原创 Efficiency well above 100%: Is it possible?

I am a staunch believer in the laws of thermodynamics and their plain-spoken corollary, "When it comes to energy, there is no free lunch." That's why I automatically dismiss any article or ad promising efficiency of 100% or better (you've likely seen the ones for products which promise to run your car on water or similar, or heard about "miracle machines" which produce more energy than they consume). Efficiency close to 100%, maybe -- but equal to or greater than 100%, sorry, I'm not even remotely buying it.
That's why I was shocked when I saw the headline “8000% efficient LED enables ultralow-power data transmission” in a recent article in Laser Focus World. Whoa…a little above 100% is one thing, but 8000% is truly an outrageous boast. I assumed it was a typographical error; they meant to write 80% and accidently added the zeroes in the headline.

It turns out, though, that the headline was correct. The improved efficiency has to do with how and where you measure input and output powers, plus the subtleties of LEDs and quantum mechanics. To quote the start of the article:

Certain LEDs, when heated and then run at a very low power, exhibit wall-plug efficiencies [electrical-to-optical power-conversion efficiency] of more than 100%; this effect is explained by the existence of thermoelectric pumping of the LED… The setup transmitted 3 kbit/sec at a bit-error rate of 3 x 10^-3 and an energy expenditure per bit of only 40 fJ.

Roughly speaking, the LED is pumped with thermal energy in advance, then the energy is parceled out via optical data bits, while the LED itself does not need to consume additional corresponding power. (This effect goes by several names: electro-luminescent or EL cooling, electroluminescent refrigeration, thermophotonic cooling, and optothermionic refrigeration.) You can read the detailed, highly academic article on what these researchers did in Applied Physics Letters.

I'll admit, I only could follow a few of the details of their work, which included a highly specialized GaInAsSb/GaSb double hetero-junction LED. The report states, "The setup was able to detect light power of only 7.66 ±1.6 pW at an LED wall plug efficiency of 8000 ±1700%." I interpreted this to mean that the efficiency parameter had a tolerance of a little over 20%, which is either pretty good or pretty bad depending on your perspective. Power measurements, especially related to an ultralow-power optical regime rather than an electrical one, are very challenging to implement and fickle in their consistency. At these low powers, there isn’t a lot of headroom or range—every photon counts.

The questions I always have when I look at serious, legitimate research such as this are: how do you measure energy and power in this extreme low-power region with confidence? How do you know what you are measuring is what you actually intended to measure, rather than a measurement artifact unintentionally distorted by wishful thinking? How do you verify, and calibrate, the fundamental credibility of your instrumentation?

Have you been involved with ultra-sensitive electrical or optical power measurements? What steps did you have to take to instrument the setup and compensate for factors such as temperature shift?