Nitrogen is a necessary element of human, animal and plant growth. Although constituting~78% of the atmosphere, nitrogen, in molecular form, is unusable to most organisms because of its strong nonpolar N≡N covalent triple bond. Thus, artificial nitrogen fixation is carried out through the Haber-Bosch process, in which hydrogen gas reacts with nitrogen gas to yield ammonia in the presence of catalysts under high pressure and temperature. Both the energy consumption and raw material costs are high for this process. Therefore, artificial nitrogen fixation under milder conditions has been a chemical issue of great significance, since the reduction in the input energy during the fixation process and no use of hydrogen from a natural gas may be preferred from the viewpoints of cost and environmental preservation. This significance has directed many chemists to find chemical1, 2, electrochemical3, 4 and photochemical routes5, 6 to fix nitrogen under mild conditions.
Nitrogen photofixation technology is considered to be a promising method to replace the traditional Haber-Bosch process. In 1977, Schrauzer et al.7 first reported that N2 can be reduced to NH3 over Fe doped TiO2 under UV light. Since then, many Tibased metal oxides and composite catalyst were reported8-12. Ranjit et al.8 investigated the nitrogen photofixation activity of precious metals modified TiO2, and found that Ru modified TiO2 exhibited the best photofixation performance. A linear relationship observed between the concentration of NH4+ and the strength of metal-H bond. Rusina et al.9 investigated the N2 photofixation performance using Fe2Ti2O7 as catalyst and ethanol as the hole-trapping agent. Hoshino et al.10, 11 prepared conducting polymer/TiO2 hybrid material for nitrogen photofixation under white light. The main product is NH4ClO4. Zhao et al.12 prepared Fe-doped TiO2 nanoparticles with highly exposed (101) facets by two-step hydrothermal method. They found the quantum yields of nitrogen photofixation depend on the partial pressure of nitrogen in the reaction. However, because of the poor visible light absorption caused by the wide band gap energy, the nitrogen fixation ability of the Ti-based metal oxides and composite catalyst is still low under visible light. Moreover, as most photoexcited electrons tend to recombine with their twinborn holes, rather than to be captured by the adsorbed N2, the interfacial charge transfer efficiency of these semiconductor photocatalysts is far from satisfactory13-15. Besides that, compared with the photocatalytic H2 evolution and CO2 reduction, photocatalytic N2 fixation is more challenging because the N2 fixation is seriously hampered by the high-energy N2 intermediates in the reduced or protonated form (N2- or N2H)16. These disadvantages limit the development and practical use of photocatalytic N2 fixation. Designing new photocatalysts is not only important but also a challenge in the promotion of the development of photocatalytic N2 fixation.
Recently, graphitic carbon nitride (g-C3N4) has been widely applied in a variety of fields, including photocatalysis17, 18, fuel cells19, 20, organic synthesis21, and gas storage22, 23. The versatile application of g-C3N4 is largely due to its unique physicochemical properties, such as moderate band gap energy, energy-storage capacity, special optical properties and gas-adsorption capacity. Ionic liquids (ILs), regarded as designer solvents, have been extensively investigated in recent years. They show great promise in organic synthesis, catalysis, separation and polymerization24-26. Their favorable properties, such as thermal stability, negligible vapor pressure, high ionic conductivity and wide electrochemical window, make them attractive as reaction media and solvents. The combination of ionic liquids with nanotechnology has led to major advances in materials science. Nanorods, nanospheres, nanotubes, and mesostructures of semiconductor materials have been synthesized using ILs as solvent, electrolyte and template27-30. More recently, ILs have also been used for synthesizing carbon nitridebased semiconductor materials. Xu et al.31 prepared graphite-like C3N4 hybridized α-Fe2O3 (g-C3N4/α-Fe2O3) hollow microspheres. It is found that ionic liquid 1-butyl-3-methylimidazolium tetrachlorideferrate (Ⅲ) [Bmim]FeCl4 is supposed to have the triple roles of reactant, dispersing media and template at the same time. Di et al.32 prepared g-C3N4/BiOBr visible-light-driven photocatalyst using ionic liquid [C16mim]Br as solvent, reactant, template and dispersing agent at the same time. Xiao et al.33 reported an economical and facile hydrothermal approach to synthesize fluorescent carbon nitride dots (CNDs) derived from ionic liquids. The results suggest that the obtained CNDs are highly water soluble and exhibit a strong fluorescence. Li et al.34 synthesized novel sphere-like g-C3N4/BiOI composite photocatalysts by a onepot EG-assisted solvothermal process in the presence of reactable ionic liquid 1-butyl-3-methylimidazolium iodine ([Bmim]Ⅰ).The g-C3N4/BiOI composite displayed enhanced photocatalytic activity for degradation of Rhodamine B (RhB), methyl blue (MB), methyl orange (MO), bisphenol A (BPA), and chlorophenol (4-CP).
Dong35 and Li36 et al. reported that the introduction of nitrogen vacancy into g-C3N4 can chemisorb and activate N2 molecules thus significantly improving the nitrogen photofixation ability. Hong et al.37 found this nitrogen-deficient graphitic carbon nitride (g-C3N4-x) can be prepared via hydrothermal treatment using ammonium thiosulfate as an oxidant. However, in their investigation, the more concentration of ammonium thiosulfate cannot introduce more nitrogen vacancies as active sites37. In this work, based on the preparation method of Hong, we introduced ionic liquid [Bmim]Br into the solvothermal system. The results show that the introduction of [Bmim]Br can produce more nitrogen vacancies in the g-C3N4 lattice. Besides that, the morphology of the asprepared g-C3N4 is also changed, leading to the markedly increased surface area. This increased surface area of as-prepared g-C3N4 causes that more nitrogen vacancies, as the active sites, are exposed on the surface, leading to the markedly promoted nitrogen photofixation ability.
All the chemicals used in this experiment were reagent grade and without further treatment. [Bmim]Br is purchased from JCNANO Tech Co., Ltd. The pure g-C3N4 was prepared using urea as the precursor. 10 g of urea was calcined at 550 ℃ for 4 h with a ramp rate of 2 ℃∙min-1. The product was denoted as G-CN. 1 g of G-CN was dispersed into 80 mL ionic liquid [Bmim]Br under vigorous stirring at 50 ℃. 10 mL of ammonium thiosulfate solution (10 g∙L-1) was added into above suspension under vigorous stirring. The formed suspension was transferred to a 100 mL Teflon-lined autoclave and maintained at 150 ℃ for 20 h. The product was washed with deionized water, dried at 80 ℃ and denoted as ATI-CN. For comparison, I-CN was prepared following the same procedure mentioned above in the absence of ammonium thiosulfate solution.When deionized water was used to replace the [Bmim]Br following the same procedure as synthesis of ATI-CN, the obtained product was denoted as ATH-CN. H-CN was prepared following the same procedure as synthesis of ATH-CN but in the absence of ammonium thiosulfate solution.
The XRD patterns of the prepared samples were recorded on a Rigaku D/max-2400 instrument (Shimadzu, Japan) using Cu-Kα radiation (λ=0.154 nm). The scan rate, step size, voltage and current were 0.05 (°)∙min-1, 0.01°, 40 kV and 30 mA, respectively. UV-Vis spectroscopy was carried out on a V-550 model UV-Vis spectrophotometer (JASCO Japan) using BaSO4 as the reflectance sample. Fourier transform infrared (FT-IR) spectra were obtained on a FT-IR spectrometer (Nicolet 20DXB, USA). The morphologies of prepared catalyst were observed by using a scanning electron microscope (SEM, JSM 5600LV, JEOL Ltd., Japan). Nitrogen adsorption was measured at -196 ℃ on a Micromeritics 2010 analyser (USA). All the samples were degassed at 393 K prior to the measurement. The BET surface area (SBET) was calculated based on the adsorption isotherm. Electron paramagnetic resonance (EPR) spectrum was monitored using a digital X-band spectrometer (EMX-220, Bruker, USA) equipped with a Bruker ER 4121VT temperature controller within the temperature range 113-273 K. Inductively coupled plasma-mass spectrometry (ICPMS) was performed on a Perkin-Elmer Optima 3300DV apparatus (USA). The XPS measurements were performed on a 250 XPS system with Al Kα radiation as the excitation source (Thermo Escalab, USA). The binding energies were calibrated by referencing the C 1s peak (284.6 eV) to reduce the sample charge effect. Temperature programmed desorption (TPD) studies were performed using a CHEMBET-3000 (Quantachrome, USA) instrument in the temperature range from 313 to 1073 K. The photoluminescence (PL) spectra were measured at room temperature with a fluorospectrophotometer (JASCO FP-6300, Japan) using a Xe lamp as the excitation source.
Isotopic labeling experiments are carried out as follows. Labeled 15N2 gas was purchased from Sigma-Aldrich Chemical Company. In the experimental process, Ar was used to eliminate air and the possible adsorbed ammonia in the reaction system. Then, 15N2 was passed through the reaction mixture for 30 min. After that, the reactor was sealed. Other experiment conditions were the same as those for 14N2 photofixation. Indophenol method was used to examine the produced 15NH4+, owing to the low mass of 15NH4+ for liquid chromatograph-mass spectrometer (LC-MS) studies. The sample for LC-MS analysis was prepared as follows. 0.5 mL of the reaction reacted with 0.1 mL of 1% phenolic solution in 95% ethanol. Then, 0.375 mL of 1% NaClO solution and 0.5 mL of 0.5% sodium nitroprusside solution were added into above solution. MS studies were carried on an Ultimate 3000-TSQ (LCMS-ESI).
The DFT simulations were performed using the program package Dmol3. The substrate is modelled by one layer of g-C3N4 separated by a vacuum layer of 1.2 nm. All the atoms in the layer and the N2 molecule are allowed to relax. The Brillouin zones of the supercells were sampled by the Gamma points. Based on the structures of g-C3N4, the g-C3N4 surface with nitrogen atom vacancy was modelled to study the N2 adsorption properties.
The nitrogen photofixation property was evaluated according to previous literature12. The nitrogen photofixation experiments were performed in a double-walled quartz reactor in air. For these experiments, 0.2 g of photocatalyst was added to a 500 mL 0.789 g∙L-1 ethanol as a hole scavenger12. The suspension was dispersed using an ultrasonicator for 10 min. During the photoreaction under visible light irradiation, the suspension was exposed to a 250 W high-pressure sodium lamp with main emission in the range of 400 to 800 nm, and N2 was bubbled at 100 mL∙min-1 through the solution. The UV light portion of the sodium lamp was filtered by a 0.5 mol∙L-1 NaNO2 solution. All runs were conducted at ambient pressure and 30 ℃. At given time intervals, 5 mL aliquots of the suspension were collected and immediately centrifuged to separate the liquid samples from the solid catalyst. The concentration of ammonia was measured using the Nessler′s reagent spectrophotometry method (JB7478-87) with a UV-2450 spectrophotometer (Shimadzu, Japan)12, 36.
The nitrogen photofixation performance over the as-prepared catalysts under visible light is shown in Fig. 1(a). The control experiment results indicate that no NH4+ is generated in the absence of irradiation, N2 or photocatalyst, indicating that nitrogen photofixation occurs via a photocatalytic process. G-CN shows the NH4+ generation rate (r(NH4+)) of 0.38 mg∙L-1∙h-1∙g-1. I-CN shows the r(NH4+) of 0.61 mg∙L-1∙h-1∙g-1, slight higher than that of GCN. When ammonium thiosulfate is added during the preparation process, the r (NH4+) for ATH-CN sharply promotes to 6.4 mg∙L-1∙ h-1∙g-1. The r(NH4+) for ATI-CN further increases to 10.4 mg∙L-1∙ h-1∙g-1, with the turn over number (number of product molecules per catalyst molecule) of 0.96 × 10-2. This hints that the introduction of ionic liquid [Bmim]Br is beneficial to the nitrogen photofixation performance of catalysts. The Fig. 1(a) insert shows the photocatalytic stabilities of ATI-CN. No obvious decrease in nitrogen photofixation ability is observed after 20 h, hinting its good stability.
The N2 photofixation ability of ANI-CN under 15N isotope-labeled N2 (purity > 98%) was carried out to further investigate the nitrogen source of generated NH4+. The produced 15NH4+ reacts with phenolic and hypochlorite to form 15N labeled indophenol, which was analyzed by LC-MS. A strong 15N labeled indophenol anion mass spectroscopy signal presents at 199 m/z in LC-MS studies (Fig. 1(b)). It is noted that this signal intensity is obviously higher than the 14N : 15N natural abundance ratio. This confirms that N2 is the nitrogen source of generated NH4+ in this N2 photofixation process. The change in the pH value of the ANI-CN suspension during the nitrogen photofixation process is analyzed. Prior to the nitrogen photofixation process, the pH value of the suspension was measured to be 6.2. However, Fig. 1(c) shows that this pH value increases to 8.5 after 24 h because of the consumption of H+ during the nitrogen photofixation process, as shown in the following equations:
In order to compare the nitrogen photofixation ability with Tibased catalysts, the Fe-TiO2, Fe2Ti2O7 and Ru-TiO2 were prepared according to the previous work8, 9, 12. The nitrogen photofixation abilities of prepared catalysts are shown in Fig. 1(d). Obviously, Ru-TiO2 shows higher nitrogen photofixation ability than Fe-TiO2 and Fe2Ti2O7, but much lower than ANI-CN. H2 production is a possible competitive reaction. Thus the photocatalytic H2 production experiment is performed according to previous work38. The result shown in Fig. 1(e) indicates that the H2 production abilities of as-prepared catalysts are very low (less than 1 μmol∙ h-1). This is probably due to the absence of a proper co-catalyst. EPR can provide direct information on monitoring various behaviors of native defects, such as oxygen and nitrogen vacancies39, 40. As shown in Fig. 1(f), G-CN shows no peaks, suggesting that no localized unpaired electrons present in the G-CN. However, for ATH-CN and ATI-CN, a resonance signal at g=2.0031 is observed, which confirms the presence of nitrogen vacancies. The stronger resonance signal for ATI-CN hints the higher nitrogen vacancies concentration compared with ATH-CN.
The XRD patterns of as-prepared catalysts are shown in Fig. 2. G-CN and I-CN show the typical characteristic peaks of g-C3N4 located at 13.1° and 27.5°. The peak at 13.1° corresponds to the in-plane structural packing motif of the tri-s-triazine units and is indexed as the (100) peak. The distance is calculated to be d=0.67 nm. The peak at 27.5° corresponds to the interlayer stacking of the aromatic segments, with a distance of 0.326 nm, and is indexed as the (002) peak. It is noted that, compared with G-CN and I-CN, a 0.2° shift to higher 2θ value is observed for ATH-CN and ATICN. This is probably due to the formation of some crystal lattice defects in g-C3N4 when ammonium thiosulfate was added as an oxidant. The C/N ratios for both G-CN and I-CN are 0.73 obtained by elemental analysis, close to the theoretical values. For ATHCN, the C/N molar ratio is 0.77. This value further increases to 0.82 forATI-CN. Combine with the XRD results, it is deduced that the crystal lattice defects in g-C3N4 should be the nitrogen vacancies. The higher C/N molar ratio for ATI-CN causes the higher nitrogen vacancies concentration compared with ATH-CN, hinting that the introduction of [Bmim]Br into the solvothermal system is helpful for the formation of nitrogen vacancies in the g-C3N4 lattice. The C/N molar ratio for H-CN is also 0.73, same as G-CN, indicating H2O as solvent cannot form the vacancy densities.
UV-Vis spectrum is used to investigate the light absorption property of as-prepared catalysts (Fig. 3). g-C3N4 shows typical semiconductor absorption, originating from charge transfer response of g-C3N4 from the valence band (VB) populated by N 2p orbital to the conduction band (CB) formed by C 2p orbital17. The obvious red shifts of absorption band are observed for ATH-CN and ATI-CN, indicating their band gap energies are decreased. This hints that the presence of nitrogen vacancies could affect the electronic structure of g-C3N4, thus changes its optical property41. The band gap, estimated from the method of Oregan42, decreases from 2.74 eV for G-CN and I-CN to 2.63 eV forATH-CN andATICN.
The FT-IR result of [Bmim]Br, G-CN and ATI-CN are shown in Fig. 4. For G-CN, a series of peaks in the range from 1200 to 1600 cm-1 are attributed to the typical stretching modes of CN heterocycles, while the sharp peak located at 810 cm-1 is assigned to the bending vibration of heptazine rings, which indicating the synthesized g-C3N4 is composed of heptazine units. The broad absorption band around 3200 cm-1 is originated from the stretching vibration of N-H bond, associated with uncondensed aminogroups43. For ATI-CN, all the characteristic vibrational peaks of g-C3N4 are observed, suggesting that the structure of g-C3N4 is not changed after post-treatment. No peak for [Bmim]Br is observed in ATI-CN, indicating that [Bmim]Br is only used as solvent but not anchored on the surface of ATI-CN.
The morphologies of the representative samples were examined by SEM analysis (Fig. 5). The results in Fig. 5(a) indicate that GCN is composed of a large number of irregular particles. These particles exhibit a layered structure similar to that of the graphite analogue. In Fig. 5(b), after hydrothermal treatment, the morphology changes from layered structure to bulk crystal. This morphological change is consistent with previous work44. When [Bmim]Br was introduced into the solvothermal system (Fig. 5(c, d)), the morphology of I-CN and ATI-CN changes to the nanoparticles with the uniform size distribution around 30-40 nm. This smaller particle size may lead to the larger specific surface area.
To characterize the specific surface area of as-prepared g-C3N4 catalysts, the nitrogen adsorption and desorption isotherms were measured (Fig. 6). The isotherm of ATI-CN is of classical type IV, suggesting the presence of mesopores. The BET specific surface areas (SBET) of G-CN is 8.6 m2∙g-1, higher than that of ATH-CN (7.2 m2∙g-1). ATI-CN and I-CN show much higher SBET than that of G-CN, 36.7 and 37.9 m2 ∙g-1. This is due to the decreased catalyst particle sizes, which is shown in SEM images. The large SBET can not only provide more reactive sites but promote adsorption, desorption and diffusion of reactants and products, which is favorable to the photocatalytic performance. The pore size distribution of ATI-CN is presented in Fig. 6 insert. The pore distribution centered around 40-80 nm is observed in the BJH pore-size distribution curve, which should be formed by the accumulation of secondary particles.
XPS was used to characterize the surface chemical compositions of the as-prepared g-C3N4-based catalysts. In Fig. 7(a), two components located at 284.6 and 287.8 eV for both catalysts. The sharp peak around 284.6 eV is attributed to the pure graphitic species in the CN matrix. The peak with binding energy of 287.8 eV indicates the presence of sp2 C atoms bonded to aliphatic amine (-NH2 or -NH-) in the aromatic rings45. In Fig. 7(b) (N 1s region), the two contributions of G-CN located at 398.5 and 400.0 eV are assigned to the sp2-hybridized aromatic nitrogen atoms bonded to carbon atoms (C-N=C) and nitrogen atoms bonded to three carbon atoms (N-C3) in the aromatic rings46. For ATI-CN, no obvious difference in peak position is observed. However, the peak area ratio of (N-C3)/(C-N=C) decreases from 0.327 for G-CN to 0.272 for ATI-CN, clearly indicating that nitrogen vacancies are primarily located at the tertiary nitrogen lattice sites.
In order to further investigate the band structure of as-prepared catalysts, the VB XP spectra were employed (Fig. 7(c)). The VB potentials of G-CN, I-CN, ATH-CN and ATI-CN are calculated to be +1.71, +1.69, +1.67 and +1.73 eV, respectively. Combined with the UV-Vis results, the optical CB potentials of G-CN, I-CN, ATHCN and ATI-CN locate at -0.92, -0.94, -1.07 and -1.01 eV, respectively. This result indicates that the formation of nitrogen vacancies influences the band structure of as-prepared catalysts. It is reported that the standard redox potential for N2/NH3 is -0.09 V (vs NHE)10. The more negative reduction potential causes the larger CB driving force. This CB driving force determines the migration rate of photogenerated holes and electrons, causing the higher N2 photofixation ability47.
Fig. 8 shows the PL spectra of as-prepared catalysts under N2 atmosphere. In general, at a lower PL intensity, the separation rate of the photogenerated electron-hole pairs is higher. A broad PL band around 470 nm is observed for all the catalysts. This is assigned to the band-band PL phenomenon which the light energy is equal to the band gap of g-C3N4. I-CN shows the slight lower PL intensity than G-CN. This is probably due to the decreased grain size of I-CN, causing a shorter migration distance which is beneficial to charge transfer from the bulk to the surface of the g-C3N4 material and leads to a higher separation rate. In the case of ATHCN and ATI-CN, the PL intensities obviously decreased compared with G-CN and I-CN. This is due to the fact that the nitrogen vacancies formed by introducing ammonium thiosulfate could trap the photogenerated electrons, causing the increased separation rate.
Chemisorption is considered to be the essential step in heterogeneous catalysis because the chemical adsorption sites are generally the reaction centers to activate reactant molecules. N2-TPD was carried out to investigate the N2 adsorption situation on the surface of G-CN, ATH-CN and ATI-CN (Fig. 9). Obviously, two adsorbed N2 species in ATH-CN and ATI-CN are observed. One peak at 110-130 ℃ is assigned to physical adsorption. The other peak at 320-360 ℃ is attributed to the strong chemisorption species of N2 molecule. In the case of G-CN, only physical adsorbed N2 species is observed. This result indicates that nitrogen vacancies could act as chemical adsorption sites to activate N2 molecule for nitrogen photofixation. This is consistent with previous results35. It is noted that, compared with ATH-CN, more N2 species chemisorbs on ATI-CN surface. This is due to the more nitrogen vacancies on ATH-CN which provides more chemical adsorption sites.
RhB degradation abilities over as-prepared catalysts under visible light are shown in Fig. 10. The reaction rate constant (k) was obtained by assuming that the reaction followed first-order kinetics (Fig. 10 insert). The results indicate that the catalyst with higher surface area displays higher RhB degradation rate. The rate constant for ATI-CN is 0.014 min-1, which is 3.3-fold greater than that of G-CN. However, the NH4+ generation rate for ATI-CN increases 27.4-fold compared with G-CN, which is much higher than the increase of RhB degradation rate. It is deduced that the enhanced nitrogen photofixation ability is not only due to the increased SBET but also due to the formation of nitrogen vacancies.
According to the XPS results, it is deduced that the nitrogen vacancies are located at the three-coordinated nitrogen, as shown in Fig. 11. To further confirm that N2 is activated by nitrogen vacancies, density functional theory (DFT) simulations were employed to investigate the interaction between N2 molecule and the g-C3N4 with nitrogen vacancies (Fig. 11). The calculation results show that the adsorption energy is -166.2 kJ∙moL-1, confirming the chemisorption occurs. When the N2 molecule adsorbs on the nitrogen vacancies, an σ bond between the N2 molecule and the nearest C atom is formed, causing the N≡N bond prolonged from 0.1107 to 0.1242 nm. This result confirms that the nitrogen vacancies can activate N2 molecule.
In summary, the possible nitrogen photofixation process over g-C3N4 with nitrogen vacancies is shown in Fig. 12. First of all, N2 species chemisorbs on the nitrogen vacancies. Under visible light irradiation, photogenerated electron-hole pairs are formed (step 1). The photogenerated electrons are trapped by the nitrogen vacancies and immediately transferred to the adsorbed N2 molecule (step 2). Because the bonding orbitals of N2 molecule are occupied by four electrons, this photogenerated-electron has to occupy the anti-bonding orbitals, leading to the nitrogen activation (step 3). The activated N2 molecule reacts with H+ in water to form NH3, and then it finally forms NH4+ and OH-.
By introducing ionic liquid [Bmim]Br as solvent into the solvothermal post-treatment, graphitic carbon nitride with larger surface area and more nitrogen vacancies is synthesized in this work. These nitrogen vacancies not only trap the photogenerated electrons to promote the separation rate, but serve as active sites to adsorb and activate N2 molecules. The adsorption energy is -166.2 kJ ∙mol-1 when N2 molecule interacts with nitrogen vacancy sites. The N≡N bond is prolonged from 0.1107 to 0.1242 nm, confirming that nitrogen vacancy can activate N2 molecule. This increased surface area of as-prepared g-C3N4 causes that more nitrogen vacancies, as the active sites, are exposed on the surface, leading to the markedly promoted nitrogen photofixation ability. Under visible light, the NH4+ generation rate of ANI-CN reaches 10.4 mg∙L-1∙h-1∙g-1, which is 27-fold higher than that of G-CN.