Disorder and illumination

No, this is not another grim post about the chaotic US research funding environment.  Instead I wanted to write a bit about a good example of empiricism in experimental condensed matter physics, the use of illumination to (somewhat but not entirely mysteriously) improve electronic transport in 2D electronic systems.  

This story goes back decades, and it's all about the role of "disorder" and its effects on electronic conduction.  It's been appreciated since the 1930s that, at low temperatures so that lattice vibrations are frozen out, conduction in ordinary crystalline metals and semiconductors is limited by the charge carriers (let's work with electrons rather than holes to make discussion simpler) scattering from disorder, deviations from an infinite periodic crystal lattice.  Grain boundaries, vacancies, impurities - these all can scatter electrons that would otherwise propagate ballistically through the material, and this is often modeled as a "disorder potential", a spatially varying potential energy \(V(\mathbf{r})\).  If you want the best transport properties (longest elastic mean free path), you want \(V(\mathbf{r})\) to be small in magnitude and as smooth as possible.  This is even more important if you want to study some delicate many-body state that is expected to arise at very low temperatures - you need the disorder potential to be small compared to the energy scale of that state to avoid messing it up.

In semiconductors, where the carrier density is low and screening is therefore not as good, charged defects are particularly effective at scattering.  In modulation doping, the dopants that are the source of the charge carriers in some nearby semiconductor 2D interface or quantum well are spatially distant from where the current is going to be flowing, to minimize the scattering from those ionized donors.  

For decades it has been known that, to get the best transport properties in GaAs-based (and other) semiconductor structures, it can be good to illuminate the devices at cryogenic temperatures with a red LED.  See, for example, this paper trying to explore the upper limits of charge mobility in GaAs 2D electron gas (2DEG), where the authors say "For measurement, our samples are loaded into a 3He cryostat, where a red light-emitting diode (LED) is used to illuminate the samples for 5 min at \(T \sim \) 10 K. Following illumination, we wait for 30 min at \(T  \sim \) 10 K after the LED has been turned off before resuming the cool down to base temperature."   The qualitative explanation for this is that the photons provide enough energy to excite charge carriers, and those mobile carriers can occupy trap states, rearrange themselves, and generally set up a better screened disorder potential.  In GaAs 2DEG, the result is higher mobilities (as inferred from conductivity + Hall effect) and much cleaner fractional quantum Hall effect data, showing that the post-illumination disorder is now sufficiently weak that more delicate states can form - see Fig. 1 of this paper (arXiv version) for the before/after.  As far as I know, there is not a deep, rigorous theory of how this works, but it is known empirically.  

Fig. 2 from this paper, showing electronic magnetotransport

before/after illumination by a UV LED at low temperatures.

This preprint on the arXiv this week shows that a similar improvement in transport properties can be found in structures where graphene is encapsulated by hexagonal boron nitride (hBN).  Sandwiching graphene and other 2D materials in hBN has been known since 2010 as a way of drastically improving the charge disorder situation compared to just putting 2D materials on top of SiO2.  (That paper has 8800+ citations on google scholar btw.)  Now, it is shown that if you shine 5 eV photons (deep UV = 248 nm wavelength) on such a sandwiched structure, the already-good charge environment can become even better.  Even though that energy scale is below the band gap of hBN, the light is able to kick enough charges around to smooth out some residual disorder.  Very cool.