Metal oxide semiconductor (MOS) gas sensors have been developed as promising gas sensing platforms owing to their compact size, low cost, rapid kinetics and high versatility. Conventional MOS gas sensors, however, are hindered by significant constrain...
Metal oxide semiconductor (MOS) gas sensors have been developed as promising gas sensing platforms owing to their compact size, low cost, rapid kinetics and high versatility. Conventional MOS gas sensors, however, are hindered by significant constraints, including limited selectivity, low sensitivity at trace-level concentrations, and high power consumption due to elevated working temperatures. To address these challenges, this dissertation presents comprehensive strategies ranging from rational material design to intelligent data analysis, with a primary focus on tungsten oxide-based nanostructures. Tungsten oxide is investigated as a representative MOS platform due to its diverse polymorphism, tunable defect chemistry, and distinctive surface reactivity that facilitates tailored gas interactions.
In the first part of this study, Pd-functionalized W18O49 nanowires were developed to attain multimodal selectivity for hydrogen and ammonia. Pristine W18O49 exhibits defect-rich chemistry and surface acidity, while the decoration of Pd nanoparticles markedly improves the sensing characteristics via the catalytic spillover effect. The sensor exhibited temperature-dependent gas selectivity, allowing for the discrimination of two separate gases using a single sensor. To further demonstrate its practical utility, the developed sensing material was integrated into a MEMS platform to reduce power consumption and validate the feasibility of the sensor for versatile applications, ranging from energy production safety to environmental monitoring.
In the second part, a highly reliable dual-signal hydrogen sensor was realized by depositing Pd-loaded WO3 nanorods onto a MEMS platform equipped with a microheater functioning as a resistance temperature detector (RTD). Addressing the issue of false alarms in traditional single-output sensors, this device generates two simultaneous but distinct signals when exposed to hydrogen—a catalytic combustion response from the microheater and a chemoresistive response from the sensing layer. This dual-mode operation provides a robust self-validation mechanism, ensuring high reliability even in complex environments.
In the third section, Ta-doped WO3 nanourchins were synthesized for high-performance breath acetone sensing, aimed at non-invasive diabetes diagnosis. The Ta doping stabilized the ɛ-phase with a non-centrosymmetric structure and facilitated strong dipole-dipole interactions with polar acetone molecules, leading to remarkable selectivity. In addition, a machine learning-assisted method was employed to analyze impulsive breath signals. A dual-output deep learning model (Bi-LSTM and MLP) successfully decoupled complex transient patterns, facilitating concurrent gas classification and concentration estimation. Finally, the integration of this sensor into a wireless portable breath analyzer demonstrated the practical viability of the study for daily personal healthcare monitoring.
In conclusion, this dissertation presents a multi-level approach to address the limitations of MOS gas sensors. From a material perspective, strategies such as defect chemistry engineering, catalytic functionalization, and crystal phase control were employed to optimize sensing properties. Complementarily, at the system level, the adoption of MEMS platforms and machine learning algorithms enabled low-power operation and accurate signal processing. Integrating these material- and device-level strategies provides a practical pathway for developing robust sensing systems suitable for next-generation advanced gas sensors.