Informatics-Driven Selection of Polymers for Fuel-Cell Applications

TL;DR

The paper tackles Nafion's shortcomings in PEM fuel cells by introducing an informatics-driven screening pipeline that links defined material requirements to rapid property predictions. It builds six Gaussian process regression models, including two multi-task models, to predict proton conductivity, water uptake, gas permeabilities, band gap, glass transition, decomposition temperature, and Young's modulus from SMILES-derived fingerprints. Applying the approach to 30,624 known polymers yields 60 candidates (including PEM1, AI1, CI1) with predicted superior proton conductivity and robust mechanical properties, offering a viable path toward experimental validation. The framework is generic and extensible, enabling accelerated discovery of polymer materials for fuel cells and other applications.

Abstract

Modern fuel cell technologies use Nafion as the material of choice for the proton exchange membrane (PEM) and as the binding material (ionomer), used to assemble the catalyst layers of the anode and cathode. These applications demand high proton conductivity as well as other requirements. For example, PEM is expected to block electrons, oxygen, and hydrogen from penetrating and diffusing while the anode/cathode ionomer should allow hydrogen/oxygen to move easily, so that they can reach the catalyst nanoparticles. Given some of the well-known limits of Nafion, such as low glass-transition temperature, the community is in the midst of an active search for Nafion replacements. In this work, we present an informatics-based scheme to search large polymer chemical spaces, which includes establishing a list of properties needed for the targeted applications, developing predictive machine-learning models for these properties, defining a search space, and using the developed models to screen the search space. Using the scheme, we have identified 60 new polymer candidates for PEM, anode ionomer, and cathode ionomer that we hope will be advanced to the next step, i.e., validating the designs through synthesis and testing. The proposed informatics scheme is generic, and can be used to select polymers for multiple applications in the future.

Figures (6)

• Figure 1: (a) Schematic of a cross-sectional view of a PEM fuel cell unit in which a PEM layer is sandwiched between two electrodes (anode and cathode), each of them contains a gas diffusion layer (GDL) and a catalyst layer (CL), and (b) general chemical structure of Nafion, a random copolymer composed of an electrically neutral semicrystalline polytetrafluoro-ethylene backbone and a randomly tethered side-chain ending with the pendant sulfonate group -SO$_3^-$. The backbone length of Nafion is $m \simeq 5.5$ while the polar, hydrophilic -SO$_3^-$ sulfonate groups are essential for capturing water molecules. The catalyst layers are created by using polymeric ionomers to bind Pt nanoparticles together.
• Figure 2: A machine-learning and multi-objective driven scheme for the selection of polymers that can be used as a PEM or (anode/cathode) ionomer in fuel cells.
• Figure 3: (a) The MT model M1, trained on two datasets of proton conductivity and water uptake, for predicting the proton conductivity $\sigma$, and (b) Proton conductivity $\sigma$, predicted by model M1 at $80^\circ$C as a function of the relative humidity RH, given in comparisons with experimental data.
• Figure 4: Chemical structrure and key properties predicted for PEM1, AI1, and CI1, the representative candidate of PEM (top row), anode ionomer (lower left), and cathode ionomer (lower right), respectively.
• Figure 5: Property radar chart of PEM1, AI1, and CI1 whose details are given in Fig. \ref{['fig:cands']}. Their key properties are given in unit of the same properties of Nafion (scale is given in grey numbers). For PEM1, the proton conductivity $\sigma$ at RH=100% is used.