Flexible Nanostructured Hydrogen Sensors

Abstract

Hydrogen is rapidly emerging as a key energy carrier and industrial feedstock in the global transition to cleaner energy systems. However, due to its high diffusivity and flammability, even minor leaks can result in significant hazards, including explosions. Hydrogen leaks are particularly dangerous because hydrogen is odourless, colourless, and has a wide flammability range in air (4–75 vol%). This necessitates the development of reliable, rapid-response hydrogen sensors that can detect small leaks before concentrations approach dangerous levels. Among the many hydrogen sensing technologies developed to date, chemiresistive sensors are particularly promising due to their relatively simple design, low power requirements, and ability to function at ambient conditions. These sensors operate based on changes in electrical resistance upon exposure to hydrogen. However, challenges remain: traditional chemiresistive sensors often suffer from low sensitivity at trace hydrogen levels, poor long-term stability, and inconsistent reproducibility across sensor batches. This research seeks to address these limitations through a novel fabrication approach aimed at developing high-performance, cost-effective, and mechanically flexible hydrogen sensors suitable for real-world deployment. The work focuses on palladium (Pd)-based nanoparticle sensors fabricated on flexible polyimide substrates. Palladium is well-known for its superior hydrogen selectivity and capacity to operate at room temperature. In its nanostructured form, Pd offers enhanced surface area and faster response times. The central challenge addressed in this thesis is the development of a scalable, low-complexity fabrication method to deposit Pd nanoparticles effectively, without the need for sophisticated equipment or high-temperature processing. Initial experiments explored the use of copper (Cu) as a base metal for galvanic replacement reactions with PdCl2 solutions. This reaction led to the spontaneous formation of bimetallic Cu-Pd deposits via a diffusion-limited aggregation mechanism. These outgrowths were found to be electrically conductive and sensitive to hydrogen, demonstrating a measurable decrease in resistance upon hydrogen exposure. However, a significant limitation of these early sensors was their lack of recovery after hydrogen removal, likely due to the mixed composition of Cu and Pd in the deposits. To improve recovery behaviour, efforts were made to increase the Pd content in the Cu-Pd structures, including optimizing precursor concentrations and geometries (e.g., trench width between Cu contacts). Despite these modifications, the sensors still failed to reliably regain their baseline resistance, prompting a shift in strategy. The focus then turned to aluminium as the base metal, based on promising results from late-stage experiments. Unlike copper, aluminium forms a stable oxide layer that impedes direct galvanic exchange with Pd²⁺. However, the acidic PdCl2 solution (which contains HCl) promotes pitting corrosion, allowing protons to penetrate the oxide and reach the underlying aluminium. Here, hydrogen atoms are generated and act as reducing agents for Pd2+ ions at the aluminium-liquid interface, leading to the formation of Pd nanoparticles. This reaction is accompanied by the generation of hydrogen gas bubbles, which rupture the oxide layer intermittently and eject Pd nuclei into the surrounding solution. These nanoparticles, now mobile, deposit onto nearby surfaces such as the polyimide substrate. It was observed that Pd (and other noble metals like Pt and Rh) can adsorb hydrogen from these bubbles, facilitating further reduction of Pd2+ in solution. This autocatalytic process leads to dense, electrically conductive Pd nanoparticle films, often with minimal contamination from the underlying aluminium. By strategically positioning the polyimide above the aluminium foil (termed the "inverted assembly" approach), deposition efficiency was maximized, as hydrogen bubbles carrying Pd nuclei rise upwards due to buoyancy. This orientation resulted in more consistent deposition with a higher yield. Further improvements were made by tuning the precursor solution. Increasing the HCl content and using thicker aluminium tape led to more vigorous hydrogen evolution, improving deposition density. However, the increased reaction extent meant that essential precursors like potassium bromide (KBr)—added for oxidation resistance—were consumed entirely during deposition. To counteract this, a post-treatment step was introduced where fresh precursor solution containing KBr was dropped on to the deposited film for a brief period. This step significantly improved oxidation resistance by passivating the surface and removing residual oxides, enabling better long-term performance in air. To enhance selectivity, a ZIF-8 (Zeolitic Imidazolate Framework-8) coating was put on some sensors. ZIF-8 possesses six-membered rings just large enough to allow hydrogen molecules to pass through, while filtering out larger interferents like O2 and CH4. The coating process was straightforward: immersing the polyimide-supported Pd nanoparticles into a ZIF-8 precursor solution resulted in spontaneous, uniform coating via heterogeneous nucleation. Sensors with ZIF-8 exhibited faster response times and better reproducibility across repeated tests, although their sensitivity was slightly reduced compared to uncoated samples. Regardless of the sensor configuration, three response characteristics were consistently observed: a drifting baseline, an undershoot at low H2 concentrations, and an overshoot at all concentrations. These features were linked to background oxygen levels and were modelled using a dual-zone surface reaction framework. The model assumed separate surface zones for hydrogen adsorption (leading to hydride formation) and oxygen-hydrogen reaction pathways, with the rate-limiting step being the formation of hydroxyl species. While the model captured qualitative trends well, it lacked the fidelity to predict sample-to-sample variations in quantitative behaviour. Finally, the sensors were tested under various mechanical and environmental stressors—including strain, aging, and exposure to potential interferent gases. The ZIF-8-coated sensors demonstrated more consistent performance, validating the protective and selective role of the MOF coating. These experiments highlighted the robustness of the fabrication process and the potential of the sensors for deployment in real-world hydrogen leak detection systems. In conclusion, this research presents a novel, low-cost, and scalable method for fabricating flexible Pd-based chemiresistive hydrogen sensors. By combining aluminium-aided nanoparticle deposition, ZIF-8 filtering, and post-treatment protocols, the developed sensors achieve fast response, reliable recovery, and good resistance to oxidation—all under ambient conditions. The insights gained here not only improve current hydrogen sensing technologies but also lay the groundwork for further innovations in nanoparticle-based sensor fabrication.