A new database brings more than 30,000 hydrogen-storage entries together

The Digital Hydrogen Platform combines data from over 4,000 experimental sources to help compare solid materials for storing hydrogen.

Editorial illustration — not from the study.

The Digital Hydrogen Platform, or DigHyd, was built using AI-assisted literature mining followed by human checking. It records how much hydrogen materials can store, as well as enthalpy and entropy changes linked to hydrogen absorption reactions. These properties can be used to estimate the pressure at which hydrogen storage materials reach equilibrium at a given temperature.

The researchers also used a selected set of mostly single-phase materials to test composition-based symbolic regression, a form of modeling that produces relatively simple mathematical relationships. The models performed comparably to more complex black-box models while identifying factors associated with storage capacity and room-temperature equilibrium pressure.

What the database contains

DigHyd contains more than 4,000 experimental literature sources and more than 30,000 data entries on hydrogen-storage materials. Alongside gravimetric hydrogen storage density, the database includes enthalpy and entropy changes derived by manually analyzing pressure-composition-temperature measurements at multiple temperatures using van’t Hoff analysis.

Statistical analysis found different patterns in these thermodynamic properties across material classes, along with wide variation in composition within representative hydride systems. In a demonstration application, symbolic-regression models produced compact relationships for storage density, room-temperature equilibrium pressure, enthalpy and entropy. The descriptor analysis pointed to recurring roles for host atomic mass, lattice geometry, elastic stiffness, metal filling factor and electronegativity-based correlations in the trade-off between storage density and equilibrium pressure.

Why the database matters

Hydrogen-storage research has lacked large, well-curated datasets, making systematic comparison and data-driven discovery more difficult. By recording thermodynamic properties rather than only equilibrium pressure at one fixed temperature, DigHyd can support evaluations tailored to different operating conditions. Its interpretable models may also help researchers examine links between material composition and storage behavior without relying only on opaque predictions.

Evidence and limits

This is a database and modeling study, not a new experimental test of a hydrogen-storage material. The data were collected from published experiments through AI-assisted literature mining and human validation, with thermodynamic values obtained by manual analysis of multi-temperature measurements. The modeling demonstration used a curated subset consisting of single-phase or near-single-phase materials, so its results do not necessarily represent every entry in the database. The abstract reports comparable performance with black-box models but does not give numerical performance measures or describe independent experimental validation of the model predictions.

Sediments show a Mexican lake recovering after decades of deforestation

Sediments and recent water samples show that Lago Verde underwent a major ecological shift during peak deforestation and erosion in the 1970s and 1980s. As forest cover partly recovered, disturbance-tolerant diatoms declined and some earlier planktonic species increased, but the lake has not fully returned to its former state.

15 hours ago·27 views

Materials & Energy

An AI system builds an open concrete database from 27,000 papers

Researchers developed an AI-powered pipeline that extracts information about concrete materials from scientific papers and organizes it into a database. Applied to a corpus screened from more than 27,000 publications, the system produced nearly 9,000 records for an open database on blended cement concrete.

3 days ago·43 views

Materials & Energy

Blinking nanocrystals help pinpoint particles to 1.2 nanometers

Researchers report that individual lanthanide-doped nanocrystals can blink collectively, even though each contains thousands of light-emitting ions. The long-lasting, high-contrast blinking enabled low-power microscopy that localized particles with a mean precision of 1.2 nanometers.

4 days ago·22 views