Hydrogen can store renewable energy and power fuel cells, but finding materials that safely store hydrogen at useful weights and pressures has been a persistent obstacle. A research team led by Tohoku University has now developed an AI-powered method that maps out which physical properties matter most for solid hydrogen storage, turning scattered experimental data into a clear design guide. The study, published in Chemical Science on May 25, 2026, offers a way to speed up the search for materials that could make green energy storage more practical.
Combining a curated database with explainable AI
The team built on DigHyd, a database of hydrogen-storage measurements gathered from the scientific literature, and paired it with a symbolic-regression tool called GoodRegressor. Rather than generating a black-box prediction, GoodRegressor searches for human-readable equations that link material properties to performance. This approach reveals the underlying physics, not just a list of candidate materials.
Two key properties, two separate controls
The analysis showed that hydrogen capacity and room-temperature equilibrium pressure are controlled by different material features. Capacity depends mainly on the average size of the metal atoms and on thermal conductivity, which reflects how the metal lattice responds when hydrogen enters its empty spaces. The most favorable materials have an average metal-atom radius tuned to about 1.47 Å and a relatively soft lattice. In contrast, the pressure at which hydrogen is absorbed or released is governed by elastic properties-particularly shear modulus and Poisson's ratio-that describe lattice stiffness.
"The model doesn't spit out suggestions - it explains why certain physical properties matter, which we can then logically apply to produce the desired outcome," said Hao Li, a distinguished professor at the Advanced Institute for Materials Research.
A practical blueprint for material design
This separation of roles gives researchers a clear strategy: adjust geometry and lattice flexibility to increase how much hydrogen a material can hold, while tuning stiffness to keep the equilibrium pressure near ambient conditions, around one atmosphere. Using that framework, the team proposed specific composition-changing routes for several major classes of interstitial hydrides, including BCC alloys, Laves phases, LaNi5-type materials, and TiFe-type materials. These are design candidates that still need experimental validation, since they are new, but they narrow the search space considerably.
"Even so, the work provides an explainable way to narrow the search space and reduce trial-and-error in developing solid hydrogen-storage materials," said Seong-Hoon Jang, an associate professor at the Unprecedented-scale Data Analytics Center.
Why this matters for science and research professionals
For researchers working on energy storage, the study translates a large collection of experimental data into a physically grounded design logic. Instead of screening thousands of materials by intuition or brute-force computation, materials scientists can now target specific atomic-scale knobs-such as average atom radius and shear modulus-to engineer hydrides that balance capacity and usable pressure. The same approach could extend to other energy materials, like ionic hydrides and hydride-based solid electrolytes, offering a template for data-driven, interpretable materials discovery.
Your membership also unlocks:
