key: cord-0020432-w7z92sif authors: Park, Jongyoon; Tabata, Hitoshi title: Gas Sensor Array Using a Hybrid Structure Based on Zeolite and Oxide Semiconductors for Multiple Bio-Gas Detection date: 2021-08-10 journal: ACS Omega DOI: 10.1021/acsomega.1c01435 sha: e0ab36c071cbf68b858388992a807829828ffad7 doc_id: 20432 cord_uid: w7z92sif [Image: see text] Semiconductor-type gas sensors, composed of metal-oxide semiconductors and porous zeolite materials, are attractive devices for bio-gas detection, particularly when used as bio-gas sensors such as electronic nose application. Previous studies have shown such detection can be obtained with a separate gas concentrator and a sensor device using zeolites and oxide semiconductors of WO(3) nanoparticles. By applying the gas concentrator, porous molecular structures alter both the gas sensitivity and the selectivity, and even can be used to define the sensor characteristics. Based on such a gas sensor design, we investigated the properties of an array of three sensors made of a layer of WO(3) nanoparticles coated with zeolites with different interactions between gas molecule adsorption and desorption. The array was tested with four volatile organic compounds, each measured at different concentrations. The results confirm that the features of individual zeolites combined with the hybrid gas sensor behavior, along with the differences among the sensors, are sufficient for enabling the discrimination of volatile compounds when disregarding their concentration. The role of bio-sniffer played by dog or human in the past to detect various types of diseases such as cancer or diabetes is modernized by using bio-gas sensors. 1−3 As part of these features, bio-gas sensors are currently studied for many kinds of bio-gases such as volatile organic compounds (VOCs) for biomarkers of disease or change of metabolism in human body. 4, 5 Especially nowadays, the epidemic of COVID-19 continues, and methods to perform fast, reliable and noninvasive tests before symptoms appear are being studied. 6 According to studies related to COVID-19, the virus does not generate gas by itself, but certain VOCs are secreted from infected cells. By setting such a bio-gas as a biomarker, there is expected a demand for the development of a bio-gas sensor that can determine whether or not it is infected by a virus. 7, 8 In the near future, bio-gas sensors, led by electronic nose, will take over the role and if such a wearable device for examination is applied to society, we can check our medical status or diseases at home without having to go to the hospital. 9−12 In this study, a device that allows individuals to diagnose quickly and simply using a bio-gas sensor using the olfactory diagnosis method, but without taking time to train a dog as bio-sniffer, was developed with the aim of miniaturizing the device to be a tentative wearable diagnosis device. For high gas sensitivity and selectivity, porous materials are used in this study. Porous materials have been widely studied, particularly for their gas adsorption and desorption characteristics. 13 A noteworthy example is the combination of zeolites and semiconductor-type gas sensors such as SnO 2 . 14 In most cases, the performance of these hybrid structures may exceed that of the individual devices. For instance, a phenomenon in which a low concentration of gas is adsorbed to a zeolite and desorbed into a high concentration has been found to lead to an enhancement of gas sensitivity and selectivity. 15 For skin gas sensing, such detection should be assumed to proceed at very low concentrations of VOCs. 16 A hybrid gas sensor array system was exposed to four types of bio-gases, that is, acetone, ethanol, acetaldehyde, and ammonia, using various mixtures of each individual gas, as shown in Table 1 . During this study, acetone, produced as a byproduct of fat metabolism, was selected as the target analyte owing to the metabolic information in acetone for use in healthcare applications when monitoring diet and exercise phases. 17, 18 Additionally, WO 3 has relatively high sensitivity and quick response time characteristics to acetone gas compared to other matching oxide semiconductors or gases. 19 For this reason, we fabricated a gas sensor using WO 3 as the main material to evaluate its sensing properties. In addition, it was confirmed in experi-ments that it behaves as very suitable analytes as a reference bio-gas for determining the gas identification ability of the electric nose. Since zeolites are extremely suitable as gas-concentrating porous materials, we are interested in investigating the sensing qualities of hybrid gas sensors composed of zeolite/WO 3 nanoparticles. 20, 21 In a previous study, it was confirmed that faujasite (FAU) zeolite has a suitable property as a gasconcentrating device because of its very high adsorption and desorption rates for acetone gas. 22, 23 In addition, we found that in a zeolite-combined hybrid gas sensor using WO 3 nanoparticles, the results depend on the properties of the zeolite, the detection of several gas molecules is inhibited, and only specific gas molecules are compressed. Further studies have shown that the gas adsorbed in the zeolite is desorbed only when a specific temperature and energy are applied, and selective compression and injection are possible. An important aspect of these hybrid structures is the fabrication process. The standard method involves completely covering the nanoparticle layer via a solution process, while continuously maintaining the hydrophilicity of the surface. In this way, by combining zeolite with a sensor made of hydrothermally grown WO 3 nanoparticles, the properties of the sensor can be completely changed, resulting in a compact and thin porous layer on the gas sensor. 24 Using the characteristics of such a hybrid sensor, a method for selectively detecting only a desired gas with high sensitivity by selectively combining several types of zeolites should be developed. 25, 26 Despite these promising results, the effective exploitation of these sensors is difficult owing to the characteristics of the gas detection. An individual hybrid gas sensor has very little response other than to a single gas, usually the target gas, thereby resulting in interference under different conditions. In other words, although one hybrid gas sensor may have high sensitivity to a single gas, the reliability of its detection is decreased. This problem can be avoided by combining and analyzing multiple gas sensors that have different characteristics into a single sensor array system. In this way, enabling a more accurate gas analysis based on various gas detection data is a key to this technology. 27, 28 In addition, a gas-sensor array has an advantage that it is possible to detect and discriminate unknown gases at an accurate concentration. 29, 30 Owing to these characteristics, gas sensor arrays are currently being studied for use in electronic nose applications, although the research field still remains focused on high gas concentrations. In this study, we investigated the sensor properties of an array of four WO 3 nanoparticles, each installed with a different zeolite. The functionalized nanoparticles were fabricated following a one-pot method using hydrothermal growth (NTT DOCOMO). 31 Three different zeolite materials are used to apply the compressive and sieve effects to organic gases like acetone and ammonia. The sensor array was tested by measuring the sensitivity to four volatile compounds, each representative of a chemical family of compounds: acetone, ethanol, acetaldehyde, and ammonia. Owing to its molecular diameter, each compound is expected to interact with the pores in the zeolites with a different blend of fundamental interactions, including van der Waals forces, hydrogen bonds, and coordination. The results indicate that these four compounds can be identified using an array of hybrid gas sensors, even when they are present at different concentrations and under mixed gas conditions. This suggests that hybrid gas sensors are a viable material for the development of highly sensitive and highly selective gas-sensor array systems. 2.1. Operating Temperature and Gas Separation Effect. We developed a single gas-sensing device with high selectivity using a hybrid layer structure. Additionally, through the single gas-desorption experiment of the hybrid gas sensor, the potential of distinguishment for multiple gases is confirmed by the experiment of zeolites that a specific gas desorbed at a specific temperature. To verify the effect of desorption, the result of measuring the gas desorption response that occurs while changing the heat applied to the hybrid gas sensor is conducted (see Figure S1 in Supporting Information). The temperature increase rate is set to 10°C/min to measure the sensitivity of the gas sensor. The original response signal of the hybrid gas sensor should show the sensitivity change of the sensor itself according to the change of temperature. However, in order to make the desorption effect clear to see, we intentionally removed the effect of free-electron generation according to the temperature change of the semiconductor in Figure S1 . Through the result, it is clearly identified that the acetone gas is desorbed in the 390HUA zeolite (Tosoh Corporation) at 189°C. Nevertheless, in the process of offsetting the sensitivity value along with the temperature change from the raw sensitivity, noise is generated in the final sensitivity value due to the sensitivity value mismatch in the interval where desorption does not occur. Beyond the compression effect of a single gas, the reason why the developed device itself exhibits high selectivity for multiple gases is that it indicates a distinctive feature in the dependence of the desorption temperature due to the difference in activation energy between the zeolite and each gases. As an example, a graph showing the desorption peak depend on temperature for each gas of a hybrid gas sensor using 390HUA is shown in Figure 1 . The temperature rise rate is set to be the same as the measurement conditions, as shown in Figure S1 . The three bio-gases (acetone, ethanol, and acetaldehyde) have different activation energies for 390HUA. Since the desorption characteristic of the hybrid gas sensor is dependent on the temperature, the activation energy of each zeolite and gas can be obtained by substituting the temperature value at which the desorption response peaks into the redhead equation 33 where E d is activation energy of desorption, R is the gas constant, T p is the temperature at which the gas component is the most desorbed, v 1 is the frequency factor, and β is the heating rate. Each value obtained from the Readhead equation is shown in Table 2 . 130 kJ/mol (189°C) in acetone, 118 kJ/mol (139°C ) in ethanol, and 108 kJ/mol (113°C) in acetaldehyde. The desorption temperature deduced from the calculated activation energy required for analyte desorption and the actual desorption temperature value matched well, and it is confirmed that the Redhead equation well reflected the desorption characteristics of the hybrid gas sensor using zeolite. The advantage of operating the device at the desorption temperature obtained from the above equation is that high gas sensitivity can be obtained by excluding moisture. First, by maintaining the operating temperature above 100°C, it is possible to exclude moisture from the surface of sensors, which accounts for most of the skin gas ratio, from adsorbing to the gas sensor surface. By excluding water molecules, there is an advantage of obtaining higher response ratio with less noise of the target gas. In addition, by using the temperature dependence of the hybrid gas sensor due to the characteristics of zeolite, the sensitivity can be optimized by maintaining the operating temperature at desorption temperature. As such, it is confirmed that the hybrid structure worked well, and it is proved that it has the sensing ability to identify multiple gases. As an objective evaluation, this method is less effective in gas selectivity than a gas sensor using a chemical receptor, but has an great advantage against deterioration and there is potential to increase the identification resolution for gases using multi-sensors by adding machine learning to it. The 390HUA/WO 3 hybrid gas sensor showed the highest sensitivity at 190°C calculated from the desorption activation energy for the target analyte, acetone gas (see Figure S2 in Supporting Information). In this way, it is possible to obtain improved sensitivity by understanding the difference in activation energy between each zeolite and analyte and applying the optimum operating temperature. By using the different characteristics in the desorption temperature to each gases of the hybrid gas sensor, it has become the basis for obtaining higher sensitivity and high selectivity to the target gas by controlling the operating temperature of the sensor. 2.2. Gas-Sensing Properties. For gas detection measurement, the fabricated gas sensors are exposed to four kinds of bio-gases, namely, acetone, ethanol, acetaldehyde, and ammonia. The properties of the skin gases which are emanated from human skin, as shown in Table 1 , and their molecular characteristics are related to the performance of the gas sensor, including the sensitivity and selectivity. In particular, it is important to understand the characteristics of each gas and the zeolite because a hybrid gas sensor is fabricated using different gas properties based on the adsorption/desorption rates, which depend on the pore size of the zeolite and/or the SiO 2 /Al 2 O 3 ratio. As Tables 1 and 3 indicate, the similar size between the pores of the zeolite and the gas molecules increases the adsorption rate through the van der Waals force. Therefore, the pore and molecular sizes are the most important factors, although the adsorption/desorption rates of the gas molecules from the hydrophilicity/hydrophobicity of the zeolite vary significantly. 34 Although the molecule size of acetone is 4.6 Å, and the molecule size of ethanol is 4.3 Å, the dipole moment of the ethanol molecule is comparably low owing to the low property of 1.6D, and thus, a low desorption rate occurs in 390HUA with a high hydrophobicity at a SiO 2 /Al 2 O 3 ratio of 500, which lowers the sensitivity in hybrid gas sensors by approximately 25% at the same gas concentration. An aldehyde molecule has a dipole moment of 2.7D, which is almost the same as that of acetone; however, owing to the difference in molecular size, the FAU-type zeolite having a pore size of 7.4 Å has a reduced adsorption rate of 40% compared to acetone, and ammonia gas is not adsorbed into the FAU-type zeolite owing to its small ratio of the zeolite is 100 or 10, the desorption rate of acetone is reduced by 97% owing to the increased hydrophilicity, and consequently, it did not function as a gas-concentrating material. As a result, the sensitivity of the sensor to ethanol is the highest with an increase of 60%. As described above, the pore size of the zeolite and the SiO 2 /Al 2 O 3 ratio have a significant influence on the adsorption/desorption properties depending on the characteristics of the gas, and in this way, the target analyte can change. 35 Using these properties of zeolites and gas molecules, various gas sensors with different gas sensitivities and selectivity characteristics can be set. As a condition of the gas concentration for measurement purposes, when reducing the gases, the concentration is approximately 1−30 ppb/min (non-drinking state), and the ammonia exhibited an emanation rate of 0.38 ppb/min. The minimum measurable concentration of WO 3 nanoparticles reached 13 ppb, and thus, at least a 15-min gas enrichment time is required. Therefore, the measurement experiment is carried out after each gas is filled with 20, 40, 60, 100, 200, 400, 600, 800, and 1000 ppb in a closed space and had a 15min concentration waiting time. Figure 2 shows scanning electron microscopy (SEM) and transmission electron microscopy (TEM) images of the dispersed WO 3 nanoparticles. As shown in SEM image of Figure 2a , WO 3 nanoparticles form a very dense layer, and the surface roughness is formed over a range of 10 to 100 nm. As a result, the gas reactivity of WO 3 nanoparticles layer is increased by increasing the surface area. The corresponding TEM image confirms existence of WO 3 nanoparticles, as demonstrated in Figure 2b . The average particle diameter is approximately 7.3 nm, which is smaller than the doubled Debye length of WO 3 used to obtain a grain-control model caused by a complete nanoparticle, which is covered by a depletion layer. 36 Furthermore, the particles are characterized by rounded borders and irregular shapes. As such, WO 3 nanoparticles having a diameter of less than 10 nm can have improved gas sensitivity. As shown in Figure 3 , from the surface AFM image of the zeolite (390HUA) deposited on the WO 3 nanoparticle layer, it can be seen that the surface of the zeolite has a high roughness. The roughness is formed between about 10 to 100 nm, which makes the surface area increase and give the advantage that it is easier to react with gas. As the above, the improved but fixed characteristics of WO 3 nanoparticles can extend the field and the number of cases by combining with zeolites. When WO 3 having such a defined characteristic is used in combination with various types of zeolites, it can lead to diversity of reactions to gases, which is well shown in Figure 4 . These graphs show the response curves of the four sensors to the four volatile compounds. Since each hybrid sensor is characterized by a different gas saturation pressure, the concentrations in the gaseous phase are different. The concentration of gas is within the interval between 20 ppb and 1 ppm. The gas reaction characteristic of the gas sensor using pure WO 3 nanoparticles without zeolite is shown in Figure 4a . When comparing with the gas reaction characteristics of the hybrid gas sensors, as demonstrated in Figure 4b , 390HUA/WO 3 indicates seven times higher sensitivity amplification characteristics for 20 ppb of acetone. In the same way, the sensitivity of 4A/WO 3 and 5A/WO 3 (4A and 5A are fabricated by Tosoh Corporation) increased by 2.8 times and 2.2 times, respectively, at 50 ppb of ammonia. In this part, it is worth noting that the acetone gas reaction of the 4A/ WO 3 and 5A/WO 3 hybrid gas sensors is lower than that of the WO 3 gas sensor. Through this, it can be seen that by applying the zeolite layer, a characteristic capable of suppressing the sensitivity by using the sieve effect to the non-target gas is afforded. As found in a previous study, a common gas sensor behavior was shown when zeolite was applied. This indicates that the difference in the material composition ratio (SiO 2 /Al 2 O 3 ) occurs between the zeolites, proving that the characteristics of the gas adsorption and desorption by the composition ratio are important factors in a gas sensor. The peculiar response to acetone may be explained by considering that it is possible to control the hydrophobicity by changing the ratio of SiO 2 and Al 2 O 3 . The higher the value of SiO 2 /Al 2 O 3 , the more hydrophobic the zeolite is, and the lower the value, the more hydrophilic it is. This occurs because, when the value of SiO 2 / Al 2 O 3 is low, the amount of Al increases; however, because Al is negatively charged, cations are required to keep it electrically neutral. This cation becomes a permanent dipole and attracts water through a dipole interaction, making it hydrophilic. As described above, the hydrophilicity of the zeolite can be adjusted to facilitate the adsorption or desorption of acetone, thereby changing the characteristics of the gas sensor. The interaction with other volatile compounds increases the resistance. This unintended interaction with the gas causes a significant decrease in the sensitivity and selectivity of the gas sensor. To prevent such a phenomenon, by selectively combining zeolite with a gas sensor, it has the effect of blocking unintended gases from the gas sensor surface through a sieve effect. However, there has not been much research into which zeolites can interact with what kind of gas to expect a sieve effect. Therefore, more studies are necessary to clarify the interaction between the volatile compound and zeolite adsorption/desorption characteristics. In this study, we investigated the interaction between three kinds of hybrid gas sensors with different zeolites (390HUA, 4A, and 5A) and four types of bio-gases. In the 390HUA hybrid sensor, acetone and ethanol do not saturate well and maintain relative linearity even above 200 ppb, whereas acetaldehyde and ammonia almost saturate at http://pubs.acs.org/journal/acsodf Article 200−400 ppb. Consequently, the linearity may also be due to the local linearization of a more general non-linear isotherm in bio-gas sensor applications owing to its extremely low gas concentration. The sensitivity of the sensors to the four compounds is calculated as the derivative of the sensor response (R air /R) with respect to the concentration of various ranges of parts per billion to part per million. All sensitivities are compared in Figure S3 . As discussed above, the sensors are characterized by a common behavior with respect to the set of VOCs. Because the sensitivity to acetone is explicitly high, it is interesting to evaluate the sensitivity to the other VOCs with respect to acetone. The ratio of sensitivity accounts for the selectivity of the sensor with respect to acetone. Figure S3b shows the sensitivity to acetone normalized to the absolute value. Excluding the sensitivity of 390HUA/WO 3 to acetone, the ratio of sensitivity is eight times larger than that of the other compounds. Moreover, the improvement of sensitivity to acetone is evident compared to the sensor using only WO 3 . The LTA-type (4A, 5A) zeolite has low sensitivity to gases other than ammonia because of its pore size and material properties. It is important to note that each sensor shows a different pattern of relative sensitivity. This feature is particularly important for sensor arrays because the sensor's http://pubs.acs.org/journal/acsodf Article capability to discriminate compounds relies on these differences. Acetone is the target analyte in this research, and ethanol, acetaldehyde, and ammonia are considered as interferents. Ethanol and acetaldehyde are chosen as interferents since the sensitivities of most oxide semiconductors to these analytes are known to be high. In addition, these analytes are a type of gas with a high concentration in bio-gas. The interferents mentioned hereafter and in the figure captions refer to a mixture of 20 ppb ethanol + 20 ppb acetaldehyde in Figure S3 . To assess the ability of the array to function as a gas sensor in a real situation, it is tested in air. The sensitivity of the array to each analyte individually and to mixtures of the target analytes (acetone) and the interferents, as shown in Figure S3 , is tested systematically. The composition of the test analytes and the sensitivity of each unit to these test analytes are also verified (see Table S1 in Supporting Information). When testing various types of gases, each of which is in a 20 ppb mixture containing acetone, 390HUA/WO 3 exhibited a higher response (see Figure S3e −h). For all gas mixtures, the sensitivity of 390HUA/WO 3 is 3.47, whereas those of 4A/ WO 3 and 5A/WO 3 are 1.11 and 0.870, respectively. Likewise, when the test mixture contained only pure ammonia gas, 4A/ WO 3 showed a strong response (see Figure S3d ) with a sensitivity of 0.43 at a 20 ppb level. This value is approximately 25 times higher than that of 390HUA/WO 3 . For ethanol (see Figure S3a ), the sensitivity of 390HUA/WO 3 is 2.41, which is approximately 26 times higher than those of 4A/WO 3 and 5A/ WO 3 . As interferents, for ethanol (see Figure S3b ), the sensitivity of 390HUA/WO 3 is 0.74, which is seven times higher than those of the LTA-type hybrid sensors. Finally, for acetaldehyde (see Figure S3c ), LTA-type hybrid gas sensors have 13 times higher sensitivity than 390HUA/WO 3 . These changes in the gas sensitivity to only the zeolite show that the gas selectivity is caused by a large change in the adsorption/ desorption characteristics owing to the pore size and material composition ratio of the zeolite, thereby changing the energy required for adsorption and desorption of the gas. Namely, 390HUA/WO 3 exhibited good selectivity for acetone and ethanol, whereas 4A/WO 3 and 5A/WO 3 showed good selectivity for ammonia and acetaldehyde, respectively. When the array is exposed to a mixture of the four analytes (see Figure S3h ), the sensitivity of each unit is over approximately 1, which is much higher than the values observed when the array is exposed to only the interferents (see Table S1 ). In summary, when the test gas contained acetone, ethanol, acetaldehyde, and ammonia, or any combination thereof, the corresponding unit(s) in the array exhibited marked responses. Hence, this array achieved a simultaneous detection of multiple VOCs, that is, acetone, ethanol, acetaldehyde, and ammonia. Furthermore, the number of units in such a sensor array could be added or removed as desired. In general, for the use of wearable applications, a sensor array containing two units for acetone detection and one for ammonia detection would be sufficient. 2.3. PCA and Gas-Sensor Array. A principal component analysis (PCA) is used to investigate the collective behavior of the three sensors. For the scope, the sensor responses are arranged in a matrix, and the PCA is calculated on an autoscaled (zero mean and unitary variance) matrix. 37 The PCA is applied to the whole set of measurements, and each concentration is measured three times. In this way, the PCA results enable an appreciation of the separation of the different VOCs with respect to the dispersion of the sensor signals. http://pubs.acs.org/journal/acsodf Article Figure 5 shows the scores and loadings of the first three principal components of the auto-scaled data matrix. The variance explained in each principal component is indicated in the title of each plot of figures. As found with the correlated sensors, the first principal component carries a large part of the total variance (93%), and all sensors almost equally contributed to PC1. By analyzing these data, we can find out the intrinsic correlation between the sensors (analyzed by referring to the pattern in Figure S3 ) and the correlation between the sample properties. Each regression line is calculated to have a minimum variance of gas measurements on each hybrid gas sensor, which allows us to determine which sensor played a crucial role in the distinction between gases. In this experiment, the last contribution is attributable to the varying concentrations of the compounds. Because the sensor responses show almost linear characteristics at low concentrations of under 100 ppb (see Figure 4 ), the variable concentration introduces a strong correlation in the data. The influence of the concentration is clearly visible in the first principal component, which is strongly correlated with the concentration regardless of the type of compound. The correlation of the scores with the concentration is attenuated in the second and third principal components. For these components, the contribution of the sensors is also rather sparse. The individual characteristics of each sensor, which are not clearly distinguished in the sensitivity comparison, can be easily understood by taking into account the main components. The second principal component graph shows that the divergence between 4A/WO 3 and the other sensors. On the other hand, the first principal component graph shows that the variance tendency of ammonia is extremely prevalent with other reducing gases. Although the divergence value of each gas sensor shows a very slight difference, it is clear that 390HUA/ WO 3 is the most important factor that distinguishes ammonia in PC1 from the tendency of dispersion. In addition, the chemical disparity between the reducing gas is captured by the first principal component. The second principal component graph shows a more dramatic difference. The variance in ethanol tends to be extremely large compared to that of acetone, meaning that it is an interval where ethanol and acetone can be distinguished based on the detection of 4A/ WO 3 , which has the largest PC2 value. In the third PCA, the graph shows a variance data set of 5A, which represents a relatively high adsorption/desorption response to ammonia and provides higher ammonia selectivity compared to 4A, which has little reactivity to acetaldehyde. Finally, in the third PCA, the graph shows that 5A, which represents a relatively high adsorption/desorption response to ammonia, provides a higher ammonia selectivity compared to 4A, which has little reactivity to acetaldehyde. By interpreting each of these main component analysis graphs, we can see that the first principal component that has a decisive role in separating ammonia from the other gases is the 390HUA/WO 3 sensor, and that acetone and ethanol each play a major role by taking advantage of the difference in variance. In addition, the graph shows that the variance data of 5A are analyzed using the gas response of 4A to enable a third PCA distinguishing ammonia from acetaldehyde. The zeolite-concentrating effect of the VOCs requires removing the influence of the concentration on the sensor responses. This will limit the sensor response to only the qualitative characteristics of the measured compounds. The qualitative segregation from quantitative information has been a well-known problem since the original studies on sensor arrays. In the case of linear sensors, this can be achieved through a simple normalization of the sensor response. 38 This consists of dividing the response of each sensor by the sum of the responses of all sensors of the array. In the case of an array of N linear sensors, the response of the i-th sensor (ΔR ij /R i ) to the j-th compound at concentration c (c j ) is where S ij is the sensitivity of the i-th sensor to the j-th compound. The normalized sensor response, (ΔR/R)*, is calculated using the following equation In practice, the signal of the i-th sensor corresponds to the weighted sensitivity. By using PCA, the matrix of the normalized data is plotted, as shown in Figure 6 as a PCA bi-plot result. Here, the scores and loadings are shown simultaneously. Through these plots, it also allows one to study not only clustering but also the relationship between data and sensors. As shown in Figure 6 , the data are related to each analyte cluster together, disregarding the concentration. Owing to its peculiar sensor responses, acetone is plotted aside. The loadings of 390HUA/ WO 3 , which is the most selective sensor to acetone gas, point toward the region where the acetone is plotted. The other compounds are gathered together, but without overlaps. Finally, the loadings of 4A/WO 3 and 5A/WO 3 point toward the ammonia group. This is not surprising, considering that 4A/WO 3 shows the largest sensitivity to ammonia. The PCA of normalized data matches the sensitivity pattern of its consistency and tendency. This indicates that even if these sensors show similar characteristics, they are quite different to distinguish the tested compounds. Gas sensor arrays have been shown to be an effective electronic nose application in many different fields, from food analysis to medical diagnosis. 39 These sensors have been almost exclusively based on mass or chemical transducers. However, from an electronic perspective, the implementation and integration of gas-sensitive resistors are the most straightforward in achieving a low gas concentration. An important advancement in the exploitation of the sensing properties of hybrid structures using zeolites can be expected if they can be matched with bio-gas sensing applications. The combination of zeolites with a conductive material is a practical approach for bio-gas sensors. This study investigated this concept, considering the behavior of hybrid gas sensors made of WO 3 nanoparticles combined with three kinds of zeolite, that is, 390HUA (FAUtype) and 4A and 5A (LTA-types). All these sensors show a peculiar sensitivity to acetone (a reducing gas with a strong electron donor), but also a different pattern of interactions with the other tested volatile compounds. These differences appear to be subtle when sensors are individually compared but are sufficiently large in identifying different volatile compounds regardless of the variation in concentration. This study focused on a demonstration in which a hybrid gas sensor array is obtained by combining different types of zeolites, which are typically used as gas concentrators, when deposited onto the surface of a WO 3 nanoparticle layer. However, it remains to be demonstrated whether the change in resistance can efficiently transduce all the adsorption and desorption interactions that can occur between the zeolites and bio-gases, which are composed of volatile compounds, and whether such an array can actually be used to discriminate between real bio-gases. Through this study, the bio-gas sensor using WO 3 nanoparticles could contain high sensitivity characteristics in one compact device through a hybrid structure, and it became possible to identify multiple gases. The countermeasures for such miniaturization and gas diversity are expected to contribute in the aspect of daily monitoring of healthcare and disease prevention. WO 3 nanoparticles having a diameter of 7.3 nm are prepared to fabricate a hybrid gas sensor. A metal mask made of stainless steel (thickness of 0.1 mm) is used, and a metal mask is brought into tight contact with a sapphire substrate (10 mm × 10 mm). Comb-shaped Cr and Pt electrode films are then formed at a thickness of 1.5 and 38.5 nm, respectively, using a sputtering apparatus. A solution is prepared using 100 mg of WO 3 nanoparticles dispersed in 10 mL of de-ionized water, and 5 μL of the prepared solution is dropped onto a combshaped electrode and vacuum-dried in a desiccator for approximately 30 min to obtain a WO 3 nanoparticle film. After the solution is generated, annealing is applied in an electric furnace at 100°C for 1 h and again at 400°C for 1 h in an air atmosphere. The fabrication of the gas sensor unit is then completed, creating a hybrid structure. As a hybrid-structured gas sensor, three kinds of zeolites were deposited on WO 3 nanoparticle films to achieve a molecular sieve effect and concentrator functionality on the WO 3 layer. LTA (4A, 5A)-and FAU (390HUA)-type zeolites are used at the uppermost layer to create a hybrid structure. A solution is processed to form a zeolite layer, and a powdered solution of zeolite is prepared at a concentration of 18% in deionized water as a solvent. In the case of the 390HUA zeolite, the SiO 2 /Al 2 O 3 ratio is 500, which shows a strong hydrophobicity. Therefore, the surface of the sensor area must be subjected to oxygen plasma surface treatment to improve the surface adhesion. For the surface treatment process, oxygen plasma treatment is carried out on a plasma generator for 15 min with a pressure of about 0.2 Pa and power of 100 W to form high surface hydrophilicity. Without this process, the cohesiveness between nanoparticles and the bonding of zeolite/nanoparticle interface are very poor; although the electrical resistance of the gas sensor is changed by damage occurred during the surface treatment, the sensitivity to the bio-gas is not significantly changed between the periods before and after the process. It is evaluated how much surface damage caused by the plasma treatment affects sensing performance by comparing between the WO 3 gas sensor with the hydrophilic treatment and the untreated one; it is verified that the characteristics difference of these sensors are less than 5% (see Figure S4 in the Supporting Information). Subsequently, to form a zeolite layer, zeolite solutions are dropped onto the WO 3 nanoparticle layer and air-dried, followed by baking at 500°C for 1 h. The fabricated device structure is shown in Figure 7 . We prepare a gas of an appropriate concentration to evaluate the gas sensor. Gas concentration is measured using gas chromatography in order to know the concentration of the atmosphere to which the gas sensor is exposed (see Figure S5 in the Supporting Information). For the next step, to get electrical characteristics of gas sensors, electrical measurement methods are divided into two ways. In the case of a hybrid gas sensor, a waiting time is required for the process of concentrating the gas on the zeolite layer and the one of adjusting the temperature each time, an impedance measurement method that cannot measure a real-time measurement and know a precise impedance value is more suitable to this kind of device. The measured impedance value is analyzed as an ideal equivalent circuit by nanoparticle modeling analysis, and parameters such as resistance, capacitance, and reactance are extracted to digitize the electrical state of the gas sensor. From the extracted parameters, the response of the gas sensor ACS Omega http://pubs.acs.org/journal/acsodf Article is calculated using the following equation: response = R a /R (R a : resistance under the air, R: resistance under the target gas). 32 Furthermore, in this study, the sensor sensitivity is defined as follows for the convenience of calculating the sensitivity comparison between gas sensors: sensitivity = (R a − R)/R. Additionally, rising time and falling time are measured as parameters of response time, which is defined as the time measured between 10 and 90% of the sensitivity interval which is used effectively in electrical signals. Upon the definitions, as can be seen in Figure 8a for real-time measurement (measuring resistance), the WO 3 gas sensor represents good repeatability with an error of less than 3% in the continuous measurements toward 50 ppb of acetone gas. In addition, in Figure 8b , it is confirmed that the WO 3 gas sensor exhibits very fast response and recovery curve characteristics through rising and falling time (3.6 and 0.3 s, respectively) measurements at the same condition. ■ ASSOCIATED CONTENT The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsomega.1c01435. Comparison of the effect of surface treatment; measurement methods; response of the hybrid gas sensor according to temperature change; sensitivity variation according to the temperature change; response of the gas sensor array to various gases; and sensitivity of three kinds of hybrid gas sensors (PDF) Using Electrochemical Enzymatic Sensors for Detection of Biomarkers in Human Breath Gas-Phase Biosensors (Bio-Sniffer and Sniff-Cam) for Volatile Chemicals Two-Step Investigation of Lung Cancer Detection by Sniffer Dogs Graphene and metal organic frameworks (MOFs) hybridization for tunable chemoresistive sensors for detection of volatile organic compounds (VOCs) biomarkers. Carbon Prediction of Breast Cancer Risk with Volatile Biomarkers in Breath Scent dog identification of samples from COVID-19 patientsa pilot study Multiplexed Nanomaterial-Based Sensor Array for Detection of COVID-19 in Exhaled Breath Sniffer Dogs as a Screening/Diagnostic Tool For COVID-19: A Proof of Concept Study Real-Time, Personalized Medicine through Wearable Sensors and Dynamic Predictive Modeling: A New Paradigm for Clinical Medicine Sensor Array Optimization and Discrimination of Apple Juices According to Variety by An Electronic Nose Response of the WO 3 gas sensor toward 50 ppb acetone gas for (a) Repeatability test and (b) rising and falling time Ultra-Low-Power Smart Electronic Nose System Based on Three-Dimensional Tin Oxide Nanotube Arrays Exhaled Breath Analysis Using Electronic Nose and Gas Chromatography−Mass Spectrometry for Non-Invasive Diagnosis of Chronic Kidney Disease, Diabetes Mellitus and Healthy Subjects Adsorption and Desorption Characteristics of Hydrophobic Hierarchical Zeolites for the Removal of Volatile Organic Compounds Zeolite Membranes for Highly Selective Formaldehyde Sensors Selective Detection of Methanol by Zeolite/Pd-WO 3 Gas Sensors Real-Time Monitoring of Skin Ethanol Gas by a High-Sensitivity Gas Phase Biosensor (Bio-Sniffer) for The Non-Invasive Evaluation of Volatile Blood Compounds Breath Sensors for Health Monitoring ZnO-CuO Core-Hollow Cube Nanostructures for Highly Sensitive Acetone Gas Sensors at the ppb Level Breath Acetone Sensors as Non-invasive Health Monitoring Systems: A Review Biomimetic Synthesis of Zeolitic Imidazolate Frameworks and Their Application in High Performance Acetone Gas Sensors Multicore−shell nanocomposite formed by encapsulation of WO 3 in zeolitic imidazolate framework (ZIF-8): As an efficient photocatalyst Adsorption Materials for Volatile Organic Compounds (VOCs) and The Key Factors for VOCs Adsorption Process: A Review Hydrocarbon Detection with Metal Oxide Semiconducting Gas Sensors Modified by Overlayer or Admixture of Zeolites Na-A, H-Y and H-ZSM-5 Sulfur Dioxide Gas-Sensitive Materials Based on Zeolitic Imidazolate Framework-Derived Carbon Nanotubes Quantitative and Qualitative Analysis of Multicomponent Gas Using Sensor Array Least Squares Neural Network-Based Wireless E-Nose System Using An SnO 2 Sensor Array A Gas Sensor Array for The Simultaneous Detection of Multiple VOCs Situ Localized Growth of Ordered Metal Oxide Hollow Sphere Array on Microheater Platform for Sensitive, Ultra-Fast Gas Sensing WO 3 Nanocubes: Hydrothermal Synthesis, Growth Mechanism, and Photocatalytic Performance High Performance Ethylene Sensor Based on Palladium-Loaded Tin Oxide: Application in Fruit Quality Detection Thermal Desorption of Gases Ultratrace Measurement of Acetone from Skin Using Zeolite: Toward Development of a Wearable Monitor of Fat Metabolism Discrimination Among Gas Translation, Surface and Knudsen Diffusion in Permeation through Zeolite Membranes High Sensitivity and Selectivity Chlorine Gas Sensors Based on 3D Open Porous SnO 2 Synthesized by Solid-State Method Selective Detection of Target Volatile Organic Compounds in Contaminated Humid Air Using a Sensor Array with Principal Component Analysis High-Performance Gas Sensor Array for Indoor Air Quality Monitoring: The Role of Au Nanoparticles on WO 3 , SnO 2 , and NiO-Based Gas Sensors Sensors for Enhanced Detection of Acetone as a Potential Tool for The authors declare no competing financial interest. The authors thank Dr. Yuki Yamada of Research Laboratories, NTT DOCOMO Inc. for their help in developing the gas sensors. This work was partially supported by the Basic Research Grant (Hybrid AI) of the Institute for AI and Beyond at The University of Tokyo and JSPS KAKENHI grant number 20H05651.