A thermoelectric generator with a shape that no human designer would likely have imagined has now been created by a computer—and it performs more than eight times better than conventional designs. Rather than relying on intuition or repeated trial and error, the breakthrough was achieved through advanced computational optimization.
A joint research team led by Professor Jae Sung Son of the Department of Chemical Engineering at POSTECH (Pohang University of Science and Technology), in collaboration with Professor Hayoung Chung of the Department of Mechanical Engineering at UNIST (Ulsan National Institute of Science and Technology), has developed a general design framework that enables computers to autonomously identify the optimal structure of thermoelectric generators, which convert waste heat into electricity. Their work was recently published online in Nature Communications.
Vast amounts of energy are continuously lost as waste heat—from automobile exhaust systems and industrial processes at steel mills and semiconductor plants to even the warmth emitted from the human body. Thermoelectric power generation has long been regarded as a promising way to recover this wasted energy, as it can produce electricity from nothing more than a temperature difference, without requiring any additional fuel. It is the same principle used by NASA to power deep-space probes.
Despite steady progress in improving thermoelectric materials, device performance in real-world operating environments has often fallen short of expectations. The reason is that efficiency depends not only on the material itself but also on the device structure. A wide range of factors—including the path of heat flow, the distribution of electrical resistance, contact losses, and load conditions—must work together in a highly coordinated way for the device to perform at its full potential. Until now, most thermoelectric generator designs have been developed largely through human intuition and repeated experimental testing.
To overcome this limitation, the research team turned to topology optimization, a computational design method that allows the computer to determine the most efficient three-dimensional geometry. Instead of starting from a preconceived shape, the computer evaluates the design conditions and generates structures that maximize efficiency while taking into account realistic operating parameters such as the thermal environment, material properties, contact resistance, and electrical load.
The resulting designs were far from conventional. Traditional thermoelectric generators are typically built in simple rectangular shapes because they are familiar and easy to fabricate. The computer, however, produced highly unconventional geometries, including I-shaped and asymmetric hourglass-shaped structures—forms that would be difficult to conceive through intuition alone. These designs were found to enhance overall system efficiency by precisely controlling heat flow, increasing the temperature difference across the device, and simultaneously minimizing electrical resistance and contact-related losses.
The team then fabricated the optimized structures using 3D-printing technology and experimentally evaluated their performance. The best-performing design achieved up to 8.2 times higher power-generation efficiency than a conventional rectangular generator. The experimental results also showed strong agreement with the computational predictions, confirming the validity of the framework.
This work points to a future in which wasted heat can be more effectively converted into useful electricity. "This study is significant in that it moves beyond the conventional focus on discovering better materials and introduces a new pathway for improving performance through design-driven optimization tailored to real thermal environments," said Professor Jae Sung Son. Professor Hayoung Chung added, "This technology can derive optimal structures directly from input conditions without human trial and error, and its range of applications and impact could expand further through integration with AI."
This research was supported by the Mid-Career Researcher Program and the Nano & Material Technology Development Program through the National Research Foundation of Korea (NRF), funded by the Ministry of Science and ICT.
