Electrochemical Nitrogen Fixation for Green Ammonia: Recent Progress and Challenges

Abstract Ammonia, a key feedstock used in various industries, has been considered a sustainable fuel and energy storage option. However, NH3 production via the conventional Haber–Bosch process is costly, energy‐intensive, and significantly contributing to a massive carbon footprint. An electrochemical synthetic pathway for nitrogen fixation has recently gained considerable attention as NH3 can be produced through a green process without generating harmful pollutants. This review discusses the recent progress and challenges associated with the two relevant electrochemical pathways: direct and indirect nitrogen reduction reactions. The detailed mechanisms of these reactions and highlight the recent efforts to improve the catalytic performances are discussed. Finally, various promising research strategies and remaining tasks are presented to highlight future opportunities in the electrochemical nitrogen reduction reaction.

The electrochemical synthetic route for nitrogen fixation has received considerable attention as it can facilitate the green production of NH 3 without the formation of harmful pollutants while using electricity generated from sustainable sources ( Figure 1C). [8,[25][26][27][28][29][30] The electrochemical nitrogen reduction reaction (eNRR) proceeds as follows: 1) N 2 molecules approach the electrode surface; 2) electrical charges are exchanged and protons are injected to break bonds and hydrogenate N 2 (bond dissociation energy = 941 kJ mol −1 ); 3) NH 3 is released. [26,31] The net nitrogen reduction reactions in acidic and alkaline electrolytes follow different reaction pathways but have the same equilibrium potential of 0.56 V SHE . [32,33] N 2 + 6H + + 6e − → 2NH 3 ( acidic electrolyte condition ) (1) There are several review articles that have summarized studies of electrochemical nitrogen reduction (eNRR). [17,[34][35][36][37][38][39][40][41][42][43][44][45][46][47][48][49][50] In this review, we aim to classify recent eNRR studies into two distinct classes, and present the current mechanistic understanding of each pathway. Based on the reaction pathways (as shown in Figure 2), studies on eNRR can be divided into two categories: direct nitrogen reduction reaction (DNRR) and indirect nitrogen reduction reaction (INRR). The use of electrons is the most significant difference between DNRR and INRR. The term Indirect was chosen for Li-mediated pathway because electron is used to reduce "Li + " to metal Li, whereas electron in DNRR pathway is used to break and activate the "nitrogen" triple bond. DNRR refers to direct electrochemical nitrogen activation on a heterogeneous catalyst surface. In this pathway, heterogeneous catalysts only provide the binding sites for dinitrogen molecules. In contrast, Figure 1. Nitrogen fixation. Three representative routes to produce ammonia from N 2 : A) Thermochemical Haber-Bosch. B) biological N 2 fixation by nitrogenase enzyme and C) electrochemical nitrogen reduction system.
INRR involves reactive metal species, such as Li, with a strong reducing ability to break triple nitrogen bonds (N≡N). While the electrons from the electrode are utilized to "directly" activate dinitrogen in the DNRR pathway, electrons in INRR are solely consumed to form Li metal films, which "indirectly" convert dinitrogen into lithium nitride. In other words, instead of the sequential activation of nitrogen on the catalyst surface, INRR harnesses metal nitride intermediates to produce NH 3 from dinitrogen in a nonaqueous system. Although both selectivity (Faradaic efficiency (FE)) and reactivity (current density) in the INRR have been remarkably improved, the high energy input required to initiate Li plating on the electrode remains a considerable roadblock.
This review highlights the most recent progress in the two eNRR methods, DNRR and INRR. We aim to provide useful guidelines for future research by identifying the most significant developments and critical challenges. Initially, we focus on the reaction mechanisms of DNRR and INRR, based on the findings from electrokinetic, in situ, and ex situ spectroscopic, and various computational studies for eNRR. The effects of the catalyst design, including the morphology and heterostructure, on the DNRR are presented, considering the binding affinity of dinitrogen molecules on the catalyst surface during catalysis. The engineering strategies for the electrochemical components of INRR are discussed, followed by an overview of recent advances in Limediated nitrogen reduction. Finally, we point to the most salient challenges currently associated with DNRR and INRR, suggesting potential research avenues to prompt research advances in this fledgling field.

Electrochemical Mechanisms for DNRR and INRR
The sluggish nitrogen reduction kinetics are primarily attributed to the thermodynamically stable triple N─N bond in N 2 , for which the bond (941 kJ mol −1 ) and ionization (15.6 eV) energies are considerably high. [51][52][53][54] Moreover, N 2 has a negative electron affinity (−1.9 eV) and requires additional electrical energy for electron injection. For instance, the first protonation of *N 2 to *NNH in the DNRR is endothermic (∆H 0 = +37.6 kJ mol −1 ). [11,55,56] Furthermore, electron excitation in N 2 is significantly hindered by the large gap between the highest occupied and lowest unoccupied molecular orbitals: 10.82 eV (or 1044.0 kJ mol −1 ). Finally, compared with other gases, N 2 has low solubility in aqueous solutions (0.6 mmol L −1 in water at room temperature). Herein, we present the recent understanding of the mechanism of each pathway.

Proposed Mechanisms for DNRR
Three elementary steps are typically considered for DNRR ( Figure 3A): 1) chemical adsorption of N 2 molecules and protons onto the cathode surface, 2) activation of N 2 and hydrogenation, and 3) rearrangement and desorption of the product NH 3 or other products (e.g., hydrazine [N 2 H 4 ] and diazene [N 2 H 2 ]) from the electrode surface and their migration into the electrolyte. [10,26,57,58] The Haber-Bosch process produces NH 3 via a dissociation pathway because of high pressure and temperature operating conditions. In contrast, biological systems adopt an associative pathway to generate NH 3 , which involves N 2 fixation via the enzymatic process of nitrogenases. [31,59] For electrochemical nitrogen reduction, an associative path would be more appropriate because ambient reaction conditions are used in electrochemical systems, necessitating nitrogen adsorption and protonation as the first step for an efficient eNRR. [60] The associative pathway can be further categorized into alternating and distal pathways based on the sequence of hydrogenation on the nitrogen atoms. Specifically, in the associative alternating pathway, hydrogenation occurs alternatively on the two nitrogen atoms, while in the associative distal pathway, it occurs on one nitrogen atom and then the other.
Because N 2 adsorption/activation is consistent with the formation of N 2 H* species on the electrode, Medford et al. proposed that Brønsted-Evans-Polanyi (BEP) relationships result in a volcano plot, which helps to visualize the ability of metals to reduce N 2 electrochemically ( Figure 3B). [25,26,98] The model focuses on the eNRR rate-determining step (RDS) on catalyst materials, depending on the eNRR pathways (solid lines for the dissociative pathway and dotted lines for the associative pathway) and the surface atomic geometry (flat surfaces for black and stepped surfaces for red). For metals on the left side of the volcano plot, the hydrogenation process is considered the RDS in both the associative and dissociative pathways (*NH 2 + H + + e − → NH 3 on the stepped surfaces and *NH + H + + e − → *NH 2 for the flat surfaces). However, for metals on the right side of the plot, the RDSs are changed depending on the reaction pathways; the first  . Reaction pathways of DNRR and INRR A) Scheme for three possible DNRR mechanisms: dissociative, associative alternating, and associative distal pathways. Reproduced with permission. [18] Copyright 2016, Elsevier. B) Volcano plot for DNRR on metal electrodes depending on the mechanistic pathways. The volcano plot for the HER is overlaid for comparison. Reproduced with permission. [26] Copyright 2012, RSC. C) SEIRAS spectra of an Rh film electrode in N 2 -saturated 0.1 m KOH. Reproduced with permission. [3] Copyright 2020, Wiley VCH. D) Overall reaction schemes of Li-mediated electrochemical nitrogen reduction on Cu electrode. Reproduced with permission. [78] Copyright 2019, Elsevier. E) Stepwise Li-mediated catalytic cycle for nitrogen reduction to ammonia. Reproduced with permission. [76] Copyright 2020, Springer Nature. F) Heterogeneous nitrogen reduction on lithium metal surface, which is similar to DNRR pathway. Lithium (or LiH, Li 3 N) layers are not involved in the cycle, instead, provide the binding sites to N 2 and nitrogen-containing adsorbates. Reproduced with permission. [61] Copyright 2020, Wiley VCH. www.advancedsciencenews.com www.advancedscience.com protonation step of N 2 (N 2 → *N 2 H) for the associative pathway and the N 2 splitting step (N 2 → *2N) for the dissociative pathway regardless of the surface atomic geometries. It is concluded that the flat surfaces of transition metals such as Sc, Y, Ti, and Zr effectively reduce N 2 to NH 3 through the dissociative mechanism with the applied potential of −1 V NHE to −1.5 V NHE , by comparing the free energies of *N (white area), *H (gray-shaded area), and intermediates (those species are stable on the left side area from the vertical lines a) *NH, b) *NH 2 , c) *N 2 H, and d) *N 2 H 2 ) on the catalyst surfaces. Furthermore, the slow rate of N 2 reduction observed on catalysts such as Pt, Pd, or Ru can be explained by free energy calculations. These calculations suggest that, under negative bias, these catalyst materials become saturated with hydrogen, hindering the adsorption of nitrogen and subsequently impeding NH 3 production. Also, N vacancy incorporated Marsvan Krevelen (MvK) mechanism has been suggested by Yang et al. In the MvK mechanism, a lattice N on the surface of transition metal nitride catalysts is consumed to produce NH 3, leaving an N vacancy, which is reoccupied by adsorbed N 2 . [62] Despite the efforts to reveal the nature of DNRR process, eNRR mechanism, however, has remained elusive. [63][64][65][66]

Spectroscopic Studies
Due to the low yields of NH 3 from DNRR, it is challenging to study the reaction intermediate species. A few reports have investigated the reaction intermediates using surface-enhanced infrared spectroscopy (SEIRAS) and differential electrochemical mass spectrometry (DEMS) ( Figure 3C). [67][68][69] Using the SEIRAS, Yao et al. discovered the potential-dependent molecular characteristics on the Rh electrode surface. Figure 3C inset plots the band center (cm −1 ) and intensity of each stretching mode of N = N & Rh-H species versus applied voltages. The authors revealed that the extent of coverages of the N 2 H x and H species on the Rh electrode strongly depends on the applied potentials. Notably, maximum N 2 H x coverage was obtained between 0 and −0.2 V, and H deposition on the electrode surface increased below −0.2 V (inset of Figure 3C). Also, by integrating SEIRAS and DEMS, it was discovered that the final protonation of *N 2 H 2 to NH 3 on the Rh electrodes is the RDS, which is consistent with the theoretical expectation for the associative eNRR mechanism, as suggested by Medford et al. N 2 H x (0 ≥ x ≥ 2) was detected using SEIRAS and DEMS, with an N═N stretching mode at ≈2020 cm −1 and a signal at m/z = 29 (N 2 H + ), respectively. However, the N 2 H 2 + species was not detected at m/z = 30. The same group also determined the DNRR mechanisms on the Ru, Au, and Pt electrode surfaces, but the N 2 H 2 + species have not been detected thus far. These studies suggest an alternative mechanism via the facile formation of N 2 H x species on the catalysts and decomposition in the electrolytes.

Development of Li-Mediated Nitrogen Reduction Methodology
The Li-mediated eNRR started back in the 1990s. After it was first suggested by Fichter et al. in 1931, [70] Tsuneto et al. reported Li-mediated electrochemical N 2 fixation via organic solutions containing a small amount of proton source as an electrolyte to suppress the competing reaction, HER. [71,72] The experiments were performed under 50 atm N 2 pressure with LiClO 4 (0.2 m) and ethanol (0.18 m) in a tetrahydrofuran (THF) electrolyte. The selectivity toward NH 3 was low, and it required a high nitrogen pressure.
McEnaney et al. proposed a strategy to separate N 2 reduction from the subsequent protonation to NH 3 in their molten-salt system. [73] The suggested phase diagram of Li─N─O─H incorporates the following Li cycle steps: 1) Reductive LiOH electrolysis generates Li in the hydrogen-free environment. The phase diagram indicates that Li exists as a stable phase above −3.3 V versus the standard hydrogen electrode (SHE). 2) Introduction of 1 bar of nitrogen gas into the activated Li surface facilitates Li nitridation. Lithium nitride (Li 3 N) is demonstrated to be more stable than Li for all potentials, implying that Li 3 N formation is energetically favorable under mild conditions. 3) Li 3 N is rapidly hydrolyzed to NH 3 upon adding water at 0 V versus SHE. Li 3 N decomposes to stable LiOH and NH 3 . Although McEnaney et al. successfully synthesized NH 3 with 88.5% FE at low N 2 pressure, their system had several practical limitations; for example, high operating conditions (> 400°C) were required for providing a continuous liquid phase of molten salt electrolyte, which inevitably resulted in the formation of liquid Li metal (melting point = 180.5°C). Thus, additional steps were necessary to separate liquid Li. Kwiyong et al. reported an alternative strategy using a Li-ion conducting glass ceramic material (LISICON) system. Although elevated temperature (220°C) and stepwise processes from different reaction chambers are required, LISICON allows Li + ion migration and Li deposition at room temperature. [74] Recently, several studies have been conducted on continuous Li-mediated nitrogen reduction catalysis under mild conditions without any separation steps. [75][76][77] Protons from external proton donors or anodic reactions are fed into the catholyte, which directly reacts with the Li 3 N layers to yield NH 3 . This approach has received significant attention as it is a one-pot synthesis without harsh operating conditions.

Proposed Mechanisms for INRR
The overall reaction of Li-mediated electrochemical nitrogen reduction on the electrode is depicted in Figure 3D. This diagram well illustrates the key components of the reaction. [78] The mechanism of continuous Li-mediated nitrogen reduction typically involves three steps ( Figure 3E): 1) a highly reactive Li layer is formed on the electrode surface, 2) Li 3 N is formed when Li reacts with dissolved N 2 , and 3) NH 3 is produced via sequential protonation, resulting in Li + ions in the electrolyte that close the catalytic cycle. Based on this concept, the Li redox couple is "mediated" to complete mixed electrolytic and chemical reactions. The detailed mechanism can be subcategorized depending on whether each pathway is concerted or proceeds stepwise.
Cai et al. experimentally demonstrated three stepwise mechanisms, as suggested in Figure 3E. [79] They observed that NH 3 continued to be produced from N 2 even when the voltage was no longer applied after a certain period of electrolysis. They claimed that the Li metal layers on the electrode remained sufficiently www.advancedsciencenews.com www.advancedscience.com reactive to continue the N 2 splitting cycle until Li was completely consumed. Several control experiments were performed to test this hypothesis. They found that the reductive current values remained almost unchanged under N 2 and Ar environments or with and without ethanol. This implies that N 2 reduction or ethanol splitting is not involved in the electrochemical process, and the injected electrons are only consumed for Li plating.
For the concerted pathway of (1) and (3), the expulsion of NH 3 from Li 3 N and Li redeposition was suggested. Andersen et al. suggested kinetic models wherein electrochemical Li deposition is accompanied by hydrogenation of the adsorbed N 2 molecules. [80] Based on this mechanism, they reported that highenergy electrons from highly negative potentials could be restored in the Li layer in the absence of an electric field and could be used to produce NH 3 from N 2 . Additionally, Ma et al. suggested a (1) and (2) concerted model for N 2 fixation in the Li-N 2 battery cycle. [81] During the discharging step, Li + reacts with nitrogen to form Li 3 N via an electrochemical process, while the Li anode provides Li + ions.
To elucidate the cycle with a concerted proton-electron transfer model, Schwalbe et al. hypothesized that the long-lived Li metal layer might provide heterogeneous active sites, which assist the coupled proton-electron transfer in the catalytic cycle. [61] Figure 3F depicts the stepwise reaction pathways to yield NH 3 from N 2 , analogous to the direct eNRR pathway. In this mechanism, Li + is no longer participating in the catalytic cycle. Instead, N 2 splitting and its hydrogenation become voltage-dependent processes on the Li metal layer. The authors suggested that the mixed Li x N y H z species (e.g., lithium imide or lithium amide) formed on the electrode surface can act as catalytically active intermediates for the eNRR.

Electrokinetic Studies
Several research groups have endeavored to elucidate the complex kinetic nature of Li-mediated nitrogen reduction. [82,83] Lazouski et al. reported continuous Li-NRR (0.2 m LiClO 4 and 0.2 m ethanol in THF) under ambient conditions. Furthermore, they attempted to analyze the overall kinetics of NRR through a series of electrokinetic studies. [75] Figure 4A depicts two possible reaction routes in the Li-NRR. After being plated on the electrode, Li can react with either N 2 to generate Li 3 N or with a proton source (HA) to produce molecular hydrogen. Li 3 N yields NH 3 by reacting with HA. The rate of each pathway can be expressed as follows The reaction order for nitrogen pressure was found to be firstorder ( Figure 4B) from the dependence of the partial current density toward NH 3 (i NH 3 ) on the nitrogen partial pressure (Equation (2)). Similarly, Figure 4C displays the log-log plot of the i NH 3 versus the concentration of ethanol, a proton donor. The NH 3 partial current first increases with ethanol concentration up to 0.1 m, and then it monotonically decreases, exhibiting a −1.5 order dependence on ethanol concentrations. The authors claimed that the obtained negative reaction order was attributed to the con-tribution from the hydrogen evolution ( Figure 4A, top), which depleted the available Li layer for nitridation.
The subsequent hydrogenation of Li 3 N to NH 3 was assumed to be a rapid step because the Li 3 N layer was scarcely observed in the cathode. Li 3 N formation was considered the RDS. Because the partial current toward hydrogen was considerably higher than that toward NH 3 , the authors further assumed that 1) the reaction rate should be r NH 3 ≪ r H 2 , 2) Li was in a quasisteady state during NRR (the rate of change in Li concentration was ≈0), and 3) most of the applied current was used to generate the Li layer. The abovementioned experimental results and assumptions lead to the following relationships Subsequently, the obtained Li concentration (Equation (4)) was employed to correlate FE NH 3 and total current. FE NH 3 showed a one-half-order dependence on the total current, resulting in the following expression, which explains the dependence of FE NH 3 on the reacting species and current The ethanol reaction order x was deduced to be unity and the Li reaction order, , and , to 2 and 3, respectively.
In the subsequent study, Lazouski et al. further endeavored to describe the role of proton donors in terms of transport kinetics. They attempted to correlate the permeability of the solid electrolyte interface (SEI) and the types/concentrations of proton donors. [77] They postulated that two different types of SEIs (permeable and impermeable) coexist in the system; the permeable SEI fraction was denoted by . The effective diffusivity of 1) nitrogen and 2) proton donors through the SEI was considered as a critical parameter of in their model Where D j is the diffusivity and N j is the total diffusion flux of species j through the SEI; C j,bulk and C j,s are the concentrations of species j at the SEI and at the reactive electrode surface beneath the SEI, respectively. At a steady state, the rate of the NRR should be equal to that of nitrogen diffusion through the SEI When the lithium layer is formed by reduction, it is spontaneously consumed through either hydrogen evolution (upper) or ammonia formation (bottom) with the help of proton donor. Log-log plot of partial current density toward NH 3 with varying B) partial pressures of N 2 and C) ethanol concentration as proton donor. Reproduced with permission. [75] Copyright 2019, Elsevier. Simulated Faradaic efficiency toward NH 3 depending on the proton donor concentration at different D) rate constant (k N 2 ) for nitrogen reduction and E) diffusivity of nitrogen (D N 2 ). Reproduced with permission. [77] Copyright 2022, American Chemical Society.
The rate of all the reactions involving the proton donor can be expressed as follows. r HA,red indicates the reduction rate toward hydrogen evolution, and r HA,prot is protonation rate of dinitrogen molecule to ammonia r HA,red = k HA C HA,S C Li (11) r HA,prot = 6k N 2 C N 2 ,S C Li (12) Similar to Equation (3), the accumulation rate of the Li layer can be expressed using the current density and consumption rate (Equation (12)) The surface concentrations of N 2 , HA, and Li species (C N 2 ,S , C HA,S C Li ) can be obtained from Equation (12) as assuming steady-state conditions and constraints from Equations (8) and (9). Considering the stoichiometry of the reacting species and the relationship between FE and the reaction rate, the following equations are derived www.advancedsciencenews.com www.advancedscience.com The rate of NH 3 evolution is limited by the diffusivity of nitrogen (Equation (8)). The FE for hydrogen is then derived by subtracting the NH 3 rate from the total HA-involved reaction rate Therefore, to minimize the FE H 2 , the proton diffusion rate should be controlled to limit NH 3 formation (Equation (15)). Based on the kinetic diffusion model, the authors attempted to reveal the effects of diffusivity and rate constants. Figure 4D,E shows that the peak FE NH 3 is expected to be more strongly affected by the relative diffusivities of N 2 and H 2 than by the overall rate constant for the NRR. Because commonly used proton donors are sterically bulky, the nitrogen diffusivity in the SEI layer is typically higher than that of the proton donor. Thus, even if the rate constant for N 2 reduction is lower than for proton donor reduction, a high FE can still be achieved as diffusion of nitrogen is favored over proton in SEI. This implies that the SEI serves as a protective layer for selective NH 3 production by suppressing undesirable hydrogen evolution and exclusively promoting N 2 transport.

Catalyst Design Strategies for Enhancing DNRR Performances
The practical applications of the eNRR are hindered by the slow reaction kinetics and competitive HER, resulting in low FE. N 2 is composed of two N atoms, each with five valence electrons in the 2s and 2p orbitals. The electrons occupy all bonding orbitals and one * orbital, making this chemically stable form. In terms of charge exchange, donating electrons from an occupied N 2 orbital to an empty metal d orbital, with the back donation from an occupied metal d orbital to an empty N 2 * orbital, weakens the triple N─N bonds and promotes eNRR. [84] Therefore, the electrocatalyst design can efficiently enhance eNRR performance because the charge exchange property is directly related to the chemical features of the catalyst surface. [85,3] Direct electrochemical NH 3 synthesis from nitrogen and water is a simple alternative to the conventional Haber-Bosch method. [3] Carbon-based materials, metals, nonmetals, metal oxides, metal sulfides, metal nitrides, bimetallic and ternary nitrides, and single-atom catalysts have been developed as catalysts for the eNRR in aqueous electrolytes. The intrinsic activity of the active sites is influenced by the electronic structure of the catalyst, which can be regulated through defect/strain engineering. [85] In contrast, the apparent activity, which depends on the yield, is influenced by the structure and morphology of the catalyst. The apparent activity of the electrocatalyst can be enhanced through surface area modification, pore engineering, and hybridization. Herein, we discuss the effects of defect engineering, surface geometry/morphology, and heterostructure formation on the eNRR performance of electrocatalysts in aqueous electrolytes.

Defect Engineering
Defect engineering is an efficient strategy for regulating the electronic structure of electrocatalysts and consequently enhancing N 2 adsorption. [86,87] Moreover, defect engineering alters the adsorption energy of the intermediates, thereby suppressing the HER and improving the selectivity of eNRR. The defects can primarily be induced by doping with heteroatoms or by forming vacancies. Doping induces charge redistribution and improves the charge transfer abilities of the electrocatalyst. [88] Boron, nitrogen, and sulfur are commonly used as dopants in electrocatalysis. When graphene is doped with 6.2% boron, the NH 3 yield rate and FE of the doped graphene increase by 5 and 10 times, respectively, compared with those of the undoped graphene. [86] This increase in performance is attributed to the difference in electronegativity between boron (2.04) and carbon (2.55). The electron-deficient boron sites have an increased tendency to bind to N 2 , thus enabling the formation of B─N bonds and the subsequent production of NH 3 . Similarly, Wang et al. reported the role of Ndoping in improving the eNRR activity of NiO nanosheet arrays on carbon cloth (NiO/CC). [88] Initially, a hydrothermal method was used to synthesize NiO/CC, which was subsequently doped with nitrogen (N-NiO/CC) by annealing in an Ar/NH 3 atmosphere. As deduced from density function theory (DFT) computations, N doping modified the electronic structure of NiO, thereby promoting the activation of N 2 , improving the stabilization of the key intermediate *NNH, and decreasing the reaction energy barrier. Consequently, the eNRR activity was enhanced. N─NiO/CC resulted in an NH 3 production rate of 22.7 μg h −1 mg −1 and 7.3% FE in 0.1 m LiClO 4 aqueous electrolyte at −0.5 V versus the reversible hydrogen electrode (RHE). In contrast, vacancies create active sites with abundant localized electrons, which can weaken N≡N. [89] Peng et al. reported the eNRR activity of carbon nitride with N 2C vacancies, which was grown on carbon fiber paper (CN/C) through chemical vapor-assisted synthesis. [90] CN with perfect planes is characterized by low catalytic activity, whereas CN with N 2C defects exhibits strong interactions with N 2 . The growth conditions can control the number of vacancies on CN. CN/C 600 synthesized at 600°C has more catalytic sites and a higher electrochemically active surface area (ECSA) than CN/C 500 . The ratio of deconvoluted peaks (C═N─C/N─C 3 ) in the N 1s spectra of CN/C 600 and CN/C 500 (Figure 5A) indicates the presence of C═N─C N 2C vacancies in CN/C 600 . In addition, the increased ratios of D/G peaks in the Raman spectrum of CN/C 600 ( Figure 5B) indicate an increased number of defects compared with those in CN/C 500 and bare carbon paper. Based on the slope of the m NH3 -time plot, the NH 3 yield and FE for CN/C 600 at −0.3 V versus RHE were evaluated as 2.9 μg mg cat −1 h −1 and 16.8% FE, respectively (5.7 times higher than those for CN/C 500 ) ( Figure 5C,D).
Jin et al. reported the eNRR activity of W 2 N 3 nanosheets with stable nitrogen vacancies. [91] Long-term electrolysis and postcharacterization techniques indicated the stability of nitrogen vacancies in the W 2 N 3 nanosheets. DFT calculations predicted that the high stability of these vacancies could be attributed to the 2D confinement effect and the high valence state of the W atoms. Furthermore, active catalytic centers could be created by incorporating anions into metal oxides, thereby improving the eNRR Figure 5. Defect engineering of electrocatalysts. A) N 1s XPS spectra of CN/C 500 and CN/C 600 , B) Raman spectra of CN/C 500 , CN/C 600 , and C paper, C) NH 3 yield rate, and D) Faradaic efficiency of CN/C 600 and CN/C 500 at different potentials. Reproduced with permission. [87] Copyright 2020, American Chemical Society. E) TEM image, F) XRD pattern, G) EPR spectrum, and H) FE and NH 3 production rate of various C-Ti x O y /C hybrids. Reproduced with permission. [89] Copyright 2019, Wiley.
performance. [92] C─Ti x O y /C structures were obtained by annealing MIL-125 (Ti) (a Ti-based metal-organic framework) at different temperatures (800-1100°C). The oxygen vacancies in TiO 2 were substituted with carbon atoms, thus forming Ti─C bonds in C─Ti x O y . The high-resolution transmission electron microscopy (TEM) image of M-1000 (MIL-125 (Ti) annealed at 1000°C) indicates the uniform dispersion of TiO 2 nanoparticles (NPs) ranging between 3 and 20 nm in an amorphous environment ( Figure 5E). In the X-ray diffraction (XRD) pattern ( Figure 5F), peaks attributed to tetragonal rutile or anatase are observed. However, as the temperature increases from 800 to 1100°C, peaks attributed to TiC are more dominant than rutile or anatase titania. The increased intensity of the peaks in the electron paramagnetic resonance (EPR) spectrum indicates a higher concentration of oxygen vacancies at higher temperatures ( Figure 5G). The eNRR performance of C─Ti x O y /C is improved, with an NH 3 yield rate of 14.8 mg h −1 mg cat −1 and 17.8% FE ( Figure 5H).

Surface Geometry/Morphology Engineering
The apparent activity of the electrocatalyst can also be enhanced by tailoring the morphology of nanostructures and engineering the surface chemistry. [93][94][95] 3D bimetallic PdRu porous nanomaterials with interconnected voids and skeletons exhibit promising eNRR performance. [93] These porous electrocatalysts provide numerous active sites for reducing N 2 to NH 3 and facilitate the mass transfer of N 2 . Porous bimetallic PdRu provides a satisfactory NH 3 yield, which declines by only 8% after long-term stability tests. Also, engineering the surface chemistry through appropriate electrolyte is important for enhancing eNRR performances. Wang  In this case, Na 2 SO 4 can be considered a promising electrolyte for the eNRR on the porous binder-free Au/NF electrode. [96] Engineering surface features on an atomic scale has shown several insightful results for improving eNRR performance. For example, MoS 2 nanoflowers with disordered atomic arrangements and crystal defects are regarded as efficient catalysts for the eNRR. [97] The disorder and defects improve the stability of the nanoflower by tuning the properties and activity of the reactive sites. Irregularly structured catalysts with high-index facets and multiple edge/corner sites are favored because of their effective binding with the reactants. [98] In contrast, Hao et al. showed that the eNRR performance of a Ni 3 S 4 catalyst with low energy and a smooth surface was higher than that of those with distorted surfaces. [94] Ni 3 S 4 prepared using two different synthetic protocols resulted in extended flat facets and irregular surfaces, as shown in the scanning electron microscopy (SEM) images in Figure 6A (right image for the flat facet and left inset image for the irregular surface in the insets). The eNRR reaction products collected after 2 h were quantified using 1 H-NMR ( 1 H nuclear magnetic resonance) spectroscopy under extended scan times, and spurious signals were eliminated based on control experiments under varied synthesis conditions. The catalytic activity of low-energy Ni 3 S 4 with flat facets was three times higher than that  [94] Copyright 2021, American Chemical Society. E) HAADF-STEM image and corresponding elemental mapping analysis of Cu 1.81 S hexagonal nanoprism. F) aspect ratio of the nanoprisms. G) NH 3 production yield and Faradaic efficiency. H) exposed facet densities (nm −1 ) at basal (EFD-B; red) and prismatic (EFD-P; blue) planes. I) simulations of H* transfer configurations on Cu 1.81 S(010) and Cu 1.81 S(100). Reproduced with permission. [99] Copyright 2022, American Chemical Society.
of Ni 3 S 4 with a distorted surface ( Figure 6B). An NH 3 production rate of 1.28 μg mg −1 h −1 and 6.8 ± 3.3% FE were achieved for Ni 3 S 4 with smooth facets, whereas the FE of the distorted Ni 3 S 4 was only 1.9 ± 3.6%. According to DFT calculations, the availability of multiple vacant Ni sites and the competing nitrogen and hydrogen adsorption influence the enhanced eNRR activity of the smooth facets ( Figure 6C,D). Coordinatively vacant Ni sites in the flat low-energy facets are readily available for N 2 binding, thus improving the eNRR performance.
Accordingly, Jin et al. emphasized the importance of the coordination environment in catalytic activity and reported the influence of the atomic geometry of NPs on the eNRR kinetics. [99] The dependence of eNRR activity on the surface geometry of copper sulfide (Cu 1.81 S) nanocatalysts with exposed facets of (100) and (010) for smooth and zigzag planes, respectively, was investigated. High-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) images indicate the uniform distribution of Cu and S with the aspect ratio differences from 0.13 for CS-10 to 4.37 for CS-120 (x and y indicate zigzag and smooth planes, respectively) ( Figure 6E,F). The NH 3 production rate and FE increase with the increasing length of the hexagonal nanoprisms ( Figure 6G); the improvement in the eNRR performance is associated with the prismatic plane density ( Figure 6H). Also, as the prismatic plane density increased, lower current density was obtained at −0.5 V RHE under Ar saturated electrolyte. This result indicates that the zigzag surface geometry limits the competitive HER, which is consistent with the results of eNRR. DFT calculations reveal that the protruded atoms on the zigzag planes of Cu 1.81 S nanoprisms have a reduced proton transfer distance from 2.81 Å on the (100) to 1.71 Å on the (010), and H efficiently migrates from H*@S to N 2 *@Cu, thus improving the eNRR kinetics ( Figure 6I).
The catalytic performance of single-atom catalysts can be improved by engineering their geometry through metal atoms or coordination anions. [100,101] For example, a tungsten singleatom catalyst with oxygen and nitrogen coordination and a metal loading exceeding 10 wt% provides satisfactory eNRR performance. [101] The O and N coordination on tungsten influences its adsorption behavior and selectivity, thus improving the performance. Li et al. reported a strategy of interfacial polarization to facilitate the breaking of N≡N bonds and enhance NH 3 production. [102] The electric fields produced from the protruding Fe single-atom catalysts immobilized on MoS 2 can polarize N 2 by attenuating the interfacial energy barrier. The resultant polarization fields between N 2 and Fe-MoS 2 drive additional electrons into the antibonding orbitals of N 2 , thus improving the NH 3 production rate. Therefore, the morphology of the nanomaterial affects the electric field that is generated to polarize N≡N.

Heterostructure
Heterostructure engineering is a facile strategy to enhance the catalytic activity of nanomaterials. Owing to the synergy from individual components in multicomponent nanostructures (hybrid structures), their eNRR activity is higher than that of the individual components. [103,104] The interface between two components in hybrid structures imparts unique properties to the individual components, promoting abundant electron channels and accelerating the electron transfer rate. [105] Interface engineering primarily includes metal (derivative)-carbon, metal (derivative)metal oxide, and intermetallic interfaces. [106] SnO 2 quantum dots (QDs) distributed on reduced graphene oxide (SnO 2 /RGO), synthesized via self-propagating combustion, are demonstrated to be efficient catalysts for producing NH 3 at a rate of 25.6 μg h −1 mg −1 and 7.1% FE in 0.1 m Na 2 SO 4 at −0.5 V (vs RHE). [107] The RGO support improves the conductivity of the SnO 2 QDs and prevents their agglomeration, thereby increasing the surface area and number of active sites for N 2 adsorption. DFT calculations reveal that the highly conductive hybrid structure lowers the energy barrier of *N 2 → *N 2 H and promotes eNRR activation. nanocomposite. [104] The interface engineering of 2D/2D materials enables face-to-face contact, facilitating strong electronic interactions at the interface. [108] In the MoS 2 /C 3 N 4 heterostructure, electrons migrate from C 3 N 4 to MoS 2 across the interface, thus creating a charge density difference. This electron redistribution significantly increases the conductivity of MoS 2 /C 3 N 4 . The crucial intermediate *N 2 H is stabilized on the Mo edge sites of MoS 2, and the reaction energy barrier decreases. Furthermore, *H is adsorbed on the S edge sites, thus protecting the NRR-active Mo edge sites from the competing HER.
The synergistic effect between Fe 3 O 4 and Au in Au-Fe 3 O 4 NPs results in a higher eNRR activity than that of Fe 3 O 4 , Au, and Au@Fe 3 O 4 (core@shell) NPs. [103] Au-Fe 3 O 4 and Au@Fe 3 O 4 NPs synthesized through a simple one-pot wet-chemical method have similar morphologies (Figure 7A). The Au 4f XPS spectra ( Figure 7B) show that the ratio of Au 3+ to Au 0 in Au-Fe 3 O 4 NPs is higher than that in the other catalysts. Additional Au─O bonds in the Au-Fe 3 O 4 NPs facilitate their strong binding to the intermediates, thereby enhancing the eNRR. The Au-Fe 3 O 4 NPs provide an NH 3 yield rate of 21.42 μg mg cat −1 h −1 and 10.54% FE at −0.2 V versus RHE ( Figure 7C). The enhanced N 2 chemisorption peak of Au-Fe 3 O 4 NPs in the N 2 temperature-programmed desorption curves demonstrates the unique role of the heterojunction structure in the eNRR performance. Theoretical calculations indicate that the N 2 to *N 2 H transition occurs on the Fe atoms, while Au optimizes the adsorption energy of NRR intermediates, making the NRR path on Au-Fe 3 O 4 along an energetically favorable process. A metastable amorphous material with dangling bonds is known to be more catalytically active than its crystalline counterpart. [109] Li et al. reported that the eNRR activity of amorphous Au NPs supported on bisubstrate CeO x -RGO (a-Au/CeO x -RGO) was higher than that of their crystalline counterparts. This hybrid catalyst with a low Au loading (1.31 wt%) was prepared via coreduction. CeO x facilitates the conversion from crystalline to amorphous Au, whereas RGO serves as a support for the uniform dispersion of the a-Au NPs. The high concentration of unsaturated coordination sites in the metastable amorphous material consequently improves the chemical reactivity with N 2 molecules. Subsequently, Au sub-nanoclusters (≈0.5 nm) embedded on TiO 2 with a loading of 1.542 wt% were demonstrated to be promising eNRR catalysts. [110] The size of the Au clusters significantly influences the formation of NH 3 from N 2 . Moreover, in an acidic electrolyte, the partially positively charged Au active centers, through the Au─O─Ti bond, preferentially adsorb electroneutral N 2 than H + cations, resulting in an efficient eNRR.
To achieve high catalytic activity, Suryanto et al. reported a strategy for the polymorphic engineering of NPs. [111] Ru clusters were decorated on the 1T and 2H polymorphs of MoS 2 to form Ru/1T-MoS 2 and Ru/2H-MoS 2 hybrids, respectively. Figure 7D depicts www.advancedsciencenews.com www.advancedscience.com STEM images of the hybrid and crystal structures of 2H and 1T-MoS 2 . The catalytic activity and selectivity of Ru-decorated semiconducting 2H-MoS 2 toward eNRR are higher than that of metallic 1T-MoS 2 ( Figure 7E). Also, evaluation of HER performances revealed the crystal structure dependent eNRR/HER relationship. (Ru)/1T-MoS 2 showed better HER performances than those of (Ru)/1H-MoS 2 ( Figure 7F). The eNRR performances show that Ru functions as N 2 active site and 1H-MoS 2 provides protons to form NH 3 at the interface of Ru/1H-MoS 2 . Thus, a high eNRR activity can be obtained by tuning the HER kinetics of MoS 2 via polymorphic engineering. The synergistic effects between different metals in intermetallic compounds enable them to be potential candidates for the eNRR. [112][113][114] Pd-, Au-, and Ru-based alloys are commonly used as intermetallic compounds. According to Tong et al., body-centered cubic PdCu NPs, obtained by annealing their face-centered cubic (FCC) counterparts, are more favorable for the eNRR because the energy barrier is lower than that of FCC PdCu. The phase transformation of PdCu positively shifts its d-band center and improves its ability to bind nitrogen molecules. DFT calculations reveal that the strong d-d coupling between Pd and Cu enhances the transfer activity, enabling an efficient eNRR. [115]

Engineering SEI Layers on the Li Cathode
As discussed in Section 2.2, the reaction is initiated after a highly reductive potential is applied to form a Li metal surface from Li + . [116] The highly reductive environment can be detrimental to electrolytes, proton sources, or other electrolyte components, which electrochemically disintegrate into the SEI, and this has been extensively investigated in Li-ion battery studies. The SEI is considered a "passivation layer" that exclusively allows Li cations to reach the electrode surface and acts as an ionic conducting separator to prevent further electrolyte decomposition or undesirable Li dendrite formation. [117,118] The SEI promotes the stability and performance of the Li-mediated NRR. [77,119] Moreover, it can control the relative diffusion rates of Li + , H + , and N 2 (r Li + , r H + , r N 2 ), which are pivotal variables to affect the rate and selectivity. Despite considerable investigation, the formation mechanism and structure of the SEI in NRR remain unclear.
Several researchers have performed real-time spectroscopic analyses to investigate the composition of the cathode surfaces during catalysis. Through the in situ XRD analysis of Au-coated carbon fiber paper, Gao et al. confirmed that Au facilitates Li-ion reduction ( Figure 8A). [120] They reported that the spontaneous initiation of nitrogen splitting to Li 3 N and protonation to NH 3 occurred immediately after the rate-determining Li-ion reduction. A predominant peak of Li (101) was observed at ≈37°-38°only for Au on carbon paper (Au/CP) electrode, indicating that Au assists in rapid Li + reduction kinetics. Only the Au/CP electrode exhibits a gradual increase in the Li 3 N (100) peak, whereas the pristine CP electrode presents no signal.
Although NRRs typically employ pure N 2 gas to maximize the N 2 content in the feed, practical applications may involve O 2 inputs from air or other sources. Li et al. discovered that adding small amounts of O 2 improves the efficiency and stability of the eNRR. [118] To analyze these effects of O 2 on the SEI layer, the changes in the phase and surface composition were monitored using XRD and X-ray photoelectron spectroscopy (XPS). The XRD data ( Figure 8B) revealed that as the O 2 level increased to 3 mol%, various oxidized products were derived from the working electrode, owing to parasitic side reactions and oxygen reduction. Interestingly, the 0.8 mol% O 2 atmosphere only showed a distinct peak in N 1s XPS spectra, corresponding to the Li 3 N layer, without forming any undesirable oxidized product. Through microkinetic modeling, the oxygen-induced changes in the SEI layer were found to delay Li + diffusion, while optimizing the availability of N 2 and H + in the system.
Blair et al. investigated the dynamic behavior of the SEI structure and its composition via in situ neutron reflectometry. [121] Scattering-length density profiles of the cathode-electrolyte interface during Li-NRR on Mo thin-film electrodes were obtained before and after chronopotentiometry (CP). Figure 8C shows that only the Mo-oxide surface layer is detected in the pre-CP electrode, whereas additional components from the interfacial layer are observed in the post-CP electrode. The secondary layer after CP was considered the decomposed electrolyte product (SEI layer).
Li et al. further attempted to reveal the effects of Li-containing electrolyte salts through theoretical modeling and depth-profiling XPS. Three different salts (LiBF 4 , LiPF 6 , and LiClO 4 ) were investigated using XPS spectra, based on the etching time ( Figure 8D). SEI layer formation was confirmed after a short reductive electrolysis in all three cases. For example, a LiF-enriched SEI layer was detected when using LiBF 4 . They reported that a LiBF 4 electrolyte induced the formation of a compact and uniform SEI layer, resulting in homogeneous Li plating and enhanced NRR efficiency. [124]

Electrode Engineering
Effective mass transport of N 2 to the electrode is crucial to facilitate the splitting of the dinitrogen bonds. However, owing to the low solubility of N 2 in nonaqueous solvents, the reactants are less accessible to the catalysts. To address the issue, Lazouski et al. designed an effective gas-liquid contacting gas diffusion electrode (GDE) in nonaqueous solvents. [76] Figure 9A depicts the typical aqueous media/GDE system, which allows direct gas contact with the catalysts on hydrophobic carbon supports. On the other hand, with the nonaqueous solvent, the carbon supports are inevitably wetted, and the catalyst is flooded, which results in limited N 2 transport and a low mass-transfer-limited current. Lazouski et al. developed a catalyst-deposited stainless-steel cloth, and a pressure gradient was maintained across the cloth. As shown in Figure 9A, N 2 was able to freely approach the catalysts without passing through the liquid phase, significantly improving NRR reaction rates and selectivity.
In the Li-NRR scheme, the working electrode provides active and stable sites for Li deposition. Tsuneto et al. tested several metals as working electrodes; Mo and Cu electrodes have typically been used in Li-mediated NH 3 synthesis because they do not form alloys with Li, [123] thus ensuring the reversible formation of the Li layer on the electrode. Figure 9B displays the SEM images of pristine Ni foam (left) and Cu electrodeposited on Ni foam Figure 8. Investigation of SEI structure via spectroscopic analysis. A) Color map of operando-XRD measurement during Li-mediated NRR at Au/CP electrode to monitor the dynamic evolution of lithium-containing intermediates. Reproduced with permission. [120] Copyright 2020, Wiley VCH. B) XRD diffractograms of postelectrolysis electrode (Mo foil) with variation of O 2 mol%. Reproduced with permission. [118] Copyright 2021, American Association for the Advancement of Science. C) Scattering-length density profiles of before (Pre-CP) and after (Post-CP) NRR electrolysis on Mo electrodes demonstrating the formation of SEI layers and corresponding schemes of Pre-CP and Post-CP model, respectively. Reproduced with permission. [121] Copyright 2022, American Chemical Society. D) Dept-profiling XPS spectra of the Cu electrodes depending on the electrolyte materials, which shows the formation of different SEI layers. Reproduced with permission. [124] Copyright 2022, Elsevier.
using the hydrogen bubble template (HBT) method (right). [122] Li et al. reported that a high-surface-area Cu electrode fabricated via the HBT method successively demonstrated a current density of −100 mA cm geo −2 at 20 bar N 2 . Owing to the porous Ni foam and the HBT method, a high ECSA of 66.5 ± 7.1 cm 2 was obtained with a geometric surface of 0.5 cm 2 . The HBT-Cu electrode exhibited a NH 3 yield rate (46.0 ± 6.8 nmol s −1 cm geo −2 ) superior to that of the pristine Cu electrode. Norskov & Chorkendorff group further attempted to use stainless steel mesh to prepare porous Cu via the HBT method. [124] Figure 9C shows cross-sectional SEM images of the porous Cu electrode on the stainless steel mesh. The three porous Cu samples prepared with different deposition times (15 s, 1 min, and 5 min) exhibited different specific capacitances and ECSA. The 5 min Cu sample exhibited the highest performance (15.4 mF cm geo −2 ; 308 cm 2 ), resulting in 95% ± 3% FE at a current density of −100 mA cm geo −2 under 20 bar N 2 . Additionally, stable ammonia production was achieved on the porous Cu/steel electrode with the use of LiBF 4 as the electrolyte condition.
In the typical Li-NRR scheme, the anodic reaction has not been thoroughly investigated compared to the cathodic reaction. Recently, several studies have been conducted to replace the anodic reaction with a hydrogen oxidation reaction (HOR) by introducing an external hydrogen feed. [125,126] In this scheme, the anodic reaction can contribute to the efficiency of the Li-NRR by providing protons to the catholyte. To determine the catalytic properties of the HOR, Hodgetts et al. investigated the poisoning on the catalyst surface by varying the water content. [125] The authors hypothesized that poisoning of the surface from the adsorption of the electrolyte species and organic solvent oxidation products can be overcome by choosing proper electrode materials and water contents. Figure 9D   , and nonaqueous/steel cloth. Proposed overall eNRR schemes with NRR at cathode and HOR on anode. Reproduced with permission. [76] Copyright 2020, Springer Nature. B) SEM images of pristine Ni foam and HBT-Cu/Ni foam. Reproduced with permission. [122] Copyright 2021, American Chemical Society. C) Cross-sectional SEM images of porous Cu electrode and its chronopotentiometry profile with LiBF 4 electrolyte. Reproduced with permission. [124] Copyright 2022, Elsevier. D) Cyclic voltammetry curves of various metal anode materials in H 2 saturated 0.1 m LiNTf 2 /THF solution with varying humidity conditions, Pd/C [H 2 O] = 9 mm (semi dry) and 0.6 mm (dry), Ru/C [H 2 O] = 5 mm (semidry) and 0.4 mm (dry), respectively. Reproduced with permission. [125] Copyright 2022, American Chemical Society.
were suggested with optimal catalytic activity and water tolerance under HOR.

Electrolyte Engineering
The recent efforts for engineering SEI compositions and electrode materials have been discussed thus far. Other approaches have been also suggested for enhancing the Li-NRR performance. [127] Manthiram et al. examined diverse proton donors to reveal the effect of the proton donor structure on the transport kinetics in Li-NRR. [77] Figure 10A shows that slight changes in the proton donor structure significantly impact the selectivity toward NH 3 . The electrolytes that form similar SEI species to that derived from THF can result in higher FEs owing to the enhanced SEI permeability, which directly affects the diffusion rates of N 2 and the proton donor to the electrode surface. Here, n-butanol exhibits the highest NH 3 FE and forms an SEI similar to that obtained from THF.
Apart from adjusting the proton donors, the Li-containing electrolyte, another key component in Li-NRR, is being investigated.
Recently, Du et al. reported almost 100% FE with 2 m LiNTF 2 . [128] According to the plot of FE and NH 3 yield rates against the conductivity of various Li compounds ( Figure 10B), LiNTF 2 has the highest NH 3 yield rate and FE. The dense ionic assembly formed at the electrode-electrolyte interface suppresses electrolyte decomposition and provides high Li + transference numbers.
In INRR scheme, the "proton shuttle" can play an important role to deliver the protons generated from the anode to the lithium cathode for the hydrogenation of lithium nitride. The proton shuttle can reversibly and reproducibly operate the catalytic cycle without participating in the side reaction during catalysis. In this vein, Suryanto et al. suggested reproducible phosphonium cation trihexyltetradecylphosphonium ([P 6,6,6,14 ] + ) as a proton shuttle. [129] The authors demonstrated the reversibility of the salts with 31 P NMR ( Figure 10C) analysis. When the [P 6,6,6,14 ][eFAP] (([P 6,6,6,14 ]phosphonium tris(pentafluoroethyl)trifluorophosphate) solution was treated with excess Li 3 N, [P 6,6,6,14 ] + cation peak at 32.9 ppm was diminished and a new peak at 9.3 ppm emerged, corresponding to the deprotonated ylide form. Recovery of the proton shuttle was achieved by adding a proton source. Reproduced with permission. [77] Copyright 2022, American Chemical Society. B) NH 3 yield rate and Faradaic efficiency versus conductivity of various electrolyte species. Reproduced with permission. [128] Copyright 2022, Springer Nature. C) 31 P NMR spectra of phosphonium proton carrier during deprotonation and reprotonation cycle. [P 6,6,6,14 ] + cation peak at 32.9 ppm diminished after the reaction with Li 3 N and a new peak at 9.3 ppm emerged, which was assigned to the deprotonated ylide form. The acetic acid treatment restored the peak at 32.9 ppm. Reproduced with permission. [129] Copyright 2021, American Association for the Advancement of Science.

Summary and Outlook
In this review, we have highlighted the recent progress in electrochemical nitrogen reduction catalysis for NH 3 production. We defined two approaches in terms of the reaction mechanisms: DNRR and INRR. DNRR follows a conventional electrocatalysis scheme and thus the development of an effective catalyst governs the overall performance of the system. We also discussed the currently adopted mechanistic pathways of nitrogen reduction and introduced several strategies for developing efficient catalysts. The DNRR requires considerable overpotentials from −0.7 to −0.2 V based on the electrolyte conditions. Among the reported catalysts, Mo-based chalcogenide catalysts showed good DNRR performances, as shown in Figure 11A; and Table S1 (Supporting Information). Although Bi-based catalysts showed a decent performance, [130] the FE and NH 3 yields of most electrocatalysts are still low for practical ammonia production due to the competing HER and the low solubility of N 2 .
On the other hand, the performance of the Li-mediated INRR is superior in terms of FE and partial current. Notably, the FE approached 100% in recent studies (purple circle in Figure 11B; and Table S2, Supporting Information). [124] It demonstrated a higher possibility of realizing the electrochemical Haber-Bosch process. However, INRR also has several limitations to be addressed. The Li/Li + redox couple requires a high reductive potential, which results in energy inefficiency and undesirable side reactions such as solvent/electrolyte decomposition and SEI layer formation. [131] In addition, the use of the scarce and expensive Li metal may limit the applicability of this method. Therefore, considering the current achievement and remaining issues, we provide the key challenges to be addressed and future research directions, as follows.

Enhancing the Availability of N 2
Because DNRR is a surface-confined reaction, enhancing N 2 affinity via surface modifications such as atomic configuration (or functionalization) and maximization of electrode surface area would be a useful strategy ( Figure 11C). Hybrid structures, such as metal-organic frameworks composed of the catalyst and functional agent, may effectively capture N 2 on the catalyst surface and accelerate N 2 dissolution based on Le Chatelier's principle. [132] Also, since eNRR in the practical membrane electrode assembly (MEA) device is operated with humidified N 2 gas due to the activation of membrane electrolyte, fine-tuning the pore structures of the catalyst layer like in the case of the fuel cell would be an effective approach to enhancing the eNRR performances by expediting gas phase reactant supply and product removal in the MEA.

Efficiently Suppressing the HER
As discussed in Section 3, most reports introduced enhanced DNRR performance, but most faradaic currents predominantly came from the HER. [48,58] As protons are obtained from water in aqueous electrolytes, modifying the wettability of the catalyst surface can effectively control proton availability. Generally, in electrochemistry, hydrophilic carbon (on hydrophobic gas diffusion layers) is used as a supporting material. The thing is that the carbon electrode typically shows marginal HER performance at the eNRR potentials or promotes the HER by adsorbing water molecules on their surfaces. Therefore, tuning the surface wettability should be considered when designing DNRR catalysts. [133,134] Furthermore, DNRR in nonaqueous electrolytes should be considered for enhancing the DNRR currents by stoichiometrically matching the number of protons for DNRR and minimizing HER currents. www.advancedsciencenews.com www.advancedscience.com

Preventing Surface Poisoning from Hydrazine or NH 3
Similar to CO and CO 2 poisoning in electrochemical hydrogen and methanol oxidation reactions, surface poisoning from species such as CO, NH 3 , and N 2 H 2 would considerably affect the DNRR performance because surface passivation may induce catalyst deactivation and degradation. For example, a graphene layer coating on the catalyst surface may effectively minimize the poisoning by controlling the surface adsorption behaviors of these species (for example, CO poisoning during hydrogen oxidation). [135]

In Situ and Operando Studies to Gain Mechanistic Insights
To gain a deeper understanding of reaction mechanisms and to elucidate the dynamic behavior that occurs at the reaction interface, it is essential to use a combination of electrokinetic, in-situ, ex-situ spectroscopy, and computational studies. In recent years, in situ techniques such as differential electrochemical mass spectrometry (DEMS), surface-enhanced infrared absorption spectroscopy (SEIRAS), surface-enhanced Raman spectroscopy (SERS), and attenuated total reflection infrared spectroscopy (ATR-IR) have been utilized in the study of ammonia oxidation, the reverse reaction of nitrogen fixation. These techniques have enabled the identification of N-containing intermediate species formed on active sites. [136][137][138] By using a combination of DEMS and cyclic voltammetry, as well as DEMS and SEIRAS, it is possible to obtain information on adsorbed intermediates during the reaction. Combining these in situ instruments can provide valuable insights for the advance of catalyst designs.

Engineering the Architecture of the SEI Layer
Engineering the architecture of the SEI layer: As discussed previously, Li plating frequently induces the growth of an undesired SEI layer owing to solvent/electrolyte decomposition ( Figure 11D, left). The permeability of the SEI layer is a crucial factor in the Li-mediated NRR, because the diffusion of N 2 and proton donor molecules to the electrode can be significantly affected by the structure of the SEI layer. Modifying the electrolyte composition and concentration can influence the formation kinetics of the SEI layer; an optimized current density might prevent undesirable dendritic Li growth and result in a uniform layer. Constructing an artificial SEI layer from electrolyte additives might suppress the randomized growth of the Li and SEI layers. According to previous studies on Li batteries, adding a fluorinated polysulfonamide electrolyte or SnF 2 stabilizes the SEI composition and inhibits dendrite formation. Similar strategies may also be used for optimizing the SEI.

Overcoming the Inferior Energy Efficiency
Li + reduction to Li metal requires a high negative potential accompanied by solvent degradation and oxidation at the anode, which results in extremely high electrolyte resistance (typically 1000-1200 ohm); therefore, the energy efficiency is low (below 3%) for NH 3 production. Decreasing the solution resistance can possibly increase energy efficiency. Another promising strategy involves coupling the INRR with hydrogen oxidation or other sacrificial anodic reactions to protect the solvent molecules. Furthermore, modifying the electrolyte composition can decrease unwanted energy loss. Other factors that need to be addressed include FE or NH 3 loss during synthesis and limited cell lifetime ( Figure 11E).

Ideal Electrolyte Composition for eNRR
Most studies involving Li-NRR have typically used THF because of its broad negative potential window. Figure 11F shows several candidate solvents that could be stable at highly negative potentials. Ionic liquid-based electrolytes could be alternative candidates as stable N 2 -soluble solvents. [139] In Section 3.2.3, we discussed that the structures of the solvent and proton donor determine the diffusivity of the reacting species. Therefore, various electrolyte/solvent/HA components must be investigated to determine the optimal combination for an efficient NRR.

Effective Proton Delivery
For continuous ammonia production, effective proton transport to the lithium nitride layer is particularly crucial. Highperformance continuous synthesis for commercialization necessitates a consistent stream of proton sources during the reaction. So far, hydrogen oxidation or anodic oxidation of proton donors have served as proton sources. [76,77,125] Facile accessibility of proton to cathode has been tried by several groups [129,140] introducing proton shuttle concepts. Additionally, in order to facilitate proton transport, it is important to regulate the solvation of protons in nonaqueous media. In this vein, Kamlet-Taft solvent parameters can provide valuable guidelines to find out the optimal composition of the electrochemical system. [141] Polarizability ( *) and hydrogen bond acceptor ability ( ) of the solvents and proton donors would have a strong correlation for the transport of proton to SEI layers. Controlling SEI structure and proton transport will be the key to maximizing the overall efficiency of the Li-mediated NRR.
In conclusion, we believe that eNRR would provide numerous possibilities to expand the applications of heterogeneous electrocatalysts and also to achieve global carbon neutrality. At present, the eNRR is technically at the initial stage, and distinct merits and drawbacks have been suggested for the two eNRR routes. In particular, the Li-mediated INRR reaches ampere-scale currents of NH 3 production, which potentially can be comparable to the commercialized electrolysis system, in near future. We hope that, with this review, researchers in the fields of nanomaterials and electrochemistry gain significant insights into the development of eNRR-related materials and technologies.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.