Mass-Controlled Direct Synthesis of Graphene-like Carbon Nitride Nanosheets with Exceptional High Visible Light Activity. Less is Better

In the present work, it is very surprising to find that the precursors mass, a long overlooked factor for synthesis of 2D g-C3N4, exerts unexpected impact on g-C3N4 fabrication. The nanoarchitecture and photocatalytic capability of g-C3N4 can be well-tailored only by altering the precursors mass. As thiourea mass decreases, thin g-C3N4 nanosheets with higher surface area, elevated conduction band position and enhanced photocatalytic capability was triumphantly achieved. The optimized 2D g-C3N4 (CN-2T) exhibited exceptional high photocatalytic performance with a NO removal ratio of 48.3%, superior to that of BiOBr (21.3%), (BiO)2CO3 (18.6%) and Au/(BiO)2CO3 (33.8%). The excellent activity of CN-2T can be ascribed to the co-contribution of enlarged surface areas, strengthened electron-hole separation efficiency, enhanced electrons reduction capability and prolonged charge carriers lifetime. The DMPO ESR-spin trapping and hole trapping results demonstrate that the superoxide radicals (•O2−) and photogenerated holes are the main reactive species, while hydroxyl radicals (•OH) play a minor role in photocatalysis reaction. By monitoring the reaction intermediate and active species, the reaction mechanism for photocatalytic oxidation of NO by g-C3N4 was proposed. This strategy is novel and facile, which could stimulate numerous attentions in development of high-performance g-C3N4 based functional nanomaterials.

the conjugated aromatic units of CN 33 . A typical (100) diffraction peak around 13.0° can be indexed to (100) peak of graphitic materials, in accordance of the in-plane tri-s-triazine units which formed one-dimension (1D) melon strands 34,35 . Further observation on an enlarged view of (002) peak ( Fig. 1b) demonstrates that the diffraction angle 2θ of (002) peak gradually increases, which can be ascribed to the more complete condensation of thiourea when the precursor mass was decreased. The typical (002) peak is 27.20° for CN-20T, 27.28° for CN-10T, 27.35° for CN-5T and 27.42° for CN-2T respectively, indicating the interplanar distance tends to decrease. The reduced interplanar distance of CN-2T will shorten the charge transport distance from the bulk to the surface, beneficial for the final products separated from the photocatalysts, which is highly desirable for photocatalytic applications 36,37 . Figure 2 depicted that the nitrogen adsorption-desorption isotherms and Barrett-Joyner-Halenda (BJH) pore-size distribution of as-prepared samples. All samples exhibit nanoporous architecture, as reflected in Fig. 2a,b. The nitrogen adsorption-desorption isotherms (Fig. 2a) for all samples are assigned to type IV (Brunauer, Deming, Deming, and Teller, BDDT classification) suggesting the presence of mesopores (2-50 nm) 37,38 . As the thiourea mass decreased, the hysteresis loops shift to the region of lower relative pressure accompanied by the enhanced hysteresis loops areas, indicating the formation of enlarged mesopores architectures 37 . The hysteresis loops assigned to type H 3 indicate the formation of slit-shaped pores from the aggregates of plate-like particles. Notably, the apparent enlargement of surface areas and mesopores volumes can be observed when diminished the precursors mass ( Fig. 2b and Table 1). These results are well in agreement with the sheet-like morphology of g-C 3 N 4 presented in SEM (Fig. 3). Besides, the correlations among surface area, pore volume, and precursor mass of g-C 3 N 4 samples are summarized in Table 1. The S BET was increased from 12 m 2 /g (CN-20T) to 66 m 2 /g (CN-2T), as well as the pore volume enlarged from 0.083 cm 3 /g (CN-20T) to and 0.39 cm 3 /g (CN-2T), when the precursor mass was diminished from 20 to 2 g. In addition, transformations of peak pore-size distribution (PSD) (Fig. 2b) of g-C 3 N 4 corresponding to diverse precursor masses further confirm the introduction of mesopores architecture. The peak intensity of mesopores with diameters about 30.0 nm gradually increased as   Table 1. The S BET , pore volume, peak diameter, band gap value, NO removal ratio and NO 2 fraction of g-C 3 N 4 samples as well as that of the references. The data for BiOI, BiOBr, (BiO) 2 CO 3 and Au/(BiO) 2 CO 3 were collected from related references.
the precursor mass was decreased. Especially, the PSD curve of CN-2T is quite broad (from 2 to 100 nm) with small mesopores (~2.7 and ~3.7 nm) and large mesopores (~30.8 nm). The small mesopores can be assigned to the porosity within the nanoscale sheets (Fig. 3a), and the large mesopores assigning to the porosity between packed layers. Less is better. The high surface area and pore volume will be beneficial for augmenting the number of active sites and accelerating the transfer of the intermediates and final products, which could exhibit higher photocatalytic capability 16 .
The typical TEM images of the as-prepared samples are shown in Fig. 3. It demonstrates that all as-synthesized samples obtained from diverse precursors are composed of irregularly curved layers. The nanosheets become thinner and the architecture turn fluffier, as the precursor mass gradually decreased. Figure 3a reflects that the CN-20T sample is composed of tightly overlapped bulk layers, further demonstrated by the magnifying view (Fig. 3b). When the precursor mass is decreased to 10 g, the structure of CN-10T turns fluffy and the nanosheets becomes thinner, producing nori-like structure (Fig. 3c,d). Besides, as the precursor mass is decreased to 5 g, the thickness of the nanosheets will further be reduced, and large amount of smooth nanosheets overlap irregularly together forming tremella-like layers (Fig. 3e,f). Unexpectedly, when the masses of the thiourea are further reduced to 2 g, thickness of CN-2T nanosheets was nearly reduced to several nanometers scale, resulting in extremely fluffy and transparent architecture (Fig. 3g,h).
Thus, a conclusion can be drawn that in a semi-closed system, less precursor mass contributes to g-C 3 N 4 with smaller 2D nanosheets, unique porous architecture and higher specific surface area. However, no solid was remained in crucible as thiourea mass was less than 2 g. On the basis of observation, we proposed a reaction mechanism to clearly elucidate the reaction processes during the pyrolysis of thiourea in a semi-closed system. In converting thiourea into g-C 3 N 4 , the synergistic effects of pyrolysis-generated self-supporting atmosphere 36 and air molecules (mainly nitrogen and oxygen) exert great significance in tailoring the microstructure and properties of g-C 3 N 4 . In semi-closed system, the retained nitrogen and oxygen molecule can be well-distributed in the pyrolysis-generated self-supporting atmosphere 39 . Fewer precursors can produce less self-supporting atmosphere. Therefore more sufficient nitrogen and oxygen concentration in the reaction system can generate an array of gas bubbles, splitting when it grows up, in resulting in stripping g-C 3 N 4 layers into smaller layers, as well as forming more fluffy structure (Fig. 3) 37 . The thickness of g-C 3 N 4 samples will be remarkably reduced by the layer-by-layer exfoliation process 12 in semi-closed system with less precursor masses. Moreover, the fewer precursors in semi-closed system, the more space can be used for self-supporting atmosphere growing and diffusion, which can be also a contribution to fluffy structure of g-C 3 N 4 .
However, for less than 2 g thiourea, the scarce pyrolysis-generated self-supporting atmosphere will be further diluted to lower concentration. These dilute pyrolysis-generated gases quickly escaped the system without further reaction, and the products would be decomposed at high temperature, leading to no solids residue. The typical AFM images and thickness analysis ( Fig. 4) reflect that the nanosheets thickness of g-C 3 N 4 is apparently decreased for about 15 nm for CN-20T, 10 nm for CN-10T, 5 nm for CN-5T to 2 nm for CN-2T. This result further demonstrates that the precursor amount is a pivotal factor in controlling the thickness of g-C 3 N 4 . Decreasing the precursors mass is favorable for generation of thinner g-C 3 N 4 nanosheets. Figure S1 in the Supporting Information shows the FT-IR spectra of CN-2T, CN-5T, CN-10T and CN-20T. We can observe that the weak absorption at 700-800 cm −1 region are assigned to the bending vibration mode of CN heterocycles, and the characteristic out of plane bending vibration mode of the triazine units at 810 cm −1 are found for all the samples 37 . The absorption bands in the range of 1200-1600 cm −1 are assigned to stretching mode of C-N heterocycles, and the broad bands in the range of the 3000-3700 cm −1 region are attributed to the adsorbed H 2 O molecules and N-H vibration 39 . For fewer samples, the polymerization reaction of thiourea can be more complete resulting in stronger vibration of the C 6 N 7 units, consistent with XRD results. In addition, absorption bands assigned to sulfur bond (such as -SH, -SN, -SC) have not been detected, indicating that the sulfur element in thiourea of CN-20T, CN-10T, CN-5T and CN-2T is completely released during thermal treatment. From XPS spectra, the final ratio between C and N for CN-20T, CN-10T, CN-5T, CN-2T ranges from 0.746 to 0.752, close to the theoretical ratio of C and N in g-C 3 N 4 , which indicates that the as-prepared samples have high purity.

Chemical composition.
The XPS measurements were carried out to determine the chemical state of the elements for all samples. Figure 5a represents the full survey spectra of C, N and O elements for CN-2T, CN-5T, CN-10T, and CN-20T. Peaks assigned to S species cannot be observed for each sample, further indicating that the sulfur in thiourea was completely released during heating treatment. The C1s spectra ( Fig. 5b) clearly reflect the similar four peaks for all the obtained samples. The peak located at 284.8 eV relates to adventitious carbon species, and the other three peaks at 286.8, 288.3 and 293.7 eV, corresponding to (C) 3 -N, C-N-C and N-C-O coordination in the g-C 3 N 4 lattice, respectively 40 . The N1s region (Fig. 5c) can be fitted into four peaks, which can be ascribed to C− N− C (398.8 eV), tertiary nitrogen N-(C) 3 (400.6 eV), and π -excitations (404.5 eV), respectively 41 . As shown in Fig. 5d, the O1s spectra of all g-C 3 N 4 samples can be clearly divided into three peaks with binding energies of 531.7, 533.1 and 534 eV. The major peak at 533.1 eV is assigned to C-N-O, assign to partial oxidation of the low polymerized CN during thermal exfoliation in air 40 . The other two peaks at 531.7 and 534 eV are assigned to surface -OH groups and adsorbed H 2 O, which is consistent with FT-IR observation 40 .
Optical properties and band structure. The optical properties of CN-20T, CN-10T, CN-5T, CN-2T samples were investigated by UV− vis DRS (Fig. 6a). As demonstrated in Fig. 6a, the optical absorption spectra shows that all samples feature an intrinsic semiconductor absorption in the blue region of the visible spectra, corresponding to band gap transitions from valence band to conduction band. The absorption edges of g-C 3 N 4 samples change with the variation of precursor masses. Figure 6b demonstrates the band gap energy, estimated from the intercept of the tangents to the plots of (α hν ) 1/2 vs photon energy. The band gap energy of the samples increase from 2.37 eV for CN-20T, 2.38 for CN-10T, 2.46 for CN-5T, to 2.50 for CN-2T depicted in Fig. 6a,b. The hypsochromic shift of the absorption edges reflects the quantum confinement effects of the thermally induced thinner nanosheet structure as is evidenced in TEM (Fig. 3) and AFM (Fig. 4), which is assigned that the size and thickness of g-C 3 N 4 are in nanoscale region 42 . The enlargement in band gap could subsequently boost the photocatalytic redox ability. To investigate the electronic structures and reduction ability of the electrons, we measure VB XPS (Fig. 7) of the as-synthesized samples. The VB XPS reflects that the valence band maximums (VBM) of all samples are the same (1.58 eV), which demonstrates that photo-oxidation abilities of the holes for all samples are equal. Figure 8 illustrates the valence band maximum (VBM) and conduction band minimum (CBM) potentials of CN-2T, CN-5T, CN-10T and CN-20T, as the band structures perform a pivotal role in the photocatalytic performance, as well-known. Generally speaking, the more positive the value of VB, the higher the mobility of holes produced, along with the better the photo-oxidation efficacy of holes. As showed in Fig. 8, the VBM potentials of all samples are the same, demonstrating that photo-oxidation abilities of the holes for all samples are equal. Nevertheless, the CBM potentials shift to more negative values. The CBM potentials values are decreased from − 0.79, − 0.80, − 0.88 to − 0.92 eV, which is associated with the degree of polymerization of π -conjugated polymeric network in fewer precursors. In general, the more negative the value of CB, the stronger reductive power of the photo-excited electrons, contributing to strengthen the photo-driven electron-holes separation efficiency due to charge carrier good transport ability 43 . PL spectra have been widely applied to investigate the charge carrier trapping, migration, and transfer of electron-hole pairs in semiconductors. PL emission results from the recombination of electron-hole pairs. Figure 9 depicts the room temperature PL spectra of g-C 3 N 4 obtained from different mass of precursors. Emission peaks centered at around 440-455 nm for all samples originate from direct band-to-band transitions. It is remarkable that CN-2T sample exhibits the highest PL emission peak, which is associated with its less structural imperfection on account of more complete condensation of thiourea, as evidenced by XRD and FT-IR. Note that more complete condensation of thiourea minimized the number of structural defects (e.g., uncondensed −NH 2 , −NH groups), which could capture the electrons or holes, hence resulting in high radiative PL emission. Figure 9 also reflects that the PL peak   (Fig. 6b), which is ascribed to quantum confinement effect 44 .
To understand the photophysical characteristics of photo-excited charge carriers, the ns-level time-resolved fluorescence decay spectra of CN-2T, CN-5T, CN-10T, and CN-20T were recorded, as   shown in Fig. 10(a-d). By fitting the decay spectra, the radiative lifetimes with different percentages can be determined ( Table 2). As the precursors masses decrease, the short lifetime of the samples is prolonged from 1.8 ns for CN-20T, 1.9 ns for CN-10T and CN-5T to 2.0 ns for CN-2T. Moreover, the long lifetime of charge carriers is increased from 9.5 ns for CN-20T to 10.4 ns for CN-2T. The prolonged radiative lifetime of the charge carriers is really crucial in improving the probability of their involvement in photocatalytic reaction before recombination. The increased lifetime of charge carriers is associated with the more negative conduct band potentials and enhanced charge transfer induced by CN-2T. This prolonged lifetime would afford to improve the probability of electrons or holes captured by reactive substrates to initiate the photocatalytic reactions.
Photocatalytic activity, stability and mechanism. The photocatalytic activity of the as-prepared g-C 3 N 4 samples was evaluated by removing gaseous NO (concentration: 600 ppb) under visible-light  irradiation in a continuous reactor for air purification. Figure 11a depicts the variation of NO concentration (C/C 0 %) with irradiation time over g-C 3 N 4 samples treated form different precursor masses. Previous investigation indicated that NO could not be degraded without photocatalyst under light irradiation or with photocatalyst for lack of light irradiation 33,45,46 . When the g-C 3 N 4 is irradiated in the absence of NO gas, no products can be detected, which indicates that the products (NO 2 , NO 3 − ) are originated from the photocatalysis of g-C 3 N 4 . In the presence of photocatalyst, the photo-generated reactive radicals react with NO, converting NO to the final product of HNO 3 40 . Figure 11a reveals that the all g-C 3 N 4 samples of CN-2T, CN-5T, CN-10T and CN-20T demonstrate visible light photocatalytic activity toward NO removal, in accordance with the facts that g-C 3 N 4 has a suitable band gap that can be directly excited by visible light. Figure 11a also shows that the NO concentration for all samples decreased rapidly in 5 min. The slight decrease in activity during 5 to 15 min can be ascribed the gradual generation of reaction intermediates that may occupy the active sites of photocatalysts. When the reaction reached equilibrium, the photocatalytic activity was kept almost unchanged. Moreover, the NO removal ratio of g-C 3 N 4 samples regularly increases from 12.7, 25.5, 33.6 to 48.3% for CN-20T, CN-10T, CN-5T and CN-2T after 30 min irradiation as a equilibrium. The optimized activity of CN-2T samples sharply outperforms that of BiOBr (21.3%) 47 , as well as the (BiO) 2 CO 3 (18.6%) and Au/(BiO) 2 CO 3 (33.8%) 48 , demonstrating that diminishing the amount of precursor is an effective strategy to enhance the photocatalytic activity of g-C 3 N 4 .
To better understand the reaction kinetics of the NO degradation catalyzed by g-C 3 N 4 photocatalysts, the experimental data were fitted by a first-order model, according to the true that the value of the rate constant k app commonly gives an indication of the capability of the nanocomposites photocatalyst. Figure 11b gives the values of the rate constants k app of all g-C 3 N 4 samples. CN-2T reflects the highest apparent k app of 1.13 min −1 , which was about 2.7 times as that of CN-20T sample (0.42 min −1 ).
The reaction intermediate of NO 2 is monitored online during photocatalytic oxidation of NO presented in Fig. 11c. Photocatalytic oxidation of NO to NO 2 is not beneficial for application as NO 2 is more toxic. To promote real application, NO conversion to NO 2 should be inhibited. The NO should be oxidized to final product of NO 3 − . As we can see in Fig. 11c, 42.2% of NO is conversed to NO 2 for CN-20T sample and the selectivity of NO to final NO 3 − is 57.8%. This is not so good for application. However, the conversion ratio of NO to NO 2 is decreased to 20.0% for CN-2T and the selectivity of NO to final NO 3 − is increased to 80.0%, which is favorable for application. This enhanced NO conversion ratio over CN-2T can be ascribed to the more negative conduction band of the graphene-like g-C 3 N 4 . The final oxidation products (nitric acid or nitrate ions) can be simply washed away by water wash. Note that the NO concentration in the outlet was decreased gradually when the photocatalytic reaction was going on, due to the continual conversion of NO to NO 3− . What is more, the NO concentration would reach minima till the photocatalytic reaction reached equilibrium. The slight rising of NO concentration was due to the accumulation of NO 3− product on the catalyst surface, occupying the active sites 45,49,50 . After long-term irradiation, the NO concentration in the outlet would reach a steady state. The stability of the optimized photocatalyst was tested in an extended experiment (Fig. 11d). The high capability was reproducible, and the material showed excellent stability during 5 times cycle testing. In addition, we also measured XRD spectra of the CN-2T samples after stability testing, and the result reflected that the architecture of the optimized samples is firmly stable (Fig. S2).
To elucidate the main reactive species responsible for the NO removal reaction, ESR technique was employed for CN-2T in reaction systems. Figure 12 shows ESR spectra measured as the effect of light irradiation on CN-2T photocatalyst at room temperature in air. DMPO (5,5-dimethyl-1-pyrroline N-oxide) is nitrone spin trap generally used for trapping radicals due to the generation of stable free radicals, DMPO-•OH or DMPO-•O 2  (1-4)), reflected by the signal of the origination and strength of the two radicals. Moreover, judging from the electrode potential (Eϕ ), the VB holes of g-C 3 N 4 sample might also oxidize NO because the Eϕ VB (about 1.58 V vs. NHE) of g-C 3 N 4 is more positive than Eϕ (NO 2 /NO, 1.03 V vs. NHE), Eϕ (HNO 2 /NO, 0.99 V vs. NHE), and Eϕ (HNO 3 /NO, 0.94 V vs. NHE) 51 . The holes as active species are further confirmed by trapping experiment as shown in Fig. S3. When the typical hole trapping agent (KI, 1 wt.%) is mixed with the CN-2T sample, the photocatalytic activity is decreased rapidly, indicating photo-generated holes are one of the active species for g-C 3 N 4 photocatalysis. Based on aforementioned discussion, the reaction mechanism of photocatalytic oxidation of NO by g-C 3 N 4 is proposed as presented in Eqs. (1)(2)(3)(4)(5)(6)(7)(8).  Above all, the exceptional high photocatalytic activities of CN-2T can be explained as the co-contribution of the 2D nanosheets-like architectures, reduced interplanar distance, higher surface area, unique porous architecture, elevated CBM, enhanced electron-holes separation efficiency, as well as prolonged lifetime of charge carriers. Firstly, for the g-C 3 N 4 sample treated at smaller mass of thiourea, the nanosheet-like structure (Fig. 3a) and reduced interplanar distance (Fig. 1b) can facilitate the photo-induced electrons and holes transportation, thus lowering the recombination rates of the photo-driving hole-electron pairs 16 . Secondly, the nanosheets with high surface area (Fig. 3) can advance the pollutant adsorption 52 and provide more active sites for intermediates diffusion 53 . Thirdly, numerous large mesopores provides more spaces for quick diffusion of reactants and intermediates 54 . Pivotally, the promotion of the conduct band minimum could largely expedite the reduction capability of electrons contributing in more •O 2 − species generation, as well as enhancing the photo-driving hole-electron separation and prolonging the lifetime of charge carriers 12 . All these favorable factors co-contribute to the exceptional high photocatalytic performances of CN-2T.
Beyond expectation, by facilely tailoring thiourea mass, we have successfully engineered the nanostructures of g-C 3 N 4 with exceptional high visible light photocatalytic performance. This strategy not only provides a facile method for optimizing g-C 3 N 4 with superior photocatalytic capacity, but also a guidance for fabricating multi-functional materials in areas such as solar energy conversion, photosynthesis, and catalyst support. We have done initial experiments on other g-C 3 N 4 precursors, such as dicyandiamide and urea. For these precursors, the results indicate that less amount of precursor is also beneficial for enhancement of photocatalytic activity of g-C 3 N 4 , which implies that the present method is general for synthesis of graphene-like g-C 3 N 4 .

Conclusion
The precursor mass, a critical factor governing the micro-architecture of g-C 3 N 4 , has long been overlooked. Considering the unique effects of precursor mass, we developed a novel strategy for direct production of graphitic carbon nitride, with excellent photocatalytic capability for NO purification under visible-light illumination. This unique effects of precursor mass on the microstructure and photocatalytic activity of g-C 3 N 4 were firstly revealed. It was surprising to find that the morphology and the photocatalytic activity of g-C 3 N 4 were highly dependent on the precursor masses. A new concept for nanomaterials synthesis was proposed: Less is better, and meanwhile this founding can offer a guidance for obtaining g-C 3 N 4 with excellent performance. When diminishing thiourea mass from 20 to 2 g, the visible light photocatalytic capability of 2D g-C 3 N 4 nanosheets toward gaseous NO purification was conspicuously enhanced. The exceptional high visible-light activity of CN-2T can be ascribed to the contribution of accelerated electrons transportations, enhanced electron-hole separation efficiency, strengthened electrons reduction capability and prolonged charge carriers lifetime. This strategy is novel, environmental benign, easily-available without the aid of co-factors, which will stimulate extensive attentions in environmental and energetic domain.

Methods
Synthesis of g-C 3 N 4 . All reagents used in this study were analytical grade. In a typical synthesis, thiourea powder of different masses (2, 5, 10 and 20 g) were put into four different alumina crucibles (50 ml in volume) with a cover, respectively, and then heated to 550 °C at a heating rate of 15 °C/min in a muffle furnace for 2 h. After the reaction, the alumina crucible was cooled to room temperature. The resultant g-C 3 N 4 were collected and ground into fine powders. The g-C 3 N 4 samples prepared from different precursor masses (2, 5, 10 and 20 g) were labelled as CN-2T, CN-5T, CN-10T and CN-20T, respectively.
Characterization. The crystal structures were analyzed by X-ray diffraction with Cu-Kα radiation (XRD: model D/max RA, Japan). The morphological structure was analyzed by transmission electron microscopy (TEM: JEM-2010, Japan). The UV-vis diffuse reflection spectra were measured through a Scan UV-vis spectrophotometer (UV-vis DRS: UV-2450, shimadzu, Japan) with an integrating sphere assembly. The 100% BaSO 4 was chose as reflectance sample. The atomic force microscopy (AFM) study in the present work was performed by means of MultiMo de-V (Veeco Metrology, Inc.). FT-IR spectra were measured with a Nicolet Nexus spectrometer embedded in KBr pellets. The Fluorescence spectrophotometer (FS-2500, Japan) was used to investigate the photoluminescence spectra taking Xe lamp as the excitation source with optical filters. The surface properties and the total density of the state (DOS) distribution of the valence band were recorded by the X-ray photoelectron spectroscopy with 150 W Al-Kα X-ray radiation (XPS: Thermo ESCALAB 250, USA). The shift of the binding energy was calibrated using an internal standard of C1s level of 284.8 eV. The specific surface areas were measured through the nitrogen adsorption-desorption on a nitrogen adsorption apparatus (ASAP 2020, USA). The samples were degassed at 150 °C before measurement. The FLsp920 Fluorescence spectrometer (Edinburgh Instruments) was utilized to obtain time-resolved photoluminescence spectra with the excitation at 420 nm. The ESR measurement (FLsp920, England) was recorded by mixing g-C 3 N 4 in a 50 mM DMPO Visible light photocatalytic DeNOx activity. The photocatalytic capability of ppb-level NO purification was investigated in a continuous flow reactor. The rectangular reactor with a volume capacity of 4.5 L (30 cm × 15 cm × 10 cm) was manufactured with stainless steel and covered with Saint-Glass. A 150 W commercial tungsten halogen lamp was vertically placed above the reactor. The UV beam was cut off by adopting a UV cut-off filter (420 nm). A certain amount of photocatalyst (0.2 g) was coated onto two dishes (12.0 cm in diameter). Then the coated dish was placed at 70 °C in a drying oven to remove water of the suspension. NO gas was supplied by a compressed gas cylinder at a concentration of 100 ppm (N 2 balance). The NO concentration of 600 ppb was acquired by the air stream dilution. The desired relative humidity (RH) level (50%) of the NO flow was controlled by passing the zero air streams through a humidification chamber. The gas streams were pre-mixed completely through a gas blender. Mass flow controllers were utilized to control the flow rate at 2.4 L/min. When the adsorption-desorption equilibrium reached, the lamp was turned on. The concentration of NO was continuously measured by a chemiluminescence NOx analyzer (Thermo Environmental Instruments Inc., 42i-TL). The removal ratio (η) of NO was determined with η (%) = (1 − C/C 0 ) × 100% (1), where C and C 0 are the concentrations of NO in the outlet steam and the feeding stream, respectively. The photocatalytic kinetics of NO purification is a pseudo-first-order reaction at low NO concentration as ln(C 0 /C) = k app t, where k app stands for the apparent rate constant.