Optical Hydrogen Sensor Built on Planar Lightwave Circuit Detects Leaks Without Electricity

Hydrogen is colorless, odorless and explosive at concentrations ranging from 4 to 75 percent in air, yet the clean energy systems of the future depend on storing and moving it through pipelines, tanks and fuel cells. In a study published in Discover Sensors, Erika Ishii, Keonhoe Kim and colleagues describe a hydrogen sensor built on a planar lightwave circuit, a flat silica waveguide platform borrowed from optical communications, coated with a hydrogen-sensitive film of platinum-loaded tungsten oxide. Because the sensing region requires no electrical energization or heating, the design avoids an ignition source sitting directly inside a potentially explosive gas stream.

The device itself is elegantly simple in concept. The researchers fabricated a planar lightwave circuit on a silicon substrate, consisting of a 15-micrometer-thick silicon dioxide lower cladding and a 6-micrometer-thick, 6-micrometer-wide core of germanium-doped silica deposited by flame hydrolysis. Using photolithography and reactive ion etching, they exposed a section of the core roughly 21 centimeters long, then immobilized a hydrogen-sensitive film on a small portion of it covering about 3 to 5 square millimeters. When light from a 1300-nanometer LED travels through the core, a small evanescent field extends into the cladding region and interacts with whatever film is deposited there. If that film reacts with hydrogen, its optical properties change, and the amount of light reaching a detector at the far end drops in proportion.

The chemistry driving the signal is a well-known one. Platinum catalyzes the dissociation of hydrogen molecules into adsorbed hydrogen atoms, which then insert into tungsten oxide to form hydrogen tungsten bronze, reducing tungsten from the six-plus to the five-plus state. That matters optically because tungsten five-plus absorbs near-infrared light far more strongly than tungsten six-plus. When hydrogen flows over the film, transmission through the waveguide plunges; when air returns and reoxidizes the film, the signal recovers. The team quantified performance using sensitivity, defined as the percentage drop in transmitted light, along with response time, recovery time and baseline drift across repeated exposure cycles of 180 seconds in hydrogen followed by 540 seconds in air, all at 22 plus or minus 2 degrees Celsius.

What distinguishes this study is its systematic comparison of three ways of laying down the sensing film from a sol-gel precursor solution. In the deposition of controlled volume method, a precise droplet was pipetted onto the substrate and left to dry naturally. In the drip-and-absorb method, a droplet was applied and the excess gently soaked up with a cotton swab. In the spray method, the solution was misted through a mask onto a defined area. Each device was then sintered on a hot plate at 300 degrees Celsius, since the attached optical fiber cables could not tolerate conventional furnace calcination. Because strong reflections from the silicon substrate interfered with direct measurement, film thicknesses were characterized on fused silica reference substrates using laser microscopy and white light interferometry.

The deposition-of-controlled-volume films revealed a fundamental trade-off. The thickest film, about 292 nanometers at a tungsten oxide density of 5 nanomoles per square millimeter, delivered the highest sensitivity at 98 percent, with the cracks running through its center apparently enlarging the reactive surface area and aiding gas diffusion. But it also showed severe baseline drift of roughly 74 percent after ten exposures, because tungsten five-plus species deep inside the film remained shielded from oxygen and accumulated instead of reverting to the original oxide. Thinner films of 152 and 113 nanometers responded faster, at 352 and 334 seconds respectively, and drifted far less, only 30 and 15 percent after ten cycles, illustrating how thickness governs both sensitivity and speed.

The drip-and-absorb approach fared worst. Despite expectations that swabbing away excess solution would leave a thin layer, the resulting film measured 264 nanometers thick, smooth and crack-free but densely packed. Hydrogen diffusion into such a structure is hindered, and the sensor managed only about 22 percent sensitivity with a serious 40 percent baseline change after just six exposures, pointing to poor repeatability. The authors attribute the failure to irreversible accumulation of reduced tungsten species within the compact film interior, a reminder that in hydrogen sensing, morphology can matter as much as chemistry.

The spray method produced the winners. A crack-free, uniform, roughly 94-nanometer film with a tungsten oxide density near 3 nanomoles per square millimeter achieved 51 percent sensitivity with a 131-second response and 81-second recovery, and drifted only 15 percent after ten exposures. Its slightly rougher surface, about 1.4 times rougher than the drip-and-absorb films, appears to have increased the reaction area and enhanced diffusion. Even more striking was a thicker sprayed film of about twice that thickness at a density of 4 nanomoles per square millimeter: it reached 99 percent sensitivity with a 106-second response on the first cycle, and after initial drift it stabilized dramatically. From the twenty-first to thirtieth exposure, response and recovery times averaged 15.3 plus or minus 2.1 seconds and 184 plus or minus 13 seconds, and the response was relatively consistent over 30 repeated exposures overall.

That same sprayed film defined the sensor’s practical detection capability. When hydrogen in air was introduced at increasing concentrations, almost no response appeared at 0.1 volume percent, but a clear signal emerged at 0.5 volume percent. A formal signal-to-noise calculation suggested a limit near 0.04 volume percent, but the authors judged this overly optimistic because linearity below 0.5 percent could not be confirmed and baseline drift obscured weak signals. They therefore report 0.5 volume percent as the experimentally demonstrated detection limit. That figure sits comfortably below the roughly 0.8 volume percent threshold, corresponding to 20 percent of hydrogen’s 4 percent lower explosive limit, that industrial safety practice generally treats as sufficient for reliable leak detection, although the authors note that conventional electrochemical and metal-oxide-semiconductor sensors still detect far lower concentrations, below 100 parts per million.

The fastest single response figure reported in the abstract, a 90 percent reduction in light propagation within approximately 110 seconds after exposure to 4 volume percent hydrogen in nitrogen, captures the device’s headline performance, and the authors emphasize that near the lower explosive limit itself the response plateaued yet recovered immediately and completely on return to air. The planar platform brings additional advantages over fiber-based optical hydrogen sensors, which can suffer from limited interaction length, mechanical fragility and bending sensitivity. A lithographically defined exposed core offers a long evanescent interaction length, excellent mechanical stability and compatibility with chemosensitive coatings, and silica glass provides broadband transmission, chemical and heat resistance.

The researchers are careful about what they have and have not shown. Each coating condition was tested on a single device, so multi-device statistical reproducibility remains future work, and distributed or multipoint leak detection along optical fibers, though a natural extension enabled by techniques like optical time domain reflectometry, was not demonstrated in this study. Long-term degradation, and the effects of temperature, humidity and cross-sensitivity to other gases, also await systematic investigation, as does the role of platinum particle distribution in hydrogen dissociation. The authors candidly state that current performance, with its relatively slow response kinetics and comparatively high detection limit, remains below the level required for practical applications. Even so, the demonstration establishes the feasibility of pairing platinum-loaded tungsten oxide films with exposed-core planar lightwave circuits, and suggests that with further optimization of catalyst materials and immobilization techniques, passive, intrinsically safe optical monitoring of hydrogen infrastructure could move from laboratory concept toward real-world deployment.

Subject of Research: An optical hydrogen gas sensor based on a planar lightwave circuit with a Pt/WO3 sensing film

Article Title: Hydrogen gas sensor using planar lightwave circuit

Article References: Ishii, E., Kim, K., Iwabuchi, A., Okazaki, S., Nishijima, Y., Arakawa, T., Usukura, T., Sato, N., Matsubara, N., & Ota, I. (2026). Hydrogen gas sensor using planar lightwave circuit. Discover Sensors, 2(1), Article 44. https://doi.org/10.1007/s44397-026-00078-6

Image Credits: AI Generated

DOI: 10.1007/s44397-026-00078-6

Keywords: hydrogen sensor, planar lightwave circuit, tungsten oxide, platinum catalyst, evanescent wave, sol-gel film, leak detection, optical sensing, hydrogen safety, spray coating, Hydrogen, sensor