Carbon nanotube based carbon dioxide gas sensors for respiratory monitoring.

Abstract

The objective of this work is to create a new sensor for monitoring the concentration of exhaled CO2 gas in human breath. Limitations such as high power, large size, lack of portability and undesirable use of sampling tubes are experienced currently during respiratory CO2 monitoring. CO2 being a very important biomarker, it is desirable to extend the scope of CO2 monitoring beyond clinical use to home and ambulatory monitoring. Due to the vast amount of prior research effort put into currently used non-dispersive infra red (NDIR) CO2 sensors, it was deemed unnecessary to further investigate this technique. The sensor development approach in this thesis has been creation of a solid-state CO2 sensor through merging of state-of-the-art research in different disciplines - namely materials science, nanotechnology, chemistry, mechanical engineering and electrical engineering. Early promise for development of such a sensor is shown by use of functionalized carbon nanotube (CNT) materials. Single-walled carbon nanotubes (SWNTs) functionalized with polyethylene imine (PEI) is used as the CO2 sensitive material. A conductivity measurement technique using surface acoustic wave (SAW) sensors enables measurement of SWNT conductivity with very high resolution. While sensitive to CO2, this embodiment is several times more sensitive to humidity in the environment. Since humidity variation happens simultaneously with CO2 variation in exhaled breath, this is found not to be a viable technique for respiratory CO2 measurement. This early failure suggested a need for a sensor that was equally or more sensitive to CO2 than to other environmental analytes. In looking for such an alternative sensor, a CO2 sensor based on stiffness measurement of bare SWNTs was reported to be sensitive and selective to CO2. However, current techniques used for film-stiffness measurement are too bulky, unreliable or expensive. Hence, a new stiffness sensing transducer is developed using an electret microphone. This stiffness measurement technique is based on the extreme sensitivity of an electret microphone's capacitance to the stiffness of its membrane. A CO2 sensor is obtained by coating such a microphone with SWNTs. This embodiment shows good sensitivity to CO2 but unpredictable response to humidity changes. While some microphones show excellent humidity resistance, others show large response to humidity. This behavior is traced to the fabrication of the microphones. Since commercial microphones are used in this work, it is not possible to control manufacturing specifications. Thus, practical difficulties with obtaining a reliable microphone are a major impediment. It was also judged that the sensitivity of stiffness changes to CO2 might be insufficient for respiratory monitoring. The above two sensor embodiments suggest the difficulty in obtaining a selective yet sensitive solid-state CO2 sensor using carbon nanotubes. Hence, an alternative approach is tested using sensitive, selective but slow commercial CO2 sensors. CO2 sensors made using an electrolytic sensing technique are commercially sold for indoor air-quality monitoring. While reliable, such sensors are too slow for respiratory monitoring. But, development of a (second order) mathematical model for the sensor's slow response enables detection of fast CO2 changes during breathing. This is achieved by inverting the mathematical model to predict the fast CO2 input based on the sensor's slow output. The resulting embodiment is the first reliable respiratory CO2 sensor developed in this work. Though better than NDIR sensors, the power requirements and size of electrolytic CO2 sensors are still unacceptable for portable and wireless respiratory CO2 monitoring. Finally, based on research into CNTs and electrolytic CO2 sensors, a new nanocomposite-material based CO2 sensor is fabricated. This sensor combines advantages of high sensitivity and fast response of CNTs with the selectivity of metal carbonates to CO2. The nanocomposite material is fabricated by attaching nanoparticles of calcium carbonate (CaCO3) to SWNTs. CO2 sensing is achieved by measuring the resistance of the SWNT film which changes due to the reaction between CaCO3 and CO2. Cross-sensitivity to humidity, while present, is small enough to be removed using a reference CNT sensor that does not respond to CO2 but responds to humidity. While reliable in operation, this sensor however suffers from slow response due to chemisorption of CO2 on some of the CNTs. Since resistance of the entire nanocomposite can be controlled by a few CNTs, such slow-responding CNTs cause very poor overall response times (>100s). Model inversion techniques developed earlier are not effective with such response times to predict breath-by-breath CO2 changes. In order to enhance the response time, a capacitance based sensor is developed using a similar nanocomposite (SWNT-BaCO3). This sensor's speed of response is found to be much better compared to the previous embodiment which results in the development of a low-power, small, fast and inexpensive CO2 sensor. However, the sensor's capacitance is still found to be sensitive to environmental humidity. Further, the developed nanocomposites are also found to require humidity in the environment for sensing CO2. Thus, the sensor needs constant humidity to respond to CO2 reliably during breath sensing. This is achieved by completely removing humidity from the exhaled breath (using a molecular sieve) before it reaches the sensor. Simultaneously, humid air sampled away from the face is supplied using a low-power pump to humidify the sensing chamber. Using these designs, a reliable respiratory CO2 sensor is fabricated that is compared with a NDIR CO2 analyzer. Results show that the sensor reliably monitors CO2 concentration in the breath. The developed embodiment could potentially be improved with drift-correcting techniques (hardware and software); but is currently unique in its ability to perform low-power, portable and low-cost respiratory CO2 sensing.