Alternative Strategies Toward Sustainable Ammonia Synthesis

As one of the world’s most produced chemicals, ammonia (NH3) is synthesized by Haber–Bosch process. This century-old industry nourishes billions of people and promotes social and economic development. In the meantime, 3%–5% of the world’s natural gas and 1%–2% of the world’s energy reserves are consumed, releasing millions of tons of carbon dioxide annually to the atmosphere. The urgency of replacing fossil fuels and mitigating climate change motivates us to progress toward more sustainable methods for N2 reduction reaction based on clean energy. Herein, we overview the emerging advancement for sustainable N2 fixation under mild conditions, which include electrochemical, photo- , plasma-enabled and homogeneous molecular NH3 productions. We focus on NH3 generation by electrocatalysts and photocatalysts. We clarify the features and progress of each kind of NH3 synthesis process and provide promising strategies to further promote sustainable ammonia production and construct state-of-the-art catalytic systems.


Introduction
Occupying 78% of the atmosphere on the earth in volume and being the fifth most abundant element in solar system, nitrogen is essential for the synthesis of nucleic acids and proteins, which are the most important building blocks of life [1]. This condition is based on the reactive nitrogen that entails dinitrogen (N 2 ) fixation to its fully hydrogenated product: ammonia (NH 3 ). However, N 2 fixation is a grand challenge because N 2 molecule is thermodynamically stable with an extremely high triple-bond energy (941 kJ/mol) and insurmountable first-bond cleavage energy (410 kJ/mol) [2]. In nature, only a small group of microorganisms could biologically fix N 2 to NH 3 with the enzyme nitrogenase [3][4][5]. The most common molybdenum (Mo)-based nitrogenase is a two-component system, with Fe protein as electron-transfer media and FeMo protein as the N 2 -binding and reduction site (Fig. 1a) [6,7]. During reaction, electrons flow from a pair of adenosine triphosphate (ATP) molecules to the Fe-S cluster within the Fe protein, pass through the P cluster on FeMo protein and finally arrive at the FeMo cofactor (FeMoco), where the reduction of N 2 to NH 3 occurs. The accumulation of 8e − results in the reduction of one N 2 molecule along with the reductive elimination of one H 2 molecule [8,9]: Figure 1b illustrates the structure of FeMoco, which contains seven Fe atoms and one Mo atom that are bonded to the sulfur bridge with an interstitial carbon atom [10].
Although nitrogenase has long existed in nature, artificial N 2 fixation was invented not until the early twentieth century by Fritz Haber, and Carl Bosch developed the necessary engineering skills [11]. This pioneering work, known as the Haber-Bosch (H-B) process, successfully converts atmospheric N 2 to NH 3 on a large scale by reacting with H 2 using an iron metal-based catalyst under high temperature (400-500 °C) and pressure (100-200 bar): (1) N 2 + 8H + + 8e − + 16MgATP → 2NH 3 + H 2 + 16MgADP + 16P i (2) N 2 + 3H 2 → 2NH 3 Δ r H 0 = − 91.8 kJ/mol 1 3 To date, the worldwide output of NH 3 by H-B process has exceeded 200 million metric tons per year, which contributes to half of the global nitrogen fixation [12]. A total of 80% of the produced NH 3 is used for fertilizer production [13], whereas the remaining reserves are used in refrigeration, explosives, pharmaceuticals, plastics and other industrial processes [14]. NH 3 is a promising chemical energy carrier candidate, and it could be converted into hydrogen at point of use, showing great application potential in fuel cell technology [15,16]. Nitrogen fertilizer has supported nearly 4 billion people born since 1908 [17], and more than half of the nitrogen found in the modern human bodies originates from the H-B process [18]. Nonetheless, H-B process consumes 3%-5% of the world's natural gas for H 2 extraction and 1%-2% of the world's energy reserves, belching out millions of tons of carbon dioxide (CO 2 ), which accounts for 1.5% of all greenhouse gas emissions, annually to the atmosphere [19,20]. Over the past 150 years, atmospheric CO 2 concentration has increased from 280 × 10 −6 to 408 × 10 −6 as a result of intensive fossil fuel consumption [21]. The urgency of limiting global warming and fossil energy use has led to intense research on N 2 reduction driven by renewable sources, thus providing effective strategies to close the gap in carbon cycle and address the food supply in the future [22]. On the other hand, H-B process is a highly centralized industry, but fertilizer consumption is distributed throughout agricultural territories [23]. This condition results in elevated transportation costs and emissions, especially for remote and underdeveloped regions. Therefore, sustainable strategies that would allow NH 3 production under benign conditions at distributed sources must be developed [22].
In addition to enzymatic N 2 fixation in nature, several alternative methods for N 2 reduction reaction (NRR), including electrocatalysis [13,[24][25][26][27][28][29][30], photocatalysis [19,23,29,[31][32][33], plasma catalysis [14,34] and homogeneous molecular catalysis, have been developed [2,35]. Compared with the H-B process that requires high temperature and pressure, the reactions of these alternative methods are induced under mild conditions, which include partially reduced temperature/pressure [36], high or intermediate temperature with atmospheric pressure [37,38] and room temperature (RT) with atmospheric pressure (ambient conditions) [39]. On the other hand, the hydrogen of synthesized NH 3 in these alternative methods originates from H 2 (same as in H-B process) or H 2 O molecules. From a thermodynamic point of view [22], NH 3 synthesis using N 2 and H 2 O at ambient conditions is extremely challenging, but such process will result in profound influence on the sustainable future once achieved.

Electrochemical Ammonia Synthesis
Motivated by the urgent targets of CO 2 emission reduction and decreasing costs of renewable energy such as solar and wind, the electrical grid is rapidly transforming toward a low-carbon system. The necessity of protons/electrons for the completion of natural enzymatic N 2 fixation (reaction 1) indicates that NH 3 could be synthesized via an electrochemical process. Investigating electrochemical N 2 reduction could not only realize distributed NH 3 production but also balance the supply in electrical grids due to the intermittency of renewable sources. Thus far, several kinds of electrolytes [13,26] are used for electrochemical NH 3 synthesis: (1) solid electrolytes, such as solid polymer or perovskites, which are operated from RT to high temperature (T > 500 °C); (2) molten electrolytes which are operated at intermediate temperature (100 °C < T < 500 °C); (3) liquid electrolytes which are operated at near RT. In the first decade of this millennium, research was focused on NH 3 synthesis in solid electrolytes or molten salts, and relatively high NH 3 production rate up to the order of 10 −8 mol/(cm 2 s) has been achieved at elevated temperature [13]; however, these methods remain inapplicable for practical use. Research on N 2 fixation at ambient conditions in liquid electrolytes, especially aqueous media, has experienced an explosive growth in the last 4 or 5 years [40], aiming at harvesting NH 3 from ubiquitous nitrogen and water by renewable electricity. Biological N 2 fixation. a Reaction pathway of N 2 reduction to NH 3 on nitrogenase. Reproduced with permission from Ref. [7]. Copyright 2016, AAAS. b Structure of the resting-state FeMoco. Reproduced with permission from Ref. [10]. Copyright 2016, American Chemical Society

Solid-State Electrochemical Ammonia Synthesis (SSAS)
The reaction kinetics of NH 3 synthesis is favored at high temperatures. However, general electrolytes could not be operated in the high-temperature region. Highly protonconductive solid-state materials working at high temperatures were discovered in 1980s [41]. In 1998, Marnellos and Stoukides [37] first demonstrated the electrochemical synthesis of NH 3 from N 2 and H 2 using a solid proton conductor. Typically, the SSAS system is composed of two porous electrodes (anode and cathode) separated by a dense solid electrolyte, which blocks gas diffusion and facilitates ion transport of protons or oxide anions [42] (Fig. 2).
The aforementioned perovskite, pyrochlore and fluoritetype electrolytes belong to high-temperature proton conductors for NH 3 synthesis at atmospheric pressure. Nonetheless, high temperature would increase energy consumption and cause NH 3 decomposition. Solid polymers, such as Nafion, are good proton conductors that can be used as electrolytes for SSAS at atmospheric pressure and low temperature. Kyriacou and coworkers [54] first introduced Nafion membrane to SSAS with Pt and Ru, which were used as anode and cathode, respectively. At − 1.02 V and 90 °C, this system achieved an ammonia formation rate of 2.12 × 10 −11 mol/ (cm 2 s) with a current efficiency of 0.24%. The performance of Nafion-based SSAS was greatly improved when metal electrodes were replaced by composite oxides. Liu's group [48] adopted fluorite-type Ce 0.8 Sm 0.2 O 2−δ and perovskitetype SmFe 0.7 Cu 0.3−x Ni x O 3 as the anode and cathode, respectively, in a Nafion-based SSAS cell. At − 2 V and 80 °C, this system yielded NH 3 with a fast rate of 1.13 × 10 −8 mol/ (cm 2 s) and a high current efficiency of 90.4%. This rate is the highest among SSAS cells to date.

O 2− -Conducting SSAS
In addition to proton-conducting electrolytes, O 2− -conducting solid electrolyte cells could realize NH 3 production from N 2 and H 2 O (Fig. 2b): Fig. 2 Schematic of SSAS systems using a solid-state proton-conducting electrolyte and b solid-state oxygen anion-conducting electrolyte. Inert carrier gas is omitted. Adapted with permission from Ref. [42]. Copyright 2014, Frontiers Production Office Stoukides's group [55]

Pros and Cons of SSAS
NH 3 synthesis in solid-state electrolytes is usually conducted at elevated temperatures, because of not only hastened reaction kinetics but also substantially boosted conductivity of electrolytes. An appreciable production rate of up to 10 −8 mol/(cm 2 s) can be achieved from N 2 and H 2 [Eq. (2)], but this exothermic reaction is not thermodynamically favorable at high temperatures, necessitating a trade-off for maximizing NH 3 yield. H 2 production from natural gas reforming is energy intensive and causes most of CO 2 release. Although NH 3 production from N 2 and H 2 O [Eq. (7)] is endothermic and theoretically carbon-free, the weak reducing power and low conductivity of O 2− greatly restrict NH 3 yield.

Electrochemical Ammonia Synthesis in Molten Electrolytes
Compared with SSAS, molten electrolytes enhance the ionic conductivity and reduce the operating temperature of the reactions. In molten alkali metal salts, N 3− is stabilized and acts as intermediate product from N 2 to NH 3 . Moreover, protic solvents are avoided, which eliminates the competitive hydrogen evolution reaction (HER).
Ito and coworkers [57,58] discovered that N 2 gas could be electrochemically reduced to N 3− in molten LiCl-KCl-Li 3 N eutectic melts. Afterward, they used the molten salt system for electrochemical NH 3 synthesis (Fig. 3a) [59]. N 3− is the conducting ion in molten LiCl-KCl-Li 3 N or LiCl-KCl-CsCl-Li 3 N electrolytes. The reactions on the two electrodes are as follows: In this system, the NH 3 synthesis rate was correlated to hydrogen partial pressure in the gas electrode instead of (7) electrolysis potential [60]. The rate-determining step was proposed to be the dissolution/diffusion of hydrogen in the molten electrolytes. The highest NH 3 production rate was 3.33 × 10 −8 mol/(cm 2 s) with a current efficiency of 72% at 400 °C [60]. In the presence of N 3− , however, a portion of NH 3 could dissolve to form imide (NH 2− ) and amide (NH 2 − ) anions, which might lower the NH 3 production rate [61]. Apart from H 2 , other hydrogen sources, such as H 2 O [62,63], HCl [64], H 2 S [65] and CH 4 [66], were explored for electrochemical NH 3 synthesis in LiCl-KCl-CsCl eutectic melts.
Licht and colleagues [36] illustrated a configuration for electrochemical NH 3 synthesis, where NH 3 was generated by electrolysis of air and water steam in molten 0.5 NaOH/0.5 KOH in the presence of nano-Fe 2 O 3 catalysts. At 200 °C and 1.2 V, NH 3 was produced under 2 mA/cm 2 of applied current with a current efficiency of 35%. At the largest current density of 200 mA/cm 2 , the production rate of NH 3 was as high as 1.0 × 10 −8 mol/(cm 2 s). The high surface area of the nano-Fe 2 O 3 that remained colloidal in electrolysis was critical to the NH 3 synthesis process, whereas macro-Fe 2 O 3 descended and accumulated at the cell bottom without discernible NH 3 production. Notably, solar thermal energy could be introduced into the system, resulting in a solar thermal electrochemical process (Fig. 3b) [67].
McEnaney et al. [68] reported an ammonia synthesis system from N 2 and H 2 O using a lithium cycling electrification strategy at atmospheric pressure (Fig. 3c). This lithium-mediated cycling process combines three steps: Step 1: Electrolyzing molten LiOH to metallic Li at 400 °C-450 °C: Step 2: Nitridation of metallic Li to form Li 3 N: Step 3: Li 3 N hydrolysis to release NH 3 and regenerate LiOH: This stepwise approach circumvents direct N 2 protonation and therefore substantially inhibits undesired HER, leading to a high initial current efficiency of 88.5%. The ease of dissociation of the strong N-N bond over metallic Li and diffusion processes to form Li 3 N at RT are keys to the demonstrated cycle. This strategy could be coupled with renewable sources of electricity to facilitate localized sustainable NH 3 synthesis.

Electrochemical Ammonia Synthesis in Liquid Electrolytes
Low-temperature electrochemical NH 3 synthesis in liquid or aqueous electrolytes has attracted considerable research attention. On the one hand, NH 3 production at low RT could substantially reduce the energy consumption. On the other hand, ubiquitous H 2 O is used as the proton source instead of H 2 , which reduces the process cost and inhibits greenhouse gas emission fundamentally. Solid or molten salt electrolytemediated N 2 reduction systems contain limited or zero water molecules, resulting in an insignificant hydrogen evolution. Therefore, research on NH 3 synthesis in solid or molten salt media mainly focuses on electrolytes and the system design to reduce operating temperatures and enhance ion conductivity. In liquids, especially aqueous media, however, the abundance of water results in an extremely competitive electrolysis to hydrogen. In theory, given a highly active electrocatalyst, NRR can proceed in a narrow region of negative potentials without inducing H 2 O reduction (line a in Fig. 4) at any pH condition [22]. However, most electrocatalysts have an insufficient activity toward NRR. Therefore, NH 3 could only be generated at more negative potentials than water reduction, where most electrons would favor H 2 generation. Normally, the current efficiency of NH 3 production in aqueous electrolyte hardly exceeds 5%. The selectivity challenge necessitates the delineation of a mechanistic understanding of catalytic dinitrogen reduction to ammonia, based on which efficient heterogeneous electrocatalysts could be reasonably designed.

Reaction Mechanisms
The hydrogenation of N 2 to NH 3 on heterogeneous catalysts can be divided into dissociative and associative mechanisms (Fig. 5) [69,70]. As for the dissociative mechanism, the N≡N molecule is cleaved into adsorbed atomic nitrogen, which is subsequently hydrogenated to   (Fig. 5a). Dissociative pathway is assumed to dominate ammonia synthesis in H-B process, where the nitrogen chemisorption/dissociation is considered as the rate-determining step [71]. In the associative pathway, protonation on adsorbed N 2 primarily occurs. The N-N linkage is maintained during the initial reduction steps. Based on the hydrogenation sequences on two nitrogen atoms in N 2 , the associative mechanism can be specifically assigned to the distal or alternating pathways. In the distal pathway ( Fig. 5b), hydrogenation is asymmetric: the distal (relative to the substrate) nitrogen atom is hydrogenated first until one NH 3 molecule is released, followed by subsequent hydrogenation on the proximal nitrogen atom and the second NH 3 release. In the alternating pathway (Fig. 5c), the two N atoms are hydrogenated simultaneously (symmetrically) until NH 3 is released, which is analogous to the mechanism of nitrogenase [6].
In a particular electrocatalytic system of NRR, the explicit pathways (Fig. 5) remain obscure. Density functional theory (DFT) is often used to calculate the free energy changes of possible intermediates and delineate the possible mechanism on specific catalysts. For example, Nørskov's group [72] systematically evaluated the electrocatalytic activity of NH 3 formation on both flat and stepped surfaces of a range of transition metals. Based on the approximate linear relations between the adsorption energies of nitrogen-containing intermediates NH x /N 2 H x and the chemisorption energy of N-adatom, the free energy changes in elementary reactions (∆G) were calculated, and the rate-limiting ∆G reflected the onset potential (U) for NH 3 synthesis. Combined volcano diagrams were plotted [72,73] to show the estimated onset potential as a function of the nitrogen-binding energy for dissociative (solid lines) Nitrogen reduction pathways on heterogeneous catalysts. a Dissociative pathway where N≡N bond is broken before hydrogenation. Associative pathways including b distal or asymmetric hydrogenation and c alternating or symmetric hydrogenation. Adapted with permission from Ref. [70]. Copyright 2014, The Royal Society of Chemistry and associative (dashed lines) mechanisms on either flat (blue) or stepped (red) surfaces (Fig. 6). On the left sides of the volcanoes, the dissociative pathway dominates, and the onset potential for NH 3 formation is slightly more positive on the flat surfaces than that on the stepped ones. This slight difference is attributed to the smaller free energy change from *NH to *NH 2 (rate-limiting step on flat surfaces) than that from *NH 2 to NH 3 (rate-limiting step on stepped surfaces). On the right sides of the volcanoes, N 2 splitting is the rate-limiting step for dissociative mechanism on either flat or stepped surfaces (overlapped as black solid line) due to the same free energy change from N 2 to *2N, whereas the first hydrogenation step on adsorbed N 2 determines the onset potential for the associative mechanism (blue and red dashed lines). Different transition metals prefer different reaction pathways. Reactants on early transition metals (Sc to Fe; Y to Ru) follow a dissociative pathway with one of the hydrogenation reactions as the rate-limiting step. Several late transition metals prefer a dissociative mechanism with N 2 splitting as the limiting step (Rh, Ir, Co and Ni), whereas others prefer an associative mechanism at more negative potentials (Pd, Cu, Ag and Au). Notably, the surfaces in most part of the diagrams are dominated by H-adatoms, resulting in the rapid formation of H 2 other than N 2 . Ru and Rh with theoretically the highest activity (top of volcano diagrams) exhibit poor selectivity due to strong H adsorption. Only the flat metal surfaces of Sc, Y, Ti and Zr (shaded green area) are assumed to be covered by N instead of H. However, the strong binding of N might lead to the difficult desorption of reaction intermediates.
Theoretical screening provides limited guidance for catalyst design. The actual surface environment on the catalyst could occasionally be located distantly from that in calculations. For example, surface reconstruction occurs during reactions, and certain metal catalysts would form surface oxide layers under reduction potentials. Nanosized catalysts also have more complexed exposed faces and defects than their bulk counterparts [74][75][76][77][78][79]. In recent years, various electrocatalysts have been explored for NH 3 synthesis from N 2 and H 2 O in liquid or aqueous media [80].

Metal Catalysts
Thus far, limited experimental reports are available regarding electrochemical NRR (ENRR) on metal bulks or films in aqueous media. One chemically deposited Au thin-film electrode produced NH 3 at potentials below 0 V versus reversible hydrogen electrode (V RHE ) [81]. The authors used surface-enhanced infrared absorption spectroscopy (SEIRAS) to successfully detect N 2 H x species in the reaction process. The current efficiency was 0.12%, and the NH 3 formation rate reached 3.84 × 10 −12 mol/(cm 2 s) at − 0.5 V RHE . N 2 H x intermediates indicated an associative mechanism on Au, which is consistent with the calculations in Fig. 6. A parallel experiment on Pt film showed no intermediate absorption bands under the same conditions. Another active metal film reported is a (110) orientation Mo, which achieved a current efficiency of 0.72% for NH 3 generation [82]. shaded area indicates the region where surface is likely to be covered by adsorbed nitrogen, and white area denotes the region where surface is likely to be covered by adsorbed hydrogen. Adapted with permission from Refs. [72] and [73]. Copyright 2012, The Royal Chemical Society. Copyright 2019, Macmillan Publishers Limited More active results were observed on nanosized metal catalysts. Yan and colleagues [83][84][85] tested a series of Au electrocatalysts with different nanostructures. Tetrahexahedral (THH) Au nanorods (NRs) by a seeded growth method showed good electrocatalytic N 2 reduction performance in 0.1 mol/L KOH [83]. THH-Au NRs exposed abundant high-index facets of (310) and (210), which facilitated N 2 activation and reduction. The THH-Au achieved the highest ammonia production rate of 2.69 × 10 −11 mol/(cm 2 s) at − 0.2 V RHE with a current efficiency of ~ 4%. Although highindex faces on THH-Au play important roles in N 2 reduction, the relatively large size of NRs (~ 50 nm in length) restricts the atomic utilization. Subnanometer Au clusters (0.5 nm) on TiO 2 support showed enhanced electrocatalytic NRR, where the low-coordination sites on small Au clusters are considered crucial. An ammonia production rate of 1.12 × 10 −10 mol/(cm 2 s) with a current efficiency of 8.11% was achieved at − 0.2 V RHE [84]. Amorphous nanocatalysts with highly unsaturated coordination sites are occasionally more active than the crystalline ones in catalysis. This condition is illustrated by anchored amorphous Au nanoparticles (a-Au NPs, ~ 5 nm) on a bisubstrate of CeO x /reduced graphene oxide (CeO x -RGO), where CeO x transforms the crystallized Au into its amorphous form, and RGO acts as the substrate to anchor and disperse the Au NPs [85]. The obtained a-Au/CeO x -RGO exhibited a current efficiency of 10.1%, which was considerably higher than that of crystalline Au on RGO.
On a carbon-supported nano-Pd catalyst, an overpotential as low as 56 mV was reported for N 2 reduction [86]. α-Palladium hydride was assumed to form during reaction and to lower the free energy barrier of N 2 hydrogenation to *N 2 H, the rate-limiting step for NH 3 electrosynthesis. A current efficiency of 8.2% was attained at 0.1 V RHE .
A record-breaking activity and selectivity for ENRR at ambient conditions were recently realized on a main-group metal bismuth nanocrystals with K + promoters in aqueous electrolyte [87]. At the optimum potential of − 0.6 V RHE , the NH 3 generation rate reached as high as 1.44 × 10 −8 mol/ (cm 2 s) with a remarkable current efficiency of 66%. This production rate outperforms other works of ENRR in aqueous media by one to three orders of magnitude and parallels the levels in molten salt electrolyte cells operating at elevated temperature or pressure. During the long-term operation of up to 50 h, the Bi nanocrystal catalyst retained its activity and metallic state. The authors proposed a distal pathway of associative mechanism (Fig. 5b), with the reductive protonation of N 2 to *NNH as the rate-limiting step. The projected density of states revealed a remarkable overlap between 2p orbitals of adsorbed N and Bi 6p bands both below and above the Fermi level. By comparison, Au as a transition metal showed minimal overlap between its 5d band and N 2p orbital. Therefore, a stronger interaction existed between Bi surface and *NNH intermediate, whose formation energy barrier (∆G *NNH ) was considerably lower compared with that on Au (Fig. 7a). More importantly, K + acted as a strong promoter in this system, boosting the current efficiency from 9.8% to 67%. On the one hand, K + shifted the 2p orbital of adsorbed N to lower energy and decreased ∆G *NNH (Fig. 7b), leading to a stronger Bi-N bond and a better N-N activation. On the other hand, highconcentration K + hindered proton migration from the bulk solution to the Bi electrode surface, suppressing HER and enhancing ENRR selectivity (Fig. 7c, d, respectively).

Fig. 7
Boosting ENRR with Bi metal catalysts and K + cations. a Free energy changes (ΔG *NNH ) for the ratelimiting step forming *NNH (* + N 2 + H + + e − → *NNH) on Bi (012), (110), (104), and Au (111) facets. b ΔG *NNH on Bi (012), (110), and (104) facets without (patterned bars) and with (filled bars) K + . Mass transfer of H + and N 2 to the catalyst surface in electrolytes without c and with d K + . Adapted with permission from Ref. [87]. Copyright 2019, Macmillan Publishers Limited As illustrated in the "Reaction Mechanisms" section, the electrocatalytic conversion of N 2 to NH 3 on transition metal relies on the linear scaling relationship between *N 2 H and *NH 2 . The overpotentials of ENRR are larger than those of HER, leading to the limited success of NRR on transition metals. Metal alloys might be an approach to meet this challenge because of their synergetic interactions with the absorbed intermediates. Nonetheless, alloys cannot break the scaling relationship for *N 2 H and *NH 2 , as each kind of metal bonds to the intermediates through a single N atom. How to design the active sites to change the intermediates' binding modes remains an important question.

Metal Compound Catalysts
As mentioned in the "Electrochemical Ammonia Synthesis in Molten Electrolytes" section, nano-Fe 2 O 3 shows excellent electrocatalytic performance for NRR in molten base electrolyte, but the possible mechanism was not discussed [36]. A DFT study on hematite (0001) revealed an associative pathway where the potential demanding step is *NNH formation from N 2 [88]. The calculated applied bias needed is − 1.1 V, which is consistent with the conducted 1.2 V for nano-Fe 2 O 3 [36]. Apart from molten electrolytes, Fe 2 O 3 electrocatalyst was also applied either in gas diffusion layer (GDL) electrode [89][90][91] or directly immersed in aqueous media [92]. Chen et al. [89,91] studied carbon nanotubesupported iron oxide (Fe 2 O 3 -CNT) as N 2 reduction electrocatalyst on GDL electrode in a three-phase cell separated by a proton exchange membrane. The liquid chamber contained aqueous electrolyte for H 2 O electrolysis to produce H + and electrons. The protons migrated through proton exchange membrane to the GDL electrode, where N 2 gas was reduced by electrons and hydrogenated to form NH 3 . The current densities were determined by proton concentration in acid electrolyte and by water transport in the base electrolyte. The 30% Fe 2 O 3 -CNT was the best electrocatalyst among all Fe 2 O 3 -CNT samples [89]. Kong et al. [90] also investigated γ-Fe 2 O 3 electrocatalyst for NH 3 synthesis in a three-phase cell separated by an anion-exchange membrane. Using KOH electrolyte, the γ-Fe 2 O 3 electrode achieved a current efficiency of 1.9% at 0 V RHE .
Through an electrospinning/calcination method, a hybrid metal oxide Bi 4 V 2 O 11 /CeO 2 with an amorphous phase (BVC-A) was fabricated and used as the cathode for electrocatalytic nitrogen reduction [93]. The amorphous Bi 4 V 2 O 11 contains abundant oxygen vacancies (O vac ), which acted as active sites for N 2 reduction. CeO 2 not only induced the amorphous structure but also established suitable band alignment with Bi 4 V 2 O 11 to facilitate interfacial charge transfer (Fig. 8a). A high current efficiency of 10.16% was achieved at − 0.2 V RHE .
MoS 2 as one hot-spot two-dimensional (2D) material has attractive catalytic properties. The NH 3 formation by MoS 2 was first reported in 0.1 mol/L Na 2 SO 4 [94]. Defects on the monolayer of MoS 2 boosted N 2 reduction [95]. A DFT simulation assumed that single Fe atom deposited on MoS 2 would mimic FeMoco and endow inactive MoS 2 the capability to convert N 2 to NH 3 electrocatalytically. Fe center yields electron charge to MoS 2 and turns into an   [96]. Interestingly, Ru instead of Fe on MoS 2 was experimentally discovered as an active catalyst for ENRR [97]. As shown previously in Fig. 6, Ru is near the top of volcano plot but exhibits considerable hydrogen evolution due to dominant H adsorption. In the work on Ru/MoS 2 [97], dispersed Ru clusters provide N 2 -binding sites, whereas nearby S-vacancies on the 2H-MoS 2 serve as centers for H + reduction to adsorbed *H (Fig. 8b). These hydrogenated S-vacancies act as the H-provider because the formed *H can be transferred directly to nearby N 2 on Ru, forming *NNH as an intermediate. The *NNH was unstable on Ru/MoS 2 and spontaneously cleaved into *N and *NH, following which a dissociative pathway was proposed. M 3 C 2 transition metal carbides (M stands for metals from d 2 , d 3 and d 4 series), as a kind of Mxene, have shown capabilities for N 2 capture and reduction in a DFT study [98]. V 3 C 2 and Nb 3 C 2 exhibited the most promising features for reduction to NH 3 . In another metal carbide system, Mo 2 C nanodots embedded in carbon nanosheets were developed for electrochemical nitrogen fixation [99]. At − 0.3 V RHE , the NH 3 production rate reached 11.3 µg/ (h mg). Carbon-supported Mo 2 C is proposed to be rich in nitrogen adsorption active sites, and the unique electronic structure is favorable to N≡N bond cleavage and hydrogenation.
Transition metal nitrides draw great attention because they have the potential to activate adsorbed N 2 via the MvK mechanism [100][101][102][103]. Specifically, one surface N atom on the nitride is hydrogenated and reduced to one NH 3 molecule, forming a lattice vacancy which is subsequently restored by reductive incorporation of a N 2 molecule. One good example was demonstrated by vanadium nitride (VN) NPs, where surface VN 0.7 O 0.45 acted as the active phase [104]. The use of 15 N 2 feed gas produced 14 NH 3 and 15 NH 3 , indicating a MvK mechanism. In a catalytic cycle, a native surface N atom of VN was extracted by hydrogen atoms and left behind a N vacancy (N vac ), which could activate/ adsorb a N 2 molecule and be healed after the desorption of one NH 3 molecule. Moreover, only the surface N sites adjacent to a surface O are active toward nitrogen reduction, and the removal of surface O would deactivate the catalyst (Fig. 8c). The 2D MoN material is also a promising candidate for ENRR in theoretical and experimental demonstrations [105,106].
As illustrated, in metal compound catalysts, introducing anion vacancies, such as oxygen, nitrogen, sulfur and selenium vacancies, plays an important role in ENRR. The anion vacancies could trap metastable electrons, which are transported into an antibonding orbital of adsorbed N 2 molecules, contributing to enhancing nitrogen triple-bond cleavage for subsequent catalytic reaction. The regeneration capability of vacancies during reaction needs further investigation.

Metal-Free Catalysts
Scalable hierarchically structured nitrogen-doped nanoporous carbon membranes with embedded CNTs were developed for ENRR [107]. In this hierarchically porous membrane structure, micropores and small mesopores provided large and accessible surface areas up to 432 m 2 /g. Meanwhile, large mesopores and macropores formed interconnected 3D conductive framework to expedite mass diffusion and enhanced N 2 reduction efficiency on active sites, which are proposed to be pyridinic and pyrrolic N atoms in N-doped carbons [107,108].
A polymeric carbon nitride (PCN) abundant in N vac was proposed as an electrocatalyst to enable ammonia synthesis under ambient conditions [109]. DFT calculations illustrated that dinitrogen molecule can be chemisorbed on N vac of PCN in a dinuclear end-on bound structure, which dramatically increases N-N bond length and improves spatial electron transfer. A high current efficiency of 11.59% was therefore obtained at − 0.2 V RHE , leading to an ammonia production rate of 9.64 × 10 −10 mol/(cm 2 s).
Black phosphorus (BP) with a few layers of nanosheet structure is a nonmetallic electrocatalyst for nitrogen reduction under ambient conditions [110]. DFT calculations propose an alternating hydrogenation pathway. The zigzag and diff-zigzag edges of BP are active centers for nitrogen adsorption and activation.

Battery Configurations for NRR
In addition to electrolysis, proof-of-concept batteries have led to new directions toward N 2 fixation. Minteer and colleagues [111] combined nitrogenase and hydrogenase into a H 2 /N 2 fuel cell, with a 3-morpholinopropane-1-sulfonic acid buffer as the electrolyte (Fig. 9). The electron transfer between cathode/nitrogenase and anode/hydrogenase was realized by using methyl viologen (MV, N,N′-dimethyl-4,4′-bipyridinium) as the sole electron donor. The coupling of this nitrogenase cathode to reduce N 2 with a hydrogenase anode to oxidize H 2 resulted in an enzyme-assisted fuel cell (EFC), generating NH 3 from H 2 and N 2 while simultaneously outputting an electrical current. A current efficiency of 26.4% was obtained on this H 2 /N 2 EFC. Ma et al. [112] reported a successful illustration of a reversible nitrogen cycle based on a rechargeable lithium-nitrogen (Li-N 2 ) battery with the proposed reversible reaction of 6Li + N 2 ⇌ 2Li 3 N. The N 2 fixation battery assembly was composed of a lithium anode, an ether-based electrolyte, and a carbon cloth cathode and exhibited a promising electrochemical current efficiency of 59%.

Methods to Suppress HER
In protic (e.g., aqueous) electrolytes, HER is a serious competitive reaction that greatly restricts the selectivity of N 2 reduction to NH 3 in the presence of protons. For example, recent publications [83-86, 93, 109, 113-115] reported electrochemical NH 3 synthesis in aqueous solutions: The highest current efficiency for NH 3 and the highest NH 3 production rate depend on small applied potentials, whereas further enlarged negative potentials result in a significantly reduced current efficiencies and NH 3 production rates (Fig. 10). − 0.2 V RHE is the most frequently reported potential [83-85, 93, 109, 113, 114] at which the fastest NH 3 generation is attained (Fig. 10a-c). For a palladium-catalyzed N 2 reduction system, the highest current efficiency of 8.2% was achieved at as high as 0.1 V RHE (Fig. 10d) [86]. At such small applied potentials, the current densities are in the range of several hundred microamperes to several microamperes, which are inapplicable to practical applications. This effect could be amplified when current densities are larger.
Nørskov and colleagues [116] presented in-depth perspectives on the selectivity challenges for electrochemical NH 3 synthesis. They suggested that limiting either proton or electron availability at the surface is a promising way to improve the selectivity of NH 3 . Detailed methods include limiting the proton transfer rates by reducing proton concentration in the electrolyte or increasing proton transfer barriers to the catalyst surface or limiting the electron transfer rates by constructing thin insulators or supplying slow streams of electrons.
By reducing proton concentration using ionic liquids as the electrolytes, MacFarlane and coworkers [117] Fig. 9 Incorporation of hydrogenase and nitrogenase proteins into a H 2 /N 2 fuel cell with MV as the electron mediator in both chambers. This fuel cell produces NH 3  [eFAP] is notably higher than that in aqueous solutions. In addition, these ionic liquids can serve as aprotic electrolytes where H 2 evolution can be effectively suppressed. Notably, a trace amount of water (20 × 10 −6 -250 × 10 −6 ) is the only proton source. As a result, an unprecedented current efficiency of 60% for NH 3 synthesis in liquid electrolytes at ambient conditions was achieved.
High current efficiency of ENRR was also realized in a solution of LiClO 4 (0.2 mol/L) in tetrahydrofuran/ethanol (99:1 V/V) on metal electrodes [118]. Lithium in the electrolyte acted as a mediator because Li + was reduced on the electrode to deposit metallic Li, which reacted with N 2 to form Li 3 N. Li 3 N underwent a subsequent ethanolysis to generate NH 3 . Moreover, in this Li-mediated system, the metal electrode can be modified with a functional layer such as superhydrophobic metal-organic framework to suppress the proton availability and accumulate N 2 molecules at the electrode surface [119].
Although reducing proton concentration surrounding the catalyst is an effective way to improve the selectivity of ENRR, the sacrifice of current density or NH 3 generation rate cannot be overlooked. The operating current density at several microamperes [117,119] is insufficient. New approaches need to be developed to solve this dilemma.

Summary and Protocols for Electrochemical Ammonia Synthesis
The past three decades have witnessed a flourishing of interest in electrochemical NH 3 synthesis. Table 1 summarizes several representative developments in this realm. For SSAS using N 2 and H 2 as the reactants, NH 3 generation rate could reach the order of 10 −9 mol/(cm 2 s) to 10 −8 mol/(cm 2 s), and the current efficiency is considerable. However, either the high temperature [37] (usually larger than 500 °C) or the large cell potential [48] limits the energy utilization; furthermore, H 2 production is energy consuming and carbon intensive. Molten electrolytes enhance the ionic conductivity and reduce the operating temperature to 200-400 °C. The current efficiency in molten electrolytes is as high as that in SSAS, whereas the NH 3 generation rate is faster at the order of 10 −8 mol/(cm 2 s). Notably, reaction from N 2 and H 2 O achieves a NH 3 production rate of 1.0 × 10 −8 mol/(cm 2 s) in molten NaOH/KOH at a moderate temperature of 200 °C [36]. From the perspective of sustainable development, harvesting NH 3 from N 2 and H 2 O at ambient conditions is the most tempting goal. Nonetheless, this process is extremely challenging because of the competitive hydrogen evolution and difficult N 2 activation. The NH 3 yield in aqueous electrolytes ranges from the order of 10 −12 mol/(cm 2 s) to 10 −10 mol/(cm 2 s) with a current efficiency less than 10%. Reducing the proton availability, such as by using aprotic solvents, is an approach to notably boost the current efficiency; however, the operating current density is extremely small, leading to a low NH 3 yield [117,119]. One Bi nanocrystal catalyst in aqueous media achieves a high NH 3 yield of 1.44 × 10 −8 mol/(cm 2 s) and a high current efficiency of 60% [87]. This result is exciting, but more study is needed to solve the problems of high overpotential and the narrow operating potential window.
In spite of the reported progress thus far, electrochemical NH 3 production will need to achieve considerable progress toward practical applications that require a minimum NH 3 production rate in the order of 10 −7 mol/(cm 2 s) [13]. More importantly, electrochemical NH 3 production needs both scientific and engineering design to make the process less energy consuming than the well-developed H-B process.
Another challenge is that the slow NH 3 generation rate of ENRR causes difficulty in attributing the detected NH 3 to real electrochemical N 2 fixation given the existence of numerous exogenous nitrogen contaminants. In addition, N 2 as a robust and nonpolar molecule is extremely difficult to fix under ambient conditions. Although ENRR has achieved great enhancement in ammonia production rates in the past 4 years, limited papers provide rigorous evidence to prove that ammonia truly comes from N 2 . In any given experiment, adventitious ammonia can be introduced in the reaction system in various ways, as illustrated by a rigorous protocol by Andersen et al. [120]. Ammonia contamination could be present in air, atmosphere, human breath or Nafion membranes or originate from nitrogen-containing compounds that are normally present in the nitrogen gas supply. Numerous electrocatalysts are also nitrogen-containing compounds or fabricated from nitrogen-containing precursors. Therefore, excluding the interference of exogenous contamination and confirming the source of fixed nitrogen are the prerequisites to reporting a positive result. Fortunately, benchmarking protocols are being progressively established to identify and eliminate contamination sources [40,73,120], to prevent false positives and standardize ENRR experiments. Upon successful detection of ammonia after running electrolysis, one must conduct control experiments with Ar gas supply under the exact same conditions and with N 2 gas in the open-circuit condition. Isotope labeling using 15 N 2 is a necessary confirmation procedure. Numerous papers reported the qualitative detection of 15 NH 3 or 15 NH 4 + from 15 N 2 , thereby alleging the successful fixing of N 2 . However, 15 N 2 gas stock normally contains 15 N-labeled nitrate or ammonia, which invalidates the detected results [121]. Therefore, gas-cleaning unit must be applied before bubbling N 2 stream to the electrolyte. Meanwhile, the yield of 15 NH 3 should be quantitatively consistent with that of 14 NH 3 during identical reduction experiments. Andersen's work demonstrates excellently how to perform quantitative isotope measurements. A copper catalyst trap was used to eliminate N-containing contaminations, and a gas circulation system was used to maximize the use of expensive 15 N 2 gas [120]. The developing rigorous protocols will enable the identification of ENRR results by preventing false positive data and contribute to the development of more efficient processes toward electrochemical NH 3 production.

Photocatalytic Ammonia Synthesis
Effectively capturing solar energy for the production of fertilizers and fuels is an ambitious and challenging goal [122]. Green ammonia from nitrogen photofixation has drawn increasing attention in recent years [19,123]. Abiotic photofixation of dinitrogen in soils and sands has been suggested to be the third most significant source of natural nitrogen fixation, apart from biological N 2 fixation and lightning discharges [23,124]. The investigation of solar-driven N 2 fixation is significant for people to comprehend and modulate the nitrogen cycle. Although early studies mainly focused on titania-based catalysts, a sharp increase occurred in recent years regarding photocatalytic N 2 fixation on emerging catalytic systems.

Titania-Based Photocatalysts
Rutile TiO 2 in sands or soils is considered to be the N 2 reduction catalyst in the presence of light and water [124]. The first experiment on N 2 photoreduction with water splitting was reported by Schrauzer and Guth [125] on outgassed rutile TiO 2 powders. H 2 evolution was notably inhibited in N 2 atmosphere, whereas iron doping enhanced the photocatalytic reactivity. Since then, titania has been intensively explored as the photocatalyst for N 2 fixation, although other metal oxide semiconductors, such as tungsten oxide [126] and iron oxide [127], were also investigated. Ranjit et al. [128] studied photocatalytic reduction of N 2 to NH 3 on noble-metal-loaded TiO 2 . They observed a correlation between NH 3 yield and the M-H bond strength, where a high-bond strength gives rise to a high NH 3 yield. Hoshino et al. [129] reported N 2 photoreduction on needle-like solid ammonium perchlorate (NH 4 ClO 4 ) using a TiO 2 /conducting polymer (poly 3-methylthiophene, P3MeT) catalyst. Under illumination, photogenerated carriers at the TiO 2 /P3MeT interface contributed to NH 3 synthesis in the presence of water. Meanwhile, ClO 4 − was doped from P3MeT driven by electrons, and the acid-base reaction formed mesoscale NH 4 ClO 4 needles. Kisch and colleagues [130] reported nitrogen fixation at nanostructured iron titanate films, where the highest NH 3 generation rate was realized on iron titanate film with Fe:Ti ratio of 1:1. Zhao et al. [131] fabricated Fe-doped TiO 2 NPs with highly exposed (101) facets by a two-step hydrothermal method. Optimal doping of Fe 3+ is essential to the improvement of photocatalytic activity.
Recently, the introduction of O vac has drawn new insight into N 2 photofixation on TiO 2 catalysts. Hirakawa et al. [132] introduced a large number of O vac in a commercially available TiO 2 by H 2 treatment. Under UV light illumination, N 2 molecules were reduced to NH 3 by Ti 3+ species on O vac (Fig. 11a), leading to a solar-to-chemical energy conversion efficiency of 0.02%. However, methods such as H 2 reduction do not avoid the introduction of O vac to the bulk to form bulk defects, which potentially act as carrier traps and induce charge recombination [133]. Therefore, the introduction of O vac on the outermost surface is highly desirable for TiO 2 and many other reducible oxides. Gong and coworkers [39] reported the conversion of N 2 to NH 3 in pure water using a plasmon-enhanced rutile TiO 2 NR array modified with surface O vac , which were created by atomic layer deposition (ALD). This method ensures the introduction of O vac on the surface without affecting bulk structure. Compared with the less active rutile TiO 2 surface, the amorphous ALD TiO 2 layer (a-TiO 2 ) with catalytic centers of surface O vac could promote N 2 adsorption and activation, greatly enhancing the N 2 photofixation rate. Meanwhile, surface plasmons of Au extended the absorption range of TiO 2 to the visible region and provided high-energy hot electrons for N 2 reduction (Fig. 11b). Accordingly, the TiO 2 /Au/a-TiO 2 photoelectrode exhibited a notably higher NH 3 production rate than bare TiO 2 , achieving 13.4 nmol/(cm 2 h) under 1 sun illumination [39]. Similarly, Yang et al. [134] illustrated a "working-in-tandem" nitrogen photofixation system, which was realized by assembling plasmonic Au nanocrystals on O vac -rich ultrathin TiO 2 nanosheets (Fig. 11c, d). The O vac on the TiO 2 nanosheets chemisorbed and activated N 2 molecules, which were further reduced to ammonia by hot electrons generated from plasmonic gold nanocrystals. The apparent quantum efficiency for the conversion of incident photons to NH 3 reached 0.82% at 550 nm. The N 2 photofixation rate can be further improved by optimizing the absorption of visible light with the mixture of Au nanospheres and NRs.

2D-Layered Photocatalysts
Layered bismuth oxyhalides for nitrogen photofixation were first reported by Zhang and coworkers [31,135]. Under visible-light illumination and ambient conditions, efficient NH 3 was generated from N 2 and water on bismuth oxybromide (BiOBr) nanosheets of O vac in the absence of any organic scavengers and precious-metal cocatalysts. O vac on the  [134]. Copyright 2018, American Chemical Society exposed (001) facets provided localized electrons for π-backdonation and activated adsorbed N 2 , which could be reduced to NH 3 by the transferred electrons from the conduction band of excited BiOBr nanosheets (Fig. 12a). The resultant N 2 photofixation rate on BiOBr was 104.2 µmol/(h g) [135]. Zhang and coworkers [136] further demonstrated that O vac on BiOCl could act as the catalytic centers and contribute to the solar light driven N≡N triple-bond cleavage via a proton-assisted electron-transfer pathway. In addition, different BiOCl facets strongly influence the N 2 reduction pathways by affecting both the adsorption structure and the activation level of N 2 .
Graphitic carbon nitride (g-C 3 N 4 ) is a burgeoning material for N 2 photofixation, either as a catalytic or a supportive material [137][138][139][140][141][142][143]. Dong et al. [137] observed that N vac endowed g-C 3 N 4 with the photocatalytic N 2 fixation capability because N vac could selectively adsorb and activate N 2 given their same shape and size with the nitrogen atom in N 2 . Cao et al. [138] reported a Z-scheme heterojunction-structured photocatalyst: 3,4-dihydroxybenzaldehyde-functionalized Ga 2 O 3 /graphitic carbon nitride (Ga 2 O 3 -DBD/g-C 3 N 4 ). The interaction between aromatic aldehydes in Ga 2 O 3 -DBD and the terminal -NH 2 groups in g-C 3 N 4 improved the dispersion of Ga 2 O 3 -DBD NPs and resulted in the formation of a well-combined interface, which enhanced the charge transfer rates. Aromatic rings with good conductivity acted as electron mediators and promoted the recombination between photogenerated electrons from the conduction band of Ga 2 O 3 and photogenerated holes from the valence band of g-C 3 N 4 , boosting the overall photovoltage [138]. Hu et al. [139,140] anchored Fe 3+ and Cu 1+ at the interstitial position of g-C 3 N 4 , where coordinative M-N bonds were formed. The metaldoped g-C 3 N 4 exhibited notably higher photoactivity for N 2 reduction compared with bare g-C 3 N 4 . DFT simulations showed that a high nitrogen adsorption energy was obtained on anchored metal sites, and N-N bond could be elongated. DOS results indicate that the electrons of σ g 2p orbital (highest-occupied molecular orbital) in nitrogen atom were substantially delocalized when N 2 adsorbed on metal-doped sites, and the orbital energy almost crossed that of π g *2p orbital (LUMO), illustrating that Fe 3+ or Cu 1+ sites can activate the N 2 molecule effectively. Jiang and colleagues [142] designed and fabricated a new TiO 2 @C/g-C 3 N 4 photocatalyst through thermal treatment of a mixture of melamine and MXene Ti 3 C 2 T x . This method endowed carbon nanosheetsupported TiO 2 with abundant Ti 3+ species that were tightly wrapped by in situ-formed g-C 3 N 4 nanosheets. This heterojunction enhanced light absorption and charge separation, where electrons were injected from g-C 3 N 4 to Ti 3+ on TiO 2 for N 2 activation and reduction (Fig. 12b).
Graphene could generate a high density of hot electrons well above the Fermi level under visible light [144,145]. Chen and colleagues [146] noted that light-generated highly energetic hot/free electrons of graphene could act as a promising reducing agent for NH 3 synthesis from N 2 and H 2 under ambient conditions. They fabricated an iron-and Schematic of the energy band structure and electron-hole separation of TiO 2 @C/g-C 3 N 4 . Adapted with permission from Ref. [142]. Copyright 2018, the Royal Society of Chemistry. c Proposed pathway for the NH 3 synthesis using Fe@3DG catalyst under light illumination. Adapted with permission from Ref. [146]. Copyright 2016, American Chemical Society graphene-based catalyst, Fe@3DGraphene, for NH 3 photosynthesis. Hot electrons from graphene induced by visible light were ejected onto the Fe catalytic sites, where N 2 activation and NH 3 /H 2 generation occurred directly, without any other agents (Fig. 12c). Alumina as a structural promoter enhanced the stability of Fe@3DGraphene up to 50 h [146]. The same group further proved that nano-Al 2 O 3 acting as a barrier among nano-Fe 2 O 3 could significantly prevent the aggregation of Fe 2 O 3 particles, improving the stability of catalysts [147].
A series of ultrathin layered-double-hydroxide (LDH) nanosheet photocatalysts of the type M II M III -LDH (where M II = Mg, Zn, Ni, Cu; and M III = Al, Cr) were synthesized by simple coprecipitation routes [148]. These LDH nanosheets were engineered with O vac defects to enhance the absorption and activation of N 2 . Especially, the CuCr-LDH photocatalyst exhibited a high activity under visible light for the photoreduction of N 2 to NH 3 . Cu 2+ ions in the LDH nanosheets were assumed to introduce additional structural distortions and compressive strain, which boosted the interaction between the nanosheets and N 2 and thereby enhanced NH 3 formation.

Other Semiconductor Photocatalysts
Bismuth oxyhalide is susceptible to photocorrosion, where surface O vac is easily oxidized to lose the catalytic sites. Wang et al. [149] resolved this problem using self-assembled 5 nm diameter Bi 5 O 7 Br nanotubes (NTs) through a low-temperature wet chemical method. The Bi 5 O 7 Br NTs contained abundant and light-switchable O vac , realizing excellent and stable photosynthesis of NH 3 in pure water. The NH 3 generation rate was as high as 1.38 mmol/(h g), with an apparent quantum efficiency of 2.3% at 420 nm.
Bismuth monoxide (BiO) quantum dot is a low-valence metal oxide semiconductor that has fewer coordination atoms than its high-valence states [150]. This condition endows BiO with high electron-donating power and empty 6d orbitals for N 2 adsorption and activation. The N 2 molecule could be stretched and activated by alternately arranged Bi atoms by donating electrons to the empty Bi 6d orbitals. Without hole scavengers, BiO quantum dots exhibited a high NH 3 generation rate of 1226 µmol/(h g).

Biohybrid and Biomimetic Photocatalysts
In a bioinorganic system, cadmium sulfide (CdS) nanocrystals were used to photosensitize the nitrogenases molybdenum-iron (MoFe) protein, where light harvesting from CdS replaced ATP hydrolysis on the Fe protein to transfer electrons for enzymatic reduction of N 2 into NH 3 [7]. This CdS:MoFe protein biohybrid system achieved an optimal turnover rate of 75 per minute, which is 63% of the ATP-coupled reaction for nitrogenase. Kanatzidis and coworkers [151] reported a nitrogenase-inspired biomimetic chalcogel system that exhibited photoactivity for N 2 reduction to NH 3 in aqueous media under ambient pressure and RT. The high-surface-area amorphous chalcogels were composed of Mo 2 Fe 6 S 8 (SPh) 3 or Fe 4 S 4 with Sn 2 S 6 clusters and exhibited strong optical absorption. Compared with chalcogels with Mo 2 Fe 6 S 8 (SPh) 3 cluster [151], Mo-free chalcogels containing only Fe 4 S 4 clusters are more efficient for N 2 reduction to NH 3 [152]. This result suggests that Fe might be the active site for N 2 binding, similar to that in nitrogenase.
The first trial was realized in a PEC cell that contained a p-GaP cathode and an Al metal anode immersed in a nonaqueous electrolyte. N 2 was reduced to NH 3 on illuminated p-GaP electrode, whereas Al was continually consumed as the reducing agent [156]. Hamers and coworkers [157] used diamond as a solid-state source of solvated electrons for N 2 reduction. The conduction band edge of diamond lay at about 1 eV above the vacuum level, which enabled the electrons to be directly ejected into the inert N 2 molecules with negligible barrier (Fig. 13a). Oshikiri et al. [158,160] investigated Au-decorated SrTiO 3 photoelectrode with Ru or Zr/ZrO x as the cocatalysts. Water was oxidized by holes on the Au side, whereas excited hot electrons were conducted through the conduction band of SrTiO 3 to the catalyst side for N 2 reduction (Fig. 13b). MacFarlane and colleagues [159] presented a solar-driven PEC cell based on plasmonenhanced black silicon nanowires (NWs) for the reduction of N 2 to ammonia. When sulfite was used as a reactant, the process could produce ammonium sulfate, an important fertilizer with high economic value (Fig. 13c). Li et al. [161] fabricated GaN NW array on a silicon substrate by plasmaassisted molecular beam epitaxy. Afterward, finely dispersed sub-nanoclusters of Ru were deposited on GaN with a high load of 5 wt% (Fig. 13d), forming a Schottky barrier junction between Ru and GaN (Fig. 13e), which resulted in partially negatively charged Ru species for elevated N 2 reduction performance under illumination. Zheng et al. [162] designed an aerophilic-hydrophilic heterostructured Si-based composite photocathode for PEC reduction of N 2 to NH 3 . Polytetrafluoroethylene (PTFE) porous framework was used as the N 2 diffusion layer, whereas Au NPs acted as the active sites and the electric contact between PTFE framework and Si. This structure formed an aerophilic-hydrophilic functional layer, which enriched N 2 concentration at the Au active sites and suppressed HER by reducing proton availability. This photocathode exhibited an NH 3 yield rate of 3.1 × 10 −10 mol/ (cm 2 s) and a current efficiency of 37.8% at − 0.2 V RHE at ambient condition. Table 2 lists the important developments in photocatalytic NH 3 synthesis. In photocatalyst systems, the NH 3 yield has enhanced from the order of µmol/(h g) to mmol/(h g). This increase is a notable improvement, but the production rate is only realized at the laboratory scale and still cannot meet the requirements for practical use. In PEC systems, the NH 3 yield is evidently smaller than that in ENRR at present. In addition to the low yields, several other issues need to be addressed for photocatalytic NH 3 synthesis. First, many photocatalytic systems require an electron-donating scavenger to consume photogenerated holes, which would otherwise oxidize H 2 O to O 2 (kinetically unfavorable) or oxidize the generated NH 3 to NO x (decrease NH 3 production). However, the hole scavenger itself (e.g., alcohols) often comes from fossil fuels and would increase costs. Second, most papers reported did not specify the light intensity, which is one of the most important parameters in photocatalytic reactions. Therefore, comparison of different works or evaluation of the energy conversion efficiency is difficult. Third, standards and protocols must be established to confirm that NH 3 is derived from N 2 , similar to the procedures in ENRR. The residual contaminations in photocatalysts, which are dispersed in the solution, are more likely to be released than those on electrodes, which might lead to false positive results.

Summary for Photocatalytic Ammonia Synthesis
Despite the persistent interest and significance of photocatalytic NH 3 synthesis, little advancement has been made in comprehending the fundamental mechanisms in the reaction. The band gap must be systematically tuned for effective absorption of sunlight and high photoactivity for solar-to-ammonia conversion. Moreover, advanced in situ or operando characterization techniques are indispensable. For instance, in situ Fourier transform infrared spectroscopy (FTIR) helps in examining the adsorbed nitrogen intermediates on photocatalyst surfaces. Photocorrosion under long-term reaction is also an issue for future applications [163]. More progress will be made with the development of characterization and calculation techniques.

Plasma Catalysis for Ammonia Production
Plasma catalysis has drawn attention in the past few decades as a possible alternative to the H-B process for NH 3 production [14,34]. By ionizing the source gases via an electric discharge, nonequilibrium plasma is generated; it contains highly excited atomic, molecular, ionic, and radical species. The energy transfer to form excited species is realized by collisions between reactant molecules and high-energy electrons, where the huge mass discrepancy results in the relatively low background temperature. For example, in nonthermal plasmas, the temperature of electrons reaches 10 5 K because of their small mass, whereas large-mass ions/molecules and background gas are observed at RT. This finding is favorable for an exothermic process such as NH 3 synthesis [164] and could also reduce sintering or coking of catalysts. The highly reactive electrons, ions, atoms, and radicals in the plasma also greatly boost the kinetics, enabling NH 3 production at RT and atmospheric pressure [165].
On the other hand, the composition and properties of plasma are complicated, posing a challenge to the control and understanding of plasma-incorporated process, especially when catalysts are involved. For example, dielectric catalyst would change electric field distribution and affect the plasma characteristics [166]. In addition, the lifetime of short-lived active species can be extended on the catalyst surface, making the media more favorable for reaction [167].
As an external stimulus, plasma could also help heterogeneous catalysts to overcome scaling relations by converting N 2 into vibrationally or electronically excited states, which decrease the activation energy for N 2 dissociation without affecting subsequent reaction steps.
In a nonthermal plasma reactor induced by dielectric barrier discharge (DBD), the highest energy efficiency of 2.3 g NH 3 /kWh was achieved at a frequency of 10,000 Hz, an applied voltage of 6000 V, and a supplied N 2 /H 2 ratio of 3:1. Ru catalyst with CNT support outperformed the others, where cesium acted as a promoter, and Molecular Sieve 13X and Amberlyst 15 served as microporous absorbents [168]. Iwamoto and colleagues [169] designed a wool-like copper electrode, which was found to be an effective catalyst for NH 3 synthesis, using nonthermal atmospheric-pressure plasma by glow discharge. The energy efficiency of NH 3 production reached 3.5% with an NH 3 production rate of 3.3 g NH 3 /kWh. The catalytic activity increased during reaction runs, and this phenomenon was also observed on a series of other wool-like metal electrodes. This finding might be due to the increased catalyst surface area [170]. Mehta et al. [38] demonstrated that DBD-induced nonthermal plasma could overcome the scaling relations through vibrational excitation of N 2 in a plasma-enabled catalytic process. Based on a kinetic model that incorporates the effect of N 2 vibrational excitation in a nonthermal plasma, they observed that NH 3 production rates could be significantly improved over thermal catalytic rates at the same temperature and pressure. Additionally, the optimal catalyst shifted to sites that bind  [162] nitrogen weakly, such as Co and Ni, instead of those for thermal catalysis, which were illustrated by both calculations and experiments (Fig. 14). This work represents the first demonstration of the computationally guided design of plasma-catalyst system for NH 3 synthesis.

Homogeneous Molecular Catalysis for Ammonia Production
Another important artificial nitrogen fixation system focuses on stoichiometric transformation of coordinated dinitrogen or homogeneous catalytic reduction of N 2 on transition metal-dinitrogen complexes [35]. As a synthetic functional analogue to nitrogenase, well-defined homogeneous catalysts allow researchers to gain more insights into the N 2 reduction mechanism because these materials can be thoroughly investigated through various spectroscopic techniques, which aid in identifying molecular reactivity and elementary reactions that occur in nitrogenases. The first report of catalytic reduction of N 2 by transition metal complexes came up in 2003, when Schrock and colleagues [171] synthesized a mononuclear molybdenum complex HIPTN 3 NMo(N 2 ) (HIPT refers to hexa-iso-propyl-terphenyl) to convert N 2 into NH 3 , with 8 equivalents of ammonia per Mo atom. Nishibayashi and coworkers [172] observed that dimolybdenum-dinitrogen complex bearing 2,6-bis(di-tert-butylphosphinomethyl) pyridine pincer ligands acted as an effective catalyst for N 2 reduction to NH 3 , with 23 equivalents of NH 3 being generated (12 equivalents of NH 3 produced per Mo atom). Early studies emphasized the effect of Mo as an essential element of nitrogenases. With the development of biochemical and spectroscopic studies on nitrogenase, Fe instead of Mo was determined as the site of N 2 binding in the FeMo cofactor [8], and the central atom in FeMo cofactor was confirmed to be carbon [173,174]. In 2011, Holland and coauthors [175] reported for the first time an example of complete stoichiometric N 2 reduction by Fe complexes, although this reaction was not a catalytic process. A tris(phosphine)boranesupported iron complex was discovered in 2013; it catalyzes the reduction of N 2 to NH 3 under mild conditions, where a single iron site can stabilize various N x H y intermediates generated during catalytic N 2 reduction. A flexible iron-boron interaction possibly plays an important role in this catalytic process [176]. Holland further studied a synthetic complex with a sulfur-rich coordination sphere, which provides structural and spectroscopic implication for FeMoco-N 2 binding and nitrogenase mechanism. The results illustrate that the sulfur-rich Fe site in the FeMoco causes N 2 activation, and that Fe-S bonds can be easily reduced and broken to allow N 2 binding [177]. In addition to Mo-and Fe-based transition metal complexes, other molecular complexes incorporating metals, such as titanium and uranium, were also investigated to understand the mechanism behind enzymatic or thermal NH 3 synthesis [178][179][180].

Summary and Outlook
Developing NH 3 synthesis using sustainable or distributed approaches is becoming important to face the challenges of energy, environment, and transport issues. Homogeneous molecular catalysts are ideal for deep mechanism studies and show high activity due to their well-defined structures and facile mass transfer in the single liquid phase. Nonetheless, the difficulty in separating and recovering costly catalysts limits their practical applications. Heterogeneous catalysts normally possess complicated structures, where the exploration of the reaction mechanism is challenging. Still, these materials are the most widely studied catalysts for ENRR at present given their easy separation properties, relatively high activities, and great practical application potential. In an era when electricity is becoming more reliable for clean energy, electrochemical NH 3 synthesis is highly expected. With a tremendous potential, this process will need considerable advancement toward practical applications and require a minimum production rate in the magnitude of 10 −7 mol/(cm 2 s) [13]. In systems using aprotic electrolytes, the investigations should focus on enhancing current densities, improving catalyst/electrode lifetime, and reducing process temperatures. In systems using aqueous electrolytes, the utmost concern is how to prohibit HER as much as possible. Although limiting proton or electron availability at the electrode surface could improve the selectivity of NH 3 , it is not feasible for application under high current conditions.
Photocatalytic NH 3 synthesis uses photo-responsive catalysts or photoelectrodes, and the yields demonstrated are still far from meeting the requirements for practical use. Standardization of the operation conditions is significant for future researchers for fair evaluation and comparison.
Electron transfer and reactions on the catalyst surface are critical steps in both electro-and photocatalytic N 2 reduction. Therefore, surface reaction mechanism must be understood, and strategies that promote surface kinetics must be designed accordingly. Inspiration from enzymes and homogeneous systems demonstrate the potential to design complex active sites to improve catalyst performance at ambient conditions [181]. In situ characterization techniques, such as attenuated total reflectance-FTIR, SEIRAS, and in situ near ambient-pressure X-ray photo-electron spectroscopy, can detect reaction intermediates or active sites [81,182]. Along with DFT simulations, the reaction mechanism and screen principles would be reasonably deduced [183]. Moreover, the structure of catalysts could be analyzed more precisely by X-ray absorption fine-structure and aberration-corrected TEM, which in turn contributes to the development of new catalytic materials such as single-atom catalysts [184][185][186]. In addition, designing new reaction configuration, such as semiconductor-biological systems [7], lithium cycling electrosynthesis [68], bio-electrochemical [187], or (solar thermal) chemical looping [188,189], could decouple the linear scaling between nitrogen-binding energy and activation barrier for N 2 dissociation [190].
Considering the small NH 3 yield at present, a significant issue in (photo)electrocatalytic NH 3 synthesis is the rigorous protocol to prove that NH 3 truly comes from N 2 reduction. Ammonia contamination is caused by various potential sources during an experiment cycle, which would lead to a fake positive result. Special attention must be paid to catalysts that are either nitrogen-containing or prepared from nitrogen-containing precursors. In the premise of eliminating every possible exogenous contamination, a proper 15 N 2 control experiment is essential before drawing a convincing conclusion. In addition, utilizing various NH 3 detection methods simultaneously could make the results more accurate (for instance, NMR/indophenol method).
In plasma-enabled catalytic NH 3 synthesis, limited knowledge is available about the fundamental reaction mechanisms of the activated species. Further development in this direction could draw inspiration from traditional heterogeneous catalysis and focus on building kinetic models, which are based on in situ characterization and computational simulations.
Although molecular catalysis for N 2 reduction still requires development toward practical use, catalytic systems for N 2 fixation have been successfully developed, and molecular structures that reveal essential components for the N 2 reduction mechanism have been offered. More insightful molecule design and experimental/theoretical investigations could broaden the understanding behind reactions, providing insights into the design of heterogeneous catalysts.
Overall, flourishing alternative catalytic systems have been making great contributions to N 2 reduction reaction. With the intensive and collaborative research worldwide, sustainable NH 3 synthesis will eventually arrive and create renewable wealth for human society.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creat iveco mmons .org/licen ses/by/4.0/.