Finding the next innovative polymer is always a tough task, but researchers are now using artificial intelligence (AI) to revolutionize this area.
Some well-known polymers like Nylon, Teflon, and Kevlar have significantly impacted our world. From Teflon-coated cookware to advancements in 3D printing, polymers are essential for improving various systems that enhance our daily lives.
Despite the challenges in identifying the next revolutionary polymer, Georgia Tech researchers are harnessing artificial intelligence (AI) to influence the future of this field. The team led by Rampi Ramprasad is developing and modifying AI algorithms to speed up the discovery of new materials.
This summer, two papers released in the Nature journal series highlight remarkable achievements stemming from years of AI-based research in polymer informatics. The first paper, published in Nature Reviews Materials, features recent advancements in designing polymers for key applications like energy storage, filtration technologies, and recyclable plastics. The second paper in Nature Communications explores how AI algorithms can uncover a new category of polymers focused on electrostatic energy storage, with successful lab synthesis and testing of the designed materials.
“Initially, the application of AI in materials science was primarily driven by curiosity, especially after the White House initiated the Materials Genome Initiative over ten years ago,” said Ramprasad, a professor at the School of Materials Science and Engineering. “Only recently have we started observing genuine success stories in AI-driven polymer discovery. These accomplishments are driving significant changes in the industrial materials research and development landscape, making this review particularly significant and timely.”
AI Opportunities
Ramprasad’s team has created innovative algorithms that can quickly predict the properties and formulations of polymers even before they are physically produced. The process starts by establishing specific property or performance criteria required for applications. Machine learning (ML) models are then trained on existing material-property data to forecast these desired outcomes. The team can also generate novel polymers, predicting their properties through ML models. The leading candidates that satisfy the property criteria are selected for real-world validation through lab synthesis and testing. Results from these experiments are integrated with the initial data to continually enhance the predictive models in an ongoing process.
Although AI can expedite the discovery of new polymers, it introduces distinct challenges. The reliability of AI predictions relies on having rich, diverse, and extensive initial data sets, making high-quality data crucial. Additionally, developing algorithms that can generate polymers that are both chemically realistic and synthesizable is a complex endeavor.
One major challenge arises after algorithms predict the properties: confirming that the designed materials can indeed be synthesized in the lab, function as expected, and demonstrate scalability for real-world applications. Ramprasad’s group specializes in designing these materials, while collaborators from various institutions, including Georgia Tech, handle their fabrication, processing, and testing. Professor Ryan Lively, from the School of Chemical and Biomolecular Engineering, collaborates regularly with Ramprasad’s team and co-authored the paper in Nature Reviews Materials.
“In our regular research, we extensively utilize the machine learning models developed by Rampi’s team,” Lively noted. “These tools expedite our work and allow us to quickly explore new concepts, showcasing the promise of ML and AI. We can make model-based decisions before investing time and resources into laboratory exploration.”
With the help of AI, Ramprasad’s team and their collaborators have achieved significant progress in areas like energy storage, filtration technologies, additive manufacturing, and recyclable materials.
Polymer Progress
A notable success mentioned in the Nature Communications paper involves developing new polymers for capacitors, crucial devices for storing electrostatic energy. These capacitors are important in applications such as electric and hybrid vehicles. Ramprasad’s group collaborated with researchers from the University of Connecticut on this project.
Traditional capacitor polymers typically offer either high energy density or thermal stability, but not both. Using AI tools, the researchers discovered that insulating materials made from polynorbornene and polyimide polymers can achieve both high energy density and high thermal stability. These polymers can also be further optimized to withstand demanding environments, such as those found in aerospace applications, while promoting environmental sustainability.
“The new polymer class with both high energy density and thermal stability exemplifies how AI can steer materials discovery,” Ramprasad stated. “This achievement is also the result of years of interdisciplinary collaboration with Greg Sotzing and Yang Cao from the University of Connecticut, along with ongoing support from the Office of Naval Research.”
Industry Potential
The opportunity for practical applications of AI-assisted materials development is further highlighted by industry contributions acknowledged in the Nature Reviews Materials article. The co-authors of this publication also include scientists from Toyota Research Institute and General Electric. To further promote AI-driven materials development in the industry, Ramprasad co-founded Matmerize Inc., a startup spun out from Georgia Tech. Their cloud-based polymer informatics software is already in use by companies across diverse sectors, including energy, electronics, consumer goods, chemical processing, and sustainable materials.
“Matmerize has turned our research into a solid, flexible, and industry-ready solution, allowing users to design materials virtually with enhanced efficiency and reduced costs,” Ramprasad explained. “What started as a research curiosity has gained substantial momentum, ushering us into an exciting new era of designing materials.”
