A new hydrogen storage database could help scientists understand why some materials hold more hydrogen but are harder to work with. A research team led by Tohoku University, with the University of Tokyo, has built the Digital Hydrogen Platform, or DigHyd. It is a curated database of more than 30,000 data entries drawn from over 4,000 experimental publications. Tohoku’s Advanced Institute for Materials Research (WPI-AIMR) announced the work on 23 October 2026. The study was published online in Applied Physics A on 4 September 2026.
Using the database, the team created a “physical map” that links material properties to two things: hydrogen storage capacity and equilibrium pressure. Equilibrium pressure helps set the conditions under which a material absorbs and releases hydrogen. The key idea is the trade-off between how much hydrogen a material can hold and the practical conditions needed to use it.
How the hydrogen storage database was built
The researchers say a reliable database needed more than automated data collection. They combined AI-assisted literature mining with careful human validation. That meant checking extracted information against the original publications, standardising units and analysing experimental curves.
For metal hydrides, which store hydrogen within their structures, the team manually curated 4,643 storage-capacity measurements and 652 pairs of enthalpy and entropy measurements. These thermodynamic quantities help predict how storage behaviour changes with temperature. Each entry links back to its original publication, so the data can be traced and verified. The DigHyd platform is publicly available online.
AI that writes equations, not black boxes
The team then applied GoodRegressor, an AI method based on symbolic regression. Instead of producing an opaque prediction, symbolic regression finds relationships in data and expresses them as mathematical equations.
In five-fold cross-validation tests, the method matched the predictive performance of leading black-box models while producing equations scientists could interpret. That let the team move beyond predicting behaviour and look at the physical mechanisms behind it.
“By combining reliable experimental data with interpretable AI, we can begin to uncover these underlying physical relationships,” said Hao Li, Distinguished Professor at WPI-AIMR.
What the hydrogen storage database revealed
The map showed an interconnected hierarchy of properties. Lighter host atoms generally favour greater hydrogen storage capacity by weight. In interstitial hydrides, where hydrogen sits in spaces between metal atoms, atomic size and the arrangement of atoms in the crystal lattice matter. Thermal expansion and thermal conductivity also give clues to how a material accommodates hydrogen.
Mechanical properties play a role too. When hydrogen enters a metal lattice, the structure must deform and expand. Stiffer materials resist this, raising the elastic energy needed to take in hydrogen. That can make the hydride less stable and raise its equilibrium pressure. The team also found links between electronegativity, thermal conductivity and resistance to compression, which helps explain why improving capacity can change the conditions needed to release hydrogen.
The researchers conclude that capacity and equilibrium pressure should be considered together when designing new materials, rather than treating each property in isolation. For more on hydrogen research, see our story on Groningen’s hydrogen production material.
The case for and against
The case for this work is that it pairs a large, traceable dataset with equations people can read. That makes it easier to ask why a material behaves as it does, not just what it will do.
The case for caution is that no new material has been made yet. Shin-ichi Orimo of WPI-AIMR said the next step is to test whether predicted properties hold up in experiments. The team plans to expand the database, identify promising compositions and synthesise candidates for testing.
My view on the hydrogen storage database
As a writer, I feel the human checking is the most valuable part. AI can scrape thousands of papers quickly, but the team still verified entries against originals and standardised units. I also like that the AI here explains itself through equations. The real test will be whether these maps lead to better storage materials in the lab.
FAQ: DigHyd hydrogen storage database
What is the DigHyd hydrogen storage database?
A curated database from Tohoku University and the University of Tokyo with more than 30,000 entries from over 4,000 experimental publications on hydrogen storage materials.
What trade-off does it highlight?
The link between hydrogen storage capacity and equilibrium pressure, which sets the conditions for absorbing and releasing hydrogen. Stiffer materials can raise equilibrium pressure.
Where was the study published?
In Applied Physics A, online on 4 September 2026.
1 thought on “Hydrogen Storage Database: DigHyd Maps Capacity vs Pressure”