University of Oklahoma researcher awarded funding to pursue AI-powered material design

(Press-News.org) NORMAN, Okla. – Mike Banad, a researcher with the University of Oklahoma, has been awarded funding from the U.S. Department of Defense (DoD) to pursue the development of advanced materials that could shape the future of energy-efficient electronics and photonics. His project uses inverse design techniques that aim to accelerate their creation and develop a systematic framework for future materials engineering that targets specific, desired properties.

While the framework can be used to rapidly design a wide range of materials, Banad’s current work centers on metal-insulator transition (MIT) chalcogenides. These materials can quickly switch between a metallic, electricity-conducting state and an insulating, electricity-blocking state, typically without altering their crystal structure. This makes them valuable for a range of applications – including computing, sensing and defense systems – but they are difficult to successfully produce without extensive and costly trial and error.

Banad, who is an associate professor in the Gallogly College of Engineering’s School of Electrical and Computer Engineering, is approaching this challenge by embedding artificial intelligence into every stage of the discovery process. His team is training AI models on existing MIT materials and using AI to find advanced materials with the optimized qualities they’re looking for, such as the ability to reliably switch states while using low amounts of energy. After finding the right matches, the team evaluates the materials’ structural stability, switching performance and ability to fulfill their targeted applications.

“We simulate the atoms to check if the structures withstand realistic operating conditions, such as high temperatures, repeated switching cycles or exposure to harsh environments relevant to defense applications,” Banad said. “Only the few candidates that pass these tests move on, which saves a lot of time and money throughout the discovery process.”

The researchers then use trained AI models to determine which procedures, or “recipes,” are most likely to succeed at making the MIT materials. From there, the team fabricates the optimized materials using the right chemicals, temperatures and other conditions recommended by the AI-guided workflow.

Banad refers to the project’s process as property-driven material discovery, design and deployment – or PMD³. And as the co-leader of OU’s INQUIRE Lab, he said the research’s pursuit of specialized MIT materials could push the lab’s efforts forward. “Our interest is in artificial synapses and neurons, which are the hardware support for next-generation energy-efficient AI applications,” he said. “Those will be the next step in this technology.”

His project team is among more than two dozen groups – four of which are led by OU researchers – to earn funding through the DoD’s Defense Established Program to Stimulate Competitive Research (DEPSCoR) 2024 Research Collaboration competition. DEPSCoR aims to strengthen higher education research infrastructure in underutilized U.S. states and territories. Through the competition, projects are awarded in areas associated with DoD initiatives.

If the PMD³ framework is successful, Banad said it could produce MIT chalcogenides that more reliably switch between their “on” and “off” conductivity states, require less power, and withstand tougher environments.

“That’s important for neuromorphic computing, where we want high-speed, low-energy switching that can operate reliably in demanding conditions,” Banad said. “If we can create materials that meet those requirements, we can move closer to hardware that behaves more like the human brain while remaining robust enough for military systems that must be small, lightweight and very reliable.”

Beyond neuromorphic systems, the PMD³ framework could enable broader advances in electronics, sensing and secure communications by offering a predictive, efficient pathway for discovering materials with tailored properties. “Our approach reduces uncertainty at every stage and creates a scalable and repeatable material-design methodology,” Banad said.

He emphasized that while the framework is currently being demonstrated with MIT chalcogenides, it can be applied to other material classes. “The combination of physics-based simulation and AI-driven optimization opens up the possibility of designing materials that previously felt out of reach.”

Banad said that he aims to continue this research through future grants, noting that with ongoing training, the AI will produce better results for developing materials. “This award allows us to accelerate a vision where material discovery is faster, more precise and better aligned with future national needs.”

END