Metal-insulator-silicon carbide (MIS) structures have been studied as gas sensors. We have investigated how the sensors detect gases, how fast they do that, and how they could be used for exhaust gas monitoring.

We have prepared simple field-effect devices, MIS-capacitors and Schottky diodes, on silicon carbide with platinum gates. We have tested the gas sensitivity by measuring the electrical characteristics of the devices in oxygen and in hydrogen and sometimes carbon monoxide (representing oxidising and reducing gases). The flat-band voltage of a MIS-capacitor is more positive in oxygen than in hydrogen. Hydrogen or carbon monoxide changes the current-voltage characteristics of a Schottky diode on n-type SiC in a way typical for a lowering of the barrier height. For Schottky diodes on p-type SiC, the apparent barrier height increases in hydrogen. We have determined values of barrier heights for different platinum-insulator combinations, but they differ from each other depending on the measurement method used. We conclude that the thermionic emission model does not describe the MISiC sensors well. In the introductory chapters, a mechanism is discussed how gas exposure might change the electrical characteristics of Schottky diodes and capacitors.

When we expose a MIS capacitor to hydrogen, not only the field effect influences the properties of silicon carbide. At temperatures above 600°C, the inversion capacitance of an n-type MIS capacitor decreases with increasing hydrogen pressure. We have measured and analysed the effectfor 6H and 4H n-type silicon carbide. We have suggested that the influence of hydrogen is due to a decreased minority-carrier generation in the semiconductor. For p-type material, the capacitors apparently do not build up an inversion layer.

We have studied sensor signals also in more complex gas mixtures. We have measured the effects of compounds of synthetic exhaust gas systematically using an experimental design. For each sensor temperature, we could identify one gas compound that influences the sensor signal most. To understand the origin of these effects, we have studied more closely a mixture of oxygen and carbon monoxide. We could explain the sensor behaviour qualitatively with the relative sizes of reaction kinetics and mass transport. For a medium temperature, e.g.400 °C, and high nominal gas concentrations, the oxidation of carbon monoxide is kinetically controlled: The sensor surface is covered with carbon monoxide even if oxygen is in excess, because carbon monoxide has a higher sticking coefficient on platinum. At a high temperature, e.g.500 °C, the oxidation on the platinum is very efficient. Only the gas inexcess is present on the sensor surface and dominates the sensor signal. The possibility is discussed to use an sensor array for the multicomponent analysis in exhaust gases, e.g. for the diagnosis of a catalytic converter. Not alone the sensor temperature determines the selectivity of the sensor but also the presence of reactive surfaces close to the sensor, which change the local gas concentrations.

We have investigated the response times of some MISiC sensors to see if they can be used as fast binary A-sensors for cylinder-specific monitoring of a car engine. We have invented a new method for measuring response times of fast gas sensors in the laboratory, i.e. the moving gas outlets. At temperatures of 650°C and higher, the MISiC sensors react to a change in the ambient in less than three milliseconds. Measurements in the exhaust of a bench engine have confirmed this finding.