Hydrogen could play an important role in cleaner energy systems, but even small leaks can create safety risks. Reliable sensors are therefore essential wherever hydrogen is produced, stored, transported, or used. A persistent problem is humidity: water molecules can cover the active surfaces of hydrogen detectors and prevent them from responding accurately, particularly in hot, moisture-rich environments.

An international research team involving Tohoku University's Advanced Institute for Materials Research (WPI-AIMR), the University of Utah, and Pacific Northwest National Laboratory has developed a catalytic hydrogen sensor that performs reliably in such conditions without an external heater. Reported in Advanced Materials, the device uses a network of palladium nanoparticles with abundant grain boundaries--tiny interfaces where neighboring crystals meet. These boundaries turn out to be the key to resisting moisture interference.

The sensor detects hydrogen across ambient temperatures from 22 to 80 °C and relative humidity levels from 0% to 90%. Remarkably, it can detect hydrogen concentrations as low as 100 parts per million at 80 °C and 90% relative humidity, when water vapor is present in overwhelming excess. It also shows strong selectivity against 21 potentially interfering gases, reproducible performance over 100 consecutive sensing cycles, and stable operation in tests spanning 125 days. The device satisfies relevant ISO 26142 humidity-response requirements under demanding conditions.

To explain these results, the researchers combined experiments that track structural changes and water desorption with molecular dynamics and first-principles calculations. "We found that grain boundaries help hydrogen enter palladium and rapidly form palladium hydride, even when the surface is wet," says Hao Li, a Distinguished Professor at WPI-AIMR. This hydride binds water weakly, leaving catalytic sites more accessible to the hydrogen-oxygen reaction that generates the measurable sensing signal. The interplay between crystal boundaries, hydride formation, and water adsorption explains why the material remains responsive when conventional catalysts struggle.

The team also incorporated a machine-learning model into a portable sensor prototype. By accounting for changing temperature and humidity, the system estimated hydrogen concentrations with deviations within 5% in prototype tests under randomized conditions. Beyond demonstrating a more robust sensor, the work points to grain-boundary engineering as a promising strategy for safer hydrogen monitoring in industrial facilities and future hydrogen-energy infrastructure.

Details of the study were published in the journal Advanced Materials on October 23, 2026.