Reactive radicals are of key importance in many fields of degradation of environmental pollutants, such as heterogeneous catalysis for the NOx/VOCs removal [1, 2], advanced oxidation for the organic pollutants treatment of wastewater [3, 4] and photocatalytic environmental remediation [5-7]. Thus, the generation and transformation of active radicals gives rise to a general concern in the environmental and catalysis fields. As in photocatalysis, the redox reactivity is brought about by photoexcited electrons (e-) and holes (h+) produced in the photocatalysts [8-10]. Several active radicals, which are the major source of oxidation capacity, are subsequently generated through the redox reaction with e- and h+. As photocatalysis normally takes place involving oxygen and water as the reactants, the reactive radicals converted by oxygen and water are thus called reactive oxygen species (ROS) [11].
Since ROS are the primary oxidants in photocatalytic environmental remediation, the generation and transformation of ROS is important to understand the photocatalysis mechanism, which improves the photocatalytic efficiency and facilitates practical applications. Four major ROS are commonly recognized, which include superoxide radical (•O2-), singlet oxygen (1O2), hydrogen peroxide (H2O2) and hydroxyl radical (•OH) [12-14]. In a typical photocatalysis reaction, since both reduction and oxidation occur simultaneously, ROS could be generated through the h+ induced stepwise oxidation path of H2O, alongwith the production of •OH, H2O2, •O2- and 1O2 [15, 16]. On the other hand, the stepwise reduction of O2 generates •O2-, H2O2 and •OH [17]. Among these, •OH and •O2- are recognized as the most stable radicals during the reaction, which can be effectively detected and are regarded as important reactants in the photocatalytic oxidation process.
In typical photocatalysts, their band structures, which possess wide bandgaps and suitable band edges to align the redox potential of O2/•O2- and OH-/•OH [18], normally provide a tolerated band range to produce both •O2- and •OH [19, 20]. Hence, the generation mechanism of ROS can be directly summarized as: O2 + e- → •O2- and OH- + h+ → •OH [21]. However, in some new types of two-dimensional photocatalysts such as graphite carbon nitride (g-C3N4) [22-26] and MoS2 [27-29], their valence band (VB) edge is not low enough to directly oxidize H2O to generate •OH [30, 31]. Thus, •OH are generated through the O2 reduction at the conduction band (CB), along the paths: O2 + e- → •O2-, •O2- + 2H+ + e- → H2O2 and H2O2 → 2•OH [32-34]. However, the rate-determining step in the reaction pathways has not been clarified yet. As a major ROS with high oxidation capacity, •OH manifests much more importance in photocatalytic remediation than •O2- [11, 35]. The generation mechanism of •OH is thus required to be further investigated. Moreover, understanding the conversion of ROS driven pollutants is highly desirable, in order to direct the design of photocatalysts toward more efficient environmental and energy applications.
In this study, we have synthesized g-C3N4 with a facile method based on the property that it exhibits photocatalytic activity toward NO removal. By a combined in situ diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) and density functional theory (DFT) method, it is determined that the H2O2 dissociation is the rate-determining step of •OH generation in g-C3N4and the reaction coordinates of ROS driven NO removal are presented. It is found that the oxidation capacity of g-C3N4 largely depends on the production and oxidation capacity of •OH, rather than •O2-. This article presents a research protocol to investigate the photocatalytic ROS generation and transformation, and clarifies the reaction mechanism of ROS driven NO removal. This approach could be extended to many environmental and catalytic applications.
All chemicals used in this study were analytical grade and no further treatment was required. The CN product was fabricated with a reported method [36]. In a typical synthesis procedure, 10 g thiourea was added to a crucible with 20 mL deionized water. The mixed solution was dried at 80 ℃ for recrystallization. Then, the precursor was placed in a semi-closed alumina crucible with a cover and roasted at 550 ℃ for 2 h with a heating rate of 15 ℃/min in static air. After the calcining treatment, the crucible was cooled to room temperature.
X-ray diffraction (XRD) was used to investigate the crystal phases with Cu Kα radiation (model D/max RA, Rigaku Co. Japan). Scanning electron microscopy (SEM, model JSM-6490, JEOL, Japan) and transmission electron microscopy (TEM, JEM-2010, Japan) were used to investigate the morphology. Fourier transform infrared (FT-IR) spectra were recorded on a Nicolet Nexus spectrometer (IRPrestige-21, SHIMADAZU, Japan) using samples embedded in KBr pellets. A scanning UV-vis spectrophotometer (TU-1901, China) was used to characterize the UV-vis diffuse reflectance spectrometry. The photoluminescence (PL) spectra were investigated using a fluorescence spectrophotometer (Edinburgh Instruments, FLSP-920) equipped with a Xe lamp with an excitation wavelength of 420 nm. Electron spin resonance (ESR) spectra of chemical radicals were obtained on a JES FA200 spectrometer to determine the involvement of the ROS in methanol dispersion for DMPO-•O2- and aqueous dispersion for DMPO-•OH, respectively.
The photocatalytic activity was investigated via the NO removal at 500 ppb in a self-designed continuous-flow reactor (Scheme S1 in the Supplementary Information). The rectangular reactor (30 cm × 15 cm × 10 cm) is made of polymeric glass and covered with Saint-Glass. The as-prepared sample (0.20 g) was dispersed onto two glass dishes for testing. A 150 W tungsten halogen lamp (0.16 W/cm2) was vertically placed above the reactor. After the adsorption-desorption equilibrium was achieved, the lamp was turned on. The NO gas was acquired from a compressed gas cylinder at a concentration of 100 ppm of NO (N2 balance). The initial concentration of NO was diluted to about 550 ppb by a zero air generator and the relative humidity (RH) level of the NO flow was controlled at 50% by passing zero air stream through a humidification chamber. The flow rates of the air stream and NO were controlled at 2.4 L/min and 24 mL/min, respectively. The NO removal ratio (η) was calculated as
where C and C0 refer to the NO concentration in the outlet stream and feeding stream, respectively. The resulting final product (nitrate) was removed by washing the dilute NaOH solution.
In situ DRIFTS measurements were performed with a TENSOR Ⅱ FT-IR spectrometer (Bruker) which was equipped with an in situ diffuse-reflectance cell (Harrick) and reaction chamber (HVC), as shown in Scheme S2. High-purity He, O2, and 100 ppm of NO were mixed and fed into the reaction system. A tri-way ball valve was used to switch between the target gas (NO) and purge gas (He). The flow rate of total gas was set to 100 mL/min, and the concentration of NO was adjusted to 50 ppm by dilution with O2. The chamber was enclosed in a dome having three windows, two for IR light entrance and detection, and one for illuminating the photocatalyst. The observation window was made of UV quartz and the IR windows were made of ZnSe. A Xe lamp (MVL-210, Japan) was used as the irradiating light source. Before the measurements, prepared samples were pretreated for 20 min at 300 ℃ to remove the adsorbed species on the catalyst surface. The figure of the reaction chamber under the working conditions is given in Fig. S1.
All the spin-polarized DFT-D2 calculations were performed with the "Vienna ab initio simulation package" (VASP 5.4.1 code) [37, 38], using a generalized gradient correlation functional with the exchange and correlation functional of the Perdew-Burke-Ernzerhof [39]. A plane-wave basis set with the cut-off energy at 450 eV within the framework of the projector-augmented wave method was used [40]. The Gaussian smearing width was set to the value of 0.2 eV. K points in the Brillouin zone were sampled with a 5 × 5 × 1 mesh. The van der Waals correction in the D2 calculations was set at the default parameters by the Grimme's method [41]. All atoms were allowed to relax at the convergence below 0.01 eV Å–1. A hybrid functional based on the Heyd-Scuseria-Ernzerhof (HSE06) method was applied to estimate the exact band structures [42]. The reaction coordinates were calculated with the climbing image nudged elastic band (CI-NEB) code, which located the minimum energy pathways (MEPs) from the initial state (IS) to its final state (FS) [43]. The transition state (TS) was verified with a single imaginary frequency [44]. A 2 × 2 × 3 supercell of g-C3N4 that includes 168 atoms was designed and relaxed. The lattice parameters were set to 14 × 16 × 20 Å with a vacuum region of 10 Å. The adsorption energy (Eads) is defined as
where Etot, Emol and ECN/OCN-K-CN depict the total energy of the adsorption complex, the isolated molecule and the CN/OCN-K-CN, respectively.
XRD was used to investigate the crystal structure of the as-prepared g-C3N4. As shown in Fig. 1(a), two characteristic peaks at 13.1° and 27.4° were prominent, which were assigned to the (100) and (002) facets of g-C3N4, respectively. This result reflects the generation of the in-plane repeating aromatic units and interlayer reflection in g-C3N4, which agrees with the published works [24]. The FT-IR spectrum was further studied to verify the basic substructures of g-C3N4 polymers. As shown in Fig. 1(b), the absorption bands in the 1200–1700 cm–1 range and the sharp peak at 810 cm–1 are both attributed to the stretching vibration of the C6N7 units [45], demonstrating the existence of basic atomic structure. Moreover, the absorption bands (3100–3500 and 890 cm–1) are maintained, which correspond to the N–H components stemming from uncondensed amino groups [46]. In addition, there is evidence of the stretching vibration of N=C=N (2212 cm–1). These results illustrate that the g-C3N4 photocatalyst has been successfully synthesized. The morphology and microstructure of g-C3N4 were examined by SEM (Fig. 1(c)) and TEM (Fig. 1(d)). These analyses showed that the typical layered nanosheets were distinctly perceived.
The PL spectrum was then studied to investigate the light response and charge dynamics (Fig. 2(a)). It was found that g-C3N4 possessed light absorption and charge separation properties in the visible light region (λ ≥ 420 nm), which manifests great potential for ROS generation and NO removal under visible-light irradiation. The light absorption properties were measured by using a combined experimental and theoretical method. As depicted by the UV-Vis DRS spectrum (Fig. 2(b)), the light harvesting capacity of g-C3N4 in the visible light region is authenticated, confirming the PL result. The calculated absorption spectrum also supports the experimental result. The band energy was measured from the intercept of tangents to the plot of (αhv)1/2 vs. photoenergy (Fig. 2(b), inset). It was estimated that the energy gap (Eg) of g-C3N4 was ca. 2.39 eV. We then calculated the density of states (DOS) using a hybrid DFT method, measuring the Eg to be 2.33 eV, which is in good accordance with the experimental value. These theoretical results reveal that the constructed DFT model of g-C3N4 (Fig. S2) agrees with the experimental prepared sample.
The energies of the valence band (EVB) and conduction band (ECB) are subsequently calculated, to investigate the exact band structure of g-C3N4 (Fig. 2(d)), by the following equations:
where X and Ee represent the electronegativity and energy of free electrons in hydrogen atoms (about 4.5 eV), respectively. Eg is the gap energy calculated from the UV-Vis DRS spectrum.
Based on the band structure of g-C3N4, it is apparent that •O2– can be directly produced via O2 reduction at the VB, while the CB position is not low enough to oxidize H2O for •OH generation. As the photo-oxidation capacity mainly originates from ROS, it is required that the generation and transformation mechanism of ROS in g-C3N4 should be further investigated.
The DMPO spin-trapping ESR experiment is utilized in methanol dispersion for DMPO-•O2– (Fig. 3(a)) and aqueous dispersion for DMPO-•OH (Fig. 3(b)), respectively. As expected, the signals of •O2– and •OH are detected and increase as the irradiation time progresses, which indicates that both •O2– and •OH can be produced in g-C3N4. Based on the band structure and ESR results, the generation and transformation pathways of ROS can be summarized as follows:
As stated before, •OH possesses higher oxidation capacity than •O2- and cannot be directly generated in g-C3N4. In order to investigate the transformation mechanism of ROS, we subsequently applied DFT calculations to estimate the reaction coordinates at the atomic level. As Fig. 3(c) shows (Refer Fig. S3 for the detailed minimum reaction pathways), energy and charge transfer took place once the isolated O2 molecule approached the g-C3N4, which leads to a significant energy release. The energy-favorable adsorption and activation of O2 contributes to the formation of •O2- (Path 1), which agrees with the ESR result. Moreover, •O2- is further reduced to generate •O22–, combining with two hydrogen ions (H+) to generate H2O2 (Path 2) and dissociation of the O–O bond in H2O2 generates •OH (Path 3). Notably, the dissociation required 0.48 eV to materialize (TS3), making it the rate-determining step in •OH generation in g-C3N4. Hence, tailoring the rate-determining step and enhancing the production of •OH are of key importance in the design of g-C3N4 based photocatalysts for environmental remediation.
The photocatalytic performance of the as-prepared g-C3N4 was evaluated for NO removal under visible-light irradiation (λ > 420 nm). As shown in Fig. 4(a), the maximum NO removal ratio (28%) was reached in approximately 5 min. We further carried out a five-time cycling test and after the first two runs, some decrease in activity was observed. But the removal ratio remained constant for the five cycles (19%), which indicates that NO can be removed by photocatalysis of g-C3N4.
In order to understand the reaction process and clarify the mechanism in detail, in situ DRIFTS was performed to measure the dynamic reaction intermediates and final products in a time sequence. The baseline was measured before the NO injection into the reaction chamber. In dark conditions, the absorption bands of NO appear once the gas flow contacts g-C3N4 (Fig. 5(a)). The absorption bands of NO (2198 cm–1), N2O (2285 cm–1) and NO2 (2067 cm–1) were prominently detected, which is assigned to the chemical adsorption of NO in g-C3N4 [47, 48]. As the adsorption time increases, the intensities of these peaks gradually increased, indicating the accumulation of reactants and reaction intermediates on the catalyst surface. It is notable that some new absorption bands were developed, namely NO2– (1153, 1133, 1088, 890 and 877 cm–1) and NO3– (1011 cm–1) [49, 50]. The accumulation of these reaction products is presumed to arise from the two-coordinated pyridine-like N atoms on g-C3N4, which possess an extra lone-pair of electrons to facilitate the NO oxidation in dark conditions [51].
After the adsorption equilibrium (Fig. 5(b)) is reached, the in situ DRIFT was subsequently recorded under visible-light irradiation. As shown in Fig. 5(b), once the light was turned on, the absorption bands of the reactants, NO and N2O, are no longer detected, which illustrates that NO could be effectively oxidized in g-C3N4. Moreover, the reaction intermediates (NO2, 2070 and 933 cm–1) are also detected, which restrain the over-all NO photocatalytic removal and hamper the generation of the final products. Over time, the absorption bands of the reaction products, NO2– (1100, 1000, 883 and 866 cm–1) and NO3– (1044 and 984 cm–1 are gradually intensified) [52]. All the observed DRIFTS bands of the adsorbed species are listed in Tables S1 and S2. As the ROS are a major source of the oxidation capacity, the primary reaction pathways are proposed as follows:
In order to elucidate the primary reaction mechanism that is proposed by the in situ DRIFTS, the ROS driven reaction coordinates were thus calculated using the CI-NEB method with a single imaginary frequency (f/i, Table 1). As shown in Fig. 6, •O2– driven NO oxidation was composed of two elementary reactions (Paths 4 and 5, Fig. 6(a)–6(c)). It was found that path 4 required an activation energy (Ea) of 0.47 eV in order to transfer NO to NO2–. However, the oxidation of NO2 (Path 5) is more difficult as the activation energy for climbing over the TS5 is as high as 0.66 eV, leading to a lower reaction rate than that in Path 4. These results confirm our experimental results that NO2 oxidation is the rate-determining step in photocatalytic NO removal.
•OH driven reactions were further conducted (Fig. 6(d)–6(f)), which manifest similar patterns with •O2–. The Ea for NO2 oxidation (Path 7, 0.46 eV) is higher than that for NO oxidation (Path 6, 0.27 eV). Specifically, as the rate-determining step, NO2 is more likely to be oxidized by •OH (0.46 eV) than •O2– (0.66 eV). Based on these results, it can be concluded that the key to enhancing the photocatalytic efficiency of NO removal in g-C3N4 is to facilitate the ROS transformation from •O2– to •OH.
In summary, we have proposed a combined experimental and theoretical approach to investigate the generation and transformation of ROS in g-C3N4 for effective photocatalytic NO removal. This work provided direct evidence that the generation of •OH in g-C3N4 arises from the O2 reduction at CB, and the H2O2 dissociation is the rate-determining step for •OH formation. Moreover, the reaction mechanism of NO oxidation is shown by in situ DRIFTS and DFT calculations. It certified that NO2 oxidation was the rate-determining step, which suppresses the over-all NO removal and leads to the accumulation of toxic intermediates. In order to overcome the rather high activation energy, the production of •OH requires to be enhanced. This work reveals that understanding and tailoring the generation and transformation of ROS is crucial for elevating the performance of photocatalysts in environmental and energy-related applications.
2018,
Vol. 39 Issue (10): 1695-1703 
PDF
2018,
Vol. 39 Issue (10): 1695-1703 PDF