The Trials and Tribulations of Thermoelectric Energy Harvesting
What I learned from my experiments with thermoelectric energy harvesters at Accenture Labs.
Did you know you can make ice in the desert? It’s surprisingly simple: make a shallow pool of water right after the sun sets and wait until dawn. The water will cool by a process called night-sky cooling, lowering below ambient temperature and freezing. Just make sure you store the ice in a dark room with thick walls before the sun comes up or you’ll be back where you started. This phenomenon works because of the way materials absorb and emit heat in the form of infrared radiation. The atmosphere absorbs some wavelengths more than others, as a result there is a small window of wavelengths where it absorbs almost nothing This is known as the sky’s radiation window and is what allows night sky cooling to work.
The issue is in the name, “night-sky” cooling. The heat of the sun during the day more than makes up for this effect naturally. But with some clever materials engineering, you can make materials that get below the temperature of the air even in the daytime. Researchers have been able to make coatings that reflect nearly all the sun’s heat and emit their own heat almost exclusively in this sky window, which allows them to hit temperatures below ambient air on a hot summer day.
I see great potential in this technology for improving thermal energy management of buildings in hot climates, but I recently wanted to explore the possibility of using it for energy harvesting. I got the opportunity with my work at Accenture Labs as an Emerging Energy Technologies Researcher. I proposed a project where, using thermoelectric materials, a temperature difference between a daytime radiative cooling surface and a solar absorbing surface could be used to drive an electric current. This would provide a constant 24h low-power source for different applications. I called it the passive thermoelectric generator. While the project was ultimately unsuccessful, I learned a lot of key points I think are worth sharing for those looking to explore this technology.
My Passive Thermoelectric Generator
Let’s start with what I made and how it did. I made a radiative cooling paint by mixing calcium carbonate powder and a clear acrylic medium in a 70–30 ratio by volume, in favor of the calcium carbonate. For the solar absorbing surface, I bought a spray can of black high-temperature paint. I ordered a pack of Peltier plates from Amazon for the thermoelectric generators. To make sure both sides of the Peltier plate could be affected by the sun, I also bought some aluminum business card blanks. These were stuck to the Peltier plate’s surfaces facing opposite ways using a thermal paste. After coating my surfaces, I pushed the generator into a bed of insulating foam to reduce the effects of heat transfer from conduction or convection and took it to my roof in Washington, DC for testing during the day and at night.
The results were pretty underwhelming. During the day, with an ambient temperature of 3°C, my solar collector managed to hit 8°C in the sun. The cold side stayed at 3°C, making the temperature difference a neat 5°C. That got me 40mV and 0.04mA, which combines for a full 1.6mW of power out. At night, things got worse. The hot and cold plates were the same temperature; no voltage, no current, no power.
This was not the result I was hoping for, but after consideration, I think there are three general themes that explain why this didn’t work.
1) Material Quality is Important
Just because something has simple components and says it can be made cheaply, it is not necessarily easy to make. A paper from Purdue reported a method of making a daytime radiative cooling paint from calcium carbonate particles and an acrylic filler for about the same cost as commercial white paints. My approach of “these materials are easy to find, I can totally make this” is like looking at an instant ramen packet and thinking the same thing. Ramen is difficult to make from scratch, and so too is this radiative cooling paint. The availability of the materials does not consider the precise morphology and processing required to achieve the reported results. Replicating this approach with off-the-shelf materials proved to be a challenging task and getting these specific materials or processing methods wasn’t within the project’s budget and time constraints.
This need for quality materials extended to the other components as well. A black paint solar collector works well during the day but becomes a black-body radiator at night, shedding all of its heat very efficiently. This cooled the “hot” surface down to the temperature of the “cold” surface leading to zero temperature difference. Coatings have been specifically designed to absorb as much heat as possible and emit as little as possible — inverting the principles behind radiative cooling — but these can be difficult to obtain in the small quantities required for this project. High quality thermoelectric materials are also hard to come by, usually generated in a lab and not within a cheap pack of Peltier plates.
2) Lack of Power for Thermoelectric to Harvest
I am not the first person to overestimate the power of thermoelectric materials for solar harvesting, even those with access to all of the quality materials will struggle to get meaningful power out of these systems. To understand why, let’s look at the thermodynamics behind a passive thermoelectric generator (warning: some equations ahead). Looking at the Carnot efficiency equation tells us the maximum efficiency of any system getting work from the heat flow between two thermal reservoirs.
For my 5°C difference on a cold January afternoon (the equation uses Kelvin but the temperature difference is the same in both units) that efficiency is slightly under 2%. Even if the temperature difference was 105°C, the efficiency is only 27%. That’s the best theoretical efficiency we can get from any system. For a thermoelectric, we calculate the efficiency based on the figure of merit, a term dependent on temperature that combines.
The thermoelectric material in my device was Bismuth Telluride, Bi2Te3, which has figure-of-merit of 0.577. For the maximum temperature range of 5°C I observed, this gives me an efficiency of only 0.37%. I can increase the efficiency to 1% by increasing the figure of merit by a factor of eight, or by making the temperature difference to 15°C.
The other issue is that there simply isn’t a lot of power to harvest from a temperature gradient of 5°C. As efficiency is the ratio of power out to power in, I can use my 1.6mW of power out and the 0.37% efficiency to calculate that my device received roughly 430mW of power from the environment. Controlling for the size of the device that’s around 27mW/cm², roughly a quarter of the power of sunlight on any given square centimeter of earth.
3) Problems are Better than Solutions
Fully understanding a problem is the best way to apply an effective solution. It is also important to identify internal biases when ideating solutions. These biases can be social (who is affected by the solution and how), intellectual (incomplete or faulty assumptions about the problem space), technical (relative understanding across different solutions), practical (relative understanding of how to implement any given solution), or simply opinion disguised as rationalized fact. Improper identification of these biases results in an at best a suboptimal solution, and at worst the exacerbation of the original problem or creation of a new one.
It is also important to differentiate between different problems, even if they appear similar at first. Radiative cooling and solar collectors still carry immense potential for low-cost, passive thermal management, a problem at its core related to the conversion of energy into a useful form. Direct energy harvesting, while a very similar core challenge, cannot easily be solved using the same solutions. It’s like trying to build a chair from wood and from sheet metal with the same set of tools.
This is where the passive thermoelectric generator project ran into trouble. My incomplete technical understanding of the proposed solution — thermoelectric materials — meant that I didn’t realize how little power they generated at low temperature differences, even with high efficiencies, until midway through the project. When planning my device, I did not account for this low power outputs of thermoelectric devices. Even theoretical designs of new materials can only reach a few milliwatts of power per square centimeter at the temperature differences I was expecting. That’s around 2–3% of the solar irradiance hitting the material. A commercial solar panel can convert ~20% of the electricity and can do it at much lower overall cost. This doesn’t take into consideration the practical aspects of building the passive thermoelectric generator, like having both sides of the thermoelectric material facing the sky. I was so focused on the theoretical elegance of the solution I didn’t think about solving the practical considerations of building it.
Technology Outlooks
While there’s a lot to be done before the combination of thermoelectric and selective heating and cooling materials reaches viability, the two technologies on their own have a lot to offer.
Passive radiative cooling has excellent applications for building and infrastructure thermal management. We can reduce energy demands for cooling by incorporating this technology into refrigeration systems, which will only become more energy intensive as the world warms. Similarly, just coating critical infrastructure would reduce the stress heating and cooling cycles place on concrete. These materials are all white at the moment which limits the color palette (though who doesn’t love the aesthetic of an all-white bridge?) there is promising work showing you can add color without losing the daytime radiative cooling effect. Zooming out, we tend to refer to materials that can shape waves due to the material’s structure instead of it’s properties as metamaterials. These kinds of materials are becoming even more important in radio and microwave communications systems like 5G, and can be adapted to infrared, visible, or even sound waves for a whole host of applications.
Thermoelectric materials have promise as energy harvesting tools, though their low power output restricts them to niche applications. They are best suited for use in places with much higher thermal gradients like exhaust pipes, engine blocks, or steam vents. It’s only under these high temperature differences that thermoelectric can hit really high efficiencies. However, sometimes you don’t need a lot of power, you just need it to be constant. In this case, a thermoelectric powered by human body heat could keep a health monitoring device running constantly. Combining a thermoelectric with a heat exchanger such as in an air conditioning unit is another option to recover energy for useful monitoring purposes, such as air quality.
Final Thoughts
I believe the only failure is failure to learn and I’m excited to take the lessons from this project into my future work. What I learned while doing this was that it’s important to look at the problem first, and make sure your personal preferences don’t influence the kind of solution you propose. As I continue to develop my understanding of energy harvesting technologies and their prospective application spaces, I’m glad to be doing so at Accenture Labs. If you’d like to know more about our work in emerging energy technologies, please contact me and my colleagues at Accenture Labs.
You can reach me at daniel.gregory@accenture.com or visit the Accenture Labs website.
