Ammonia (NH₃) is not only an essential chemical in producing various chemicals, like fertilizers, in modern society, but it is also an important clean energy carrier and fuel. It is currently manufactured by the well-known industrial Haber-Bosch process using H₂ and N₂ under extremely high-pressure and high-temperature conditions, typically over an iron-based catalyst [1-5]. The discovery of the Haber-Bosch process for ammonia synthesis in 1909 resulted in 4 times multiplication of crop yield, which enabled the global population to nearly quadruple since the rapid implementation of the process in the early 20th century [6, 7]. However, the dominance of the Haber-Bosch process leads to massive consumption of energy, ~1%-2% of the world's annual energy consumption, and generates large amounts of CO₂, putting enormous pressure on the world's energy demand and environmental issues [8]. Therefore, it is of great significance to develop green and sustainable strategies for NH₃ synthesis, especially using renewable energy under ambient conditions. Until now, various strategies that can be operated under mild conditions, including electrocatalysis [9-12] and photo(electro)catalysis [8, 13-16], have been explored in regard to N₂ reduction for NH₃ synthesis, and some promising progress has been achieved. Artificial photosynthesis of ammonia directly from sunlight, N₂ and H₂O via photocatalysis, is regarded as an ideal, energy-saving and environmentally-benign process for NH₃ production since it can be performed at normal temperature and atmospheric pressure using renewable solar energy. The study of nitrogen fixation via artificial photosynthesis was first reported in 1977 by Schrauzer et al. [17], who employed the model photocatalyst TiO₂ as a proof-of-concept to convert N₂ into NH₃ and to oxidize H₂O to oxygen simultaneously under UV light irradiation. Since then, the research in this field had attracted many attentions in the last century, but further study has become much more challenging. With the increase in demands for renewable energy in the world, photocatalytic nitrogen fixation for ammonia synthesis has recently gathered momentum in terms of development.

The principle of artificial photosynthesis for nitrogen fixation is similar to photocatalytic water splitting and CO₂ reduction [18-21] For artificial photosynthesis on semiconductor-based photocatalysts, the first step is photoexcitation, in which the electrons are excited onto the conduction bands leaving the photogenerated holes remain within valence band. Following is the dissociation of the photogenerated charges, so that the electrons and holes separate and diffuse to the surface of photocatalysts (Fig. 1). Finally, the electrons and holes participate in the redox reactions with the adsorbed molecules (e.g., N₂, H₂O) onto the reactive sites for the production of ammonia. Thermodynamically, photocatalytic N₂ fixation for NH₃ synthesis is an endothermic reaction, with a Gibbs free energy of 339 kJ/mol (Eq. (1)). In this reaction, H₂O is oxidized to O₂ and protons, while N₂ and protons are reduced to NH₃. As a result, NH₃ is produced from water and N₂ under ambient conditions, using sunlight as the energy source. Compared with water splitting reactions, the nitrogen fixation needs more energy to drive the reaction and is much more challenging than water splitting. The whole reaction can be separated into two sub-reactions, the initial reaction being H₂O splitting to H₂ and O₂, and the second being NH₃ synthesis using H₂ and N₂. Therefore, the first step for artificial photosynthesis for N₂ fixation is photocatalytic water oxidation to generate O₂ and protons, and the prerequisite for a potential photocatalyst is that it should satisfy the energy requirement for efficient water oxidation.

(1)

The key reactions in artificial photosynthesis, that is photocatalytic water splitting, CO₂ reduction and nitrogen fixation, share many common challenges. For example, the primary step, water oxidation, is the most challenging, which directly determines whether the reactions can occur or not, and how efficient they will be. Therefore, a photocatalyst, which possesses good water oxidation ability and suitable band structures, is the prerequisite for all these artificial photosynthesis reactions. For CO₂ reduction and N₂ fixation, one of the most challenging issues is the activation of CO₂ and N₂ molecules, both of which are chemically stable molecules and need a high input energy to activate them. Therefore, CO₂ reduction and N₂ fixation reactions are more difficult than photocatalytic water splitting, and the reported efficiencies are consequently much lower than those of water splitting.

For the research of artificial photosynthesis for nitrogen fixation via photocatalytic processes, many semiconductor-based photocatalysts have been investigated and reported to be active, e.g., titanium oxides [15, 22], bismuth oxyhalides [23, 24], bismuth oxides [25], carbon nitrides [26], cadmium zinc sulfides [27], cadmium sulfides [28], and even diamond [29]. One of the big challenges in photocatalytic nitrogen fixation is the adsorption and activation of inert nitrogen molecules under ambient conditions, which is recognized to be the rate-determining step, as the cleavage of the N≡N triple bond needs an extremely high dissociation energy (~941 kJ/mol).

Oxygen vacancy-based semiconductors, with abundant localized electrons in oxides, can act as active sites and exhibit superior performances in various photocatalytic reactions [30-34]. Oxygen vacancies have been reported to be important in enhancing the adsorption and activation of O₂ molecules; oxygen vacancies in TiO₂ are able to activate O₂ to reactive oxygen species, such as peroxide and superoxide species [34, 35] Vacancies and defect states always coexist in a specific photocatalyst, and there are many different opinions when it comes to the intrinsic role of oxygen vacancies or defect states in photocatalysis. In some cases, vacancies or defect states on the surface of photocatalysts can act as trapping sites for photogenerated electrons or protons to inhibit the charge recombination, so that the surviving charges can participate in the following photocatalytic reactions.

Oxygen vacancies have been demonstrated to be essential in the adsorption and activation of nitrogen molecules [16, 23]. Li et al. [23] introduced oxygen vacancies in visible-light-responsive BiOBr nanosheets exposed with {001} facets, and found that the designed BiOBr photocatalyst displayed efficient photocatalytic activity in nitrogen fixation to ammonia, even without any cocatalysts and sacrificial reagents. The unit of BiOBr is a layered structure composing of [Bi₂O₂] slabs interleaved with double slabs of bromine atoms. Its {001} facets are terminated with high-density oxygen atoms, which create oxygen vacancies on the surface when solvothermal treatment was used to synthesize BiOBr nanosheets (Fig. 2). The work revealed that oxygen vacancies of BiOBr could activate N₂ molecules by elongating the N≡N triple bond from 1.078 Å in the original N₂ molecule to 1.133 Å in the N₂ molecule, which had been adsorbed onto the oxygen vacancies, via an end-on configuration. By means of this activation, electrons from the conduction band of photoexcited BiOBr could be facilely injected into the π anti-bonding orbitals of the adsorbed N₂. As a result, N₂ reduction to ammonia, catalyzed by oxygen vacancies in BiOBr, required an ultralow energy barrier and the vacancy-containing BiOBr nanosheets show a very high photocatalytic activity for ammonia synthesis under visible light irradiation using N₂ and H₂O as the reactants, however, no ammonia was produced on BiOBr nanosheets without the presence of oxygen vacancies. Such vacancy-containing BiOBr nanosheets, with the availability of localized electrons for π-back-donation, have the ability to adsorb and subsequently activate N₂ molecules, which can thus be efficiently reduced to NH₃ by the interfacial electrons transferred from the excited BiOBr nanosheets.

Oxygen vacancies, acting as active sites for N₂ fixation, can also be generated by light treatment of photocatalysts. Very recently, Wang et al. [36] reported that Bi₅O₇Br nanotubes with sufficient oxygen vacancies, induced by visible light irradiation, could realize efficient and stable photofixation of atmospheric N₂ into NH₃ in pure water, giving an apparent quantum efficiency over 2.3% at 420 nm. As shown in Fig. 3, for Bi₅O₇Br nanotubes without light irradiation, no obvious signal in the electron spin resonance characterization was observed, however, a very strong signal of oxygen vacancies can be observed under light condition. It was also found that the performance of photocatalytic N₂ fixation is closely correlated with the concentration of oxygen vacancies. Such a photo-induced variation of oxygen vacancies on Bi₅O₇Br nanotubes plays a vital role in the activation and fixation of nitrogen molecules. The proposed mechanism for the whole process can be summarized into four steps [36]: (1) part of the O atoms will escape in the form of O₂ from the surface of Bi₅O₇Br, creating surface oxygen vacancies under visible light irradiation; (2) nitrogen molecules are chemisorbed and activated on the sites of the oxygen vacancies; (3) the photoexcited electrons are injected into the activated nitrogen molecules to form ammonia; (4) after the reaction, the oxygen vacancies can be refilled by seizing O atoms from water, leading to a recovery of the original state to complete a circulation. It can be seen that the oxygen vacancies in this reaction have a similar role to that of catalysts and intermediates in catalytic reactions.

For photocatalytic N₂ fixation on the model photocatalyst TiO₂, both oxygen vacancies and Ti³⁺ species, existing in the bulk or on the surface, have been reported to be significant in activating N₂ molecules in photocatalytic ammonia synthesis [22]. Hirakawa et al. [22] found that oxygen vacancies and Ti³⁺ species are inherently created on the surface defects of commercially available TiO₂ and behave as active sites for photocatalytic N₂ reduction, which greatly promotes the efficient reduction of N₂ to NH₃; the solar-to-chemical energy conversion efficiency of this system is much higher than other reported ones. Fig. 4 shows the intrinsic mechanism of how Ti³⁺ species and oxygen vacancies activate the N₂ molecules on the surface of rutile TiO₂. Taking the rutile (110) surface as an example, it is characterized by alternating rows of 5-fold coordinated Ti⁴⁺ and bridging O (O_b) that run in the (001) direction. Surface defects are the O_b vacancies, where two excess electrons associated with O_b are transferred to the empty 3d orbitals of the neighboring Ti⁴⁺, resulting in the generation of two exposed Ti³⁺. The donor levels of these Ti³⁺ lie at 0.1-0.3 eV below the conduction bands of TiO₂ and, therefore, they can act as trapping sites for the photogenerated electrons. The yield of NH₃ under photocatalytic conditions can be enhanced by increasing the flow rate of N₂ gas, and the photocatalytic activity is much higher than the majority of the reported photocatalyst systems. Evidence of Ti³⁺ species acting as active sites in the reaction is provided by electron spin resonance (ESR) analysis. The Ti³⁺-containing TiO₂ catalyst shows a distinctive ESR signal located at g = 2.004, which can be assigned to the bridging oxygen (O_b) vacancies at -196 ℃ in a vacuum. After adding a certain amount of N₂ in the system, the ESR signals completely disappeared, indicating that the surface Ti³⁺ leads to adsorption of N₂ via the electron donation from Ti³⁺. These results clearly show the irreplaceable roles that Ti³⁺ species and oxygen vacancies played in the adsorption and activation of N₂ molecules for photocatalytic N₂ fixation in NH₃ synthesis.

Layered double hydroxides (LDHs) display great potential in photocatalytic applications due to their unique layered structure, with versatility in composition, morphology and architecture. Zhao et al. [37] reported that oxygen vacancies could be easily generated on the surface/edge of ZnAl-LDH nanosheets when they were prepared to be a few nanometers thick, and the vacancy-containing ZnAl-LDH nanosheets are found to be active in photocatalytic CO₂ reduction. Very recently, they also found that, using simple co-precipitation methods, a certain amount of oxygen vacancies could be introduced within the ultrathin LDH nanosheets which lead to distortions in the MO₆ octahedra (Fig. 5(a)) [38]. A series of LDH nanosheet photocatalysts, strategically engineered with oxygen vacancies, can promote the chemisorption and activation of N₂ molecules at ambient temperatures and pressures; the CuCr-LDH nanosheet photocatalyst displayed an extraordinarily high photocatalytic activity for N₂ reduction to NH₃, both under UV and visible light irradiation (Fig. 5). As far as we know, this is the first photocatalyst that can drive photocatalytic N₂ reduction for NH₃ synthesis under wavelengths greater than 500 nm using H₂O directly as the proton source. The CuCr-LDH nanosheet, which contains oxygen vacancies, also exhibits excellent photo-stability under reaction conditions, with no obvious degradation after several successive cycles. The introduction of oxygen vacancies within the ultrathin LDH nanosheets results in the distortions in the MO₆ octahedra, which is responsible for promoting the adsorption and activation of N₂ molecules for NH₃ synthesis. This work not only verifies the essential role of oxygen vacancies for photocatalytic N₂ reduction, but also demonstrates that LDHs are a class of promising photocatalysts that are capable of NH₃ synthesis using visible light under mild conditions.

In natural systems, N₂ is fixed by the enzyme nitrogenase, which is responsible for catalyzing nitrogen fixation and reducing the energy barrier of N₂ to NH₃ so that the reaction can take place in ambient conditions. The natural nitrogenase complex consists of two proteins: the homodimeric Fe protein, and the heterotetrameric MoFe protein [39-41]. The function of the Fe protein is to transfer electrons from a reducing agent, such as ferredoxin or flavodoxin, to the MoFe protein, that is, it is responsible for the supply of electrons. The transfer of electrons requires an input of chemical energy, which comes from the binding and hydrolysis of ATP. The function of the MoFe protein is as a nitrogenase, which uses the electrons to reduce N₂ to NH₃, that is, the MoFe protein is the active site for N₂ reduction. However, in the popular Haber-Bosch process, nitrogen reduction for NH₃ synthesis is accomplished at high temperatures and pressures over metal or metal oxides catalysts, which is significantly different to natural nitrogen fixation. To overcome such an obstacle between natural and artificial nitrogen fixation, Brown et al. [28], inspired by nature, first introduced an inorganic semiconductor to mimic the role of the Fe protein in nitrogenase, combining an artificial photocatalyst, cadmium sulfide (CdS), with the nitrogenase MoFe protein for N₂ fixation. As shown in Fig. 6, CdS was used to photosensitize the nitrogenase MoFe protein, where light harvesting replaces ATP hydrolysis in the natural system to drive the enzymatic reduction of N₂ into NH₃. Such a hybrid system was demonstrated to be successful and efficient, and the turnover rate for NH₃ production was measured to be 75 per minute, which is more than 60% of the ATP-coupled reaction rate for the nitrogenase complex under optimal conditions. Although some problems still exist in this system, e.g., hydrogen production is the main competing reaction with a rate of almost 10 times that of ammonia production in this system, and the photocatalytic activities obviously degraded after several hours' irradiation, the system displayed a blueprint for the successful transfer of photogenerated charges from inorganic semiconductors to natural nitrogenase and can be further used for nitrogen reduction reactions. The work also highlights the possibility of creating hybrid systems with natural and artificial photocatalysts for photocatalytic N₂ fixation.

Understanding the role of the unique structure of natural nitrogenase and then mimicking it to construct artificial photocatalytic systems is beneficial for boosting research into artificial photosynthesis for N₂ fixation. Learning from natural nitrogenase, Liu et al. [42] synthesized a synthetic nitrogenase that mimicked the form of chalcogel, composed of molybdenum and iron-containing ([Mo₂Fe₆S₈(SPh)₃]) and single-cubane (Fe₄S₄) biomimetic clusters, that can accomplish photocatalytic N₂ fixation and its conversion to NH₃ at ambient temperatures and pressures (Fig. 7). They introduced a bottom-up synthesis of chalcogels, which revealed sponge-like and porous morphologies. To recognize the role of Mo in the catalytic mechanism, Mo-free chalcogels (Fe₄S₄) were also synthesized for comparison. They found that the redox-active iron-sulfide- containing materials, Mo₂Fe₆S₈(SPh)₃, could activate the N₂ molecules upon visible light irradiation, which can be reduced all of the way to NH₃ using protons and sacrificial electrons in aqueous solution. Surprisingly, Fe₄S₄ clusters, without Mo, exhibit a higher photocatalytic activity, indicating that Mo itself is not necessary to carry out this process [40]. The mechanism of photocatalytic conversion of N₂ over chalcogels was investigated by means of an in situ diffuse-reflectance Fourier transform infrared spectroscopy in combination with a moist stream of N₂ in isotopically labeled experiments. It can be clearly found that two dominant absorption bands at 1753 and 1746 cm^-1 appeared, however, both bands disappeared without a flow of N₂ or without light irradiation, indicating that the corresponding species are most probably the intermediates participating in the photocatalytic N₂ reduction reaction. These absorption bands can be attributed to the N-N stretching mode of the Mo-N=N structure. In the isotopically labeled experiments, when D₂O was used in place of H₂O, the absorption band at 1746 cm^-1 shifted to 1724 cm^-1, while the band at 1753 cm^-1 band remained unchanged, showing that hydrogen is only involved in the stretching mode of Mo-N=N species at 1753 cm^-1. The results provide an clear evidence that the produced NH₃ is derived from N₂ and H₂O under light irradiation. Both the CdS-MoFe hybrid system and Mo₂Fe₆S₈(SPh)₃ clusters consequently reinforce the belief that integrating natural catalysts or mimicking the unique structure of active sites in natural N₂ fixation processes can greatly expand the scope of material design and engineering for the exploration of highly efficient artificial photocatalysts for ammonia synthesis, which can be operated under ambient conditions.

The industrial production of ammonia consumes massive amounts of energy, ~1%-2% of the world's annual energy consumption, and generates large amounts of CO₂, therefore, exploring and establishing new technologies capable of capturing solar energy and producing fertilizers and fuels have become increasingly attractive and critical. Furthermore, artificial photosynthesis for nitrogen fixation also has the potential to significantly impact how we understand and engineer the nitrogen cycle, both in natural and industrial systems. Significant interests and progress in artificial nitrogen fixation have been made. Despite the substantial developments of N₂ fixation under ambient conditions so far, several fundamental challenges have emerged and require addressing in order to realize artificial photosynthesis for nitrogen fixation at a practical level. The reaction rates and turnover frequencies of photocatalytic nitrogen fixation at ambient temperatures and pressures reported to date have been far beyond the expectation, and the achieved efficiency for photocatalytic nitrogen fixation is at a very low level (the highest apparent quantum efficiency is only 2.3% at 420 nm until now, and the estimated solar-to-chemical conversion efficiency is less than 0.1%), which is far below the requirement for practical industrial applications.

Based on the knowledge of the author, several remarks about the development of artificial photosynthesis for ammonia synthesis in the future may be included as follows:

(1) The fundamental mechanism for photocatalytic nitrogen fixation requires better understanding in order to design and explore efficient photocatalyst systems. The widely-studied mechanism of ammonia synthesis in heterogeneous catalysis can provide useful guidance to understand the intrinsic processes that the photocatalytic reaction follows. Nitrogen reduction to ammonia on a heterogeneous catalyst surface is generally recognized to be controlled by two broad classes of mechanism: associative and dissociative. In an associative mechanism, the two nitrogen centers in N₂ remain bound to each other as the molecule is hydrogenated, with NH₃ being released only once the final N≡N bond is broken. In a dissociative mechanism, the N≡N bond is broken before any hydrogenation takes place, leaving individual N-atoms on the surface, which are converted into NH₃ independently [43]. In photocatalytic nitrogen fixation reaction, how and when the N≡N bond breaks and N-H forms requires clarification on several model photocatalysts, which will be beneficial and essential for the research of photocatalytic ammonia synthesis.

(2) Engineering oxygen vacancies in semiconductor-based photocatalysts has been demonstrated to not only affect the adsorption of N₂ molecules, but also benefit the activation of N₂ and H₂O molecules in the reaction. Rational construction of the oxygen vacancies or surface defect states, building the relation between vacancies and photocatalytic reactions, and investigating the intrinsic roles of these vacancies will be helpful not only for the reorganization of the reaction mechanism, but also for the design of novel photocatalyst systems for N₂ fixation. Natural nitrogenases in plants are most active for efficient fixation of N₂ gas in air under ambient condition, which is ascribed to its unique complex structure with molybdenum and iron-containing biomimetic clusters. Although synthesis of nitrogenase itself is not very easy, mimicking the core structure of its active sites and understanding the roles of every unit in the unique clusters will be a useful strategy. Considering the low solubility of nitrogen molecules in aqueous solutions, a big challenge is that it inevitably lowers the encountering frequency between N₂ and H₂O molecules at the surface of photocatalysts. Therefore, strategies for improving the N₂ adsorption and encountering frequency of reactant molecules are of equal importance in artificial photosynthesis for N₂ fixation.

(3) Exploring new materials with a wide light absorption range and suitable band gaps and developing new strategies to improve the adsorption and activation of N₂ molecules are still significant topics in the field of artificial photosynthesis for N₂ fixation. The establishment of new methodologies for synthesizing visible-light-responsive photocatalysts with novel structures for efficient photocatalytic N₂ fixation is desired. Morphology engineering, vacancy engineering and the modification of suitable cocatalysts to rationally tune the surface structures and active sites for N₂ adsorption and activation will be essential for constructing efficient artificial photocatalyst systems.

(4) Ammonia oxidation is an inevitable back-reaction in photocatalytic nitrogen fixation, which is indeed a big obstacle. The produced ammonia in the solution can be easily oxidized to nitrogen-containing compounds, such as NO₃^-, NO₂^-, by photogenerated protons in the presence of O₂. Moreover, the back-reaction could be inhibited via a photoelectrochemical cell with a proton-exchange membrane in-between, where water could be oxidized on a photoanode to produce protons and electrons which could then be used to reduce and protonate nitrogen to form ammonia at the cathode side. Other strategies for inhibiting the back-reaction of ammonia oxidation also require urgent development.

(5) Theoretical calculations are useful tools to simulate the intrinsic photocatalytic behavior and to gain a fundamental understanding of the intrinsic reaction mechanism. Many key scientific issues in artificial photosynthesis for ammonia production require consideration, including how the nitrogen adsorption and activation takes place on the surface of photocatalysts, how the dissociation of nitrogen triple bonds and the formation of ammonia occurs, etc. Although only a few relative studies on the theoretical simulations of photocatalytic nitrogen fixation are currently available, it can be anticipated that more research will focus on such a challenging, but attractive field.

(6) It should also be noticed that the generally used experimental method for the detection of NH₃ production in the reported literatures is Nessler's reagent, which may be easily affected by various additions to the solution (e.g., sacrificial regents, buffer solutions) and lead to error-prone results. To make the obtained experimental data more convincing, it is better to confirm NH₃ production by other methods, such as ion chromatography, and nuclear magnetic resonance, together with isotropic labeled experiments. Furthermore, new methods with high sensitivity for the accurate measuremtent of low amounts of NH₃ are urgently required.

It is promising that the continuous development and eventual realization of efficient approaches for artificial photosynthesis N₂ fixation under ambient conditions using sunlight shows strong potential for producing NH₃, while significantly reducing global energy costs and carbon emissions. As more and more progresses in the key challenges in this field are gradually achieved, it is likely that artificial photosynthesis for ammonia synthesis may lead to a dramatic change in the present infrastructure of the Haber-Bosch process, toward a more convenient and widely-available process in the future.