Oxygen vacancy-rich MoO_3−x nanobelts for photocatalytic N₂ reduction to NH₃ in pure water

Abstract

Photocatalytic nitrogen fixation is a promising sustainable and green strategy for NH₃ synthesis. The development of an efficient photocatalyst is the key to make this approach applicable. In this work, MoO_3−x nanobelts are synthesized by a hydrothermal method, and they are firstly utilized as photocatalysts for N₂ fixation without using sacrificial reagents and precious-metal co-catalysts at room temperature and atmospheric pressure. It is found that oxygen vacancies (O_VS) are located on the (001) and (100) planes by STEM analysis. The surface O_VS can chemisorb the N₂ molecules by a side-on model on (001) while an end-on model on the (100) plane and elongate their bond length, playing a critical role in the photocatalytic N₂ fixation activity. This result provides a new insight into developing efficient photocatalysts with original oxygen vacancies.

---

Introduction

Nitrogen fixation such as N₂ reduction to NH₃ is an important artificial synthesis in the chemical industry.¹ However, the industrial synthetic strategies consume a large amount of the world's annual energy supply and release large quantities of CO₂ aggravating global climate change.² After the first report by Schrauzer and Guth in 1977, the photocatalytic N₂ reduction to NH₃ has attracted extensive research interest and shown high potential as a sustainable and green solution compared with the industrial Haber–Bosch process.³ However, the photocatalytic activity of N₂ fixation is still not satisfactory due to the high reaction activation energy barriers and hard cleavage of the strong N≡N triple bond energy (941 kJ mol⁻¹) of N₂ molecules.⁴

It is well-known that crystal defect engineering has shown positive effects on many kinds of photocatalytic reactions,⁵ and oxygen vacancies (O_VS) are one of the most widely used defects.^6–8 There are three main kinds of ways to generate O_VS for photocatalytic experiments. Generally, O_VS could be formed via hydrothermal or solvothermal treatment naturally. For example, Tian et al. reported oxygen vacancy-rich BiPO₄ through a one-pot solvothermal treatment for MB degradation.⁹ Additionally, thermal treatment in hydrogen or inert gases is also a usual method to modulate the surface O_VS of samples.¹⁰ Chen and co-workers enhanced the photocatalytic performances of TiO₂ nanocrystals for methylene blue degradation and water splitting by generating O_VS in a hydrogen atmosphere and proved that the generation of O_VS was positive in enhancing light absorption.¹¹ Moreover, Ding et al. demonstrated that a triple atom defect cluster (Br,Bi,O) on BiOBr could promote the generation of singlet oxygen through an energy transfer process. It resulted in an excellent conversion rate and selectivity in the sulfoxidation reaction.¹²

It has been demonstrated that the assistant electron-donating center can firstly capture the N₂ molecules and weaken the N≡N triple bond, and then the photogenerated electrons accumulate in the capture sites to reduce the N₂ into NH₃.^13,14 Many research results have revealed that oxygen vacancies can also act as electron trap sites and activate the inert N₂. For example, Shiraishi et al. reported that the original O_VS in TiO₂ directly determined the efficiency of photocatalytic nitrogen fixation from water and N₂.⁴⁵ Zhang and co-workers found that the energy barrier of N₂ reduction to ammonia could be remarkably decreased by elongating the N≡N triple bond via an end-on configuration between the adsorbed N₂ and O_VS on the BiOBr surface.^2,15 It was also reported that the O_VS in Bi₂MoO₆ played an important role in the N≡N triple bond splitting.^16,17 The O_VS in LDH nanosheets enhanced the adsorption and activation of N₂ and H₂O, which made them show excellent photocatalytic activity for N₂ reduction.¹⁸

Molybdenum trioxide (MoO₃) has been extensively explored as a photochromic and electrochromic material due to its unique energetic and electrical properties.^19–21 Oxygen vacancy-rich MoO_3−x has an unusual oxygen defect structure and novel properties in the nanometer regime.^22,23 It has been successfully used to photocatalyze isopropyl alcohol to propylene under visible light.²⁴ Besides, electrocatalytic nitrogen fixation of MoO₃ has also shown good activity in a recent study.²⁵ However, there are few reports about photocatalytic N₂ reduction by MoO_3−x. In this work, MoO_3−x nanobelts were synthesized by a simple one-pot hydrothermal method as previously described.²⁴ In comparison with the annealed sample in air, it was found that the O_VS on the surface of MoO_3−x can chemisorb and reduce the activation energy of N₂ molecules, playing a critical role in determining the photocatalytic activity for N₂ reduction.

Experimental

Materials and methods

In a typical synthesis, 0.36 g of molybdenum acetylacetonate was added into 40 ml of distilled water. Then the obtained solution was loaded into a stainless steel Teflon-lined autoclave and the sealed autoclave was kept at 180 °C for 20 h. After naturally cooling to room temperature, the blue product was washed with distilled water several times to remove any ions and possible remnants and subsequently dried at 50 °C for 12 h in a vacuum oven. For comparison, the as-synthesized products were annealed at 400 °C in a tube furnace in air atmosphere for x minutes (x = 0, 1, 5, and 30) to reduce surface oxygen vacancy concentration, and the final reactants are denoted as MoO_3−x-x min.

Characterization of the photocatalysts

X-ray diffraction (XRD, Bruker, D8 Advance X-ray diffractometer) was used to characterize the crystal structure of the samples with the scanning range of 2θ from 10° to 50°. The morphology of the samples was observed on a field emission scanning electron microscope (FE-SEM, ZEISS SUPRA-®-55). The STEM and EELS analyses were performed on a Double-Cs aberration-corrected Themis G2 microscope at 300 kV with a Quantum GIF detector. UV-vis absorption spectroscopy was carried out on a UV-2450 (Shimadzu). X-ray photoelectron spectroscopy (XPS) was performed on a Thermo Scientific ESCALAB 250 Xi system. The Raman spectra were measured using a Horiba iHR 320 with a 532.1 nm laser as the excitation light, and the accumulation time was 25 s. The specific surface areas of the samples were determined through N₂ adsorption–desorption isotherms at 77 K using an accelerated surface area and porosimetry system (ASAP, 2020 HD88). Electron spin resonance (ESR) spectroscopy was carried out using a Bruker A300, German under a magnetic field from 3400 to 3600 G. The temperature-programmed desorption of N₂ (N₂-TPD) was tested with an Autosorb-iQ-C chemisorption analyser (Quantachrome, USA) with the temperature from 0 to 773 K.

Photocatalytic N₂ fixation

The photocatalytic performance of the sample for N₂ reduction was evaluated by the production rate of NH₄⁺ under full spectrum illumination without any sacrificial agents at room temperature and atmospheric pressure. The light source was a 300 W xenon lamp (CEL-HXF300, wavelength: 300–2500 nm, radiant output 50 W) and LEDs with wavelengths of 365 nm, 385 nm, 400 nm and 427 nm (UVEC-4II). A water circulation system was used to maintain the temperature at 25 °C. In each experiment, 50 mg of the photocatalyst was added and dispersed in 100 mL of distilled water, and then the mixture was kept under stirring and irradiated from the top-window with a N₂ flow at a rate of 60 mL min⁻¹. 3 mL of the solution was taken out after every 2 h during the illumination. The catalyst in the sample was removed by centrifugation with a rotation rate of 16 000 r min⁻¹ for 5 min and then was filtered two times. Finally, the photocatalytic product was determined by ion chromatography (Metrohm IC). The photocatalyst was recycled after the photocatalytic experiment by centrifugation and then was filtered several times. And the recycling photocatalytic experiment was carried out with a 300 W xenon lamp.

DFT calculations

All calculations were performed with the plane-wave projector-augmented wave method,^26,27 as implemented in the CASTEP package. The Perdew–Burke–Ernzerhof (PBE) form of generalized gradient approximation (GGA) was chosen as the exchange–correlation potential.^28,29 The PBE + U approach with spin polarization was employed to take account of the strong on-site Coulomb interaction (U), present in the localized 4d electrons of Mo with U values being set to 2 eV. The cutoff energy for the plane wave was 340 eV. The integration in the Brillouin zone was carried out on a set of k-points determined by the Monkhorst and Pack scheme, 4 × 1 × 4. Convergence was assumed when the total energy difference was less than 10⁻⁶ eV between cycles of self-consistent iterations. Based on the results of the STEM analysis, a unit cell with one O vacancy at the 2c site (Mo₄O₁₁) was used to perform the geometry-optimization calculations. The atomic populations (Mulliken) were analyzed after optimization. Furthermore, the binding energy of O-deficient Mo atoms in this structure model was further calculated by removing another coordinating O atom. In comparison, the perfect unit cell of MoO₃ (Mo₄O₁₂) was also calculated with the same calculation configurations.

The calculation of the N–N bond was performed using the first-principles density functional theory (DFT) + U calculations with the exchange–correlation energy functional, which were described by a generalized gradient approximation with the Perdew–Burke–Ernzerhof (PBE) exchange–correlation function.²⁸ We also took van der Waals interactions into correction by the DFT-D2 method of Grimme.³⁰ To simulate N₂ activation on the MoO₃ (001) and (100) surfaces, (2 × 1 × 3) and (3 × 1 × 2) supercells were used, respectively, enabled by a VASP code in which the projector-augmented wave (PAW) method presented the electron–ion interaction with a kinetic energy cutoff of 520 eV.^27,31 The k-points were 8 × 4 × 1, 4 × 4 × 1 and 4 × 8 × 1 for each supercell. During optimizations, the energy and force converged to 10⁻⁴ eV per atom and 0.02 eV Å⁻¹, respectively.

Results and discussion

The XRD patterns of the as-synthesized and annealed samples with different times are shown in Fig. 1a. All the peaks can be indexed to orthorhombic MoO₃ (o-MoO₃) whose 2θ values are located at 12.8°, 23.4°, 25.7°, 25.8°, 27.3° and 39.0°. The result is consistent with previous reports.^32,33 After annealing, the XRD patterns of the annealed samples don't show an obvious shift as presented in Fig. 1a. The SEM images of the as-synthesized and annealed products are shown in Fig. S1.† It can be observed that the products are all composed of nanobelts, and the morphologies remain stable even after annealing. In addition, as shown in the HRTEM image, a MoO_3−x nanobelt is single crystalline and the inserted diffraction pattern in Fig. 1c proves the single-crystalline nature of the MoO_3−x nanobelt. Detailed analysis for fringing the formation of oxygen vacancies was performed using atomic scale HAADF Z-contrast images in Fig. 1d. Ordered oxygen vacancies residing at certain crystal plane surfaces often show up as dark-lines in a Z-contrast image due to the reduced average atomic density.³⁴ As shown in Fig. 1d, the white arrows point to the ordered oxygen vacancies residing at the (001) & (100) surfaces of MoO_3−x, where we observed such dark-lines along these surface planes. In addition, the red lines mark the dislocation formed in MoO_3−x. The dislocations formed in the MoO_3−x may partially release the internal strain induced by the formation of such oxygen vacancies.

Fig. 1 (a) XRD patterns of the as-synthesized and annealed MoO_3−x nanobelts for various times; (b) HAADF STEM image of the overall morphology of the MoO_3−x nanobelts; (c) HRTEM image of MoO_3−x with the insert showing the electron diffraction pattern; (d) atomic scale HAADF image showing the ordered oxygen vacancies at the (001) & (100) planes.

In order to prove the changes of O_VS, the fine structures of the electron energy loss spectroscopy (EELS) edges were further measured. The EELS edges reflect the unoccupied local density of states and therefore, provide useful local oxidation states and coordination chemistry of materials.³⁵ The Mo and O EELS results of the as-obtained MoO_3−x and MoO_3−x after 30 min annealing were acquired and plotted in Fig. 2 for comparison. The lattice distortion and reduced symmetry at the oxygen-deficient MoO_3−x certainly caused changes in the oxygen K edges. As reported by previous literature studies, the enhancement of a peak labelled by the red arrow is a signature of the formation of oxygen vacancies.³⁶ Another interesting feature to notice is the decreased intensity of the third peak as pointed by the purple arrow in the MoO_3−x sample compared with the MoO_3−x after 30 min annealing. Based on the multiple-scattering theory, the decreased intensity is caused by a decrease in the number of backscattering O²⁻ ions at the first or second-nearest oxygen neighbours, further proving the presence of oxygen vacancies in the MoO_3−x sample.³⁷

Fig. 2 EELS analysis of the as-synthesized MoO_3−x and MoO_3−x after 30 min annealing in air atmosphere.

The UV-vis absorption spectra of the samples in Fig. 3a were transformed from their reference spectra based on the Kubelka–Munk formula. There is a tail absorption for each sample, which results from the O_VS,¹⁷ and decreases gradually as the annealing time increases. The colour of the sample gradually turns white after the heat treatment as shown in Fig. S2.† These results indicate that the annealing treatment can remove the O_VS in the samples.

Fig. 3 (a) UV-vis absorption spectra, (b) Raman spectra, (c) Mo 3d-related XPS spectra, and (d) O 1s-related XPS spectra of the as-synthesized and annealed MoO_3−x for different times.

The Raman spectroscopy characterization results are shown in Fig. 3b. The Raman peaks of the as-synthesized MoO_3−x nanobelts are in agreement with those of previous research.³⁸ The peak at 819 cm⁻¹ shows a strong and well-defined line, assigned to the doubly coordinated bridging oxygen (Mo–O–Mo units). A couple of weak peaks at 461 and 676 cm⁻¹ are attributed to the triply coordinated oxygen (Mo₃–O). The peak at 996 cm⁻¹ consists of the terminal oxygen (Mo=O) stretching mode. Both the XRD patterns and Raman spectra indicate that the prepared MoO_3−x nanobelts belong to an orthorhombic phase. However, there is no obvious shift in the Raman spectra of the annealed samples for different times, and it may result from the fact that the reduced O_VS by annealing are not capable enough to change the Raman characteristics. Besides, we have tried to choose lower annealing temperatures (150 °C, 200 °C, and 300 °C) in our study before ensuring 400 °C as the final annealing temperature. And it is found that there was no obvious change in the sample colour although the annealing time was 1 h at 300 °C (Fig. S3†), indicating that the O_VS do not decrease very much. On the other hand, the structure of the samples is stable after 400 °C annealing, because the XRD and Raman results are similar (Fig. 1a and 3b) for the as-synthesized and annealed samples. However, the absorption coefficient of the UV-vis tail absorption (Fig. 3a) decreases as the annealing time increases. This proves that O_VS were drastically removed and the crystal structure remained stable after annealing at 400 °C.

The X-ray photoelectron spectra (XPS) are presented in Fig. 3c and d. For the as-synthesized MoO_3−x nanobelts, the peaks at 232.3 eV and 235.6 eV could be ascribed to the 3d_5/2 and 3d_3/2 orbitals of Mo 3d, respectively.^39,40 In Fig. 3d, the peaks at 531.4 eV originate from the lattice oxygen anions.⁴¹ After annealing, the peaks of Mo 3d gradually left-shift as the annealing time is increased. Similar phenomena are observed for the O 1s orbital as shown in Fig. 3d. In order to understand the deep reasons, the Mulliken electronegativity of the Mo and O elements of MoO₃ with 25% or without O_VS was calculated using CASTEP (see details in the Experimental section). Mo₄O₁₂ and Mo₄O₁₁ were chosen as the unit cells in this theoretical calculation, corresponding to the non-defect and defect MoO₃, respectively. The calculation results are shown in Table S1.† The average charges for Mo and O in the non-defect MoO₃ are 1.44 and −0.451, respectively. The corresponding values are 1.37 and −0.499 for the defect MoO₃, respectively. Obviously, Mo and O in the defect MoO₃ show more negative charges than those in the non-defect MoO₃, which result in the increasing binding energy for Mo 3d and O 1s in the decreasing oxygen defect samples as shown in Fig. 3c and d.⁴² Besides, the results of binding energy for Mo₄O₁₂ and Mo₄O₁₁ have also been calculated. In comparison with Mo₄O₁₂, the binding energy of Mo₄O₁₁ (7.64 eV) becomes smaller than that of Mo₄O₁₂ (8.52 eV), which also confirms the weaker chemical bond of Mo–O in the sample with higher O_VS concentration.

The photocatalytic performance was characterized under the irradiation of a 300 W xenon lamp with full spectra, and the quantity of the photo-fixed NH₄⁺ was determined using an ion chromatograph. Compared with the traditional method (Nessler's reagent), the ion chromatograph could detect NH₄⁺ more accurately and the measurement results would not be influenced by any organic reagents during the reaction.⁴³ Pure water was used as the solvent and proton source in this work. Generally, the chemical reaction from N₂ to NH₃ in neutral solution can be described as the following equation:⁴⁴

The photocatalytic nitrogen fixation activity is shown in Fig. 4 (the representative original data from ion chromatography are shown in Fig. S4†). The as-synthesized MoO_3−x exhibits the highest NH₃ production rate reaching nearly 11.1 μmol L⁻¹ g⁻¹ h⁻¹, while the production rate is just a quarter remaining for the MoO_3−x-1 min sample. In addition, there is no detectable N₂ reduction product for the MoO_3−x-5 min and MoO_3−x-30 min samples. As we know, the BET surface area is also a significant parameter for the photocatalytic activity. However, the surface areas are all around 20 m² g⁻¹ (Fig. S5†). Accordingly, the surface area can't obviously affect the photocatalytic N₂ reduction activity. Furthermore, the light absorption of MoO_3−x-5 min doesn't show great difference in comparison with that of the as-synthesized sample, but its photocatalytic product is undetectable. To find out the reason, electron spin resonance (ESR) spectroscopy and temperature-programmed desorption of N₂ (N₂-TPD) were used to investigate the O_VS and N₂ adsorption on the surface O_VS. In Fig. 5a, a signal exists at g = 2.001 caused by O_VS in the as-synthesized and annealed samples.⁴⁵ However, there are no obvious changes even if the samples are annealed for 30 minutes at 400 °C. This must result from the fact that the O_VS have not been totally removed in the whole sample. In Fig. 5b, a single desorption peak of N₂ begins at 180 °C and centers at 320 °C for the as-synthesized sample, which is ascribed to the chemisorbed N₂.² However, the peak becomes weaker obviously for sample MoO_3−x-5 min and disappears for sample MoO_3−x-30 min. These results indicate that the surface O_VS were gradually removed but there were still O_VS in the inner part of the annealed samples. Compared with the photocatalytic activities in Fig. 4, samples MoO_3−x-5 min and MoO_3−x-30 min don't show detectable products, which must be caused by the disappearance of enough surface O_VS. In other words, the surface O_VS are the key factor for the efficient photocatalytic N₂ reduction.

Fig. 4 Photocatalytic experiments for nitrogen fixation of the as-synthesized and annealed MoO_3−x nanobelts (light source: 300 W xenon lamp; photocatalyst: 0.05 g; solution: 100 ml of pure water. The error bar results came from photocatalytic experiments of the same samples performed three times.)

Fig. 5 (a) ESR spectra and (b) N₂-TPD profiles of the as-synthesized and annealed samples for different times.

The reported photocatalytic activities of typical metal oxides were compared, which were characterized in pure water under UV-vis light irradiation (Table S2†). Hirakawa and co-workers created Ti³⁺ as active sites in TiO₂ which exhibited an NH₃ production rate of 35.0 μmol g⁻¹ h⁻¹.⁴⁵ Zhang et al. reported that the NH₃ production of Mo-doped W₁₈O₄₉ reached 61.9 μmol g⁻¹ h⁻¹.⁴⁶ Most other reported metallic oxides exhibited production rates around 0.41–28.4 μmol g⁻¹ h⁻¹.^47–51 The as-synthesized MoO_3−x nanobelts show this production in the middle range.

It could be seen that there was strong IR absorption even at 850 nm from the UV-vis absorption (Fig. 3a). The photocatalytic performance was further characterized under LED irradiation with different wavelengths to find out the connection between the light wavelength and NH₃ synthesis. LEDs with wavelengths of 365 nm, 384 nm, 400 nm, and 427 nm were used as the light source. The results have been normalized because of the different powers of LEDs. As shown in Fig. 6a, the LED with a wavelength of 365 nm exhibits the highest NH₃ production rate reaching nearly 3.3 μmol L⁻¹ g⁻¹ h⁻¹ W⁻¹, while NH₃ production is not detected in the reaction irradiated by 400 and 427 nm LEDs. This means that the sample could only fix N₂ under ultraviolet light irradiation (λ < 400 nm) although there was still light absorption in the longer wavelength range (λ ≥ 400 nm). In addition, the quantum efficiency (QE) was also calculated by the following equation:⁵²

where P is the optical power density of the LED, S is the area of the illumination, t is the reaction time and α is the light absorption. Here it has been supposed that all of the light energy was absorbed during the photocatalytic reaction. The optical power was measured to be 325 mW for the 365 nm LED, and the calculated value of QE at 365 nm was about 0.013%.

Fig. 6 (a) Photocatalytic activities for nitrogen fixation of the as-synthesized MoO_3−x nanobelts under LED light irradiation with various wavelengths; (b) control experiment in N₂ and Ar gas in succession for photocatalytic nitrogen fixation (light source: 300 W xenon lamp).

To ensure the source of the N element, argon was used to replace the N₂ as the gas source after illumination for 6 h under a N₂ flow. Obviously, the N₂ fixation reaction hardly occurs under the Ar flow because the amount of NH₄⁺ doesn't increase (Fig. 6b). This result indicates that the N atoms in NH₃ come from N₂. Besides, recycling photocatalytic experiments were also carried out to characterize the photochemical stability of MoO_3−x. It has been found that the NH₃ production is just 70% remaining during the first cycle, and 65% remaining for the second cycle (Fig. S6a†). The possible reason is that some O_VS are inactive for the further photocatalytic N₂ fixation after each cycle, because there are no obvious change for the XRD and TEM results of the sample before and after reaction (Fig. S7 and S8†). Moreover, we have also studied the effect of catalyst dosage on photocatalytic activities. As shown in Fig. S6b,† the rate of NH₃ production would be enhanced obviously if the quantity of the catalyst is increased (nearly 1 μmol L⁻¹ for 50 mg of catalysts, while 4 μmol L⁻¹ for 1 g of catalysts). However, the rates of NH₃ production decrease if the value is converted to one gram. It is reasonable that the larger dosage of the catalyst cannot be exposed enough to light for photocatalytic N₂ fixation. Then a larger dosage of the catalyst exhibits a lower photocatalytic activity when converting to μmol L⁻¹ g⁻¹.

Based on the result of the electron diffraction pattern in Fig. 1c, the (100) and (001) planes exist in the MoO_3−x nanobelts. Theoretical calculation was carried out to further study the formation energy of O_VS in these two planes. As shown in Fig. 7, there are three and four kinds of O-sites on the (001) and (100) planes, respectively. The calculation results are shown in Table S3.† For the (001) plane, the O_VS formation energy on the O1 site is equal to that on the O2 site (1.05 eV), which is more stable than that on the O3 site (1.29 eV). This means that it is easier to form O_VS on the O1 (O2) site. For the (100) plane, the formation energy on the O1 and O2 sites is equal to −1.97 eV, and the value is 2.24 eV on the O3 and O4 sites. These results indicate that the O_VS are easier to form on the O1 and O2 sites of the (001) plane, and the O_VS can spontaneously form on the O1 and O2 sites of the (100) plane.

Fig. 7 (a) Three kinds of O sites on the (001) plane and (b) four kinds of O sites on the (100) plane.

The density of states of the (001) plane with or without O_VS was calculated to explain the electron transfer processes during O_VS-mediated photocatalytic nitrogen fixation (Fig. S9†). When there are O_VS on the (001) plane, some defect levels appear between the valence and conduction bands. This also explains why there is tail absorption for defect MoO_3−x (Fig. 3a and 8a). Due to the lower energy of the defect levels, the photoexcited electrons on the conduction band will move to the defect levels, and the O_VS thus act as the trap sites of electrons. Subsequently, the photoexcited electrons with enough energy will reduce the chemically adsorbed N₂ by O_VS into NH₃.

Fig. 8 (a) Electron transfer processes during O_VS-mediated nitrogen fixation. (b) N₂ adsorbed on the pristine surface; (c) terminal side-on adsorption on (001) and (d) end-on bridging adsorption on (100) for N₂ (yellow means increasing charge and blue means decreasing charge).

Additionally, the theoretical calculation was carried out to further study the function of surface O_VS for N₂ photofixation. When there are no O_VS on the surface of MoO₃, the N₂ molecule floats above the surface of the catalyst, and the N≡N triple bond length (1.113 Å) doesn't change (Fig. 8b). There are two models including end-on and side-on ways for nitrogen adsorption on the O_VS on the (001) surface and (100) surface where the O_VS are located as confirmed by atomic scale HAADF (Fig. 1d) and the above discussion. Based on the theoretical calculation results, there is little possibility for end-on ways on the (001) surface and side-on ways on the (100) surface for nitrogen adsorption, respectively, due to the reason that their adsorptions are too unstable to produce models for further calculation.

The bond length N1–N2 is 1.152 Å and 1.157 Å as observed on the (001) surface in Fig. 8c and on the (100) surface in Fig. 8d, respectively.

Obviously, the bond length of the N–N bond is elongated for the adsorbed N₂ molecule. In other words, the N₂ molecule chemisorbs on the O_VS and its activation energy can be decreased, which is beneficial to photocatalytic N₂ reduction to NH₃. Fig. 8c and d show the charge distribution of nitrogen by side-on adsorption on the (001) plane and end-on adsorption on the (100) plane, respectively. It is well-known that N₂ molecule reduction needed electrons to change into NH₃. The Bader charge analysis in Fig. 8 demonstrates that there will be 0.34|e| electron on the (001) plane and 0.09|e| on the (100) plane transferring from the unsaturated Mo atom (O_VS) to the N₂ molecule, respectively. When the semiconductor is irradiated by light, the photoexcited electrons can be successfully injected from the O_VS into the N₂ molecule for its reduction reaction.

Conclusions

In summary, we have successfully synthesized o-MoO_3−x by a simple one-pot hydrothermal method and tested its photocatalytic nitrogen fixation activity in pure water for the first time. We found that the O_VS were located on the (001) plane and (100) plane by STEM measurement. The surface O_VS play a critical role in the photocatalytic N₂ reduction by chemisorption and the activation energy decrease of nitrogen molecules. The N₂ molecule prefers to adsorb on the O_VS by the side-on model on the (001) plane and the end-on model on the (100) plane, respectively. Our findings can provide useful information for further development of other efficient photocatalysts for N₂ reduction.

Conflicts of interest

There are no conflicts of interest to declare.

Acknowledgements

This research was financially supported by Guangdong key lab Project (2017B030301013), Shenzhen Science (JCYJ20151015162256516), Technology Research Grant (JCYJ20160226105838578), National Key R&D Program of China (2016YFB0700600), and the Shenzhen Peacock Plan (KQTD2014062714543296).

Notes and references

1. D. E. Canfield, A. N. Glazer and P. G. Falkowski, Science, 2010, 330, 192–196.
2. H. Li, J. Shang, Z. Ai and L. Zhang, J. Am. Chem. Soc., 2015, 137, 6393–6399.
3. G. N. Schrauzer and T. D. Guth, J. Am. Chem. Soc., 1977, 99, 7189–7193.
4. H. Hirakawa, M. Hashimoto, Y. Shiraishi and T. Hirai, ACS Catal., 2017, 7, 3713–3720.
5. Z. Dai, F. Qin, H. Zhao, J. Ding, Y. Liu and R. Chen, ACS Catal., 2016, 6, 3180–3192.
6. J. Ding, Z. Dai, F. Qin, H. Zhao, S. Zhao and R. Chen, Appl. Catal., B, 2017, 205, 281–291.
7. Q. Wang, W. Wang, L. Zhong, D. Liu, X. Cao and F. Cui, Appl. Catal., B, 2018, 220, 290–302.
8. Q. Wang, Z. Liu, D. Liu, G. Liu, M. Yang, F. Cui and W. Wang, Appl. Catal., B, 2018, 236, 222–232.
9. F. Tian, H. Zhao, G. Li, Z. Dai, H. Zhao, Y. Liu and R. Chen, ChemSusChem, 2016, 9, 1579–1585.
10. D. Tio, Nanoscale, 2013, 5, 3601–3614.
11. X. Chen, L. Liu, P. Y. Yu and S. S. Mao, Science, 2011, 331, 746–751.
12. J. Ding, Z. Dai and F. Tian, J. Mater. Chem. A, 2017, 5, 23453–23459.
13. K. C. MacLeod, D. J. Vinyard and P. L. Holland, J. Am. Chem. Soc., 2014, 136, 10226–10229.
14. S. Wang, X. Hai, X. Ding, K. Chang, Y. Xiang, X. Meng, Z. Yang, H. Chen and J. Ye, Adv. Mater., 2017, 29, 1–7.
15. J. Li, H. Li, G. Zhan and L. Zhang, Acc. Chem. Res., 2017, 50, 112–121.
16. C. Shi, X. Dong, Y. Hao, X. Wang, H. Ma and X. Zhang, RSC Adv., 2017, 7, 50040–50043.
17. Y. Li, Y. Zhao, H. Cheng, Y. Hu, G. Shi, L. Dai and L. Qu, J. Am. Chem. Soc., 2012, 134, 15–18.
18. Y. Zhao, Y. Zhao, G. I. N. Waterhouse, L. Zheng, X. Cao, F. Teng, L. Z. Wu, C. H. Tung, D. O'Hare and T. Zhang, Adv. Mater., 2017, 29, 1–10.
19. H. Natori, K. Kobayashi and M. Takahashi, J. Oleo Sci., 2009, 58, 203–211.
20. Y. Yang, Y. Cao, B. Loo and J. Yao, J. Phys. Chem. B, 1998, 102, 9392–9396.
21. M. Greenblatt, Chem. Rev., 1988, 88, 31–53.
22. S. G. Processes, N. Pinna and M. Niederberger, Angew. Chem., Int. Ed., 2008, 47, 5292–5304.
23. M. A. Ibrahem, F.-Y. Wu, A. Mengistie, C. Chang, D. A. Mengistie, L. Li, C. W. Chu and V. A. Online, Nanoscale, 2014, 6, 5484–5490.
24. H. Bai, W. Yi, J. Li, G. Xi, Y. Li, H. Yang and J. Liu, J. Mater. Chem. A, 2016, 4, 1566–1571.
25. J. Han, X. Ji, X. Ren, G. Cui, L. Li, F. Xie, H. Wang, B. Li and X. Sun, J. Mater. Chem. A, 2018, 6, 12974–12977.
26. P. E. Blöchl, Phys. Rev. B: Condens. Matter Mater. Phys., 1994, 50, 17953–17979.
27. D. Joubert, Phys. Rev. B: Condens. Matter Mater. Phys., 1999, 59, 1758–1775.
28. J. P. Perdew, K. Burke and M. Ernzerhof, Phys. Rev. Lett., 1996, 77, 3865–3868.
29. V. I. Anisimov, M. A. Korotin, J. Zaanen and O. K. Andersen, Phys. Rev. Lett., 1992, 68, 345–348.
30. S. Grimme, T. O. Chemie and O. I. D. U. Münster, J. Comput. Chem., 2006, 16, 1787–1799.
31. G. Kresse and J. Furthmüller, Phys. Rev. B: Condens. Matter Mater. Phys., 1996, 54, 11169–11186.
32. N. Pinna and M. Niederberger, Angew. Chem., Int. Ed., 2008, 47, 5292–5304.
33. S. Hu and X. Wang, J. Am. Chem. Soc., 2008, 130, 8126–8127.
34. R. F. Klie and N. D. Browning, Microsc. Microanal., 2018, 8, 475–486.
35. D. A. Muller, Nat. Mater., 2009, 8, 263–270.
36. D. Wang, D. S. Su and R. Schlögl, Z. Anorg. Allg. Chem., 2004, 630, 1007–1014.
37. N. D. Browning, H. O. Moltaji and J. P. Buban, Phys. Rev. B: Condens. Matter Mater. Phys., 1998, 58, 8289–8300.
38. L. Seguin, M. Figlarz, R. Cavagnat and J. C. Lassegues, Spectrochim. Acta, Part A, 1995, 51, 1323–1344.
39. M. A. Bica De Moraes, B. C. Trasferetti, F. P. Rouxinol, R. Landers, S. F. Durrant, J. Scarmínio and A. Urbano, Chem. Mater., 2004, 16, 513–520.
40. Q. Huang, S. Hu, J. Zhuang and X. Wang, Chem. – Eur. J., 2012, 18, 15283–15287.
41. J. Wang, Y. Xia, H. Zhao, G. Wang, L. Xiang, J. Xu and S. Komarneni, Appl. Catal., B, 2017, 206, 406–416.
42. Z. Mei, M. Zhang, J. Schneider, W. Wang, N. Zhang, Y. Su, B. Chen, S. Wang, A. L. Rogach and F. Pan, Catal. Sci. Technol., 2017, 7, 982–987.
43. X. Gao, Y. Wen, D. Qu, L. An, S. Luan, W. Jiang, X. Zong, X. Liu and Z. Sun, ACS Sustainable Chem. Eng., 2018, 6, 5342–5348.
44. K. Tennakone, S. Punchihewa, S. Wickramanayake, C. A. N. A. N. Fernando, O. A. Illeperuma, O. A. Ileperuma and S. Punchihewa, J. Chem. Soc., Chem. Commun., 1987, 1078–1080.
45. H. Hirakawa, M. Hashimoto, Y. Shiraishi and T. Hirai, J. Am. Chem. Soc., 2017, 139, 10929–10936.
46. N. Zhang, A. Jalil, D. Wu, S. Chen, Y. Liu, C. Gao, W. Ye, N. Zhang, A. Jalil, D. Wu, S. Chen, Y. Liu, C. Gao and W. Ye, J. Am. Chem. Soc., 2018, 140, 9434–9443.
47. Q. Li, K. Donmen, S. Naito, T. Onishi and K. Tamaru, Chem. Lett., 1983, 2, 321–324.
48. M. M. Khader, N. N. Lichtin, G. H. Vurens, M. Salmeron and G. A. Somorjai, Langmuir, 1987, 3, 303–304.
49. J. Soria, J. C. Conesa, V. Augugliaro, L. Palmisano, M. Schiavello and A. Sclafani, J. Phys. Chem., 1991, 95, 274–282.
50. S. Sun, X. Li, W. Wang, L. Zhang and X. Sun, Appl. Catal., B, 2017, 200, 323–329.
51. K. T. Ranjit and B. Viswanathan, Indian J. Chem., Sect. A: Inorg., Bio-inorg., Phys., Theor. Anal. Chem., 1996, 35, 443–453.
52. M. Liu, R. Inde, M. Nishikawa, X. Qiu, D. Atarashi, E. Sakai and Y. Nosaka, ACS Nano, 2014, 8, 7229–7238.

---

Footnote

† Electronic supplementary information (ESI) available: SEM, BET, XRD, TEM, colour change of samples, density of states (DOS) of the (001) plane, ion chromatography images and tables of oxygen formation energy, atomic populations (Mulliken) of the Mo₄O₁₂ unit cell and Mo₄O₁₁ unit cell, and summary of some metal oxides. See DOI: 10.1039/c8cy02357c

---