July 30, 2026
Using machine learning and physics-based modeling, ME researchers are discovering new materials for flexible electronics and wearables.
Through tests that included thermal imaging (shown here), the researchers found that a new material they discovered through machine learning and physics-based modeling could serve as a high-performing material in flexible electronics. Photo credit: Lijun Zhou/iMatter Lab.
Soft materials that combine mechanical flexibility with high thermal or electrical conductivity are essential for wearables, stretchable electronics and soft robotic systems. With both solid and liquid phase inclusions, hybrid filler composites provide a versatile platform for engineering these multifunctional properties.
To identify new composite materials with those properties, researchers typically create and test many different combinations, a process that can be time-consuming, expensive and lead to waste. Researchers in ME have developed a new inverse design framework that reverses the standard design process to speed up the discovery of multifunctional materials.
The framework starts with the desired material properties for a specific application — such as wearable electronics — and works backward to determine the best material composition using physics-based modeling and machine learning.
“Our framework guides experiments and helps us identify optimal material compositions with far fewer experiments,” says ME Assistant Professor Mohammad Malakooti. “This approach can significantly reduce development time, cost and material waste while enabling faster innovation in flexible electronics, wearable devices and many other advanced material systems.”
The researchers experimentally validated only the most promising candidates, including a material identified by the framework that consists of liquid metal and aluminum oxide in a silicone rubber matrix.
They created and integrated the material as part of a flexible thermoelectric generator. Experiments showed that the material achieved about 60% higher thermal conductivity while reducing material cost by about 10%, compared to materials that were previously used. Through tests that involved stretching the material and thermal imaging, the researchers found that the material could serve as a high-performing material in flexible electronics.
“While the results aligned well with our expectations, it was especially rewarding to see the multifunctional composite designed using inverse design outperform existing liquid metal composites for thermal management while maintaining softness and stretchability at lower cost,” Malakooti says. “These findings demonstrate how AI can search a large design space and identify material compositions that might otherwise be overlooked using traditional R&D approaches.”
The researchers received funding from the National Science Foundation and the AI Institute in Dynamic Systems. ME Ph.D. student Lijun Zhou led the research, and the project was overseen by ME Assistant Professors Malakooti and Krithika Manohar.
Through this project, Zhou gained experience in physics-based modeling, finite element analysis, machine learning, optimization, composite synthesis and wearable device fabrication.
“The next step is to expand the framework to more complex composite systems with additional design parameters and more sophisticated models,” Zhou says. “Ultimately, we hope to use machine learning to navigate the vast design space of composites, develop materials with improved performance and identify applications where their unique combinations of properties can have the greatest impact.”