Surface State Models For Conductance Response Of Metal Oxide Gas Sensors During Thermal Transients

In the last decades a large amount of research work has been devoted to understanding the sensing mechanism of metal oxide gas conductometric sensors (Wang et al. 1995; Barsan and Weimar 2001, 2003; Korotcenkov 2008; Moseley et al. 2008; Pokhrel et al. 2008; Barsan et al. 2010, 2011; Yamazoe and Shimanoe 2010; Hübner et al. 2011a). Many results have been obtained with special reference to the most often used oxides such as, e.g., SnO2 (Barsan and Weimar 2001; Barsan et al. 2011) and WO3 (Guérin et al. 2006; Bendahan et al. 2007). Indeed, it can be stated that for these materials the sensing principle is understood in its essential features. Nevertheless, even for these materials that have been applied in gas sensors for more than 20 years, and that are the basis of some successful and widespread commercial products, some aspects are still under study, and it can be affi rmed that an exhaustive knowledge of their behavior has not yet been achieved. Due to this lack of knowledge, and to the absence of a reliable and simple sensor input–output relationship, at present, sensor-based measurement systems are mainly designed on the basis of experimental characterization and still suffer from some unsolved problems, such as drift and heavy sensitivity to operating conditions. In this context it is clear how the development of a dynamic model for metal oxide gas sensors would be of the utmost utility from many points of view. First, the availability of a simulation tool, able to predict with suffi cient accuracy the sensor behavior, would enable us to replace experimental tuning with simulations, and could signifi cantly speed up sensor-based system development and guarantee better performance. Moreover, and perhaps most important, comparison of the model outputs with experimental data could greatly help us to understand the behavior of metal oxide sensors, because it would allow us to explore the relevance of the different mechanisms involved in sensing, to validate some commonly accepted assumptions, or to assess their validity ranges. Computational modeling is, in fact, a powerful tool to gain information about the behavior of complex systems which is diffi cult or impossible to obtain by direct measurements and observations. Sensor models have to incorporate all the phenomena that contribute to the sensing mechanism and these, in the case of metal oxide conductometric sensors, are very complex in that they comprise chemical solid–gas reactions and physical phenomena related to electronic conduction. Especially the fi rst are very diffi cult to observe and often remain the subject of hypotheses that are diffi cult to assess with independent measurements. It must be underlined, also, that the chemical-physical behavior of the sensor depends, of course, on the material—its composition as well as bulk and surface structure—but also heavily on the crystal bulk and surface defect population, and fi nally on the sensor micro- and macrostructure. All these aspects are crucial to the electronic conduction in the sensors (Korotcenkov et al. 2007), and should be taken into account when building a computational model. In detail, a sensor model has to describe, fi rst, the interaction of the target gases with the sensing material, resulting from both the surface chemical reactions and the possible gas diffusion in it. It must be stressed that, for resistive sensors, only reactions involving electron exchange between the sensing material and the gas can be sensed, even if other possible reactions (absorption, adsorption or desorption), though not sensed directly, affect the sensor response. Note that the sensor response also depends on the interactions between different gases (target gases and other gases) that take place on the surface. Among important interfering gases, the effect of water vapor has to be taken into account; in fact, water vapor is always present in normal environments in very large concentrations (tens of thousands of parts per million) and heavily affects the response of binary oxides (Hübner et al. 2001b; Henderson 2002; Korotcenkov et al. 2007; Gaman et al. 2008). The reaction of water with oxide surfaces has long been studied but is not yet completely understood. The second contribution to the sensor response is the mechanism that causes a variation of the sensor resistance in relationship to the quantity of adsorbed gas. Different mechanisms can occur in different fi lm microstructures (Wang et al. 1995; Brynzari et al. 1999; Korotcenkov 2007, 2008; Rettig and Moos 2008; Yamazoe and Shimanoe 2008a; Shaalan et al. 2011). In particular, for many metal oxides, used in the temperature range 200–400°C, the electronic conduction, and hence the resistance value, is mainly determined by the surface chemical reactions, which are responsible for the creation of some “extrinsic” surface energy states. The presence of adsorbates affects the resistance in a way that depends on the metal oxide fi lm microstructure. For poorly sintered porous fi lms, for instance, the most relevant phenomenon responsible for the sensor response is the surface electric fi eld, which is established in the surface region where free carriers are trapped by the adsorbed molecules. In this chapter, some research carried out in the fi eld of simulation based on surface chemical reactions of the dynamic response of resistive metal oxide sensors is summarized, and the different points of view and assumptions are discussed. The models presented in the literature have proven to well describe some devices under particular operating conditions, but different points of view are often chosen by the researchers. Particular attention will be devoted in this work to SnO2 sensors interacting with mixtures of oxygen (O2) and carbon monoxide (CO), and a possible approach followed by the authors of this review for modeling their behavior will be briefl y described. This approach leads to the development of a gray-box model that, starting from a physical-chemical description of the surface reactions, provides a compact description of the sensor behavior in dry O2, and in the presence of both O2 and CO (Fort et al. 2006a, 2006b, 2006c, 2007, 2010; Bicelli et al. 2009) The effect of water vapor is also accounted for (Fort et al. 2011a). The model explains the sensor dynamics by means of oxygen adsorption and ionization processes, and of CO direct adsorption and reaction with the ionized adsorbed oxygen, as suggested also by many other researchers. The model applies to large-grained thick-fi lm sensors or to nanowires (quasi-onedimensional) bundles. All the developed models take into account sensor operation under dynamic thermal conditions.