Effect of Working Atmospheres on the Detection of Diacetyl by Resistive SnO2 Sensor

Nanostructured metal oxide semiconductors (MOS) are considered proper candidates to develop low cost and real-time resistive sensors able to detect volatile organic compounds (VOCs), e.g., diacetyl. Small quantities of diacetyl are generally produced during the fermentation and storage of many foods and beverages, conferring a typically butter-like aroma. Since high diacetyl concentrations are undesired, its monitoring is fundamental to identify and characterize the quality of products. In this work, a tin oxide sensor (SnO2) is used to detect gaseous diacetyl. The effect of different working atmospheres (air, N2 and CO2), as well as the contemporary presence of ethanol vapors, used to reproduce the typical alcoholic fermentation environment, are evaluated. SnO2 sensor is able to detect diacetyl in all the analyzed conditions, even when an anaerobic environment is considered, showing a detection limit lower than 0.01 mg/L and response/recovery times constantly less than 50 s.


Introduction
Gas sensors are of great interest due to their numerous applications and the possibility for a real-time analysis of several analytes [1][2][3]. Among these, resistive gas sensors exhibit attractive advantages compared with other gas sensors, such as fast and accurate gas detection, flexibility, low cost and small size [4]. The development of high-performance resistive gas sensors requires suitable gas-sensing materials in terms of both physical and chemical properties, such as morphology, crystalline structure, specific surface area and active sites. Since, as is well known, these characteristics all affect the performance of gas sensors, exploring and developing innovative materials has received attention in scientific research in recent years [5]. Attention is concentrated on the development of nanostructured materials, endowed with better sensing properties if compared to the same bulk material, such as carbon-based materials (in the form of carbon dots, nanotubes and graphene) [6][7][8][9], polymeric fibers (hybrid nanofibers) [10,11] and metal oxides semiconductor (MOS, as nanospheres, nanowires and nanosheets) [12][13][14][15][16][17]. Indeed, the large surface areas of nanomaterials ensure more active sites, enabling fast charge transfers and efficient gas-sensitive reactions [4].
Due to their widespread use in a variety of applications, in the recent years timeconsuming, complex, and expensive traditional techniques have been substituted by these types of sensors. It has been especially evident in the field of volatile organic compound (VOC) detection, where traditional techniques, such as chromatography-mass spectrometry [18][19][20][21], generally lack ease of use, need specialized personnel and elaborate protocols, and are characterized by high cost of the devices and insufficient flexibility. When early warning and quality monitoring applications are requested, resistive sensors are valuable alternatives. Indeed, in some specific applications, VOCs, such as diacetyl, need to be continuously monitored [22]. During the alcoholic fermentation processes of wine, beer and distilled beverages, diacetyl is naturally produced in small quantities, and it confers a characteristic butter-like aroma [23][24][25][26][27]. However, high diacetyl concentration (>5 mg/L) imparts an undesirable flavor indicating an alteration in the production or storage process. The time-monitoring of diacetyl concentrations can contribute to identifying and characterizing the quality of products [22]. Moreover, during this process, the so-called alcoholic fermentation, carbon dioxide and ethanol are produced too, and the influence of the atmosphere composition was reported to be fundamental in diacetyl detection and monitoring [28]. Till now, few works reported the detection of gaseous diacetyl mainly by means of complex array sensors [27,29,30]. Itoh et al. used variously-doped SnO 2 and CeO 2 to detect diacetyl among other volatile compounds [27], whereas PPy-V 2 O 5 and PPy-ZnO nanocomposite fibers were successfully used by Pirsa et al. to discriminate the presence of diacetyl in fermented products [29]. Bailey et al. used an array of conducting polymers mounted on an electronic chip to discriminate between beers and to recognize the presence of volatile compounds, such as diacetyl [30].
In this work, tin oxide (SnO 2 ) is used to develop a high-performance resistive MOS sensor for the detection of VOCs [17,31], diacetyl in particular, thanks to the peculiar electrical properties of its microsphere structure [13]. The detection of diacetyl is performed by sampling the head space above the liquid solutions at different diacetyl concentrations (0.2-3.2 mg/L). The sensor has been tested using both aqueous and alcoholic (5% ethanol) diacetyl solutions in different carrier/regeneration gases, able to reproduce the typical alcoholic fermentation environment, rich in carbon dioxide (CO 2 ) and poor of oxygen. The revealed performance shows the promising ability of the of SnO 2 sensor to discriminate between different diacetyl concentrations in anaerobic atmospheres and in the contemporary presence of ethanol, generally characterized by a strong signal, identified as the major constituent of the headspace of alcoholic beverages [32]. Response and recovery time are also considered for a better investigation.

Tin Oxide Preparation and Charactetrization
A facile hydrothermal procedure was used to prepare SnO 2 powder, as reported elsewhere [13]. In detail, SnCl 2 (II), used as the tin oxide precursor, was solubilized in ethanol (7.57 g/L). In order to fully dissolve SnCl 2 , the mixture was firstly sonicated for 20 min and then it was transferred into a Teflon-lined stainless-steel autoclave at 200 • C for 6 h and 40 min. The solution was finally cooled down to room temperature. The obtained yellow solid was collected by centrifugation and washed in ethanol until the total removal of chloride ions. After ethanol washing the sample was dried at 80 • C overnight. The sample preparation procedure is schematically depicted in Figure 1.

Sensor Preparation and Testing
The sensing film was realized using a tin oxide/ethanol mixture sonicated for 15 min, in order to obtain a homogeneous paste, and deposited on an alumina planar substrate (6 mm × 3 mm) supplied with interdigitated Pt electrodes on the front side and a heating element on the back side [33]. The paste was dried at room temperature and finally annealed at 80 °C for 1 h. Before measurements, in order to improve the deposited film stability, the sensor was conditioned in air at 400 °C for 2 h. Tests were performed by posi- The sample was investigated using X-ray diffraction (XRD) analysis in the 2θ range from 10 • to 65 • (Cu Kα λ = 1.54056 Å) using steps of 0.02 • and a count time of 5 s per step, with a temperature effect evaluation within 25-500 • C by using a PANalytical Empyrean Appl. Sci. 2022, 12, 367 3 of 10 diffractometer equipped with an Anton Paar heat chamber. Peaks were attributed in accordance with the Crystallography Open Database (COD). The microstructure's morphology was studied by scanning electron microscopy, SEM, (Phenom Pro X).

Sensor Preparation and Testing
The sensing film was realized using a tin oxide/ethanol mixture sonicated for 15 min, in order to obtain a homogeneous paste, and deposited on an alumina planar substrate (6 mm × 3 mm) supplied with interdigitated Pt electrodes on the front side and a heating element on the back side [33]. The paste was dried at room temperature and finally annealed at 80 • C for 1 h. Before measurements, in order to improve the deposited film stability, the sensor was conditioned in air at 400 • C for 2 h. Tests were performed by positioning the sensor in a testing cell and flowing a total gas stream of 100 sccm. To assess the different working conditions' effects on the performance of the SnO 2 sensor, different pure gas atmospheres, namely Air, N 2 and CO 2 were evaluated. Diacetyl was chosen as the target analyte, both in aqueous and in alcoholic (5% ethanol) solutions. Analyte vapor was obtained by bubbling carrier gases in the solution that was maintained at 20 • C during the tests. All gas fluxes were measured by computer-controlled mass flow meters, and humidity was monitored and controlled to oscillate between 5% and 10%. The sensors' resistance data were collected in four-point mode by an Agilent 34970A (Santa Clara, CA, USA) multimeter, while sensor temperature was controlled by using a dual-channel power supplier instrument (Agilent E3632A, Santa Clara, CA, USA).
The sensor working temperature was fixed at 200 • C. Sensor response was defined as R/R 0 when the sensor showed an n-type response and the sensor resistance R measured in the presence of the reducing analyte gas decreased in respect to the sensor resistance R 0 of the sensor exposed to the carrier flow. Differently, when the sensor showed a p-type response and the sensor resistance R in the presence of the same analyte gas increased, the response was defined as R 0 /R [34]. The response and recovery time of the sensor were defined as the time needed for the sensor to reach 90% of its saturation limit after the exposure to the analyte gas and as the time needed for the sensor to reach the 10% of its original resistance value once the target gas was switched off and the sensor was exposed to the carrier gas only, respectively [14].

SnO 2 Powder Characterization
The crystalline microstructure of the prepared SnO 2 was studied by XRD analysis, as shown in Figure 2a. The main characteristic peaks are centered at 26.7 • , 34.2 • , 37.8 • , and 51.9 • and correspond to the (110), (101), (200), and (211) SnO 2 crystal planes, respectively. The peaks appeared broad and with weak intensity, thus resulting with small average crystallite size and low crystallinity. This latter one was ascribed to the presence of defects on the material surface, which greatly affect the reactive sensing sites and electronic structure of gas-sensing material, influencing the gas sensing properties [35]. XRD spectra acquired at increasing temperatures, from 25 • C up to 500 • C, proved the high stability of the sample, being all the registered patterns superimposable.
The morphology of the powder was studied by SEM analysis. Figure 2b shows that sample is composed of homogeneous small particles of spherical form with size distribution ranging from 0.8 to 2.1 µm. A compact oxide thin film was used as the sensing layer.

Gas Sensor Measurements
First, to study the behavior of the tin oxide sensor in different working atmospheres, the sensing properties toward vapors of diacetyl in aqueous solution (0.4 mg/L) were tested in oxidant, inert and reducing atmospheres, using as carrier and regeneration flow air N2 and CO2.
Tin oxide is well known as an MOS whose behavior is n-type. Indeed, the electrical resistance of the layer decrease when the sensor is exposed to reducing gas such as NH3 [36], CO [37], CH4 [38], H2S [39] or SO2 [40] due to the oxidation of the target gas on the SnO2 surface. Measurements carried out under air and N2 atmospheres (Figure 3a), confirmed the n-type behavior of the SnO2 sensor with R < R0, whereas the measurement carried out in CO2 ambient showed an inverse behavior, with R > R0, proving indeed that in CO2, tin oxide worked as a p-type sensor.

Sensor Response in Air Atmosphere
SnO2 response in air was widely studied and reported in literature [14]. Generally, in air the surface of SnO2 is covered with negatively charged oxygen adsorbates ( , , and ). The formation of such oxygen adsorbates extracts electrons from the conduction band of SnO2, bulk forming an electron depletion layer on the SnO2 grains surface (space-charge region) and a potential barrier at the grain boundaries ( Figure 3b) [41]; the sensor has a high resistance value (~30 MΩ). As soon as the sensor interacts with the analyte vapor, the diacetyl molecules are oxidized by oxygen species on the surface and the depleted layer electrons are fed back into SnO2 bulk, thus a narrowed depletion layer and a reduction of the space-charge region is detected. Therefore, the sensor resistance decrease. In this case, exposing the sensor to 0.4 mg/L diacetyl solution vapors, the sensor response is equal to 1.43 and it results faster in the first seconds after the exposure, with a total response time of 3.0 min. Despite a very fast response being first observed, a delay in the sensor signal rise is detected; it demands a longer time to reach the final value. Indeed, the recording of the dynamic behavior of a sensor can be confusing, because the

Gas Sensor Measurements
First, to study the behavior of the tin oxide sensor in different working atmospheres, the sensing properties toward vapors of diacetyl in aqueous solution (0.4 mg/L) were tested in oxidant, inert and reducing atmospheres, using as carrier and regeneration flow air N 2 and CO 2 .
Tin oxide is well known as an MOS whose behavior is n-type. Indeed, the electrical resistance of the layer decrease when the sensor is exposed to reducing gas such as NH 3 [36], CO [37], CH 4 [38], H 2 S [39] or SO 2 [40] due to the oxidation of the target gas on the SnO 2 surface. Measurements carried out under air and N 2 atmospheres (Figure 3a), confirmed the n-type behavior of the SnO 2 sensor with R < R 0 , whereas the measurement carried out in CO 2 ambient showed an inverse behavior, with R > R 0 , proving indeed that in CO 2 , tin oxide worked as a p-type sensor.

Sensor Response in Air Atmosphere
SnO 2 response in air was widely studied and reported in literature [14]. Generally, in air the surface of SnO 2 is covered with negatively charged oxygen adsorbates (O − 2 , O − , and O 2− ). The formation of such oxygen adsorbates extracts electrons from the conduction band of SnO 2, bulk forming an electron depletion layer on the SnO 2 grains surface (spacecharge region) and a potential barrier at the grain boundaries (Figure 3b) [41]; the sensor has a high resistance value (~30 MΩ). As soon as the sensor interacts with the analyte vapor, the diacetyl molecules are oxidized by oxygen species on the surface and the depleted layer electrons are fed back into SnO 2 bulk, thus a narrowed depletion layer and a reduction of the space-charge region is detected. Therefore, the sensor resistance decrease. In this case, exposing the sensor to 0.4 mg/L diacetyl solution vapors, the sensor response is equal to 1.43 and it results faster in the first seconds after the exposure, with a total response time of 3.0 min. Despite a very fast response being first observed, a delay in the sensor signal rise is detected; it demands a longer time to reach the final value. Indeed, the recording of the dynamic behavior of a sensor can be confusing, because the measured change in signal depends both on the intrinsic reaction of the gas sensing mechanism and on a delayed change in test gas concentration [42]. Moreover, it should be highlighted that response/recovery times may be dependent on the measurement procedure.
On the contrary the recovery time is constant and slow (10.2 min).
Appl. Sci. 2022, 12, x FOR PEER REVIEW 5 of 10 measured change in signal depends both on the intrinsic reaction of the gas sensing mechanism and on a delayed change in test gas concentration [42]. Moreover, it should be highlighted that response/recovery times may be dependent on the measurement procedure. On the contrary the recovery time is constant and slow (10.2 min).

Sensor Response in N2 Atmosphere
Under N2 atmosphere the sensor showed a lower baseline resistance than in air (~3 MΩ). This result is compatible with the reduction of the density of oxygen species bound to the surface and then the increment of free carrier electrons into the conduction band that reduce the space-charge region at SnO2 grains boundaries. The response to 0.4 mg/L diacetyl exposure confirms a typical n-type sensing, with a resistance decrement due to the interaction of diacetyl molecules with the surface. In addition, sensors tested under N2 conditions shows a response comparable in terms of shape and magnitude with that observed under air, even if the sensor in N2 atmosphere do not show a complete recovery. A similar trend was already observed with a SnO2 sensor in argon atmosphere [43]. Measurements performed in air and N2 appeared with the same shape. Previews works indicated that despite a traditional n-type response was observed, in absence of oxygen the sensing mechanism is related to the direct adsorption of VOC on the MOS surface [43].
The sensor response in N2 is 1.60, but both response and recovery time result longer than in air measurements cause the poor interaction between N2 carrier and the sensor surface.

Sensor Response in CO2 Atmosphere
In a pure-CO2 atmosphere the resistance R is greater than R0, revealing a p-type behavior. In this condition the sensor is able to detect diacetyl and shows a total recovery in short time.
In a CO2 atmosphere R resulted smaller than both in air and N2 (~30 kΩ). Indeed, according to Wang et al., when SnO2 works at 240 °C in high CO2 concentrations and less

Sensor Response in N 2 Atmosphere
Under N 2 atmosphere the sensor showed a lower baseline resistance than in air (~3 MΩ). This result is compatible with the reduction of the density of oxygen species bound to the surface and then the increment of free carrier electrons into the conduction band that reduce the space-charge region at SnO 2 grains boundaries. The response to 0.4 mg/L diacetyl exposure confirms a typical n-type sensing, with a resistance decrement due to the interaction of diacetyl molecules with the surface. In addition, sensors tested under N 2 conditions shows a response comparable in terms of shape and magnitude with that observed under air, even if the sensor in N 2 atmosphere do not show a complete recovery. A similar trend was already observed with a SnO 2 sensor in argon atmosphere [43]. Measurements performed in air and N 2 appeared with the same shape. Previews works indicated that despite a traditional n-type response was observed, in absence of oxygen the sensing mechanism is related to the direct adsorption of VOC on the MOS surface [43].
The sensor response in N 2 is 1.60, but both response and recovery time result longer than in air measurements cause the poor interaction between N 2 carrier and the sensor surface.

Sensor Response in CO 2 Atmosphere
In a pure-CO 2 atmosphere the resistance R is greater than R 0, revealing a p-type behavior. In this condition the sensor is able to detect diacetyl and shows a total recovery in short time.
In a CO 2 atmosphere R resulted smaller than both in air and N 2 (~30 kΩ). Indeed, according to Wang et al., when SnO 2 works at 240 • C in high CO 2 concentrations and less than 14% relative humidity conditions, CO 2 acts as an electron donor, as a weak reducing agent [14]. Such a response-type transition is also observed for other materials, and it is due to the inversion of the conduction type of major carriers, which limits the dynamic range of the sensor at high concentration [44]. In the pure-CO 2 atmosphere the sensor response to 0.4 mg/L diacetyl exposure is 1.13, with response and recovery times of 5.5 min and 7.8 min, respectively. Recovery time results are shorter than in air and in N 2 atmosphere cases.

Sensor Response towards Diacetyl in Aqueous and Alcholic Solutions
Once the effect of the different carrier/regeneration atmospheres on the performance of SnO 2 towards diacetyl was evaluated, just oxygen-deficient atmospheres were considered. Indeed, alcoholic fermentation, producing ethanol and carbon dioxide, usually occurs in anaerobic environments, since it does not require oxygen. Hence, tests were further performed in oxygen-deficient atmospheres, N 2 and CO 2 , in: (i) aqueous diacetyl solution

Sensor Response towards Diacetyl in Aqueous and Alcholic Solutions
Once the effect of the different carrier/regeneration atmospheres on the performance of SnO2 towards diacetyl was evaluated, just oxygen-deficient atmospheres were considered. Indeed, alcoholic fermentation, producing ethanol and carbon dioxide, usually occurs in anaerobic environments, since it does not require oxygen. Hence, tests were further performed in oxygen-deficient atmospheres, N2 and CO2, in: (i) aqueous diacetyl solution (0.4 mg/L); (ii) alcoholic (5% ethanol) diacetyl solution (0.4 mg/L), and (iii) alcoholic (5% ethanol) solution, for comparison purposes. The response values along with the response and recovery time are summarized in Figure 4.
The sensor in both N2 and CO2 environments shows a higher response to ethanol than diacetyl, reaching 1.80 and 1.22, respectively, with shorter response and recovery times. In alcoholic diacetyl solution, the response in N2 ambient registers a value of 1.61, similar to the value found with the only-diacetyl solution, and response and recovery time results are similar, showing that under N2 atmosphere it is very difficult to distinguish diacetyl effects in aqueous or alcoholic solutions. Instead, in a CO2 atmosphere an increment in the response towards diacetyl alcoholic solution, with a final value of 1.27, is detected. In turns, the sensor, despite the lower response than in CO2, succeeds in distinguishing diacetyl under fermentation conditions. Moreover, response time in CO2 shows a marked decrement (46%) compared to only diacetyl. Similarly, recovery time is reduced too.

Effect of Diacetyl Concentrations in Alcholic and Aqueous Solutions
The fabricated SnO2 sensor was investigated by varying the diacetyl concentration both in aqueous and alcoholic solutions in a CO2 environment ( Figure 5). The response The sensor in both N 2 and CO 2 environments shows a higher response to ethanol than diacetyl, reaching 1.80 and 1.22, respectively, with shorter response and recovery times. In alcoholic diacetyl solution, the response in N 2 ambient registers a value of 1.61, similar to the value found with the only-diacetyl solution, and response and recovery time results are similar, showing that under N 2 atmosphere it is very difficult to distinguish diacetyl effects in aqueous or alcoholic solutions. Instead, in a CO 2 atmosphere an increment in the response towards diacetyl alcoholic solution, with a final value of 1.27, is detected. In turns, the sensor, despite the lower response than in CO 2 , succeeds in distinguishing diacetyl under fermentation conditions. Moreover, response time in CO 2 shows a marked decrement (46%) compared to only diacetyl. Similarly, recovery time is reduced too.

Effect of Diacetyl Concentrations in Alcholic and Aqueous Solutions
The fabricated SnO 2 sensor was investigated by varying the diacetyl concentration both in aqueous and alcoholic solutions in a CO 2 environment ( Figure 5). The response curves for different diacetyl concentrations (0.05-3.2 mg/L) are reported in Figure 5a,b. It can be seen that the response of the SnO 2 sensor increases by increasing diacetyl concentration with an exponential trend. The response of the sensor (Y) can be fitted as a logarithmic function of Y = 1.14 + 0.07logX with X representing the diacetyl concentration in mg/L and with a regression coefficient R 2 of 0.97, as shown in the inset of Figure 5a. An exponential trend of the response depending on the diacetyl concentration in alcoholic solution is instead reported in Figure 5b. The sensor response in the alcoholic solution results higher than diacetyl in a water solution, following a logarithm profile. This is due to the high sensitivity of SnO 2 to ethanol as reported in previous works [37,38]. Moreover, the presence of ethanol markedly enhanced the gas sensing performance toward diacetyl in the alcoholic solution.
curves for different diacetyl concentrations (0.05-3.2 mg/L) are reported in Figure 5a and b. It can be seen that the response of the SnO2 sensor increases by increasing diacetyl concentration with an exponential trend. The response of the sensor (Y) can be fitted as a logarithmic function of Y = 1.14 + 0.07logX with X representing the diacetyl concentration in mg/L and with a regression coefficient R 2 of 0.97, as shown in the inset of Figure 5a. An exponential trend of the response depending on the diacetyl concentration in alcoholic solution is instead reported in Figure 5b. The sensor response in the alcoholic solution results higher than diacetyl in a water solution, following a logarithm profile. This is due to the high sensitivity of SnO2 to ethanol as reported in previous works [37,38]. Moreover, the presence of ethanol markedly enhanced the gas sensing performance toward diacetyl in the alcoholic solution. By fitting the sensor response with a logarithmic function Y = 1.28 + 0.13logX (R 2 = 0.99) it is evident how the slope of the curve is greater than the response of diacetyl in water solution. From the fit it is possible to extract the lower detection limit (LDL), the minimum concentration of gas detectable by gas sensor [45,46].
For SnO2 sensor LDL can be estimated by the extrapolation of the linear fit of the response curve down to the minimum response Rmin/R0 where Rmin = R0 − 3σ, and σ is the standard deviation of the baseline resistance before analyte exposure. For diacetyl in an aqueous solution LDL results 0.1 mg/L, while for measurements performed on diacetyl in an alcoholic solution LDL is less than 0.01 mg/L. In turn, ethanol presence increases the sensitivity of the sensor, proving the capability of the SnO2 sensor to detect diacetyl in alcoholic solution in a CO2 atmosphere.

Response and Recovery Time in Alcholic Diacetyl Solution
Response and recovery times of diacetyl in alcoholic solution under a CO2 atmosphere are analyzed, being important to understand the performance of the sensor. Figure 6a shows response/recovery times of the SnO2 sensor in different diacetyl concentrations. Their analysis shows that for all the evaluated diacetyl concentrations, recovery time is higher than response time. Response and recovery time values decrease with an exponential law upon increasing the concentration; response time is almost halved, from 3.2 min to 1.7 min with a diacetyl increment from 0.4 mg/L to 3.2 mg/L. Analogously, the recovery time decreases from 6.6 min to 2.2 min in the same working conditions. According to literature data [42], basic effects of surface covering kinetics and diffusion may be the cause of a dependence on concentration change. Moreover, since the semiconductor By fitting the sensor response with a logarithmic function Y = 1.28 + 0.13logX (R 2 = 0.99) it is evident how the slope of the curve is greater than the response of diacetyl in water solution. From the fit it is possible to extract the lower detection limit (LDL), the minimum concentration of gas detectable by gas sensor [45,46].
For SnO 2 sensor LDL can be estimated by the extrapolation of the linear fit of the response curve down to the minimum response R min /R 0 where R min = R 0 − 3σ, and σ is the standard deviation of the baseline resistance before analyte exposure. For diacetyl in an aqueous solution LDL results 0.1 mg/L, while for measurements performed on diacetyl in an alcoholic solution LDL is less than 0.01 mg/L. In turn, ethanol presence increases the sensitivity of the sensor, proving the capability of the SnO 2 sensor to detect diacetyl in alcoholic solution in a CO 2 atmosphere.

Response and Recovery Time in Alcholic Diacetyl Solution
Response and recovery times of diacetyl in alcoholic solution under a CO 2 atmosphere are analyzed, being important to understand the performance of the sensor. Figure 6a shows response/recovery times of the SnO 2 sensor in different diacetyl concentrations. Their analysis shows that for all the evaluated diacetyl concentrations, recovery time is higher than response time. Response and recovery time values decrease with an exponential law upon increasing the concentration; response time is almost halved, from 3.2 min to 1.7 min with a diacetyl increment from 0.4 mg/L to 3.2 mg/L. Analogously, the recovery time decreases from 6.6 min to 2.2 min in the same working conditions. According to literature data [42], basic effects of surface covering kinetics and diffusion may be the cause of a dependence on concentration change. Moreover, since the semiconductor can be considered as an RC filter with a time constant τ, electronic effects may have influence too. Due to the increasing of diacetyl concentration, the resistance decreases and thus also the time constant.
Indeed, the sensor resistance as a function of time during the exposure to the analyte and during the recovery can be described through the following equations [42]: where R(t) is the resistance after the analyte exposure, R 0 is a constant, R max is the saturated resistance, t is the time, and τ resp and τ rec are the response and recovery time constants, respectively. Figure 6b shows the plot of time constants as a function of diacetyl concentration. The curves show that the time constants decreased as diacetyl concentration increased. According to response and recovery time analysis, the range of τ values recorded at low diacetyl concentration is much wider than that recorded at high diacetyl concentration, where τ is less than 50 s. The parameter τ is therefore considered a more suitable parameter to characterize sensors response and recovery times. can be considered as an RC filter with a time constant τ, electronic effects may have influence too. Due to the increasing of diacetyl concentration, the resistance decreases and thus also the time constant. Indeed, the sensor resistance as a function of time during the exposure to the analyte and during the recovery can be described through the following equations [42]: where R(t) is the resistance after the analyte exposure, R0 is a constant, Rmax is the saturated resistance, t is the time, and τresp and τrec are the response and recovery time constants, respectively. Figure 6b shows the plot of time constants as a function of diacetyl concentration. The curves show that the time constants decreased as diacetyl concentration increased. According to response and recovery time analysis, the range of τ values recorded at low diacetyl concentration is much wider than that recorded at high diacetyl concentration, where τ is less than 50 s. The parameter τ is therefore considered a more suitable parameter to characterize sensors response and recovery times.

Conclusions
A SnO2-based MOS sensor is used to investigate the detection of VOCs in fermented beverages such as beer, wine and distillates. Synthesized and characterized SnO2 is used to discriminate the concentration of diacetyl vapors in different operating atmospheres, as to replicate a typical alcoholic fermentation scenario.
Under air, N2 and CO2 gases, SnO2 shows different behavior and baseline resistances due to different interactions with its surface. In particular, measurements carried out under air and N2 atmospheres show the n-type behavior of the SnO2 sensor, whereas measurements carried out in CO2 ambient prove that SnO2 acted as a p-type sensor.
Tests performed in aqueous and alcoholic diacetyl solutions show good response in terms of diacetyl detection in the range of concentration 0.2-3.2 mg/L at 200 °C in CO2 atmosphere, with LDL of 0.1 mg/L and 0.01 mg/L, respectively.

Conclusions
A SnO 2 -based MOS sensor is used to investigate the detection of VOCs in fermented beverages such as beer, wine and distillates. Synthesized and characterized SnO 2 is used to discriminate the concentration of diacetyl vapors in different operating atmospheres, as to replicate a typical alcoholic fermentation scenario.
Under air, N 2 and CO 2 gases, SnO 2 shows different behavior and baseline resistances due to different interactions with its surface. In particular, measurements carried out under air and N 2 atmospheres show the n-type behavior of the SnO 2 sensor, whereas measurements carried out in CO 2 ambient prove that SnO 2 acted as a p-type sensor.
Tests performed in aqueous and alcoholic diacetyl solutions show good response in terms of diacetyl detection in the range of concentration 0.2-3.2 mg/L at 200 • C in CO 2 atmosphere, with LDL of 0.1 mg/L and 0.01 mg/L, respectively.
Response and recovery times trends reveal the diacetyl concentration dependance, with response and recovery times constant less than 50 s. Author Contributions: Conceptualization, A.G., A.F., L.B. and A.D.; methodology, A.G., A.F. and L.B.; formal analysis, A.G. and A.F.; investigation, A.G. and A.F.; data curation, A.G.; writingoriginal draft preparation, A.G. and A.F.; writing-review and editing, A.G. and L.B.; supervision L.B. and A.D. All authors have read and agreed to the published version of the manuscript.
Funding: This research was co-financed with the support of the European Commission, the European Social Fund and the Calabria Region. The author is solely responsible for this research and the European Commission and the Calabria Region decline any responsibility for the use that may be made of the information contained therein.