The catalytic MIS structure of the Pd-TiO₂ thin film gas sensor was fabricated on low resistivity p-type or n-type heavily doped Si as Pd/TiO₂ nanoporous/SiO₂/Si layers with a surface area of 1.3 × 1.3 cm². Figure 1 shows the schematic diagram of the device fabrication steps. First the Si was cleaned with sulfuric-peroxide mixture (SPM) to remove the native oxide layer, and then a 100 nm thick SiO₂ insulator layer was grown on the Si using wet oxidation. The backside of the Si was etched using reactive ion etching to form the ohmic contact. The surface morphology of the TiO₂ and Pd affects the device performance. Therefore, a nanoimprinting process [13] was used to pattern the anodic aluminum oxide (AAO) to expand the surface area. Figure 1 also shows the fabrication procedure of the well-ordered nanopored titania structures. For this fabrication, an Al plate (1 × 3 cm²) was first degassed? in acetone and electropolished in a 1:5 mixed solution of perchloric (60%) acid and ethanol (85%) under a constant voltage condition (15 V) at 3 °C for 2 min to create a mirrored surface. Al was anodized at a constant voltage of 50 V in a 0.3 M oxalic acid solution at 15 °C for 4 h. Thereafter, the anodic oxide layer was removed using a mixture of phosphoric acid (50 mL, 85%), chromic acid (18 g, Aldrich) and deionized water (950 mL) at 60 °C for 2 h. The second anodization step was performed under the same conditions for 40∼60 sec. The pores were widened by dipping the nanoporous alumina template in a 0.5 M phosphoric acid solution for 1 hour.

Polymethyl methacrylate (PMMA, Aldrich) with a molecular mass of 350 kg/mol was dissolved in chlorobenzene and poured onto the nanoporous alumina template to fabricate a nanopoled polymer. The PMMA infiltrated into the nanopores during heating at 150 °C. The PMMA nanopoles formed after the alumina template was removed through wet etching in a 1.4 wt% FeCl₃/5 M HCl solution, and the remaining Al and alumina was removed with a 10 wt% NaOH solution. The titania sol-gel solution was prepared by mixing titanium (IV) ethoxide (1.5 g), HCl (3 g, 38%) and 2-propanol (10 g). The titania sol-gel solution was spin-coated onto a SiO₂/Si, and then the substrate was embossed with the PMMS nanopoles before the solution was dried. The embossed titania film was sintered in a dry oven at 100 °C for 1 hour. The nanoporous titania film was obtained by removing the PMMA nanopoles with acetonitrile. Subsequently, a Pd electrode with a thickness of 60 nm and an area of 1 cm² was thermally evaporated using a shadow mask on the pored TiO₂ surface. A high current was applied for 600 sec for the deposition using a thermal evaporator because Pd has a high melting point, i.e., above 1,500 °C. Finally, a metal Al contact, with a thickness of 0.1 μm, was evaporated on the opposite side of the device as an ohmic contact.

The surface morphology of the titania film was analyzed with an atomic force microscope (AFM, Parksystems, XE-100) using a Pt-Ir coated SiN₃ tip under the non-contact mode. The composition of the titania films was determined using semiquantitative energy-dispersive X-ray (EDX, Hitachi, S-3500N) analysis. The crystalline phase of the titania films prepared under various conditions was determined using an X-ray diffraction instrument. For this analysis, the titania layer was prepared over the ITO surface [13] at different sintering temperatures. The X-ray diffraction (XRD) diffractograms of these layers were compared with titania powder to analyze the possible presence of the peak related to the anatase phase. A field emission scanning electron microscope (FE-SEM) was also used to examine the cross section of the gas sensor device. For this analysis, the device was first coated with Pt and inserted in the FE-SEM chamber under a vacuum of 10⁻⁵ torr.

A gas chamber was designed with a provisional gas flow through a Mass Flow Controller (MFC) with a flow rate range of 1 sccm to 100 sccm. The working volume of the gas chamber was approximately 2 L. There were two gas flow lines within the chamber. One line was used to supply the gas and the other line was used to collect gas samples in a Tedlar® bag in order to confirm the concentration of the gases with a gas chromatography (GC) unit. Before the target gases were injected, nitrogen was injected into the chamber to remove any other gases present and to create atmospheric conditions (STP). The electronic interface of the detection unit had a Keithley^® 236 voltage source and a current measurement meter that could supply (–3 V) static and (–3 to 0 V) sweep voltages. In this study, the static potential mode was used to determine the sensing time, and the sweep voltage was used to obtain the I-V curve.

Titania is a well studied transition metal oxide. In the last few years, interest in both the application and the fundamental research of this material has increased because of its remarkable optical and electronic properties and its good catalytic activity, which is similar to those of Au and Pt [9,13,14]. The thickness and roughness of the TiO₂ layer in a gas sensor significantly influence the sensitivity of device [9,14]. Therefore, studying the surface morphology of TiO₂ layer was important. The XRD spectra of the TiO₂ film (Figure 2) revealed the effect of the sintering temperature on the film formation. The anatase peak in this figure was seen at about 25 degrees. This phase has been shown to be linked with gas sensor sensitivity in previous reports [9,16]. Also, anatase (101) was not present under a sintering temperature of 500 °C. Therefore, processing of the device as maintained over 500 °C for 1 hour to form the TiO₂ anatase phase seen in Figure 2.

The roughness and the thickness of Pd layer that was deposited using a thermal evaporator affected the gas sensing performance in a similar manner to TiO₂. Therefore, the cross sectional images of non-nanoimprinted device were examined using FE-SEM (Figure 3), and revealed a palladium thickness of about 60 nm. In the latter part of the study, the nanoimprinting method was used to expand the surface area of the sensing layer. Therefore, the device fabricated with the nanopored TiO₂ layer had a higher surface roughness than the non-nanopored process (flat surface). The palladium surface in these devices was monitored using the AFM lithography process. Figure 4 shows the topography of the Pd surface monitored using AFM and FE-SEM. The AAO template (Figure 4a) was prepared to make the PMMA mold. The imprinted TiO₂ nanopore size was about 70∼90 nm, as seen in Figure 4b. The SEM image of the imprinted TiO₂ layer (Figure 4c) illustrated the surface morphology. From these data, the surface area of the imprinted TiO₂ layer was larger than the non-imprinted layer.

First, the fabricated devices with the non-imprinted Pd layer were placed in the gas chamber and tested for detection of N₂. The detection was performed under a low vacuum and under a standard air condition (80%:20% N₂:O₂ respectively). The standard diagnostic test of the sensor operation was a voltage sweep from the negative to the positive direction while measuring the current. During this diagnosis, the devices showed the typical Schottky diode I-V characteristics, as previously reported [3,9]. N₂ and O₂ gases decrease the sensors’ output current level because these gases react with palladium and TiO₂ on the gas sensor device [9]. The experiments were repeated on the device with nitrogen gas that was diluted with standard air. Various concentrations of nitrogen ranging between 50 ppm to 200 ppm were injected into the detection chamber to determine R-V characteristics. The gas sensor’s resistance increased proportionately with the N₂ concentration. The nitrogen gas sensor operated typically in the forward bias region at a bias of 1.5∼1.8 V, with a current change from 47 mA to 67 mA at an N₂ flow rate of 100 sccm. The difference in the sensor output with nitrogen gas compared to standard air was 20 mA with a forward bias of 1.7 V. The sensor response time was only 2 minutes.

The gas sensor with the non-imprinted Pd layer was subsequently tested to detect toluene in N₂ using the same reaction chamber. The adsorption of toluene gas at the Pd-TiO₂ interface of the sensor resulted in a change of the gas sensor’s I-V characteristics, and the relationship between the sensor output and toluene concentration was linear (Figure 5). However, the sensor output in terms of current level for the toluene gas injection was smaller compared to nitrogen, but the gas sensing characteristics were excellent in the former case. This gas sensor operated even at room temperature, and the detection time was only a few minutes. Additionally, the sensor was reused with a full recovery to the baseline level (Figure 5, stretch-d) by drawing the air out of the vessel after the reaction reached steady state (Figure 5, stretch-b to c) by pumping fresh air into the chamber. At high toluene concentrations, the response time was shorter for longer recovery times, and vice versa. The higher recovery period for the device could be reduced by applying an opposite voltage [9,10].

In addition to the operability of the sensor at room temperature, this study also attempted to improve the sensitivity of the gas sensor device towards toluene. Therefore, the surface appearance was changed using a nanoimprinting method to effectively enhance the surface roughness (to about 30%), and the gas sensing characteristics improved due to an increased interface reaction between the palladium surface and the charged toluene ions. The nanoimprinted device was more sensitive towards toluene gas than the device fabricated using a non-imprinted (normal) method as evident in Figure 6, thereby validating the relationship between the surface roughness and the device sensitivity. On the other hand, the sensitivity of the AAO device, in Equation 1 where R is the change in the resistance from the baseline, was not affected by increasing the toluene concentration (Figure 7) [7–9]: (1) S = R nitrogen / R toluene

In other words, the device had a higher sensitivity at a low toluene concentration while using the AAO nanoimprinting method.