Rishi Gurnani1,Stuti Shukla2,Deepak Kamal1,Chao Wu2,Chris Kuenneth1,Yang Cao2,Gregory Sotzing2,Rampi Ramprasad1
Georgia Institute of Technology1,University of Connecticut2
Rishi Gurnani1,Stuti Shukla2,Deepak Kamal1,Chao Wu2,Chris Kuenneth1,Yang Cao2,Gregory Sotzing2,Rampi Ramprasad1
Georgia Institute of Technology1,University of Connecticut2
A record-setting new polymer dielectric that displays the highest-ever electrostatic energy density (8.9 J/cc) at 200 C has been designed using advanced AI algorithms, and subsequently validated through synthesis and breakdown testing. This new flexible, all-organic polymer, polyoxanorbornene-1 (PONB1), beats the current industry standard for high energy density capacitors--biaxially oriented polypropylene (which degrades rapidly above 25 C)--by a significant margin of 3.5 J/cc. Concurrent developments, also by us, in polymer informatics played a key role in the discovery of PONB1. These developments include an advanced deep learning framework that reduces the time required for polymer screening by more than 50 times compared to traditional approaches, and two algorithms for generating realistic but never-before-synthesized polymers. Our progress, especially in the development of algorithms for polymer generation and in the real-world validation of PONB1, was made possible by lively collaboration and iteration with experimental colleagues. Looking forward, the polyoxanorbornene family--which includes PONB1--expands the scope and effectiveness of capacitor-based energy storage, a technology that may ultimately power the world's electrification. Given the results of our work, it is likely that other valuable candidates exist in our massive data set of never-before-synthesized polymers for many applications. We offer this data set to the community, as well as the open source code for our advanced deep learning framework.