SO₂ Detection over a Wide Range of Concentrations: An Exploration on MOX-Based Gas Sensors

Abstract

Noxious gases such as sulfur-containing compounds can inflict several different adverse effects on human health even when present at extremely low concentrations. The accurate detection of these gases at sub-parts per million levels is imperative, particularly in fields where maintaining optimal air quality is crucial. In this study, we harnessed the capabilities of nanostructured metal-oxide semiconducting materials to detect sulfur dioxide, since they have been extensively explored starting from the last decades for their effectiveness in monitoring toxic gases. We systematically characterized the sensing performance of seven chemoresistive devices. As a result, the SnO₂:Au sensor demonstrated to be the most promising candidate for sulfur dioxide detection, owing to its highly sensitivity (0.5–10 ppm), humidity-independent behavior (30 RH% onwards), and selectivity vs. different gases at an operating temperature of 400 °C. This comprehensive investigation facilitates a detailed performance comparison to other devices explored for the SO₂ sensing, supporting advancements in gas detection technology for enhanced workplace and environmental safety.

1. Introduction

Sulfur oxides (SO_x) are one of the major contributors to atmospheric pollution, including acid rain and smog formation [1]. Indeed, the oxidation of SO₂ leads to the formation of sulfuric acid and fine particulate matter (PM2.5, particulate matter less than or equal to 2.5 μm in diameter). This can have ecological impacts on soil, forests, and freshwater. Moreover, this category of pollutants is strongly harmful to human beings because inhaled SO₂ can easily be hydrated in the respiratory tract, producing sulfurous acid that subsequently dissociates to form its derivatives, bisulfite and sulfite anions [2], which can cause damage and increase chronic obstructive pulmonary diseases, e.g., pneumonia and acute bronchitis, and daily mortality [3]. The impact on human health motivated the low Threshold Limit Values, i.e., the Short-Term Exposure Limit (TLV-STEL) and the Permission Exposure Limit Time-Weighted Average (PEL-TWA) for SO₂ are 0.25 parts per million (ppm) and 5 ppm, respectively [4,5]. Therefore, many countries have established or are enacting stricter regulations to reduce the SO₂ emissions into the atmosphere [1].

In nature, the release of SO_x is caused by the combustion of carbonaceous fuels, which occurs in forest fires and volcanoes [6]. In the last case, SO₂ fluxes can be substantial in terms of atmospheric source strengths, for instance, the mean SO₂ flux from Mt. Etna in Sicily is equivalent to the total of anthropogenic SO₂ emissions from France, making it arguably the world’s largest sustained point source emitter of SO₂. In addition to these inevitable natural contributions, there are also many anthropogenic sources of SO_x releases such as the combustion of fossil fuels in power plants [7], for which the permissible concentration levels are 0.03 ppm (1 h) and 0.14 ppm (24 h) [8] in the case of electricity generation and industry boilers, respectively. Further contributions to SO_x emission came from oil refineries, domestic boilers, and motor vehicles, e.g., the European permissible limit of SO₂ produced from the automobiles exhaust ranges between 20 and 120 ppm/km [9]. Moreover, the extensive use of food additives leads to residual SO₂, generally in the range of 50–250 ppm [10], while the use of pesticides such as ammonium sulfate and potassium sulfate leads to the inevitable decomposition of SO₂ gas. Exceeding 2 ppm of SO₂ can have detrimental effects on crops, as it can potentially destroy chloroplasts essential for photosynthesis [11].

To minimize the impact on the environment, it has become a primary goal to reduce anthropogenic SO_x, particularly by desulfurization of flue gas during combustion, as well as to monitor SO_x concentrations in situ at the sources and in the ambient atmosphere. Moreover, air quality monitoring of SO₂ is critical for safety in closed spaces and for leak identification.

Hence, an effective monitoring system is crucial to maintaining SO₂ levels within safe thresholds. To analyze and quantify SO₂ and other sulfur compounds, the conventionally used laboratory technique is gas chromatography coupled with mass spectrometry (GC-MS). GC-MS is acknowledged for its precision and reliability in identifying gaseous compounds, establishing itself as the industry standard. Despite its accuracy, GC-MS presents practical challenges such as the need for bulky and expensive equipment, specialized personnel, and complex pre-treatment processes, resulting in being unsuitable for continuous real-time detections. This underscores the impracticality of such techniques for widespread use in on-site gas monitoring. Therefore, there is a clear need for alternative methods that offer cost-effectiveness, swift processing, rapid response, and reduced reliance on solvents and samples for routine detections.

To date, gas sensors have emerged as both a cost-effective alternative and a complementary system to laboratory-based analyses, providing a rapid and promising solution for SO₂ detection. Notably, electronic nose (e-nose) systems have gained attention for odor assessments, leveraging their versatility and resemblance to the human olfactory system. The implementation of e-noses involves training procedures, with qualified samples to build a database of reference, as well as sensor array optimization, which requires the selection of appropriate devices among the different classes of gas sensors. Recent developments involved the use of electrochemical and non-dispersive infrared (NDIR) sensors for SO₂ detection. However, some limitations by NDIR devices, e.g., short device lifetime, low selectivity, and spectral interference, have prevented their widespread usage.

Metal-oxide semiconductor (MOX)-based chemoresistive gas sensors are a compelling alternative, showcasing not only high sensitivity but also cost-effectiveness, scalable manufacturing, durability, and seamless integration into Internet of Things (IoT) systems [12]. Previous experimental and theoretical works proved that MOX-based nanostructured materials, such as SnO₂, WO₃, and ZnO are potential candidates to detect SO₂ [7,13,14,15,16,17,18]. Nonetheless, the investigation of MOX for SO₂ detection has been hampered by evidence of irreversible interactions between this analyte and the active groups over the sensing layers [19], resulting in sensor instability and a short lifetime. For instance, Berger et al. [20] demonstrated irreversible sulfate formation on the surface of SnO₂, which was identified as the cause of the irreversibility of the device response after initial SO₂ detection.

Hence, to enhance their sensitivity and selectivity, various approaches, including the do** and/or decoration effect and heterojunction formation, have been investigated for tuning the material properties. In these perspectives, SnO₂-based sensors, containing small amounts of metals such as Ni and Cu, have already been characterized and tested to achieve selective and sensitive detection of SO₂ gas [21,22].

Consequently, to further investigate the potentialities of chemoresistive gas sensors in monitoring SO₂, this study assesses the sensing capability of seven distinct MOX materials. Three of these are pure MOXs, namely, SnO₂, WO₃ and ZnO, to identify the most promising among them within our measurement system. The best candidate, in terms of sensitivity and selectivity, has been functionalized with noble metals, such as Au, Pt, Pd, and Ag.

2. Materials and Methods

2.1. Chemoresistive Gas Sensors

ZnO, SnO₂, WO₃, SnO₂:Au, SnO₂:Pt, SnO₂:Pd, and SnO₂:Ag were synthesized in powder form by the sol-gel method [23,24,25]. The powders were mixed with organic excipients, namely, α-terpineol, ethyl cellulose, and silica, to form homogeneous pastes that facilitate the screen-printing technique. The resulting pastes were sonicated for 2 h and then screen-printed onto alumina substrates (defined area of 2.54 × 2.54 mm²) [26,27]. The semiconducting MOX thick films (around 20–30 μm of thickness) were sintered in air for 2 h at 650 °C to remove the organic precursors and improve mechanical stability and film adhesion. The substrates were equipped with interdigitated gold contacts on the front side, to provide the bias voltage for measuring the conductance of the film, and a platinum heater was integrated on the back side to thermally activate the sensing material at its optimal working temperature. Finally, these substrates were bonded to a standard TO-39 support through thermo-compression of 60 μm of diameter gold wires (Figure S1, Supporting Information). The integration of electrodes and heaters in the alumina substrate allows for precise control and measurement of the sensing material’s response.

2.2. Lab-Test Setup

The MOX-based sensors were electrically characterized in a sealed test chamber of volume ≈ 622 cm³ in which temperature and relative humidity (RH%) were measured by a commercially available Honeywell (Charlotte, NC, USA) HIH-4000 sensor (accuracy ± 3.5 RH%). Firstly, the devices were exposed to synthetic dry air (20% O₂ and 80% N₂) for almost one day in order to obtain the baseline stabilization. Secondly, the mixture of gases from certified cylinders were injected inside the chamber through mass-flow controllers, achieving a total flow of 500 standard cubic centimeters per minute (sccm). The filling time of the test chamber was about 1 min and 15 s, as it depends on the size, the geometry of the chamber, and the velocity of the gas flow. The humidity conditions were obtained by fluxing a fraction of the total synthetic air into a gas bubbler filled with deionized water. Power suppliers from Aim TTi (Glebe Rd, Huntingdon, Cambridgeshire, UK) were used to furnish the requisite electric current to the sensor heater, while a multimeter, the K2000 from Keithley (Cleveland, OH, USA), was utilized to measure the electrical conductance of the sensing film. A constant bias of 5 V was maintained between the two interdigitated gold electrodes. The circuitry employed for data acquisition was structured around an operational amplifier (OA), which serves as a pivotal component in processing the signals. Thanks to the configuration in which the input voltage (V_in) and output voltage (V_out) values corresponded to the ends of the sensor resistor (R_s) and the applied load resistor (R_f), respectively, the gain was determined by the relationship $\frac{V_{out}}{V_{in}}$ = $-\frac{R_f}{R_S}$. This fundamental expression elucidates the relationship between the V_in and V_out, facilitating accurate interpretation of the sensor’s response. According to A. Rossi et al. [26], the expression for the sensor conductance G_S is given as follows:

$$G_S=\frac{1}{R_S}=-\frac{V_{out}}{R_f\cdot V_{in}}$$

(1)

2.3. Sensing Measurements

The sensors performances were proven by making a comprehensive electrical characterization.

First, the best operating temperature of each MOX film was investigated considering their response to 1 ppm of SO₂ at increased working temperatures, ranging from 200 to 450 °C.

Then, in order to study the sensitivity, the films were exposed to several concentrations of SO₂ ranging from 0.5 to 10 ppm under dry conditions. Specifically, to verify possible surface poisoning effects, once achieved the plateau of the signal after the gas injection, the sensors were again exposed to a fully dry air flux to assess the complete recovery of the baseline The sensor response was defined as the following equations for reducing and oxidizing gases, respectively:

$$R=\frac{G_{gas}-G_{air}}{G_{air}}\mathrm{for}\mathrm{reducing}\mathrm{gas}$$

(2)

$$R=\frac{G_{air}-G_{gas}}{G_{gas}}\mathrm{for}\mathrm{oxidizing}\mathrm{gas}$$

(3)

where G_gas and G_air are the conductance values in gas and in air, respectively. The error on the response value was calculated applying error propagation, considering the load resistance tolerance and the instrument uncertainty on the output voltage acquisition. The result was an error of 5% [28].

Due to instrumental limitations, the sensors’ response at the TLV-STEL concentration, 0.25 ppm, was not experimentally evaluated. Otherwise, the theoretical Limit of the Detection (LOD) of the best device was determined as 3(RMS noise/R_air), where RMS noise is the root mean square deviation and R_air is the resistance in dry air [29].

Humidity is a very common interference due to its ability to interact with the film, affecting its conductance and generating competitive −OH groups that limit the adsorption sites on the reactive surface. To explore its effect on the sensor performance, 5 ppm of SO₂ was injected into the gas chamber in presence of different percentages of humidity (2–56 RH%).

The stability and repeatability of the best SO₂ devices at its operating temperature was assessed under 6 ppm of the target gas in dry and wet (30 RH%) conditions. The response (τ_res) and the recovery (τ_rec) times of the sensors were calculated as the times needed to attain 90% of the steady-state response and required to switch back to 90% of the baseline value, respectively.

To verify the influence of possible interfering compounds, they were also exposed to different volatile analytes, such as benzene, ethanol, nitrogen dioxide (NO₂), dimethyl disulfide (DMDS), carbon monoxide (CO), and carbon dioxide (CO₂). The gases and their concentrations were chosen in order to test the surface reactivity of the semiconductors vs. different chemical functional groups. In particular, the concentrations of the potential inteferents were chosen based on TLV-TWA and the American Conference of Governmental Industrial Hygienists (ACGIH) TLV values [30,31,32,33,34,35].

Moreover, to evaluate the ability of the sensors to detect SO₂ in presence of different analytes, we also performed cross-selectivity measurements. Therefore, the sensors were exposed to 5 ppm of SO₂ in an atmosphere of 1 ppm of ethanol in dry and wet (30 RH%) conditions.

3. Results

3.1. Operational Temperature

The sensors’ optimal working temperature was determined by measuring conductance variations before and after introducing 1 ppm of SO₂ in dry conditions across a range of working temperatures from 200–450 °C (Figure 1). The response of ZnO, SnO₂:Pd, SnO₂:Au, SnO₂, SnO₂:Pt, and SnO₂:Ag sensors increased with the temperature rise, reaching the highest value at 400 °C. High operating temperatures are commonly used to lower response and recovery times and to optimize the MOX-based sensors catalytic activity towards SO₂, facilitating surface redox reactions. Nevertheless, beyond the optimal temperature, gas desorption rates rise significantly, leading to further lower sensor response [24].

In contrast to the other sensors, WO₃ exhibited two peak responses at 200 °C and 350 °C, suggesting distinct optimal conditions for detecting the analyte, likely due to different active sites on the surface for SO₂ chemisorption, namely, O₂⁻ at 200 °C and O⁻ at 350 °C, as explained in our previous work [26] and also deepened in the Discussion section of the present study. However, since the response value at 200 °C and 350 °C is almost the same, the lowest working temperature (200 °C) was selected due to alignment with requirements for low-power consumption in portable and IoT applications.

The optimal temperature for each sensor is summarized in

Table 1. Optimal operating temperature of the sensors for SO₂ detection, obtained from Figure 1.

Sensor	Optimal Working Temperature [°C]
WO3	200
ZnO	400
SnO2:Pd	400
SnO2:Au	400
SnO2	400
SnO2:Pt	300
SnO2:Ag	400

and was selected for subsequent comprehensive electrical characterization.

Notably, Figure 1 also shows that the SnO₂-based sensor had greater response values than the other pure MOXs, namely, ZnO and WO₃. That had been the underlying motivation for decorating SnO₂ with noble metals rather than ZnO or WO₃.

3.2. Sensitivity

To investigate the sensitivity, we exposed the sensors towards several concentrations of the analyte, in particular, SnO₂:Au, SnO₂:Pd and ZnO from 0.5 to 10 ppm, instead of SnO₂, SnO₂:Pt and SnO₂:Ag from 1 to 10 ppm. Only WO₃ was tested from 0.5 to 7 ppm since the response tended to remain constant and low. Figure 2 highlights that SnO₂:Au, SnO₂:Pd and SnO₂ sensors at each tested concentration exhibited higher responses than the others and they did not saturate at the relative highest concentration tested. On the other hand, WO₃, ZnO, SnO₂:Ag, and SnO₂:Pt sensors showed a constant response for concentrations above 2 ppm, which can be attributed to the saturation of active sites on the materials and sulfur poisoning of the surface [36]. Therefore, SnO₂:Au, SnO₂:Pd and SnO₂ sensors exhibited a remarkable capability to detect SO₂ across a wide range of concentrations, which make them appropriate for diverse industrial, environmental, and safety monitoring applications.

The calibration curves of the sensors were fitted with asymptotic function ($y=a-b\cdot c^x$), except for the SnO₂:Ag device, for which it was not possible to identify a suitable curve fit. All the metrics parameters of the function are listed in Table S1 (Supporting Information).

In particular, it turned out that the SnO₂:Au sensor showed better responses than the other devices for SO₂ detection with a theoretical LOD of ∼ 0.48 ppm. In Figure S2 (Supporting Information), we reported the SnO₂:Au sensor response at three SO₂ concentrations close to its LOD. One can observe that below the LOD, the sensor signal (at 0.2 and 0.3 ppm) was noisy and not stable.

3.3. Humidity Effect

The impact of humidity on the sensing performance was assessed by fluxing 5 ppm of SO₂ while exposing the devices to escalating levels of water vapor (2–56 RH%). As depicted in Figure 3, the responses exhibited a significant drop at 17 RH% due to the competitive reaction for active sites between water molecules and SO₂ [27]. However, for SnO₂-based sensors and ZnO, the response to SO₂ was clearly discernible, with SnO₂:Au exhibiting the strongest signal. Only the WO₃ was not able to detect SO₂ in humid conditions for values higher than 17 RH%. Starting from 45 RH%, all sensors maintained an almost constant response. This behavior is explainable with the Grotthuss mechanism, in which the formation of continuous water layers over the sensing film hinders the reaction between the analytes and the functional material active sites [37].

To the best of our knowledge, the only work that deeply discussed the humidity effect in SO₂ detection was developed by Gaiardo et al., who proposed a novel approach using nanostructured silicon carbide (SiC) as a sensing film [28]. It resulted in a highly selective SO₂ sensor even if it operated at a high temperature (600 °C). In this work, they tested five different relative humidities vs. 10 ppm of SO₂. Starting from 2 to 60 RH%, the response remained rather stable around 0.25, reaching a peak of 0.27 at 13 RH%. Moreover, in a consecutive study, M. Della Ciana et al. supported the previous work by Gaiardo et al. applying operando Diffuse Reflectance Infrared Fourier Transform (DRIFT) spectroscopy. With this advanced technique, they demonstrated that the enhancement of SO₂ response under wet conditions was attributed to the formation of SO_x species, which increased the number of charge carriers. However, despite these advancements, especially in SO₂-sensing mechanisms in wet conditions, SiC sensors still exhibited a low response (0.20 at 8 RH%) even to higher concentrations of SO₂, such as 25 ppm [37]. On the contrary, the response at 8 RH% of SnO₂:Au sensor to a SO₂ concentration five times less (5 ppm) is equal to 3.7.

3.4. Repeatability

The stability of the dynamic response vs. 6 ppm of SO₂ during four consecutive cycles was investigated in dry and wet conditions for the best candidate, SnO₂:Au (Figure 4). The film exhibited a clear reversible interaction and good repeatability for detection over time. A doubling increase in the conductance baseline of the sensitive film was observed when exposed to a humid atmosphere due to water’s homolytic dissociation. This chemical process involved water interacting with metal sites and lattice oxygen, leading to the formation of terminal and anchored hydroxyl groups (−OH) on the film’s surface. These −OH groups acted as surface donors, releasing electrons into the conduction band. This alteration in the surface chemistry affected the film’s catalytic properties regarding SO₂ detection by replacing the oxygen ion active sites, typically present in dry conditions, with the newly formed hydroxyl groups.

However, the variation of the conductance values over the four test cycles under SO₂ exposure was similar in dry and wet conditions. Instead, the values of the conductance baseline mainly increased during the test in dry air, while they recovered almost the initial levels during the measurement in humid air. This confirmed that SnO₂:Au sensor can be reliable in real-like conditions limiting the effect of humidity on the signal drift.

The response and recovery times of SnO₂:Au sensors are 3 and 17 min, respectively, in dry conditions, whereas in wet conditions, the response and recovery times are 5 min and 19 min, respectively. These values are in line with those exhibited by common MOX-based gas sensors [24,26].

In addition, Figure 4 shows a similar response shape of the sensor in dry and humid conditions, i.e., a rapid increase when the target gas was introduced. This sharp peak can be ascribed to the dynamic equilibrium between gas diffusion into deeper layers of the sensing material and the kinetic processes of gas adsorption/desorption at its surface [38]. At the initial stage, SO₂ diffuses along the sensing film, starting to influence the dynamic equilibrium of redox reactions occurring at the surface of SnO₂:Au, decreasing, for example, the number of ionized oxygens, resulting in an increase in conductance. Nevertheless, at equilibrium, several factors may influence the conductance in presence of SO₂, slightly lowering its initial value. For instance, the competitive diffusion of new analyte and non-reactive reaction products inside and outside the sensing film should be considered, as well as the kinetic regeneration of active sites.

In order to investigate the long-term stability of the SnO₂:Au sensor, we verify its response value by comparing three different measurements with 3 ppm of SO₂ over a period of 5 months (Figure S3, Supporting Information).

A deepened investigation on SO₂ chemoresistive gas sensors was conducted in order to compare the performance of SnO₂:Au studied in this work with devices already discussed in the literature; see

Table 2. Comparison of gas detection features exhibited by SO₂ chemoresistive sensors identified in the literature.

Sensing Materials	Operating Temperature [°C]	Concentration [ppm]	Response Formula	Response in Dry/Wet Air	LOD	Reference
rGO/TAPP	RT	5	(Rgas − Rair)/Rair	-/0.45 at 30 RH%	5 ppm	[39]
Ni/SnO2	RT	5	Rgas/Rair	-/2.4 at ∼55–65 RH%	-	[19]
rGO	RT	5	(Rair − Rgas)/Rair	3.21/-	-	[40]
ITO(In2O3/SnO2)	240	5	Rair/Rgas	-/1.8 at 35 RH%	-	[41]
Ag NWs@SnO2/CuO	80	50	Rair/Rgas	2.1/-	0.25 ppm	[42]
GO/ZnO NR	RT	25	Rgas/Rair	2.97/-	-	[43]
RGO/SnO2 nanocomposites	60	10	(Rair/Rg) − 1	1.2/-	-	[44]
SnO2/NiO	180	500	(Rair/Rg) − 1	56/50 at dry/70 RH%	-	[45]
Ru/Al2O3/ZnO	350	100	(Rair − Rgas)/Rair	65/-	5 ppm	[46]
NiO:SnO2	240	100	Rair/Rgas	10/9.1 at dry/70 RH%	-	[47]
SnO2:Au	400	5	(Ggas − Gair)/Gair	10.4/2 at dry/30 RH%	0.48 ppm *	This work

The “-” stands for not mentioned in the respective article. R_air and R_gas are the resistance in dry air and in presence of SO₂, respectively. * Theoretical value.

. Even if it operates at a high temperature, the SnO₂:Au sensor can be obtained with an easier and cheaper synthesis process, such as sol-gel synthesis, enabling a potential large-scale production than the semiconducting composites explored by other authors mentioned. Moreover, the complete inorganic nature of our sensing material guarantees a strong stability with respect to the organic-based semiconductors. The SnO₂:Au sensor can detect SO₂ over a wide range of concentrations (0.5–10 ppm) and, according to the PEL-TWA (5 ppm), it showed a response at least three times higher than the other devices. Under wet conditions, among the sensing material listed in Table 2, only the Ni/SnO₂ sensor exhibited a comparable response to SnO₂:Au ones, even if the latter demonstrated a negligible influence by humidity over 30 RH%, which makes it suitable for real applications.

The SnO₂:Au sensor showed outstanding sensitivity even at a very low concentration. Indeed, the theoretical limit of detection calculated is equal to 0.48 ppm, a value remarkably comparable to that obtained in the Ag NWs@SnO₂/CuO sensor, which reported a LOD of 0.25 ppm.

Concerning the response and recovery times, it was not possible to compare the performance since these values are strongly correlated with geometry and arrangement of the experimental setup, e.g., test chamber volume and gas flow rate.

Therefore, the analysis carried out in this work revealed superior performance of the Au-decorated SnO₂ for the sulfur dioxide detection with respect to the current MOX-based sensors investigated from the literature.

3.5. Selectivity

Moreover, in the practical implementation, one should consider the co-presence of other gases whose physicochemical properties may affect the sensor performance. Therefore, the selectivity of the devices was investigated by exposing the sensors to various interferents, such as benzene (0.5 ppm), ethanol (5 ppm), NO₂ (3 ppm), DMDS (0.25 ppm), CO₂ (5000 ppm) and CO (25 ppm) under dry air. The dynamic signal of output voltage of the sensors for all the measurements are shown in Figures S4–S10 (Supporting Information), together with the load resistances in Table S2. The bar graph in Figure 5 illustrates the response values determined by Equations (2) and (3), to better evaluate the sensor’s selectivity.

As expected, almost all sensors exhibited a high response to ethanol [26,48,49,50], in particular, SnO₂:Pd, which therefore cannot be defined as selective for SO₂ detection. Among the decorated sensor-based on SnO₂, the SnO₂:Au is the one that exhibited the lowest response value to ethanol. In contrast, all the sensors exhibited negligible responses to CO, CO₂ and benzene. Only SnO₂-based sensors showed superior response to sulfur-containing gases such as DMDS and SO₂, especially SnO₂:Au, which proved to be particularly selective vs. SO₂. Then, WO₃ and ZnO, two of the most common pristine nanostructured MOX materials, were considered unsuitable for SO₂ detection due to their low sensitivity (Figure 2) and stronger selectivity for interfering gases such as ethanol and NO₂ than for the analyte of interest (Figure 5).

3.6. Cross-Selectivity

The cross-selectivity of the sensors towards 5 ppm of SO₂ in the presence of 1 ppm (baseline) of ethanol was performed in dry and wet conditions (30 RH%), as shown in Figure 6. The presence of ethanol as an interferent during the exposure to SO₂ had a notable impact on the response of SnO₂:Au sensors. Comparing Figure 6 to Figure 2, one can observe that the response to 5 ppm of SO₂ decreased from 10.4 to ~2.5 in dry conditions and ~0.5 in wet conditions. This effect can be attributed to competitive reactions between the reactive gases. Indeed, both ethanol and H₂O can react with the material active sites, making them ineffective for detecting SO₂. Furthermore, the products of such reactions, for example chemisorbed alkyl chains, might accumulate over the surface, hysterically hindering nearby active sites. However, SnO₂:Au demonstrated the effective ability to discriminate against SO₂ among ethanol interfering gas.

4. Discussion

Gas-sensing mechanisms of metal oxides, such as SnO₂, ZnO, and WO₃ presented in this study, involve complex surface reactions, which are influenced by various intrinsic and extrinsic factors. The intrinsic ones are inherent factors such as (i) activation of chemisorbed active sites from oxidative and reductive reactions, (ii) the creation of acidic or basic centers, and (iii) hydrate–hydroxyl layers, that, synergically combined, play a crucial role [51]. Indeed, (iv) the formation of extrinsic active centers on the surface are due to the modifications by catalytic clusters of noble metals, e.g., Au, Pd, Pt, and Ag.

4.1. Intrinsic Properties

4.1.1. Oxygen Chemisorption

The most common sensing mechanism explaining the response of MOX-based sensors to reducing gases involves oxygen chemisorption and subsequent redox reactions of the analytes with the oxygen-based surface ions. In principle, the oxygen molecules, which are chemisorbed on the semiconductors, ionize into O₂⁻, O⁻, and O²⁻ anion species by accepting the electrons from MOX materials at different temperatures and decreasing the overall conductance of the sensing film. The oxygen adsorption has been investigated by several works in literature [52], as it follows:

$$(O_2{)}_{gas}\leftrightarrow (O_2{)}_{adsorbed}$$

(4)

$$(O_2{)}_{adsorbed}+e^{-}\leftrightarrow O_2^{-}$$

(5)

$$O_2^{-}+e^{-}\leftrightarrow {2O}^{-}$$

(6)

$$O^{-}+e^{-}\leftrightarrow O^{2-}$$

(7)

Oxygen anions O₂⁻, O⁻, and O²⁻ are stable below 100 °C, 100–300 °C, and above 300 °C, respectively [52]. Based on this consideration, since in the current work, the operating temperature for most of the sensors was 400 °C, it is likely that they exhibited the formation of O²⁻ ions (Figure 7a), except for WO₃, which operates at a low temperature of 200 °C, where the formation of chemisorbed O₂⁻ and O⁻ ions was more probable. Differently from the other sensors, WO₃ displayed two maxima responses at 200 °C and 350 °C, meaning that there are two distinct conditions for optimal detection of SO₂, probably characterized by different active sites, namely, O₂⁻ and O⁻ ions in the first case and O⁻ and O²⁻ in the second case. In the temperature range between these two optimal values (200 °C and 350 °C), an unfavorable competition between different reaction mechanisms of the analyte absorption and desorption would lead to a decrease in WO₃-sensing capabilities [26].

When the sensors were exposed to SO₂ gas molecules in dry conditions, they reacted with the chemisorbed oxygen species on the surface and released the electrons to the conduction band of the MOX material.

In particular, in the case of the sensor that exhibited the best performance in the present work, i.e., SnO₂:Au, the overall surface reaction mechanism was given by the catalytic oxidation of SO₂ to SO₃ at high temperature (400 °C) [28] (see Figure 7b):

$$S{O}_2+O^{2-}\leftrightarrow S{O}_3+2e^{-}$$

(8)

From the above reaction, it is clear that when the target gas reacts with the oxygen species adsorbed on the SnO₂:Au surface, the number of free electrons increases, resulting in an increase in conductivity of the material.

4.1.2. Acid and Basic Centers

Beyond the ionosorption and interaction of oxygen species, surface acidity and basicity play a pivotal role in gas-sensing mechanisms. Indeed, functional material active sites can be basic in nature, such as surface oxygen anions and hydroxyl species, or acidic, like coordinately unsaturated cations and −OH Brønsted acid states [53]. The concentration of different active sites influences the adsorption and reaction of target gases on the sensor’s surface, thereby affecting its sensitivity and selectivity, as deeply discussed by Marikutsa et al. in [53]. The concentration of Brønsted and Lewis acid sites followed the order ZnO < SnO₂ < WO₃, consistent with the increase in metal–oxygen bond energy, electronegativity, and charge/radius ratio of cations. This sequence also reflects the decreasing optical basicity of oxides [53]. The basic nature of ZnO resulted in selectivity for VOCs like ethanol, but poor detection capabilities for sulfur-containing compounds. On the other hand, WO₃, the most acidic in nature, displayed a high response for oxidizing gas, such as NO₂, and very low signals for the other reducing gases. Between the three MOXs, SnO₂ was the most suitable for the development of chemoresistive SO₂ sensors. Berger et al. demonstrated that an increase in surface acidity for SnO₂-based material (probably related to sulfate formation) after SO₂ exposure may explain the increase in the sensor’s sensitivity [20]. Furthermore, enhanced sensitivity of SnO₂-based material was ascribed to the presence of noble metals, such as Au and Pd, which elevate the inherent acidity of pristine SnO₂. Therefore, the intrinsic acid centers of SnO_2, in combination with the extrinsic active centers, i.e., noble metals, effectively affect the sensing performance towards SO₂.

4.1.3. Hydrated Layers

Concerning the gas-sensing mechanism in wet conditions, the presence of H₂O molecules affects the reaction scenario (Figure 7c). Indeed, H₂O molecules (Lewis base) can be adsorbed as are on the surface or interact with cations (Lewis acid) by a strong binding resulting in water molecules dissociation and then producing the −OH groups. In this second case, strong Lewis acidic cations, i.e., W⁶⁺, Sn⁴⁺, Zn²⁺, can bind more than one −OH group, then acquire an acidic proton and behave as Brønsted acid site. This leads to an increase in the semiconductors conductance due to the redox reaction, which provides electrons to the sensing materials.

In our case, before the interaction with the sensing films, water molecules can react with SO₂ in the atmosphere by producing H₂SO₃ and further by-products (see Figure 7d). Subsequently, SO₂ and the other gaseous compounds, resulting from the interaction between SO₂ and H₂O, can interact with oxygen and hydroxyl ions adsorbed on the surface of the materials. The competitive reaction between −OH groups and pre-adsorbed oxygen species contribute to reducing the sensitivity of the materials and, as a result, negatively affect the performance of the sensors with respect to the dry ambient.

Furthermore, the consecutive creation of continuous layers of water molecules would hinder the functional material’s active sites for reaction with the analyte, thus reducing the sensitivity of the MOX films [37].

4.2. Extrinsic Properties

The presence of noble metals nanoparticles, in particular for SnO₂:Au sensors, exhibited a good catalytic dissociation performance for oxygen, which promoted the formation of chemically adsorbed oxygen on the material surface [54]. Thus, the reaction rate between chemically adsorbed oxygen and SO₂ increased, and the space charge layer narrowed accordingly. This further improved the response of the sensor to SO₂.

A further crucial aspect concerns the fact that the SnO₂:Au, thanks to the low chemical affinity between Au nanoparticles and sulfur atoms, is capable of preventing irreversible poisoning of the sensing film surface due to the formation of strong metal–sulfide bonds [36].

On the other hand, among SnO₂-based gas sensors tested, SnO₂:Ag demonstrated a constant response with increasing SO₂ concentrations, which is probably due to the production of sulfur stable compounds with Ag metals such as Ag₂S consequently poisoning the reactive surface from 2 ppm.

5. Conclusions

The experimentation performed in the present work could guide future developments on modified MOX-based sensors and open new frontiers for olfactory systems dedicated to SO₂ monitoring in a wide range of concentrations and in real-world applications. ZnO and WO₃ were proven to be selective to different gases and therefore not appropriate for SO₂ detection. In contrast, SnO₂ was more suitable for the detection of sulfur-containing gases. Therefore, SnO₂ was decorated with different noble metals. Addition of Ag or Pt seems to poison the surface, leading to worse sensing performances than pure SnO₂. On the other hand, Pd and Au have been demonstrated to be effective in improving the performance of pure SnO₂. In particular, SnO₂:Au highlighted good sensitivity, cross-selectivity, and humidity-independent behavior for concentrations above 17 RH%. The enhanced reactivity could be explained by the decoration of the SnO₂ surface with Au nanoparticles that positively affected the concentration of active sites and acid-base properties of the surface. The effect of catalytic additives on the intrinsic and extrinsic active centers of SnO₂ is an important factor that determines the proper conditions for an effective gas-solid interaction. In fact, modification with Au has different effects on the concentration and reactivity of the active sites of pristine SnO₂. As a result, modification with Au may cause an increase in the concentration of chemisorbed oxygen species and acid sites, which are the centers of interaction with SO₂. Therefore, for the future works, additional studies should be performed under controlled conditions in the presence of the target molecules with the use of in situ or operando techniques, such as operando DRIFT spectroscopy, or chemisorption analysis to explore the active sites of the material in depth.