Rational design of photocatalysts for ammonia production from water and nitrogen gas

Photocatalytic N2 reduction has emerged as one of the most attractive routes to produce NH3 as a useful commodity for chemicals used in industries and as a carbon-free energy source. Recently, significant progress has been made in understanding, exploring, and designing efficient photocatalyst. In this review, we outline the important mechanistic and experimental procedures for photocatalytic NH3 production. In addition, we review effective strategies on development of photocatalysts. Finally, our analyses on the characteristics and modifications of photocatalysts have been summarized, based on which we discuss the possible future research directions, particularly on preparing more efficient catalysts. Overall, this review provides insights on improving photocatalytic NH3 production and designing solar-driven chemical conversions.


Introduction
Recently, there has been an increasing focus on ammonia (NH 3 ) as not only a key commodity for chemicals widely used in industries [1,2], but also as a liquid energy carrier [3] that enables transport and supply of hydrogen (H 2 ) gas through cracking. Furthermore, NH 3 is an alternative carbon-free energy source that can be utilized in energyconversion devices, for example, direct NH 3 fuel cells [4,5]. Traditionally, NH 3 production has mainly relied on the Haber-Bosch process, which is energy intensive and generates considerable volume of CO 2 into the atmosphere because of these requirements, along with the needs for H 2 and the extreme operation condition [6].
Photochemical NH 3 production using N 2 and water as a hydrogen source at ambient temperature and pressure can offer a promising alternative route that is less energy intensive and reduces CO 2 emission, thereby significantly reducing environmental concerns [7]. NH 3 production paired with water oxidation is a thermodynamically uphill reaction; therefore, external energy input is necessary [8]. Photocatalysts can utilize solar energy, which is supplied to Earth with sufficient solar power and abundance and facilitate NH 3 production. Since TiO 2 photocatalysts applied to N 2 reduction, new photocatalysts and their modifications have been intensively suggested to improve photocatalytic NH 3 production [9].
Major advances in metal oxides such as TiO 2 , WO 3 , and SrTiO 3 , bismuth oxyhalides, and polymeric carbon nitride photocatalysts have resulted in interests in the development of photocatalysts and their applications for NH 3 production [10][11][12][13][14]. This review is composed of mainly two sections: (1) fundamentals for photocatalytic NH 3 production, (2) strategies to develop photocatalysts. The first section focuses on the principles and mechanisms involved in the overall photocatalytic reaction. The second section contains recent progresses in the development of catalysts and their major advantages with regard to catalytic performance. The insights and discussions provided in this review will serve as useful resources for further development of photocatalysts and provide future research direction for photocatalytic NH 3 production.
2 Fundamentals for photocatalytic NH 3 production 2.1 Principle and mechanism NH 3 production by N 2 reduction (N 2 (g) + 6H + + 6e − ⇌ 2NH 3 (g), E 0 = 0.092 V vs. SHE) paired with O 2 production by water oxidation (2H 2 O(l) ⇌ O 2 (g) + 4H + + 4e − , E 0 = 1.229 V vs. SHE) is a thermodynamically uphill reaction that requires a potential of at least 1.137 eV [15] per electron [16]. Moreover, three electrons are required to produce an NH 3 molecule, and four holes are required to produce an O 2 molecule according to eqns. (1) and (2): A simplified schematic diagram for photocatalyic NH 3 production and challenges in designing photocatalysts are summarized in Fig. 1. For overall NH 3 production, photons with energies > 1.137 eV are required to generate photoexcited electron-hole pairs that carry out the desired redox reactions at the active sites on the surfaces of the photocatalysts. Thus, a photocatalyst must have an energy band gap, (E g = E CBM -E VBM , where CBM and VBM represent conduction band minimum and valence band maximum), larger than the energy required for the overall NH 3 production. In addition, a photocatalyst must have a suitable conduction and valence band alignment to drive two half-reactions using photoexcited electrons (e − ) and holes (h + ) under solar light irradiation.
The major factors to consider for photocatalyst designs are (1) photon absorption followed by photoexcited carrier generation, (2) their migration and separation, and (3) their consumption by redox reactions on the surfaces of photocatalysts. The dominant photon energy spectrum obtained from the sunlight is in the range of 390-700 nm. Furthermore, because photocatalytic redox reactions require kinetic overpotentials, photocatalysts have a bandgap energy in the range of 1.6-2.4 eV. The generated photoexcited carriers can migrate to active sites on the surface where redox reactions carried out. However, since the kinetics of carrier recombination in the interiors of the photocatalysts is fast, considerable amounts of photoexcited carriers are recombined [17]. Moreover, photoexcited carriers can be trapped in the Absorption of light : Nitrogen reduction (NRR) : interior and surface of the photocatalyst, which causes photocorrosion. Thus, a photocatalyst design to improve electron-hole pair migration and separation is important for further development of photochemical NH 3 production [18]. Most semiconductors that act as light absorbers in photocatalysts are inactive in the redox reactions with the photoexcited carriers; thus, the active sites on the surface of semiconductor are decorated with cocatalysts [19]. For example, cocatalysts, such as RuO x , CoO x , and Co-pi [20][21][22], combined on the surface of the semiconductor for facile water oxidation by providing active sites. Another critical point to note in NH 3 production via photochemical N 2 reduction is the facile competitive hydrogen evolution reaction by water reduction. NH 3 production is kinetically much more complex than water reduction, although the thermodynamic requirements for those reactions are similar [23]. This indicates that photochemical H 2 production may be predominant, resulting in inactive NH 3 production. Thus, it is critical to modify the surface that provides active sites for the desired redox reaction using suppressing competitive reactions on the semiconductors.
It is also necessary to note the importance of rate balance for photocatalytic reactions on photocatalysts. Since electrons and holes are generated in pairs under illuminations, their consumption must be coupled as well. For instance, if oxidation reactions involving photogenerated holes are sluggish, the total NH 3 production rate should be similar. In this case, scavengers, such as ethanol, methanol, and sulfites, can be employed as sacrificial reagents to facilitate oxidation reaction using photoexcited holes, thereby preventing from photoexcited carrier recombination [24]. However, since scavengers may interfere with counter reactions, here N 2 reduction, or product evaluation methods, the target scavenger should be chosen carefully. It is discussed in detail in the next section.

Experimental process
The simplest photocatalytic NH 3 production set-up is a particulate PC system where photocatalysts are dispersed in a medium, typically pure water, with N 2 bubbling under simulated solar light illumination, as shown in Fig. 1b. NH 3 and O 2 are generated on a single photocatalyst. The produced NH 3 is technically evaluated by the spectrophotometric method via the indophenol blue method or quantified using proton nuclear magnetic resonance spectroscopy ( 1 H-NMR) with isotope labeled nitrogen 15 N as shown in Fig. 2.
The indophenol blue method is based on the principle of the Berthelot reaction, where typically, salicylic acid, citrate, hypochlorite, and ferricyanide are used in an alkaline solution. In the presence of NH 4 + , the color of the yellowish indicating solution, which contains the reagents, turns to blue, corresponding to 655 nm. As shown in Fig. 2a and b, thus, the concentration of NH 3 can be quantified by measuring the absorbance change at 655 nm using a UV-Vis spectrophotometer [28]. However, to quantify the concentration of NH 3 generated by photocatalytic N 2 reduction, multiple possible contaminants that interfere with the evaluation should be fully controlled [28]. For example, methanol, a widely used as a hole scavenger, and its derivatives cause incorrect color changes and significantly decrease the accuracy of NH 3 quantification as shown in Fig. 2c and d. Another critical point for NH 3 , which is photocatalytically produced, quantification is to eliminate the parts from NH 3 contaminants, unintentionally present introduced through catalyst, components in set-up [26] or even the air. In fact, NH 3 present in human breath, sanitizers, and so can rapidly accumulate in water [29].
The most rigorous procedure to quantify photocatalytically generated NH 3 from N 2 reduction is to use isotope labeled 15 N 2 gas. The photocatalytic reactions are conducted in a simple particulate system with 15 N 2 Fig. 1 a The schematic illustration of photochemical nitrogen reduction reaction [15]. b The challenges and strategies for photocatalytic NH 3 production in particulate photocatalytic (PC) ammonia production system bubbling. Since the splitting of 1 H resonance in 15 NH 4 + and 14 NH 4 + resulting from the difference in the scalar interaction between N ( 15 N or 14 N) and H is differentiated, the 15 NH 4 + produced, which is dissolved in an aqueous reacting solution, can be quantified by 1 H-NMR [30]. For instance, the 1 H resonance in 15 NH 4 + can be split into two symmetric signals with a spacing of 73 Hz, whereas the resonance in 14 NH 4 + can be split into three symmetric signals with a spacing of 52 Hz as shown in Fig. 2e. Thus, false positive evaluations for the produced NH 3 can be prevented effectively.

Strategies to develop photocatalysts
This section is divided into mainly two parts: (1) semiconductors comprising photocatalysts, which have been widely applied as a light absorber in Fig. 3a, and (2) (black), respectively. Reproduced with permission [27]. Copyright 2021, The Royal Society of Chemistry. f The peaks for 14 NH 4 + corresponding to its concentration. Reproduced with permission [8]. Copyright 2020, American Chemical Society strategies to modify the catalysts, which have been effectively used to further increase the photocatalytic NH 3 production.

Semiconductors
Titanium(Ti)-based oxides Titanium dioxide, TiO 2 , possesses several advantages that make it attractive for use in photocatalytic NH 3 production. TiO 2 has a bandgap of ~ 3.2 eV, which absorbs UV light, and a CBM and VBM that offer sufficient overpotential for the photoexcited electrons to reduce N 2 to NH 3 and holes to oxidize water to O 2 , respectively. In addition, the inexpensive, environmentally benign, and stable nature of TiO 2 is an additional advantage. Historically, the primary limitation of using TiO 2 has been rapid photoexcited electron-hole recombination, which is similar to many other oxidebased photocatalysts. Strontium titanate, SrTiO 3 [31], is another candidate showing two major benefits over c The color of Ni-x-TiO 2 gradually deepens with the increase of the amount of Ni 2+ . d The N 2 -TCD spectrums of Ni-0.8-TiO 2 (red) and TiO 2 (black), respectively. e Bader charge density difference of Ni doped TiO 2 ; Ti(silver), Ni(blue), and O(red). Reproduced with permission [45]. Copyright 2020, Americal Chemical Society TiO 2 for use in the NH 3 production. The CBM of SrTiO 3 is slightly more negative than that of TiO 2 , thereby providing a stronger driving force to reduce N 2 using the photoexcited electrons. In addition, intrinsic charge mobility in SrTiO 3 is higher than that in TiO 2 , thereby suppressing photoexcited charge recombination [32]. However, despite this thermodynamic advantage, the photocatalytic NH 3 production using TiO 2 or SrTiO 3 is limited because of ineffective light absorption resulting from their wide bandgaps and inactive redox reactions due to the limited active sites on their surface. Recently, many attempts have been made to address these limitations, which include doping, coupling with electrocatalysts (cocatalysts), surface defect engineering, and so on [33][34][35].
Bismuth oxyhalide (BiOX) BiOX (X = Cl, Br, and I) has a layered structure where bismuth and oxygen layers (BiOx) are connected with halide (X) layers alternatingly via Van der Waals forces [36]. Since the VBM is determined by O 2p and X δ p ( δ = 3, 4, and 5, corresponding to X = Cl, Br, and I, respectively) and the CBM is mainly evaluated by Bi 6p, their band alignments and bandgap are varied depending on the element comprising the semiconductors [37]. For example, BiOCl, BiOBr, and BiOI have bandgap energies of 2.92, 2.65, and 1.75 eV, respectively [38], and are thermodynamically suitable for NH 3 production, they enable the production of NH 3 and O 2 by the overall photocatalytic N 2 reduction. However, despite these advantages, BiOX semiconductors have severe drawbacks when using them as photocatalysts, mainly due to the insufficient photogenerated charge separation in the bulk and surface. Recently, their fundamental limitations have been investigated, and strategies to improve their photocatalytic performances have been explored.
Graphitic carbon nitride (g-C 3 N 4 ) g-C 3 N 4 is a polymeric semiconductor with a bandgap energy of 2.7 eV and appropriate positions for the CBM and VBM for the NH 3 and O 2 production, respectively [39]. However, since rapid photoexcited carrier recombination in the bulk results in limited charge separation in the redox reactions and ineffective N 2 adsorption on the surface, as an initial step for N 2 reduction, multiple advancements to develop g-C 3 N 4 have been suggested [40].

Strategies to enhance photocatalytic NH 3 production
As described in the introduction section, photocatalysts comprising semiconductors can produce NH 3 using N 2 and water under illumination. However, the typical NH 3 production rate and efficiency of intrinsic semiconductors using solar energy have been far lower than those expected from their bandgap energies. The major limiting factors include (1) limited light absorption because of the intrinsically large bandgaps, (2) significantly fast photoexcited electron-hole pairs recombination, and (3) slow redox reactions including N 2 reduction and water oxidation. Recently, many strategies have been made to address one or more of these challenges, which include engineering defects by extrinsic doping or vacancy introducing, composites forming using multi-light absorbers, and coupling with electrocatalysts. This section will give an overview of each of these attempts and discuss how these strategies affected the photocatalytic NH 3 production.

Defect engineering
Defect engineering has been widely suggested to control bulk and surface properties of catalysts, for example, by heteroatoms doping, vacancies formation, and so on. This defect engineering enhances photocatalytic NH 3 production by changing the light absorption properties, improving charge separation efficiencies, or facilitating surface redox reactions, and it generally influences the multi-mechanistic steps of light harvesting simultaneously [41].
Heteroatom doping Heteroatom doping involves intentional introduction of impurity atoms into the lattices of materials to change their optical and electrical properties [42][43][44]. Depending on the species and degree (concentration) of dopants, changes, in the bandgap energy, band alignment, charge separation efficiency, and N 2 adsorption ability are expected. However, since dopants can provide electron-hole recombination sites or prohibit charge transports, identifying types and concentrations of dopants appropriate for the photocatalytic process is important.
Zaicheng Sun and coworkers prepared a nickel-doped TiO 2 (Ni-x-TiO 2 , where x = concentration of the Ni precursor controlled in the preparation step) photocatalyst using the sol-gel method [45] show Their optimized sample (Ni-0.8-TiO 2 ) exhibited an NH 3 production rate of 46.80 µmol·g −1 ·h −1 under simulated solar light irradiation, which is seven times higher than the rate using pure reference TiO 2 . The Ni-0.8-TiO 2 sample extends the light absorption range to the visible light region with a bandgap energy of 2.92 eV, whereas pure TiO 2 absorbs only UV light with a bandgap energy of 3.2 eV as shown in Fig. 3b. As Ni atoms having 2 + valence that replace the Ti site having 4 + valence, oxygen vacancies (Ov) are created naturally for charge neutrality within the host semiconductors, and they vary the band position of the CBM and VBM in TiO 2 , which results in a decrease in the bandgap energy. Besides changes in the optical properties, N 2 adsorption, the initiation step for NH 3 production from N 2 , is enhanced by Ni doping on TiO 2 , as investigated using a N 2 temperature programmed desorption spectroscopy (N 2 -TPD) and a computational analysis.
Tierui Zhang and coworkers prepared a copper-doped TiO 2 nanosheet (x%-TiO 2 , where x = molar ration of Cu/Ti controlled in the preparation step) photocatalyst using a hydrothermal method [46]. The optimized sample, 6%-TiO 2 , exhibited an NH 3 production rate of 78.9 µmol·g −1 ·h −1 , whereas pristine TiO 2 nanosheets exhibited a rate of 0.34 µmol·g −1 ·h −1 . As Cu heteroatoms replaced the sites of Ti in the TiO 2 nanosheets, small amounts of Ti 3+ valence states in Ti 4+ -dominant TiO 2 and Ov sites are created, as observed using X-ray photoelectron spectroscopy and X-ray absorption fine structure spectroscopy. Furthermore, owing to the difference in size between the Cu 2+ dopants and Ti 4+ hosts, Cudoped TiO 2 has a compressive strain, which is evaluated by DFT calculations, and it causes changes in the distribution in the electron densities around O and Ti in the materials. The combination of non-stoichiometry and lattice strain in Cu-doped TiO 2 narrows its bandgap energy, corresponding to the extended light absorption range of 600-800 nm, and electrons are accumulated around the sites of O atoms, thereby affecting the changes in the N 2 adsorption affinity.
Jing Zhang and coworkers prepared a Fe-doped SrTiO 3 (Fe x Sr 1-x TiO 3 , where x = the stoichiometric ratio of Fe precursor and Sr precursor controlled in preparation step) photocatalyst using the hydrothermal method followed by calcination [47]. The Fe 0.1 Sr 0.9 TiO 3 sample exhibited the best NH 3 production rate of 30.1 μmol·g −1 ·h −1 , which is 3.2 times higher than that of pristine SrTiO 3 . By replacement of the Fe 3+ ion, which has a smaller size, on the Sr 2+ site in SrTiO 3 , the size of particles is reduced along with the corresponding increase in the surface area. Furthermore, the Fe 0.1 Sr 0.9 TiO 3 photocatalyst exhibits significantly enhanced N 2 chemisorption and activation ability, which Fe dopants are the major contributor to.
Vacancies Vacancy, one of defect formation strategies, leads to changes in the band structure and chemical adsorption nature on the surface [48]. This section gives an overview of intentionally induced anion vacancies in semiconductor photocatalysts and discusses how they affected the photocatalytic NH 3 production.
Zhong Jin and coworkers compared the NH 3 production activities using BiOBr semiconductors in the presence (Vo-BiOBr) or absence (BiOBr) of oxygen vacancies [49]. The Vo-BiOBr samples were prepared using the hydrothermal method with the addition of polyvinylpyrrolidone surfactants leading to in-situ generation of abundant vacancies, whereas pristine BiO-Brs were synthesized without adding the surfactants. The Vo-BiOBr photocatalysts produced NH 3 at a rate of 54.70 μmol·g −1 ·h −1 under UV-Vis irradiation, which is 10 times higher than the production rate using BiOBr. The major contribution of this significant improvement is enhanced N 2 adsorption that initiates the intermediate formation for NH 3 production. For example, the amount of adsorbed N 2 , estimated based on N 2 adsorption isotherms, on Vo-BiOBr is considerably higher than that using BiOBr as shown in Fig. 4a. Furthermore, by introduction of the oxygen vacancies, the bandgap energy is reduced with the shifts in the CBM and VBM positions, which leads to increased light absorption as shown in Fig. 4b.
Chuanyi Wang and coworkers prepared metal-free g-C 3 N 4 semiconductors by thermal decomposition of melamine and induced nitrogen vacancies (V-g-C 3 N 4 ) within the semiconductors by additional calcination under a N 2 atmosphere [50]. The V-g-C 3 N 4 samples produced NH 3 at a rate of 1240 μmol·g −1 h −1 using pure N 2 and water under visible light irradiation, whereas pristine g-C 3 N 4 produced negligible amounts of NH 3 . Further, morphologies and bandgap energies were not significantly varied depending on the intentional creation of N vacancies; however, two factors changed considerably. Due to the presence of N vacancies on the semiconductor, the signal collected using photoluminescence spectra decreased, indicating the reduction in the carrier recombination and enhancement of their separation. Furthermore, the N vacancies enable selective adsorption and activate N 2 , and it was experimentally confirmed that negligible amount of NH 3 produced on g-C 3 N 4 selectively blocked the vacancies using Pd particles.

Coupling with multi-light absorbers
Coupling with multi-light absorbers to form heterojunctions has been widely investigated to improve light absorption capacity under single illumination and separation of photoexcited electron-hole carriers through their designed band alignments. This strategy for catalyst design is effective in improving the photocatalytic NH 3 production.
Zhong Jin and coworkers prepared Bi 2 MoO 6 /Ov-BiOBr composites, where Ov denotes the oxygen vacancies using two-step methods, and synthesized Bi 2 MoO 6 by solution-phase reflux process followed by the hydrothermal process to couple Ov-BiOBr on as-prepared Bi 2 MoO 6 [51]. Despite minimal difference in light absorption before and after coupling, the Bi 2 MoO 6 /Ov-BiOBr composites showed a significantly enhanced NH 3 production rate (90.7 μmol·g −1 h −1 ), which is almost 30 and 3 times higher than the rates evaluated using Bi 2 MoO 6 and Ov-BiOBr, respectively as shown in Fig. 5a. The major role of coupling is the formation of cascaded band alignments at the interface between two semiconductors. The photoexcited holes are transported and oxidize the electron donor on Bi 2 MoO 6 where the VBM is located relatively higher, whereas the photoexcited electrons are transported and reduce N 2 on Ov-BiOBr where the CBM located relatively lower. Furthermore, oxygen vacancies on BiOBr show synergistic effects on the NH 3 production by providing N 2 adsorption and activation sites as shown in Fig. 5b. Jimmy Yu and coworkers prepared Ov-TiO 2 using the hydrothermal method followed by loading Au nanoparticles. The Au nanoparticles generate hot electrons under visible light illumination [51], which is known as plasmonic phenomena. The hot electrons are injected into the CBM of TiO 2 and then trapped in vacant sites where N 2 is reduced into NH 3 . The generated holes remain on Au and oxidize the electron donors, methanol, in solution as shown in Fig. 5c and d. The optimized Au/TiO 2 -Ov samples produced NH 3 with a rate of 78.6 μmol·g −1 ·h −1 under visible light irradiation, which is at least 98 and 35 times higher those using Au/TiO 2 and TiO 2 -Ov photocatalysts, respectively.

Surface modification and reaction engineering
As discussed above, photocatalytic activities have been developed significantly by modifying photocatalysts, for example, defects engineering and composite formation. However, although photogenerated charge generation and separation are advanced by multiple possible strategies, if the surfaces of the semiconductors exhibit slow charge injection and are poorly catalytic for N 2 reduction, the NH 3 production rate may not be enhanced directly. The surface modification on the semiconductors using a robust electrocatalyst (cocatalyst) for N 2 to NH 3 conversion is the simplest way to improve the limiting factor at the interface between the semiconductors and liquid where reactants are present. Therefore, the modification of the semiconductor surface with various NH 3 production catalysts, such as Ru, Cu, and Au, has been investigated [20,34,53]. In contrast, since the rates of reduction and oxidation reactions always correspond to one another, the sluggish photo-oxidation using holes may become a limiting factor for the photocatalytic NH 3 production. To overcome the possible limitation, it has been widely suggested to load appropriate electrocatalysts (cocatalysts) for water oxidation on the semiconductor or provide a more facile hole acceptor (electron donors) than water, such as methanol or ethanol (Table 1).

Summary and outlook
In summary, we reviewed the mechanistic and experimental steps for photocatalytic NH 3 production and examined various strategies to modify photocatalysts employed to enhance the production activities. This review clearly shows that although significant development of photocatalyts has been reported thus far, further advances for practical NH 3 production are required. First, the efficiency for charge generation via solar light absorption should be enhanced. Considering that the majority of incident photons are in the range of visible light, development of absorbers, can harvest the incident solar light, is required by determining appropriate materials and engineering semiconductors.
The charge separation effect is also required for photocatalytic NH 3 production through suppression of their recombination. For example, strategies for doping and composite formation will exhibit advances in the future and may increase the photon absorption efficiency simultaneously. In addition, morphology engineering [72][73][74][75][76] for photocatalysts will be promising and can be achieved by controlling the distances for carrier transportation and lifetime duration between the photo-induced generation and recombination. Furthermore, this may change facets and surface area, thereby providing active sites for the redox reactions using transported carriers. Moreover, if the facets are optimized with increasing surface area, morphology engineering can be significantly beneficial for NH 3 production.
With improvements in photon absorption and charge separation, surface engineering will be able to provide active sites where N 2 reduction using electrons and oxidation using holes is important. Thus, defect engineering, for example, formation of vacancies, cocatalyst coupling, and applying sacrificial reagents have been suggested. It is noteworthy that N 2 diffusivity control close to photocatalysts make an important contribution to practical photocatalytic NH 3 production. For instance, . This indicates that even if highly robust catalysts are prepared, if the supply of N 2 gas and water is insufficient at the region close to the catalyst, the NH 3 production is limited. Thus, environmental engineering, for example, via hydrophobicity control on the catalysts, will be significantly effective. The mechanism, experimental methods, discussion, insights, and suggestions contained in this review provide a good fundamental and foundation for further improvement of photocatalytic NH 3 production. SC  Nessler's reagent [70]