Preparation of Powders Containing Sb, Ni, and O for the Design of a Novel CO and C₃H₈ Sensor

Abstract

In this work, powders of NiSb₂O₆ were synthesized using a simple and economical microwave-assisted wet chemistry method, and calcined at 700, 800, and 900 °C. It was identified through X-ray diffraction that the oxide is a nanomaterial with a trirutile-type structure and space group P4₂/mnm (136). UV–Vis spectroscopy measurements showed that the bandgap values were at ~3.10, ~3.14, and ~3.23 eV at 700, 800, and 900 °C, respectively. Using scanning electron microscopy (SEM), irregularly shaped polyhedral microstructures with a size of ~154.78 nm were observed on the entire material’s surface. The particle size was estimated to average ~92.30 nm at the calcination temperature of 900 °C. Sensing tests in static atmospheres containing 300 ppm of CO at 300 °C showed a maximum sensitivity of ~72.67. On the other hand, in dynamic atmospheres at different CO flows and at an operating temperature of 200 °C, changes with time in electrical resistance were recorded, showing a high response, stability, and repeatability, and good sensor efficiency during several operation cycles. The response times were ~2.77 and ~2.10 min to 150 and 200 cm³/min of CO, respectively. Dynamic tests in propane (C₃H₈) atmospheres revealed that the material improved its response in alternating current signals at two different frequencies (0.1 and 1 kHz). It was also observed that at 360 °C, the ability to detect propane flows increased considerably. As in the case of CO, NiSb₂O₆’s response in propane atmospheres showed very good thermal stability, efficiency, a high capacity to detect C₃H₈, and short response and recovery times at both frequencies. Considering the great performance in propane flows, a sensor prototype was developed that modulates the electrical signals at 360 °C, verifying the excellent functionality of NiSb₂O₆.

1. Introduction

In recent years, atmospheric pollution due to CO, CO₂, NO₂, SO₂, and H₂S has caused a decrease in air quality [1,2]. Emissions of toxic gases into the atmosphere by industries and internal combustion engines are considered the main sources that cause air pollution and the emergence of various respiratory diseases [2,3]. An alternative to somehow counteract these problems is to constantly control and monitor the concentration of toxic gases by designing simple electronic devices that meet the features of high-performance environmental gas sensors [4,5]. Several research groups and sensor manufacturing companies have worked on a wide variety of materials, mainly semiconductors, that could act as detectors for different polluting gases [4,6]. With such materials, toxic gas (such as CH₄, LPG, and H₂) leak detectors have been manufactured to monitor the emission of pollutants thrown by automobiles, etc. [6].

Nowadays, there is a great variety of gas sensors that can be classified according to their electrolytic, catalytic, electrochemical, and semiconductor properties [7,8]. The latter have been the most studied due to their excellent detection properties, having a high degree of sensitivity, good efficiency, wide detection of different gases (selectivity), as well as good performance at high operating temperatures (200–400 °C) in unsuitable humidity conditions [4,5,6]. To determine if a material can be used as a gas detector, it must detect low gas concentrations, detect a specific gas (selectivity), obtain a signal in the presence of the test gas in a relatively short time, and return to its initial state when finishing the detection process [9,10]. It has been reported in the literature that the most effective materials that can be applied as gas detectors are semiconductor metal oxides because they possess great advantages, such as thermal stability and high sensitivity [9,11]. It has been recognized that the detection ability of semiconductor oxides is based on reduction–oxidation reactions that occur between the material’s surface and the test gas [11,12]. As a result of these reactions, an electronic variation is produced on the surface that can be verified by the change in capacitance, electrical resistance, work function, or changes in its optical properties [9]. Among the wide variety of investigations on gas sensors, the binary compounds SnO₂ and ZnO [13] and the ternaries SmCoO₃ and ZnAl₂O₄ have been the most studied [14]. In fact, several studies have indicated that the trirutile-type antimonates ZnSb₂O₆ and CoSb₂O₆ are good candidates for application as gas sensors due to their interesting detection properties [15].

As for the nickel antimoniate, NiSb₂O₆, this is a trirutile-type oxide that conforms to the formula of the antimonates ASb₂O₆ (where A is replaced, in this case, by the Ni cation) [16]. This family of compounds has interesting optical, catalytic, and electrical properties that allow them to be applied, for example, in lithium-ion batteries and in many other fields [16,17]. These compounds have traditionally been synthesized using the ceramic method, obtaining different microstructures [17]. However, better results have been achieved using chemical methods, obtaining particles of the order of nanometers, which is a crucial factor in the ability to detect different types of gases [17,18]. We conducted extensive literature research, finding out that NiSb₂O₆ has already been studied as a potential sensor of static (for CO and C₃H₈) and dynamic (for LPG) atmospheres. However, we found no evidence that NiSb₂O₆ has yet been dynamically tested in CO and C₃H₈. Therefore, to contribute to the studies of this material as a gas sensor, NiSb₂O₆ powders were synthesized to evaluate their ability to detect carbon monoxide and propane gases at different concentrations and temperatures, achieving excellent results. Additionally, based on the ability of the studied oxide to detect high and low concentrations of both toxic gases, an electronic circuit was successfully designed for a high-efficiency and fast response gas sensor in such atmospheres.

2. Materials and Methods

2.1. Synthesis

The synthesis of the NiSb₂O₆ powders was performed using a microwave-assisted wet chemical method. An outline of the preparation process can be seen in Figure 1. For this, ~1.4533 g of Ni(NO₃)₂·6H₂O (Sigma-Aldrich, 99%), ~2.2803 g of SbCl₃ (Aldrich, 99%), 2 mL of ethylenediamine C₂H₈N₂ (Sigma-Aldrich, 99%), and ethyl alcohol (CTR, 99.5%) as a solvent were used. Each salt was dissolved in 5 mL of ethyl alcohol, and the ethylenediamine was dissolved in 10 mL of the solvent. The three solutions were left stirring on a magnetic plate at room temperature for approximately 20 min. Once the salts were completely dissolved, the solutions were mixed with ethylenediamine to form a suspension, which was left under moderate stirring for 24 h at room temperature. After that, the solvent was evaporated by microwave radiation. This was accomplished using a domestic microwave oven (General Electric, model ST0912C01352). Evaporation consisted of applying a power of 70 W at exposure intervals of 60 s. After this, a green paste was obtained, which was dried in a Novatech muffle at 200 °C for 8 h. Then, the powders were calcined at 700, 800, and 900 °C for 5 h in the same muffle. The calcination ramp was 100 °C/h.

2.2. Physical Characterization

The crystal analysis of the NiSb₂O₆ powders was performed at 700, 800, and 900 °C through X-ray powder diffraction (Panalytical Empyream). Cuα radiation (λ = 1.5406 Å) was used, scanning from 10 to 90° at a rate of 0.02°/second. To determine the forbidden band gap of the calcined oxide powders at 700, 800, and 900 °C, a spectrophotometer (UV–Vis–NIR, model UV-3600 Plus) was used coupled to a deuterium and tungsten-halogen lamp. The absorbance spectra were taken in a range of 200 to 800 nm. The microstructure was analyzed with field-emission scanning electron microscopy (FE-SEM, Tescan MIRA 3 LMU) with an acceleration voltage of 10 kV in a high vacuum.

2.3. Tests in CO

To evaluate the capacity and efficiency of the NiSb₂O₆ powders calcined at 900 °C for detecting air–CO concentrations, experiments were carried out in static and dynamic atmospheres at different temperatures (100–300 °C). In both cases, pellets were made with 0.4 g of powder using a hydraulic press (Simplex-Italy Equip-25 Ton) compressing at 10 tons for 20 min. The dimensions of the pellets were 12 mm in diameter and 0.5 mm thick. Then, two colloidal silver paint ohmic contacts (Alfa Aesar 99%) were placed on the pellets for having good contact with the system’s electrodes (see Figure 2). Subsequently, the pellets were taken to the measurement chamber.

In static atmospheres, changes in NiSb₂O₆’s electrical resistance were recorded as a function of gas concentration (1, 50, 100, 200, and 300 ppm of CO) and operating temperature (100, 200, and 300 °C). The sensitivity magnitude in the static tests was estimated using the equation S = (G_G − G_O)/G_O [19], where G_G and G_O are the electrical conductances in CO and air, respectively. The electrical conductance (1/electrical resistance) was considered according to reference [20]. For the tests in dynamic atmospheres at a constant temperature of 200 °C, an adaptation was made to the measurement system by placing a metal box (volume = 19 cm³) inside of it. The box had two perforations: one for the CO inlet and the other one for the test gas outlet and for introducing the electrodes in contact with the pellets’ surface (see in Figure 2). Subsequently, the gas was extracted using a pump (with a capacity of 10⁻³ Torr) incorporated in the measurement chamber. The detection tests consisted of supplying two different flows (300 and 400 cm³/min) of extra dry air (21% O₂ + 79% N₂) as stabilizing gas for the pellets’ surface. The CO flows were 100, 150, and 200 cm³/min. The partial pressure of the CO flows in static and dynamic atmospheres was monitored with a Leybold TM20 electronic detector. Changes in the test gas electrical resistance were recorded with a Keithley 2001 multimeter coupled to a data acquisition and control system using the LabView 8.6 software (National Instruments). Brooks Instruments mass flow regulators with capacities of 2600 cm³/min (GF100CXXC-SH452.6L) and 10 cm³/min (GF100CXXC-SH40010C) were used for the CO flows in the dynamic tests.

2.4. Tests in Propane

For the dynamic tests in air–propane atmospheres, thick films were prepared with the NiSb₂O₆ powders calcined at 900 °C. For this, a ceramic base was used with a circular cavity in the center and 4 ceramic columns around its perimeter. A hole was made in the middle of each column through which high purity platinum wires (0.006 in diameter) were introduced, functioning as electrodes connected to the detection system. For the preparation of the thick films, approximately 0.4 g of the material was placed in a vial with 3 mL of isopropyl alcohol and dispersed with an ultrasonic generator (Brason 2510 Ultrasonic). The dispersed material was then drip deposited into the circular cavity, forming films with dimensions of ~0.5 mm thick and ~0.3 mm in diameter. Subsequently, the films were subjected to a heat treatment at 300 °C for 4 h in a programmable muffle (Vulcan, model 5‒550).

The films were placed on an aluminum plate inside a quartz tube and in a tubular oven with programmable temperature control (Lindberg/blue). Air and propane flows were controlled by the same mass flow regulators as for CO. The changes in the sensitivity magnitude |Z| were recorded with an Agilent 4263B multimeter controlled through the LabView 8.6 software (National Instruments).

3. Results and Discussion

3.1. XRD Analysis

Figure 3 shows the crystal evolution of the NiSb₂O₆ powders calcined at 700, 800, and 900 °C. In these diffraction patterns, it is observed that the material showed its characteristic peaks at the three calcination temperatures.

However, portions of secondary phases to the compound’s main phase were identified for each heat treatment. At 700 °C, although the peaks of the material’s crystalline structure were obtained, secondary phases were identified for Sb₂O₃ (PDF # 71‒0383) and Sb₂O₅ (PDF # 50‒1376), located at the 2θ points 28.14°, 28.96°, 30.11°, and 33.78° corresponding to the fewer intensity peaks. At 800 °C, crystal planes of NiSb₂O₆ with more intense peaks of greater purity and crystallinity were observed. These reflections were identified at 2θ = 19.22°, 21.43°, 27.14°, 33.50°, 34.99°, 38.80°, 40.19°, 44.73°, 53.22°, 56.03°, 60.23°, 62.77°, 63.32°, 67.20°, 67.78°, 73.89°, 80.89°, and 86.83° (PDF # 38‒1083). In addition, a small phase associated with NiO (PDF # 44‒1159) was located at 2θ = 37.19°. At 900 °C, planes attributed to NiO were again found, but with less intensity than at 800 °C. As expected, comparing the results of Figure 3 with PDF # 38‒1083, it was found that NiSb₂O₆ belongs to the trirutile-type family of materials [21], with a tetragonal structure (network parameters: a = 4.641 Å, c = 9.223 Å) and space group P4₂/mnm [22].

Based on the diffractograms’ peaks, the crystallite size was estimated with Scherrer’s equation [23] as follows:

$$\mathbf{t}=\frac{0.9\mathsf{\lambda }}{\mathsf{\beta }\mathbf{c}\mathbf{o}\mathbf{s}\mathsf{\theta }}$$

(1)

where λ is the wavelength of the radiation (in this case, Cu = 1.5406 Å), β is the width of the peak measured at half the maximum intensity, and θ is the Bragg angle. For the calculation, the most intense peak corresponding to the plane (110) located at 2θ = 27.14° was considered. From the diffractograms depicted in Figure 4a, at 700 °C, the crystal size was estimated at ~36.43 ± 1.65 nm; at 800 °C, it was ~38.40 ± 0.96 nm; and at 900 °C, it was ~44.02 ± 0.50 nm. Accordingly, it was verified that when increasing the temperature, the crystallite size increased. This is attributed to the fact that a particles’ agglomeration occurred due to the increasing calcination temperature [24,25] and to the material’s residence time in the muffle at the given temperature. It is worth mentioning that the time for the thermal treatments was relatively short (5 h), thus achieving that meant the crystal size stayed in the order of nanometers [25].

Figure 4b,c show a projection of the octahedral gap formed by the Ni²⁺ and Sb⁵⁺ cations, which are coordinated with the oxygen anions O²⁻. Additionally, Figure 4c depicts the atomic arrangement of NiSb₂O₆’s crystal structure seen from the plane of greatest intensity (110). Both structures were obtained with the Diamond software using PDF # 38‒1083 and reference [26]. It is important to mention that the cations’ arrangement forms a triplicate rutile-type structure in the c-axis and, according to the literature, this modification on the c-axis of NiSb₂O₆ represents a trirutile-type structure [16,17,27].

Comparing our results with those of reference [17], where the same compound was synthesized, we obtained the crystalline phase at a shorter exposure time (5 h) than that of such reference, where at 800 °C, the residence time was of 3 days. Furthermore, in our case, a shorter preparation time was required than those reported in references [16,17], where they used the solid-state reaction method, requiring significantly longer preparation times.

3.2. UV–Vis Analysis

Figure 5a shows the characteristic UV‒Vis absorption spectra of NiSb₂O₆ calcined at 700, 800, and 900 °C in a wavelength range of 200 to 800 nm (1.55 to 6.2 eV). NiSb₂O₆’s absorption bands were identified in the range from 200 to 500 nm [16,28], which are characteristic of semiconductor oxides with a trirutile-type structure [29,30].

In the range from 600 to 700 nm, the bands corresponding to the electronic interactions occurring in the oxygen–metal (O‒Ni) bond [31,32,33] were recorded. To estimate the values of the forbidden bandwidth (Eg) of the NiSb₂O₆ powders, Tauc’s equation was used [34].

$$\propto \left(\mathit{E}\right)=\mathit{A}\frac{{\left(\mathit{E}-{\mathit{E}}_{\mathit{g}}\right)}^{\frac{1}{2}}}{\mathit{E}},$$

(2)

where E is the energy of the incident photon, α is the optical absorption coefficient, Eg is the forbidden bandwidth, and A is the proportionality constant. In our case, a direct transition (n = 1/2) was considered for NiSb₂O₆ [16,28]. Tauc’s equation was used to elaborate the graph of (αℎυ)² vs. E, superimposing a straight line on the spectrum’s final part (Figure 5b). With this, the bandgap energy of the calcined material at 700, 800, and 900 °C was estimated at ~3.10 ± 0.02, ~3.14 ± 0.04, and ~3.23 ± 0.03 eV, respectively. According to Figure 5, the material’s crystal size strongly influenced the variation of the forbidden band [25,28]. In our case, as the calcination temperature increased, the forbidden band value had a slight increase. This is attributed to the fact that, when the calcination temperature rose from 700 to 900 °C, the crystal size increased (see Figure 4a), provoking the energy variations of NiSb₂O₆.

Then, the crystal size, whether it increases or decreases, has a very significant effect on the compound’s bandgap value. In the literature, bandgap values of ~2.7–2.8 eV at a calcination temperature of 700 °C and ~2.9 eV at 800 °C for materials such as our oxide were reported [16]. In reference [28], a bandgap value of ~3.92 eV was estimated for NiSb₂O₆. In comparison, our results are within the accepted energy range by different studies for the same or similar material to ours (see Figure 5) [16,25,29,35].

3.3. SEM Analysis

Five SEM micrographs of NiSb₂O₆ powders calcined at 900 °C are shown in Figure 6. These images show that the material’s surface was made up of polyhedron-shaped particles and irregular particles of different sizes (Figure 6a–e). Figure 6a,b show a large agglomeration of these particles, which were distributed over all the analyzed areas. Additionally, it is clear to observe that the polyhedra were composed of smaller, agglomerated particles, which gave rise to the microstructures shown in Figure 6c–e.

The morphologies shown in Figure 6 are attributed to the fact that the microstructural features are strongly related to the synthesis method [24]. Reference [36] reported that the preparation route, the calcination temperature, and the application of microwave radiation favor obtaining specific morphologies, such as octahedral structures, spheres, wires, and hexagonal nanostructures, among others. On the other hand, other studies informed that chemical methods are desirable in the formation of specific microstructures adequate for gas detection because the nucleation and growth processes that take place with them favor the control of the particles’ nanometric size [37,38]. In our case, the use of a wet-chemistry process assisted with microwave radiation in the presence of ethylenediamine favored the growth of morphologies such as those presented in Figure 6a‒e. The results followed the nucleation and growth principles proposed in reference [18].

From Figure 6, the particle size was estimated considering areas of the material’s surface where the particles were clearly identifiable. A size in the range of 40–150 nm was estimated, with an average of ~92.30 nm and a standard deviation of ~± 24.64 nm (see Figure 7a). The size of the octahedral polyhedra was estimated in the range of 80–240 nm, with an average of ~154.78 nm and a standard deviation of ~±32.55 nm (Figure 7b).

3.4. CO Analysis

The gas response was analyzed with the material synthesized at 900 °C, where a minor presence of the secondary phase was obtained. To evaluate NiSb₂O₆’s detection capacity at different temperatures (100–300 °C) and CO concentrations (1–300 ppm), tests were conducted in static atmospheres using pellets made with powders of the material calcined at 900 °C. The results are shown in Figure 8 as a function of the test gas concentration (Figure 8a) and operating temperature (Figure 8b).

According to the analysis, the pellets did not show changes in electrical resistance at 100 °C regardless of the increase in the CO concentration. In contrast, at 200 °C, a small variation in resistance was recorded, obtaining a sensitivity magnitude of ~3.14 at 300 ppm CO. When increasing the operating temperature to 300 °C, very significant changes in electrical resistance took place, with a maximum sensitivity magnitude of ~72.67 at 300 ppm of CO. During the experiments, it was observed that the increase in the CO concentration caused the increase in the pellets’ response (see

Table 1. Response of NiSb₂O₆ pellets at different concentrations and operating temperatures.

Temperature (°C)	CO (ppm)	Response (R)	Temperature (°C)	CO (ppm)	Response (R)
	100	0.87		100	35.52
200	200	0.67	300	200	55.50
	300	3.14		300	72.67

).

This was due to the fact that when the test gas was in contact with the oxygen on the pellets’ surface, it oxidized, causing a transfer of charge carriers on the material’s surface [11,39,40]. In the literature, it is reported that the increase in response in materials such as NiSb₂O₆ at temperatures above 150 °C is due to the ${\mathbf{O}}^{-}$ and ${\mathbf{O}}^{2-}$ ionic oxygen species [37,39], which are very reactive at relatively high temperatures (in our case, at 300 °C) [11,39]. At temperatures below 200 °C, the oxygen species that appear the most are of the ${\mathbf{O}}_2^{-}$ type [11,39]. Such species are less reactive at temperatures below 100 °C because the thermal energy supplied is not enough to favor chemical reactions between the oxygen species, the CO, and the surface of the pellets, causing low kinetic interaction between the material nanoparticles and the CO molecules [40,41,42]. This resulted in not recording significant increases in the pellets’ response at 100 °C (see Figure 8a,b). It is important to mention that the excellent response and good stability recorded at 300 °C were mainly associated with the microstructural features obtained, which led to an increase in catalytic sites for the gas reaction with the nanoparticles’ surface [43,44,45].

The dependence of gas sensing on the operating temperature is due to several facts. One of them is that the type of oxygen species adsorbed on the oxide depends on temperature [28]. As mentioned, atomic species, which are more reactive than molecular species, predominate at temperatures above 150 °C [13]. Another fact is that the oxidation reaction is an activated process, where its rate increases with temperature. Thermal energy is needed to activate the adsorption of ionized oxygen species and overcome the barriers of the sensing reactions [20]. Additionally, desorption and diffusion processes are also temperature dependent. Thus, the sensitivity can rise appreciably with increasing temperature [24]. Different literature reports have shown that the gas response at low temperatures starts to disappear at around 200 °C and is replaced by the typical sensitivity response at high temperatures [13,20]. The optimum temperature depends on the sensing material and the sensed gas. Therefore, as in previous reports [29,43], a noticeable response variation was observed at 300 °C. Comparing our results with references [29,42,43,46], where similar experiments were carried out in static gas atmospheres, we found that our pellets had a greater response and a better performance in detecting CO at different operating temperatures (100–300 °C).

We also carried out tests in dynamic atmospheres as a function of time, applying direct current (d. c.) signals at 200 °C. For this, extra dry air flows (21% O₂ + 79% N₂; 400 cm³/min for 7 min at the test’s start and 300 cm³/min for 13 min at the test’s end) were supplied to the measuring chamber to stabilize the pellets’ surface, while the CO flows were 100, 150, and 200 cm³/min. The total time for the dynamic cycles was approximately 20 min. The results are shown in Figure 9.

It is observed in Figure 9a that, as the CO flow increases, pellets’ electrical resistance decreases significantly. The estimated drop in electrical resistance was ~1008.59 MΩ at 100 cm³/min, ~1273.27 MΩ at 150 cm³/min, and ~1474.49 MΩ at 200 cm³/min. The response and recovery times were ~2.24 and ~1.39 min at 100 cm³/min, and ~2.77 and ~2.00 min at 150 cm³/min. At 200 cm³/min, the response time was ~2.10 min. The response and recovery times were calculated according to reference [47], where 90% of the electrical resistance in the test gas and 10% of the resistance in the air are considered. We observed that the response and recovery times in the dynamic tests (see Figure 9) were short. That was because by introducing the CO into the measurement chamber, it reacted immediately with the material’s surface due to the effect of temperature, with the consequent immediate change in electrical resistance (that is, the response time). On the other hand, when the test gas stopped injecting and the air was introduced (during the recovery time), the material tended to return almost immediately to its initial resistance values, which meant that the detection process was reversible.

The changes in electrical resistance shown in Figure 9a are typical for n-type semiconductors [11]; however, it is reported in the literature that NiSb₂O₆ tends to be a p-type semiconductor when exposed to atmospheres (for example, CO₂ and LPG) such as ours (CO) [28]. We also made measurements in large air flows (400 and 300 cm³/min), obtaining a supersaturation of oxygen species $\left({\mathbf{O}}^{-}\right)$ on the pellets’ surface, originating an oxidation process with the test gas [45]. This caused all the charge carriers (electrons and holes) to interact with each other, allowing competition between them and resulting in a majority flow of electrons on the pellets’ surface [48]. Another factor that influenced the trends shown in Figure 9a was the operating temperature [12,40]. According to the literature, the oxygen species on the pellets’ surface are adsorbed by the effect of temperature (in our case at 200 °C) [11], releasing electrons that are then captured to form ionized species (${\mathbf{O}}_2^{-},{\mathbf{O}}^{-},\mathrm{or} {\mathbf{O}}^{2-}$) [49]. This means that, when the pellets’ surface comes into contact with the CO molecules at 200 °C, the latter react with the ionized species, causing an electron transfer during the adsorption and desorption of the gas molecules on the material’s surface [11] and, with it, the decrease in electrical resistance depicted in Figure 9a. Shifts in percent sensitivity as a function of time are shown in Figure 9b. The response (sensitivity) was determined considering the variations in electrical resistance for 100, 150, and 200 cm³/min of CO at 200 °C when the surface of the pellets was stabilized by means of extra-dry air flows, as in the case shown in Figure 9a. To obtain the graphs of Figure 9b, the following formula was considered [50]: S(%) = (G_G − G_O)/G_O × 100, where G_G and G_O are the electrical conductances (1/electrical resistance) in CO and air, respectively. From the results, for the flow of 100 cm³/min, the estimated sensitivity was ~57.10% (first curve). For the 150 cm³/min flow, the sensitivity increased to ~84.77% (second curve). Finally, for the flow of 200 cm³/min, the percentage obtained was ~113.75%. As expected, increasing the CO flow increased the pellets’ percent sensitivity. With this, it can be corroborated that the increase in time-dependent sensitivity in a material such as NiSb₂O₆ is closely related to the variation in the test gas concentration [29]. Indeed, it is reported in the literature that as the concentration of the test gases increases, the sensitivity of the material increases [32,34]. This is due to the fact that the higher the gas concentration value is, the more easily it reacts with the oxygen species (${\mathbf{O}}^{-}$) on the pellets’ surface due to the temperature effect (in our case, at 200 °C), causing an increase in the gas oxidation and, thus, an electron release, causing NiSb₂O₆’s sensitivity to rise [20,34]. Some studies report that the adsorption of oxygen species on a semiconductor’s surface due to the operating temperature effect is what causes the high sensitivity of materials such as ours [28,29,35,48,49,51]. In particular, our results (displayed in Figure 9) show better thermal stability, high sensitivity, reproducibility, and efficiency in the detection of different CO concentrations compared to those reported in references [14,28,29,31,34,35,42,46,49], where tests in static and dynamic CO, CO₂, and LPG atmospheres were performed.

A mechanism that explains the interaction between a material such as NiSb₂O₆ and a gas such as CO is based on the drop of the depletion layer due to oxygen adsorption [51,52]. When a gas such as CO is adsorbed on the material’s surface, it reacts with the chemisorbed oxygen species [52], producing CO₂ molecules and a release of electrons that move to the conduction band [20]. This produces changes in the material’s conductance (or electrical resistance) [53] and, therefore, variations in the electrical response in CO atmospheres (see Figure 9a,b). Therefore, since the reaction mechanism between a gas such as CO and the surface of a semiconductor such as NiSb₂O₆ depends largely on the operating temperature [20,52], a reaction that may be occurring on the pellets’ surface at 200 °C is as follows [20,41,49,52,54]:

$${\mathbf{O}}_2^{-}+2\mathbf{C}\mathbf{O} \to 2\mathbf{C}{\mathbf{O}}_2+{\mathbf{e}}^{-}$$

(3)

where ${\mathbf{O}}_2^{-}$ is the oxygen ions on the material’s surface and ${\mathbf{e}}^{-}$ is the electrons in the bulk material [20,49,54]. Furthermore, the response depicted in Figure 9 depended on the partial pressure of the oxygen in the extra dry air [44,49,54], as well as on the porosity, morphology, and particle size [55]. If the average particle size decreases, NiSb₂O₆’s surface area increases [4] and, consequently, the gas–solid interaction is favored [55]. In our case, we obtained a material with an average particle size of ~95.63 nm, which meant better oxygen adsorption and desorption on the material’s surface and, therefore, a high NiSb₂O₆ response, as shown in Figure 9a,b.

3.5. Propane Analysis

Figure 10 shows the results of the dynamic response at 0.1 and 1 kHz of NiSb₂O₆ films exposed to propane at 360 °C. In Figure 10a,b, the results are shown for films stabilized through a flow of extra dry air (1500 cm³/min) for 5 min. Then, without removing the air, 600 ppm of propane was injected into the measuring chamber for 5 min, recording the decrease in sensitivity magnitude |Z| by varying the frequency from 0.1 to 1 kHz. The values of |Z| at 0.1 kHz were in the range of 6.6–7.05 MΩ, with response and recovery times of 42.96 and 35.56 s, respectively. At 1 kHz, the |Z| values were 1.74–1.75 MΩ, with response and recovery times of 11.91 and 5.88 s, respectively. The response and recovery times were calculated for both frequencies considering 90% of the response in the test gas and 10% of the value when exposed only to air [47]. Although NiSb₂O₆ is a p-type semiconductor [28], the trend in Figure 10a,b is attributed to the fact that the decrease in |Z| is caused by the continuous flow of oxygen-rich air (21% O₂), which causes oxidation (${\mathbf{O}}^{-}$) [48,54,56] and an oxygen saturation (${\mathbf{O}}^{-}$ and ${\mathbf{O}}^{2-}$) [29,35,43,54,57] that reacts with the propane on the films’ surface, causing the excellent response shown in Figure 10a–e.

We observed that the good dynamic response of the NiSb₂O₆ films in propane atmospheres was caused by the rapid adsorption of the gas that reacted with the film’s surface, causing an electron exchange between the oxygen species adsorbed on the surface and the surface itself [48]. It has been reported in the literature that the electrons on the material’s surface are the most probable cause of the electrical response variation and, therefore, of the improvement in the sensor response [11,48,50,58]. The variations in the impedance magnitude (|Z|) and in the response and recovery times were due to the chemical reaction between the surface of the thick films and the propane gas. The reaction occurred because there was a high kinetic activity of the gas molecules due to the operating temperature, causing an immediate response (response time) to 600 ppm of the gas. As in the case of CO, when the propane stopped injecting and the air was introduced into the system, the impedance values tended to immediately return to their stable values (recovery time). It was corroborated that at 0.1 and 1 kHz, the response and recovery times decreased significantly compared to other semiconductors tested as gas sensors [14,28,34,58].

The reaction mechanism that could explain the adsorption and desorption of propane gas molecules on a semiconductor’s surface, such as the one studied here, is not yet fully established. Some authors report that the propane molecules on the film’s surface react with adsorbed oxygenated species, producing CO₂, water vapor, and the release of electrons on the surface [59]. Other authors believe that a reaction that could be occurring between a given semiconductor and propane gas could be the following: C₃H₈ + 10${\mathbf{O}}^{-}$→ 3CO₂ + 4H₂O + e⁻ [35,60]. This means that when the propane gas is adsorbed, it dissociates on the material’s surface because of the temperature (in our case at 360 °C), causing changes in the sensor’s electrical response (see Figure 10a–e).

At this point, it is convenient to address several sensing mechanisms of metal oxide semiconductors in more detail. These mechanisms can be classified into the following two categories: microscopic and macroscopic [61]. The first one comprises the mechanisms of Fermi level control, grain boundary barriers control, and electron-depletion and hole-accumulation layers. Additionally, the changes in electrical properties are accompanied by changes in physical properties such as energy bands and work functions.

The macroscopic category includes mechanisms of adsorption-desorption patterns, bulk resistance control, and gas diffusion control. In the case of bulk resistance control, the change in the electrical properties of the sensors can be caused by phase transformations, which occur in certain types of materials. The gas diffusion control mechanism dictates that the most important factor in the diffusion process is related to the microstructure of the material, which, in turn, affects the surface chemical reactions and the sensing properties [61,62].

The adsorption/desorption model is the current conventional mechanism, which states that the electron capture by adsorbed species, and their subsequent desorption, is responsible for changes in the sensor’s electrical properties. This model includes oxygen adsorption, chemical adsorption/desorption, and physical adsorption/desorption. The theory of electron depletion and hole accumulation layers is an extension of the oxygen adsorption model. As mentioned in the discussion section, the material’s sensitivity is primarily related to the oxygen chemisorption, where the formed ionic or molecular species are temperature dependent. Chemical adsorption/desorption also extends to the material’s exposure to different gases [11,20,61,62].

Various theories have been developed on the adsorption of gases. For example, Freundlich and Kuster’s mathematical modeling of an isotherm for gas adsorbates, Langmuir’s isotherm theory of adsorption of the unimolecular layer, the BET theory of multilayer adsorption, and dynamic adsorption and desorption processes based on Langmuir’s theory [63]. Additionally, the adsorption of toxic small molecules has been theoretically investigated through adsorption configurations, adsorption energy, charge transfers, and electronic properties. Eight possible orientations of CO adsorption are documented for pentagraphene-based materials, where chemical adsorption is favored more than physical adsorption [64]. Likewise, in CO sensing using GaN, four stable gas adsorption configurations have been reported [65]. Theoretical studies on propane adsorption have also been conducted, where it did not have a significant site preference on the studied surfaces [66].

For the design of our 600-ppm propane detector, changes in $\left|\mathbf{Z}\right|$ at 1 kHz and 360 °C were considered. The calculations shown in Figure 10c–e were conducted with the following formula:

$$\mathit{Z}\left(\mathit{t}\right)=\left|\mathit{Z}\left(\mathit{t}\right)\right|{\mathit{e}}^{\mathit{\theta }\left(\mathit{t}\right)}=\mathit{r}\left(\mathit{t}\right)+\mathit{i}{\mathit{X}}_{\mathit{c}}=\frac{1}{2\mathit{\pi }\mathit{f}\mathit{C}}$$

(4)

where $\mathit{i}$ is an imaginary operator, $\mathit{r}\left(\mathit{t}\right)=\left|\mathit{Z}\left(\mathit{t}\right)\right|\cos \mathit{\theta }\left(\mathit{t}\right)$ is the resistance (Figure 10c), ${\mathit{X}}_{\mathit{c}}=\left|\mathit{Z}\left(\mathit{t}\right)\right|\mathrm{sen}\mathit{\theta }\left(\mathit{t}\right)$ is the reactance (Figure 10d), and $\mathit{C}\left(\mathit{t}\right)=\frac{1}{2\mathit{\pi }\mathit{f}\left|\mathit{Z}\left(\mathit{t}\right)\right|\mathrm{sen}\mathit{\theta }\left(\mathit{t}\right)}$ is the capacitance in faradays (Figure 10e).

Considering the values in Figure 10c–e, we developed our electrical circuit using the following equation:

$${\mathit{z}}_{\mathit{s}}\left(\mathit{t}\right)={\mathit{z}}_{\mathit{c}}+{\mathit{z}}_{\mathit{d}}\left(\mathit{t}\right)={\mathit{z}}_{\mathit{c}}+\frac{{\mathit{r}}_{\mathit{d}}\left(\mathit{t}\right)}{{\mathit{r}}_{\mathit{d}}\left(\mathit{t}\right)+{\mathit{X}}_{\mathit{d}}\left(\mathit{t}\right)}$$

(5)

where ${\mathit{z}}_{\mathit{s}}\left(\mathit{t}\right)$ is the sensor’s impedance, ${\mathit{z}}_{\mathit{c}}$ is the constant impedance, and ${\mathit{z}}_{\mathit{d}}\left(\mathit{t}\right)$ is the sensor’s dynamic impedance in the test gas. Therefore, if we consider an operation point at $\left[\overline{\mathit{t}}, \mathit{z}\left(\overline{\mathit{t}}\right)\right]$ and $\left[\overline{\mathit{t}}, \mathit{\theta }\left(\overline{\mathit{t}}\right)\right]$ [67], the impedance would be as follows:

$${\mathit{z}}_{\mathit{s}}\left(\mathit{s}\right)={\mathit{z}}_{\mathit{c}}+{\mathit{z}}_{\mathit{d}}\left(\mathit{s}\right)={\mathit{z}}_{\mathit{c}}+\frac{{\mathit{r}}_{\mathit{d}\mathit{s}}}{{\mathit{r}}_{\mathit{d}\mathit{s}}+{\mathit{X}}_{\mathit{d}\mathit{s}}}.$$

(6)

Substituting the identity ${\mathit{X}}_{\mathit{d}\mathit{s}}=\frac{1}{\mathit{s}{\mathit{C}}_{\mathit{d}\mathit{s}}}$, the impedance takes the following form:

$${\mathbf{z}}_{\mathbf{s}}\left(\mathbf{s}\right)={\mathbf{z}}_{\mathbf{c}}+{\mathbf{z}}_{\mathbf{d}}\left(\mathbf{s}\right)={\mathbf{z}}_{\mathbf{c}}+\frac{{\mathbf{r}}_{\mathbf{d}\mathbf{s}}{\mathbf{C}}_{\mathbf{d}\mathbf{s}}\mathbf{s}}{{\mathbf{r}}_{\mathbf{d}\mathbf{s}}{\mathbf{C}}_{\mathbf{d}\mathbf{s}}\mathbf{s}+1}.$$

(7)

If $\overline{\mathit{t}}=300\mathit{s}$, we obtain ${\mathit{z}}_{\mathit{c}}=1754362{\mathit{e}}^{-74.3106}$, ${\mathit{r}}_{\mathit{d}\mathit{s}}=18167 \mathbf{\Omega }$, ${\mathit{C}}_{\mathit{d}\mathit{s}}=9.37 \mathrm{pF}$, and the impedance as follows:

$${\mathit{z}}_{\mathit{s}}\left(\mathit{s}\right)=1754362{\mathit{e}}^{-74.3106}+\frac{0.0001704\mathit{s}}{0.0001704\mathit{s}+1}.$$

(8)

From Equation (8) and using the impedance ${\mathit{z}}_{\mathit{d}}\left(\mathit{s}\right)$, we obtained the Bode diagram shown in Figure 11a.

Comparing this diagram with NiSb₂O₆’s electrical response (Figure 10a,b), the results are consistent because $\left|\mathit{Z}\right|$ decreases when the frequency increases from 0.1 to 1 kHz.

The detector’s electronic diagram in propane atmospheres is shown in Figure 11b. NiSb₂O₆ pellets working as detectors were placed in one of the Wheatstone bridge’s arms, whose output voltages were ${\mathit{V}}_{\mathit{a}}=\frac{{\mathit{z}}_1}{{\mathit{z}}_{\mathit{s}}+{\mathit{z}}_1}{\mathit{V}}_{\mathit{a}\mathit{c}}$ and ${\mathit{V}}_{\mathit{b}}=\frac{{\mathit{z}}_2}{{\mathit{z}}_{\mathit{c}\mathit{a}}+{\mathit{z}}_2}{\mathit{V}}_{\mathit{a}\mathit{c}}$, where ${\mathit{z}}_{\mathit{c}\mathit{a}}$ is the variable impedance employed to calibrate the bridge, ${\mathit{z}}_1$ and ${\mathit{z}}_2$ are impedances, and ${\mathit{V}}_{\mathit{a}\mathit{c}}$ is the supply voltage. Voltages ${\mathit{V}}_{\mathit{a}}$ and ${\mathit{V}}_{\mathit{b}}$ are the input signals of an adder‒subtractor circuit, whose output voltage is ${\mathit{V}}_{\mathit{a}\mathit{b}}={\mathit{V}}_{\mathit{a}}-{\mathit{V}}_{\mathit{b}}=\frac{{\mathit{z}}_1}{{\mathit{z}}_{\mathit{s}}+{\mathit{z}}_1}{\mathit{V}}_{\mathit{a}\mathit{c}}-\frac{{\mathit{z}}_2}{{\mathit{z}}_{\mathit{c}\mathit{a}}+{\mathit{z}}_2}{\mathit{V}}_{\mathit{a}\mathit{c}}$. This output voltage is the input signal of the comparator circuit, whose inverter’s input and supply $-{\mathit{V}}_{\mathit{c}\mathit{c}}$ were connected to a common for avoiding negative voltages.

The detector’s operating principle is shown in Figure 11c, which is based on the following two basic stages: calibration and detection. In the calibration stage, the sensor is exposed to the air-atmosphere, where the Wheatstone bridge was calibrated through the variable impedance ${\mathit{z}}_{\mathit{c}}$ until the condition ${\mathit{z}}_1={\mathit{z}}_2={\mathit{z}}_{\mathit{c}}={\mathit{z}}_{\mathit{s}}$ was met, yielding ${\mathit{V}}_{\mathit{a}\mathit{b}}=0 \mathit{V}$ and no alarm signal: ${\mathit{V}}_{\mathit{A}\mathit{l}\mathit{a}\mathit{r}\mathit{m}}\approx 0 \mathit{V}$. In the test gas detection stage, the sensor showed electrical variations in propane atmospheres, whereby the Wheatstone bridge was decalibrated, causing ${\mathit{V}}_{\mathit{a}\mathit{b}}\ne 0$, ${\mathit{V}}_{\mathit{a}}>{\mathit{V}}_{\mathit{b}}$, and, as a result, the comparator circuit to be positively saturated with ${\mathit{V}}_{\mathit{A}\mathit{l}\mathit{a}\mathit{r}\mathit{m}}\approx \mathit{V}\mathit{c}\mathit{c}$, which was the detector’s alarm signal in the test gas.

In general, it is well known that the performance of semiconductors such as ours (NiSb₂O₆) depends, to a great extent, on the microstructure and the crystal size. The decrease in crystal size in semiconductors increases the specific surface area [10,20,24]. Consequently, there is an increase in the number of potentially active catalytic sites that favor the high adsorption of gases on the semiconductor’s surface. That is because the space charge zone, or depletion layer, which is a function of charge carriers, has a thickness L_S of less than 100 nm [61,62]. According to some references, the charged outer layer depends on the partial pressure and the concentration of the test gas [34,55]. However, the conductance (or electrical resistance) is closely related to the crystal size D [34,68], which means that when D is less than 2L (in our case 44.02 ± 0.50 nm), the response of semiconductor oxides is significantly increased [62]. Then, if the crystal (or particle) size is D < 2L, the crystals will participate in the electronic transport during sensing [34,54,68,69,70], provoking variations in the electrical resistance and, therefore, increases in the sensor’s response, high electrical sensitivity, thermal stability, and high efficiency in the gas detection [68,70]. Thus, it is evident that obtaining nanometric particle sizes during the synthesis process will positively impact the detection properties, as reported in this work (see Figure 8, Figure 9 and Figure 10).

4. Conclusions

The synthesis of NiSb₂O₆ powders was conducted successfully by means of a wet chemistry process assisted with microwave radiation. This method allowed us to obtain different morphologies for use as a gas sensor (in this work: CO and propane). Using UV–Vis, forbidden bandgap values of ~3.06, ~3.14, and ~3.25 eV were estimated for the powders of NiSb₂O₆ calcined at 700, 800, and 900 °C, respectively. It was observed that the bandgap value was closely related to the crystallite size (~32, ~40, and ~44 nm at 700, 800, and 900 °C, respectively) and the calcination temperature in the synthesis process. By means of scanning electron microscopy (SEM), microstructures in the form of polyhedra and irregular particles of different sizes, distributed over the entire surface of the material, were observed. This was largely due to the chelating agent (ethylenediamine), the heat treatment, and the synthesis method. In static and dynamic tests using pellets made with powders of the oxide, a good response was obtained at different concentrations and flows of CO and operating temperatures, obtaining a maximum sensitivity of ~72.4 at 300 ppm of CO and 300 °C. In dynamic tests conducted using thick films prepared with the material’s powders, an excellent response was obtained at 360 °C in 600 ppm of propane and applying an alternating current (at 0.1 and 1 kHz). Good stability, reproducibility, performance, and short response and recovery times were found. The excellent material’s dynamic response in propane allowed us to design an efficient prototype for detecting atmospheres of such gas at different working frequencies. The synthesis method used and the high material’s crystallinity contributed greatly to obtaining excellent results in the CO and propane detection tests; therefore, we can affirm that NiSb₂O₆ calcined at 900 °C is a strong candidate for being applied as a gas detector.