NO₂ Detection Using Hierarchical WO₃ Microflower-Based Gas Sensors: Comprehensive Study of Sensor Performance

Abstract

Monitoring nitrogen dioxide (NO₂) in various scenarios is crucial due to its significant environmental impact as a hazardous gas which is emitted by several industrial sectors. This study reports the optimized synthesis of WO₃ flower-like structures using the microwave-assisted hydrothermal method under various experimental conditions, resulting in the optimized sample designated MF-WO₃-K2. Structural, morphological, and chemical characterizations revealed that WO₃ microflowers (MF-WO₃-K2) exhibit a hexagonal crystalline phase, a bandgap of 2.4 eV, and a high specific surface area of 61 m²/g. The gas-sensing performance of WO₃ microflowers was investigated by electrical measurements of six similarly fabricated MF-WO₃-K2 sensors. The MF-WO₃-K2 sensors demonstrated a remarkable sensor signal of 225 for 5 ppm NO₂ at 150 °C and response/recovery times of 14.5/2.4 min, coupled with outstanding selectivity against potential interfering gases such as CO, H₂, C₂H₂, and C₂H₄. Additionally, the sensors achieved a low detection limit of 65 ppb for NO₂ at 150 °C. The exceptional sensing properties of WO₃ microflowers are attributed to the abundance of active sites on the surface, large specific surface area, and the presence of pores in the material that facilitate the diffusion of NO₂ molecules into the structure. Overall, the WO₃ microflowers demonstrate a promising ability to be used as a sensitive layer in high-performance chemiresistive gas sensors due to their high sensor performance and good reproducibility for NO₂ detection.

1. Introduction

Industrial and technological progress has promoted rapid social advancements across numerous areas. However, the gas emissions from industrial waste have simultaneously led to environmental pollution and degraded air quality. Among these gases, NO₂ is a polluting and toxic compound that can contribute to the formation of photochemical smog and acid rain when it reacts with atmospheric water vapor. Additionally, NO₂ can adversely affect human health, even at parts per million (ppm) levels, impacting the respiratory and nervous systems. The Conference on Industrial Hygiene established tolerable short-term and long-term exposure levels to NO₂ for humans: 5 ppm for 15 min and 3 ppm for 8 h, respectively [1]. Meanwhile, the World Health Organization (WHO) guideline set a limit of 0.1 ppm for 1 h of exposure to NO₂ [2].

Gas sensors based on semiconducting metal oxides (SMOx) have been extensively studied, and the findings reveal their great potential for application in monitoring systems for industrial processes [3] automotive emissions [4], and human health tracking [5,6]. Recent advances in NO₂ sensing have focused on both n-type (ZnO, SnO₂, In₂O₃, TiO₂, WO₃) and p-type (CuO, NiO, Co₃O₄) metal oxide semiconductors (MOS), which have demonstrated high sensitivity, fast response, and potential for room-temperature operation. In particular, n-type MOS materials such as ZnO nanowalls, SnO₂ nanotubes, mesoporous In₂O₃, and TiO₂ nanoparticles exhibit short response/recovery times, low detection limits (down to ppb levels), and enhanced performance due to large surface areas and abundant defect sites that facilitate NO₂ adsorption and charge transfer. Similarly, p-type oxides including CuO, NiO, and Co₃O₄ show promising results, mainly attributed to their hierarchical and porous morphologies, which promote efficient gas diffusion and high surface reactivity. Despite these significant advances, challenges persist in achieving high selectivity, long-term stability, and resistance to environmental cross-sensitivity. These limitations continue to drive efforts toward heterojunction engineering, surface functionalization, and composite nanostructures aimed at developing reliable, low-power NO₂ sensors suitable for practical air-quality monitoring. [7]

SnO₂ is widely recognized as a benchmark material for gas sensing applications; however, WO₃ has demonstrated superior potential as a sensing element in the active layers of SMOx sensors [1,8,9,10]. WO₃ is an alternative n-type semiconductor to SnO₂ with several advantages, including low toxicity, cost-effectiveness, scalability for miniaturization, and good performance for real-time sensing [11]. The benefits of the material include an enhanced surface-to-volume ratio, the ability to modulate band structure, and increased surface reactivity, which favors oxygen adsorption from the atmosphere and enables efficient redox reactions with target gas molecules [12]. Several WO₃-based nanomaterials, each presenting unique characteristics such as surface area, optical bandgap, and crystalline structure, have been examined for gas sensing, with particular emphasis on NO₂ detection. These materials include porous and hollow spherical architectures [13,14,15], nanowires [16], nanorods [17], nanoplates [18], and surface-functionalized arrangements [19]. Despite the extensive number of studies on WO₃-based materials for gas-sensing applications, this material remains a topic of considerable scientific interest, with many recent works still being published in this field [20,21,22,23,24,25,26,27,28,29,30,31]. Recently, we have reported the synthesis of WO₃ nanofibers via the electrospinning technique and their application as gas sensing elements. This material achieved exceptional sensor response for NO₂ detection, coupled with high selectivity against potential interferents such as H₂ and CO [10].

Hierarchical structures, composed of tiny nanocrystals as building blocks, have attracted significant attention due to their shorter gas diffusion length, higher charge carrier mobility, and relatively larger specific surface area compared to agglomerated nanoparticles. Consequently, the development of 3D hierarchical micro- and nanostructures offers promising alternative materials for manufacturing efficient gas-sensing devices. Several studies have reported the synthesis of WO₃ nanostructures using various chemical and physical methods, including solvothermal and hydrothermal techniques [32,33], chemical bath deposition [34], chemical vapor deposition [35], RF sputtering [36], and the colloidal chemical method [37]. The microwave-assisted hydrothermal method (MAHM) has advantages over the conventional hydrothermal approach, such as higher yield, shorter synthesis time, reduced solvent usage [38,39], and improved control over morphology and crystalline structure [40]. The sensor performance of WO₃ can be enhanced by adjusting the synthesis parameters to refine its morphology, resulting in an effective response to target gases [41]. In particular, hierarchical flower-like structures prepared via solvothermal and hydrothermal methods have shown promising results in detecting multiple gases, including NO₂ [1,32,42,43,44,45,46].

Although numerous studies have reported the synthesis of WO₃ morphologies via hydrothermal routes, the microwave-assisted hydrothermal method (MAHM) enables faster synthesis with tunable morphological and structural properties, resulting in enhanced gas-sensing performance. Most previous reports have focused either on synthesis or on sensor application; in contrast, our study integrates both aspects, offering a comprehensive understanding that may guide future developments. Furthermore, the use of self-heating devices provides a promising strategy for compact and integrable gas sensor systems, in contrast to conventional externally heated configurations.

This work presents the gas sensing performance of WO₃ flower-like microstructures synthesized via the microwave-assisted hydrothermal method. The morphological and crystalline characteristics of the microflowers, prepared under several synthesis parameters, were investigated using appropriate techniques. The gas sensing performance was evaluated for detecting several target gases, such as NO₂, CO, H₂, C₂H₂, and C₂H₄, at different operating temperatures (ranging from 100 to 300 °C) using a self-heating system. Key performance metrics, including sensor signal, response/recovery time, selectivity, limit of detection, and reproducibility, were determined. Based on the results, a conduction mechanism related to the reception and transduction phenomena occurring on the surface of the hierarchical WO₃ structures was proposed.

2. Materials and Methods

2.1. Synthesis of WO₃ Microflowers by the Microwave-Assisted Hydrothermal Method

The synthesis of WO₃ microflowers was carried out in five steps: (I) 1.0 g of tungstic acid (H₂WO₄, 99% purity, Sigma Aldrich, St. Louis, MO, USA) was dissolved in 10 mL of hydrogen peroxide (H₂O₂, 30%) for 30 min using an ultrasonic bath; (II) 2.0 g of potassium sulfate (K₂SO₄, 99% purity, Synth, Rio Verde, Brazil) was dissolved in 40 mL of deionized water under vigorous stirring until a homogeneous solution was obtained; (III) the solutions from steps (I) and (II) were mixed, followed by the addition of 20 mL of a 2M hydrochloric acid solution (HCl); (IV) the resulting solution from step (III) was placed in a Teflon reactor, and the synthesis was conducted using a modified microwave system (Panasonic, Kadoma-shi, Japan; model Piccolo Style NN-ST341WRUK, 850 Watts) at 150 °C for 30 min with a heating rate of 10 °C/min; (V) the as-synthesized material was washed with deionized water and centrifuged at 10,000 rpm, followed by drying at 80 °C for 12 h. Subsequently, a heat treatment at 400 °C in air was performed with heating and cooling rates of 10 °C/min. The resulting material was named MF-WO₃-K2. The influence of the synthesis parameters on the WO₃ materials was studied by varying the amount of K₂SO₄, HCl concentration, and synthesis time. The experimental parameters were as follows:

(a)

Different amounts of K₂SO₄ (0 g, 1 g, 2 g, 3 g, and 4 g; referred to as MF-WO₃-K0, MF-WO₃-K1, MF-WO₃-K2, MF-WO₃-K3, and MF-WO₃-K4, respectively) were used. The HCl concentration was 3M, and the synthesis time was 30 min.

(b)

Different HCl concentrations (0 M, 1 M, 2 M, 3 M, and 4 M; referred to as MF-WO₃-H0, MF-WO₃-H1, MF-WO₃-H2, MF-WO₃-H3, and MF-WO₃-H4, respectively) were tested. The amount of K₂SO₄ used was 2 g, and the synthesis time was 30 min.

(c)

Different synthesis times (15 min, 30 min, 60 min, 90 min, and 120 min; referred to as MF-WO₃-T15, MF-WO₃-T30, MF-WO₃-T60, MF-WO₃-T90, and MF-WO₃-T120, respectively) were examined. The amount of K₂SO₄ used was 2 g, and the HCl concentration was 3 M.

2.2. Structural, Morphological, and Thermal Characterization of the WO₃ Microflowers

The crystallinity of the synthesized materials was analyzed using X-ray diffraction (XRD; Rigaku, Akishima-shi, Japan, Rint 2000) with Cu Kα radiation and Raman spectroscopy (Horiba Jobin Yvon, Villeneuve d’Ascq, France, model Lab Ram HR) with a 632 nm excitation source. The optical bandgap was estimated by ultraviolet-visible spectroscopy (UV-Vis; Lambda 1050, PerkinElmer, Waltham, MA, USA). The specific surface area and pore size distribution were determined through the N₂ adsorption/desorption experiments using the Brunauer-Emmett-Teller (BET) method (Micromeritics instrument, Norcross, GA, USA, model ASAP 2000). Thermogravimetric analysis (TG; TA Instruments, New Castle, DE, USA, model SDT 2960) was performed to monitor mass changes in the as-synthesized material over a temperature range of 25 °C to 500 °C, with a heating rate of 10 °C/min. Morphological and structural characterizations were conducted using field-emission scanning electron microscopy (FE-SEM; JEOL, Akshima, Japan, model JSM-7500F), dual-beam microscopy (FIB-SEM; FEI, Eindhoven, The Netherlands, model Helios NanoLab 600i), and high-resolution transmission electron microscopy (HRTEM, Philips, Eindhoven, The Netherlands, model CM200).

2.3. Methods for Gas Sensor Measurements

The sensor devices were fabricated by dropping a solution containing WO₃ microflowers dispersed in isopropyl alcohol (30 mg/mL) onto alumina substrates coated with interdigitated platinum electrode arrays (200 μm Pt fingers spaced 150 μm apart). 200 μL of the solution was used, distributed in 10 μL aliquots. On the opposite side of the substrate, a platinum circuit was used to heat the sensors within an operating temperature range of 100 °C to 300 °C. Gas sensing measurements were conducted in a stainless-steel chamber by monitoring the electrical resistance of the materials using a multichannel system with a data acquisition unit (Agilent, Santa Rosa, CA, USA; model 34972A). Additionally, synthetic dry air was used as the baseline gas, the analyte gases (NO₂, H₂, CO, C₂H₂, and C₂H₄) were diluted in synthetic dry air to a concentration of 100 ppm, and the gases were introduced into the chamber using mass-flow controllers (MFC; MKS, Andover, MA, USA, model GV50A). The sensors were initially equilibrated in dry air for 12 h, followed by exposures to several concentrations (ranging from 2 to 100 ppm) of NO₂, H₂, CO, C₂H₂, and C₂H₄ in 20 min cycles. After each exposure, 60 min dry air cycles were performed, maintaining a constant total gas flow of 100 sccm throughout all the experiments. The sensor signals were calculated using the following equations:

$${\mathrm{S}}_{\mathrm{o}\mathrm{x}\mathrm{i}}={\displaystyle \frac{{\mathrm{R}}_{\mathrm{g}\mathrm{a}\mathrm{s}}}{{\mathrm{R}}_{\mathrm{a}\mathrm{i}\mathrm{r}}}}$$

(1)

$${\mathrm{S}}_{\mathrm{r}\mathrm{e}\mathrm{d}}={\displaystyle \frac{{\mathrm{R}}_{\mathrm{a}\mathrm{i}\mathrm{r}}}{{\mathrm{R}}_{\mathrm{g}\mathrm{a}\mathrm{s}}}},$$

(2)

where ${\mathrm{S}}_{\mathrm{o}\mathrm{x}\mathrm{i}}$ and ${\mathrm{S}}_{\mathrm{r}\mathrm{e}\mathrm{d}}$ represent the sensor signals for oxidizing and reducing gases, respectively. ${\mathrm{R}}_{\mathrm{g}\mathrm{a}\mathrm{s}}$ and ${\mathrm{R}}_{\mathrm{a}\mathrm{i}\mathrm{r}}$ correspond to the electrical resistance of the materials in the presence of the analytes and dry air, respectively. In addition, the response time and recovery time were calculated. The response time corresponds to the time required to reach 90% of the maximum resistance in the presence of the analyte gas. The recovery time is defined as the time needed to reach 90% of the resistance in dry air after the analyte’s removal [47].

3. Results and Discussion

3.1. Structural, Morphological, Chemical, and Thermal Characterization of the WO₃ Microflowers

Figures S1–S3 show the SEM images of multiple samples synthesized by varying the synthesis parameters. Figure S1 presents hierarchical flower-like structures assembled from nanostructures on their surface, which were obtained using different amounts of K₂SO₄ (0 g, 1 g, 2 g, 3 g, and 4 g), a 3M HCl concentration, and a synthesis time of 30 min. The SEM images revealed a progressive increase in the size of microflowers along with a simultaneous refinement of the nanostructures on the surface with the incremental addition of K₂SO₄. The MF-WO₃-K1 sample consists of microflowers with smaller diameters and nanorods on their surface. In contrast, sample MF-WO₃-K2 contains larger microflowers with nanostructured plates. The MF-WO₃-K3 sample shows greater and agglomerated microflowers resembling urchin-like nanostructures. The MF-WO₃-K4 sample results in clusters with irregular shapes composed of nanostructures of varying thickness. Figure S2 illustrates the effect of changing HCl concentrations (0 M to 4 M) on samples synthesized with 2 g of K₂SO₄ and a 30 min synthesis time. Clusters of microflowers were obtained for HCl concentrations between 0 M and 2 M, showing improved shape patterns as the HCl concentration increased. Microflowers with well-defined spherical shapes were produced with 3 M HCl, whereas irregular nanostructures were observed at 4 M HCl. Figure S3 shows SEM images of materials synthesized at different synthesis times (15, 30, 60, 90, and 120 min) using 2 g of K₂SO₄ and a 3 M HCl concentration. Clusters of microstructures with thin nanostructures were observed after 15 min of synthesis. However, well-defined microflowers formed at 30 and 60 min. Longer synthesis times (90 and 120 min) led to the formation of different flower morphologies and the breakdown of the microflower structure into dispersed nanostructures.

To understand the formation of the observed structures, it is essential to consider the role of the precursor, tungstic acid (H₂WO₄). In aqueous solution, H₂WO₄ can generate several polytungstate ions, including the following:

$${\mathrm{H}}^{+}+{\mathrm{W}\mathrm{O}}_4^{2-}={\mathrm{H}\mathrm{W}\mathrm{O}}_4^{-},$$

(R1)

$$7{\mathrm{H}}^{+}+6{\mathrm{W}\mathrm{O}}_4^{2-}=\mathrm{H}{\mathrm{W}}_6{\mathrm{O}}_{21}^{5-},$$

(R2)

For pH > 7, the ${\mathrm{W}\mathrm{O}}_4^{2-}$ species predominates, whereas for pH < 7, $\mathrm{H}{\mathrm{W}}_6{\mathrm{O}}_{21}^{5-}$ becomes the dominant species. The latter can dissociate into ${\mathrm{W}\mathrm{O}}_4^{2-}$, maintaining reaction equilibrium [48,49]. According to Bai et al. [48], the controlled addition of an inorganic acid, such as 3M HCl, to an aqueous solution containing ${\mathrm{W}\mathrm{O}}_4^{2-}$ gradually induces gelation of these ions and the subsequent formation of WO₃ nanoparticles during the hydrothermal reaction, which can lead to the formation of unique structures depending on the synthesis conditions.

The SEM analysis (Figures S1–S3) revealed the influence of synthesis parameters on the characteristics of the obtained microflowers. Previous studies have reported that ions act as selective adsorption agents on specific crystalline faces, favoring growth in particular directions [42,50,51]. Increasing the HCl concentration decreases the size of the microflowers, highlighting the relationship between the microflowers nucleation and HCl concentration. Additionally, the acid concentration may affect the thickness of plate-like nanostructures on the microflowers surface. Regarding the synthesis time, SEM images indicate that longer reaction times result in smaller microflowers. Conversely, particle aggregation increases, forming thicker plates, likely due to the dissociation and recrystallization of the structures [52]. Based on the SEM images, the MF-WO₃-K2 sample was considered the optimized material, presenting a well-defined microflower structure with nanoplates on its surface.

Figure 1A shows a SEM image of the as-synthesized microflowers (MF-WO₃-K2), exhibiting a spherical shape with diameters ranging from 1 μm to 3 μm. A zoom-in image (Figure 1B) reveals nanometric plates on the surface of these structures. After heat treatment at 400 °C, the microflowers maintained their morphology, as indicated in Figure 1C,D. Figure S4 presents the energy-dispersive X-ray spectra (EDS) with peaks corresponding only to oxygen and tungsten, which indicates the absence of impurities in the MF-WO₃-K2 sample. SEM images were used to determine the distributions of microflower diameters and nanoplate thicknesses. Gaussian curves fitted to these distributions yielded diameter and thickness values of 1.78 μm and 18.2 nm, respectively (Figure 1G,H). The cross-section of a single microflower (MF-WO₃-K2) was studied, revealing a predominantly solid core with a surface covered by plates (Figure 1E). Tiny pores were also observed near the surface, possibly due to the spaces between the plates. Figure 1F displays a high-resolution TEM image of plates on the surface of the WO₃ microflowers. The image reveals a high crystalline material, with plates exhibiting an interplanar distance of 3.2 Å, corresponding to the (200) growth direction. Additionally, the plane (200) is associated with the concentric ring observed in the selected-area electron diffraction (SAED) pattern, shown in the inset of Figure 1F.

XRD analyses (Figure 2A) were conducted to investigate the crystalline phase of the synthesized material. The results revealed that the as-synthesized microflowers consisted of hydrated tungsten oxide composition ($\mathrm{W}{\mathrm{O}}_3•0.33{\mathrm{H}}_2\mathrm{O}$), which crystallized in an orthorhombic structure (JCPDS card nº 87-1203). However, heat treatment at 400 °C caused water removal from the structure, leading to a phase transition to hexagonal tungsten trioxide (WO₃) (JCPDS card nº 85-2459). This transition was confirmed by the disappearance of the peak at 18° in the diffractogram [11]. The TG analysis (Figure 2B) indicates the water removal, showing a mass loss of 4.1% between 28 °C and 360 °C, which is attributed to two processes: (I) the removal of adsorbed H₂O molecules from room temperature to 210 °C [11], and (II) the elimination of H₂O molecules from the oxide structure between 210 °C and 360 °C, which occurs concurrently with the crystallization process. The condensation of adjacent $\mathrm{W}-\mathrm{O}\mathrm{H}$ groups led to the formation of water molecules and $\mathrm{W}-\mathrm{O}-\mathrm{W}$ bonds, resulting in the hexagonal phase transition observed in the XRD analysis [11]. Additionally, a 2.5% mass loss related to the crystalline phase change was estimated based on the chemicals used in the synthesis. No further mass loss was observed between 360 °C and 490 °C.

Figure 3A shows the Raman spectra of the MF-WO₃-K2, exhibiting the bands associated with the stretching vibrations of tungsten-oxygen bonds. The bands at 251 cm⁻¹ and 323 cm⁻¹ correspond to the $\mathrm{v}\left(\mathrm{W}-\mathrm{O}-\mathrm{W}\right)$ vibrations [53], while those at 689 cm⁻¹ and 781 cm⁻¹ are related to $\mathrm{v}\left(\mathrm{O}-\mathrm{W}-\mathrm{O}\right)$ vibrations [53]. Additionally, the band at 940 cm⁻¹ is assigned to $\mathrm{v}\left({\mathrm{W}}^{6+}=\mathrm{O}\right)$ [54]. The bands below 200 cm⁻¹ are typically attributed to crystal lattice vibrations [55]. Furthermore, diffuse reflectance measurements were performed using the Kubelka-Munk function, and a bandgap of 2.4 eV was estimated (Figure 3B). From BET (Figure 3C) and BJH adsorption methods (Figure 3D), the specific surface area and mean pore size were calculated to be 61 m²/g and 43 nm, respectively.

3.2. Gas Sensing Measurements

Figure 4A shows the resistance of the MF-WO₃-K2 sensor over time during exposure to NO₂ concentrations ranging from 2 ppm to 100 ppm at operating temperatures between 100 °C and 300 °C. The baseline resistance in dry air decreased as the temperature increased. This phenomenon arises from two competing processes: (i) exponential thermal excitation of electrons, which promotes them to the semiconductor’s conduction band as temperature increases, and (ii) oxygen ionosorption under heating, leading to the formation of ionic species on the oxide surface, thereby reducing conductivity [9,56]. Under NO₂ exposure, the sensor resistance increased with higher gas concentrations. This behavior is typical of n-type semiconductors and results from electron trapping in the conduction band due to interactions between NO₂ and the WO₃ surface [57]. The sensor signals were calculated from the resistance data using Equation (1), and the calibration curves are shown in Figure 4B. The sensor signals increased with rising NO₂ concentration at all operating temperatures, with the highest responses obtained at 100 °C and 150 °C. The decrease in sensor response at elevated operating temperatures is attributed to the increased desorption rate and reduced diffusion depth of NO₂ on the WO₃ surface. At 100 °C and 150 °C, the sensor showed similar sensor signals below 50 ppm (approximately 62, 250, 500, and 1200 for 2, 5, 10, and 25 ppm of NO₂, respectively). However, at 50 ppm and 100 ppm, the sensor response was higher at 100 °C than at 150 °C. Specifically, for NO₂ concentrations of 50 and 100 ppm, the sensor signals were 3225/6250 and 1711/2400 at 100 °C and 150 °C, respectively.

The high sensor response achieved in the range of NO₂ concentrations studied can be attributed to several factors: (I) the structure formed by well-aligned nanoplates, which offer abundant active sites on the surface. Furthermore, the spaces on the surface due to the loosely connected nanoplates facilitate the diffusion of NO₂ molecules into the microstructure, improving the sensor response [58]; (II) the ratio between the electron-depleted volume and the total volume directly affects the sensor response, especially for thin nanoplates [58]; and (III) the high crystallinity of the material, which ensures favorable conditions for electronic transport by minimizing electron traps such as crystal defects, enabling more efficient electron transfer [57].

For further analysis, the gas sensing measurements were performed at 150 °C. The MF-WO₃-K2 device exhibited similar sensor signals at 100 °C and 150 °C; however, at 150 °C, smaller response and recovery times were obtained. The selectivity of the MF-WO₃-K2 device towards potential interferent gases was also evaluated. In Figure 5A, the sensor resistance at 150 °C is plotted over time during exposures to reducing gases such as CO, H₂, C₂H₂, and C₂H₄, with concentrations ranging from 2 to 100 ppm. The sensor signals for 5 ppm of reducing gases and NO₂ were comparatively analyzed, and the results are presented in Figure 5B. The sensor signal for NO₂ (${\mathrm{S}\mathrm{S}}_{{\mathrm{N}\mathrm{O}}_2}$) is over 200-fold higher than those for the reducing gases (${\mathrm{S}\mathrm{S}}_{{\mathrm{N}\mathrm{O}}_2}/{\mathrm{S}\mathrm{S}}_{{\mathrm{H}}_2}=223$; ${\mathrm{S}\mathrm{S}}_{{\mathrm{N}\mathrm{O}}_2}/{\mathrm{S}\mathrm{S}}_{{\mathrm{H}}_2}=221$; ${\mathrm{S}\mathrm{S}}_{{\mathrm{N}\mathrm{O}}_2}/{\mathrm{S}\mathrm{S}}_{{\mathrm{C}}_2{\mathrm{H}}_2}=208$; and ${\mathrm{S}\mathrm{S}}_{{\mathrm{N}\mathrm{O}}_2}/{\mathrm{S}\mathrm{S}}_{{\mathrm{C}}_2{\mathrm{H}}_4}=211$). These findings demonstrate the exceptional selectivity of the MF-WO₃-K2 device for detecting NO₂ in the presence of potential interfering gases. The sensor responses for reducing gases can be improved by both increasing the partial pressure of the reducing gases and the oxygen gases [59]. Furthermore, semiconductor materials such as WO₃, when exposed to humidity, typically exhibit a decrease in the sensor’s baseline resistance, which in turn reduces the response toward oxidizing gases such as NO₂. In this process, water molecules are adsorbed onto the WO₃ surface and release electrons back into the conduction band [60,61].

To examine the reproducibility of the results under optimized operating conditions, six MF-WO₃-K2 sensors were analyzed at 150 °C for NO₂ concentrations ranging from 2 to 100 ppm. The sensors exhibited a consistent resistance profile in response to varying NO₂ concentrations (Figure 6A). The average sensor signals (S) versus NO₂ concentrations on a linear scale (Figure 6B) followed the power-law function [47]:

$$\mathrm{S}=\mathrm{a}·{\mathrm{c}}_{{\mathrm{N}\mathrm{O}}_2}^{\mathrm{b}}$$

(3)

where $\mathrm{S}$ is the sensor signal, $\mathrm{a}$ and $\mathrm{b}$ are constants, and ${\mathrm{c}}_{{\mathrm{N}\mathrm{O}}_2}$ is the analyte concentration.

Table 1. Sensor signal with the standard deviation of the MF-WO₃-K2 sensor at 150 °C for NO₂ concentrations ranging from 2 to 100 ppm.

NO2 Concentration (ppm)	Sensor Signal (${\mathbf{R}}_{{\mathbf{N}\mathbf{O}}_2}/{\mathbf{R}}_{\mathbf{g}\mathbf{a}\mathbf{s}}$) (SD)
2	59 (7)
5	225 (44)
10	464 (117)
25	927 (231)
50	1560 (117)
100	2372 (97)

summarizes the average sensor signals, along with their respective standard deviations. The interaction between the gas and the semiconductor, which follows a power-law relationship, can be readily identified from the log–log plot of the sensor signal versus gas concentration from the linearization of Equation (4) [62,63]. In this case, Figure 6C shows a log-log plot for log (S) versus ${\mathrm{c}}_{{\mathrm{N}\mathrm{O}}_2},$ and the calibration curve with an R-square value of 0.98. The calibration curve function corresponds to the following equation:

$$\log \left(\mathrm{S}\right)=\log \left(\mathrm{a}\right)+\mathrm{b}·\log \left({\mathrm{c}}_{{\mathrm{N}\mathrm{O}}_2}\right)$$

(4)

The values of $\mathrm{a}$ and $\mathrm{b}$ are 39.21 and 0.90, respectively.

Figure 7A shows curves of stability tests at 150 °C for 25 ppm of NO₂. For the tests, the sensors were stabilized under 120 min of dry air and 20 min of NO₂ exposure, maintaining a constant total gas flow of 100 sccm. The MF-WO₃-K2 sensor exhibited an average sensor signal of 1014 (192), which is similar to the sensor signal value presented in Table 1. The average response and recovery times for the six MF-WO₃-K2 sensors at 150 °C were calculated from the sensor response curves while varying NO₂ concentrations from 2 to 100 ppm (Figure S5), and the results are shown in Figure 7 and summarized in

Table 2. Average response and recovery times, including standard deviations, of the MF-WO₃-K2 sensor at 150 °C for NO₂ concentrations ranging from 2 to 100 ppm.

NO2 Concentration (ppm)	Average Response Time (min) (SD)	Average Recovery Time (min) (SD)
2	14.4 (0.9)	2.3 (0.6)
5	14.5 (0.5)	2.4 (0.5)
10	11.6 (1.9)	3.0 (0.3)
25	7.6 (2.5)	3.7 (1.1)
50	7.5 (1)	4.4 (1.3)
100	7.4 (1.7)	9.1 (1.5)

. The analysis revealed that the response times decreased with increasing NO₂ concentration, changing from 14.4 min (for 2 ppm) to 7.4 min (for 100 ppm). In contrast, recovery times increased from 2.3 min (2 ppm) to 9.1 min (100 ppm). The decrease in response times and increase in recovery times may be related to enhanced sensor responses at higher NO₂ concentrations, indicating a faster depletion of charge carriers of microflowers and slower desorption of NO₂ molecules from the surface. The response and recovery times can be decreased by increasing the gas flow rate in the test chamber or by decreasing the thickness of the sensing layer through deposition of a smaller amount of WO₃ microflowers. Table 2 summarizes the obtained response and recovery times.

The limit of detection (LOD) for the MF-WO₃-K2 sensor was estimated from the resistance curve [47] and represents the lowest NO₂ concentration that the sensor can detect. Figure 8A shows the average resistance of six MF-WO₃ sensors as a function of NO₂ concentrations at 150 °C on a linear axis, which follows a power-law relationship (Equation (4)). The corresponding log-log plot (Figure 8B) shows a linear relationship (${\mathrm{R}}^2=0.99$) described by the following equation:

$$\mathrm{l}\mathrm{o}\mathrm{g}\left(\mathrm{R}\right)=\mathrm{l}\mathrm{o}\mathrm{g}\left(\mathrm{a}\right)+\mathrm{b}·\log \left({\mathrm{c}}_{{\mathrm{N}\mathrm{O}}_2}\right)$$

(5)

where $\mathrm{R}$ is the resistance of sensor response, $\mathrm{a}$ and $\mathrm{b}$ are constants, and ${\mathrm{c}}_{{\mathrm{N}\mathrm{O}}_2}$ is the analyte concentration.

The LOD was calculated using Equation (5) by replacing $\mathrm{R}$ with ${\overline{\mathrm{R}}}_0+\left(3·{\mathsf{\sigma }}_0\right)$, resulting in Equation (6) [47]. Here, ${\overline{\mathrm{R}}}_0$ corresponds to the mean baseline resistance for the six MF-WO₃ sensors, and ${\mathsf{\sigma }}_0$ is the standard deviation of ${\overline{\mathrm{R}}}_0$ (see

Table 3. Constant values for LOD calculation at 150 °C.

Constants	$\mathit{a}$	$\mathit{b}$	${\overline{\mathit{R}}}_0$	${\mathit{\sigma }}_0$
Values	$3.84·{10}^7$	$0.94$	$1.25·{10}^6$	$5.6·{10}^5$

). Consequently, the calculated LOD of the MF-WO₃-K2 sensor reached 65 ppb of NO₂ at 150 °C, which is much lower than the values reported by the Conference on Industrial Hygiene (3 ppm) [1] and the WHO (0.1 ppm) [2].

$${\mathrm{f}}^{-1}\left(\mathrm{R}\right)=\mathrm{l}\mathrm{o}\mathrm{g}\left({\mathrm{c}}_{{\mathrm{N}\mathrm{O}}_2}\right)={\displaystyle \frac{\log \left({\overline{\mathrm{R}}}_0+3·{\mathsf{\sigma }}_0\right)-\log \left(\mathrm{a}\right)}{\mathrm{b}}}$$

(6)

Table 4. Summary of the performance of the MF-WO₃-K2 sensor and a comparison with literature reports on WO₃-based sensors for NO₂ detection.

WO3 Material	Operating Temperature (°C)	NO2 Concentration (ppm)	Sensor Signal	Response/ Recovery Times (s)	LOD	Reference
MF-WO3-K2 sensor	150	5	225 * 226 ** 22,578% ***	870/144	65 ppb	This work
WO3 Nanofibers	150	10	220 *	-	0.5 ppm	[10]
WO3 Microflowers	150	100	55.5 *	9/225	1 ppm	[64]
WO3 nanosheets	100	10	460 *	54/63	-	[9]
WO3 Layer-like	160	5	106 **	32/13	-	[65]
GaSe	150	2	83% ***		80 ppb	[66]
NiO	200	5	4.2% ***	6/78		[67]
In2O3/SnS2	180	5	15.61 *	34/65	25 ppb	[68]
CuO	200	10	30% ***		1 ppm	[69]
NiMoO4/NiO	160	1	16.18 *	32/717	4.7 ppb	[70]
WO3 S: WO3	150	5	5.6 * 578 *	186/>3000 185/130	- 54	[71]
MoS2 nanopetal	100	100	61.78% ***	22/225	1 ppm	[72]
PTh/MWCNT/SnO2	150	100	76.8% ***	23/75	-	[73]
WO3-rGO heterostructure	100	5	77 **	24/457	-	[74]
Pd-WO3-Ov	160	10	106.44 *	37.2/28.5	51 ppb	[75]
ZnO nanorods	250	50	8.8 *	28/20	-	[76]

* $({\mathrm{R}}_{{\mathrm{N}\mathrm{O}}_2}/{\mathrm{R}}_{\mathrm{a}\mathrm{i}\mathrm{r}})$, ** $({\mathrm{R}}_{{\mathrm{N}\mathrm{O}}_2}-{\mathrm{R}}_{\mathrm{a}\mathrm{i}\mathrm{r}})/{\mathrm{R}}_{\mathrm{a}\mathrm{i}\mathrm{r}}$, *** $100\%·({\mathrm{R}}_{{\mathrm{N}\mathrm{O}}_2}-{\mathrm{R}}_{\mathrm{a}\mathrm{i}\mathrm{r}})/{\mathrm{R}}_{\mathrm{a}\mathrm{i}\mathrm{r}}$.

summarizes the sensor performance of the MF-WO₃-K2 device for a NO₂ concentration of 5 ppm at 150 °C and compares it with current WO₃ sensors with various morphologies and other SMOx presented in the literature. The comparison reveals that the MF-WO₃-K2 device exhibited the highest sensor signal and a lower limit of detection than most of the sensors. Additionally, the response and recovery times of the MF-WO₃-K2 device were higher than those reported in other studies, which can be attributed to its high sensor response. Thus, the MF-WO₃-K2 device exhibited excellent sensor performance.

3.3. Gas Sensing Mechanisms

Tungsten trioxide exposed to a synthetic air environment without humidity undergoes oxidation due to the ionosorption of oxygen on its surface, which can occur predominantly through molecular oxygen species (${\mathrm{O}}_2^{-}$, at 150 °C) [77,78,79,80]. Thus, at the operating temperature of 150 °C, the oxygen adsorption in molecular form on the WO₃ surface can be expressed in equilibrium conditions, such as:

$${\mathrm{O}}_{2\left(\mathrm{g}\right)}+{\mathrm{V}}_0^{••}+{\mathrm{e}}_{\mathrm{C}\mathrm{B}}^{-}={\left({\mathrm{O}}_2^{-}-{\mathrm{V}}_0^{••}\right)}_{\mathrm{a}\mathrm{d}\mathrm{s}\left({\mathrm{W}\mathrm{O}}_3\right)},$$

(R3)

where ${\mathrm{e}}_{\mathrm{C}\mathrm{B}}^{-}$ represents the electrons in the conduction band, ${\mathrm{V}}_0^{••}$ is a double-charged positive oxygen vacancy, and ${({\mathrm{O}}_2^{-}-{\mathrm{V}}_0^{••})}_{\mathrm{a}\mathrm{d}\mathrm{s}\left({\mathrm{W}\mathrm{O}}_3\right)}$ is the oxygen chemisorbed coupled to ${\mathrm{V}}_0^{••}$ [81]. The oxygen adsorption at the WO₃ grain boundaries generates potential barriers, and its change leads to variations in the barrier potential energy (${\mathrm{E}}_1$), which regulates the surface conductivity of WO₃ [9], as illustrated in Figure 9A.

At temperatures equal to or lower than 150 °C, it is unlikely for NO₂ molecules to interact with ${\mathrm{O}}_2^{-}$ [4]. Generally, NO₂ molecules interact with the metal oxide surface by reacting with oxygen vacancies [82]. Using Density Functional Theory (DFT), Han and Yin [83] reported that NO₂ molecules fill vacancy sites during adsorption at the WO-terminated surface. This indicates the presence of numerous surface vacancies [61] in the crystalline structure of WO₃ microflowers available for NO₂ interaction. Thus, direct NO₂ adsorption on available WO₃ surface sites is much more probable than its interaction with pre-adsorbed oxygen species, and the reaction can be expressed as follows:

$${\mathrm{N}\mathrm{O}}_{2\left(\mathrm{g}\right)}+{\mathrm{V}}_0^{••}+{{\mathrm{e}}^{-}}_{\left({\mathrm{W}\mathrm{O}}_3\right)}\to {({\mathrm{N}\mathrm{O}}_2^{-}-{\mathrm{V}}_0^{••})}_{\mathrm{a}\mathrm{d}\mathrm{s}\left(\mathrm{W}{\mathrm{O}}_3\right)},$$

(R4)

Additionally, Blackman et al. and Mirabella et al. proposed that changes in the ‘intrinsic’ oxygen vacancy concentration may drive the conductivity variation, in agreement with the surface conductivity model [59,84], while recent operando DRIFTS studies by Suman et al. provide direct evidence supporting the involvement of surface-adsorbed oxygen species [85]. The adsorption of NO₂ (Reaction (4)) further increases the electron depletion region, raising the material’s electrical resistance [4]. Consequently, electron conduction through the grains is modulated by NO₂ concentrations [86]. Figure 9B illustrates the modulation of the space-charge layer and potential barrier energy (${\mathrm{E}}_2$) due to the adsorption of NO₂ on the WO₃ surface, which increases its electrical resistance. Additionally, the dissociation of NO₂ species into NO molecules and oxygen adsorbed on the WO₃ surface is also possible, as shown in Reaction (5).

$${\left({\mathrm{N}\mathrm{O}}_2^{-}-{\mathrm{V}}_0^{••}\right)}_{\mathrm{a}\mathrm{d}\mathrm{s}\left({\mathrm{W}\mathrm{O}}_3\right)}\to {\mathrm{N}\mathrm{O}}_{\mathrm{g}\mathrm{a}\mathrm{s}}+{\left({\mathrm{O}}^{-}-{\mathrm{V}}_0^{••}\right)}_{\mathrm{a}\mathrm{d}\mathrm{s}\left({\mathrm{W}\mathrm{O}}_3\right)},$$

(R5)

Moreover, the formation of nitrite [87,88] and nitrate [88] species through electron trapping in the material is possible. The following reactions express these processes:

$${\mathrm{N}\mathrm{O}}_{2\left(\mathrm{g}\mathrm{a}\mathrm{s}\right)}+{\mathrm{e}}^{-}\to {\mathrm{N}\mathrm{O}}_{2\left(\mathrm{a}\mathrm{d}\mathrm{s}\right)}^{-},$$

(R6)

$${2\mathrm{N}\mathrm{O}}_{2\left(\mathrm{g}\mathrm{a}\mathrm{s}\right)}+{\displaystyle \frac{1}{2}}{\mathrm{O}}_{2\left(\mathrm{g}\mathrm{a}\mathrm{s}\right)}+2{\mathrm{e}}^{-}\to {\mathrm{N}\mathrm{O}}_{2\left(\mathrm{a}\mathrm{d}\mathrm{s}\right)}^{-}+{\mathrm{N}\mathrm{O}}_{3\left(\mathrm{a}\mathrm{d}\mathrm{s}\right)}^{-},$$

(R7)

$${\mathrm{N}\mathrm{O}}_{2\left(\mathrm{a}\mathrm{d}\mathrm{s}\right)}^{-}+{\displaystyle \frac{1}{2}}{\mathrm{O}}_{2\left(\mathrm{g}\mathrm{a}\mathrm{s}\right)}\to {\mathrm{N}\mathrm{O}}_{3\left(\mathrm{a}\mathrm{d}\mathrm{s}\right)}^{-},$$

(R8)

The adsorption of reducing gases CO, H₂, C₂H₂, and C₂H₄ on the n-type WO₃ surface at 150 °C occurs through reactions with negatively charged oxygen species (${\mathrm{O}}_2^{-}$) chemisorbed on the surface. This process releases electrons into the conduction band, thereby increasing the conductivity of WO₃ and decreasing the space-charge layer. The reactions involved for CO, H₂, C₂H₂, and C₂H₄ gases can be described as follows [80,89,90,91]:

$$2{\mathrm{C}\mathrm{O}}_{\left(\mathrm{g}\mathrm{a}\mathrm{s}\right)}+{\mathrm{O}}_{2\left(\mathrm{a}\mathrm{d}\mathrm{s}\right)}^{-}\to {2\mathrm{C}\mathrm{O}}_{2\left(\mathrm{g}\mathrm{a}\mathrm{s}\right)}+{\mathrm{e}}^{-},$$

(R9)

$${\mathrm{H}}_{2\left(\mathrm{g}\mathrm{a}\mathrm{s}\right)}+{\displaystyle \frac{1}{2}}{\mathrm{O}}_{2\left(\mathrm{a}\mathrm{d}\mathrm{s}\right)}^{-}\to {\mathrm{H}}_2{\mathrm{O}}_{\left(\mathrm{g}\mathrm{a}\mathrm{s}\right)}+{\mathrm{e}}^{-},$$

(R10)

$$2{\mathrm{C}}_2{\mathrm{H}}_{2\left(\mathrm{g}\mathrm{a}\mathrm{s}\right)}+{\mathrm{O}}_{2\left(\mathrm{a}\mathrm{d}\mathrm{s}\right)}^{-}\to 2{\mathrm{C}}_2{\mathrm{H}}_2{\mathrm{O}}_{\left(\mathrm{a}\mathrm{d}\mathrm{s}\right)}+{\mathrm{e}}^{-},$$

(R11)

$$2{\mathrm{C}}_2{\mathrm{H}}_{4\left(\mathrm{g}\mathrm{a}\mathrm{s}\right)}+{\mathrm{O}}_{2\left(\mathrm{a}\mathrm{d}\mathrm{s}\right)}^{-}\to 2{\mathrm{C}}_2{\mathrm{H}}_4{\mathrm{O}}_{\left(\mathrm{a}\mathrm{d}\mathrm{s}\right)}+{\mathrm{e}}^{-},$$

(R12)

Figure 9C illustrates the reaction between CO gas and a negatively charged oxygen species chemisorbed on the surface of WO₃, which releases electrons into the conduction band, thereby decreasing its electrical resistance. The reaction is similar for the other reducing gases, as reported in Reactions (10)–(12). A previous study reported the temperature dependence of chemisorbed oxygen species on the WO₃ surface, indicating that the density of these species decreases as the temperature decreases [80]. At temperatures ranging from 25 °C to 200 °C, the concentration of chemisorbed oxygen species shows minimal variation but increases significantly above 200 °C. Conversely, the density of the oxygen lattice increases at higher temperatures [80]. Therefore, the low density of ${\mathrm{O}}_2^{-}$ species at 150 °C may be related to the low sensor response to reducing gases, which explains the high sensor selectivity to NO₂. In situ FTIR analyses are necessary for future investigations into the detection mechanism of NO₂ and reducing gases, which may corroborate the reactions mentioned above. Specifically, the interaction described by Reaction (5) between NO₂ and WO₃ surface can be confirmed as the predominant mechanism.

4. Conclusions

Hierarchical WO₃ materials with flower-like morphology and micrometer sizes were synthesized using the microwave-assisted hydrothermal method. The material exhibited a WO₃ composition, hexagonal crystalline phase, a bandgap of 2.4 eV, and a high specific surface area of 61 m²/g. From the several systems studied, the sample MF-WO₃-K2, synthesized at 150 °C for 30 min, using HCl 3M and 2 g of K₂SO₄, demonstrated extraordinary sensor performance in detecting NO₂ at concentrations ranging from 2 to 100 ppm, with elevated selectivity against interferent gases, including CO, H₂, C₂H₂, and C₂H₄. The MF-WO₃-K2 sensor achieved a signal of 225 for 5 ppm NO₂ at 150 °C, with response/recovery times of 14.5/2.4 min, respectively. It exhibited good reproducibility and stability and a low detection limit of 65 ppb at this temperature. Long-term sensing stability tests should be conducted as additional analyses to evaluate sensing performance. These impressive results can be attributed to the unique morphology of the microflowers composed of nanoplates on the surface, which can provide abundant active surface sites, a large specific surface area, better diffusion of gas molecules into the material, a high proportion of depleted volume, and high crystallinity with fewer electron traps. Therefore, the MF-WO₃-K2 microflowers exhibited high sensor performance, making this material potentially promising for gas sensing applications. Furthermore, additional life cycle investigations and a device prototype can be developed.