Theory Meets Material Science
A great presentation on thermoelectrics
Today I wish to talk about a beautiful presentation he just gave at TTI/Vanguard in San Francisco.
Thermoelectrics is an effect that seems to have nothing to do with our usual topics. But Boukai uses a mathematical trick to make a “new” type of material. This material has to have quite special properties, and he is able to make it by using ideas that we are familiar with in theory. This is a great example, I believe, of theory interacting with technology.
Boukai presented his work at TTI/Vanguard, which is a conference I have talked about before—see here. It is oriented toward the future of technology of all kinds, with a special emphasis on electronic and computer technology. The talks often highlight new technologies, many of which are being developed by startups. This is the case with Boukai, who is co-founder of the company Silicium Energy. They are attempting to build components that will radically change how we power small devices. This is especially relevant to IoT—that is, the “Internet of Things.” Think watches, for example, that never need to be recharged.
In order to understand the math problem we need at least a high level understanding of the Seebeck effect, named after Thomas Seebeck. He discovered in 1821 that a compass needle is deflected, if it is connected to a loop that contains two metals, provided there is a temperature difference between the metals. Wikipedia’s diagram illustrates the underlying phenomenon:
The compass needle moves because the electrons in the metals act differently owing to their temperature difference, and thereby create an electrical current. This current then induces a magnetic field that moves the needle. Seebeck named this phenomenon the thermomagnetic effect, which is really wrong. The primary effect is the creation of an electrical flow—this was renamed to “thermoelectricity” by Hans Ørsted. Wrong or not, it is still called the Seebeck effect—he may have guessed how it worked incorrectly, but he discovered the effect.
A Special Material Is Needed
Thus, the goal is to try and extract energy from a small heat difference. For example, Silicium Energy plans to use this method to build watches that need no recharging. The watches would exploit that while on your wrist there is a natural source of a heat difference: we are warm and the air around us is usually cooler. So by the Seebeck effect there will be an electrical current. The amount of energy created is tiny, but it will be large enough to power the processor in a modern digital watch.
This sounds doable. Yet it is tricky. The problem is getting a material that is a great conductor of electrons, but a poor conductor of heat. The insight that Boukai’s company is based on is that this can be made out of silicon. The advantage of using silicon and not some exotic materials, which have been used before, is cost. Silicon devices can be made using standard technology, for pennies per device, while exotic materials can be very expensive.
Being able to turn silicon into a thermoelectric material and do it at low cost is quite a feat. Silicon has good electrical properties, but also is a pretty good conductor of heat. The trick is to find a way to lower the thermal conductivity of silicon in order to increase its thermoelectrical efficiency. Lowering the thermal conductivity makes it easier to keep the cold side of the device cold to create that temperature difference needed by the Seebeck effect.
Randomness to the Rescue
Boukai and his co-workers’ clever idea—finally—is to fabricate a piece of silicon that uses its structure to make a material that conducts electrons well and conducts heat poorly. Here is how he does this: Imagine a square of silicon, with the top side hot and the bottom cold. Initially—by the Seebeck effect—electrons will move from the top to the bottom and create a current. This is wonderful. However, the problem is that heat will quickly also flow from the hot top to the cold bottom and will make the Seebeck effect stop.
The trick is to make random defects, essentially holes, in the silicon. The point is this:
- Electrons will still move easily from top to bottom. The holes will not affect them.
- On the other hand, heat will make a kind of random walk bouncing off the holes, as it transports from the top to the bottom. This means that the thermal conductivity of the modified silicon is much lower. And the Seebeck effect continues.
We thank him for sending the following picture:
Note: in physics, a phonon is a collective arrangement of atoms or molecules in a solid. They play a key role in the transport of heat. Boukai’s trick depends on the size of phonons, which are much larger than electrons. This explains why electrons are pictured as scooters and phonons as trucks.
I know this is a rough explanation, but I believe it is a reasonable description of what happens. And the fact that heat flows as a random-walk type process yields in practice a fold decrease in the silicon’s thermal conductivity. This keeps the watch running. See his joint paper for more technical details.
In What Ways Algorithmic?
The ideas of random behavior and statistical mechanics have been around in physics for a long time. Karl Pearson coined the term “random walk” in 1905, the same year as Albert Einstein’s famous paper on Brownian motion. Ising models partly motivated the concept of , and Markov chains were long studied in physics before becoming a staple of computer theory. So there is no chicken-egg question about which methods came first where.
What strikes Ken and me as distinctively algorithmic, however, is the way the silicon materials are being programmed to have a physical property directly. This is different and feels more qualitative than programming logic gates on silicon. Of course there are other cases of mathematical structure and algorithmic behaviors being used to create new materials—witness the recent Nobel Prizes for work on graphene and quasicrystals.
I really liked the trick used here. Is there some other application where we could imagine using it to make some other new material, or even to use the trick abstractly in some algorithm?