Materials created by AI could cool cities and spacecraft

New materials developed using machine learning and artificial intelligence can, among other things, keep your home cool and reduce your energy bill.

Researchers from the University of Texas at Austin, Shanghai Ziaoton University, Singapore National University, and UMEA University in Sweden have developed a new machine learning-based approach to creating complex, three-dimensional thermal metamitters. This study has been published in the journal Nature.

Using this system, researchers have developed over 1,500 different materials and can selectively release heat at different levels and different manners, making them ideal for energy efficiency through more accurate cooling and heating.

“Our machine learning framework represents an important leap in the design of thermal meta-emitters,” said Yuebbing Zen, professor at Walker School of Mechanical Engineering at the Cockrell School of Engineering and co-leader of the research.

“By automating the process and expanding the design space, we can create materials with superior performance that we previously could not have imagined.”

To test their platform, researchers produced four materials for design verification. They also applied one of the materials to the model house and compared them to commercial paints with cooling effects.

After 4 hours of daytime exposure to direct sunlight, the roofs of the meta-emitter-coated buildings were on average cooler than 5-20 degrees Celsius Celsius, respectively, than those with white and grey paint.

Researchers estimated that this level of cooling could amount to 15,800 kilowatts per year in hot climate apartments like Rio de Janeiro and Bangkok. A typical air conditioning unit uses approximately 1,500 kilowatts per year.

However, applications go beyond improving energy efficiency in your home and office. Using a machine learning framework, researchers developed seven classes of meta emissions, each with different strengths and applications.

Thermal metame emitters can be deployed to help reduce urban temperatures by reflecting sunlight and emitting heat at certain wavelengths. This reduces the effectiveness of the city’s heat islands. This effect is that large cities have higher temperatures than surrounding areas due to lack of vegetation.

Additionally, thermal meta-emitters can be useful in space to manage the temperature of a spacecraft by reflecting solar radiation and efficiently releasing heat.

Beyond the application of this research, thermal metame emitters can become part of many things that we use every day. Integrating them into textiles and fabrics could improve cooling techniques for clothing and outdoor equipment. Wrapping the car and embed it in the material inside it can reduce the heat that accumulates when you sit in the sun.

The laborious and traditional processes of designing these materials have hindered them from mainstream adoption. Other automated options struggle to address the complexity of the three-dimensional hierarchical structure of meta-emitters, limiting the results to simple geometry such as thin film stacks and planar patterns, and performance can be short on some measurements.

“Traditionally, the design of these materials is slow, labor intensive and relies on trial and error methods,” Zheng said. “This approach often leads to suboptimal designs and limits our ability to create materials with the required properties.”

Researchers continue to improve this technology and apply it to more aspects of the field of nanophotonics. This is the interaction of light and matter on the smallest scale.

“Machine learning may not be the whole solution, but the unique spectral requirements of thermal management make it particularly suitable for designing high-performance thermal emitters,” said Kan Yao, co-author of the work and a researcher with Zheng’s group.