A University of Texas at Dallas physicist has teamed with Texas Instruments Inc. to demonstrate thermoelectric generators (TEGs) created using nanostructured silicon thermopiles fabricated on an industrial silicon complementary metal–oxide–semiconductor (CMOS) process line.

Described in a paper in the journal Nature Electronics, these TEGs exhibit a high specific power generation capacity (up to 29 μW cm⁻² K⁻²) near room temperature, which is competitive with typical (Bi,Sb)₂(Se,Te)₃-based TEGs.

The researchers said that the high power capacity results from the ability of CMOS processing to fabricate a very high areal density of thermocouples with low packing fraction and to control electrical and thermal impedances carefully. TEG power was also found to increase significantly when thermocouple width was decreased, providing a path to further improvements.

(Bi,Sb)₂(Se,Te)₃ TEGs, our silicon integrated circuit TEGs could be seamlessly integrated into large-scale silicon CMOS microelectronic circuits at very low marginal cost.

—Hu et al.

Unlike

The findings could greatly influence how circuits are cooled in electronics, as well as provide a method of powering the sensors used in the growing internet of things.

In a general sense, waste heat is everywhere: the heat your car engine generates, for example. That heat normally dissipates. If you have a steady temperature difference—even a small one—then you can harvest some heat into electricity to run your electronics.

—Professor Mark Lee, corresponding author

Sensors embedded beneath a traffic intersection provide an example of convenient thermoelectric power.

The heat from tires’ friction and from sunlight can be harvested because the material beneath the road is colder. So no one has to dig that up to change a battery.

—Professor Lee

The primary hurdles for widespread thermoelectric harvesting have been efficiency and cost, he said.

Thermoelectric generation has been expensive, both in terms of cost per device and cost per watt of energy generated. The best materials are fairly exotic—they’re either rare or toxic—and they aren’t easily made compatible with basic semiconductor technology.

—Prof. Lee

Silicon, upon which so much technology relies, is the second-most abundant element in Earth’s crust. It has been known since the 1950s to be a poor thermoelectric material in its bulk, crystalline form. But in 2008, new research indicated that silicon performed much better as a nanowire.

Efforts to make a useful silicon thermoelectric generator haven’t succeeded. One barrier is that the nanowire is too small to be compatible with chip-manufacturing processes. To overcome this, Lee and his team relied on “nanoblades”: 80 nanometers thick but more than eight times that in width. While that is still much thinner than a sheet of paper, it’s compatible with chip-manufacturing rules.

Study co-author Hal Edwards, a TI Fellow at Texas Instruments, designed and supervised fabrication of the prototype devices. He turned to Lee and UT Dallas to further study what the devices could do.

A deep dive for these novel measurements, detailed analysis and literature comparisons requires a university group. Professor Lee’s analysis identified key metrics in which our low-cost silicon technology competes favorably with more exotic compound semiconductors.

—Hal Edwards

Lee explained that the nanoblade shape loses some thermoelectric ability relative to the nanowire. However, he noted, that using many at once can generate about as much power as the best exotic materials, with the same area and temperature difference.

The team’s circuit-design solution combined an understanding of nanoscale physics with engineering principles. One key realization was that some previous attempts failed because too much material was used.

When you use too much silicon, the temperature differential that feeds the generation drops. Too much waste heat is used, and, as that hot-to-cold margin drops, you can’t generate as much thermoelectric power. There is a sweet spot that, with our nanoblades, we’re much closer to finding than anyone else. The change in the form of silicon studied changed the game.

—Prof. Lee

Lee said that the advanced silicon-processing technology at Texas Instruments allows for efficient, inexpensive manufacturing of a huge number of the devices.

You can live with a 40% reduction in thermoelectric ability relative to exotic materials because your cost per watt generated plummets. The marginal cost is a factor of 100 lower.

—Prof. Lee

Lee said the work was also novel because they used an automated industrial manufacturing line to fabricate the silicon integrated-circuit thermoelectric generators.

We want to integrate this technology with a microprocessor, with a sensor on the same chip, with an amplifier or radio, and so on. Our work was done in the context of that full set of rules that govern everything that goes into mass-producing chips. Over at Texas Instruments, that’s the difference between a technology they can use and one they can’t.

—Prof. Lee

Lee’s research is supported by the National Science Foundation through the Grant Opportunities for Academic Liaison with Industry (GOALI) program, and the work at UT Dallas was performed in the Texas Analog Center of Excellence (TxACE) laboratory.

Resources

• Gangyi Hu, Hal Edwards & Mark Lee (2019) “Silicon integrated circuit thermoelectric generators with a high specific power generation capacity” Nature Electronics volume 2, pages 300–306 doi: 10.1038/s41928-019-0271-9