非金属电催化剂环境条件下氮还原反应的研究进展

许桐，马奔原，梁杰，岳鲁超，刘倩，李廷帅，赵海涛，罗永岚，卢思宇，孙旭平,＊

1 电子科技大学，基础与前沿研究院，成都 610054

2 郑州大学，化学学院与绿色催化研究中心，郑州 450001

1 Introduction

NH3plays an important role in modern society as essential building block in manufacturing fertilizer, aqueous ammonia,plastic, explosive and dyeetc1,2. It also receives attention as green alternative fuel because of its carbon-free nature, large hydrogen capacity (17.6% (w)), high energy density (4.3 kW·h·kg−1) and easy transportation (−33 °C, liquid)3. Currently,industrial NH3production mainly relies on Haber-Bosch process at high pressure and temperature using N2and H2as feeding gas4.However, this process accounts for ～2% global energy consumption and ～1% CO2emission, which significantly stimulates the exploration of alternative approaches.Encouragingly, electrochemical N2reduction reaction (NRR)offers a promising route towards sustainable and environmentally-benign ammonia synthesis at ambient conditions, and its applied potential is compatible with intermittent solar, wind and other renewable energies5–8.However, it needs efficient electrocatalysts to drive N2-to-NH3conversion because of the extremely inert N≡N bond (945 kJ·mol−1bond energy).

To date, great efforts have been put into exploring highperformance catalysts towards high efficiency and selectivity.Generally, noble-metal catalysts perform efficiently for the NRR, but the scarcity and high cost limit their large-scale applications9–12. Therefore, much attention has focused on earthabundant transition-metal (TM) catalysts which can activate N2by an acceptance-donation process, using empty or unoccupied orbitals to accept the lone-pair electrons of N2meanwhile donating the abundantd-orbital electrons into the antibonding orbitals of N213–35. However, these catalysts may release metal ions, leading to environmental pollution36,37. Nevertheless, most of these TM electrocatalysts show moderate proton adsorption free energy and thed-orbital electrons of TM may also favor the formation of TM―H bonds, facilitating the hydrogen evolution reaction (HER) during the electrocatalytic reaction. Recent years, it is a hot pot to explore metal-free catalysts (MFCs)38,39.MFCs mainly include carbon-based catalysts (CBCs) and some boron-based and phosphorus-based catalysts. Usually, CBCs have porous structure and high surface area, which are favorable to expose more active sites and provide rich accessible channels for mass/electron transfer. Furthermore, heteroatoms doping(such as N, S, O, B, P and F), defect engineering, edge engineering and heterojunction can induce more intrinsic active sites or construct favorable catalytic environment by redistributing charge density and tailoring electronic structure38–41.More importantly, the graphitic carbon of CBCs can greatly enhance the electron conductivity which is necessary to supply enough reactive electrons to active site and antibonding orbital of N241,42. The unsaturated carbon atoms (spandsp2) may favor heteroatom doping by accepting foreign electrons in unoccupied orbitals43,44. Finally, CBCs are generally made up of earthabundant elements (C, B, N, O, S) with good durability in acidic environment, taking the advantages of low cost and stability38,45Moreover, the Lewis acid sites of most metal-free compounds could also accept the lone-pair electron of N2and adsorb N2molecules by forming nonmetal―N bond, further widening their potential in electrocatalytic NRR. However, compared with metal-based catalyst, the occupied orbitals of metal-free catalysts can only form covalent bond or conjugatedπbond,hindering the electron donation from electrocatalyst to N2and molecular activation.

This review summarizes recent advances in designing and developing MFCs towards NRR under ambient conditions in aqueous media, including compounds, heteroatom-doping, and organic polymers. We also discuss the strategies of boosting NRR performances and outlook the development perspectives of MFCs.

2 Fundamentals of electrocatalytic N2 reduction

The inertness of the N2molecule partly originates from the high bond energy (941 kJ·mol−1) of the triple bond. On account of the negative electron affinity, low proton affinity and nonpolarity render N2be Lewis base. The large energy gap between the highest occupied molecular orbital (HOMO) and the lowest unoccupied molecular orbital (LUMO) of N2is not benefit for electron transfer processes. There are two reactive mechanisms responsible for the N2-to-NH3conversion, including dissociative and associative pathways41,46. The dissociative pathway mainly concerns the Haber-Bosch process, by high-energy cleaving N≡N to obtain isolated N atoms which then react with H atoms to form NH3molecules. The associative pathway involves the biological and electrochemical N2fixation, for which hydrogenation process occurs but keeping the two N atoms of N2bonding with each other before the first NH3is generated.According to N2adsorption pattern and hydrogenation order, the associative mechanism can be further classified into distal,alternating and enzymatic pathways. In distal pathway,hydrogenation firstly occurs on the N atom away from catalyst surface to generate one NH3, and then another N atom of N2bonded on catalyst surface. As to alternating pathway,hydrogenation occurs on the two N atoms alternatively. The enzymatic pathway also has an alternatively hydrogenation process but simultaneously keeping the two N atoms bonded to catalyst surface before the first NH3desorption. For the associative mechanism, the catalyst is vitally important in favoring N2adsorption, activation, hydrogenation and NH3desorption.

3 Carbon-based catalysts

3.1 Heteroatom doping

3.1.1 N atom doping

Fig. 1 (a) Schematic of NPC preparation. (b) Contents of pyridinic,pyrrolic and graphitic N in NPCs. (c) NH3 yields of NPC-850 and NPC-1 at −0.9 V. Reproduced with permission from Ref. 47. Copyright 2018 American Chemical Society. (d) NRR reaction pathway on pyridinic N3 moiety site. Reproduced with permission from Ref. 48.Copyright 2018 Elsevier.

Recent studies show that N doping plays a significant role in the NRR performances of CBCs. Density functional theory(DFT) calculations confirm that pyridinic N and pyrrolic N are the key active sites for NRR, and graphitic N is unfavorable for N2adsorption. As shown in Fig. 1a, N-doped porous carbons(NPC) were synthesizedviathe pyrolysis of zeolite imidazolate framework (ZIF) at 750, 850, and 950 °C (denoted as NPC-750,NPC-850, and NPC-950, respectively)47. The obtained catalysts not only preserve ZIF’s features of high surface area and porous structure, but also have abundant active sites enriched by thermally induced defects and charge polarization. Among them,NPC-750 catalyst has the smallest surface area and the highest defect content. In 0.05 mol·L−1H2SO4, it achieves the highest FE of 1.42% and a NH3yield of 1.4 mmol·g−1·h−1at -0.9 V. In this regard, the highest NRR performances of NPC-750 mainly benefit from its intrinsic catalytic activity. XPS results show the highest N content of NPC-750, especially pyridinic (6.2% (x,atomic fraction)) and pyrrolic (5.3% (x)) N. For example as shown in Fig. 1b and 1c, NPC-1 sample of higher content of pyridinic N (2.8% (x)) than NPC-850 (2.1% (x)) has better NRR performance even though its total N content is slightly lower than NPC-850 (5.2% (x)vs. 5.5% (x)). To verify NH3not reduced from the N atoms of catalyst, NPC-750 catalyst is performed under Ar for NRR. The resulting electrolyte shows no NH3,confirming this hypothesis. Wu and co-workers48also reported N-doped nanoporous carbons (C-ZIF)viaheating ZIF-8 at 700–1100 °C for different hours. As temperature and time increasing,N doping content reduces with more disordered carbon structures induced. In 0.1 mol·L−1KOH, the sample thermally treated at 1100 °C for 1 h exhibits a maximum NH3yield of 3.4 ×10−6mol·cm−2·h−1with a Faradic efficiency (FE) of 10.2% at−0.3 V. DFT results reveal that the energetically favorable active sites are thermally induced pyridinic N3moieties embedded in graphitic layer. As shown in Fig. 1d, it consists of one vacancy linked with two pyridinic N and one H linked pyridinic N. The vacancy may result from the removal of one pyridinic N caused by high temperature thermal treatment. It suggests that the vacancy can effectively adsorb and reduce N2molecule with the assistance of pyridinic N. In Song's work49, the catalyst was synthesized from C2H2and NH3gas by a plasma method. The catalyst surface is composed of sharp spikes (～50–80 nm length,～1 nm width), which can concentrate the electric field at tips,promoting N2polarization and electrons injection into the antibonding orbitals of N2. In 0.25 mol·L−1LiClO4, the catalyst achieves a high NH3yield of 97.18 µg·h−1·cm−2with a FE of 11.56% at −1.19 V. In Wang's study50, N-doped nanoporous graphitic carbon membrane (NCM) catalyze the NRR in 0.1 mol·L−1HCl, attaining a NH3yield of 0.08 g·m−2·h−1and a FE of 5.2%. After functionalized by Au nanoparticles, the catalytic efficiency can be greatly improved by attaining the value of 0.36 g·m−2·h−1and 22%.

Fig. 2 (a) Schematic of the synthesis process of bamboo shoots-derived NC. (b) HRTEM image of NC-800. Reproduced with permission from Ref. 51. Copyright 2019 Multidisciplinary Digital Publishing Institute.(c) SEM image of NCF-900. (d) Schematic of the synthesis process of cicada slough-derived NCF. Reproduced with permission from Ref. 54.Copyright 2018 The Royal Society of Chemistry. (e) Schematic of the synthesis process of alfalfa-derived NPC. (f) SEM image of NPC-500.Reproduced with permission from Ref. 52. Copyright 2019 American Chemical Society.

Biomass-derived catalysts also have promising NRR application perspectives because of the extensively accessible,renewable and inexpensive features. As shown in Fig. 2a,b, Liet al.51synthesized N-doped porous carbon (NC) from bamboo shootsviadrying, grinding into powder, hydrothermal carbonization and 600–1000 °C pyrolysis. The obtained NC-800 catalyst (800 °C pyrolysis) shows a graphitic carbon structure with total N content of 3.05% (w, mass fraction). XPS spectra reveal the ratios of pyridinic, pyrrolic and graphitic N of 25.7%,6.7% and 67.6%, respectively. Among them, pyridinic N and pyrrolic N are considered as active sites to activate *N252. The large electronegativity of N (3.04) can effectively induce charge density redistribution and tailor electronic structure53.Meanwhile, the large specific surface area and the hierarchically micro-meso-macroporous structure are favorable for active site exposure and mass transport. In 0.1 mol·L−1HCl, the NC-800 catalyst can promote N2-to-NH3conversion, attaining a NH3yield of 16.3 µg·h−1·mg−1and a FE of 27.5% at −0.35 V. In Yang’s study54, N-doped hierarchical porous carbon foams(NCF, Fig. 2c,d) are obtained by thermal treatment the ballmilled mixture of cicada slough and ZnCl2, following by acid washing away the Zn2+ions. The ZnCl2plays a role of chemical activating agent and pore-fabricating agent, which can fabricate architecture with micropores (～1.92 nm) and mesopores (～25.3 nm). The NCF-900 catalyst (900 °C thermal treatment) presents a large Brunauer-Emmet-Teller (BET) surface area of 1547.13 m2·g−1. These features are favorable for electrolyte entering and shortened transport distance. Furthermore, the NCF-900 catalyst has high content of pyridinic N (1.96% (x)) and abundant defects as confirmed by XPS analysis and Raman spectra, respectively.In 0.1 mol·L−1HCl, it reaches a NH3yield of 15.7 µg·h−1·mg−1and a FE of 1.45% at −0.2 V. In Zhao’s study52, N-doped porous graphitic carbon (NPGC, Fig. 2e,f) was synthesized from alfalfa with CaCO3and K2C2O4. The NPGC-500 catalyst (500 °C heat treatment) has 6.35% (x) N content and 3D porous structure composed of micro-, meso- and macropores. The temperatureprogrammed desorption (TPD) measurements confirm the strongest N2adsorption capacity of NPGC-500, which is ascribed to the synergy effect between large surface area and high contents of pyridinic and pyrrolic N. In 0.005 mol·L−1H2SO4, it achieves a NH3yield of 1.31 mmol·h−1·g−1and a FE of 9.98% at −0.4 V. Furthermore, it has long-term durability though15N isotopic labeling experiments and XPS analysis confirm some pyridinic N escaping from catalyst surface. DFT calculations further reveal that the pyridinic N can be reduced into NH3during NRR, with the generated N vacancies as new active sites. It suggests that both N and C vacancies contribute to the remained NRR catalytic activity.

3.1.2 O atom doping

Oxygen has a larger electronegativity of 3.44 than that of carbon of 2.55. Oxygen doping leads to positively charged carbon site by forming polarized C―O bond. This Lewis acid site is beneficial to adsorb N2of weak Lewis base41. Our group developed oxygen-doped graphene (O―G, Fig. 3a)viathe pyrolysis of sodium gluconate with Na2CO3at 950 °C for 10 min under Ar gas flow, followed by acid washing to remove remnant Na2CO355. XPS (Fig. 3d) and FTIR results reveal that O doping into C nanosheets mainly adopt C―O, C＝O and O―C＝O functional groups. As shown in Fig. 3e, this catalyst achieves a NH3yield of 21.3 µg·h−1·mg−1at −0.55 V and a FE of 12.6% at−0.45 V in 0.1 mol·L−1HCl. The NRR performances show negligible decrease after 5 cycling tests and 22 h long-term electrolysis, indicating its strong electrochemical stability. As shown in Fig. 3f, three oxygen-doping models are designed to study the NRR performancesviaDFT method. In C―O model,oxygen atom replaces one carbon atom in graphene. Mulliken charge analysis shows that oxygen atom has a negative charge of−0.562 ǀeǀversusnearby carbon atom of 0.252 ǀeǀ. This electronic structure should not only provide dispersed active sites for NRR by stabilizing N2on positively charged carbon atom, but also suppress HER by prohibiting direct H+adsorption41. DFT results accord well with this assumption. The positively charged carbon atom can effectively adsorb N2, and NRR follows alternative pathway with a low free energy barrier of 0.77 eV from *NN to*NNH. In C＝O or O―C＝O model, C−OH or ―COOH functional group deviates from graphene plane to form a hole defect. The unsaturated carbon atoms around the hole defect adsorb N2using side-on configuration. DFT calculations suggest that both C＝O and O―C＝O groups lead to the decrease of free energy barrier of 0.34 eV.

Fig. 3 (a) TEM image of O-G. Reproduced with permission from Ref. 55. Copyright 2019 The Royal Society of Chemistry. (b) TEM image of O-CN. Reproduced with permission from Ref. 46. Copyright 2019 Wiley-VCH. (c) TEM image of O-KFCNTs. Reproduced with permission from Ref. 39. Copyright 2019 The Royal Society of Chemistry. (d) XPS spectra of O-G in C 1s and O 1s regions. (e) NH3 yields and FEs of O-G at different potentials. (f) Optimized C―O, C＝O and O―C＝O models and corresponding geometric structures of ＊NNH. Gray, red, blue and white balls represent the C, O, N and H atoms, respectively. Color online. Reproduced with permission from Ref. 55.Copyright 2019 The Royal Society of Chemistry.

Oxygen-doped porous carbon nanosheets were also synthesizedviaheating sodium citrate at 750 °C for 1 h under Ar gas flow, followed by acid washing to remove residue solid38.Beside of C―O, C＝O and O―C＝O functional groups, the catalyst has defect-rich amorphous phase and hierarchically porous structure with an average pore size of 13 nm. As a result,it exhibits a NH3yield of 18.03 µg·h−1·mg−1at −0.55 V and a FE of 12.6% at −0.45 V in 0.1 mol·L−1HCl. Furthermore, we developed oxygen-doped carbon materialsviadirectly carbonization of biomass materials under inert gas39,46. The oxygen-doped amorphous nanosheets derived from tannin (OCNs, Fig. 3b) shows NRR performances of a NH3yield of 20.15µg·h−1·mg−1and a FE of 4.97% at −0.6 V in 0.1 mol·L−1HCl.The catalyst derived from kapok fibers (O-KFCNTs, Fig. 3c) has homogeneous oxygen doping, graphitic carbon and hollow microtube morphology, attaining a NH3yield of 25.12µg·h−1·mg−1and a FE of 9.1%. In addition, the two catalysts have excellent selectivity without N2H4detection in electrolytes and good stability after control experiments.

3.1.3 S atom doping

Sulfur is important component element of nitrogenase for biological N2fixation and S doping introduces more defects56.Our group developed S-doped carbon nanosphere (S-CNS, Fig.4a) as a NRR catalystviaa hydrothermal reaction of glucose and benzyl disulfide followed by annealing in Ar atmosphere57. XPS results reveal that S doping into CNS mainly adopts ―C―S―C― and ―C―SOx(x= 2, 3, 4) functional groups. Based on Raman spectra shown in Fig. 4e and f, the calculatedID/IGvalue of 1.16 of S-CNS is larger than the 0.91 of CNS. BecauseIDrepresents the peak intensity of defects whileIGcorresponds to thesp2carbon atom vibrate in plane, it concludes that S doping introduces more defects which may induce abundant active sites to promote N2adsorption and reduction. Compared to pristine CNS, S-CNS offers enhanced N2adsorption capability (Fig. 4d).In 0.1 mol·L−1Na2SO4, S-CNS (19.07 µg·h−1·mg−1) shows higher NH3yield than CNS (3.76 µg·h−1·mg−1).

This strategy also works effectively for graphene. Our group suggests that S-doped graphene achieves a high NH3yield of 27.3 µg·h−1·mg−1and a FE of 11.5% in 0.1 mol·L−1HCl58. DFT results reveal that the carbon atoms near sulfur atoms are active sites in both S-replace-C model (Fig. 4g) and S-hole model (Fig.4h). The two models of active sites can effectively promote N2-to-NH3conversion, and NH3desorption occurs spontaneously nearly without free energy barrier, suggesting the easy release of active sites for next catalytic cycle. Following our studies, Wang developed S-doped three-dimensional graphene (Fig. 4c) as a metal-free NRR electrocatalyst, capable of attaining a large NH3yield of 38.81 µg·h−1·mg−1with a FE of 7.72% in 0.05 mol·L−1H2SO459. Of note, sulfur dots–graphene nanohybrid (Fig. 4b)also performs efficiently for ambient N2-to-NH3conversion with a NH3yield of 28.56 µg·h−1·mg−1and a FE of 7.07% in 0.5 mol·L−1LiClO460.

Fig. 4 (a) TEM image of S-CNS. Reproduced with permission from Ref. 57. Copyright 2018 Wiley-VCH. (b) TEM image of sulfur dots-graphene.Reproduced with permission from Ref. 60. Copyright 2019 The Royal Society of Chemistry. (c) TEM image of sulfur-doped 3D-graphene.Reproduced with permission from Ref. 59. Copyright 2020 The Royal Society of Chemistry. (d) N2-TPD curves of S-CNS and CNS. Raman spectra of (e) S-CNS and (f) CNS. Reproduced with permission from Ref. 57. Copyright 2018 Wiley-VCH. Free energy profiles for NRR on sulfurdoped graphene based on (g) model 1 and (h) model 2. Gray and yellow spheres denote C and S atoms, respectively. Related charge density differences of adsorption configurations are also shown. Yellow and blue denote the charge accumulation and depletion, respectively.Reproduced with permission from Ref. 58. Copyright 2019 The Royal Society of Chemistry. Color online.

3.1.4 B atom doping

In a pioneering work of Légaréet al., it was found that the boron atom of borylene molecular can fix N2effectively61and the dopant of boron introduces more defects. Previous literatures also suggested that B atom with electron-deficient nature prohibits the binding of Lewis acid H+under acidic conditions62–64. In Yu’s work64, B-doped graphene (BG) was synthesizedviathermal reduction of H3BO3and graphene oxide(GO) at 900 °C in H2/Ar gas. The catalysts with different B doping contents are obtained by controlling the mass ratio of H3BO3to graphene oxide, labeled as BG-1 (5 : 1, Fig. 5a), BG-2 (1 : 10), G (undoped) and BOG (5 : 1, with high oxygen content). XPS results reveal that B-substitution-C (BC3, Fig. 5b)is the main B doping form with BG-1 sample having the highest content, while defect form (B4C) and nanosheet edge defect form(BC2O, BCO2) take only small percentage. As shown in Fig. 5c,the calculatedID/IGvalues of BG-1, BOG, BG-2 and G are 1.3,1.3, 1.1 and 0.91, respectively. Because of the larger electronegativity of carbon (2.55) than boron (2.04), B-doping position has electron-deficient environment, which is beneficial to adsorb Lewis base N2and suppress HER. TPD curves confirm the gradually enhanced N2adsorption capability with increasing B doping content (Fig. 5d). The NRR performances are in line with B doping content, with BG-1 showing the highest NH3yield rate of 9.8 µg·h−1·cm−2and a FE of 10.8% at −0.5 V in 0.05 mol·L−1H2SO4(Fig. 5e,f).

Fig. 5 (a) TEM image of BG-1. Reproduced with permission from Ref. 64. Copyright 2018 Elsevier. (b) Schematic of the sp3 orbitals of BC3 for binding N2 (left). LUMO (blue) and HOMO (red) of undoped G (G) and BG (right). The position of a single doped boron atom was labeled (right).Reproduced with permission from Ref. 45. Copyright 2019 American Chemical Society. (c) Raman spectra, (d) N2-TPD curves, (e) NH3 yields and(f) FEs of BG-1, BOG, BG-2 and G. Reproduced with permission from Ref. 45. Copyright 2018 Elsevier.

B atom has three outermost electrons with an electronic configuration of 2s22p1. As B atom is anchored into carbon materials, one ofselectrons will be excited into aporbital meanwhile with thesandporbitals hybridized by the form ofsp,sp2orsp3. As shown in Fig. 5b, bothspandsp2have unoccupied orbitals and onepempty orbital, which can drive the acceptance-donation process with N2. Andpempty orbital accepts the lone-pair electrons of N2meanwhile unoccupied orbitals donates electrons into the antibonding orbital of N≡N.Liuet al.studied boron-anchored graphyne (GY) for NRRviaDFT method45. The B adsorption on GY configuration shows a poor NRR performance, because the adsorbed B atom has similar electronic structure to free B atom with twopempty orbitals to accept electrons from N2, possessing strong interaction with the N-related intermediates to desorption, e.g.,−6 eV adsorption energy of *NH2. In contrast, the B substitution acetylenic C configuration hassp-hybrid B atom, only with onepempty orbital freely accepting electrons from N2. Its NRR profile shows a free energy barrier only of 0.44 eV at the hydrogenation step from *N to *NH. In fact,sp2- andsp3-hybrid B atoms also have NRR performances, e.g., the above discussed B-doped graphene and the B-decorated graphitic-carbon nitride(B/g-C3N4)64,65. Furthermore, Sun’s group highlights that chemical environment around boron site also plays a key role in NRR process because the abundant electrons in substrate may increase the possibility of electron injection and following reduction63.

3.1.5 P atom doping

P has a smaller electronegativity of 2.19 than 2.55 of C,thereby P doping into carbon materials provides an effective strategy to boost NRR performances. P atoms can be successfully incorporated into graphene by P―O and P―C bonds. Our group synthesized P-doped graphene (PG, Fig. 6a)viaannealing of the mixture of GO and triphenyphosphine at 1000 °C in Ar atmosphere66. The calculatedID/IGvalue of 1.07 is higher than 1.02 of pristine graphene. By DFT calculations,the NRR catalytic activities of P―O and P―C binds are compared. The results show that P―O bind has not NRR activity while the P site of P―C bind can effectively promote N2-to-NH3conversion. Density of states (DOS, Fig. 6b) shows that P doping introduces new electron state around Fermi energy, which has a high activity to interact with surface molecule. More importantly, the free energy of *H on P active site is as high as 0.34 eV. The charge variation analysis shows that P atom acts as a role of electron transmitter between nearby three C atoms and adsorbed NxHy. As a result, the P-doped graphene can catalyze NRR along a preferable distal pathway with the first hydrogenation as the rate-limiting step (1.21 eV). As shown in Fig. 6c, it attains a NH3yield rate of 32.33 µg·h−1·mg−1and a FE of 20.82% at −0.65 V in 0.5 mol·L−1LiClO4.

3.1.6 F atom doping

Fig. 6 (a) TEM image of PG. (b) Density of states of P atom in PG and C atom in graphene. (c) NH3 yields and FEs of PG/CP. Reproduced with permission from Ref. 66. Copyright 2020 The Royal Society of Chemistry. (d) HRTEM image of d-FG. (e) Top view of d-FG. (f) NH3 yields and FEs of d-PG/CP with different fluorination time at −0.7 V. Reproduced with permission from Ref. 69. Copyright 2019 The Royal Society of Chemistry.

Previous literatures suggested that the strong incorporation of fluorine atoms into graphene might break somesp2carbon bonds and induce defects, thereby providing effective active sites67,68.F atoms can be successfully incorporated into graphene by C−F and C ＝C―F bonds. Our group synthesized defect-rich fluorographene nanosheet (d-FG, Fig. 6d)viaa hydrothermal reaction of graphene and HF at 180 °C69. The hydrothermal fluorination time can change F doping content which has a significant effect on the catalytic ability of d-FG (Fig. 6f).However, high F content is unfavorable because the large amount F may breaksp2conjugated system and reduce electrical conductivity. The d-FG with 3.12% (x) F doping content presents a defect-rich structure, as confirmed by the highID/IGof Raman spectrum and the pronounced signal (g= 1.918) of electron spin resonance (ESR) spectrum. DFT calculations reveal that the two carbon atoms near F-induced pore defects are active sites (Fig.6e), which adsorb and activate N2effectively by a side-on model with N≡N elongated from 0.112 to 0.126 nm. The first hydrogenation from *N2to *N2H even shows downhill free energy change. The preferable mixed pathway for NRR shows a maximal free energy barrier of 0.37 eV from *N*NH to*N*NH2. In 0.1 mol·L−1Na2SO4, it exhibits good NRR performance and stability, with a NH3yield of 9.3 µg·h−1·mg−1and a FE of 4.2% at −0.7 V. Liuet al.also developed a F-doped 3D porous carbon framework. Zirconium chloride and terephthalic acid were used to the synthesis of octahedral carbon template, with polytetrafluoroethylene as F source70. Because of the different electronegativity between F (3.98) and C (2.55), F bonding to C atom creates Lewis acid site to facilitate the repulsive interaction between Lewis acid site and H+,suppressing HER. In 0.05 mol·L−1H2SO4, it exhibits a NH3yield of 197.7 µg·h−1·mg−1at −0.2 V and a FE of 54.8% at −0.3 V.

3.1.7 Co-doping

For electrochemical NRR, N2molecule activation is usually difficult and the first hydrogenation step from *N2to *N2H generally shows uphill free energy (ΔGN2-N2H) as a rate-limiting step in overall pathway71,72. In Chen’s study73, carbon-based catalysts were developed, including graphene (G), N-doped graphene (NG), B-doped graphene (BG), graphene with separated B and N doping (BNG-S), graphene with B−N bonds(BNG-B), BCN with B―N bonds (pyrolyzed from the mixture of PEG-2000, boric acid and urea). It suggests that pristine G and mono-doping G are ineffective for N2molecule activation with positive ΔGN2-N2Hof 2.01, 1.99 and 2.04 eV for G, NG and BG catalysts, respectively. However, BNG-B can activate the N2molecule with ΔGN2-N2Hreduced to 1.72 eV. The incorporation of B―N bonds is the active trigger and the edge C atoms near B―N bonds are the active sites for NRR. The BNG-B and BCN catalysts with abundant B―N bonds can greatly promote the NRR performances with obvious increasement of NH3yields and FEs as shown in Fig. 7a. Moreover, BNG-B and BCN offer higher HER overpotentials versus pristine G and mono-doping G catalysts in Fig. 7b, indicating suppressed HER. DOS calculations confirm that the synergistic effect of B―N bonds raises the highest electron state of catalyst towards Fermi energy level, which triggers the beneficial electron exchange between*N2H and active sites. B―N bonds also play a key role in the NRR performances of B,N co-doped porous carbon nanofiber(B/N-CNF, Fig. 7c), which was fabricatedviaelectrospinning of the mixture of boric acid and polyacrylonitrile (PAN) followed by 250 °C air stabilization and 900 °C NH3annealing74. The obtained nanofiber has a diameter of ～100 nm, consisting of amorphous and graphitic carbons (Fig. 7d,e). This B/N-CNF catalyst possesses abundant defects and vacancies as confirmed by the high ratio ofID/IGof Raman spectrum. Compared with mono-doping CNT, the synergistic effect of B and N of B/NCNF catalyst favors robust NRR partial current density, small charge transfer resistance of electrochemical impedance spectroscopy (EIS), lower Tafel slope and higher electrochemical double layer capacitance (Cdl). It is believed that B atoms contribute to N2adsorption and charge transfer while N atoms enhance electrical conductivity. In 0.1 mol·L−1KOH, the B/N-CNF catalyst attains a NH3yield of 32.5 µg·h−1·mg−1at−0.7 V and a FE of 13.2% at −0.5 V.

Fig. 7 (a) NH3 yields, FEs and (b) HER performances of G, BG, NG, BNG-S, BNG-B and BCN. Reproduced with permission from Ref. 73.Copyright 2019 Wiley-VCH. (c) NH3 yields of B/N-CNF, B-CNF and N-CNF. (d) FESEM and (e) HRTEM images of B/N-CNF.Reproduced with permission from Ref. 74. Copyright 2019 The Royal Society of Chemistry.

The large electronegativity of N (3.04) greatly influences the electron density of adjacent C atoms and leads to beneficial Lewis acid sites for N2adsorption. N, P co-doping induces charge redistribution and creates active sites. N, P co-doped porous carbon (NPPC) also harvests NRR performances with a NH3yield of 0.97 µg·h−1·mg−1and a FE of 4.2% at −0.2 V in 0.1 mol·L−1HCl75. The NPPC catalyst was synthesizedviaannealing of polyaniline aerogels at 900 °C in Ar atmosphere.TEM results reveal a foam architecture structure with massive graphitic carbons existing in it, indicating beneficial high conductivity. The resulting NPPC catalyst promotes NRR to follow an associative pathway as confirmed byin situFourier transform infrared spectroscopy (FTIR). In Tian’s study, N,S codoped graphene (NSG) was synthesizedviamicrowave treatment (2450 MHZ, 800 W) of thiourea and GO76. It suggests that N, S co-doping is beneficial to the NRR of graphene by promoting N2adsorption. In 0.1 mol·L−1HCl, it achieves a NH3yield of 7.7 µg·h−1·mg−1and a FE of 5.8% at −0.6 V.

3.2 Organic polymers

Conductive polymers such as polyaniline (PAN), polypyrrole(PPY) and polythiophene (PTH) also attract attention in the applications of electrocatalysts77,78. In our study79, PAN was electrochemically deposited on carbon paper (PAN/CP). The existence of pyridinic and pyrrolic N play an important role in the electrocatalytic activity of N-doped carbon frameworks75,80.In 0.1 mol·L−1HCl, the PAN/CP catalyst harvests a NH3yield of 5.45 × 10−11mol·s−1·cm−2at −0.7 V and a FE of 3.76% at −0.6 V.In our another study81, perylene-3,4,9,10-tetracarboxylic acid(PTCA) nanorods, which have extendedπ–πelectronic interaction and high electron mobility82,83, were decorated on reduced graphene oxide (rGO) (PTCA-rGO, Fig. 8a).Differential charge density shows obvious charge transfer between PTCA and rGO. Compared with PTCA and rGO, the synergy role of PTCA and rGO greatly promotes N2activation and reduction with reduced free energy barriers. The N2-to-NH3conversion follows both distal and alternating pathways because of the negligible difference of maximum free energy barriers. As shown in Fig. 8b, it attains a high NH3yield of 24.7 µg·h−1·mg−1with FE of 6.9% at −0.5 V in 0.1 mol·L−1HCl. In Chen’s work84,Li+ions were incorporated into poly(N-ethyl-benzene-1,2,4,5-tetracarboxylic diimide) (PEBCD) by immersing in 0.5 mol·L−1Li2SO4solution. The formed O-Li+plays the role of active site,which greatly retards HER. (Fig. 8e). As shown in Fig. 8d, the PEBCD/C (deposited on carbon cloth, Fig. 8c) catalyst has high HER Tafel slopes of 968.5 and 356.3 mV·dec−1in 0.5 mol·L−1Li2SO4and 0.5 mol·L−1H2SO4, respectively. Benefited from the role of Li+incorporation, PEBCD/C can promote the N2-to-NH3conversion by an alternating pathway, attaining a NH3yield of 1.58 µg·h−1·cm−2and a FE of 2.85% at −0.5 V.

Fig. 8 (a) TEM image of PTCA-rGO. (b) NH3 yields of different electrodes. Reproduced with permission from Ref. 81. Copyright 2019 The Royal Society of Chemistry. (c) SEM image of PEBCD. (d) HER Tafel plots of PEBCD/C in 0.5 mol·L−1 Li2SO4 and 0.5 mol·L−1 H2SO4. (e) Schematic of Li+association with O sites of PEBCD. Reproduced with permission from Ref. 84. Copyright 2017 American Chemical Society.

3.3 Carbon nitride

Lvet al.42synthesized polymeric carbon nitride (PCN)viaannealing polycondensation melamine at 620 °C in Ar atmosphere. The PCN-NV4 sample (4 h annealing, Fig. 9a) has more nitrogen vacancies (NVs) and smaller transfer resistance as confirmed by electron paramagnetic resonance (EPR, Fig. 9b)spectra and Nyquist plots, respectively. DFT calculations reveal that N2molecule can be activated with N≡N elongated to 0.126 nm after adsorption, and the NRR follows a preferable alternating pathway. Because of its 2D nature, abundant NVs and high conductivity, PCN-NV4 promotes the N2-to-NH3conversion with a NH3yield of 8.09 µg·h−1·mg−1and a FE of 11.59% at −0.2 V in 0.1 mol·L−1HCl. Many theoretical works were also done on carbon nitrides for NRR, including C2N and C9N4monolayers85–87. In Ji’s work85, B-decorated C2N monolayer was designed as catalyst for NRR. DFT calculations reveal that boron atom can be adsorbed steadily on C2N monolayer by linking to the two N atoms of hole edge (Fig. 9c).The B atom with positive charge and magnetic moment plays a role of active site to activate N2molecule through the acceptance-donation mechanism. As shown in Fig. 9d,enzymatic model is the favorable NRR pathway with an energy barrier of 0.15 eV. In Zhang’s work87, B atom was steadily decorated on the bridge site of C9N4monolayer, by linking to the two N atoms of hole edge (Fig. 9e). DOS calculations indicate the metallic nature of B-C9N4monolayer, which is favorable for promoting charge transfer during NRR (Fig. 9f). The B active site can effectively activate N2molecule and catalyze NRR by following a preferable alternating pathway, with the first hydrogenation as the rate-limiting step.

3.4 Defect engineering

Fig. 9 (a) TEM image of PCN-NV4. (b) EPR spectra of PCN and PCN-NV4. Reproduced with permission from Ref. 42. Copyright 2018 Wiley-VCH.(c) B decorated C2N monolayer. (d) Free energy profiles for NRR on B decorated C2N monolayer along enzymatic pathway. Reproduced with permission from Ref. 85. Copyright 2019 The Royal Society of Chemistry. (e) B-C9N4 monolayer. (f) Charge density difference of B-C9N4 monolayer with adsorbed N2 via end-on pattern. Red and green denote the electrons accumulation and depletion, respectively. Reproduced with permission from Ref. 87. Copyright 2020 The Royal Society of Chemistry. Color online.

Hereinbefore, we summarized the N, O, S, B, P, F atom dopant and co-dopants. Pyridinic N and pyrrolic N can increase the electron-donating property of the catalyst, further indicating the tendency of the electron migration from the carbon atoms to the antibonding orbitals of adsorbed N2. The doping of oxygen atoms could transform hybridization state of the carbon-based materials fromsp2intosp3. And synergistic effect in co-dopants could induce an asymmetrical charge distribution on carbonbased materials, thereby enhancing the electron-donor property of the carbon atoms for weakening the N―N bond and facilitating the N2activation. Besides, defect engineering also is an effective strategy to boost the NRR performances of catalysts.In fact, defects break theπconjugation and play the role of active sites.

Fig. 10 (a) TEM images of pristine CC and CC-450. (b) NH3 yields and FEs of different electrodes. Reproduced with permission from Ref. 88.Copyright 2018 The Royal Society of Chemistry. (c) Schematic of the synthesis process of layered defective graphene. (d) Raman spectra of PG,DG-700, DG-800 and DG-900. (e) N2 adsorption energies of different catalysts with side-on configuration. (f) TEM image of DG-800. Reproduced with permission from Ref. 89. Copyright 2020 The Royal Society of Chemistry. TEM images of (g) CNT and (h) O-CNT. Reproduced with permission from Ref. 90. Copyright 2019 The Royal Society of Chemistry.

In Li’s study88, carbon cloth with rich defects was obtained by air annealing at 450 °C (CC-450). Raman spectra of CC-45 show a significantly improvedID/IGof 1.25versusthe values of 1.04,1.15 and 1.21 of pristine CC, CC-250 and CC-350, respectively.As observed in Fig. 10a. HRTEM images confirm the abundant defects of CC-450. Moreover, its hierarchical porous structure contributes to abundant edge defects and better N2adsorption capacity. EIS analysis also shows smaller charge transfer impedance. In NRR tests, CC-450 presents a more positive onset potential in N2-staturated electrolyte than pristine CC, CC-250 and CC-350, indicating its favorable NRR kinetics. As shown in Fig. 10b, it offers the best electrocatalytic activity for NRR, with a NH3yield of 2.59 × 10−10mol·cm−2·s−1and a FE of 6.92% at−0.3 V in 0.1 mol·L−1Na2SO4and 0.02 mol·L−1H2SO4.Graphene is an ideal substrate to load heteroatoms by breaking its integrity ofπconjugation. In Du's study89, bulk graphite oxide is annealed in molten salts of LiCl/KCl (mass ration = 45 :55) at 800 °C to exfoliate graphene by salt permeability and peeling off (DG-800, Fig. 10c). As shown in Fig. 10f, they obtain few-layer ultrathin nanosheets (～3 nm) with defective and porous structures (1–5 nm) which expose more edges and active sites. The abundant defects are identified by broaden XRD peaks, highID/IGof Raman spectra (Fig. 10d) and highsp3/sp2ratio of XPS spectra. Compared with pristine graphene (PG),DG-800 can effectively adsorb N2molecules and catalyze NRR because of point defects and edge defects. As shown in Fig. 10e,the calculated N2adsorption energy shows more negative value at edge defect. In 0.01 mol·L−1H2SO4, DG-800 obtains a NH3yield of 4.31 µg·h−1·mg−1and a FE of 8.51% at −0.4 V. Chemical oxidation is also an effective way to produce oxygen-containing groups and defects in carbon framework. In our previous work90, carbon nanotube (CNT, Fig. 10g) was simply kept in HNO3at 80 °C for 12–48 h to obtain oxidized CNT (O-CNT,Fig. 10i). The higherID/IGof O-CNT compared to CNT confirms its defective structure. XPS spectra identify the existence of C―OH, ―COOH, C＝O and C―O groups. DFT calculations reveal that C−O group is the active site which has the most negative N2-adsorption energy of −0.32 Vversus−0.06, −0.08 and −0.08 of C―OH, ―COOH, C＝O and C―O, respectively.When tested in 0.1 mol·L−1LiClO4, the O―CNT catalyst exhibits a high NH3yield of 32.33 µg·h−1·mg−1and a FE of 12.50%.

3.5 Surface modification engineering

For the defects engineering mentioned before, carbon materials suffer from high-temperature treatment or strong acid for materials preparation. In fact, there is a surface modification method which can be operated in mild conditions. In our previous work91, rGO was surface modified with oxygen-rich tannic acid (TA)viasimple ultrasonic mixture under ambient conditions. As shown in Fig. 11a,b, TA can greatly disperse rGO from bulk suspension to ultrafine emulsion. TEM images and Raman spectra show no obvious change of intrinsic properties.XPS results (Fig. 11d,e) reveal enhanced O＝C peak intensity which is attributed to the ―COOH group of TA attached to rGO. It is believed that theπ–πstacking interaction betweenπrich TA and rGO results in intimate contact of the oxygen groups of TA and rGO, which favors the effective manipulation of the electronic properties of rGO. In 0.5 mol·L−1LiClO4, TA-rGO has better NRR performances, with a NH3yield of 17.02 µg·h−1·mg−1and a FE of 4.83% at −0.75 V higher than the values of 3µg·h−1·mg−1and 1% of rGO. Furthermore, Fig. 11f,g show the coupled TA-rGO catalyst has good cycling stability and longterm durability.

Fig. 11 Photos of (a) rGO and (b) TA-rGO in water. (c) TEM image of TA-rGO. C 1s XPS spectra of (d) TA-rGO and (e) rGO. (f) NH3 yields and FEs of TA-rGO at −0.75 V in N2/Ar-saturated electrolyte. (g) Chronoamperometry curve of TA-rGO/CP at −0.75 V. Reproduced with permission from Ref. 91. Copyright 2019 American Chemical Society.

4 Boron-based catalysts

4.1 Boron atom doping

Boron has three valence electrons with an electronic configuration of 2s22p1, thereby it has electron-deficient nature with two emptyporbitals to accept foreign electrons. The boron monolayer, composed of triangular and hexagonal motifs, is energetically stable. There are many structures with different hexagon hole densities of 0, 1/4, 1/5, 1/6, 1/7, 1/8 and 1/992–94.Among them, α sheet (1/9) andβ12sheet (1/6) are the stable structures95,96. In Liu's work92, NRR onαsheet andβ12sheet were theoretically investigated. The boron sites with 5 or 6 coordination numbers forαsheet, 4 or 5 coordination numbers forβ12sheet, present catalytic activities for NRR performances.Bader charge analysis shows that N2adsorption is achieved through electron transfer from boron to N2(emptyπ* state) but requiring additional energy input, which is ascribed to the electron-deficient nature of boron. However, the metal substrate(Ag or Au) can facilitate electron injection into boron sheet and reduce the N2adsorption energy barrier. In our previous work97,2D boron nanosheets (BNS, Fig. 12a) were exfoliated from bulk boron powders in isopropyl alcohol using ultrasonic cell pulverizer. The obtained BNS have average thickness of ～3 nm and large BET surface area of 16.38 m2·g−1. Its 2D nature can favor the exposure of active sites. DOS calculations indicate the favorable conductor nature of BNS versus the semiconductor nature of bulk boron. This is in accord with the smaller charge transfer resistance of EIS. It achieves a NH3yield 13.22µg·h−1·mg−1and a FE of 4.04% at −0.8 V in 0.1 mol·L−1H2SO4(Fig. 12b). By DFT method, 1B, 2B and 3B are considered as N2adsorption sites (Fig. 12c). The calculated results show that 3B atom is the preferable N2active site where an obvious hybridization occurs between *N2and the 2porbitals of B atom.The *N2adsorption free energy at 3B site is identified to be−0.66 eV which is more negative than the *H adsorption free energy of 0.23 eV, which suggests excellent NRR selectivity.Moreover, 1B and 2B sites present strong *H adsorption with the free energies of −1.67 and −1.87 eV, indicating that 1B and 2B sites may be deactivated by H atoms. Interestingly, 3B site after the H deactivation of 1B and 2B sites (3B(H), Fig. 12d) shows better N2activation and reduction, with a more negative *N2adsorption free energy of −0.69 eV. The NRR follows distal pathway with the second NH3desorption as the rate-limiting step. Considering the detected O―H bond of BNS by XPS, the*OH adsorption is also investigated. The results show that the 3B site prefers to be saturated by *OH. However, after this *OH deactivation, the adjacent B atom (Fig. 12e) presents good NRR performances by following alternating pathway.

4.2 Boron-based compounds

Fig. 12 (a) TEM image of BNS. (b) NH3 yields and FEs of BNS/CP.Optimal structures and charge density differences of N2 adsorption on(c) 3B site, (d) 3B(H) site and (e) oxidized B (104) surface. Pink, blue,red and white balls represent the B, N, O and H atoms, respectively.Reproduced with permission from Ref. 97. Copyright 2019 American Chemical Society. Color online.

B4C has focused much attention as electrode materials or catalyst substrates for batteries and fuel cells due to its high mechanical strength, good conductivity and electrochemical stability98–101. In fact, it also has good catalytic activity for NRR.In our work6, B4C bulk was exfoliated in ethanol to form nanosheets using ultrasonic cell pulverize. When tested in 0.1 mol·L−1HCl at −0.75 V, B4C nanosheets offer good electrochemical stability and achieve a high FE of 15.95% and a NH3yield of 26.57 µg·h−1·mg−1. DFT calculations reveal that*N2adsorption and the first hydrogenation proceed spontaneously at B active site with released free energies,indicating a favorable N2activation. As shown in Fig. 13a, the hydrogenation step of *NH2−*NH2→ *NH2−*NH3is the ratelimiting step with a free energy barrier of 0.34 eV.

Fig. 13 (a) Free energy profiles of NRR on B4C (110) surface. Reproduced with permission from Ref. 6. Copyright 2018 Springer Nature.(b) Free energy profiles of NRR on the zigzag edge of h-BNNS. (c) Nyquist plots of h-BNNS/CP and bulk h-BN/CP. Reproduced with permission from Ref. 104. Copyright 2019 Springer. (d–g) Structural configurations of N2 bonding to end-on B@ZZBN, side-on B@ZZBN,end-on B@ACBN and side-on B@ACBN. Pink, blue, gray and white balls represent the B, N, C and H atoms, respectively.Reproduced with permission from Ref. 106. Copyright 2019 The Royal Society of Chemistry.

Hexagonal boron nitride (h-BN) has advantages of high mechanical strength, good thermal conductivity and chemical stability102,103. In our work, h-BN nanosheets (h-BNNS) were exfoliated form bulk h-BN in ethanol using ultrasonic cell pulverize for an anticipation of NRR104. AFM high profiles confirm the layered structure of h-BNNS with thickness of ～1.3 nm. EIS shows a smaller charge transfer resistanceversusbulk h-BN (Fig. 13c). In 0.1 mol·L−1HCl, the h-BNNS catalyst catalyzes NRR by following an enzymatic pathway, attaining a FE of 4.7% and a NH3yield of 22.4 µg·h−1·mg−1at −0.75 V. The catalytic activity is ascribed to the unsaturated boron atoms at the edges. DFT calculations reveal that N2can be readily adsorbed to the two boron atoms at zigzag edge by forming a 5-membered ring. The unsaturated boron atoms effectively activate and reduce *N2, with the first hydrogenation proceeding spontaneously and the hydrogenation step of *NH*NH →*NH2*NH as the rate-limiting step (Fig. 13b). The catalytic mechanism and kinetics of boron-terminated zigzag (B-ZZ)edges were also confirmed by Leeet al.viatheoretical calculations105. It is believed that the boron atoms at B-ZZ edges act as the active sites for NRR because the chemical environment enables the flexible conversion between B (sp2) and B (sp3)during NRR. In Mao’s study106, BN edge was investigated for NRRviaDFT method. At the nitrogen-terminated zigzag edge of pristine BN, the calculated *N2adsorption energy is as high as −3.02 eV, which also means difficult NH3desorption (−1.45 eV). Hence, single B atom is decorated at BN zigzag or armchair edge to redistribute the charge density around the active site. The decorated B atom prefers to bond with the two edge N atoms to form a 4-membered ring for zigzag edge (B@ZZBN, Fig. 13d and e), while it bonds at the hollow site to generate 5-membered ring for armchair edge (B@ACBN, Fig. 13f,g). The decorated B atoms act as active sites, and end-on model is favorable for *N2adsorption for both the edge cases.Viacalculating the distal and alternating pathways for both B@ZZBN and B@ACBN, the NRR performances are compared. The results show the distal pathway at B@ACBN is the most favorable way for NRR, with an energy barrier of 0.29 eV from *NH to *NH2. Moreover, the*N2adsorption and NH3desorption free energies for B@ACBN are greatly re-optimized to −1.97 and 0.85 eV respectively.

5 Phosphorus-based catalysts

5.1 P atom doping

Orthorhombic black phosphorus has a layered structure composed of the monolayers with puckered honeycomb-like top view. Because of the week van der Waals force, black phosphorus can be easily exfoliated into monolayer or few-layer structure, which has extensive applications in electronics, fieldeffect transistor and gas senor107–110. In Zhang’s study111, black phosphorus was exfoliated in ice-bath sonicator to form fewlayer nanosheets (FL-BP NSs). AFM analysis confirms the layered morphology with a thickness of ～4.1 nm (5–7 layers).Benefited from the more exposed active sites, sufficient electrons of 2D nature and weak HER, the FL-BP NSs catalyst has good activity and selectivity for NRR. In 0.01 mol·L−1HCl,it achieves a FE of 5.07% at −0.6 V and a NH3yield of 33.37µg·h−1·mg−1at −0.7 V. DFT calculations reveal that the P atoms at zigzag and diff-zigzag edges are active sites, which catalyze NRR by following a preferable alternating pathway.

Fig. 14 (a) TEM image of BP nanoparticles. (b) NH3 yields and FEs of BP/CP at each potential. Densities of active sites on (c)elementary B and (d) BP. Reproduced with permission from Ref. 115.Copyright 2019 The Royal Society of Chemistry. (e) Free energy profiles of NRR on BP (111) surface. Green and purple balls represent the B and P atoms, respectively. Reproduced with permission from Ref. 113. Copyright 2019 The Royal Society of Chemistry. Color online.

5.2 Phosphorus compounds

Because of good stability, high mechanical hardness, modest band gap (2.0 eV) and high dielectric constant, cubic BP attracts much attention as catalyst112,113. In Chen’s study113, NRR performances on the (100), (110) and (111) surfaces of BP were investigatedviaDFT method. The (111) surface has the highest catalytic activity towards NRR by following an energetically enzymatic pathway. As shown in Fig. 14e, the hydrogenation from *NH2to *NH3is the rate-limiting step with an energy barrier only of 0.12 eV, suggesting the robust intrinsic catalytic activity of BP. Moreover, the adsorption energy of *N2H (−3.38 eV) is more negative than that of *H (−1.69 eV) and *H2O(−0.95 eV), indicating favorable N2activation and sluggish HER. The adsorbed *H2O prefers to react with N2molecule with the help of H+and electron to form *N2H (*H2O + N2+ H++e−→ *N2H + H2O) with negative free energy, excluding the H2O poisoning effect and suggesting the favorable role of ionic liquid which can dissolve more N2molecules113,114. Our recent work provided a strong experimental support for BP-enabled N2fixation (Fig. 14a,b)115. In 0.1 mol·L−1HCl, BP nanoparticle affords a high NH3yield of 26.42 µg·h−1·mg−1and a high FE of 12.7% at −0.6 V. It also has good cycling stability and long-term durability. As shown in Fig. 14c,d, BP is more favorable for the exposure of active sites than the elementary B catalyst.Furthermore, the P atoms can indirectly weaken the N≡N triple bond and strengthen the B―N bond by transferring sufficient electrons to B atoms and then to N atoms. Obviously, elemental phosphorus and phosphorus compounds show promising potential as eligible catalysts for NRR.

6 Conclusion and outlook

This review provides an overview of recent progress in designing and developing MFCs towards sustainable and environmentally-benign NH3synthesis at ambient conditions.As can be seen, MFCs show benign NRR electrocatalytic performances (Table 1). Although, extensive attempts and progress have been made in previous works towards MFCs electrocatalysts, compared with TM-based electrocatalysts,MFCs lack the combination of unoccupied and occupied valence orbitals, hindering the electron transfer from catalysts to N2and thus showing poor ability in nitrogen activation. To overcome this limitation, asymmetrical charge distribution on the surface active site of MFCs is effective for promoting the charge exchange between catalysts and N2. Additionally, deeper insights into catalysts design and significant improvement of NRR performances are necessary. In the future studies, the following aspects should be considered:

(I) Theoretical prediction. The possible active sites,adsorption energies, possible catalytic pathways are generally predicted. These predictions can greatly cut down experimental cost. A high-throughput screening is strongly recommended toquickly select out the effective catalysts. The free energy barriers at the hydrogenation steps of *N2→ *N2H and *NH2→ *NH3are set upper limitation to exclude the sluggish catalysts because these two steps usually have the most positive free energy changes among the whole NRR process for most catalysts.

Table 1 NRR performances of Metal-free electrocatalysts.

(II) Promotion of the intrinsic activities of catalysts. An ideal catalyst should enable robust N2adsorption, activation, smooth hydrogenation and NH3desorption. The catalyst kinetics can be modulated by redistributing charge density and tailor electronic structureviaintroducing heteroatoms, vacancies and edges unsaturated atoms,etc. For example, heteroatom of large electronegativity can induce the positive charge of adjacent atoms, which as Lewis acid sites are favorable to adsorb the N2of weak Lewis base, meanwhile these Lewis acid sites can effectively suppress HER by prohibiting binding with the Lewis acid H+41,64. In addition, catalysts after modification of heteroatoms or vacancies,etc. may generate spin magnetic mo ment or raise the highest electron state towards Fermi energy,which may trigger the beneficial electron exchange between *N2and active sites as well as the enhanced electron transfer73. The unsaturated coordination sites are also attractive to N2molecule because these sites mean unpaired electrons which are prone to pair with the foreign electrons.

(III) Exposing more active sites. Catalysts with hierarchical pore structure and high surface area can expose rich active sites and provide accessible paths for mass/electron transfer. In addition, metal-organic frameworks (MOFs) derived metal-free nanomaterials should be paid more attention as an effective way to boost NRR performances due to their good N2adsorption,adjustable morphologies and tunable coordination atoms.

(IV) Developments of data reproducibility and accuracy. The amount of ammonia produced is usually so small that it cannot be precisely attributed to NRR. Even though the contemporary methods of analysis are highly sensitive, quantification of the amounts of ammonia at levels that hardly exceed the level of the ubiquitous nitrogen-containing contaminants. Firstly, there are many possible sources of NOxcontaminants in the process of NRR experiments, such as NOxcompounds, NOximpurities and gaseous NOxcompounds that are typically present in the nitrogen gas stream. Secondly, electrocatalysts, electrolytes,solvents, ambient air and any other component of the experimental preparation that is nitrogen-based or was prepared using N-containing precursor can be a source of contamination.

Therefore, given the contaminations above, it is necessary to quantify the concentrations of NOxcompounds in the N2gas supply before NRR experiments. Among other options, alkaline solutions is a good choice for removal of NOx. In the presence of alkali, HNO2and HNO3, can be neutralized to NO2−and NO3−, impeding the reformation of gaseous NOx. Thus, to remove NOxcontaminations from the cell, electrodes and other labware, washing with alkaline solutions is an efficient method.Chemicals, electrolytes and catalytic materials contaminated with nitrogenous species can be purified through recrystallisation or annealing at an appropriate temperature,while subsequent storage should be under vacuum or clean Ar atmosphere.