The sensor design can be seen in Figure 3. All bold highlighted terms are explained in the following. A transducer ① was made in High-Temperature-Cofired-Ceramic technology (HTCC). Two alumina tapes (Keral 99, Kerafol GmbH, 99% Al₂O₃) were laminated with 15 MPa at 70 °C for 10 minutes. The “ring” in the transducer (see Figure 3b and Figure 4a) was cut by a 355 nm Nd:YAG laser (arc lamp pumped with Q-switching in a LPKF Microline 350L system). After lamination and sintering at 1560 °C, 2 μm platinum were sputtered on one side of the alumina ceramic. Later, the electrodes for the two-sensor-setup in the form of interdigital electrodes (IDEs) with 20 μm line and space were laser-patterned, using laser parameters as explained in [17]. After a subsequent thermal treatment at 1000 °C to restore laser damages, the transducer ① was completed.

For the sensitive layers ②, screen-printable pastes either of Ga₂O₃ or of 1% Nb-doped SrTiO₃ were prepared by mixing the oxide powder and an organic binder consisting of terpineol and ethyl cellulose. For the preparation of the Ga₂O₃ paste 99.999% pure Ga₂O₃ powder was calcinated at 1150 °C for 1 h and dry milled in a planetary ball mill using milling balls of zirconium dioxide to decrease grain size to approximately 1–2 μm. 1% Nb-doped SrTiO₃ powder was prepared in mixed-oxide technique as described in [18] for La-doped SrTiO₃.

The sensitive layers were then printed on the IDE structure, dried, sintered and activated (for activation see below).

An 8% yttrium doped zirconia (YSZ) substrate was used for the electrochemical pumping cell ③. The electrode material, a platinum/YSZ cermet (8 wt% Y₂O₃, Fraunhofer IKTS) was screen-printed on both sides of the substrate. The inner electrode is connected by a laser-drilled via. The electrochemical pumping cell was then fixed to the transducer with a commercial glass solder (Aremco 617; Figure 3c). The inner pumping electrode fits the cut ring at this joint and is thereby connected to the sensor chamber.

The covering ④ consists of an alumina substrate and has a cavity and a diffusion channel, which were milled and drilled, respectively, by laser. The cavity is located above the sensor and provides the volume, in which the oxygen concentration has to be adjusted. The exhaust gas enters the platform through the diffusion hole, located above the center of the circular gas sensitive area. The covering was joined to the transducer by a ceramic paste (QM42, DuPont). It should be noted here, that at this stage the sensor setup has to be heated externally to operation temperature in a small tube furnace. In the next step, a screen-printed platinum heater structure was included between the tapes. See results for the modeling and testing of the heater in Section 3.1.

One of the two sensitive films is activated by impregnation with 10% hexachloroplatinic (IV) acid similar to a procedure described in [20]. During heating up to operation temperature, the acid evaporates and catalytically active small platinum clusters are generated on the surface of the metal oxide grains.

In Figure 4a, the resulting uncovered IDE structure is shown, whereas Figure 4b depicts the same sensor after applying the screen-printed sensitive films. The successful activation of the left side can be seen by the changed color. The SEM picture (Figure 4c) shows an impregnated and thermally treated Ga₂O₃ sensor layer. The nanosized platinum clusters (white) are dispersed very well on the surface.

The two-sensor-setup was heated in a tube furnace to 650 °C. The gas flow of 60 L/h was supplied by a gas mixture facility and saturated with water in a fritted wash-bottle. The total H₂O content amounted to approx. 2.3%. 25 ppm up to 1,000 ppm of the test gas propane were added to a base gas mixture of N₂ and 2% or 5% O₂. The tests of the oxygen cross-sensitivity started with 10% O₂ and were reduced down to 0.1% O₂. The resistive sensor signals were either recorded with a Keithley 2700 Multimeter/Data Acquisition System or a Novocontrol Alpha-A High Performance Frequency Analyser. Simultaneously to the sensitivity investigations, a platinum heater to adjust the sensor platform to working temperature was developed and tested. The temperature distribution over the heater was measured with a thermo camera (Varioscan 3011 ST, InfraTec).

In order to heat the sensor element itself, a buried heater was developed. Its design was modeled by using the commercial software Comsol Multiphysics. The software provides the possibility to couple different physical problems. In our case, heat transfer was coupled with electrical conductivity. Details of the calculation method (but for a completely different system) can be found in [19]. Figure 5a shows the calculated temperature distribution of the transducer with a rosette-like heater, which turned out (at as a first order approximation) to be the best design with respect to temperature homogeneity. The temperature distribution as obtained from a realized sensor (screen-printed platinum thick-film paste; LPA 88/11; line width 300 μm; laminated as described above) is depicted in Figure 5b. The temperature distribution was measured with the thermo camera. First of all, modeling and simulation agree quite well. Especially the calculated temperature in the area where the sensitive film is located (between s_beg and s_end) is well simulated. For the application, it is important that the temperature distribution is homogenous over a wide range of the sensor area, which obviously is the case in both measurement and simulation.

A typical measurement result of a two-sensor-setup is plotted in Figure 6a. The resistance signals of an activated and a non-activated 1% Nb-doped SrTiO₃ layer are stable very soon after the measurement in the base gas has started. They also follow the propane concentration steps very quickly. The activated layer shows only a very slight reaction towards propane, whereas the non-activated sensor signal exhibits a resistance change up to 40%. This can be seen even better when the normalized resistance, R_norm, (R_norm = R / R_0ppmHC) is plotted versus the concentration of the test gas propane (Figure 6b). Similar results can be observed for Ga₂O₃ (Figure 7). The non-activated layer is very sensitive towards propane, while the resistance of the activated one changes only marginally.

Both the activated and the non-activated layer should be sensitive towards oxygen. On the one hand, the idea of the two-sensor-setup depends on the oxygen cross-sensitivity and on the other hand, the signal of the activated layer should be used to control the oxygen current through the electrochemical pumping cell.

In Figure 8, the dependency of the film resistance towards the oxygen concentration is displayed. The normalized resistance signal of activated and non-activated films behaves almost identical. Due to the 1% Nb-doping, the sensor is in the n-type region but reaches the so called intrinsic conductivity minimum at high oxygen concentrations of 10%. This result agrees completely with defect chemistry data of the literature [21].

Thus, two components of the sensor platform could be produced and tested. It was shown that it is possible to pattern platinum IDEs by laser. Secondly, the integration of the two-sensor-setup into the new platform design worked out. The tests showed that the sensor signal of activated and non activated sensor layer attained comparable results to [12].

It should be noted here that this platform may also be used for other types of high-temperature sensors that need a defined oxygen concentration for proper operation, i.e., for mixed potential sensors. Since all three components, the transducer, the electrochemical pumping cell, and the covering are prepared separately before being joined without prohibitively high temperature steps, a modular platform is achieved.