New grant to support AI-guided design of redox polymers

Chemical and biomolecular engineering professors Theresa Schoetz and Diwakar Shukla have received a seed grant to develop an artificial intelligence–guided approach to designing redox-active polymer materials aimed at safe, flexible and sustainable energy storage. The project will establish a proof‑of‑concept framework that uses large language‑model (LLM) technology to accelerate the design and fabrication of redox‑active polymers for next-generation battery applications.

The work addresses a growing need for alternatives to conventional lithium-ion batteries, which can be rigid, pose safety risks, and rely on materials with sustainability challenges. These limitations are particularly problematic for technologies such as wearable or reconfigurable electronics. Redox‑active polymers offer a promising path forward, the researchers said.

“Redox-active polymers can store and release electrical energy through many repeating redox sites along their molecular chain, which can make batteries charge faster and be made from lightweight, potentially low-cost organic materials,” Schoetz said. “However, they present challenges in that many of these polymers conduct electricity poorly and can dissolve or degrade in the electrolyte, which reduces battery lifetime.”

In addition, the rational design of redox-active polymers is difficult due to the enormous chemical design space and the limited amount of available experimental data.

To overcome these challenges, Schoetz and Shukla will create an AI system capable of efficiently designing promising redox‑polymer candidates. Instead of training an entirely new model from scratch, the team will develop a novel system called RedoxPolymerLLM that integrates a pretrained polymer language model with a second redox-focused language model. Using cross‑attention mechanisms, the new system will learn how polymer structure and electrochemical performance are connected, enabling it to represent and design redox polymers more effectively than either model could achieve on its own.

“Pre-trained models have already internalized vast amounts of knowledge about polymers and redox molecules,” Shukla said. “Therefore, combining models allows us to focus on the chemicals that form polymeric structures while also enabling us to store electricity simultaneously.”

The RedoxPolymerLLM will guide the targeted electropolymerization of redox-polymers for aluminum-based batteries, linking AI‑generated designs to laboratory synthesis and electrochemical testing. By integrating computation, simulation, synthesis, and testing into one closed loop, the project aims to deliver a validated end‑to‑end workflow for AI-guided materials discovery and fabrication.

This research aligns closely with the mission of the Molecule Maker Lab Institute, which awarded the grant, to advance AI enabled polymer science and foundational AI methodology. The project will also help establish a framework for future autonomous materials discovery efforts.