Thermoelectric pumping - how to exceed 100 % efficiency

Physicists have known for decades that it is possible, in principle, to get more light power from an LED than the input electrical power. In other words, an LED is capable of achieving an efficiency greater than 100 %, but how does that even make any sense?

LED rainbow by steeph-k

Let's start from how LEDs work. The name LED is an abbreviation of the words Light Emitting Diode. More specifically, LEDs are usually p-n junction diodes. N-type diodes have an abundance of electrons, while p-types have holes (because if you take an electron out, it leaves a hole, right?). The electrons have a negative charge and holes positive, so they attract each other. When you make a junction out of these two types, you get a rather unique material, where the holes dominate one side and electrons the other. Now, you might think that when we slam these two materials together, the electrons will occupy the holes and the material as a whole becomes electrically neutral. Actually, this happens only at the interface of the junction, which is called the depletion region, and the rest of the material retains its original charge. There is a potential that does not allow the electrons and holes to cross the depletion region.

The interesting thing is that if we apply a voltage over the junction, we can move the electrons over the neutral region to the p-side. There the holes and electrons can recombine, and boom! We have light emission. This effect was actually experimentally observed way back in 1907 by H. J. Round, only three years after the incandescent light bulb attained it's modern design. It took a century of research until LEDs were versatile and cheap enough for the markets to widely adopt them and abandon the classical light bulb. It's not a huge surprise that happened, given the many advantages of LEDs: they are bright, they come in many colors, and above all, they are extremely efficient.

LED energy diagram, familiar to most physics students!

So where can we go from here? Surprisingly, there is still a lot of room for improvement for LEDs, and one of the most interesting possibilities is overcoming the 100 % efficiency limit! If we return to the p-n junction from earlier, there is a neat trick that you can do with them. As we now know, applying a voltage across the junction will move the electrons over the neutral region. In other words, the applied voltage will increase the energy of the electrons and that energy is radiated out as light once the electron recombines with a hole. But that energy does not necessarily have to come from electricity, it can also be heat.

Thermoelectric pumping is the idea that you supply some of the required energy in the form of heat to produce light, and then the amount of required electrical power will decrease. You would still need a source of electrons (current source), but effectively this means that thermoelectrically pumped LEDs would cool their surroundings when they are on, which feels rather counter-intuitive. This effect has been witnessed experimentally, and one group achieved an efficiency of 230 %. That is, their LED produced 2.3 times more light power than what it was supplied in electrical power, with the rest of the energy coming from surrounding heat.

This idea could very well be the next big thing in LED research, and the applications could span a much larger area than just lighting. For example, embedding them into ceilings could increase the efficiency of the whole building, since the LEDs would absorb heat energy from the hottest part of the room, turn that into light and radiate it down to the floor, where the energy gets absorbed again. Such a device would transfer heat from the ceiling to the floor, thus reducing the amount of required heating. But there are some teeny tiny problems.

First of all, while the 230 % efficiency LED required a mere 30 picowatts of electrical input power, it radiated only about 70 picowatts of light. To put that into perspective, you need about 10 watts  of output light power to have a relatively bright white LED bulb, which is 142 billion times more power than 70 picowatts! For now the thermoelectric pumping works best at very low input powers, because the electrons that get their energy from heat also tend to lose it as heat. Another major problem is that ambient room temperature is nowhere near hot enough to achieve these efficiency figures, unless you've set it at a smoldering 135 degrees Celsius. To make matters worse, the thermoelectric LEDs emit at infrared wavelength, so we can't even see the light they produce!

In spite of these, not so insignificant problems, I see great promise in this technology. Once someone invents a more efficient method and/or a better material for thermoelectric pumping, we may get to see yet another revolution in lighting. It may take another century of research, but some of the technologies we use today had a worse start than this.