Pilot-Scale Studies of WO3/S-Doped g-C3N4 Heterojunction toward Photocatalytic NOx Removal

Due to the rising concentration of toxic nitrogen oxides (NOx) in the air, effective methods of NOx removal have been extensively studied recently. In the present study, the first developed WO3/S-doped g-C3N4 nanocomposite was synthesized using a facile method to remove NOx in air efficiently. The photocatalytic tests performed in a newly designed continuous-flow photoreactor with an LED array and online monitored NO2 and NO system allowed the investigation of photocatalyst layers at the pilot scale. The WO3/S-doped-g-C3N4 nanocomposite, as well as single components, were characterized by X-ray diffraction (XRD), scanning electron microscopy (SEM), Brunauer–Emmett–Teller surface area analysis (BET), X-ray fluorescence spectroscopy (XRF), X-ray photoemission spectroscopy method (XPS), UV–vis diffuse reflectance spectroscopy (DR/UV–vis), and photoluminescence spectroscopy with charge carriers’ lifetime measurements. All materials exhibited high efficiency in photocatalytic NO2 conversion, and 100% was reached in less than 5 min of illumination under simulated solar light. The effect of process parameters in the experimental setup together with WO3/S-doped g-C3N4 photocatalysts was studied in detail. Finally, the stability of the composite was tested in five subsequent cycles of photocatalytic degradation. The WO3/S-doped g-C3N4 was stable in time and did not undergo deactivation due to the blocking of active sites on the photocatalyst’s surface.


Introduction
The development of industry and intensified agricultural activities, energy production, and transport has led to increased NO x concentration in the air and contributed to environmental degradation resulting in acid rain or fog formation. The presence of NO x in the air can give rise to various toxicological responses in humans and animals, including respiratory system damage, weakening the immune system, and genetic mutation [1][2][3]. Due to harmful effects on human health, research attention is focused on controlling and decreasing the NO x concentration in the air. One of the promising methods of nitrogen oxides removal is heterogeneous photocatalysis, in which the combination of semiconductor materials and light allows the generation of reactive oxygen species (ROS) to degrade emerging contaminants [4,5]. Among various methods of NO x decomposition, the main At the top, the chamber is closed with a plate of quartz glass, which enables exposure of the samples to the full spectrum of simulated solar light. The light spectrum of the light source is presented in Figure 1b. The light flux in the range of 310-380 nm was 5 mW·cm −2 . The air is supplied to the chamber using mass flow controllers (Bronkhorst F-201CV-10K-RGD-00-V), enabling digital flow control in the range of 0.2-10 nL/min (with an accuracy of ±0.5% of reading plus ±0.1% range). The connected chambers on both sides end with diffusers (2), ensuring even gas distribution. Gas chemical sensors (3) are installed in the diffusers, enabling the measurement of the following parameters: the 4-electrode electrochemical sensor for nitrogen oxide concentration measurement, the 4-electrode electrochemical sensor for nitrogen dioxide concentration measurement, temperature, humidity, and pressure sensors.
The degradation mechanism and effect of process parameters (humidity, gas flow rate, NO2 concentration, and amount of photocatalyst) on NOx removal efficiency were studied in detail.

Materials
Melamine (ACS reagent, 99.0-100.5%) and thiourea (ACS reagent, ≥99%) for the synthesis of S-doped-g-C3N4, sodium tungstate dihydrate (ACS reagent, ≥99%), and oxalic acid (98%) for the synthesis of WO3 were purchased from Sigma-Aldrich (Poznan, Poland). Ethanol (96%) for WO3/S-doped-g-C3N4 synthesis was purchased from POCh. All reagents were used as received without further purification. Deionized water (DI) was used in all experiments. At the top, the chamber is closed with a plate of quartz glass, which enables exposure of the samples to the full spectrum of simulated solar light. The light spectrum of the light source is presented in Figure 1b. The light flux in the range of 310-380 nm was 5 mW·cm −2 . The air is supplied to the chamber using mass flow controllers (Bronkhorst F-201CV-10K-RGD-00-V), enabling digital flow control in the range of 0.2-10 nL/min (with an accuracy of ±0.5% of reading plus ±0.1% range). The connected chambers on both sides end with diffusers (2), ensuring even gas distribution. Gas chemical sensors (3) are installed in the diffusers, enabling the measurement of the following parameters: the 4-electrode electrochemical sensor for nitrogen oxide concentration measurement, the 4-electrode electrochemical sensor for nitrogen dioxide concentration measurement, temperature, humidity, and pressure sensors.
The degradation mechanism and effect of process parameters (humidity, gas flow rate, NO 2 concentration, and amount of photocatalyst) on NO x removal efficiency were studied in detail.

Materials
Melamine (ACS reagent, 99.0-100.5%) and thiourea (ACS reagent, ≥99%) for the synthesis of S-doped-g-C 3 N 4 , sodium tungstate dihydrate (ACS reagent, ≥99%), and oxalic acid (98%) for the synthesis of WO 3 were purchased from Sigma-Aldrich (Poznan, Poland). Ethanol (96%) for WO 3 /S-doped-g-C 3 N 4 synthesis was purchased from POCh.  3 of DI water and 2M HCl with the volume ratio of 1:1 and marked as solution A. Next, 0.35 g of oxalic acid was dissolved in 90 cm 3 of DI water and marked as solution B. Solution A was added to solution B and mixed for 15 min. The mixture was transferred to a 200 cm 3 Teflon-lined stainless steel reactor, and solvothermal synthesis was performed at 90 • C for 3 h. The obtained material was rinsed three times with DI water, once with methanol, and then dried at 80 • C for 2 h. Finally, the powder was calcined at 500 • C for 1 h with a heating rate of 5 • C/min.

Synthesis of S-Doped-g-C 3 N 4
In the first step, 7.5 g of melamine and 7.5 g of thiourea were ground together in a mortar. Then, the material was calcined at 580 • C for 3 h with a heating rate of 10 • C/min.

Synthesis of WO 3 /S-Doped-g-C 3 N 4 Composite
Typically, 3 g of S-doped-g-C 3 N 4 was dispersed in a 100 cm 3 mixture of DI water and ethanol with the volume ratio of 1:1 and ultrasonicated for 1 h. Then, 3 g of WO 3 was added and ultrasonicated for another 1 h. The resulting suspension was transferred to a 200 cm 3 Teflon-lined stainless steel reactor, and solvothermal synthesis was performed at 150 • C for 6 h with a heating rate of 5 • C/min. The obtained material was rinsed a few times with DI water and dried at 80 • C for 3 h.

Material Characterization
The structure and phase composition of synthesized photocatalysts were investigated using X-ray powder diffraction (XRD). X-ray diffraction patterns were recorded on the Rigaku MiniFlex 600 X-ray diffractometer (Tokyo, Japan) with Cu KA radiation (λ~1.54Å) in the 2θ = 10-80 • range. Moreover, the chemical structure and bond identification in samples was determined by Fourier-transform infrared spectroscopy (FTIR) in the transmittance mode. The FTIR Nicolet iS10 (Thermo Fisher Scientific, Waltham, MA, USA) spectrometer was used at room temperature in the wavenumber range from 4000 to 400 cm −1 . The pellets containing 95% of potassium bromide and 5% (wt.) of a photocatalyst were analyzed in each measurement.
The morphologies of prepared powders were characterized by scanning electron microscope (SEM) with field emission gun, model Quanta FEG 250 (FEI, Hillsboro, OR, USA). The content of sulfur and WO 3 in the composite were measured using a WDXRF S8 TIGER analyzer (Bruker, S8 TIGER, Bruker, Karlsruhe, Germany) using X-ray fluorescence spectroscopy (XRF). The analyses were performed according to the standard PN-EN 15309:2010. To determine the parameters of the porous structure of the single components and the composite, including Brunauer-Emmett-Teller (BET) surface area and pore volume using low-temperature nitrogen sorption, a Micromeritics Gemini V apparatus (model 2365; Norcross, GA, USA) was used. Before each measurement, the analyzed photocatalysts were degassed at 200 • C for 2 h. The specific surface areas were determined by the multipoint BET method in the p/p 0 range from 0.05 to 0.30. Total pore volumes were estimated from the nitrogen adsorbed at p/p 0 = 0.995.
The detailed chemical composition of the samples and nitrates identification were analyzed by the X-ray photoemission spectroscopy method (XPS). XPS measurements were carried out at room temperature under ultrahigh vacuum conditions with pressures below 1.1 × 10 −6 Pa using Omicron NanoScience equipment with an Argus hemispherical analyzer (ScientaOmicron, Uppsala, Sweden). Spectra deconvolution was performed with the CASA XPS software (Casa Software Ltd., ver. 2.3.23., Devon, UK) package using the Shirley background subtraction and Gauss-Lorentz curve fitting algorithm by the least-squares method-GL (30). The XPS spectra were calibrated to obtain binding energy of 285.00 eV for the C 1 s peak.
The optical properties, including absorbance and Kubelka-Munk functions, were studied by a UV-vis spectrophotometer (Thermo Fisher Scientific Evolution 220, Waltham, MA, USA) to measure diffuse reflectance (DRS). During measurements, barium sulfate was used as a standard. Based on the obtained spectra, the bandgap energy calculations were performed by Tauc's method.
The photoluminescence spectra were recorded using an FLS980 fluorescence spectrophotometer from Edinburg Instruments (Livingston, Wielka Brytania) equipped with a 450 W xenon lamp as an excitation source and a Hamamatsu 928 PMT detector. Measured PL spectra were corrected for the sensitivity and wavelength of the experimental setup. The slit width was 2 µm (excitation spectra) or 0.3 µm (emission spectra). The same equipment was used for decay measurements to calculate the charge carrier lifetime, and the lamp was changed to a 150 W pulse xenon lamp. All measurements were recorded at room temperature. The average experimental lifetimes of obtained photocatalysts were calculated according to Hu et al. [23].

Coating Methodology
Firstly, 2 g of the photocatalyst was mixed with deionized water (30 cm 3 ) in Falcon tubes (50 cm 3 ) using the vortex mixer. The homogenous sample was transferred onto glass (30.5 cm × 30.5 cm) and evenly distributed on the whole surface. Finally, the coatings were dried at room temperature for 24 h.

Photocatalytic Setup
The photocatalytic experiments were carried out in a flat quartz glass photoreactor in a volume of 1.5 dm 3 with continuous NO 2 flow. The photoreactor was used for NO 2 photocatalytic conversion over WO 3 , S-doped g-C 3 N 4, and WO 3 /S-doped g-C 3 N 4 , as depicted in Figure 1a. The plate with the photocatalyst is parallel to an air-tight optical window for photoirradiation. The gas with pollution passes only through the space between the test plate and the window. The flow rates of the reactant gases (100 ppm NO 2 in synthetic air, Linde Gas) and the pretreatment gas (synthetic air, Linde Gas) were adjusted with gas flow meters. The relative humidity (RH) level of the NO 2 flow was controlled at 40% by passing the zero air stream through a humidification chamber. The air stream, NO 2 , and humidity flow rates were controlled at 0.9 dm 3 /min, 0.1 dm 3 /min, and 0.3 dm 3 /min. The concentration of NO 2 was detected continuously by employing a NO x analyzer (AlphaSense, NO-B4 4-electrode nitric oxide sensor, Alphasense 000-0NO2-B43F nitric dioxide sensor). A LED panel light was used in the photocatalytic reaction as a light source. The light flux in the range of 310-380 nm was on average 5 mW·cm −2 .
Before each measurement, the test gas was allowed to flow into the photoreactor without photoirradiation to reach adsorption-desorption equilibrium. After 30 min, the lamp was turned on to start the photocatalytic process. Finally, the photocatalytic performance of obtained photocatalysts NO 2 was demonstrated in terms of NO 2 conversion rate (Equation (1)):

Characterization of Photocatalysts
The structural parameters and phase compositions of photocatalysts were examined using X-ray diffraction. XRD patterns are presented in Figure 2. The XRD patterns of the single-phase components monoclinic WO 3 and S-g-C 3 N 4 are in good agreement between the recorded and calculated patterns. No signals originating from impurities were detected. The diffractogram of WO 3 /S-g-C 3 N 4 was similar to pure tungsten(VI) oxide patterns.
However, a significant difference is noted in the range of 2θ = 25 • -30 • (inset in Figure 2). In the presented composite, the peaks (120) and (112) originating from monoclinic WO 3 and the main peak (002) from S-g-C 3 N 4 can be distinguished. This analysis confirmed the presence of a hybrid photocatalyst of graphitic carbon nitride and WO 3 .
Materials 2022, 15, 633 6 of 17 Figure 2). In the presented composite, the peaks (120) and (112) originating from monoclinic WO3 and the main peak (002) from S-g-C3N4 can be distinguished. This analysis confirmed the presence of a hybrid photocatalyst of graphitic carbon nitride and WO3. The chemical bonding identification of obtained samples was investigated by infrared spectroscopy. The presence of characteristic bands in single components of WO3 and S-g-C3N4 confirmed the previous phase identification of WO3/S-g-C3N4 (see Figure 3). In the case of pure WO3, a broad absorption band in the wavenumber range of 420-1000 cm −1 attributed to the vibration modes of the W-O bond verifies the formation of tungsten(VI) oxide. The strong bands in 745 cm −1 and 818 cm −1 are assigned to the stretching mode of W-O-W [24]. These peaks probably overlap with a band located at 946 cm −1 , attributed to W=O stretching vibration [25,26]. In the case of S-doped graphitic carbon nitride, a narrow band at 808 cm −1 can be attributed to the triazine ring mode, which corresponds to condensed CN heterocycles [27,28]. The distinct absorption bands at 1572 and 1639 cm −1 are attributed to C=N stretching, while the four bands at 1231, 1314, 1396, and 1453 cm −1 to aromatic C-N stretching [29]. In the composite, FTIR spectra are similar to the doped carbon nitride. The difference can be noticed below 1000 cm −1 , in which triazine ring modes from S-g-C3N4 overlap with bands assigned to the stretching mode of W-O-W. Therefore, FTIR spectra confirmed the presence of two phases in the WO3/S-g-C3N4 composite. The chemical bonding identification of obtained samples was investigated by infrared spectroscopy. The presence of characteristic bands in single components of WO 3 and S-g-C 3 N 4 confirmed the previous phase identification of WO 3 /S-g-C 3 N 4 (see Figure 3). In the case of pure WO 3 , a broad absorption band in the wavenumber range of 420-1000 cm −1 attributed to the vibration modes of the W-O bond verifies the formation of tungsten(VI) oxide. The strong bands in 745 cm −1 and 818 cm −1 are assigned to the stretching mode of W-O-W [24]. These peaks probably overlap with a band located at 946 cm −1 , attributed to W=O stretching vibration [25,26]. In the case of S-doped graphitic carbon nitride, a narrow band at 808 cm −1 can be attributed to the triazine ring mode, which corresponds to condensed CN heterocycles [27,28]. The distinct absorption bands at 1572 and 1639 cm −1 are attributed to C=N stretching, while the four bands at 1231, 1314, 1396, and 1453 cm −1 to aromatic C-N stretching [29]. In the composite, FTIR spectra are similar to the doped carbon nitride. The difference can be noticed below 1000 cm −1 , in which triazine ring modes from S-g-C 3 N 4 overlap with bands assigned to the stretching mode of W-O-W. Therefore, FTIR spectra confirmed the presence of two phases in the WO 3 /S-g-C 3 N 4 composite.
The morphology of the WO 3 /S-g-C 3 N 4 composite and single components was determined by scanning electron microscopy. SEM images of the photocatalysts are presented in Figure 4. WO 3 possessed a polydisperse nature in which small nanoparticles and discshaped nanostructures can be distinguished (Figure 4a). In the case of S-g-C 3 N 4 (Figure 4b), a flake-like morphology was observed, which is in agreement with previous reports [6,7,13]. Comparing SEM images of the composite (Figure 4c) with the morphology of single components, features of two phases can be indicated. The two-dimensional flakes of S-g-C 3 N 4 are covered with nanoparticles of WO 3 . Therefore, each phase is connected with the other, which results in creating a heterojunction. The morphology of the WO3/S-g-C3N4 composite and single components was determined by scanning electron microscopy. SEM images of the photocatalysts are presented in Figure 4. WO3 possessed a polydisperse nature in which small nanoparticles and disc-shaped nanostructures can be distinguished (Figure 4a). In the case of S-g-C3N4 (Figure 4b), a flake-like morphology was observed, which is in agreement with previous reports [6,7,13]. Comparing SEM images of the composite (Figure 4c) with the morphology of single components, features of two phases can be indicated. The twodimensional flakes of S-g-C3N4 are covered with nanoparticles of WO3. Therefore, each phase is connected with the other, which results in creating a heterojunction. Moreover, the surface area of obtained samples was calculated using the BET method. The values of specific surface areas and pore volumes are presented in Table 1. All materials possessed surface area in the range of 11.5-14.5 m 2 /g. Moreover, to confirm the presence of sulfur in the WO3/S-g-C3N4 composite, X-ray fluorescence spectroscopy (XRF) was performed. The content of S in the binary WO3/S-g-C3N4 (1:1) photocatalyst was about 0.03-0.05% (m/m). Lin et al. [30] have shown that the high dopant concentration can hinder photocatalytic activity. In the case of Fe-doped TiO2, it was demonstrated that the extent of photodegradation diminishes with increasing dopant concentration, and only the 0.01 mol% Fe-TiO2 film showed enhanced photoactivity. It resulted from the presence  The morphology of the WO3/S-g-C3N4 composite and single components was determined by scanning electron microscopy. SEM images of the photocatalysts are presented in Figure 4. WO3 possessed a polydisperse nature in which small nanoparticles and disc-shaped nanostructures can be distinguished (Figure 4a). In the case of S-g-C3N4 (Figure 4b), a flake-like morphology was observed, which is in agreement with previous reports [6,7,13]. Comparing SEM images of the composite (Figure 4c) with the morphology of single components, features of two phases can be indicated. The twodimensional flakes of S-g-C3N4 are covered with nanoparticles of WO3. Therefore, each phase is connected with the other, which results in creating a heterojunction. Moreover, the surface area of obtained samples was calculated using the BET method. The values of specific surface areas and pore volumes are presented in Table 1. All materials possessed surface area in the range of 11.5-14.5 m 2 /g. Moreover, to confirm the presence of sulfur in the WO3/S-g-C3N4 composite, X-ray fluorescence spectroscopy (XRF) was performed. The content of S in the binary WO3/S-g-C3N4 (1:1) photocatalyst was about 0.03-0.05% (m/m). Lin et al. [30] have shown that the high dopant concentration can hinder photocatalytic activity. In the case of Fe-doped TiO2, it was demonstrated that the extent of photodegradation diminishes with increasing dopant concentration, and only the 0.01 mol% Fe-TiO2 film showed enhanced photoactivity. It resulted from the presence Moreover, the surface area of obtained samples was calculated using the BET method. The values of specific surface areas and pore volumes are presented in Table 1. All materials possessed surface area in the range of 11.5-14.5 m 2 /g. Moreover, to confirm the presence of sulfur in the WO 3 /S-g-C 3 N 4 composite, X-ray fluorescence spectroscopy (XRF) was performed. The content of S in the binary WO 3 /S-g-C 3 N 4 (1:1) photocatalyst was about 0.03-0.05% (m/m). Lin et al. [30] have shown that the high dopant concentration can hinder photocatalytic activity. In the case of Fe-doped TiO 2 , it was demonstrated that the extent of photodegradation diminishes with increasing dopant concentration, and only the 0.01 mol% Fe-TiO 2 film showed enhanced photoactivity. It resulted from the presence of lattice distortions at high dopant amounts, which increased the number of recombination centers, hindering photocatalytic activity [25]. To study the chemical composition of the surface WO 3 and WO 3 /S-doped g-C 3 N 4 photocatalysts after the photocatalytic NO x conversion, high-resolution X-ray photoelectron spectroscopy (XPS) analyses were performed. Further analysis of the final composite material confirmed the presence of sulfur (Figure 5a). The peaks at 165.5 eV and 168.5 eV originate from N-S groups and H-SO 3 groups present in the S-doped carbon nitride [31]. On the basis of survey spectra, it can be concluded that the sulfur content is relatively low compared with other elements, which is in agreement with XRF results. In the case of XPS spectra of the C1s regions (Figure 5b), the experimental curve can be deconvoluted into two peaks with binding energies at 285 eV and 288.5 eV, which could be ascribed to C-C and/or graphite C=C surface coordination of carbon, and sp 2 -bonded carbon (C=N), respectively, which are typical groups for g-C 3 N 4 [32,33]. The doublet peaks in Figure 5c, located at a binding energy of 35.5 eV and 37.6 eV, are attributed to W4f 7/2 (W +5 ) and W4f 5/2 (W +6 ), respectively [14]. No sub-stoichiometry of tungsten(VI) oxide and the WO 3 /S-g-C 3 N 4 was observed. The N1s signal (Figure 5d) may be deconvoluted to two peaks at 399 eV and 401 eV, which corresponds to triazine rings (C=N-C) and C-N-H, respectively [32,34,35]. For pure WO 3 N1s region weak broad peak about 400 eV may be assigned to the unreacted NO adsorbed on the surface [36]. The DR/UV-vis spectra of WO3, S-doped g-C3N4, and WO3/S-g-C3N4 are pres Figure 6. All materials revealed the highest absorbance in the range from 200-These photocatalysts are also capable of absorbing visible light at higher wave The DR/UV-vis spectra of WO 3 , S-doped g-C 3 N 4 , and WO 3 /S-g-C 3 N 4 are presented in Figure 6. All materials revealed the highest absorbance in the range from 200-450 nm. These photocatalysts are also capable of absorbing visible light at higher wavelengths, which is consistent with the fact that all samples are colorful (WO 3 is lime-green, whereas S-doped g-C 3 N 4 is yellow, which corresponds to light absorption in the 530-590 nm range). Based on this observation and the calculated bandgap values, it was assumed that all samples could efficiently utilize visible light, which is in the majority of the solar light spectrum. The DR/UV-vis spectra of WO3, S-doped g-C3N4, and WO3/S-g-C3N4 are presented in Figure 6. All materials revealed the highest absorbance in the range from 200-450 nm. These photocatalysts are also capable of absorbing visible light at higher wavelengths, which is consistent with the fact that all samples are colorful (WO3 is lime-green, whereas S-doped g-C3N4 is yellow, which corresponds to light absorption in the 530-590 nm range). Based on this observation and the calculated bandgap values, it was assumed that all samples could efficiently utilize visible light, which is in the majority of the solar light spectrum.  Moreover, photoluminescence spectra and luminescence decay curves were determined, as presented in Figure 7. The highest PL emission was observed for S-doped g-C 3 N 4 with the maximum at~455 nm (blue light) under UV-light excitation (Figure 7a). Similar observations were noticed when the sample was excited in visible light (Figure 7b). According to Zhang et al. [7], carbon nitride can be characterized by enhanced photoresponse and an extremely high quantum yield up to 19.6%. Although these properties are attractive in many potential applications, e.g., bioimaging, in the case of photocatalysis, the high PL emission may hinder the photocatalytic activity. The emission is connected with radiative recombination, which is not desirable because this process lowers the charge carriers' concentration [4]. Therefore, the photocatalytic processes with g-C 3 N 4 as a single compound are insufficient and require further photocatalyst modification [37]. The lower PL intensity for WO 3 /S-g-C 3 N 4 in comparison with pure S-doped carbon nitride was noticed. The most visible differences are especially in Figure 6a under λ exc = 300 nm conditions. These results confirm the limited recombination rate due to efficient charge carriers' separation in the heterojunction [38]. Furthermore, charge carriers' lifetimes were calculated based on luminescence decay curves (Figure 7c,d). The average carrier lifetimes <τ> are presented in Table 1. All photocatalysts possessed the <τ> value of 14-16 ns. It is worth highlighting that the charge carrier lifetimes do not decrease when the light excitation wavelength is changed from 300 nm (UV) to 415 nm (visible). This is another confirmation that these materials are photoactive under solar irradiation. Moreover, there is no significant decline in the <τ> value when S-doped carbon nitride is introduced into the composite with WO 3 . Under 300 nm excitation conditions, the carrier lifetime is prolonged in comparison with single S-g-C 3 N 4 , which can be explained by efficient charge carriers' separation. This process resulted in enhanced photocatalytic performance. This is another confirmation that these materials are photoactive under solar irradiation. Moreover, there is no significant decline in the <τ> value when S-doped carbon nitride is introduced into the composite with WO3. Under 300 nm excitation conditions, the carrier lifetime is prolonged in comparison with single S-g-C3N4, which can be explained by efficient charge carriers' separation. This process resulted in enhanced photocatalytic performance.

Photocatalytic Activity and Degradation Mechanism
The photocatalytic activity of the WO 3 /S-g-C 3 N 4 composite, as well as single-phase photocatalysts, were demonstrated towards NO 2 conversion in a continuous-flow system under simulated solar light irradiation. The initial concentration of NO 2 was about 5 ppm. As shown in Figure 8a, WO 3 /S-doped g-C 3 N 4 composite exhibited near 100% NO 2 removal within 5 min of the process. The high efficiency of this composite is almost constant within time. In the case of pure WO 3 and S-doped g-C 3 N 4, rapid removal of NO 2 concentration was also observed; however, photocatalytic activity decreased within the time, which was particularly noticeable for S-doped g-C 3 N 4 . S-doped g-C 3 N 4 was supposed to transform NO 2 to forms, which caused blocking of the active sites on the photocatalyst's surface [39]. These results were compared with NO production (Figure 8b). Noticeably, the NO 2 → NO conversion is favorable for the WO 3 single compound. Che-Chin Yu et al. [40] reported the phototransformation of NO 2 into NO in the N 2 presence over WO 3 photocatalyst regarding the following reaction: NO 2 + N 2 → 2NO. It is consistent with the probable presence of unreacted NO adsorbed on the surface, indicated by XPS spectra. Therefore, reduction processes are predominant in the case of tungsten(VI) oxide.
Considering the overall degradation of NO x species, only the composite WO 3 /Sdoped g-C 3 N 4 allowed the complete degradation of NO 2 and did not cause the production of NO. According to the available literature reports, two possible explanations of this observation for the composite are possible. First, the surface of the binary composite is more reactive than for single components because of the more efficient charge carriers' separation. A similar effect was described by Méndez-Román and Cardona-Martínez [41], who investigated SiO 2 -TiO 2 photocatalysts towards oxidation of toluene. The composite exhibited better photocatalytic performance and lower deactivation than TiO 2 due to higher toluene adsorption capacity and changes in Brønsted acidity of intermediates. Therefore, new surface sites would be responsible for the enhanced photocatalytic performance. The second possibility is a change of photocatalytic mechanism when WO 3 and S-doped g-C 3 N 4 are present in the heterojunction. However, changes in charge carriers' transport in the composite caused the promotion of NO 2 reduction to gaseous NO and N 2 . In this case, less NO 3 − was adsorbed on the photocatalyst surface than from NO 2 oxidation [42]. Therefore, according to Dalton et al. [36], XPS spectroscopy is needed as a technique suitable for surface analysis. However, the accurate indication of the NO 2 conversion pathway is difficult for WO 3 /S-doped g-C 3 N 4 using XPS spectroscopy because of the predominant C-N signals in N1s spectra. This inconvenience was described precisely by Li et al. [43], who modified graphitic carbon nitride using nitric acid or ammonia gas atmosphere. Remarkably, there were no differences in XPS spectra of N1s signals, although the adsorption of nitrates or ammonia was certain. In this study, the negligible NO production during the photocatalytic process in the presence of composite photocatalyst may result from advanced photodecomposition without NO as a main intermediate product.
The second possibility is photooxidation to nitrates. However, in this case, blocking active sites and a further decrease in activity should be observed, whereas the performed subsequent degradation cycles showed the long-term stability of the WO 3 /S-doped g-C 3 N 4 composite material [44,45].

Effect of Process Parameters
The photocatalytic process, including NO2 removal, depends on different factors, including temperature, relative humidity, pollutant concentration, and the amount of photocatalyst [31,32]. Therefore, in this study, the effect of these parameters on photocatalytic activity was evaluated. Three relative humidity values were set as 25%, 40%, and 55%, respectively. As shown in Figure 9a for the WO3/S-doped g-C3N4 composite, relative humidity had no significant effect on the NO2 conversion. Other observations were reported by Maggos et al. [46] during photocatalytic NOx removal in the presence of TiO2. In their study, humidity plays a crucial role in the photooxidation process since water contributes to reactive oxygen species (ROS) formation. From the technological point of view, eliminating the humidity effect can allow photocatalysis without the dependence on external weather conditions. Afterward, the effect of gas flow rate was evaluated and fixed on 1.3 dm 3 /min and 2.6 dm 3 /min. In contrast to observations from the influence of humidity, the gas flow rate was a significant parameter affecting the photocatalytic efficiency, as shown in Figure 9b. A two-fold increase in gas flow rate caused a remarkable decrease in NO2 conversion. This may be related to the short contact time between pollutant and photocatalyst surface during the passage of higher gas flow rate [47]. A similar effect was observed by Li et al. [48] in photocatalytic depolymerization of the oxidized β-O-4 molecule in the continuousflow reactor. The authors noticed that too fast flow rate reduced photocatalytic efficiency due to the short contact time between the reactant and photocatalyst, while when the flow rate was too slow, performance was limited due to the possible mass transfer resistance.
Subsequently, the effect of photocatalyst dose was investigated (Figure 9c). Firstly, the chosen amount of photocatalyst-2 g for the glass plate in the size 30.5 cm × 30.5 cm-

Effect of Process Parameters
The photocatalytic process, including NO 2 removal, depends on different factors, including temperature, relative humidity, pollutant concentration, and the amount of photocatalyst [31,32]. Therefore, in this study, the effect of these parameters on photocatalytic activity was evaluated. Three relative humidity values were set as 25%, 40%, and 55%, respectively. As shown in Figure 9a for the WO 3 /S-doped g-C 3 N 4 composite, relative humidity had no significant effect on the NO 2 conversion. Other observations were reported by Maggos et al. [46] during photocatalytic NO x removal in the presence of TiO 2 . In their study, humidity plays a crucial role in the photooxidation process since water contributes to reactive oxygen species (ROS) formation. From the technological point of view, eliminating the humidity effect can allow photocatalysis without the dependence on external weather conditions.

Stability Tests
The stability of the WO3/S-doped g-C3N4 composite was evaluated in five subsequent cycles, as presented in Figure 10. The decline of NO2 conversion to ca. 80% was observed. This is consistent with the fact that adsorption of the NOx and water molecules are necessary to initiate the photocatalytic process, which can block the active sites. However, the second and fifth cycles with increased photocatalytic efficiency confirmed that the WO3/S-doped g-C3N4 material is capable of desorbing the adsorbed species and unblocking the active species for better NO2 conversion. This observation confirms that the obtained composite showed long-term stability. Afterward, the effect of gas flow rate was evaluated and fixed on 1.3 dm 3 /min and 2.6 dm 3 /min. In contrast to observations from the influence of humidity, the gas flow rate was a significant parameter affecting the photocatalytic efficiency, as shown in Figure 9b. A two-fold increase in gas flow rate caused a remarkable decrease in NO 2 conversion. This may be related to the short contact time between pollutant and photocatalyst surface during the passage of higher gas flow rate [47]. A similar effect was observed by Li et al. [48] in photocatalytic depolymerization of the oxidized β-O-4 molecule in the continuous-flow reactor. The authors noticed that too fast flow rate reduced photocatalytic efficiency due to the short contact time between the reactant and photocatalyst, while when the flow rate was too slow, performance was limited due to the possible mass transfer resistance.
Subsequently, the effect of photocatalyst dose was investigated (Figure 9c). Firstly, the chosen amount of photocatalyst-2 g for the glass plate in the size 30.5 cm × 30.5 cmallowed for the highest conversion of NO 2 . Other investigated doses: 1 g and 3 g caused the depletion of NO 2 removal within the experimental time. The observed reduction of NO 2 conversion during the process with the smaller amount of the photocatalyst could be explained by the smaller presence of available active sites. On the other hand, the rise of the catalyst dose could cause the covering of some photocatalysts particles, thereby cutting off the radiation access.
Finally, the NO 2 initial concentration effect on NO 2 conversion was examined. The selected initial concentrations of NO 2 were as follows: 2.5 ppm, 5.0 ppm, and 10.0 ppm. As shown in Figure 9d, 100% of NO 2 conversion was maintained after 5 min and 30 min, respectively, for 5 ppm and 2.5 ppm. On the contrary, during the process with the initial concentration of 10 ppm, the effectiveness of NO 2 removal decreased over time from 100% after 5 min up to 37% after 100 min.

Stability Tests
The stability of the WO 3 /S-doped g-C 3 N 4 composite was evaluated in five subsequent cycles, as presented in Figure 10. The decline of NO 2 conversion to ca. 80% was observed. This is consistent with the fact that adsorption of the NO x and water molecules are necessary to initiate the photocatalytic process, which can block the active sites. However, the second and fifth cycles with increased photocatalytic efficiency confirmed that the WO 3 /S-doped g-C 3 N 4 material is capable of desorbing the adsorbed species and unblocking the active species for better NO 2 conversion. This observation confirms that the obtained composite showed long-term stability. In addition, XRD analysis and FTIR spectra after the photocatalytic process were performed. As presented in Figure 11a, no signals on XRD patterns, which can originate from the presence of an additional phase, were noticed. Similar conclusions can be noted from FTIR spectra in Figure 11b; no additional bands were detected after the photocatalytic process. Figure 11. (a) XRD patterns and (b) FTIR spectra of WO3/S-doped g-C3N4 before and after the photocatalytic process.

Conclusions
In this study, the photocatalytic performance of WO3/S-doped g-C3N4 nanocomposite In addition, XRD analysis and FTIR spectra after the photocatalytic process were performed. As presented in Figure 11a, no signals on XRD patterns, which can originate from the presence of an additional phase, were noticed. Similar conclusions can be noted from FTIR spectra in Figure 11b; no additional bands were detected after the photocatalytic process. In addition, XRD analysis and FTIR spectra after the photocatalytic process w performed. As presented in Figure 11a, no signals on XRD patterns, which can origi from the presence of an additional phase, were noticed. Similar conclusions can be no from FTIR spectra in Figure 11b; no additional bands were detected after photocatalytic process.

Conclusions
In this study, the photocatalytic performance of WO3/S-doped g-C3N4 nanocompo towards NOx removal was studied for the first time. The experiment was performed continuous-flow photoreactor on the pilot scale. Simulated solar light emitted by a ne Figure 11. (a) XRD patterns and (b) FTIR spectra of WO 3 /S-doped g-C 3 N 4 before and after the photocatalytic process.

Conclusions
In this study, the photocatalytic performance of WO 3 /S-doped g-C 3 N 4 nanocomposite towards NO x removal was studied for the first time. The experiment was performed in a continuous-flow photoreactor on the pilot scale. Simulated solar light emitted by a newly designed LED array was used as a light source. The highly efficient composites were successfully obtained without additional phases, confirmed by X-ray diffraction. SEM images of the nanomaterials showed the nanocomposite morphology as two-dimensional. Sdoped g-C 3 N 4 nanosheets were covered by WO 3 nanoparticles. XPS and XRF spectroscopy confirmed the presence of sulfur in the final photocatalyst. High absorbance and suitable bandgap value proved that the obtained photocatalysts utilize solar light efficiently. The photoluminescence spectra and carriers' lifetime calculations showed better separation of charge carriers in the composite than in a single component. Finally, WO 3 , S-doped g-C 3 N 4, and WO 3 /S-doped g-C 3 N 4 were characterized by fast photocatalytic NO 2 removal. Almost 100% of NO 2 was degraded in less than 5 min of the process. For S-doped g-C 3 N 4 nanosheets and WO 3 nanoparticles, active sites were further blocked, which caused a decrease in NO 2 removal. This effect was not observed for the WO 3 /S-doped g-C 3 N 4 composite. Additionally, the newly developed composite efficiently degraded NO 2 without the simultaneous production of NO. Remarkably, the composite photocatalyst was tested in the subsequent degradation cycles, which confirmed the long-term stability of the composite. No structural changes were noticed after the photocatalytic processes.
To summarize, the pilot-scale studies of the photocatalysts towards NO x conversion under simulated solar light were demonstrated. The high photocatalytic activity of WO 3 /Sdoped g-C 3 N 4 and simple photocatalyst synthesis encourages the industrial application of the obtained nanocomposites in air purification processes.

Conflicts of Interest:
The authors declare no conflict of interest.