Respiration monitoring is crucial for detecting various health issues. Humidity sensing offers a promising approach for this purpose, but conventional sensors suffer from limitations like low sensitivity, slow response times, temperature drift, and parasitic capacitance. While various sensor types exist (resistive, optical, capacitive), capacitive sensors are favored due to their accuracy and integration convenience. However, capacitive sensors face challenges stemming from the availability of high-quality humidity-sensitive materials, parasitic capacitance from the substrate, and water condensation leading to slow recovery times. The use of materials like polyimide (PI) offers some advantages, but its limited capacitance output requires high-precision circuitry. Porous silicon and ceramics offer better sensitivity but have complex fabrication processes. Parasitic capacitance from the substrate and temperature-induced dielectric constant changes in the humidity-sensitive material further compromise accuracy. This paper proposes a novel sensor design to address these limitations.

The literature review section explores the existing challenges in humidity sensing for respiration monitoring. It highlights the limitations of traditional capacitive humidity sensors, focusing on issues like low sensitivity (particularly with materials such as polyimide), parasitic capacitance from the substrate (SiO2 and Si), and slow recovery times due to water condensation. It contrasts different sensor types (resistive, optical, capacitive), emphasizing the advantages of the capacitive approach for respiration applications while acknowledging existing material-related and design-related limitations. The review sets the stage for the proposed triple-layer design as a solution to these problems.

The study presents a novel triple-layer humidity sensor design. This sensor integrates a nanoforest-based sensing capacitor, a reference capacitor, a microheater, and a thermistor. The nanoforests provide high hygroscopicity due to their fully open nanostructures, leading to enhanced sensitivity. The reference capacitor helps eliminate parasitic capacitance, improving measurement accuracy. The microheater accelerates water molecule desorption, shortening recovery time. The thermistor enables temperature compensation. The sensor's theoretical capacitance model is derived, accounting for the individual capacitances and parasitic elements. The fabrication process is detailed, including steps like SiO2 deposition, aluminum sputtering for the microheater and thermistor, additional SiO2 insulation layer deposition, IDE formation, Si3N4 deposition for the reference capacitor, PI spin coating, and reactive ion etching to create the nanoforests. Four samples with varying numbers of interdigital electrodes (IDEs) and one control PI-based sample were prepared for testing. SEM images are provided to illustrate the sensor's structure. The experimental setup involved measuring capacitance changes under controlled humidity conditions (10-90% RH) and temperatures. The data were analyzed to determine sensitivity, response time, recovery time, repeatability, temperature dependence, and long-term stability.

The nanoforest-based sensor demonstrated significantly higher sensitivity (0.11 pF/%RH) compared to the PI-based control sensor (0.014 pF/%RH) within the 40-90% RH range—an approximately 8-fold improvement. The sensor exhibited a fast recovery time of 5 seconds. The reference capacitor effectively compensated for parasitic capacitance, maintaining stable performance across various humidity levels. The sensor demonstrated excellent repeatability and stability under fluctuating humidity conditions. Temperature analysis revealed a positive temperature coefficient, which is consistent with typical capacitive humidity sensors. Long-term stability testing showed consistent performance over time. When integrated into an N95 mask for respiration monitoring, the sensor, in conjunction with machine learning algorithms, achieved an accuracy of 94% in distinguishing different respiratory states. The resolution of the sensor was 0.72% RH in the 10-40% RH range and 0.075% RH in the 40-90% RH range. The dimensions of the humidity sensor are 1100 µm × 890 µm, and the areas of the nanoforest and the Si3N4 layer are exactly the same (430 µm × 820 µm).

The results demonstrate the success of the proposed triple-layer design in overcoming the limitations of conventional humidity sensors. The significantly enhanced sensitivity, rapid response and recovery times, and effective temperature compensation represent substantial improvements. The high accuracy achieved in respiration state classification validates the sensor's potential for practical applications. The use of nanoforests as the humidity-sensitive material proves to be highly effective in increasing sensitivity. The integration of the reference capacitor successfully addressed the issue of parasitic capacitance. The combination of the sensor's improved performance and machine learning-based data analysis opens opportunities for advanced respiration monitoring systems.

This study successfully developed a high-performance nanoforest-based humidity sensor for respiration monitoring, addressing key limitations of traditional sensors. The sensor's enhanced sensitivity, fast response time, temperature compensation, and accurate respiration state classification demonstrate its significant potential for both consumer electronics and medical applications. Future work could focus on exploring different nanomaterials for further sensitivity improvements, miniaturizing the device for improved wearability, and integrating the sensor into more sophisticated diagnostic systems.

While the sensor showed significant improvements, some limitations exist. The current study focused on a specific humidity range (40-90% RH) for respiration monitoring; further investigation is needed to evaluate performance across a broader range. The long-term stability was evaluated over a limited timeframe; longer-term studies are recommended to confirm its durability. The integration with an N95 mask was a proof-of-concept; further optimization might be necessary for integration into various wearable devices.