Skip to main content


The goal of HTP is to discover novel polymer electrolytes for safer batteries.  For decades, it has been recognized that a polymer – rather than a liquid electrolyte – could be a game changer for battery design by enabling high energy density with excellent safety. However, lithium conductivity in solid polymers is too low to be practical in batteries. While conductivity has improved, there has not been a breakthrough in performance.

We must develop more efficient tools that leverage scientific intuition, big data, and machine learning to expand our polymer design space. Some of the key components we are developing at TRI and through our research collaboration with MIT are:

  • Predictive and generative AI models for optimizing polymer design and expanding our molecular design space, 
  • Automated high-throughput molecular dynamics simulations (HTP-MD) to compute transport properties and evaluate mechanisms of hypothetical materials,
  • A high-throughput experimental characterization platform (located at MIT) that can measure the transport properties of new polymers across temperatures and salts.

Key to our work is leveraging recent advances in generative AI, such as models based on Chat-GPT, to generate hypothetical polymers that have desirable target physical properties. Just as Chat-GPT learns about written text, our models can learn the “language” and “syntax” of molecules. 

The model can propose new alternative chemical structures that are checked for chemical validity, synthetic accessibility, and novelty. These polymers are then passed to our automated molecular dynamics (MD) simulation pipeline, called HTP-MD. We use closed-loop feedback with these simulations to continuously generate better and better polymers. Even though many of these polymers only exist in simulation, we can learn from ion transport mechanisms in these new polymer systems and use this information to improve the design of real-world polymers. 

Ultimately, once AI-generated candidates are evaluated in simulation, we can go to the lab to attempt to synthesize the most promising polymers and then rapidly complete a full evaluation across different operating conditions (temperatures, salts, and concentrations). We envision a polymer discovery framework that can improve the design of existing polymers and eventually enable breakthroughs. 

While AI may be able to inspire new polymers or optimize existing ones, the expert scientist must remain central to the design process. This integration of human expertise and computational power holds excellent potential for accelerating the discovery of breakthrough polymers.

htpmd image

Web App

A web app for exploring data easily and intuitively is found at

Public Release Code Library

The platform includes an open-source, publicly-released code library containing analysis methods at


Z. Yang, W. Ye, X. Lei, D. Schweigert, H.-K. Kwon, and A. Khajeh. De novo Design of Polymer Electrolytes with High Conductivity using GPT-based and Diffusion-based Generative Models. Dec. 2023.

X. Lei, W. Ye, Z. Yang, D. Schweigert, H.-K. Kwon, and A. Khajeh. A Self-Improvable Polymer Discovery Framework Based on Conditional Generative Model. Dec. 2023.

T. Xie et al. “A cloud platform for sharing and automated analysis of raw data from high throughput polymer MD simulations.” APL Machine Learning, vol. 1, no. 4, p. 046108, Dec. 2023, doi: 10.1063/5.0160937.