Using machine learning to create high-performance thermoelectric devices

Researchers from the University of Notre Dame have developed a machine-learning-assisted superfast way to create high-performance, energy-saving thermoelectric devices. The novel process uses intense pulsed light to sinter thermoelectric material in less than a second (conventional sintering in thermal ovens can take hours). The researchers, led by Yanliang Zhang, associate professor of aerospace and mechanical engineering, accelerated this method of turning nanoparticle inks into flexible devices by using machine learning to determine the optimum conditions for the ultrafast but complex sintering process. The research findings were published in Energy and Environmental Science.

Flexible thermoelectric devices offer many opportunities for direct conversion of waste heat into electricity, as well as solid-state refrigeration. They also have benefits as power sources and cooling devices, as they don’t emit greenhouse gases and are durable and quiet, since they don’t have moving parts. Despite their potential broad impact in energy and environmental sustainability, thermoelectric devices have not achieved large-scale application because of the lack of a method for fast and cost-effective automated manufacturing. Machine-learning-assisted ultrafast flash sintering now enables the production of high-performance, eco-friendly devices, faster and at a lower cost.

“The results can be applied to powering everything from wearable personal devices, to sensors and electronics, to industry Internet of Things. The successful integration of photonic flash processing and machine learning can be generalised to highly scalable and low-cost manufacturing of a broad range of energy and electronic materials,” Zhang said.

Image credit: University of Notre Dame