Hydrogen is often described as a promising energy carrier, but storing it safely and efficiently remains a central engineering challenge. A recent study led by Alauddin Ahmed at the University of Michigan College of Engineering proposes a simpler way to evaluate hydrogen storage materials, one that could shorten the path from theoretical screening to practical deployment.
Ahmed, A. (2025). Physics-Informed Analytical Models for Interpretable and Deployable Hydrogen Storage Prediction in MOFs. PRX Energy, 4(4), 043009. https://doi.org/10.1103/glnw-drn4
The work focuses on metal-organic frameworks, or MOFs, a class of crystalline materials made from metal nodes connected by organic linkers. Their appeal lies in their extreme internal porosity. Even a small amount of MOF material contains a vast internal surface area, allowing hydrogen molecules to adhere to pore walls through physical adsorption rather than chemical bonding. This makes MOFs attractive for reversible hydrogen storage under controlled temperatures and pressures.
Alauddin Ahmed at the University of Michigan College of Engineering stated,
“More broadly, this work shows that physics-informed symbolic regression can be a practical bridge between large-scale simulations and real-world design. If we can replicate this success across other classes of energy materials, we will have a powerful new way to turn data into insight and, ultimately, into better technologies.”
Until now, predicting how much hydrogen a given MOF can store under real-world conditions has required either detailed molecular simulations or large machine-learning models. While accurate, these approaches are computationally expensive and often difficult to interpret. Engineers screening thousands of candidate materials are left with predictions that may perform well numerically but offer little insight into the physical drivers behind storage performance.
The Michigan research team took a different approach. Instead of relying on black-box models, they developed physics-informed equations that link hydrogen storage capacity directly to a small number of measurable material properties. After analyzing a curated dataset of more than 88,000 MOFs, they found that two parameters dominate performance: void fraction, which measures how much empty space exists inside the crystal, and pore volume, which reflects how that space is distributed.
Using symbolic regression techniques, the researchers searched through billions of mathematical forms to identify the simplest equations that still matched the accuracy of advanced simulations. The resulting models were then tested against a much larger independent dataset containing approximately 600,000 additional MOF structures. Across this expanded set, the equations remained reliable under practical operating conditions, covering pressure ranges relevant to hydrogen storage systems at cryogenic temperatures.
The results show that void fraction alone can predict how efficiently a MOF uses volume to store hydrogen, while a combination of void fraction and pore volume provides a strong estimate of weight-based storage capacity. Adding more structural descriptors led to only marginal improvements, suggesting that much of the variation in hydrogen uptake can be explained with surprisingly simple physical measures.
This finding is particularly relevant in the context of hydrogen storage targets, which are defined in terms of both mass and volume efficiency. While no known MOF currently satisfies all performance benchmarks simultaneously, the new equations make it easier to identify which structural directions are most likely to close the gap.
Beyond hydrogen storage, the study illustrates a broader shift in materials engineering toward interpretable modeling. In energy systems where safety and reliability are critical, engineers benefit from models that explain their predictions rather than obscure them behind complex computation. By embedding physical constraints directly into the mathematical search process, the researchers demonstrated that transparency and accuracy do not have to be traded off against each other.
The work also suggests a practical route for integrating such models into automated materials design pipelines. Because the equations are fast to evaluate and require minimal computational resources, they can be deployed early in the screening process to filter large material libraries before more detailed analysis is applied.
As hydrogen continues to gain attention across transportation, grid storage, and industrial energy systems, approaches like this may play an important role in deciding which materials progress from databases to prototypes. By reducing reliance on heavy computation and emphasizing physical insight, the study offers a more direct way to connect materials discovery with engineering design.
Adrian graduated with a Masters Degree (1st Class Honours) in Chemical Engineering from Chester University along with Harris. His master’s research aimed to develop a standardadised clean water oxygenation transfer procedure to test bubble diffusers that are currently used in the wastewater industry commercial market. He has also undergone placments in both US and China primarely focused within the R&D department and is an associate member of the Institute of Chemical Engineers (IChemE).
