A conventional approach to sensing is to choose a temperature a priori at which the sensitivity of the film to a target concentration of a given gas is maximized. This allows a very simple temperature control scheme, but does not address the issue of selectivity. Microhotplates are sensor platforms with very low thermal mass that allow for rapid temperature control (10^6 °C/s). Therefore, high-resolution TPS spectra spanning a range on the order of 500°C can be collected many times per minute. Thus, obtaining much more information about the gas environment.
When one looks at the underlying chemical principles that drive the metal oxide thin film gas sensor, significant insight can be obtained. Fundamentally, within the metal oxide film, contiguous grains exist, forming a conducting path across the film. As the grains interact with the ambient environment at elevated temperatures oxidation and/or reduction reactions occur that change the surface oxidation state of the film. For an n-type semiconductor such as SnO2, the electrical resistance increases when the film is highly oxidized because of a depletion of electron charge carriers near the surface of the film. Conversely, if oxygen is removed from the surface by reaction with reducing species, the electrical conductivity of the film increases.
A computer model was developed for a metal oxide sensing film to simulate the steady-state coverage of oxygen as a function of surface temperature and gas concentration. A set of differential equations using Langmuir-Hinshelwood mechanisms was developed for a general reducing reaction such as CO oxidation, with gas concentration (flux), activation energies, sticking probabilities, and temperature as variables. For the initial modeling, two fictitious gases were chosen whose activation energies were similar, differing by approximately 15%, to determine whether it is feasible to classify and quantify such gases with a single microhotplate. Families of curves were generated for the steady-state oxygen coverage as a function of film temperature, gas present (i.e., activation energy), and concentration (in parts per million). These were generated for the two different (but similar) gases, and compared. Several significant observations were made upon the examination of the curves. Notably, at low temperature, the oxygen coverage of the film is low, regardless of the type or concentration of target gas present. Conversely, at high temperature, the oxygen coverage is high, independent of the type or concentration of gas present. Hence, even in the presence of gases, it is possible to establish two baselines for a microhotplate film, a lower resistance, (at low temperature) that corresponds to low oxygen coverage, and a high resistance, (at high temperature) corresponding to high oxygen coverage. It is of key importance that these endpoints can be established, even in the presence of target gases of unknown concentrations. A further observation is that the curves in the family are "S" shaped and change from low resistance to high resistance as temperature increases. Although the curves are "S" shaped, the temperature range where the resistance transitions from the low value to the high value is relatively small and is strongly a function of gas species and concentration. The shape and characteristics of the curves make it possible to define an entirely new methodology for quantifying the response of a sensing film. Instead of having a response be the resistance of the film at a given temperature (or the resistances at a set of temperatures), the response is the TEMPERATURE at which the resistance is centered between the upper and lower saturation values. This temperature is strongly a function of gas concentration. Furthermore, the rate of change of the resistance in the vicinity of this temperature is also a function of the activation energy. Hence it is possible with a microhotplate to obtain readings from a film regardless of baseline resistance (or even in baseline resistance changes), to obtain multiple dimensions of data (in our case two, but this could easily be extended), and to obtain a broader dynamic range than is possible with a fixed temperature system.
A computer on a chip is often referred to as a microcontroller. These devices provide the ability to control and monitor an electrical system using a high level programming language, such as C. Although the processing power remains far below that of a desktop PC, it is sufficient to control the temperature of a microhotplate and monitor the resistance of the metal oxide film. Microcontrollers have the added benefits of being small in size, and of consuming little power. This allows for a hand-held, battery powered, mobile sensing instrument. Although microcontrollers have a number of advantages, including small size and low power, their processing power and user interface abilities are similarly small. To provide a processing system that can classify and quantify the gas species present, using artificial neural networks and other soft computing techniques one needs more processing horsepower. We have created a system that combines the benefits of a low power, portable, hand-held, low cost device with the sophistication and computational power of a desktop computer. This system has proven the viability of the "midpoint temperature" method of data analysis. The system consists of three parts. One part simulates the microhotplate itself. Through a set of user selectable inputs, a target gas and concentration is determined. This device acts as a microhotplate with a metal oxide film in the sense that a microhotplate controller commands it to go to a temperature and it responds with a resistance value (with added noise). The controller follows a calibration sequence where it samples the resistance at a low temperature, samples at a high temperature, and from those two resistances, calculates the midpoint resistance value. It then performs a binary search to find the temperature at which this resistance is achieved. Once the resistance is found, the slope of resistance with temperature is estimated by varying the temperature 10°C above and below the midpoint temperature and calculating the change in resistance. The temperature and slope values are communicated wirelessly to a PC using a Bluetooth serial connection (100m range) and processed using nearest neighbor classification functionality from the Neural Network Toolbox in Matlab.
It has been found that this approach works well even in the presence of random noise, allowing the accurate classification and quantification of two similar gases (activation energies differing by approximately 15%) in the range from 1 to 100 ppm. This system represents a novel, demonstrated, method by which microhotplates can be used in a portable, hand-held, battery powered device. Furthermore, the effects of baseline drift, film aging, and noise (often of significant concern in sensor systems) can be addressed in a simple and effective manner. Moving the operating temperature to the region of maximum sensitivity can increase dynamic range of the film. Sophisticated processing techniques, archival storage, networking, user interface, and alarming can be accomplished with the processing power of a desktop PC, without requiring the instrument to be tethered to the PC.
Chwieroth, B., Patton, B. R., Wang, Y. (2000) "Conduction and Gas-Surface Reaction Modeling in Metal Oxide Gas Sensors" Journal of Electroceramics 6:1, 27-41, 2001
Ivanov, P., Stankova, M., Llobet, E., Vilanova, X., Brezmes, J., Gracia, I., Cane, C., Calderer, J., Correig, X. (2005) "Nanoparticle Metal-Oxide Films for Micro-Hotplate-Based Gas Sensor Systems" IEEE Sensors Journal, Vol. 5, NO. 5, October 2005
Meier, D. C., Evju, J. K., Boger, B., Raman, B., Benkstein, K. D., Martinez, C. J., Montgomery, C. B., Semancik, S. (2006) "The potential for and challenges of detecting chemical hazards with temperature-programmed microsensors" Sensors and Actuators B 232 (2007) 282-294
Meier, D. C., Taylor, C. J., Cavicchi, R. E., White E. V, Ellzy, M. W., Sumpter, K. B., Semancik, S. (2005) "Chemical Warfare Agent Detection Using MEMS-Compatible Microsensor Arrays" IEEE Sensors Journal, Vol. 5, NO. 4, August 2005
Pilling, R. S., Bernhardt, G., Kim, C. S., Duncan, J., Crothers, C. B. H., Kleinschmidt, D., Frankel, D. J., Lad, R. J., Frederick, B. G. (2003) "Quantifying gas sensor and delivery system response time using GC/MS" Sensors and Actuator B 96 (2003) 200-214