High-temperature proton exchange membrane fuel cells (HT-PEMFCs) are highly promising for next-generation aviation, as they can operate above 160 °C and tolerate impurities in the fuel. However, they still struggle to achieve the single-cell power density of over 1.5 W·cm⁻² required for practical aircraft, primarily because dynamic phosphoric acid (PA) migration and poisoning lead to severe performance losses in the catalyst layer. A new study published in Engineering by Xin Wang, Jian Yao, Jin Zhang and their colleagues proposes a machine-learning-guided strategy that combines data-driven descriptor screening with three-dimensional multiphysics and lattice Boltzmann simulations. The authors elucidate how the PA content and its non-uniform distribution within the catalyst layer govern proton transport, platinum utilization, and local current generation. Guided by these insights, they introduce an EDTMPA–PA dual-proton-conducting system and construct a gradient dual-proton-conductor catalyst layer, enabling single cells to achieve record power densities at 160 °C while maintaining long-term operational stability under aerospace-relevant conditions.

The machine learning method was employed to identify the key factors affecting the performance of HT-PEMFCs. The authors found that, among numerous factors, the content of phosphoric acid within the catalyst layer (PA in CL) was the most significant indicator influencing HT-PEMFC performance, accounting for 25.5% of the overall performance contribution—far exceeding traditional parameters. Moreover, after contrasting extensive literature, it was discovered that the PA content within the catalyst layer exhibits a volcano-type relationship with the power density of the cell, which laid a theoretical foundation for the subsequent optimization of the catalyst layer.

Using a three-dimensional multiphysics model coupled with a lattice Boltzmann method, the team reveals that PA redistribution under operating conditions gives rise to a pronounced gradient from the membrane toward the gas diffusion layer (GDL). This gradient causes progressive deactivation of Pt in membrane-distal regions and severely reduces the local current density. Experiments with gradient-designed electrodes confirm that PA-deficient zones near the GDL experience up to a 111.5% loss in Pt utilization compared with PA-rich regions, directly limiting the power output of the cell.

Based on these insights, the researchers designed three gradient cathode catalyst layers using Pt/C catalysts and XC-72 carbon powder. The Pt/C catalysts were placed on the PEM side (IN) and the GDL side (OUT) respectively. The catalysts were uniformly mixed with the carbon powder (MIX) as a blank control group.

The authors discovered that a series of electrochemical properties, such as electrochemical active area, ionic impedance, and internal resistance to oxygen transport, were significantly better for the IN electrode compared to the OUT electrode. Under hydrogen/air operation at 160 °C, the power density of the IN electrode (389.5 mW·cm⁻²) is twice that of the OUT electrode (184.1 mW·cm⁻²). These results fully validate the multiphysics simulation results regarding the non-uniform distribution behavior of PA.

To address the issue caused by the uneven distribution of PA, the authors introduced a dual proton conductor system based on the organic phosphonic acid additives EDTMPA and PA. The strong hydrogen-bond interaction between EDTMPA and PA forms a robust proton conduction network. Compared with the pure PA system, the EDTMPA–PA mixture exhibits higher proton conductivity and lower activation energy over a wider temperature range, ensuring continuous proton pathways even in regions with initially lower proton conductor content in the catalyst layer. The research team first applied this strategy to the OUT electrode with the poorest performance.

The team found that the various electrochemical properties of the optimized OUT electrode using a dual proton conductor mixed system were significantly improved. In particular, its power density increased by 62.3%. This indicates that EDTMPA enhances the availability of proton conduction pathways, and that the additional EDTMPA promotes acid enrichment in regions lacking PA, thereby forming a continuous proton conduction network within the catalyst layer. Therefore, this dual proton strategy effectively alleviates the local proton deficiency observed in the outer sublayer of the OUT electrode.

By integrating this dual-proton conductor system into the adjacent region of the full-size electrode (ALL + EDTMPA), the researchers achieved a peak power density of 2.16 W·cm⁻² at 160 °C, 1.5 bar back pressure, under hydrogen/oxygen conditions, fully meeting and exceeding the power density target of aerospace fuel cells. After more than 15 cycles of high-intensity dynamic load cycling tests, the improved battery demonstrated excellent stability.

This study provides clear mechanistic insight and practical design guidelines for regulating phosphoric-acid distribution and engineering gradient dual-proton-conducting catalyst layers in HT-PEMFCs, ultimately advancing the development of lightweight, high-power fuel cell systems for aviation applications.

The paper “Machine Learning-Guided Gradient Dual-Proton Conducting Catalytic Layers for High Temperature Proton Exchange Membrane Fuel Cells in Aviation” is authored by Xin Wang, Jian Yao, Zhenguo Zhang, Jialin Zhang, Baohua Liu, Wen Liu, Wen Li, Shanfu Lu, Yan Xiang, Haining Wang, San Ping Jiang, and Jin Zhang. Full text of the open access paper:https://doi.org/10.1016/j.eng.2025.10.031. For more information about Engineering, visit the website at https://www.sciencedirect.com/journal/engineering.