⚡ electrical engineering

Gas Diffusivity in Fibrous Gas Diffusion Layers: Linking Pore Structure to Transport

This study presents a computationally efficient model based on effective medium theory that accurately predicts gas diffusivity in fibrous gas diffusion layers by transforming standard porosimetry data into a representative pore structure and deriving a compact correlation validated against experimental and literature data.

Original authors: Amin Mohammadi, Majid Bahrami

Published 2026-09-02

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Amin Mohammadi, Majid Bahrami

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Inside a fuel cell, a device that turns hydrogen and oxygen into electricity and water, the journey of gas is a race against time. For the cell to work efficiently, gases must travel quickly through a porous, sponge-like material called a gas diffusion layer. This layer acts as a bridge, carrying reactants to the heart of the cell where the chemical reaction happens. If the gas gets stuck or moves too slowly, the entire system loses power. Scientists have long known that the amount of empty space, or porosity, in this material matters. A material with more holes generally lets gas pass through more easily. However, this simple measure of "how much space" tells only part of the story. It ignores the shape, size, and arrangement of those holes, which can create hidden bottlenecks that slow down the flow even when the total space seems ample.

Researchers at Simon Fraser University have developed a new way to look at this problem, moving beyond the simple count of empty space to understand the actual architecture of the pores. They found that the standard way of measuring these materials, which assumes the holes are perfect cylinders, fails to capture the true nature of the fibrous structure found in real fuel cell components. By creating a model that translates these measurements into a more realistic representation of the material's internal geometry, they were able to predict gas movement with much greater accuracy. Their work suggests that the specific distribution of pore sizes is just as critical as the total amount of porosity, offering a clearer path to designing better fuel cells.

The team focused on a specific type of material used in fuel cells, made of tiny carbon fibers woven together. To understand how gas moves through this tangled web, they first needed to know the size of the gaps between the fibers. They used a standard laboratory technique involving a liquid called octane. By slowly letting the liquid evaporate from the material while measuring how much remained, they could calculate the pressure required to push air into the pores. This pressure tells scientists the size of the openings, but the standard calculation assumes every opening is a smooth, straight cylinder. In reality, the carbon fibers are curved and arranged in complex patterns, meaning the actual paths for gas are far more irregular than a simple cylinder suggests.

To fix this mismatch, the researchers created a mathematical transformation. They took the data from the standard measurements and converted it from a list of cylinder sizes into a list of distances that represent the actual gaps between the fibers in their model. Imagine the material as a grid of tiny boxes, where the size of each box represents a pore. Instead of guessing the size based on a cylinder, they calculated the exact length of the gap between two fibers that would create the same pressure resistance as the measured pore. This step allowed them to build a digital map of the material that looked and behaved like the real thing, capturing the true complexity of the fibrous structure.

Using this new map, they constructed a virtual three-dimensional network of these tiny boxes. They assigned a resistance value to each box based on its size and orientation, simulating how hard it would be for gas to move through that specific spot. By running gas through this digital network, they could calculate the overall ease of flow, known as diffusivity. The simulations revealed a crucial insight: two materials with the exact same amount of empty space could have very different abilities to transport gas if their internal pore sizes were distributed differently. A material with many tiny, restrictive gaps would slow the gas down significantly more than a material with a few large, open channels, even if the total volume of empty space was identical. This finding directly challenges older models that rely solely on total porosity, which would predict the same performance for both materials.

To make this complex simulation useful for engineers who need quick answers, the team developed a compact formula. This equation takes the total porosity and the statistical details of the pore size distribution to predict gas flow without needing to run a full computer simulation every time. They tested this formula against real-world experiments using a specialized device that measures how gas moves through a sample under controlled conditions. The results showed that their new method matched the experimental data with a high degree of accuracy, with differences of no more than eight percent. This level of precision confirms that accounting for the specific shape and size of the pores is essential for understanding how gas moves through these materials.

The study concludes that the internal architecture of the gas diffusion layer is a key factor in fuel cell performance. By shifting the focus from a simple count of holes to a detailed understanding of their size and arrangement, the researchers have provided a more reliable tool for predicting how these materials will behave. Their work demonstrates that the way pores are distributed within the material dictates the flow of gas just as much as the total amount of space available. This deeper understanding allows for the design of better fuel cell components, potentially leading to more efficient and powerful energy systems in the future.

Technical Summary: Gas Diffusivity in Fibrous Gas Diffusion Layers: Linking Pore Structure to Transport

Problem Statement
Gas transport through the gas diffusion layer (GDL) is a critical source of mass transfer resistance in proton exchange membrane fuel cells (PEMFCs), directly impacting cell performance. While the effective gas diffusivity depends on microstructural characteristics—specifically pore size distribution (PSD) and fiber diameter—conventional analytical models often rely solely on porosity. These porosity-based approaches (e.g., Bruggeman) tend to overpredict diffusivity because they fail to distinguish between different microstructures sharing the same porosity. Conversely, while numerical approaches like pore network modeling (PNM) and direct numerical simulation (DNS) offer high accuracy by resolving pore-scale morphology, they are computationally expensive and require detailed structural inputs, limiting their utility for rapid estimation or integration into larger-scale system models. There is a need for a modeling approach that captures key microstructural features, particularly PSD, while remaining computationally efficient.

Methodology
This study develops a new model based on effective medium theory to predict relative gas diffusivity in dry GDLs. The methodology proceeds through three primary stages:

Pore Size Transformation:
Standard porosimetry measurements (using octane) characterize GDLs based on an equivalent cylindrical pore geometry. However, GDLs consist of fibrous structures. To bridge this gap, the authors developed a transformation method. They modeled the GDL as an assembly of cylindrical carbon fibers (8 µm diameter) arranged orthogonally, forming a repeating unit cell. A capillary model was derived for this unit cell geometry to relate capillary pressure to the unit cell length ( $2l$ ). By mapping the capillary pressure derived from the standard porosimetry (via the Washburn equation) to the unit cell geometry, the authors converted the original cylindrical pore radius distribution into an equivalent unit cell length distribution. This transformed distribution was fitted using a cumulative log-normal function characterized by a logarithmic mean ( $\mu$ ) and standard deviation ( $\sigma$ ).
Resistance Network Modeling:
A statistical three-dimensional network of unit cells ( $30 \times 30 \times 30$ ) was constructed. Each cell's length was randomly sampled from the transformed log-normal distribution. Directional diffusive resistances were assigned to each cell based on its geometry:
- In-plane ( $x, y$ ): Resistance depends only on local porosity.
- Through-plane ( $z$ ): Resistance accounts for varying cross-sectional areas due to fiber curvature, resulting in higher resistance for smaller unit cells.
  The network was solved to determine the equivalent resistance and, subsequently, the effective diffusivity ( $D_{eff}$ ). This approach explicitly captures the anisotropic nature of transport and the impact of pore size distribution.
Compact Correlation Development:
To avoid the computational cost of solving the full network for every application, a compact correlation was derived. Using a parametric dataset generated by varying $\mu$ and $\sigma$ within physically relevant ranges, the authors fitted the network model results to a function of overall porosity ( $\varepsilon$ ), $\mu$ , and $\sigma$ .

Key Contributions

Pore Size Transformation: A novel method to convert standard porosimetry data (cylindrical pore assumption) into a unit cell length distribution consistent with fibrous GDL geometry.
Physics-Based Network Model: A resistance-network model that incorporates the transformed PSD to predict anisotropic gas transport in dry GDLs, demonstrating that diffusivity varies significantly even when porosity is constant but the pore size distribution differs.
Compact Correlation: A simple, high-accuracy correlation ( $D^* = 0.60\varepsilon^{1.14}\exp(0.37\mu - 2.03\sigma)$ ) that reproduces the network model results, enabling rapid estimation of diffusivity for system-level modeling.

Results

Microstructural Sensitivity: The network model successfully predicted different relative diffusivities for synthetic cases with identical porosity but different PSDs. Specifically, distributions with a higher fraction of smaller unit cells (broader $\sigma$ or lower $\mu$ ) resulted in significantly lower diffusivity due to increased local resistance, a nuance missed by porosity-only models.
Correlation Accuracy: The compact correlation achieved an $R^2$ of 0.985 and a root mean square error (RMSE) of 0.014 when compared to the full resistance-network model. Sensitivity analysis confirmed that increasing $\mu$ (larger characteristic cell size) increases diffusivity, while increasing $\sigma$ (broader distribution) decreases diffusivity.
Experimental Validation: The model was validated against experimental data obtained using the symmetrical modified Loschmidt cell (SMLC) method for five GDL samples (various Toray TGP-060/090 and SGL 10 BA substrates) and literature data. The predictions showed good agreement with experimental measurements, with a maximum relative difference of 8%.

Significance
The paper claims that incorporating the transformed pore size distribution allows for a more accurate prediction of gas diffusivity in GDLs than models based on porosity alone. By linking the experimentally accessible PSD to a physically representative fibrous unit cell geometry, the proposed approach bridges the gap between detailed numerical simulations and simple analytical correlations. The resulting compact correlation offers a computationally efficient tool for estimating transport properties in dry GDLs, suitable for integration into larger-scale fuel cell models without sacrificing the ability to capture microstructural effects. The study emphasizes that while the model is validated for dry conditions, the explicit accounting of pore interconnectivity limitations in current analytical frameworks remains a constraint for predicting behavior under partially saturated conditions.

More like this