Evaluation of Oxygen Adsorption Based on the Electric Properties of SnO2 Semiconductor Gas Sensors

*Corresponding author: e-mail: shimanoe.kengo.695@m.kyushu-u.ac.jp †Present address: Chemical and Texture Research Institute, Fukuoka Industrial Technology Center, Chikushino, Fukuoka 818-8540, Japan ††Present address: Faculty of Humanity-Oriented Science and Engineering, Kinki University, Iizuka, Fukuoka 820-8555, Japan †††Present address: Department of Applied Chemistry and Biochemistry, Kumamoto University, Kumamoto 860-8555, Japan


Introduction
Tin dioxide (SnO 2 ) is the main semiconductor used to detect combustible gases in commercial applications as well as in research on gas sensors. Such resistive-type gas sensors have been utilized as gas leakage detectors and for air pollution monitoring, (1-3) and can also be used as detectors for breath gases. (4,5) Combustible gases are typically detected by measuring the change in the electric resistance of SnO 2 particles. In air atmosphere, SnO 2 particles adsorb the dissociative oxygen with negative charge, and an electron depletion region is formed on the particle surface. In contrast, combustible gases consume the adsorbed oxygen via a surface combustion reaction and thus reduce the electron depletion region. In short, oxygen adsorption increases the electric resistance, whereas a surface combustion reaction decreases the electric resistance of SnO 2 particles. A difference in the electric resistance indicates the presence of combustible gases. Therefore, the key to gas detection is the oxygen adsorption amount and species. Thus understanding the oxygen adsorption ability of the SnO 2 surface is very important for gas sensors.
Recently, we have proposed the existence of a relationship between the oxygen adsorption and the electric properties, on the basis of the production of an electron depletion layer. (6,7) According to the proposal, the electric resistance of SnO 2 gas sensors is proportional to the square root of the oxygen partial pressure (P O2 1/2 ), or fourth root of the oxygen partial pressure (P O2 1/4 ), and these relationships determine the oxygen adsorption species as being either O − or O 2− . (7) For the evaluation of such a relationship, we have recently proposed that a gas treatment system equipped with a catalytic chamber and an absorbent chamber has to be put on the conventional gas flow system. (8) The resulting oxygen adsorption species on the SnO 2 surface at 350 °C in extremely dry atmosphere was O 2− . Under this condition, the oxygen adsorption is expressed by the following reaction, and the electric resistance of SnO 2 gas sensors is explained by the following equation. (7,8) Here, R/R 0 is the relative resistance, R and R 0 are the electric resistances in the atmosphere with and without oxygen, respectively, N D is the donor density, a is the crystallite radius, K (O 2− ad) is the oxygen adsorption equilibrium constant of O 2− , and c is a constant. The equilibrium constant ad) is expressed as the ratio of k 1 to k −1 (K (O 2− ad) = k 1 /k −1 ); here, k 1 and k −1 are the rate constants of the forward and reverse reactions of Eq. (1), respectively. The K (O 2− ad) value indicates the oxygen adsorption ability on the SnO 2 surface as O 2-, and determines the intrinsic properties on the sensitivity of SnO 2 gas sensors. However, this value has been difficult to evaluate experimentally. Therefore, in this study, we attempted to evaluate the K (O 2− ad) value on the basis of the relationship between R/R 0 and P O2 1/4 from experimental results using SnO 2 nanoparticles of two different sizes. Basically, the K (O 2− ad) value should be independent of the SnO 2 crystallite size when the calcination temperature of SnO 2 is the same, because the surface state of the SnO 2 crystals is determined by the calcination process. SnO 2 nanoparticles of different sizes were prepared by hydrothermal synthesis and calcination, which is the same for each SnO 2 . (9) Additionally, we can assume that these donor densities are almost in the same range, because the calcination temperature strongly affects the donor density. (10) Here, donor density is the intrinsic factor of SnO 2 nanoparticles and is not dependent on the oxygen adsorption. The density of electron charge dependent on the oxygen adsorption is defined as the "carrier concentration". Moreover, we compared the sensor responses to hydrogen using size-controlled SnO 2 nanoparticles, and discussed the responses on the basis of the characteristics of SnO 2 nanoparticles such as a and K (O 2− ad) .

Preparation and evaluation of materials
The size-controlled SnO 2 nanoparticles solutions were prepared by hydrothermal synthesis starting from SnCl 4 ·5H 2 O and NH 4 HCO 3 . (9,11) For the preparation of small-SnO 2 (S-SnO 2 ) nanoparticles, an aqueous solution of SnCl 4 (1 mol/L) was added dropwise into the aqueous solution of NH 4 HCO 3 (1 mol/L) and left for half a day to obtain a white precipitate. The thus obtained stannic acid gel was dispersed in deionized water after washing to remove the Cl − by centrifugation, and the solution pH was adjusted to 10.5 by mixing with an ammonia solution. Subsequently, the S-SnO 2 dispersed solution was prepared by hydrothermal synthesis at 200 °C for 3 h at 10 MPa with stirring at 600 rpm.
Large-SnO 2 (L-SnO 2 ) nanoparticles were prepared by hydrothermal synthesis via seedmediated growth using S-SnO 2 nanoparticles as seed particles. The stannic acid gel, as described above, was dispersed in deionized water, and the pH was adjusted to 13.5 by mixing with tetramethylammonium hydroxide (TMAH) to prepare the growth solution. An appropriate amount of S-SnO 2 sol was added to the growth solution, and this mixed solution was hydrothermally treated at 250 °C for 10 h at 10 MPa with stirring at 600 rpm. Detailed conditions for obtaining the sizecontrolled SnO 2 nanoparticle solutions are described in a previous report. (9) The size of the nanoparticles in the solutions was determined by dynamic light scattering (DLS) analysis using a DLS spectrophotometer (LPA-3000/3100, Otsuka Electronics). The crystal structure was analyzed by an X-ray diffraction (XRD; RINT 2100, Rigaku) with copper Kα radiation, and the average crystallite size was estimated using Scherrer's equation on the basis of the XRD patterns at (110), (101), and (211) peaks. The obtained nanoparticles were observed using field-emission scanning electron microscopy (FE-SEM; JSM-6340F, JEOL).

Sensor fabrication and evaluation of electronic resistance
A thin-film-type sensing layer was fabricated by the spin-coating method on an alumina substrate printed with a comb-type Au electrode (90 mm distance between lines, 64 mm 2 sensing area). The thickness of the sensing layers was controlled to ca. 200 nm. The obtained devices were attached to the gas flow apparatus equipped with an electric furnace, and calcined at 400 °C for 3 h to obtain the sensor device. The measurement system for this study was constructed with various chambers for the gas treatment system, electric resistance measurement (sensor chamber), and oxygen sensor, as described in Fig. 1(a). The gas treatment system is necessary for evaluating the electric resistance accurately. (8) First, the sample gases of various concentrations of oxygen with nitrogen were passed through the gas treatment system, which is equipped with a catalyst and adsorbent chambers, to eliminate the impurities of the gases. Subsequently, the purified gases were passed through the sensor chamber and oxygen sensor chamber to evaluate the electric resistance of the device and P O2 in the atmosphere, respectively. Oxygen partial pressure was controlled by changing the mixing ratio of oxygen (or synthetic air) to nitrogen gases from commercial gas cylinders, and the total gas flow rate of sample gases was set at 80 cm 3 /min with mass flow controllers (SEC-series; HORIBA STEC). Here, the electric resistances were evaluated at 350 °C.
Finally, the sensor responses to 200 and 1000 ppm hydrogen at 350 °C for each sensor device were evaluated using a conventional gas flow apparatus equipped with an electric furnace and without using the gas treatment system, as shown in Fig. 1(b). The sample gases of the hydrogen in air were prepared by mixing with synthetic air gas from commercial gas cylinders. The sensor response to hydrogen was defined as R a /R g , where R a is the electric resistance in synthetic air and R g is that in air containing hydrogen.

Results and Discussion
Figure 2(a) shows the colloidal particle distribution (DLS) of S-SnO 2 and L-SnO 2 nanoparticles; no significant particle aggregation appeared in each nanoparticle. The particle distribution range is clearly different for each nanoparticle. The average colloidal particle sizes of the S-SnO 2 and L-SnO 2 nanoparticles were 5 and 17 nm, respectively. The XRD patterns of the dried and calcined powders using S-SnO 2 and L-SnO 2 are shown in Fig. 2(b); powder drying and calcining were carried out under the conditions of 120 °C for 12 h and 400 °C for 3 h, respectively. These XRD patterns correspond well to the cassiterite-type SnO 2 structure (JCPDS 41-1445), and the SnO 2 was successfully crystallized during the hydrothermal treatment at 200 °C because SnO 2 peaks clearly appeared on the XRD pattern of the dried SnO 2 powders. The estimated crystallite sizes of the dried nanoparticles of S-SnO 2 and L-SnO 2 were 4 and 17 nm, respectively. These sizes are consistent with the average colloidal particle sizes from the DLS measurement. Thus, S-SnO 2 and L-SnO 2 nanoparticles are highly dispersed in each solution. In addition, the estimated crystallite sizes of calcined nanoparticles for S-SnO 2 and L-SnO 2 were 7 and 17 nm, respectively. Figure 2(c) shows the SEM images of S-SnO 2 and L-SnO 2 nanoparticles calcined at 400 °C, and no apparent aggregation of the nanoparticles are observed in both sizes of SnO 2 . Accordingly, we obtained SnO 2 nanoparticles of two different sizes under the same calcination conditions, which indicates that the donor densities of both SnO 2 nanoparticles are almost in the same range.
The electric resistance at various oxygen partial pressures [P O2 = 0 (=N 2 ), 0.01, 0.05, 0.1, 0.4, 1 atm] was measured at 350 °C using S-SnO 2 and L-SnO 2 thin-film sensors. The relative resistance is determined on the basis of the ratio of the electric resistance in the atmosphere, including oxygen, to that in nitrogen (R/R N2 ). The relationships of the relative resistance to both P O2 1/2 and P O2 1/4 are shown in Figs. 3(a) and 3(b), respectively. Clearly, the relative resistance shows a nonlinear relationship with P O2 1/2 , while it shows a linear relationship with P O2 1/4 in the operated P O2 range regardless of the particle size. Thus, the oxygen is adsorbed as O 2− on the particle surface; in addition, the amount of O − adsorption on the particle surface is negligibly small, as described in a previous report. (7) The K (O 2− ad) value was calculated from the theoretical equation Eq. (2) using the slopes of the approximated straight lines in Fig. 3(b). The obtained slope and estimated K (O 2− ad) values of S-SnO 2 and L-SnO 2 nanoparticles are compiled in Table 1. Here, N D is assumed as 5.0 × 10 18 cm −3 (5.0 × 10 −3 nm −3 ). Accordingly, these K (O 2− ad) values are almost in the same range. The K (O 2− ad) value of S-SnO 2 nanoparticles was 3.2 times as large as that of L-SnO 2 nanoparticles. However, the reason for such a small difference is as yet unclear. Nevertheless, the obtained ad) values indicate that the oxygen adsorption abilities of both SnO 2 surfaces are the same, and are independent of the crystallite size. On the other hand, the relative resistance is significantly affected by crystallite size throughout the entire measured P O2 range. The estimated surface-area-tovolume ratios of S-SnO 2 (7 nm) and L-SnO 2 (17 nm) are shown in Table 1, and the ratio of S-SnO 2 is significantly larger than that of L-SnO 2 . Thus, the crystallite size affects the ratio of surfacearea-to-volume, and causes the difference in the relative resistance. Therefore, the obtained results correspond well to the theoretical equation Eq. (2) and the K (O 2− ad) values successfully obtained from the experimental result.
Finally, we evaluated the sensor responses to 200 and 1000 ppm hydrogen using S-SnO 2 and L-SnO 2 thin-film sensors at 350 °C, as shown in Fig. 4. The sensor response of S-SnO 2 nanoparticles is significantly higher than that of L-SnO 2 nanoparticles, regardless of the hydrogen concentration, even though the K (O 2− ad) values are in the same range. Such difference in the sensor response is caused by the difference in crystallite size, which indicates a difference in the surfacearea-to-volume ratio as shown in Table 1. The sensor response using S-SnO 2 is about 3 times larger than that using L-SnO 2 , and the ratio of surface-area-to-volume of S-SnO 2 is about 2.5 times larger than that of L-SnO 2 . Thus, the enhancement of the sensor response to hydrogen is caused by increasing the surface-area-to-volume ratio when the donor density and oxygen adsorption equilibrium are the same, and it fundamentally agrees with a previously reported theory of the gas sensing mechanism.

Conclusions
To calculate the oxygen adsorption equilibrium constant K (O 2− ad) on the basis of the evaluation of electric properties, we measured the electric resistance at various oxygen partial pressures using highly pure gases passed through a gas treatment system. SnO 2 nanoparticles of two sizes, namely 7 and 17 nm, were prepared by hydrothermal synthesis and calcination. Thin-film-type sensing layers were fabricated by the spin-coating method, and the sensor devices were obtained by sintering at 400 °C. The relative resistance (R/R N2 ) of both sensors had a linear relationship with P O2 1/4 at 350 °C in extremely dry atmosphere, which indicates that oxygen is adsorbed as O 2− on the particle surface. Moreover, according to the slope of the approximate straight line between R/R N2 and P O2 1/4 , the K (O 2− ad) values of both sensors were almost in the same range. This indicates that the K (O 2− ad) value is independent of the crystallite size of SnO 2 . On the other hand, the sensor responses to hydrogen of small SnO 2 nanoparticles were clearly higher than those of large SnO 2 nanoparticles even though these K (O 2− ad) values were almost the same. This is because the surface-area-to-volume ratio of SnO 2 crystals significantly affected the sensor response. According to our results, the K (O 2− ad) value, which is important surface information of the gas sensors, can be evaluated easily from the relationship between the electric resistance and the P O2 . Moreover, the sensor response to hydrogen is strongly affected by the crystallite size because the phenomenon of the surface area is susceptible to the whole particle region. Therefore, these results confirm a material design to enhance the sensor response; for example, increasing the oxygen adsorption equilibrium boosts the sensor response of the semiconductor gas sensors. In the future, it is necessary to investigate the effect of the surface modification such as elemental doping on the oxygen adsorption equilibrium, to understand the importance of the surface modification for semiconductor gas sensors.