Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Accelerated design of proton exchange membranes for green hydrogen production with artificial intelligence

Huan Tran, Akhlak Mahmood, Harshal Chaudhari, Kuldeep Mamtani, Chiho Kim, Rampi Ramprasad, Anand N. Krishnamoorthy, Abhirup Patra

TL;DR

The paper addresses the challenge of designing sustainable, halogen-free proton-exchange membranes for PEM electrolyzers to enable green hydrogen production. It introduces a scalable AI-driven workflow that combines a virtual forward synthesis (VFS) approach with multi-task Gaussian Process Regression models to predict key membrane properties, screening millions of polymers to identify 1,738 promising candidates. The strategy is validated against experimental data for Nafion, Pemion, and sPPO and extended to uncharted polymers, uncovering design rules tied to sulfonated and aromatic blocks and highlighting a path toward interactive AI agents for iterative materials design. The work promises to accelerate the discovery of robust, high-performance PEMs and offers a blueprint for AI-assisted, large-scale polymer discovery in energy materials.

Abstract

Water electrolysis is an eco-friendly method for hydrogen production that has reached significant levels of technological maturity. Among commercialized water-electrolysis technologies, proton-exchange membrane electrolyzers offer high current density, fast dynamic response, and compact system design, among other advantages. On the other hand, managing their high capital cost and the ``forever-chemistry'' nature of Nafion, a perfluorinated proton-exchange membrane widely used in such devices, remains a major challenge. Searches for fluorine-free replacements for Nafion, pursued largely through physical experimentation, have been active for decades with limited success. In this work, we develop and demonstrate an AI-based strategy for designing new proton-exchange membranes for electrolyzers. Two key components of this strategy are an implementation of the virtual forward-synthesis approach and a set of machine-learning predictive models for essential application-inspired membrane properties; the former generates a vast space of millions of synthesizable polymers, which are then evaluated and screened by the latter. The strategy is validated against experimental data for known membranes and then applied to design over 1,700 new synthesizable candidates. This article concludes with a forward-looking vision in which the strategy could be elevated into an interactive and iterative scheme that are based on large language models to facilitate materials design in multiple ways.

Accelerated design of proton exchange membranes for green hydrogen production with artificial intelligence

TL;DR

The paper addresses the challenge of designing sustainable, halogen-free proton-exchange membranes for PEM electrolyzers to enable green hydrogen production. It introduces a scalable AI-driven workflow that combines a virtual forward synthesis (VFS) approach with multi-task Gaussian Process Regression models to predict key membrane properties, screening millions of polymers to identify 1,738 promising candidates. The strategy is validated against experimental data for Nafion, Pemion, and sPPO and extended to uncharted polymers, uncovering design rules tied to sulfonated and aromatic blocks and highlighting a path toward interactive AI agents for iterative materials design. The work promises to accelerate the discovery of robust, high-performance PEMs and offers a blueprint for AI-assisted, large-scale polymer discovery in energy materials.

Abstract

Water electrolysis is an eco-friendly method for hydrogen production that has reached significant levels of technological maturity. Among commercialized water-electrolysis technologies, proton-exchange membrane electrolyzers offer high current density, fast dynamic response, and compact system design, among other advantages. On the other hand, managing their high capital cost and the ``forever-chemistry'' nature of Nafion, a perfluorinated proton-exchange membrane widely used in such devices, remains a major challenge. Searches for fluorine-free replacements for Nafion, pursued largely through physical experimentation, have been active for decades with limited success. In this work, we develop and demonstrate an AI-based strategy for designing new proton-exchange membranes for electrolyzers. Two key components of this strategy are an implementation of the virtual forward-synthesis approach and a set of machine-learning predictive models for essential application-inspired membrane properties; the former generates a vast space of millions of synthesizable polymers, which are then evaluated and screened by the latter. The strategy is validated against experimental data for known membranes and then applied to design over 1,700 new synthesizable candidates. This article concludes with a forward-looking vision in which the strategy could be elevated into an interactive and iterative scheme that are based on large language models to facilitate materials design in multiple ways.
Paper Structure (12 sections, 5 figures, 3 tables)

This paper contains 12 sections, 5 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Methods
  3. Design criteria
  4. Training data and ML models
  5. Design space
  6. Results
  7. Machine-learning models
  8. Discovered PEM candidates
  9. Known membranes: experimental validations
  10. Uncharted territory: PEM candidates
  11. Outlooks
  12. Summary

Figures (5)

  • Figure 1: (a) A (schematic) cross-sectional view of a PEM electrolyzer in which a direct-current power source is used to split provided water molecules into hydrogen and oxygen gases in cathode and anode, (b) the chemical structure and a visualization of Nafion membrane, (c) the performances of Nafion, Pemion, and sPPO, given with respect to the application-inspired thresholds, and (d) the employed design strategy for PEMs. In (c), the design criteria are represented by a threshold polygon, outside of which good candidates for PEMs are expected to be.
  • Figure 2: ML models for (a) proton conductivity and water uptake, (b) glass transition temperature, and (c) thermal decomposition temperature.
  • Figure 3: Chemical structure and predicted proton conductivity (a) and water uptake (b) of Nafion, given in comparison with experimental data, taken from Refs. feng2018characterizationzhang2018sulfonatedpeng2017preparationsi2012synthesiswang2012clusteredsilva2004tangential and Refs. wang2012clusteredzhang2018sulfonated, respectively. In (b), (c), (d), and (e), the predicted proton conductivity of Pemion, sPPO, and 2 sulfonated aromatic poly(ether sulfone) copolymers SPAES1 and SPAES2 are given. Measured proton conductivity of Pemion was taken from Refs. nguyen2021hydrocarbonpermionionomr. For sPPO, $x=0.2$ and data were measured in this work. In SPAES1, Y is -S-, Z is -SO$_2$-, and $n/(m+n)=0.3$, while for SPAES2, Y is -S-, Z is -C(=O)-, and $n/(m+n)=0.3$. Shaded areas in panels (b)-(f) represent the uncertainty of the predictions for "water vapor" (given in solid lines).
  • Figure 4: Ten most-frequent functional groups/blocks found in 1,738 candidates identified from subset 2, containing 66M polymers generated using VFS.
  • Figure 5: Schematic illustration of the envisioned interactive AI agent workflow for application-driven polymer design. User specify performance requirements through a natural language interface, the AI agent translates the requirements into quantitative design criteria, generates polymer candidates using RxnChainer (VFS approach), predicts their properties using trained ML models, and engages in iterative refinement loop based on user feedback. The agent provides explanations of structure-property relationships and presents ranked candidates with predicted property values, uncertainty estimates, and synthesizability scores.