Using nanoscale science to unlock new approaches for storing hydrogen as a fuel alternative

As global fuel shortages and rising energy prices continue to make headlines, researchers at the University of Otago are exploring fuel alternatives that are cheap, clean, and sustainable.

Establishing a hydrogen economy is one possible answer to the global dependence on scarce fossil fuels and highly variable non-conventional resources.

Lekshmi Dinachandran, working in Dr Anna Garden’s group in the Department of Chemistry, recently focused her PhD on FeTi-based hydrogen storage materials, combining density functional theory, thermodynamic modelling, and mechanical property analysis to better understand hydride formation and stability.

Running computational models on the REANNZ High Performance Computing (HPC) Platform has been essential for this work.

“Thermodynamics-based calculations are significant at the nanoscale to consider surface energies and interface energies,” she says. “High-performance computing facilities were essential to this work, enabling large-scale simulations and data processing that would have been computationally impractical on standard systems.”

Why models matter

Hydrogen can be stored as a solid using a class of materials called metal hydrides. Commonly used metal hydride models assume idealized behaviour at the bulk scale. However, in actual hydrides, the hydrogenation and dehydrogenation involve many nanoscale processes influenced by interfaces and surfaces that may affect the sorption.

“Confined hydride alloys are subjected to significant mechanical stresses that affect their thermodynamics,” says Lekshmi. “There is extrinsic stress induced by the container, internal stresses that can result from volume expansion, as well as other influencing factors such as reaction pathways and contaminants.”

These stresses take a toll, weakening the hydride bed over time and causing it to fatigue or even crumble. In addition, the presence of impurities in the alloy can also affect its hydrogen storage performance.

This is where computational models can offer the most benefit. They enable researchers to adjust and capture real‑world effects that can then help them predict and understand how materials will behave.

Pictured above: A graphical abstract from a related project of Lekshmi's, published in a paper titled, Predicting hydrogen storage capacity loss of FeTi by ab initio thermodynamic modelling of Fe–Ti–Si phase equilibria.

How machine learning can help

While the HPC solved the computational capability limitations for DFT, these simulations also generate a structured dataset that could be extended using machine learning to accelerate predictions and explore a wider range of materials.

Although machine learning was not used in this study, it represents a natural next step where models trained on DFT data can stand in for computationally expensive calculations.

Access to HPC is therefore critical not only for scaling the underlying simulations, but also for producing the high-quality data needed to enable and support such approaches.

From ideas to reality

Eventually, the results from this project are expected to contribute to the development of more accurate models for the nanoscale behaviour of metal hydride hydrogen storage materials.

As the world searches for alternative energy options, research like Lekshmi’s demonstrates how computational science is paving the way for smarter, more efficient hydrogen storage technologies.

Do you have an example of how our HPC platforms, network services, or support team have helped advanced your work? We’re always looking for success stories to feature as a case study. Get in touch.