Cool houses, cool clothes: Researchers develop meta-emitters with AI

A team of scientists from the University of Texas, Shanghai Jiao Jong University, the National University of Singapore and Umea University have used artificial intelligence (AI) to develop complex three-dimensional thermal meta-emitters that can be applied to houses or clothing, for example, to provide passive cooling.

The researchers used a machine learning approach to develop around 1,500 different materials that can emit heat in different ways at different levels. The scientists describe the method in the study “Ultrabroadband and band-selective thermal meta-emitters by machine learning”, which has been published in Nature. In principle, these materials are suitable for passively cooling objects coated with them by reflecting the heat of the sun. This can lead to considerable energy savings, as houses coated with meta-emitters no longer have to be cooled with air conditioning systems, which is so energy-intensive.

Using conventional methods, the development of meta-emitters was a lengthy process based on trial-and-error methods. Even automated methods did not lead to success, as they cannot sufficiently derive the complex 3D structures of meta-emitters. They can only generate simple shapes and thin layers or flat patterns. This has often led to less good designs that were not effective enough.

The AI-based method that has now been developed, on the other hand, has led to an expansion of the design space, so that “materials with outstanding performance” can be produced that were “previously unimaginable”.

Energy savings through meta-emitters

From the 1500 materials proposed by the AI, the scientists selected the four most promising to produce and test in practice. They coated a model house with them. The use of meta-transmitters is similar to the application of paint. The roof of the model house was exposed to four hours of direct midday sun and was then between 5 and 20 °C cooler than roofs without this coating.

The scientists calculate that a typical residential building under hot climatic conditions can save around 15,800 kW of electricity per year for active cooling with an air conditioning system and thus reduce costs.

The researchers have divided the meta-emitters into seven classes so that they can be better assigned to different areas of application according to their specific strengths and weaknesses.

Some of these meta-emitters can be used, for example, to reduce temperatures in large cities in the summer heat by reflecting sunlight from buildings and surfaces and emitting heat in certain wavelengths. It is also possible to use them to coat clothing to keep the wearer cooler. The same applies to cars and spacecraft, for example, which are exposed to the sun and need to be kept cool.

The scientists now want to further optimize the AI for determining meta-emitters to be able to create even more effective meta-emitters.

(olb)