Materials created by AI can reduce your cooling costs

Scientists have designed more than 1,500 materials using artificial intelligence that can selectively emit heat at different levels, making buildings cooler and cutting energy bills.

Machine learning methods detailed in nature represent a significant advancement in creating custom thermal materials that can be applied to roofs, walls, and even clothing like paint.

Researchers at the University of Texas at Austin and international partners have developed thermal emitters that have performed excellent cooling performance in real-world tests. When applied to a model building, a material keeps the roof temperature at 5-20 degrees Celsius cooler than traditional paints after four hours of sunshine, saving 15,800 kWh per year in hot climates such as Rio de Janeiro.

Beyond repeated trials

The traditional development of these specialized thermal materials is very slow, depending on trial and error methods that often produce suboptimal effects. The new AI framework will change the optimal design by automatically exploring the possible combination of millions of three-dimensional structures and materials.

Yuebing Zheng, a professor in the Department of Mechanical Engineering at UT Austin, co-led the study, highlighted this point: “Our machine learning framework represents a significant leap in thermal element emitter design. By automating processes and expanding the design space, we can create exceptional performance with exceptional performance that was previously unimaginable.”

The researchers created a comprehensive library of 32 basic three-dimensional structures inspired by nature – transparent, cylinders, ridges and triangular prisms – and 30 different materials. Their AI systems can combine these elements in countless ways, resulting in tens of thousands of unique heat emitter designs.

Seven categories of smart materials

The team developed seven different categories of thermal element transmitters, each optimized for a specific application:

• Broadband transmitter: High thermal emission at all infrared wavelengths
• Band Selective Launcher: Targeted emissions in specific atmospheric windows for maximum land cooling efficiency
• Dual-band transmitter: Launch in two separate atmospheric windows to enhance cooling under various conditions
• Hot camouflage material: Designed to hide hot signatures for military applications

What makes these materials particularly effective is that they are able to reflect sunlight while emitting heat at precise infrared wavelengths. This dual action (stands cooling in the sun while actively radiating heat into the space) produces a powerful cooling effect without electricity.

Real-world performance

The researchers fabricated and tested four representative materials to verify their AI predictions. The results are surprising: the measured performance closely matches the computer predictions, confirming the accuracy of their machine learning approach.

In the optimal atmospheric window, a lineage material with 96% solar reflectivity and 92% emissivity achieves near perfect performance. Another broadband emitter maintains a solar reflectivity of more than 96%, while discharging 92% of the absorbed heat throughout the infrared spectrum.

In outdoor cooling tests, these materials keep temperatures below ambient air even during the midday peak of the sun. Under transparent sky conditions, a material reaches a temperature drop of 5.9°C at noon. In an urban environment, surrounding buildings emit heat, band-selective materials outperform broadband transmitters and commercial white paint.

From the laboratory to the roof

Perhaps most importantly, these materials can be made using a simple room temperature process and applied like traditional paint. One formulation involves mixing powdery ingredients into solutions that can be brushed, sprayed or spin-coated on various surfaces, including metal, glass, plastic and bricks.

The energy saving potential is huge. Computer simulations of a four-story apartment building in a tropical climate show that annual energy savings of 75 megabytes per square meter are equal to about 15,800 kWh of the entire building roof. In the context, a typical air conditioning unit consumes about 1,500 kWh per year.

Co-author Kan Yao notes the broader applicability of the technology: “Machine learning may not be the solution to all problems, but the unique spectral requirements of thermal management make it particularly suitable for designing high-performance thermal emitters.”

Beyond the Buildings

The scope of application is far beyond the scope of building cooling. These materials can be integrated into textiles for better cooling of clothing, applied to vehicle surfaces to reduce heat buildup or used in spacecraft thermal management systems. Urban planners can deploy them to combat the heat island effect of cities that are hotter than the surrounding areas.

The AI framework itself represents a significant advance in material discovery. Traditional optimization methods may generate dozens of design candidates over months of work. The system generates 2500 candidates per second, greatly speeding up the discovery process.

The study also reveals the basic physics of managing heat emission, where AI automatically recognizes clusters of materials and structural best suited to different wavelength ranges. This knowledge can even guide future material development beyond thermal applications.

As it becomes more common due to climate change, materials that can provide passive cooling without power consumption become increasingly valuable. These AI-designed thermal emitters provide a glimpse into how machine learning can help solve real-world problems by discovering solutions that individual intuitions may never find.

Related