In the fast-changing field of materials science, researchers at Georgia Tech are making big advances in AI-driven polymer research. They are using artificial intelligence (AI) to help. Prof. Rampi Ramprasad leads the team. They have created advanced computer algorithms, changing how scientists study the huge variety of chemicals in polymers.
The Challenges of Polymer Research
Polymers, large-molecule chemical compounds, have revolutionized numerous industries, from Teflon-coated frying pans to 3D printing. However, finding the next groundbreaking polymer is a daunting task due to the endless combinations of materials. Researchers have long relied on manual methods, such as fingerprinting, to understand the relationships between polymer structure, properties, and performance.
Introducing polyBERT: AI-driven polymer research Algorithm
To overcome these challenges, Ramprasad’s team created a new AI-driven polymer research ML model called polyBERT. They trained polyBERT using a huge dataset of 80 million polymer chemical structures. PolyBERT views the chemical structures and atom connections as a kind of chemical language. It uses advanced methods inspired by how computers understand human language, a field called natural language processing.
Key Features of polyBERT:
- Ultrafast fingerprinting: PolyBERT is much faster than traditional fingerprinting methods. It is over 100 times faster. This speed makes it perfect for high-throughput polymer informatics pipelines. High-throughput pipelines handle large amounts of data quickly.
- Multitask deep neural networks: PolyBERT can predict many properties of polymers at the same time. It uses hidden connections in the data to make better predictions than models that predict only one property.
- Vast dataset: There is now a dataset with 100 million hypothetical polymers. It includes their predictions for 29 properties. This dataset is available for academic use. Researchers have many chances to study polymers.
Breakthroughs in Polymer Design
The impact of polyBERT is clear in recent publications in the Nature journals. The first paper, in Nature Reviews Materials, shows breakthroughs in polymer design. These breakthroughs are in important areas like energy storage, filtration technologies, and recyclable plastics. The second paper, in Nature Communications, talks about using AI algorithms to find a subclass of polymers for electrostatic energy storage. The researchers used AI tools to find that insulation materials made from polynorbornene and polyimide polymers can achieve high energy density and high thermal stability at the same time. This is a big step forward for capacitor technology.
“The new class of polymers with high energy density and high thermal stability is one of the most concrete examples of how AI can guide materials discovery.” – Rampi Ramprasad
Collaborations and Industry Partnerships
Ramprasad’s team works with researchers from different institutions. They include Georgia Tech. The team uses AI to discover new polymers. Polymers are materials made of long chains of molecules.
Professor Ryan Lively is from the School of Chemical and Biomolecular Engineering. He often works with Ramprasad’s group. He helped write a paper published in Nature Reviews Materials.
The article shows that AI can help develop materials for real-world use. Companies like Toyota Research Institute and General Electric also took part in writing the article. Read more about this here.
The Future of Polymer Research
To speed up the use of AI in developing new materials for industry, Ramprasad co-founded Matmerize Inc. This is a new software startup company that was recently created by separating it from Georgia Tech. Their software runs on the cloud, meaning you can access it via the internet. It helps companies work with polymers, which are materials made of long, repeating chains of molecules. Companies in energy, electronics, consumer products, and sustainable materials are already using this software.
As the research progresses, the team has a long-term goal. They want to use the data they extract. With this data, they will train models to predict material properties. This can speed up the discovery of new polymers and materials. Polymers are large molecules made up of repeating units, and materials refer to substances used to make things.
For more intriguing insights into other STEM-related topics, visit ENTECH Online. Explore our digital magazine dedicated to inspiring teenagers and young adults to pursue their passions in science, technology, engineering, and mathematics.
Reference: DOI 10.1038/s41578-024-00708-8
