal to noise ratio is all pervasive problem in many measurements. Even a simple Thermogravimetry has this. There was a paper on superconductivity as a preprint claimed superconductivity at Room temp, by reading noise more than the signal. Here is a Prof Tim Palmer explaining very well about a graph on climate science to Sabine Hossenfelder and clearing all questions about the signal/noise ratio. This misreporting of signal to noise ration aspect is also responsible for some questionable papers even in High impact journals

    https://www.youtube.com/watch?v=-fkCo_trbT8

    ReplyDelete