Mixed-Potential Gas Sensors Using an Electrolyte Consisting of Zinc Phosphate Glass and Benzimidazole

Mixed-potential gas sensors with a proton conductor consisting of zinc metaphosphate glass and benzimidazole were fabricated for the detection of hydrogen produced by intestinal bacteria in dry and humid air. The gas sensor consisting of an alumina substrate with platinum and gold electrodes showed good response to different hydrogen concentrations from 250 parts per million (ppm) to 25,000 ppm in dry and humid air at 100–130 °C. The sensor response varied linearly with the hydrogen and carbon monoxide concentrations due to mass transport limitations. The sensor responses to hydrogen gas (e.g., −0.613 mV to 1000 ppm H2) was higher than those to carbon monoxide gas (e.g., −0.128 mV to 1000 ppm CO) at 120 °C under atmosphere with the same level of humidity as expired air.


Introduction
Expired air contains over 100 different gases in a humid atmosphere at low concentrations, ranging from sub parts per billion (ppb) to parts per million (ppm), including hydrogen (H 2 ), carbon monoxide (CO), and various volatile organic compounds (VOCs) [1][2][3][4][5][6]. Expired air can provide information that is useful for monitoring human health. The concentration of H 2 in expired air has been used to monitor microbial metabolism in the colon [1][2][3][4], since H 2 is produced by intestinal bacteria. High-performance gas sensors provide a simple and non-invasive method to monitor expired air, and are therefore one of the most promising approaches for the early detection of diseases.
Mixed-potential gas sensors using yttria-stabilized zirconia (YSZ) electrolyte have been investigated for the detection of H 2 , CO, nitric oxide, ethanol, and methane [7][8][9][10][11]. These gas sensors must be operated at temperatures above 500 • C to provide both satisfactory response rates against target gases and good gas selectivity. These gas sensors require a heater to maintain the operating temperature above 500 • C, which results in high power consumption. Mixed-potential gas sensors that use electrolytes with high proton conductivity-such as Nafion, zirconium phosphate, and antimonic acid-have been reported to function at reduced operating temperatures [12][13][14][15]. However, these electrolytes must be impregnated with water to achieve high proton conductivity, which necessitates the inclusion of a complicated humidity control system to allow these sensors to operate at room temperature and under high humidity. Moreover, these gas sensors show low selectivity for H 2 against CO. One approach to overcoming this challenge is to use anhydrous proton conductors at an intermediate temperature between 100 and 300 • C. However, to the best Sensors 2017, 17, 97 2 of 8 of our knowledge, no mixed-potential gas sensors using a high-proton-conductivity electrolyte at intermediate temperatures have been reported (Table 1).
Oine et al. and Kato et al. have successfully prepared an anhydrous electrolyte that demonstrates high proton conductivity at temperatures between 100 and 200 • C by the reaction of zinc metaphosphate glass with benzimidazole [16,17]. This material has been investigated as an electrolyte in electrochemical devices that operate at intermediate temperatures, such as fuel cells. Thus, we suggest that this anhydrous electrolyte has the potential to be applied as a mixed-potential gas sensor.
In this study, we prepared a mixed-potential gas sensor using an electrolyte consisting of zinc metaphosphate glass and benzimidazole, and the sensor's response to H 2 gas in both humid and dry air was investigated. An initial investigation of the sensor's response to CO in humid air was also conducted to determine its selectivity for H 2 against other gases.

Materials and Methods
A glass with a nominal molar ratio of 1:1 ZnO:P 2 O 5 was prepared by melting a batch mixture of the commercially available reagent-grade chemicals ZnO and H 3 PO 4 in a platinum crucible at 1100 • C in air for 30 min. The melt was poured onto an iron plate and quenched by iron pressing to make glass flakes. The glass flakes were milled using an alumina mortar and pestle to below 100 µm in diameter. The glass powder was mixed with benzimidazole, and the mixture was heated at 170 • C for 12 h to make an electrolyte. The weight ratio of the glass powder and benzimidazole was 1:3. Details of the procedures and electrolyte properties have been previously reported [17].
The sensor element we constructed is shown schematically and as a photograph in Figure 1. The electrolyte was deposited on an alumina substrate with platinum (Pt) and gold (Au) electrodes, with both the electrode gap and electrode width being 1 mm. Finally, the electrolyte-coated substrate was heated from behind by placing the assembly on a hotplate, resulting in the formation of an electrolyte membrane on the substrate. However, to the best of our knowledge, no mixed-potential gas sensors using a high-proton-conductivity electrolyte at intermediate temperatures have been reported (Table 1). Oine et al. and Kato et al. have successfully prepared an anhydrous electrolyte that demonstrates high proton conductivity at temperatures between 100 and 200 °C by the reaction of zinc metaphosphate glass with benzimidazole [16,17]. This material has been investigated as an electrolyte in electrochemical devices that operate at intermediate temperatures, such as fuel cells. Thus, we suggest that this anhydrous electrolyte has the potential to be applied as a mixed-potential gas sensor.
In this study, we prepared a mixed-potential gas sensor using an electrolyte consisting of zinc metaphosphate glass and benzimidazole, and the sensor's response to H2 gas in both humid and dry air was investigated. An initial investigation of the sensor's response to CO in humid air was also conducted to determine its selectivity for H2 against other gases.

Materials and Methods
A glass with a nominal molar ratio of 1:1 ZnO:P2O5 was prepared by melting a batch mixture of the commercially available reagent-grade chemicals ZnO and H3PO4 in a platinum crucible at 1100 °C in air for 30 min. The melt was poured onto an iron plate and quenched by iron pressing to make glass flakes. The glass flakes were milled using an alumina mortar and pestle to below 100 μm in diameter. The glass powder was mixed with benzimidazole, and the mixture was heated at 170 °C for 12 h to make an electrolyte. The weight ratio of the glass powder and benzimidazole was 1:3. Details of the procedures and electrolyte properties have been previously reported [17].
The sensor element we constructed is shown schematically and as a photograph in Figure 1. The electrolyte was deposited on an alumina substrate with platinum (Pt) and gold (Au) electrodes, with both the electrode gap and electrode width being 1 mm. Finally, the electrolyte-coated substrate was heated from behind by placing the assembly on a hotplate, resulting in the formation of an electrolyte membrane on the substrate. The sensor element was placed in a test chamber and heated to 80-130 °C in an electrical tube furnace. Synthetic dry air was introduced into the chamber for 20 min, and a gas mixture of H2 or CO in synthetic dry air was then injected for 20 min. The gas mixture flow rate was 200 mL/min. Synthetic dry air and H2 or CO in synthetic air were prepared by mixing 99.99% N2 gas, The sensor element was placed in a test chamber and heated to 80-130 • C in an electrical tube furnace. Synthetic dry air was introduced into the chamber for 20 min, and a gas mixture of H 2 or CO in synthetic dry air was then injected for 20 min. The gas mixture flow rate was 200 mL/min. Synthetic dry air and H 2 or CO in synthetic air were prepared by mixing 99.99% N 2 gas, N 2 -balanced 5% H 2 gas, N 2 -balanced 5% CO gas, and 99.5% O 2 gas. The O 2 content of synthetic dry air and that of H 2 or CO in synthetic dry air were both 20 vol %. The H 2 or CO gas concentrations in synthetic dry air were controlled at values of 250, 500, 1000, 2500, 5000, 10,000, and 25,000 ppm. Synthetic humid air was prepared by passing air through ion-exchanged water at 25 • C before flowing to a test chamber.
The potential difference (electromotive force, EMF) between Pt and Au electrodes of the sensor element was recorded using a data logger (midi Logger GL220; Graphtec Corp., Yokohama, Japan). During the potential difference measurements, the Pt electrode was always connected to the positive terminal of the data logger. The EMFs of the sensor element in the air and gas mixtures are denoted as V a and V g , respectively. The sensor response (∆V) is defined as ∆V = V a − V g . Figure 2 shows that upon exposure to 25,000 ppm H 2 in humid and dry air at 80 • C, 100 • C, 120 • C, and 130 • C, the sensor's response increased with increasing operating temperature. Miura et al. described the potential response upon exposure to hydrogen in humid gas of a solid-state sensor that used a proton-conductor [12]. The EMF depends on the electrochemical oxidation and reduction reactions occurring at the Pt and Au electrodes:

Results and Discussion
A mixed potential is established when the electron transfer rates of the hydrogen oxidation and the oxygen reduction are same. Since the kinetics of the sensor's Pt and Au electrodes are different from those of the oxidation of H 2 to water, the mixed-potential generated at each electrode varies, and thus ∆V varies in accordance with the H 2 gas concentration. We reported that the sensor response to H 2 in synthetic dry air at 150 • C is higher than that at 130 • C [18]. However, in this paper, when the sensor response to H 2 was investigated at 140 • C, the EMF was unstable and gradually decreased over the measuring time, and the sensor showed no response to H 2 gas (not pictured). This behavior may result from changes in the electrochemical properties of the electrolyte due to volatilization of the benzimidazole in the electrolyte, because the flash point of benzimidazole was 143 • C [19]. Therefore, the optimum operating temperature of the sensor is below 140 • C. In general, the sensor responses are similar at each temperature. At 120 • C, the sensor response in dry air is higher, whereas for the other temperatures, the sensor response in humid air is higher. The electrolyte derived from zinc metaphosphate glass and benzimidazole is an anhydrous proton-conducting material, and has good water durability [16,17]. Therefore, the electrical properties of the electrolyte were not changed by water vapor, and the sensor responses using the electrolyte was considered to be seen even in humid air. However, the difference between the sensor responses to H 2 in dry air and humid air was significant.
In future study, we will investigate the cycle characteristics and error evaluation of the sensor response.
Sensors 2017, 17, 97 3 of 8 N2-balanced 5% H2 gas, N2-balanced 5% CO gas, and 99.5% O2 gas. The O2 content of synthetic dry air and that of H2 or CO in synthetic dry air were both 20 vol %. The H2 or CO gas concentrations in synthetic dry air were controlled at values of 250, 500, 1000, 2500, 5000, 10,000, and 25,000 ppm. Synthetic humid air was prepared by passing air through ion-exchanged water at 25 °C before flowing to a test chamber. The potential difference (electromotive force, EMF) between Pt and Au electrodes of the sensor element was recorded using a data logger (midi Logger GL220; Graphtec Corp., Yokohama, Japan). During the potential difference measurements, the Pt electrode was always connected to the positive terminal of the data logger. The EMFs of the sensor element in the air and gas mixtures are denoted as Va and Vg, respectively. The sensor response (ΔV) is defined as ΔV = Va − Vg. Figure 2 shows that upon exposure to 25,000 ppm H2 in humid and dry air at 80 °C, 100 °C, 120 °C, and 130 °C, the sensor's response increased with increasing operating temperature. Miura et al. described the potential response upon exposure to hydrogen in humid gas of a solid-state sensor that used a proton-conductor [12]. The EMF depends on the electrochemical oxidation and reduction reactions occurring at the Pt and Au electrodes:

Results and Discussion
A mixed potential is established when the electron transfer rates of the hydrogen oxidation and the oxygen reduction are same. Since the kinetics of the sensor's Pt and Au electrodes are different from those of the oxidation of H2 to water, the mixed-potential generated at each electrode varies, and thus ΔV varies in accordance with the H2 gas concentration. We reported that the sensor response to H2 in synthetic dry air at 150 °C is higher than that at 130 °C [18]. However, in this paper, when the sensor response to H2 was investigated at 140 °C, the EMF was unstable and gradually decreased over the measuring time, and the sensor showed no response to H2 gas (not pictured). This behavior may result from changes in the electrochemical properties of the electrolyte due to volatilization of the benzimidazole in the electrolyte, because the flash point of benzimidazole was 143 °C [19]. Therefore, the optimum operating temperature of the sensor is below 140 °C. In general, the sensor responses are similar at each temperature. At 120 °C, the sensor response in dry air is higher, whereas for the other temperatures, the sensor response in humid air is higher. The electrolyte derived from zinc metaphosphate glass and benzimidazole is an anhydrous proton-conducting material, and has good water durability [16,17]. Therefore, the electrical properties of the electrolyte were not changed by water vapor, and the sensor responses using the electrolyte was considered to be seen even in humid air. However, the difference between the sensor responses to H2 in dry air and humid air was significant. In future study, we will investigate the cycle characteristics and error evaluation of the sensor response.  Figure 3 shows the sensor responses to H2 concentrations ranging from 250 ppm to 25,000 ppm in humid air at 120 °C. These results show that the ΔV of the sensor began to decrease rapidly on exposure to H2, and the response was stronger for higher concentrations of H2. Figure 2 shows that the difference between the sensor responses to H2 in dry air and humid air was large at this stage. In order to detect H2 with high accuracy, error evaluation of the sensor response under different humidity conditions must be investigated. The model designed by Miura et al. suggests that a linear relationship exists between the logarithm of the H2 concentration and ΔV, assuming a Tafel-like equation for the polarization behavior of the oxidation and reduction reactions at the electrode [12]. The ΔV can thus be expressed in the form where a and b are constants, and CH2 is the concentration of H2. Figure 4 shows the relationship between the sensor response and the logarithm of the H2 concentration. The least squares regression analysis of the sensor can be expressed as follows: where the CH2 is the concentration of H2, and R is the correlation coefficient. The sensitivity of the sensor was −4.7 mV/decade. However, no good linear relationship between the logarithm of H2 concentration and ΔV of the sensor was obviously found.   Figure 3 shows the sensor responses to H 2 concentrations ranging from 250 ppm to 25,000 ppm in humid air at 120 • C. These results show that the ∆V of the sensor began to decrease rapidly on exposure to H 2 , and the response was stronger for higher concentrations of H 2 . Figure 2 shows that the difference between the sensor responses to H 2 in dry air and humid air was large at this stage. In order to detect H 2 with high accuracy, error evaluation of the sensor response under different humidity conditions must be investigated. The model designed by Miura et al. suggests that a linear relationship exists between the logarithm of the H 2 concentration and ∆V, assuming a Tafel-like equation for the polarization behavior of the oxidation and reduction reactions at the electrode [12]. The ∆V can thus be expressed in the form: where a and b are constants, and C H2 is the concentration of H 2 . Figure 4 shows the relationship between the sensor response and the logarithm of the H 2 concentration. The least squares regression analysis of the sensor can be expressed as follows: where the C H2 is the concentration of H 2 , and R is the correlation coefficient. The sensitivity of the sensor was −4.7 mV/decade. However, no good linear relationship between the logarithm of H 2 concentration and ∆V of the sensor was obviously found.  Figure 3 shows the sensor responses to H2 concentrations ranging from 250 ppm to 25,000 ppm in humid air at 120 °C. These results show that the ΔV of the sensor began to decrease rapidly on exposure to H2, and the response was stronger for higher concentrations of H2. Figure 2 shows that the difference between the sensor responses to H2 in dry air and humid air was large at this stage. In order to detect H2 with high accuracy, error evaluation of the sensor response under different humidity conditions must be investigated. The model designed by Miura et al. suggests that a linear relationship exists between the logarithm of the H2 concentration and ΔV, assuming a Tafel-like equation for the polarization behavior of the oxidation and reduction reactions at the electrode [12]. The ΔV can thus be expressed in the form where a and b are constants, and CH2 is the concentration of H2. Figure 4 shows the relationship between the sensor response and the logarithm of the H2 concentration. The least squares regression analysis of the sensor can be expressed as follows: where the CH2 is the concentration of H2, and R is the correlation coefficient. The sensitivity of the sensor was −4.7 mV/decade. However, no good linear relationship between the logarithm of H2 concentration and ΔV of the sensor was obviously found.   According to Garzon et al., if the concentration of the sensed gas (H2) is much lower than the concentration of O2, a linear relationship between the H2 concentration and ΔV exists because of the diffusional mass transport limitation [8]. In this case, ΔV can be expressed as where a and K are constants. Figure 5 shows the relationship between sensor response and the H2 concentration. The sensor response was adequately linear. The least squares regression analysis of the sensor can be expressed as follows: where CH2 is the concentration of H2, and R is the correlation coefficient. The sensitivity of the sensor was −0.00046 mV/ppm. R for Equation (6) was higher than that for Equation (4). This suggested that the oxidation reaction of H2 may be mass transport-limited.  Figure 6 shows the sensor's responses to CO in humid air at 120 °C and at concentrations ranging from 250 to 25,000 ppm. When the sensor was exposed to these conditions, its ΔV decreased rapidly, and the response decreased as the CO concentration increased. In CO gas sensing, the oxidation reaction of CO can take place simultaneously at both the Pt and Au electrodes:  According to Garzon et al., if the concentration of the sensed gas (H 2 ) is much lower than the concentration of O 2 , a linear relationship between the H 2 concentration and ∆V exists because of the diffusional mass transport limitation [8]. In this case, ∆V can be expressed as where a and K are constants. Figure 5 shows the relationship between sensor response and the H 2 concentration. The sensor response was adequately linear. The least squares regression analysis of the sensor can be expressed as follows: where C H2 is the concentration of H 2 , and R is the correlation coefficient. The sensitivity of the sensor was −0.00046 mV/ppm. R for Equation (6) was higher than that for Equation (4). This suggested that the oxidation reaction of H 2 may be mass transport-limited. According to Garzon et al., if the concentration of the sensed gas (H2) is much lower than the concentration of O2, a linear relationship between the H2 concentration and ΔV exists because of the diffusional mass transport limitation [8]. In this case, ΔV can be expressed as where a and K are constants. Figure 5 shows the relationship between sensor response and the H2 concentration. The sensor response was adequately linear. The least squares regression analysis of the sensor can be expressed as follows: where CH2 is the concentration of H2, and R is the correlation coefficient. The sensitivity of the sensor was −0.00046 mV/ppm. R for Equation (6) was higher than that for Equation (4). This suggested that the oxidation reaction of H2 may be mass transport-limited.  Figure 6 shows the sensor's responses to CO in humid air at 120 °C and at concentrations ranging from 250 to 25,000 ppm. When the sensor was exposed to these conditions, its ΔV decreased rapidly, and the response decreased as the CO concentration increased. In CO gas sensing, the oxidation reaction of CO can take place simultaneously at both the Pt and Au electrodes:  Figure 6 shows the sensor's responses to CO in humid air at 120 • C and at concentrations ranging from 250 to 25,000 ppm. When the sensor was exposed to these conditions, its ∆V decreased rapidly, and the response decreased as the CO concentration increased. In CO gas sensing, the oxidation reaction of CO can take place simultaneously at both the Pt and Au electrodes:  where CCO is the concentration of CO and R is the correlation coefficient. Since R is slightly larger for Equation (9) than for Equation (8), the oxidation reaction of CO may be mass-transport-limited, as seen for the H2 gas sensing. The sensor responses to 1000 ppm H2 and 1000 ppm CO were determined to be −0.613 mV and −0.128 mV, respectively, in humid air at 120 °C. These responses are lower than those seen for mixed-potential gas sensors that used proton conductor operating at around room temperature (−190 mV for 1000 ppm H2 and −120 mV for 1000 ppm CO) [12]. The value of H2 selectivity against CO (SH2/CO) was determined using the equation SH2/CO = ΔVH2/ΔVCO, where ΔVH2 is the sensor's response at 1000 ppm H2, and ΔVCO is the sensor response at 1000 ppm CO. The SH2/CO of our sensor  where C CO is the concentration of CO and R is the correlation coefficient. Since R is slightly larger for Equation (9) than for Equation (8), the oxidation reaction of CO may be mass-transport-limited, as seen for the H 2 gas sensing.  Figure 7 shows the relationship between the sensor response and the logarithm of the CO concentration, and the relationship between the sensor response and the CO concentration in humid air at 120 °C. The equations of the relationship between the sensor response and CO concentration were determined from Equations (3)  where CCO is the concentration of CO and R is the correlation coefficient. Since R is slightly larger for Equation (9) than for Equation (8), the oxidation reaction of CO may be mass-transport-limited, as seen for the H2 gas sensing. The sensor responses to 1000 ppm H2 and 1000 ppm CO were determined to be −0.613 mV and −0.128 mV, respectively, in humid air at 120 °C. These responses are lower than those seen for mixed-potential gas sensors that used proton conductor operating at around room temperature (−190 mV for 1000 ppm H2 and −120 mV for 1000 ppm CO) [12]. The value of H2 selectivity against CO (SH2/CO) was determined using the equation SH2/CO = ΔVH2/ΔVCO, where ΔVH2 is the sensor's response at 1000 ppm H2, and ΔVCO is the sensor response at 1000 ppm CO. The SH2/CO of our sensor The sensor responses to 1000 ppm H 2 and 1000 ppm CO were determined to be −0.613 mV and −0.128 mV, respectively, in humid air at 120 • C. These responses are lower than those seen for mixed-potential gas sensors that used proton conductor operating at around room temperature (−190 mV for 1000 ppm H 2 and −120 mV for 1000 ppm CO) [12]. The value of H 2 selectivity against Sensors 2017, 17, 97 7 of 8 CO (S H2/CO ) was determined using the equation S H2/CO = ∆V H2 /∆V CO , where ∆V H2 is the sensor's response at 1000 ppm H 2 , and ∆V CO is the sensor response at 1000 ppm CO. The S H2/CO of our sensor was 4.8, and that of the sensor using a proton conductor operating at around room temperature was 1.6 [12].
Semiconductor gas sensors using n-type SnO 2 have been investigated for several decades for their high response to inflammable gases, such as H 2 , CO, CH 4 , and C 2 H 5 OH [20,21]. However, their low selectivity for H 2 over CO was regarded as a problem to be solved. Moon et al. focused on the use of a CuO-doped SnO 2 -ZnO composite material to improve the gas selectivity, and reported a selectivity of 4 for H 2 (200 ppm) over CO (200 ppm) at 350 • C [21]. The sensor reported herein-which used an electrolyte consisting of zinc metaphosphate glass and benzimidazole-showed a higher selectivity for H 2 against CO than the sensor reported by Moon et al. However, our sensor's response to H 2 is currently inferior, thus optimization of the sensor design is required to achieve a greater response. Further investigations of the sensor's response to other interfering species (such as CH 4 and VOCs) will be conducted in the future.

Conclusions
In this study, we have investigated the responses of a mixed-potential gas sensor that used a proton conductor composed of zinc metaphosphate glass and benzimidazole for the detection of H 2 (produced by intestinal bacteria) and CO gas (an interference in expired air). The gas sensor showed good sensor response to H 2 in the concentration range 250-25,000 ppm under atmosphere having the same level of humidity as expired air at 120 • C. The mixed-potential gas sensor without a heater can detect gases under the atmosphere at temperatures in the range 100-130 • C. The sensor response varied linearly with the hydrogen and carbon monoxide concentrations, as expected under mass-transport-limited conditions. The gas sensor showed high gas selectivity for H 2 against CO (S H2/CO = 4.8).