Graphitic Carbon Nitride for Photocatalytic Air Treatment

Graphitic carbon nitride (g-C3N4) is a conjugated polymer, which recently drew a lot of attention as a metal-free and UV and visible light responsive photocatalyst in the field of solar energy conversion and environmental remediation. This is due to its appealing electronic band structure, high physicochemical stability and earth-abundant nature. In the present work, bulk g-C3N4 was synthesized by thermal decomposition of melamine. This material was further exfoliated by thermal treatment. S-doped samples were prepared from thiourea or further treatment of exfoliated g-C3N4 by mesylchloride. Synthesized materials were applied for photocatalytic removal of air pollutants (acetaldehyde and NOx) according to the ISO 22197 and ISO 22197-1 methodology. The efficiency of acetaldehyde removal under UV irradiation was negligible for all g-C3N4 samples. This can be explained by the fact that g-C3N4 under irradiation does not directly form hydroxyl radicals, which are the primary oxidation species in acetaldehyde oxidation. It was proved by electron paramagnetic resonance (EPR) spectroscopy that the dominant species formed on the irradiated surface of g-C3N4 was the superoxide radical. Its production was responsible for a very high NOx removal efficiency not only under UV irradiation (which was comparable with that of TiO2), but also under visible irradiation.


Introduction
Graphitic carbon nitride (g-C 3 N 4 ) is a polymeric material based on melon units. The extraordinary diamond-like properties of g-C 3 N 4 are thermal, chemical and photochemical stability due to heptazine-based building blocks and strong covalent bonds between carbon and nitrogen atoms. Another important property of g-C 3 N 4 is its band gap energy of 2.7 eV, which enables absorption of visible light (λ ≤ 460 nm). This is why various applications of g-C 3 N 4 in photochemistry, including photocatalysis, have been studied. Except PVD (physical vapour deposition), CVD (chemical vapour deposition) and synthetic procedures, g-C 3 N 4 can be prepared by the thermal condensation of nitrogen-rich precursors, such as cyanamide, dicyandiamide, melamine, urea and so forth [1]. However, the photocatalytic applications of g-C 3 N 4 are limited by fast recombination of photoinduced electrons and holes, which causes low quantum efficiency. It can be overcome by doping with metals and non-metals. Doping with non-metals, such as phosphorus, sulphur, oxygen, boron, carbon, nitrogen and halogens, leads furthermore to a red shift of the absorption edge [2].
One of the promising applications of semiconductor photocatalysis is improvement of air quality in interiors. Aldehydes and other VOC (volatile organic compounds), which are released by furniture, may cause the sick building syndrome including non-specific diseases, e.g., respiratory irritation symptoms, headaches, dizziness, and fatigue [3]. Removal of acetaldehyde and formaldehyde is also part of the ISO 22197 methodology for evaluating the photocatalytic efficiency of pollutant removal from the gaseous phase [4]. The mechanism of photocatalytic removal of formaldehyde is often explained by the attack of a HO · radical to form an acetyl radical, followed by a subsequent radical reaction to generate the final products CO 2 and formaldehyde [5]. Formaldehyde can be further photocatalytically oxidized to CO 2 and H 2 O via a mechanism described in [6].
CH 3 COO · → · CH 3 + CO 2 (3) Another application of photocatalytic processes in air treatment is the removal of NO x . This process is based on the oxidation of NO and NO 2 (NO x ) in the gaseous phase into NO 2 − /NO 3 − in the aqueous phase. This process is also used as a standard ISO method for the determination of the photocatalytic efficiency of NO x removal (ISO 22197-1) [6,7]. In the case of TiO 2 as photocatalyst, the photocatalytic oxidation of NO in the inlet gaseous stream proceeds according to reactions (6)(7). Potential accumulation of nitric acid on the photocatalyst surface may cause generation of NO 2 (according to reaction 8), which is more toxic than NO. This can eventually lead to a steady state where the rate of NO decomposition is matched by the rate of NO 2 production. It follows that nitric acid must be periodically washed out (e.g., by rain), to minimise the NO 2 production.
In a previous work, Praus et al. [8] developed a novel procedure based on the post-synthetic derivatization of g-C 3 N 4 with methanesulfonyl (mesyl (CH 3 SO 2 − ) chloride) and thus synthetized S-doped g-C 3 N 4 and exfoliated g-C 3 N 4. The obtained nanomaterials were characterized in detail but their photocatalytic activity was tested only in the liquid phase and by means of the decomposition of Acid Orange 7 under ultraviolet A (UVA) irradiation. Therefore, the aim of the present work was the evaluation of the influence of exfoliation and doping on the photoactivity of g-C 3 N 4 in the gaseous phase using both UVA and VIS light. As model gaseous pollutants, acetaldehyde and NO were selected and the results were compared with the TiO 2 photocatalyst (P25), which is considered as a benchmark for practical applications.

Sample Preparation and Characterization
Bulk g-C 3 N 4 (CN) was prepared by thermal treatment of melamine at 550 • C and further exfoliated at 500 • C using a procedure described in [9] (sample marked Ex-CN). In this way, the specific surface area increased from 8 to 93 m 2 g −1 . Sulphur-doped g-C 3 N 4 photocatalysts were prepared by (i) thermal Materials 2020, 13, 3038 3 of 14 treatment of thiourea (sample S-CN) and (ii) derivatization of previously prepared bulk and exfoliated g-C 3 N 4 with mesylchloride (sample marked Mes-Ex-CN). The procedure is described in ref. [8].
For comparison, two TiO 2 photocatalysts, namely the well-established P25 (Evonik) and pure anatase CG100 (Precheza) of comparable BET surface area as g-C 3 N 4 , were involved in the study (Table 1). The immobilization of the synthetized g-C 3 N 4 powder and commercial TiO 2 materials on glass substrates was performed by drop casting from suspension and drying. The amount of photocatalytic material has been kept constant as 0.5 mg/cm 2 .

Photodegradation Process Setup-NO and Acetaldehyde
The efficiency of the photocatalytic removal of NO and acetaldehyde (AcAh) from the gaseous phase was measured in a single-pass flow-through reactor according to standard ISO methods (ISO 22197-1 removal of NO x [4] and ISO 22197-2 removal of acetaldehyde [7]). The use of ISO methods is essential for possible comparison of results from different laboratories. An overview of ISO methods in photocatalysis was given recently by Mills et al. [6]. Defined process parameters (input gas concentration, flow rate, humidity, radiation source, etc.) are listed in Table 2. The content of g-C 3 N 4 in all immobilized photocatalyst layers was fixed (0.5 mg/cm 2 ). As pollutant sources, calibrated gas mixtures from SIAD (Rajhradice, Czech Republic) were used: 100 ppm NO in N 2 and 250 ppm acetaldehyde in N 2 . These mixtures were further diluted with humidified air using mass flow controllers (see Figure 1). The final mixture, as admitted to the reactor, was either 5 ppm acetaldehyde or 1 ppm NO in air. The relative humidity (RH) was set to 50%. Analysis of the gaseous mixture was performed by a chemiluminescence analyzer (NO and NO 2 ) and a gas chromatograph equipped with a flame ionization detector (GC-FID) in case of acetaldehyde. flow rate (dm 3 /min) 3 1 UV irradiance (mW/cm 2 ) 1 1 analysis chemiluminescence analyzer GC  The active part of the reactor (photocatalyst-coated plate) was 5 cm × 10 cm and the distance from the glass cover was 0.5 cm. The emission spectra of the incident light for all light sources at the location of the reactor were measured using a spectroradiometer (Specbos 1211, Jeti, Germany) and are shown in Figure 2. As a UV light source, a 25 W blacklight (UVA) tubular light source (351 nm centre wavelength) was used. The irradiance at the location of the reactor was 1 mW/cm 2 between 340 and 400 nm.

EPR Measurements
The X-band electron paramagnetic resonace (EPR) spectra of powder samples were monitored at room temperature by an EMXPlus EPR spectrometer (Bruker, Rheinstetten, Germany) operating at 100 kHz field modulation using a High Sensitivity Probe-head (Bruker). The g-factors were determined with an uncertainty of ± 0.0005 exploiting a nuclear magnetic resonance teslameter (ER 036TM, Bruker, Rheinstetten, Germany) and integrated frequency counter. The samples were placed in thin-walled quartz EPR tubes (Bruker) and the spectra were monitored in dark and upon continuous in situ irradiation using an UV LED source (λmax = 365 nm; Bluepoint LED, Hönle UV  NaNO 2 enabling transmission of light with λ > 400 nm), with an irradiance of 0.4 mW/cm 2 between 400 and 480 nm, was used.

EPR Measurements
The X-band electron paramagnetic resonace (EPR) spectra of powder samples were monitored at room temperature by an EMXPlus EPR spectrometer (Bruker, Rheinstetten, Germany) operating at 100 kHz field modulation using a High Sensitivity Probe-head (Bruker). The g-factors were determined with an uncertainty of ± 0.0005 exploiting a nuclear magnetic resonance teslameter (ER 036TM, Bruker, Rheinstetten, Germany) and integrated frequency counter. The samples were placed in thin-walled quartz EPR tubes (Bruker) and the spectra were monitored in dark and upon continuous in situ irradiation using an UV LED source (λ max = 365 nm; Bluepoint LED, Hönle UV Technology, Gräfelfing/München, Germany; irradiance of 15 mW cm −2 within the EPR cavity determined by a UVX radiometer (UVP, Upland, CA, USA)) or using a visible-light source (KL 1600LED (T = 5600 K); Schott, Mainz, Germany). Standard EPR spectrometer settings were as follows: microwave frequency, 9.426 GHz; microwave power, 10.66 mW; center field, 336 mT; gain, 2.52 × 10 5 ; modulation amplitude, 0.1 mT; sweep time, 41 s; time constant, 20.48 ms; number of scans, 10. The generation of oxygen-containing reactive radicals upon in situ LED@365 nm exposure of aerated aqueous suspensions of photocatalysts was studied using the EPR spin-trapping technique with 5,5-dimethyl-1-pyrroline N-oxide (DMPO, Sigma-Aldrich, Buchs, Switzerland, distilled prior to the application). The spin-trapping experiments were performed at least in triplicate and the experimental spectra were processed and analyzed analogously as described in [11]. The spin-Hamiltonian parameters and relative concentration of individual DMPO-adducts were calculated using the EasySpin program [12].

Results
According to XRD phase analysis, TiO 2 P25 contains mostly the anatase phase (70%) with a crystal size around 22 nm. The remaining phase is rutile with a similar crystal size. The second commercial TiO 2 marked CG100 contains only the anatase phase with a smaller crystal size (12 nm), which corresponds to a higher surface area (70-110 m 2 /g) compared to TiO 2 P25 (45 m 2 /g).
In the case of g-C 3 N 4 materials, two diffraction patterns corresponding to a CN hexagonal phase were observed. (the diffractogram is shown as Figure 3 in ref. [8]). Compared to TiO 2 , the crystal size of the g-C 3 N 4 material is much lower (about 6 nm, detailed information given in [8]). The crystal size after exfoliation does not change. This observation is not in agreement with the specific surface area obtained by BET, which significantly differs between bulk material (CN) and exfoliated material (Ex-CN), (see Table 1). This can be explained by partial delamination of g-C 3 N 4 particles providing non-diffracting nanosheets.
Prepared materials were characterized using the FTIR-ATR technique. Spectra are shown in Figure S1 (Supplementary Information). The broad weak spectral band in region (3300-3000 cm −1 ) can be assigned to stretching vibrations of N-H bonds. Other bands in region 1630-1232 cm −1 correspond to stretching vibration of C=N, and a C-N heterocyclic ring. The band at 804 cm −1 is attributed to the breathing vibration of triazine units.
Photoelectron spectroscopy was used to prove the presence of carbon, oxygen and sulfur. An example of an S 2p spectrum is shown in Figure S2. Two types of incorporated sulfur were indicated. A first binding energy around 164 eV indicates sulfur in a low oxidation state (sulfide or thiol). The second couple of 2p binding energies corresponds to sulfur in a lower oxidation state (mesyl groups). A detailed description of XPS and FTIR spectra of g-C 3 N 4 has been reported previously [8].
For the C 3 N 4 materials, the calculated band gap energies vary from the lowest values (2.63 eV) observed for S-CN, to 2.77 eV observed for Ex-CN. From that, it follows that the excitation wavelength must be lower than 471 nm (S-CN) and 448 nm (Ex-CN), respectively [8]. As it is well known, the band gap energy for anatase is 3.2 eV and for rutile 3.0 eV, which means that anatase absorbs only UV light (lower than 388 nm), while rutile can absorb also a small part of visible light (lower than 414 nm). From the emission spectra of VIS fluorescent tubes (Figure 2), it follows that C 3 N 4 materials can utilize both peaks (404 and 435 nm) for excitation of electrons, while TiO 2 P25 can utilize only one peak (403 nm) due to the presence of the rutile phase. utilize only one peak (403 nm) due to the presence of the rutile phase.

Acetaldehyde Removal
The concentration of acetaldehyde as a function of time at the output of the flow through system during the test for sample Ex-CN is shown in Figure 3. After reaching a stable input concentration of AcAh, the test gas was introduced to the reactor chamber in the dark, resulting in a decrease in concentration caused by adsorption. After reaching the adsorption equilibrium, the initial concentration was restored. After switching the UV light on, no decrease in the AcAh concentration was observed, which indicates that the exfoliated sample (Ex-CN) did not show a measurable conversion of acetaldehyde. Similar results were obtained for S-CN and Mes-Ex-CN.

Acetaldehyde Removal
The concentration of acetaldehyde as a function of time at the output of the flow through system during the test for sample Ex-CN is shown in Figure 3. After reaching a stable input concentration of AcAh, the test gas was introduced to the reactor chamber in the dark, resulting in a decrease in concentration caused by adsorption. After reaching the adsorption equilibrium, the initial concentration was restored. After switching the UV light on, no decrease in the AcAh concentration was observed, which indicates that the exfoliated sample (Ex-CN) did not show a measurable conversion of acetaldehyde. Similar results were obtained for S-CN and Mes-Ex-CN. Figure 4 shows the removal of acetaldehyde for an immobilized TiO 2 P25 layer. We can see a significant decrease in the acetaldehyde concentration after the UV light is switched on. The conversion of acetaldehyde on the TiO 2 layer was about 80%.

NOx Removal
The concentration of NO, produced NO2 and the sum of NO and NO2 (NOx) as a function of time for an immobilized layer of Ex-CN is shown in Figure 5. In contrast to acetaldehyde, after switching the UV light on, the decrease in the NO concentration immediately set in (to 0.5 ppm) accompanied

NO x Removal
The concentration of NO, produced NO 2 and the sum of NO and NO 2 (NO x ) as a function of time for an immobilized layer of Ex-CN is shown in Figure 5. In contrast to acetaldehyde, after switching the UV light on, the decrease in the NO concentration immediately set in (to 0.5 ppm) accompanied by an increase in the NO 2 concentration (to 0.2 ppm). These concentrations remained fairly constant during irradiation. It can be seen that the conversion of NO was about 50% and the conversion of NO x was about 30%.   The dependence of NO, NO 2 and NO x concentration is shown for an immobilized layer of TiO 2 P25 in Figure 6. We can see a similar trend, only the increase in the NO 2 concentration was slower but the constant value reached after 1 h of irradiation was 0.2 ppm and the concentration of NO achieved was 0.5 ppm.   The amounts of removed NO, produced NO 2 and removed NO x , respectively, under UV irradiation for various photocatalysts are summarized in Figure 7. In the case of bulk material (CN), the removal rate of NO x is only about 0.3 µmol/h. An even lower removal rate was observed for a material prepared from thiourea (S-CN). An increase in the removal rate of NO x was observed for exfoliated C 3 N 4 (Ex-CN) (0.7 µmol/h), but this is still only half of the NO x removal rate for P25 TiO 2. Significant improvement in the removed NO x can be seen for an exfoliated sample treated using mesyl chloride (Mes-Ex-CN). Although the amount of removed NO was lower than that for P25 TiO 2 , the amount of produced NO 2 was two times lower for C 3 N 4 , resulting in a higher overall NO x removal capability. In comparison with pure anatase (CG100) an even more significant improvement in the removal rate of NO x (two times) was observed, mainly due to the lower NO 2 production.  In the next step, a monochromatic LED light source, which had a maximal intensity at 450 nm (which is closer to the C3N4 band gap threshold), was used to assess the photocatalytic activity in the visible light. The exfoliated sample (Ex-CN) and TiO2 P25 were compared and results are shown in Figure 9. Due to the absence of photons below 420 nm (Figure 2), TiO2 P25 showed no NOx conversion. In the case of Ex-CN material, we can see a high level of NOx removal, which is approximately six times higher compared to the NOx removal in Figure 8. It should be mentioned that the VIS fluorescent source and VIS LED source differ dramatically in intensities. In the case of the VIS LED, it was 10 times higher so that the increase in NOx removal was proportional to the increase in light intensity.
The C3N4-based nanomaterials usually possess paramagnetic centers traceable by EPR spectroscopy, which reflect the applied synthetic procedure and post-synthetic treatments. All the studied nanostructures exhibited the presence of an EPR signal prior to the photoexcitation and the EPR spectra obtained at room temperature (RT) represent a single Lorenztian-line with g = 2.0038. An analogous signal was monitored previously in C3N4-based nanomaterials [13,14] and was assigned to the unpaired electrons in the aromatic rings of carbon atoms in the localized π-states of typical heptazine.   In the next step, a monochromatic LED light source, which had a maximal intensity at 450 nm (which is closer to the C 3 N 4 band gap threshold), was used to assess the photocatalytic activity in the visible light. The exfoliated sample (Ex-CN) and TiO 2 P25 were compared and results are shown in Figure 9. Due to the absence of photons below 420 nm (Figure 2), TiO 2 P25 showed no NO x conversion. In the case of Ex-CN material, we can see a high level of NO x removal, which is approximately six times higher compared to the NO x removal in Figure 8. It should be mentioned that the VIS fluorescent source and VIS LED source differ dramatically in intensities. In the case of the VIS LED, it was 10 times higher so that the increase in NO x removal was proportional to the increase in light intensity.  The performance of the studied nanomaterials upon continuous exposure was also followed by EPR, as the photogenerated charge carriers and their behaviour are coupled with the emergence of paramagnetic species. The UVA (λmax = 365 nm) as well as visible-light (λ > 420 nm) exposure of the studied nanomaterials led to an increase in the intensity of the Lorentzian single line at g = 2.0038, while a more prominent effect was induced by visible-light exposure (Figure 10a). The increase in the intensity of the Lorentzian signal monitored upon photoactivation reflects the process of charge carrier formation/trapping. The signal is ascribed to the electrons in the localized states, especially the surface-trapped photogenerated electrons with preferentially carbon character, and the missing hyperfine interaction of nitrogen nuclei is attributed to the EPR line broadening [14,15]. To compare the effect of the light exposure on all the studied nanomaterials, the ratio of EPR integral intensity elucidated from the double-integrated X-band EPR spectra of the samples measured at RT upon UVA The C 3 N 4 -based nanomaterials usually possess paramagnetic centers traceable by EPR spectroscopy, which reflect the applied synthetic procedure and post-synthetic treatments. All the studied nanostructures exhibited the presence of an EPR signal prior to the photoexcitation and the EPR spectra obtained at room temperature (RT) represent a single Lorenztian-line with g = 2.0038. An analogous signal was monitored previously in C 3 N 4 -based nanomaterials [13,14] and was assigned to the unpaired electrons in the aromatic rings of carbon atoms in the localized π-states of typical heptazine.
The performance of the studied nanomaterials upon continuous exposure was also followed by EPR, as the photogenerated charge carriers and their behaviour are coupled with the emergence of paramagnetic species. The UVA (λ max = 365 nm) as well as visible-light (λ > 420 nm) exposure of the studied nanomaterials led to an increase in the intensity of the Lorentzian single line at g = 2.0038, while a more prominent effect was induced by visible-light exposure (Figure 10a). The increase in the intensity of the Lorentzian signal monitored upon photoactivation reflects the process of charge carrier formation/trapping. The signal is ascribed to the electrons in the localized states, especially the surface-trapped photogenerated electrons with preferentially carbon character, and the missing hyperfine interaction of nitrogen nuclei is attributed to the EPR line broadening [14,15]. To compare the effect of the light exposure on all the studied nanomaterials, the ratio of EPR integral intensity elucidated from the double-integrated X-band EPR spectra of the samples measured at RT upon UVA or VIS-light exposure and in the dark was calculated, providing the values of the relative EPR integral intensity increase presented in Figure 10.
Materials 2020, 13, x FOR PEER REVIEW 11 of 16 or VIS-light exposure and in the dark was calculated, providing the values of the relative EPR integral intensity increase presented in Figure 10.

Mechanism of Photocatalytic Removal
The different efficiencies of photocatalytic removal of NOx and acetaldehyde on the C3N4 and TiO2 photocatalyst can be explained by the particulate positions of their valence-band (VB) and conduction-band (CB) edges. The position of the valence-band edge (VBedge) of C3N4 is only 1.4 V (vs. NHE, pH 7) [16], while for the formation of hydroxyl radical a more positive potential (about 2 V vs. NHE, pH 7) is needed (Figure 11) [17]. This means that the oxidation potential of h + in the VB of C3N4 is not sufficient to produce hydroxyl radicals according to Equation (9) (as in the case of TiO2 [18]), which are considered a primary species in the oxidation of acetaldehyde [19].
In ref. [20], the authors reported the position of the VB (1.57 V vs. NHE) and Eg (2.7 eV) for g-C3N4 materials. After hydrothermal delamination, VB shifted to 1.27 V vs NHE and Eg decreased to 2.3 eV. Thus, the position of VB and band gap energy depends on the preparation procedure and the subsequent treatment). h + + H2O → HO • + H + (9) As was discussed above, although irradiated C3N4 does not produce HO • radicals, we observed a significant NOx conversion. Consequently, under the given experimental conditions, NO and HNO2

Mechanism of Photocatalytic Removal
The different efficiencies of photocatalytic removal of NO x and acetaldehyde on the C 3 N 4 and TiO 2 photocatalyst can be explained by the particulate positions of their valence-band (VB) and conduction-band (CB) edges. The position of the valence-band edge (VB edge ) of C 3 N 4 is only 1.4 V (vs. NHE, pH 7) [16], while for the formation of hydroxyl radical a more positive potential (about 2 V vs. NHE, pH 7) is needed (Figure 11) [17]. This means that the oxidation potential of h + in the VB of C 3 N 4 is not sufficient to produce hydroxyl radicals according to Equation (9) (as in the case of TiO 2 [18]), which are considered a primary species in the oxidation of acetaldehyde [19].
or VIS-light exposure and in the dark was calculated, providing the values of the relative EPR integral intensity increase presented in Figure 10.

Mechanism of Photocatalytic Removal
The different efficiencies of photocatalytic removal of NOx and acetaldehyde on the C3N4 and TiO2 photocatalyst can be explained by the particulate positions of their valence-band (VB) and conduction-band (CB) edges. The position of the valence-band edge (VBedge) of C3N4 is only 1.4 V (vs. NHE, pH 7) [16], while for the formation of hydroxyl radical a more positive potential (about 2 V vs. NHE, pH 7) is needed (Figure 11) [17]. This means that the oxidation potential of h + in the VB of C3N4 is not sufficient to produce hydroxyl radicals according to Equation (9) (as in the case of TiO2 [18]), which are considered a primary species in the oxidation of acetaldehyde [19].
In ref. [20], the authors reported the position of the VB (1.57 V vs. NHE) and Eg (2.7 eV) for g-C3N4 materials. After hydrothermal delamination, VB shifted to 1.27 V vs NHE and Eg decreased to 2.3 eV. Thus, the position of VB and band gap energy depends on the preparation procedure and the subsequent treatment). Figure 11. Position of valence-band and conduction-band edge for g-C3N4 and TiO2 at pH 7 [17]. h + + H2O → HO • + H + (9) As was discussed above, although irradiated C3N4 does not produce HO • radicals, we observed a significant NOx conversion. Consequently, under the given experimental conditions, NO and HNO2  Figure 11. Position of valence-band and conduction-band edge for g-C 3 N 4 and TiO 2 at pH 7 [17].
In ref. [20], the authors reported the position of the VB (1.57 V vs. NHE) and E g (2.7 eV) for g-C 3 N 4 materials. After hydrothermal delamination, VB shifted to 1.27 V vs NHE and E g decreased to 2.3 eV. Thus, the position of VB and band gap energy depends on the preparation procedure and the subsequent treatment).
h + + H 2 O → HO · + H + As was discussed above, although irradiated C 3 N 4 does not produce HO · radicals, we observed a significant NO x conversion. Consequently, under the given experimental conditions, NO and HNO 2 cannot be oxidized by HO · radicals (Equations (6) and (7)), as is the case for TiO 2 . Most probably, other oxygen-containing radical species are involved in the oxidation pathway. The CB-edge of C 3 N 4 is more negative compared to TiO 2 , which predestines strong reduction properties of photogenerated e − . In the presence of molecular oxygen, the photoproduced electrons are readily scavenged, forming the superoxide radical anion (Equation (10)) [21].
The EPR spin-trapping technique was applied to evidence the generation of O 2 ·− /HO 2 · in the aerated aqueous suspension of photocatalysts. Figure 12 illustrates the ability of various types of C 3 N 4 materials irradiated by UV light to produce DMPO spin-adducts. For all dispersed C 3 N 4 photocatalysts, the monitored EPR spectra (inset in Figure 12) Accepting the fact that the hydroxyl radicals are not produced via oxidation by holes, (Equation (9)), the · DMPO-OH spin-adduct may originate from different sources (e.g., hydroxyl radicals from hydrogen peroxide photo/photocatalytic reactions, as dismutation of O 2 ·− /HO 2 · in water produces H 2 O 2 [22], or by the · DMPO-O 2 /O 2 H transformation in aqueous media [23]). Scheme 1 illustrates the generation of both DMPO spin-adducts along with their simulated EPR spectra. The concentration of · DMPO-O 2 /O 2 H followed by EPR spin trapping is also in good agreement with the NO x removal efficiency. Bulk C 3 N 4 (CN) shows the lowest NO x conversion in accordance with the lowest concentration of · DMPO-O 2 /O 2 H. On the other hand, the S-doped exfoliated g-C 3 N 4 (Mes-Ex-CN) exhibits the highest NO x conversion and the highest concentration of · DMPO-O 2 /O 2 H. The irradiation of pure TiO 2 under identical experimental conditions (DMPO, water, air, UV light) resulted in the generation of the · DMPO-OH spin-adduct only, as published in previous work [24] confirming the dominant role of photogenerated hydroxyl radicals for this metal oxide photocatalyst.  (6) and (7)), as is the case for TiO2. Most probably, other oxygen-containing radical species are involved in the oxidation pathway. The CB-edge of C3N4 is more negative compared to TiO2, which predestines strong reduction properties of photogenerated e -. In the presence of molecular oxygen, the photoproduced electrons are readily scavenged, forming the superoxide radical anion (Equation (10)) [21]. The EPR spin-trapping technique was applied to evidence the generation of O2 •-/HO2 • in the aerated aqueous suspension of photocatalysts. Figure 12 illustrates the ability of various types of C3N4 materials irradiated by UV light to produce DMPO spin-adducts. For all dispersed C3N4 photocatalysts, the monitored EPR spectra (inset in Figure 12) represent the superposition of two spin-adducts, i.e., the dominant twelve-line signal of • DMPO-O2/O2H (aN = 1.424 ± 0.002 mT, aH β = 1.132 ± 0.003 mT, aH γ = 0.124 ± 0.002 mT; g = 2.0058 ± 0.0001) and a four-line signal (aN = 1.514 ± 0.005 mT, aH β = 1.472 ± 0.006 mT; g = 2.0057 ± 0.0001) characteristic for the • DMPO-OH adduct. The presence of • DMPO-O2/O2H unambiguously evidences the UV-light-induced generation of O2 •-/HO2 • in the photocatalytic system. Accepting the fact that the hydroxyl radicals are not produced via oxidation by holes, (Equation (9)), the • DMPO-OH spin-adduct may originate from different sources (e.g., hydroxyl radicals from hydrogen peroxide photo/photocatalytic reactions, as dismutation of O2 •-/HO2 • in water produces H2O2 [22], or by the • DMPO-O2/O2H transformation in aqueous media [23]). Scheme 1 illustrates the generation of both DMPO spin-adducts along with their simulated EPR spectra.  The concentration of • DMPO-O2/O2H followed by EPR spin trapping is also in good agreement with the NOx removal efficiency. Bulk C3N4 (CN) shows the lowest NOx conversion in accordance with the lowest concentration of • DMPO-O2/O2H. On the other hand, the S-doped exfoliated g-C3N4 (Mes-Ex-CN) exhibits the highest NOx conversion and the highest concentration of • DMPO-O2/O2H. The irradiation of pure TiO2 under identical experimental conditions (DMPO, water, air, UV light) resulted in the generation of the • DMPO-OH spin-adduct only, as published in previous work [24] confirming the dominant role of photogenerated hydroxyl radicals for this metal oxide photocatalyst.
As described in [5], the superoxide radical is able to oxidize NO to NO3 -(Equation (11)). Another possibility is the formation of hydrogen peroxide (Equation (12)) which is then further reduced to HO • radicals (Equation (13)) [25]. During photocatalytic removal of NO, the production of a typical by-product, NO2, was observed. NO2 is formed most probably, as in the case of TiO2, by its reaction with nitric acid (Equation (8)). Production of NO2 was observed for both UV and VIS irradiation and its production was proportional to the conversion of NO. In the case of photocatalysts, which exhibit lower conversion of NO, production of NO2 was lower (CN, S-CN). On the other hand, in the case of samples where the conversion of NO was significantly higher (TiO2, Ex-CN, Mes-Ex-CN), we observed a higher production rate of NO2. In the case of TiO2, the surface area had no significant effect on NO2 production. The production rate in TiO2 CG 100 (crystal size 12 nm, S BET 70-110 m 2 /g) is almost the same as in the case of P25, which exhibits a slightly lower surface area (45 m 2 /g, crystal size 22 nm). Interesting is the comparison of NO2 formation and overall NOx removal, respectively, on CG100 and Mes-Ex-CN under UV irradiation. CG100 exhibited a slightly higher BET surface area compared to the S-doped exfoliated g-C3N4 sample (67 m 2 /g). The removed amount of NO was similar, but in the g-C3N4 sample the amount of produced of NO2 was lower, which influenced the NOx overall removal rate. It seems that in the case of g-C3N4, the direct oxidation of initial NO to nitrate (Equation (11)) is the predominant pathway of NO removal, while two step oxidation of NO to NO3 -(Equations (6) and (7)) is more typical for TiO2 samples.
O2 + e -→ O2 •- As described in [5], the superoxide radical is able to oxidize NO to NO 3 − (Equation (11)). Another possibility is the formation of hydrogen peroxide (Equation (12)) which is then further reduced to HO · radicals (Equation (13)) [25]. During photocatalytic removal of NO, the production of a typical by-product, NO 2 , was observed. NO 2 is formed most probably, as in the case of TiO 2 , by its reaction with nitric acid (Equation (8)). Production of NO 2 was observed for both UV and VIS irradiation and its production was proportional to the conversion of NO. In the case of photocatalysts, which exhibit lower conversion of NO, production of NO 2 was lower (CN, S-CN). On the other hand, in the case of samples where the conversion of NO was significantly higher (TiO 2, Ex-CN, Mes-Ex-CN), we observed a higher production rate of NO 2 . In the case of TiO 2 , the surface area had no significant effect on NO 2 production. The production rate in TiO 2 CG 100 (crystal size 12 nm, S BET 70-110 m 2 /g) is almost the same as in the case of P25, which exhibits a slightly lower surface area (45 m 2 /g, crystal size 22 nm).
Interesting is the comparison of NO 2 formation and overall NO x removal, respectively, on CG100 and Mes-Ex-CN under UV irradiation. CG100 exhibited a slightly higher BET surface area compared to the S-doped exfoliated g-C 3 N4 sample (67 m 2 /g). The removed amount of NO was similar, but in the g-C 3 N 4 sample the amount of produced of NO 2 was lower, which influenced the NO x overall removal rate. It seems that in the case of g-C 3 N 4 , the direct oxidation of initial NO to nitrate (Equation (11)) is the predominant pathway of NO removal, while two step oxidation of NO to NO 3 − (Equations (6) and (7)) is more typical for TiO 2 samples.

Conclusions
Synthesized exfoliated and S-doped g-C 3 N 4 photocatalysts were characterized and the photocatalytic removal of air pollutants (acetaldehyde and NO x ) was evaluated according to the ISO 22197 and ISO 22197-1 methodology. Photocatalytic conversion of acetaldehyde was negligible for g-C 3 N 4 materials. This can be explained by the valence-band edge position of the g-C 3 N 4 photocatalyst, as the photogenerated holes are not able to generate hydroxyl radicals, which are considered to act as a primary oxidation species in acetaldehyde removal.
In contrast, in the case of g-C 3 N 4 , the most significant photoinduced species is the superoxide radical anion (O 2 ·− ), as proved by the EPR spin-trapping technique. S-doped exfoliated g-C 3 N 4 showed comparable NO and higher NO x removal as TiO 2 under UV irradiation. Besides, both exfoliated and S-doped exfoliated g-C 3 N 4 showed very promising NO x removal efficiencies under VIS irradiation. The pathway of NO removal on g-C 3 N 4 was different from that on TiO 2 , as direct oxidation of NO to NO 3 − by the superoxide radical anions played a major role. Formation of NO 2 was thus diminished, which resulted in an improved total NO x removal efficiency. g-C 3 N 4 exhibited a very high NO x removal efficiency not only under UV irradiation (which was comparable with that of TiO 2 ), but also under visible irradiation.

Funding:
The authors acknowledge the financial support from the Czech Science Foundation (project number 19-15199S). This study was financially supported by Scientific Grant Agency of the Slovak Republic (VEGA Project 1/0026/18). VB and DD thank the Ministry of Education, Science, Research and Sport of the Slovak Republic for funding within the scheme "Excellent research teams".