ll

Review
Electrochemical ammonia synthesis: Mechanistic
understanding and catalyst design
Huidong Shen,1 Changhyeok Choi,2 Justus Masa,3 Xin Li,1 Jieshan Qiu,1 Yousung Jung,2,*
and Zhenyu Sun1,*

SUMMARY The bigger picture

NH3 production is dependent on the century-old Haber-Bosch pro- Ammonia, the second largest
cess, which is energy and capital intensive and relies on H2 from synthetic chemical
steam reforming, hence, contributing to greenhouse gas emissions. commercialized worldwide, is
Electrochemical NH3 synthesis can be realized by reaction of N2 and widely used as a fertilizer and is a
a proton source under mild conditions powered by renewable elec- key intermediate for production of
tricity, which offers a promising carbon-neutral and sustainable all nitrogen-atom-containing
strategy. However, N2 has remarkable thermodynamic stability chemicals. It could also be
and requires high energy to be activated. Implementation of this employed for fueling applications.
‘‘clean’’ NH3 synthesis route therefore still requires significant Electrochemical N2 reduction
enhancement in energy efficiency, conversion rate, and durability, reaction (NRR) offers a renewable
which is only achievable through the design of efficient electrocata- and distributed route for NH3
lysts. This article provides a timely theoretical and experimental production. Heightened research
overview of recent advances in the electrocatalytic conversion of efforts have focused on the design
N2 to NH3 underlining the development of novel electrocatalysts. and development of advanced
Advances of in situ and operando studies for mechanistic under- electrocatalysts to enhance the
standing of the reaction and the main challenges and strategies efficiency of NRR to make it
for improving electrocatalytic N2 reduction are highlighted. competitive against the Haber-
Bosch process from the economic
and ecological viewpoints. We
INTRODUCTION
describe the latest advances in the
Ammonia plays a key role in sustaining life and the global chemical economy with an NRR from both theoretical and
annual production exceeding 200 million tons.1 The bulk of industrial NH3 is primar- experimental aspects and provide
ily used to make fertilizers in agriculture (~80%) and to produce explosives, pharma- a guide on how electrocatalysis of
ceuticals, refrigerants, and cleaning products (~20%). NH3 is also being reckoned to NRR could be improved. We
be a potential fuel as well as ideal hydrogen carrier with a high gravimetric hydrogen discuss the roles of emerging in
content (~17.6 wt %) and large volumetric hydrogen energy density (10.7 kg H2/100 situ and operando methods in
L), in addition to advantages of easy liquefaction for handling, storage, and transpor- elucidating the dynamic catalyst
tation. NH3 fuel produces zero CO2 and low overall emissions. N2 is regenerated at structure and other reaction
the point of use and released into the atmosphere in a closed-cycle process. parameters. The possible reaction
Currently, about 90% of the NH3 produced worldwide still relies on the century- pathways and the major
old, fossil-fuel-powered Haber-Bosch process, which entails thermocatalytic conver- challenges in improving the NRR
sion of N2 and H2 (N2 + 3H2 / 2NH3 with a standard enthalpy of formation DHf0 = are also highlighted.
45.9 kJ mol 1 and standard Gibbs free energy DGf0 = 16.48 kJ mol 1) at high
temperature (>300 C) and intense pressure (>15 MPa), over Fe- or Ru-based cata-
lysts (utilized in the Kellog, Brown, and Root [KBR] advanced ammonia process
[KAAP]) with promoters (such as Al2O3 and K). Recently, a nickel-loaded LaN catalyst
was reported to be capable of accelerating the dissociation of N2 (the kinetically
determining step), comparable with ruthenium-based catalysts.2 The Haber-Bosch
process is among the top largest industrial chemical processes. The reaction has
been claimed to be one of the greatest inventions of the 20th century and has
been the subject of three chemistry Nobel prizes. However, it is capital intensive,

1708 Chem 7, 1708–1754, July 8, 2021 ª 2021 Elsevier Inc.

 ll
Review

Figure 1. Schematic representation of nitrogen fixation processes, including biological,

industrial, photocatalytic, and electrocatalytic NRR

requiring large and centralized plant infrastructure, and energetically demanding,

requiring an energy input of ~485 kJ mol 1 (responsible for over 1% of the world’s
annual energy consumption). The process combines N2 from the air with pure H2
derived from endothermic steam-methane reforming (i.e., CH4 + H2O / CO +
3H2), consuming 3%–5% of global natural gas or other fossil resources (e.g., coal),
which emits huge quantities of the greenhouse gas CO2 into the atmosphere
(from water gas shift, i.e., CO + H2O / CO2 + H2, the global average is ~2.86
tons of CO2 released per ton of NH3 and 1.6 tons of CO2 released per ton of NH3
in the most efficient plants).3 The Haber-Bosch process also has a drawback of low
energy efficiency with an NH3 conversion of less than 15% per cycle (limited by ther-
modynamics). Demand for NH3 continues to increase to support the growing global
population. Hence, a high-efficiency, mild (avoiding unfavorable equilibrium issues),
sustainable, and eco-friendly alternative approach to manufacturing NH3 is of signif-
icant importance for both scientific research and industrial applications.

From these scenarios, three major clean routes for ammonia synthesis involving bio-
catalysis, photocatalysis, and electrocatalysis have sparked increasing research in-
terest in recent years, as illustrated in Figure 1. Biological nitrogen fixation in nature
is attained under mild conditions (<40 C, atmospheric pressure) by metalloenzyme
nitrogenases that are composed of FeMo, FeV, or FeFe cofactor as active sites with
FeMo being the most active and abundant enzyme for N2 reduction.4 A minimum of
16 moles of adenosine triphosphate (ATP) is necessary to reduce one mole of N2
(N2 + 8H+ + 8e + 16ATP / 2NH3 + H2 + 16ADP + 16PO43 where ADP is adenosine
diphosphate) with concomitant formation of one mole of H2 and a corresponding
1State Key Laboratory of Organic-Inorganic
transfer of 8(e /H+), not 6 (which could result in dissipative hydrolysis of 4 ATP).5 Composites, College of Chemical Engineering,
As a consequence, production of one NH3 consumes 8 ATP, requiring an energy Beijing University of Chemical Technology,
Beijing 100029, People’s Republic of China
input of 244 kJ mol 1. However, biological nitrogen fixation occurs only in a select
2Department of Chemical and Biomolecular
group of microorganisms, and the nitrogenases are susceptible to deactivation by
Engineering, Korea Advanced Institute of Science
oxygen.6 In addition, biological conversion of nitrogen into NH3 has a low space- and Technology (KAIST), Daejeon 34141,
time yield. Noteworthy, NH3 production via natural processes cannot meet the cur- Republic of Korea
3Max Planck Institute for Chemical Energy
rent and future NH3 needs. Photocatalytic NH3 synthesis only requires solar energy,
Conversion, Stiftstr. 34-36, 45470 Mülheim an der
water, and N2, encompassing two coupled redox half reactions, i.e., oxidization of Ruhr, Germany
water by photogenerated holes (3H2O [l] + 6h+ / 6H+ [aq.] + 3/2O2 [g]) in the *Correspondence: [email protected] (Y.J.),
valence band (VB) and reduction of N2 via photogenerated electrons (N2 [g] + 6H+ [email protected] (Z.S.)
[aq.] + 6 e / 2NH3 [g]) in the conduction band (CB). Nonetheless, the overall https://doi.org/10.1016/j.chempr.2021.01.009

Chem 7, 1708–1754, July 8, 2021 1709

 ll
Review

solar-to-chemical conversion efficiency is far from satisfactory owing to poor light

utilization, low density of active sites, and rapid recombination of photoexcited elec-
tron-hole pairs.

Electrochemical synthesis of NH3 was first demonstrated by Humphrey Davy in 1,807,7

while a relevant patent was provided in 1908. Reliable quantification of NH3 was
achieved late in 1922 by Fichter and Suter.8 Electrochemical synthesis of NH3 via N2
reduction is attractive because of (1) potentially higher energy efficiency than the
Haber-Bosch process,5 (2) environmental compatibility through coupling with car-
bon-free renewable energy resources (solar, tidal, and wind), (3) elimination of fossil
fuels as H2 sources whereby the required protons (H+) are generated in situ from water
oxidation, (4) flexible control of the reactions by adjusting external parameters (such as
electrochemical voltage), being conducive to modular and small-scale operation, and
(5) scalability and on-demand, on-site NH3 production. This direct N2 reduction pro-
cess is supposed to be less capital intensive than a combined water electrolysis with
Haber-Bosch process, which is only about 40% energy efficient.9 Electrochemical N2
reduction reaction (NRR) to generate NH3 was mostly undertaken at high temperatures
(above 500 C) using proton-conductive solid electrolytes before 2000.10 However, this
high-temperature route suffers from bottlenecks of low electronic and/or ionic conduc-
tivity in the electrolyte and NH3 decomposition, in addition to the parasitic hydrogen
evolution reaction (HER). To facilitate electrochemical NH3 synthesis at lower temper-
atures (100 C–500 C), molten electrolytes with higher ion conductivity such as
eutectic-based systems were attempted. Unfortunately, operations at such intermedi-
ate temperatures usually require large overpotentials, thus, decreasing energy effi-
ciency at high current densities. Poor durability in the long-term operation is another
issue of concern. The synthesis of NH3 below 100 C using aqueous electrolytes has
been the focus of interest since 2000, which was motivated by Nørskov and coworkers
to mimic the FeMo cofactor in the nitrogenase enzyme.11 This low-temperature syn-
thetic process significantly reduces equipment and operational costs and increases sta-
bility of NH3 produced for distributed deployment. However, binding and activation of
N2 under ambient conditions remains a grand challenge because the molecule is ther-
modynamically very stable and kinetically inert. Homogeneous (e.g., nitrogenase en-
zymes and molecular catalysts) and heterogeneous catalysts have been applied to
accelerate this up-hill reduction reaction. Although high turnover number is attained
in homogeneous catalysis, most homogeneous systems have high cost, toxicity,
poor stability, and involve complex post-separation steps, but they are less likely the
case in heterogeneous catalysis, thereby limiting their prospects for industrial applica-
tion. Therefore, major endeavors have been devoted to developing heterogeneous
electrocatalysts based on rational design approaches. However, despite recent ad-
vances achieved in this effort (Figure 2; Table S1), breakthrough progress is still
hampered with (1) low faradic efficiency (FE) (typically not more than 15% due to the
overwhelming HER catalyzed at similar or even lower overpotentials), (2) large overpo-
tential (or low energetic efficiency), (3) slow kinetics resulting in small exchange current
density, and (4) deactivation of electrodes in less than 100 h, restricting practical use
and technological commercialization. High current efficiency (the fraction of electric
charge that is used for the formation of NH3) is usually obtained at the expense of a
low NH3 production rate, compromising the overall economic viability of the process.
The majority of N2 reduction electrocatalysts reported thus far operate below 20 mA
cm 2, which is, however, far less than that required for commercial electrolyzers. Under
these circumstances, heightened research efforts have focused on the design and
development of advanced electrocatalysts, toward lower energy cost to compete
with the Haber-Bosch process and attainment of higher current density to minimize
capital costs.

1710 Chem 7, 1708–1754, July 8, 2021

 ll
Review

Figure 2. Summary of state-of-the-art various metallic and nonmetallic materials for aqueous
electrocatalytic NRR under ambient conditions
(A) NH 3 yield rate at different applied potentials.
(B) FE of NH 3 at different applied potentials.
Full-filled circles and half-filled circles represent the process using purified N2 and unpurified N 2 ,
respectively. Acidic, basic, and neutral electrolytes are labeled by (a), (b), and (n) following the
electrocatalysts, respectively.

Because of the high interest and rapid advances pertaining to this reaction, this
article aims to provide a comprehensive and up-to-date review of nanostructured
heterogeneous catalysts for the electrochemical NRR, with emphasis on relation-
ships between structures and properties. A summary of the most important develop-
ments in catalyst design, and analytic tools and techniques for electrocatalytic N2
reduction, as well as possible reaction mechanisms and pathways are presented.
Next, we discuss the strategies available to tune the electronic structure of electro-
catalysts and other relevant parameters in order to further enhance the performance
and efficiency of electrochemical N2 reduction. We also introduce emerging in situ
and operando methods for investigations of the dynamic catalyst structure and other
reaction parameters during the NRR. Finally, future challenges and opportunities of
NH3 synthesis by NRR are outlined.

FUNDAMENTALS OF ELECTROCHEMICAL N2 REDUCTION

Electrochemical reduction of N2 takes place at electrode-electrolyte interfaces and
requires protons transferred from the electrolyte, electrons transferred from the
electrode (catalyst), and optimized sites on the electrode surface onto which N2 mol-
ecules can be adsorbed and subsequently activated. Three major elementary steps
should be taken into account when modeling the fundamental processes: (1) diffu-
sion and chemical adsorption of N2 molecules and protons onto the cathode surface;
(2) activation of N2 and reductive addition of hydrogen atoms; and (3) rearrangement
and desorption of the product, NH3, or other products (e.g., hydrazine [N2H4] and
diazene [N2H2]) from the electrode surface and migration into the electrolyte.

The presence of oxygen in the cathodic compartment deteriorates NRR perfor-

mance, resulting in higher overpotentials. Moreover, aqueous electrocatalytic
NRR is plagued by concomitant production of molecular hydrogen, similar to the
cases in biological N2 fixation and electrochemical CO2 reduction.12 This leads to
low selectivity toward NH3 formation and loss of efficiency, especially because the
HER has intrinsically much faster kinetics compared with the NRR. For all metals,
the HER proceeds at a more positive potential than for NH3 synthesis.13 Moreover,
the rate of hydrogen production was modeled to be the first order in the electron
and proton concentrations, while the rate of NH3 production was zeroth order in
both.14 Hence, to boost the NRR, the availability and activity of protons at the

Chem 7, 1708–1754, July 8, 2021 1711

 ll
Review

electrochemical interface, properties of the electrode, reaction conditions, and na-

ture of the electrolyte should be manipulated to minimize the undesirable HER. Miti-
gating hydrogen evolution can be addressed by (1) selection of an electrolyte with
reduced proton donor activity,15 (2) addition of soluble coordination complexes to
facilitate N–H bond formation by mediating net H atom transfers,16 (3) optimization
of reaction conditions (pH, applied potential, and reactor configuration),17 (4) Li+ as-
sociation to decouple N2 fixation and NH3 evolution,18 (5) engineering of electrode
surface and electrode-electrolyte interface to regulate hydrophobicity (Figure 3A),19
and (6) tuning of electrocatalysts to favor adsorption and binding of nitrogen instead
of protons.20 Figure 3B illustrates that diminishing proton concentrations by using an
aprotic (or very alkaline) solvent offers an effective means to impede proton transfer
thermodynamically. Figure 3C shows an alternative route to kinetically inhibit proton
transfer by creating an aprotic and hydrophobic protection layers.

Another issue in NRR is the extremely low solubility of N2 in aqueous electrolytes

(water: ~0.00061 M at 25 C and P = 1 atm) because of its nonpolar nature, strong
triple bond, and low polarizability, dramatically limiting the amount of N2 available
for reaction. To enhance N2 dissolution, several strategies can be adopted: (1) em-
ploying low operating temperatures to reduce the Henry constant, (2) increasing N2
feed gas pressure, (3) using non-aqueous electrolytes (such as aprotic ionic liquids),
and (4) designing a hydrophobic mesoporous structure with high gas sorptivity that
can adsorb N2 but weaken interactions of water and the electrode surface. However,
lowering the reaction temperature markedly restricts N2 diffusion, and increasing
operating pressure would add technological complexity and cost. Conversely, a
flow rate of N2 into catholyte below 10 sccm affects reaction kinetics, meanwhile,
a higher N2 flow rate should in principle result in larger quantitative rates and FEs,
which, however, may level off above a flow rate of 20 sccm.21 At higher flows, no ef-
fect of the gas flow rate and the position of the gas inlet was observed on the NRR
results.22 Coupling electrolytes with a high N2 solubility and engineered electroca-
talysts with a high density of N2 adsorption sites appears to be a method of choice
for enhanced NRR performance. In contrast, the low solubility and slow transport of
N2 can be overcome by design and use of gas diffusion electrodes to allow intimate
contact between the gas, electrolyte, and catalyst.23

Parameters that may be useful in mechanistic analysis and benchmarking of electro-

catalysts for the NRR include the following: (1) NH3 production rate (mgNH3 h 1 cm 2
or mg h 1 mgcat. 1); (2) FE (FE = 3nF/Q, where 3 is the number of electrons trans-
ferred per NH3 molecule, n is the number of moles for NH3 produced, F is Faraday’s
constant (96,485 C mol 1), and Q is all the charge passed during the electrolysis pro-
cess); (3) overpotential (h, defined as the difference between the thermodynamic po-
tential of NRR and the applied potential required to achieve a desired current den-
sity); (4) NRR current density at a specific electrode potential of cell voltage; (5)
energy efficiency (EE% = DGm0/E = (1000 3 FE 3 339.2)/[3 3 F 3 (1.23 h)]
assuming an ideal nonpolarizable anodic oxygen evolution reaction with no overpo-
tential and kinetic limitation, where DGm0 represents the standard Gibbs free energy
of NH3 formation and E is the average mole energy input (kJ mol 1)); and (6) turnover
frequency (TOF, s 1), a measure of per-site activity of catalysts (io [A cm 2] 3 FE/
active site density (sites cm 2) 3 [1.602 3 10 19 (C/e 1) 3 6e 1/N2], where io refers
to exchange current density). For the sake of accurate comparison between different
materials, a combined figure of merit is preferred. In some cases, a single metrics
may fail to accurately represent the catalytic property. FE is commonly used as a
measure of the amount of charge that is effectively used for a specific faradic reac-
tion relative to the total charge that flows. However, only comparing FE is unlikely

1712 Chem 7, 1708–1754, July 8, 2021

 ll
Review

Figure 3. Strategies to improve selectivity of electrochemical NH3 synthesis

(A) Transforming water into NRR proton source by surface modification with oleylamine and
addition of butanol. Reprinted with permission from Koh et al. 19 Copyright 2020 American
Chemical Society.
(B) Limiting proton transfer rate by reducing the concentration of protons in the bulk solution.
(C) Limiting proton transfer rate by increasing the barrier for proton transfer to the surface.
Reprinted with permission from Singh et al. 14 Copyright 2017 American Chemical Society.

to give a complete picture of catalyst performance. Note that an improvement in FE

may be not necessarily accompanied with an increase in yield rate. The latter is
linked to product partial current density. A production rate normalized to electro-
chemical surface area (ECSA) provides information for intrinsic performance of a
catalyst. While normalization based on geometric surface area is fundamental
from the viewpoint of cost of a practical NRR cell. Additionally, high area-normalized
NH3 yield rate does not always mean large mass-normalized NH3 yield rate. Report-
ing of both mass- and area-normalized NH3 production is reasonable to compare N2

Chem 7, 1708–1754, July 8, 2021 1713

 ll
Review

Figure 4. Simplified molecular orbital of N2 and its binding modes

(A) Molecular orbital diagram.
(B) Proposed location of N 2 activation at the FeV cofactor (Top) and side-and top-views of the
binding modes prior to the formation of m-nitrides (i.e., linear Fe 2 ) on Fe(111) surfaces (Bottom). The
top, second, and third layers of Fe are depicted with orange, dark gray, and light gray spheres,
respectively. Reprinted with permission from Liu et al.4 Copyright 2020 American Chemical Society.

fixation among different catalysts. To attain an efficient N2 electrolyzer, it is indis-

pensable to maximize NH3 generation rate per unit of energy input, NH3 partial cur-
rent density at the highest energy efficiency as well as the energy efficiency at the
largest partial current density.

The overall cell voltage required for NRR involves potentials for both the anode and
cathode processes (Ecell = Eanode – Ecathode), where water oxidation (3 H2O # 6 H+ +
3/2 O2 + 6e–, E0[298 K] = +1.229 V [versus standard hydrogen electrode, SHE]) is
recognized as the default anode process. Coupling this to cathodic NRR generates
a minimum of several hundred millivolt overpotential for real-world electrocatalytic
NRR. The role of the anode in NRR should not be neglected because it consumes
almost half of the electrical input. Lowing oxygen evolution reaction overpotential
and improving the anode efficiency can reduce the total energy cost of electrochem-
ical NH3 synthesis, thus, promoting the prospects of its practical implementation.

Thermodynamics and kinetics of NRR

An N2 molecule comprises two nitrogen atoms bound by a disproportionately strong
homonuclear triple bond. Each atom possesses a pair of electrons in the 2s orbital
with opposite spin direction and three lone-pair electrons dispersed in the 2p orbitals
with the same spin direction. Hybridization of the s-p atomic orbitals leads to formation
of four bonding orbitals (two s and two p orbitals) and four antibonding orbitals (two s*
and two p* orbitals), with the shared electrons in the p and 2s orbitals forming an NhN
bond (Figure 4A).24 From a thermodynamic perspective, NRR is feasible with overall
negative Gibbs free energy. Nevertheless, activating N2 at ambient conditions is a formi-
dable challenge because of (1) large energy gap between the highest occupied molec-
ular orbital (HOMO) and lowest unoccupied molecular orbital (LUMO) of N2 (10.82 eV),
impeding electron transfer; (2) high enthalpy of the first H atom addition to form N2H+
(DH0 = +37.6 kJ mol 1) before breakage of the N–N bond; (3) extreme stability and inert-
ness of N2 with high cleavage energy (945 kJ mol 1) and first-bond breaking energy
(410 kJ mol 1), non-polarity (absence of permanent dipole), and large triplet state en-
ergy (6.17 eV); and (4) negative electron affinity ( 1.9 eV), low proton affinity (5.12
eV), and large ionization potential (15.85 eV).

Electrochemical NRR proceeds through coupled or sequential proton/electron transfer

processes, as summarized in Table 1. The former process seems to occur more

1714 Chem 7, 1708–1754, July 8, 2021

 ll
Review

Table 1. HER and electrochemical NRR processes with corresponding equilibrium potentials
Equation Reaction E0 (V)25,26
1 N2 (g) + 6H (aq.) + 6e # 2NH3 (g)
+
0.0577 (versus SHE)
2 N2 (g) + 8H+ (aq.) + 6e # 2NH4+ (aq.) +0.274 (versus SHE)
3 N2 (g) + 8HBase+ (MeCNa) + 6e # 2NH4+ +0.361 – 0.079 pKa (versus Fc+/0)
(MeCN) + 8Base
4 N2 (g) + 2H2O (l) + 6H+ (aq.) + 6e # +0.092 (versus SHE) or +0.23 (versus RHE)b
2NH3$H2O (aq.)
5 N2 (g) + 6HBase+ (MeCN) + 6e # 2NH3 +0.035 – 0.059 pKa (versus Fc+/0)
(MeCN) + 6Base
6 2H+ (aq.) + 2e # H2 (g) 0 (versus SHE)
7 N2 + 6H2O (l) + 6e # 2NH3 (g) + 6OH (aq.) 0.736 (versus RHE, pH 14)
8 2H2O (l) + 2e # H2 (g) + 2OH (aq.) 0.828 (versus normal hydrogen electrode,
NHE, pH 14)
9 N2 (g) + H+ (aq.) + e # N2H (g) 3.2 (versus RHE)
10 N2 (g) + 2H+ (aq.) + 2e # N2H2 (g) 1.10 (versus RHE)
11 N2 (g) + 2HBase+ (MeCN) + 2e # N2H2 1.22 – 0.059 pKa (versus Fc+/0)
(MeCN) + 2Base
12 N2 (g) + 4H+ (aq.) + 4e # N2H4 (g) 0.33 (versus RHE)
13 N2 (g) + 4HBase+ (MeCN) + 4e # N2H4 (g) + 0.398 – 0.059 pKa (versus Fc+/0)
4Base
14 N2 (g) + 4H2O (l) + 4e # N2H4 (g) + 4OH 1.16 (versus NHE, pH 14)
(aq.)
15 N2 (g) + 5H+ (aq.) + 4e # N2H5+(aq.) 0.23 (versus RHE)
16 N2 (g) + 5HBase+ (MeCN) + 4e # 0.153– 0.074 pKa (versus Fc+/0)
N2H5+(MeCN) + 5Base
17 N2 (g) + e # N2 (aq.) 4.16 (versus NHE) or 3.37 (versus RHE, pH
14)
a
Acetonitrile.
b
Thermodynamic equilibrium potential is calculated based on Nernst equation.

favorably than the latter one with lower energetic barriers, while the disruptive two-elec-
tron HER is also more severe at similar potentials in aqueous electrolytes (Equations 4
versus 6; Equations 7 versus 8). Multiple intermediates (such as N2H4 and N2H2) may be
involved during the concerted proton-electron transfer steps.25,26 The addition of the
first H atom to form N2H (Equation 9) demands a rather negative equilibrium potential
( 3.2 V versus reversible hydrogen electrode, RHE), while an even more negative po-
tential is needed for the first electron transfer to yield N2 (Equation 17). At a high
pH of 14, Equation 17 may compete with Equation 9 provided a weak affinity for
*N2H (* represents an active site on a catalyst) but a stabilizing interaction of the catalyst
with the N2 . The preferred pathway between concerted proton-electron transfer and
sequential proton-electron transfer was recently reported to be pH dependent. Howev-
er, the majority of theoretical calculations of NRR thus far did not consider the pH
impact. Hence, further exploration in this regard is necessary.

An advanced catalyst likely entails catalytic centers in favor of electron transport that
are in close proximity to sites providing protons. However, it is very unlikely for elec-
trochemical reduction of N2 to be kinetically limited by proton concentration in elec-
trolyte, which would occur if the mass transport of protons cannot match the NRR
rate that reaches the highest value for a specific system. In a typical electrochemical
process, the activity increases exponentially with overpotential. However, this is not
the case for the NRR, in which the performance first increases with overpotential in
the small overpotential regimes but declines at relatively larger overpotentials.
The drop of the NRR rate may be due to two possibilities: (1) the mass-diffusion lim-
itation and (2) fewer available *N2 molecules for further hydrogenation in light of the
fact that the electron and proton transfer steps of *N2 molecules are more rapid at

Chem 7, 1708–1754, July 8, 2021 1715

 ll
Review

higher reducing voltages.27 This is likely to be associated with the competitive HER,
in which higher surface coverage of *H occurs at larger overpotentials, while the N2
adsorption rate does not lean on potential. The surface coverage of *N2 decreases
owing to occupation of active sites by *H, thus, resulting in the decline of NRR
activity.

Adsorption of N2 and the linear scaling relationships

Adsorption and binding of N2 is essential for initiation of the NRR. In Fe7MS9C cofac-
tors (M = Mo, V, or Fe) of nitrogenase enzymes, a ‘‘belt’’ position that spans two low-
coordinate Fe ions (dFe–Fe = ca. 2.6 Å, Figure 4B) has been proposed for N2 adsorp-
tion.28 In industrial NH3 synthesis, it is supposed that specific multi-iron centers bind
and activate N2. Following initial physisorption through end-on N2 binding (g-N2,
Figure 4B), the adsorbate transforms into a bridging, side-on binding mode (a-N2,
Figure 4B) preceding N–N bond scission.4

A high-performing NRR electrocatalyst should have an ‘‘ideal’’ binding strength for

N according to the Sabatier principle. A too strong bonding implies high coverage
and surface poisoning, while a too weak adsorption entails very low coverage and
thus low reaction rate. Transition metals such as Ti, Y, Sc, and Zr were calculated
to possess a higher affinity toward N compared with hydrogen, benefiting NRR.29
Fe, Mo, and many other transition metals were predicted to chemically adsorb N2
forming metal-nitrogen bonds as a result of the interaction between the metals
and the N 2p states.30 Noble metals including Rh, Ru, Au, Pt, and Ir exhibit appro-
priate N binding strength, but they also have a strong ability to bind hydrogen. Pt
and Ir even display stronger binding of hydrogen relative to N2, detrimental to
NRR. Cu and Ag were calculated to have weak binding for N2, while Re showed an
exceptionally large adsorption energy for N2, hindering the reaction in both cases.31
Unlike pure metals, transition metal nitrides such as ZrN, NbN, chromium nitride
(CrN), and vanadium nitride (VN) were believed to trigger NRR by way of a Mars-
van Krevelen mechanism,32,33 in which a surface N atom is reduced to NH3 and
the catalyst later regenerated with dissolved N2, rather than adsorbing N2 to the
catalyst surface in the first step. Apart from the aforementioned N2 binding config-
urations, new active sites to adsorb N2, such as surface carbon34 and oxygen va-
cancies in metal oxides21,35 as well as electron-deficient atoms (as Lewis bases) in bo-
ron-doped graphene,36 have recently been identified and demonstrated to facilitate
NRR.

The multi-step six-electron NRR involves several different transition states. N2 bound
to the surface may be directly broken into adsorbed N atoms or first partly reduced
by generating an N2–H bond followed by further hydrogenation and dissociation.
The adsorption energy of the intermediates and activation barrier for N2 dissociation
were encoded by the Brønsted-Evans-Polanyi (BEP) linear correlation typically on the
surfaces of metals.13 Transition metals were suggested to be limited by the scaling
relations between the binding energy of *N2H generation and *NH protonation.
Transition metal nitrides are also limited by the weak binding for *N2H formation
but strong binding for *N protonation. This renders it difficult to design an active
catalyst that can cleave the N–N bond to yield the *N2H intermediate but not
bind NHx intermediates too strongly for reasonable catalytic turnovers. Indeed, no
metals were found to display a combination of intermediate nitrogen binding and
low energy N2 scission transition state, as depicted at the bottom middle of the
plot in Figure 5.37 On the other hand, selectivity for NRR is restricted by the scaling
between the free energy changes of the HER step and *N (in the nitride case) or *NH
(in the metal case) protonation or *N2H formation. Consequently, to attain efficient

1716 Chem 7, 1708–1754, July 8, 2021

 ll
Review

Figure 5. Linear scaling between nitrogen-binding energy and activation barrier for N2
dissociation (TOF: turnover frequency; Ea: activation energy)
Reprinted with permission from Medford et al. 37 Copyright 2015 Elsevier.

NRR, it is required to break the energetic linearity constrains between key interme-
diates (*H, *NHx, or N2Hx species where x = 0–2). To this end, new or coordinated
catalyst design schemes (through geometric and electronic effects as well as elec-
tronic dynamics) need to be developed. Exploration of promoters and solvents
beyond H2O may be another alternative route.38

Protocols for N2 electrolysis

Confirmation of product origin (from N2 or not)
Particular caution should be taken when performing NRR due to ambient ammonia
contamination.1,39 As early as 1984, minor amounts of nitrate or nitrite impurities in
alkali electrolytes were reported to undergo cathodic reduction very readily,
contributing predominantly to the reported yield of NH3.40 Therefore, stringent
experimental protocols should be adapted for NRR to avoid misinterpretation of re-
sults (false positives). This precaution was reiterated recently by Chorkendorff and
coworkers, who proposed a measurement protocol to alleviate the possibility of
misinterpretation of NRR data due to adventitious NH3 contamination, as shown
in Figure 6.1 Any labile nitrogen-containing contaminants that may stem from elec-
trocatalysts (especially those that involve nitrogen species originating from their
preparation), electrochemical reactor, chemicals (solvents, electrolytes), air, sup-
plied N2 gas, human breath, sample tubing, Nafion membranes, lab coats, nitrile
gloves, glassware, downstream traps, and stale Milli-Q water, among others must
be excluded.41 Virtually, high levels of NO3–, NO2–, and nitrides of up to 1,610 G
48 ppm were detected in some commercial metal oxides/metallic irons.42 To fully
verify a genuine NRR, at a minimum, a set of rigorous control and verification exper-
iments (for multiple repeats at each condition) need to be conducted in either Ar-
saturated electrolyte (under exactly the same conditions as the NRR experiments),
or under N2 without catalyst, or with the binder, or at an open circuit potential, or
N2-Ar alternate cycling, or by exposing the electrolyte in the air for extended pe-
riods.43 It is essential to quantify and confirm the levels of ionic and gaseous NOx

Chem 7, 1708–1754, July 8, 2021 1717

 ll
Review

Figure 6. Suggested experimental protocol for NRR

compounds before carrying out NRR measurements. Isotopic labeling using mole-
cules such as 15N2 in combination with isotope-sensitive proton nuclear magnetic
resonance spectroscopy (1H NMR) should ideally be carried out to confirm the actual
reduction of N2 (especially for catalysts that either contain nitrogen in their structures
or are prepared from nitrates or ammonium precursors), namely detection of solely
15
NH4+ (featuring a symmetric doublet at 7.12 ppm with a spacing of ~73.2 Hz as
opposed to a triplet for 14NH4+ at 7.12 ppm with a spacing of ~52.2 Hz). If appre-
ciable 14NH4+ is generated, nitrogen is most likely present in other extraneous sour-
ces. Under the same conditions, the amount of NH3 produced by the 15N2 test
should be consistent with the corresponding 14N2 test. Also, repeats of measure-
ments (over 3 times) should be performed for each catalyst type/composition to vali-
date that NH3 production is repeatable and reliable.

Purification of N2 feed gas, electrolyte salts, and electrocatalysts

Perhaps more importantly, the purged N2 gas (14N2 and 15N2) should be scrubbed to
remove any possible NH3 and nitrogen oxides (NOx) prior to starting NRR. The NOx

1718 Chem 7, 1708–1754, July 8, 2021

 ll
Review

species (NO, NO2, and N2O) are thermodynamically and kinetically more facile to be
reduced than N2.44 A 10 ppm of reducible N-impurities could potentially result in
about 4.0 mg per hour of NH3 or other nitrogen-containing compounds.39 This
dramatically affects precise quantification of NH3 from NRR in light of the fact that
typical NH3 yields are in the range of 10–1,000 nmol. It is worth noting that the level
of NH3 and NOx impurities in almost all 14N2 gas (with a purity %99.999%) reported
for NRR is uncertain or not specified. In addition, the isotopically labeled 15N2 feed-
stock (with a purity mostly %99.9%) can be contaminated with impurities (as high as
0.1%), mainly in the forms of 15N-ammonia, 15N-nitrite/nitrate, and 15N-nitrous ox-
ide.45 These point to the necessity of additional N2 purification. Removal of extra-
neous NH3 contamination can be readily addressed by passing the feed gas through
H2SO4 as adsorbent. The nitrite/nitrate and NO2 are soluble in water and therefore
can be effectively absorbed and eliminated from the stock gas by the aqueous solu-
tion. NO and other sparingly soluble NOx pollutants can be transformed into N2 and
H2O via selective catalytic reduction (SCR) of NOx with NH3 in a fixed-bed quartz
tube reactor over Cu-based catalysts or Ce0.1Ti0.9O246 at 300 C for 1 h, followed
by repeated absorption and removal of NH3 in H2SO4 solutions (confirmed by detec-
tion before NRR test). Alternatively, these N-species can be oxidized (to NO3 ) and
removed by use of a dual oxidant solution (H2O2/S2O82 ) at 50 C and pH of 11.47
During gas purification, special care should be taken to avoid new adventitious
contamination.

In light of the possible presence of impurities (such as NOx–, nitrides) in electrolyte

salts and electrode materials, they should be treated at high temperatures or
washed with KOH solution before use to remove the contaminants. The actual
amounts of NOx and nitrides should be determined and reported after the
treatments.

Cleaning of pretreated PEMs right before use

A proton exchange membrane (PEM) (such as Nafion 211 and Nafion 117) is utilized
in a typical electrochemical NRR cell allowing the crossover of protons and inhibiting
the diffusion of O2 from the anodic compartment to the cathodic compartment.
Before applying a Nafion membrane in the electrochemical cell, it is commonly pre-
treated using (3.0–5.0)% H2O2 aqueous solution at 80 C to eliminate organic impu-
rities and subsequently (0.1–1.0) M H2SO4 at 80 C to remove metallic species and to
protonate the membrane, which is further washed with ultrapure water. However,
some amounts of NH3 from the atmosphere especially in highly populated and
polluted regions may be adsorbed on the membrane, thus, interfering with the
NRR results. Hence, directly prior to use in NRR, the pretreated membrane should
be cleaned by ultrasonication (or extensive soaking) in diluted H2SO4 and ultrapure
water. This enables effective extraction of NH4+ from Nafion into the solution, mini-
mizing the effect of environmental ammonia contamination.

Quantification of ammonia
Six main types of techniques, including (1) spectrophotometry, (2) ion chromatog-
raphy, (3) ion-selective electrode (ISE), (4) fluorescence, (5) 1H NMR, and (6) ultra-
high-performance liquid chromatography mass spectrometry (UPLC-MS) have
been proposed to determine the amount of NH3 produced during NRR. Current
electrochemical NRR studies are heavily reliant on aqueous-based spectrophoto-
metric/colorimetric assays using indophenol blue48 and Nessler’s reagent,49 which
are well-established with advantages of good sensitivity (0–0.6 mgNH3-N L 1) and
low cost. The indophenol blue method, or the salicylate method, is based on the
modified Berthelot reaction, in which ammonia reacts with sodium salicylate and

Chem 7, 1708–1754, July 8, 2021 1719

 ll
Review

hypochlorite in an alkaline solution to produce indophenol blue that absorbs at

655 nm.50 The Nessler’s reagent method follows a reaction of ammonia with iodide
and mercury ions to generate a reddish-brown complex absorbing at 420 nm.
Although this method requires less reagents and shorter detection time than the sa-
licylate method, the Nessler’s reagent is highly toxic and has a lifetime of only
3 weeks. Additional pretreatments, for example by sodium potassium tartrate (KNa-
C4H4O6$4H2O) solution, are needed to eliminate metallic ions (Fe3+, Ni2+, Co2+,
Cr3+, and Ag+), sulfides (S2 ), and organics that could interfere with the Nessler’s re-
agent (a solution of K2HgI4 and KOH).

Ion chromatography is a well-known alternative method. It provides benefits of low

detection limit (3 3 10 7 mol L 1), good reproducibility, precision, and rapid detec-
tion of multiple ions at the same time, provided with suitable columns (such as Dio-
nex IonPac CS16-4mm Column) and eluents.41 However, Na+-involving electrolytes
are unsuitable attributed to overlapping of the Na+ and NH4+ peaks as well as its
short retention time.51 Additionally, challenges remain in isotopically differentiating
14
NH3 from 15NH3 because 15NH4+ and H3O+ share the same nominal m/z, which
cannot be addressed even by using advanced coupled ion chromatography mass
spectrometry (IC-MS).

Two types of ISEs have been demonstrated for measuring ammonia nitrogen.52 One is
ammonia-selective electrode via testing potential difference emanating from diffusion
of generated NH3 gas through a hydrophobic membrane (typically PTFE). The
ammonia/ammonium concentration in the solution can, thus, be determined according
to the Nernst equation. The other is ammonium-ion-sensing electrode with a polyvinyl-
chloride (PVC) membrane containing an ammonium carrier. All NH3 in samples is con-
verted to NH4+ by acidification for measurement. The ISE method offers advantageous
features of convenience, large detection range (0.03–1,400 mgNH3-N L 1) with high con-
centration of ammonia nitrogen, and suitability for continuous monitoring.52 However,
the inherent poor sensitivity of ISE especially for ammonia nitrogen at low concentra-
tions (< 0.5 mgNH3-N L 1) limits its widespread application.

The fluorometric method involves the reaction of ammonium with o-phthaldialde-

hyde (OPA) and sulfite featuring high sensitivity (detection limit of 1 nmol L–1) and
was first developed by Cohn and Lyle in 1966.53 Despite attempts to improve sensi-
tivity, this method suffers from interference by amino acids and amines, affecting
ammonium measurements.

1
H NMR has been widely used for quantification of NH4+ in biocatalytic and homo-
geneous catalytic N2 reduction.54 In contrast to indirect spectrophotometry and ion
chromatography protocols, the 1H NMR method is direct and highly selective for
analysis of both 14NH4+ and 15NH4+ (concentration in the range 5–10 mM with a
600-MHz NMR spectrometer and 1 mM with a 900-MHz NMR spectrometer).55 How-
ever, it suffers from laborious ammonia isolation via distillation. In addition, subop-
timal NMR settings lead to lengthy data acquisition time and influence accurate
quantification.

UPLC-MS was recently reported to distinguish and quantify 14NH3 (MS peak at m/z =
251.0854 for 14N) and 15NH3 (MS peak at m/z = 252.0825 for 15N).56 This analytical
method was developed based on a dansyl chloride derivatization process by mixing
dansyl chloride and ammonia solutions at optimal pH (~9.5 G 0.5). An ammonia con-
centration in the range of 0.6–43.5 mM was claimed to be rapidly measured for both
14
NH3 and 15NH3.

1720 Chem 7, 1708–1754, July 8, 2021

 ll
Review

Note that each of these methods has advantages and limits for assaying ammonia.
To evaluate the overall level of accuracy and accountability, it is strongly recommen-
ded to use a combination of different approaches. Moreover, it is imperative to
develop more selective, sensitive, accurate, and robust protocols for ammonia
quantification, as well as in situ and continuous processes for monitoring the NRR.

Quantification of by-products
From the Equations 10 and 12 (Table 1), N2H2 and N2H4 are also produced during
NRR as by-products, respectively. Such by-products may reduce the FE of ammonia;
however, these are important indicators for identifying reaction pathway of NRR. The
observation of N2H2 and N2H4 is an evidence for that the alternative pathway is
preferred to the distal pathway (Figure 8), and thus, the observation of by-products
has been employed for in situ mechanistic studies of NRR (discussed in detail in the
subsequent Mechanistic understanding of NRR section).57,58 Diazene is unstable
and can be easily decomposed into ammonia and nitrogen in the electrolyte.
Thus, diazene has not been quantified during the NRR while it has been detected
by in situ mechanistic studies on NRR.57,58 For example, in a spectroscopic method
to quantify the hydrazine in electrolytes, known as the method of Watt and Chrisp,59
a mixture of p-dimethylaminobenzaldehyde, ethanol, and concentrated HCl is used
as a color reagent. Calibration curve is obtained by using a mixture of color reagent
and N2H4 solution in dilute HCl (0.1 M) with different N2H4 concentrations. The
absorbance of the resulting solution is obtained at 458 nm. In this method, the op-
timum concentration range of hydrazine is 0.06 to 0.47 ppm, where the relative error
does not exceed 1%.

Electrolytes
Electrolytes serve as a hydrogen source and also provide a medium to trap N2 and
transfer protons and intermediates. Engineering of electrolytes can decrease the
ionic resistance in the electrochemical reactor, which is particularly critical for energy
efficiency at high currents. Akin to many other electrochemical reactions, the proton
activity, donor identity, the acid dissociation constant, and ion conductivity of elec-
trolytes affect activation barriers and reaction rates of NRR. However, further in-
depth experimental and theoretical investigations of electrolyte effects are
required. Electrolytes are generally classified into six types, including (1) liquid elec-
trolytes (aqueous solutions, organic solvents, and ionic liquids with operating tem-
perature %50 C),15,41,60 (2) molten salt electrolytes (molten 0.5 NaOH/0.5 KOH,
molten LiOH, eutectic of LiCl, KCl, and CsCl with Li3N operated between 200 C
and 500 C),60 (3) polymer membrane electrolytes (Nafion N117 and 102 mem-
branes, sulfonated polysulfone membrane operated from room temperature to
80 C),60 (4) composite membrane electrolytes (Na, K, and Li carbonates mixed
with LiAlO2 operated at 400 C),60 (5) O2 -conducting membrane electrolytes
(doped ZrO2, doped Bi2O3, doped CeO2, doped LaGaO3 at A and B sites, doped
BaSrO3 and SrCeO3, and Y2O3-ZrO2 operated at 650 C),60 and (6) ceramics/inor-
ganic proton conducting solid electrolytes (Yb doped Sr-cerate, SrCe0.95Yb0.05O3 d,
BaCe0.9Sm0.1O3 d + BaCe0.8Gd0.1Sm0.1O3 d complex perovskites, Gd-doped
barium cerate, Ce0.8M0.2O2 d [M = La, Y, Gd, Sm], Ba3(Ca1.18Nb1.82)O9 d,
Ba3CaZr0.5Nb1.5O9 d, Ba3Ca0.9Nd0.28Nb1.82O9 d, La1.9Ca0.1Zr2O6.95, Ca2+-doped
La2M2O7 [M = Ce, Zr], La1.95Ca0.05Zr2O7, La1.95Ca0.05Ce2O7, and SrZr0.95Y0.05O3 d
operated in the range 450 C–750 C).60 The majority of ambient electrocatalytic
NRR studies are conducted in liquid electrolytes. Recent progress in this regard
will be discussed in the following. For the other electrolytes, readers are referred
to earlier reviews.60

Chem 7, 1708–1754, July 8, 2021 1721

 ll
Review

Aqueous solutions (acidic, neutral, and alkaline solutions)

Most studies on electrochemical NRR at mild conditions have concentrated on
acidic, neutral, or basic N2-saturated aqueous electrolytes with OH , SO42 , or
Cl anions and alkali metal cations (e.g., Na+ and K+).61 The pH and counterions
of electrolytes impact the energetics of adsorbed intermediates and solvated
ions. Acidic electrolytes favor HER while alkaline electrolyte suppresses proton sup-
ply and transfer. Electrolysis in acidic media can cause accumulation of ammonium
salts in the electrolyte and alteration of the solution’s pH, rendering them poorly
amenable to successive and long-term operation. Lower electrolyte pH may lead
to a positive shift of equilibrium potential resulting from easier protonation of as-
formed NH3 into NH4+. However, decreasing pH also results in enhanced HER
through proton reduction. Manipulation of proton concentrations in electrolytes en-
ables one to increase surface coverage of N2 instead of H and accelerate the rate of
formation and consumption of early NRR intermediates,16 thereby promoting NRR
over HER. However, the impact of electrolytes on NRR is apparently not universal.
This may be associated with variations in cell configurations, inherent properties
of electrocatalysts, impurity effects in electrolytes, and local pH environments,
among others.41

Specifically, adsorbed alkaline cations in the inner Helmholtz plane (IHP) alter the po-
tential of the outer Helmholtz plane (OHP) and can also block hydrogen adsorption,
thereby limiting the HER.17,62 Meanwhile, cations were proposed to be solvated by
water in the OHP, favoring N2 adsorption. Unlike CO2 reduction, cations with a
smaller radius were found to favor better NRR activity, following the trend Li+ >
Na+ > K+ > Cs+.63 In addition, the 2p orbitals of adsorbed nitrogen were calculated
to shift to lower energies in the presence of K+, which facilitated electron transfer to
*NNH for N–N activation.17

Hitherto, the effects of anionic species on NRR have been rarely explored. Anions
with different buffer capacities (e.g., HCO3 and H2PO4 ) are likely to affect the local
pH at the electrode and hence the NRR selectivity. Also, PO43 was calculated to
easily adsorb N2 via bonding thus increasing N2 solubility in water.64 Adsorbed
phosphate and halides (i.e., Br , Cl , or I ) may form a bond with metal electrodes,
thus, promoting electron transfer from these anions to the vacant orbital of N2 and
boost N2 reduction. Additionally, specifically adsorbed halides may hinder proton
adsorption and induce a larger HER overvoltage.

Organic electrolytes
Most organic electrolytes exhibit an N2 solubility that is 1 or 2 orders of magnitude
higher than that of water. Use of organic solutions affords substantially enhanced N2
dissolution (toluene: 5 mmol L 1, trifluorotoluene: 10 mmol L 1, cyclohexane:
7 mmol L 1, 1H,1H,2H-heptafluorocyclopentane: 12 mmol L 1, heptane:
9.1 mmol L 1, perfluoroheptane: 17.9 mmol L 1, perfluorotributylamine:
11.8 mmol L 1, ethanol: 20.0 mmol L 1, 2-propanol: 23.0 mmol L 1)65,66 and also
regulates the supply of H+ to favor N2 adsorption onto electrode surface, benefiting
NRR. In this context, a number of organic-based electrolyte mixtures, such as 0.2 M
LiClO4 in ethanol/tetrahydrofuran,67 0.1 M LiClO4 in methanol/0.03 mol L 1
H2SO4,68 0.01 M H2SO4 in 2-propanol/water (9: 1, v/v),69 0.1 M LiCl in ethylenedi-
amine,70 0.2 M LiCF3SO3 in ethanol/dry tetrahydrofuran,71 and 1 M LiBF4 in
ethanol/dry tetrahydrofuran72 have been applied to enhance NRR over the
competing HER. In lithium (Li)-mediated N2 reduction, Li+ ions from solution in a po-
lar, aprotic solvent are reduced to Li metal, which easily splits the nitrogen triple
bond to form lithium nitride (Li3N). The Li3N then reacts with a proton source/carrier

1722 Chem 7, 1708–1754, July 8, 2021

 ll
Review

Figure 7. ILs and their interaction and solubilities for N2

(A) Structures and abbreviations of phosphonium-based ILs paired with highly fluorinated anions.
(B) N 2 solubilities (molar concentration, mmol L 1 ) of fluorinated ILs at 30  C and P = 1 atm. Data for
water and [C 2 mim][NTf 2 ] (at 25  C) are shown for comparison. Reprinted with permission from Kang
et al. 73 Copyright 2018 American Chemical Society.
(C) Structures of the [P6,6,6,14 ][eFAP] ions and their interaction with N2 . Reprinted with permission
from Zhou et al. 15 Copyright 2017 Royal Society of Chemistry.

(such as ethanol) to yield NH3. The proton carrier was claimed to play important roles
in promoting both Li nitridation to form Li3N and its further protonation to evolve
NH3.72 However, the stability and mechanistic understanding of organic electrolytes
need to be further elucidated.

Ionic liquids
Ionic liquids (ILs) with wide electrochemical stability windows are promising elec-
trolytes for efficient NRR.15 A family of phosphonium-based ILs with highly fluori-
nated anions were reported to have remarkable N2 solubility (Figures 7A and
7B).73 The already high N2 solubility of these ILs could be further enhanced by
the addition of fluorinated solvents, trifluorotoluene, 1H,1H,2H-heptafluorocyclo-
pentane, and 1H,1H,5H-octafluoropentyl 1,1,2,2-tetrafluoroethyl ether (FPEE)
that lack strong interactions (i.e., from hydrogen bonding, or functional groups).66
The ionic conductivity and fluidity were improved as well. Notably, the (1-butyl-1-
methylpyrrolidinium tris(pentafluoroethyl)trifluorophosphate ([C4mpyr][eFAP])-
FPEE binary electrolytes were observed to enable NRR with a high FE (32.0%)
and yield rate (2.35 3 10 11 mol cm 2 s 1) for NH3 formation over a-Fe nano-
rods.74 Despite this success, 15N2 isotopic labeling experiments need to be per-
formed to confirm the accuracy of the activity data. Manipulation of IL structures
by changing electric potential or nanopore confinement provide another way to
further improve the solubility of N2.75 Apart from high solubility for N2, the IL,
[C4mpyr][eFAP], was also found to stabilize the N2H intermediate via interaction
of the F–CF2 atom in [FAP] with the N2H group, thus, increasing both NRR activity
and selectivity (Figure 7C).76

The high price and viscosity (causing diffusional transport limitations) are downsides
of ILs for practical use in N2 electrolysis. Nevertheless, the high viscosity of ILs can be
circumvented by dilution with a controlled quantity of an aprotic solvent (or a mixture
of solvents), while retaining their functionality. Despite the high cost of ILs, some can
be recovered and reused, rendering electrolyzers that use ILs economic to a certain
extent. To date, only a limited number of ILs have been examined for NRR, further

Chem 7, 1708–1754, July 8, 2021 1723

 ll
Review

1724 Chem 7, 1708–1754, July 8, 2021

 ll
Review

Figure 8. Schematics of possible reaction mechanisms of ammonia synthesis

(A) Dissociative pathway. Reprinted with permission from van der Ham et al.26 Copyright 2014 Royal
Society of Chemistry.
(B) Distal pathway and alternative pathway.
(C) Enzymatic pathway. Reprinted with permission from Li et al.77 Copyright 2016 American
Chemical Society.
(D) Mars-van Krevelen mechanism on transition metal nitride. Gray, blue, and white balls indicate
transition metal, nitrogen, and hydrogen atoms, respectively. Reprinted with permission from
Abghoui et al. 33 Copyright 2015 Royal Society of Chemistry.

search and development of specially tailored N2-philic IL systems in combination

with advanced in situ mechanistic studies would be suitable subjects of focus in
the future.

Mechanistic understanding of NRR

Reaction mechanisms
Conventional mechanism of nitrogen reduction in industry. To date, most of
ammonia has been produced by the Haber-Bosch process, for which there exist
many extensive reviews. Here, we very briefly summarize the widely accepted mech-
anism of NH3 synthesis by the Haber-Bosch process for comparison purposes with
the electrochemical NRR, the focus of this review. In the conventional Haber-Bosch
process, N2 and H2 react on heterogeneous catalysts such as Fe and Ru. Ammonia
synthesis by the Haber-Bosch process has been generally believed to follow a disso-
ciative mechanism where the adsorbed N2 (*N2) dissociates into N adatoms (*N)
before undergoing hydrogenation (Figure 8A).26 The *N is then hydrogenated to
*NH3 via *NH and *NH2. This dissociative mechanism is generally considered for ni-
trogen reduction at high temperature (e.g., Haber-Bosch process) to overcome the
energy barrier for dissociation of the highly stable N–N triple bond (bond energy =
226 kcal/mol). Since electrochemical NRR is usually conducted at low temperature
(<100 C), sluggish dissociation of the N–N triple bond makes the dissociative mech-
anism highly unlikely in electrochemical NRR.

Associative mechanism. In nature, enzymes called nitrogenases can produce

ammonia from nitrogen, solvated protons, and electrons, via a so-called nitrogen fix-
ation process. Contrary to Haber-Bosch process, nitrogenases can facilitate the
reduction of N2 more efficiently even at ambient conditions. The active site of nitro-
genases is an FeMo cofactor consisting of an Mo/Fe/S cluster.78 The overall reaction
of nitrogen reduction by nitrogenases is:

N2 + 8H+ + 8e– / 2 NH3 + H2. (Equation 13)

In addition to nitrogenases, many homogeneous catalysts (to mimics of nitroge-

nases) have been studied for ammonia synthesis at ambient conditions.79 Three
different mechanisms have been proposed, namely, distal pathway, alternating
pathway, and enzymatic pathway, which are variations of the associative mechanistic
pathway (Figures 8B and 8C).26,77 Unlike the dissociative mechanism (Figure 8A), in
which the N–N triple bond is instantaneously ruptured, the adsorbed N2 molecule
(*N2) is protonated sequentially, breaking the triple bond consecutively, hence
requiring relatively less energy to drive the reaction. Since N2 can be adsorbed in
two different geometries called end-on (h1-N2) and side-on (h2-N2), both geometries
should be considered. In case of end-on adsorbed N2, both the distal pathway and
the alternating pathway are considered. In the distal pathway, the first NH3 is des-
orbed simultaneously with hydrogenation of *NNH2 (*NNH2 + (H+ + e–) / *N +
NH3), while the first NH3 is desorbed simultaneously with hydrogenation of ad-
sorbed hydrazine (*NH2NH2 + (H+ + e–) / *NH2 + NH3) in the alternating pathway.

Chem 7, 1708–1754, July 8, 2021 1725

 ll
Review

Figure 9. Possible NRR mechanism on Ru surface including various N–N bond dissociation steps
and reaction free energy obtained by DFT calculations
(A) Activation energy of N–N bond dissociation in all reaction intermediates of NRR on Ru surface.
Gray, blue, and ivory atoms indicate Ru, N, and H atoms, respectively.
(B) A summary of various possible reaction mechanisms of NRR on Ru surface. PDS and U L indicate
PDS and limiting potential, respectively.
Reprinted with permission from Back et al.80 Copyright 2016 Royal Society of Chemistry.

In the enzymatic pathway, each nitrogen atom in side-on adsorbed N2 is reduced

alternatively similar to the alternating pathway.

Reaction mechanism of NRR on heterogeneous catalysts. Similar to homogeneous

catalysts, heterogeneous catalysts generally follow the associative mechanism, and
thus, most theoretical studies in literatures have also considered the conventional
associative mechanism. However, other reaction pathways are also possible on het-
erogeneous catalysts. Contrary to single-metal sites (or small metal clusters) in ho-
mogeneous catalysts, metal surfaces can dissociate the N–N bond without the addi-
tion of a proton and electron (H+ + e–) pair.

A theoretical study on NRR over Ru surface showed that reaction pathways involving
various N–N bond dissociation steps yield the NRR activity that is comparable with
the associative mechanism with a similar limiting potential, suggesting the impor-
tance of considering all possible reaction pathways in addition to the conventional
associative mechanism (Figure 9).80 For example, *NNH2, *NHNH2, and *NH2NH2
can be dissociated into *N + *NH2, *NH + *NH2, and *NH2 + *NH2, respectively,

1726 Chem 7, 1708–1754, July 8, 2021

 ll
Review

with low activation energy (0.20~0.52 eV), while these reaction steps are not
involved in the conventional associative mechanism. Also, in situ studies of NRR
on Ru thin film using surface-enhanced infrared absorption spectroscopy (SEIRAS)
found N2H2 (diazene) as an intermediate product where the N2H2 can be decom-
posed or further protonated to ammonia. It can be noted that the latter reaction
steps have not been considered in conventional associative mechanism. Thus, it is
important to consider various potential pathways in order to describe correctly the
mechanism and activity of NRR on heterogeneous catalysts.

Now, we turn to the strategies for designing efficient NRR catalysts informed by in-
depth understanding of the reaction mechanism of NRR. The overall reaction and
equilibrium potential for NRR involving one N2 molecule and six proton-electron
pairs can be expressed below:

N2 + 6 (H+ + e–) / 2 NH3. (E0 = 0.05 V versus NHE) (Equation 1)

The equilibrium potentials of several reaction intermediates (N2H, N2H2, and N2H4)
in the associative pathway are known, and are presented in Table 1.

Although the formation of NH3 (g) from N2 (g) and H2 (g) is slightly exothermic
(DGf0 = 16.4 kJ/mol), NRR is thermodynamically hindered by the negative equilib-
rium potentials of these reaction intermediates. Notably, the equilibrium potential
of N2H (g) formation, corresponding to the first protonation step in NRR, is highly
negative, indicating that NRR via the associative mechanism is limited by the first
protonation step.26 Theoretical volcano plots of NRR also have shown that the first
protonation step (*N2H formation) is the thermodynamic limiting step for most cat-
alysts, and hence, catalysts that can strongly bind N2H favor NRR (Figures 10A and
10B).26,29,38 In addition to stabilizing adsorbed N2H (*N2H), destabilization of *NH2
is also important to design catalysts with low overpotential for NRR. A previous theo-
retical study has shown that the limiting step for NRR is the first protonation step
(N2 + [H+ + e–] / *N2H) on weak N-binding metal surfaces, meanwhile, for strong
N-binding metal surfaces, the limiting step is the final protonation step (*NH2 +
(H+ + e–) / NH3) (Figure 8).38 Thus, the strategies for lowering NRR overpotential
include stabilizing *N2H and destabilizing *NH2 independently to reduce the free
energy change for N2 + (H+ + e–) / *N2H and *NH2 + (H+ + e–) / NH3, respectively.

Recently, a surface-hydrogenation pathway is suggested for the NRR on noble-metal

catalysts (right-leg of the volcano), such as Pd and Au.81 The potential-determining
step (PDS) is *NNH formation on Pd and Au surfaces with 1~3 eV of free energy (Fig-
ure 10A), and the experimental NRR activity is observed at lower overpotentials (Ta-
ble S1). Using DFT calculations, Ling et al.81 showed that the activation energy of
*N2H2 formation via the surface-hydrogenation step (N2(g) + 2*H / *N2H2) is lower
than the formation energy of *NNH (N2(g) + [H+ + e ] / *NNH) on Pd and Au sur-
faces (Figures 10C and 10D), indicating that the N2 is activated by surface *H rather
than surface metal atoms. After *N2H2 formation via surface-hydrogenation, the
following reactions proceed by the addition of (H+ + e ) exothermically (Figure 10E).

NRR mechanism on transition metal nitrides (Mars-van Krevelen mechanism)

The reaction mechanism of NRR on transition metal nitrides is different from NRR of
pure or alloy transition metal surfaces, since surfaces of transition metal nitride are
already covered by nitrogen atoms (nitride). On transition metal nitrides, not only
N2 feed gas but also surface nitride can be reduced to ammonia. Thus, transition
metal nitrides could involve surface nitride atoms as reactants in NRR and follow a

Chem 7, 1708–1754, July 8, 2021 1727

 ll
Review

Figure 10. Volcano plot for limiting potentials (UL) of NRR as a function of *N-binding energy descriptor for fcc metal surfaces and surface-
hydrogenation mechanism on noble-metal catalysts
(A) UL for NRR as a function of *N binding on fcc(111) surfaces.
(B) U L for NRR as a function of *N binding on fcc(211) surfaces. Reprinted with permission from Montoya et al. 38 Copyright 2015 Wiley-VCH.
(C) *NNH formation free energy via * + N2 (g) / *NNH on Au and Pd surfaces.
(D) Minimum-energy pathway of *N 2 H2 formation via the hydrogenation of N 2 (N 2 (g) + 2*H / *N 2 H 2 ) on Au(100) surface.
(E) Schematic of NRR on noble-metal-based catalysts via the surface-hydrogenation mechanism. Reprinted with permission from Ling et al. 81 Copyright
2019 American Chemical Society.

so-called Mars-van Krevelen mechanism, which is generally considered for catalytic

oxidation reactions on transition metal oxides. In this mechanism, surface nitride
rather than N2 feed gas is first reduced to ammonia leaving a surface nitride vacancy.
Then, the N2 feed gas regenerates the surface recovering the nitride vacancy and
can be further reduced to ammonia. Skúlason and coworkers first suggested that
several transition metal nitrides including CrN and VN can electrochemically reduce
nitrogen to ammonia with low potential via the Mars-van Krevelen mechanism based
on DFT calculations.33 More recently, quantitative isotope (15N) experiments com-
bined with DFT calculations demonstrated that VN indeed reduces nitrogen to
ammonia via the Mars-van Krevelen mechanism.82

Insights from in situ and operando studies

In situ/operando techniques enable one to probe reaction intermediate states, iden-
tify real active sites, and provide insight into their dynamic evolution. Operando
investigation also permits identification of degradation modes of a catalyst. These

1728 Chem 7, 1708–1754, July 8, 2021

 ll
Review

Figure 11. FTIR spectra and reaction mechanisms of NRR on Au, Ru, and Rh thin film electrodes
(A) FTIR spectra during NRR on an Au film electrode in N 2 -saturated 0.1 M KOH solutions.
(B) Proposed NRR mechanism on Au surface. Reprinted with permission from Yao et al. 83 Copyright
2018 American Chemical Society.
(C) Proposed NRR mechanism on Ru surface. Reprinted with permission from Yao et al. 57 Copyright
2019 American Chemical Society.

are highly important for fundamental understanding of reaction mechanisms and

rational design of improved catalysts. Despite growing interest in this regard, in
situ mechanistic research of NRR is still in its early stages.

In situ (surface-enhanced) infrared absorption spectroscopy. In 2018, Shao et al.

detected reaction intermediates of NRR on Au thin films using a SEIRAS technique.83
They confirmed the bands in the infrared (IR) spectrum that were attributed to H–N–
H bending, –NH2 wagging, and N–N stretching of *N2Hy (3 % y % 4) species with
increasing electrode potential and identified reaction intermediates during NRR
(Figure 11A). One of the key findings was that Au follows an alternative pathway
of the associative mechanism, which involves N2H4 as an intermediate (Figure 11B).

More recently, the SEIRAS technique revealed the NRR mechanism on a Ru surface. Dur-
ing NRR, the IR signal of N=N in *N2Hx (0 % x % 2) species was detected at potentials
below 0.2 V in 0.1 M HClO4 solutions.57 However, these bands were not observed in
alkaline solutions, indicating that N2 molecules are not easily adsorbed on Ru in alkaline
electrolytes. Unlike Au (Figure 11B), Ru involved N2H2 as reaction intermediate not
N2H4. The N2H2 on the Ru surface could then be decomposed or further reduced to
ammonia (Figure 11C). In contrast to Ru that is inactive in alkaline media, Rh was shown
to manifest apparent IR signals from the end-on *N2Hx between 0.2 and 0.4 V in KOH
solutions.58 The NRR on Rh was inferred to proceed via a two-electron transfer pathway
to form an N2H2 intermediate, which subsequently decomposed to produce NH3 in the
electrolyte. Based on in situ Fourier-transform infrared (FTIR) investigations, an associa-
tive alternating mechanism was also deduced for NiFe-MoS2 nanocubes,84 Fe-doped
ReS2,85 Al-doped Co3O4 nanospheres,86 and nonmetallic N, B dual doped carbons,87
while the NRR was speculated to proceed through an associative distal pathway over
phosphorus-doped carbon nanotubes.88

Chem 7, 1708–1754, July 8, 2021 1729

 ll
Review

Figure 12. In situ Raman characterization for NRR

(A) Schematic illustration of the tailor-made electrolytic cell with a working electrode (WE), counter electrode (CE), and reference electrode (RE) for in
situ test.
(B and C) In situ Raman spectra (left) and corresponding vibration modes (right) of Re 2 MnS 6 (B) and ReS 2 (C) at different working potentials. Reprinted
with permission from Fu et al. 89 Copyright 2020 American Chemical Society.
(D and E) In situ Raman spectra of the catalyst immersed into a thin electrolyte film with applied potential from 0.1 to 1.1 V (versus RHE).
(F) Schematic depiction for the in situ electroreduction-driven transformation from nanorods to nanoparticles. Reprinted with permission from Yao
et al. 90 Copyright 2020 Wiley-VCH.

In situ Raman spectroscopy. In situ Raman spectroscopy has been employed to

distinguish intermediates at molecular level and to further uncover the N2 reduction
pathway over Re2MnS6 (Figure 12A).89 A pronounced Raman mode at 658 cm 1 was
visualized for Re2MnS6 (Figure 12B), the signal of which increased with increasing
overpotential and reached a maximum at 0.30 V. In contrast, only a weak Raman
peak was discernible for ReS2 which blue-shifted to 709 cm 1 (Figure 12C). The
two peaks at 658 and 709 cm 1 could be assigned to N–H stretching of adsorbed
*NH2*NH2 on the Re2MnS6 surface and *NNH2 on the ReS2 surface, respectively.
Detection of the *NH2*NH2 species provided evidence that the NRR was driven
by dual-metal sites following an associative alternating pathway. Identification of
*NNH2 rather than *NH2*NH2 suggested a distal pathway over single-metal ReS2.
A recent work by Qiao and coworkers showed that Bi metal organic framework
(MOF) nanorods would undergo significant structural and chemical transformation
to Bi nanoparticles (NPs) at potentials more negative than 0.5 V during electro-
chemical NRR, which was corroborated by in situ Raman spectroscopy (Figures
12D–12F).90 The in situ reduced metallic Bi species were conjectured to be the domi-
nant active sites for NRR. In situ Raman spectroscopy was used to confirm the forma-
tion of NH3 during NRR on porous PdH0.43 where it featured as a broad peak at
1,644.4 cm 1.91 The peak increased in intensity with applied overpotentials ascribed
to accumulation of generated NH3. Surprisingly, the potential where ammonia was

1730 Chem 7, 1708–1754, July 8, 2021

 ll
Review

detectable in this work was as low as 50 mV (versus RHE) for PdH0.43. Laser sources
are required in Raman spectroscopy and thus may affect the formation and detection
of NRR intermediates. To minimize such impact and avoid other optical responses
(such as fluorescence) from catalysts, near-infrared laser is preferable.

In situ DEMS. The NRR process can be tracked with the aid of online differential
electrochemical mass spectroscopy (DEMS) to monitor the volatile intermediates
and products.90 Such analysis provides important complementary direct information
on electrocatalytic mechanisms. The co-existence of mass-to-charge ratio (m/z) sig-
nals at 27, 30, 31, and 33 indicates the formation of N2H4, while the signal at 15 can
be attributed to either NH3 or N2H4. Although the signals of NH3 and water overlap
(m/z = 17 and 16, respectively), the signal intensity of NH3 varies with applied poten-
tial, which is not the case for water. Recently, Shao and co-workers58 detected the
signal of N2H+ (m/z = 29) with the H2 signal (m/z = 2) at potentials below –0.3 V
on Rh, whereas the signals of N2H2+ (m/z = 30) and N2H3+ (m/z = 31) were undetect-
able during the NRR. This provides evidence that a two-step reaction pathway may
occur over Rh involving a two-electron transfer to form N2H2 followed by a subse-
quent decomposition in electrolyte (KOH) generating NH3.

Operando XAS. Operando X-ray absorption spectroscopy (XAS) allows one to

derive useful information on the density of unoccupied states, electronic structure,
and bonding geometry of X-ray absorbing atoms. This technique has been used
to characterize the surface species of VN in NRR.82 The position and intensity of
the vanadium K-edge white line at –0.1 and –0.2 V (versus RHE) were observed to
be constant. A characteristic pre-edge peak at 5,468.4 eV was identified, indicating
the presence of oxynitride species. Furthermore, the peak became gradually weak-
ened with time during NRR, suggesting the conversion of VN0.7O0.45 to VN. The con-
version rate increased with overpotentials, reaching 57.8% after 1 h at % 0.2 V.
These results support the hypothesis that the surface VN0.7O0.45 species was the
active center in NRR, and the transformation of this phase to VN led to deactivation.

In situ XPS. By virtue of in situ X-ray photoelectron spectroscopy (XPS), the evolu-
tion of valence states of catalysts and adsorbed intermediate species can be probed.
This helps to provide insight on possible active components and reaction pathways.
Valov et al. demonstrated that N2 can be electrochemically activated at the interface
between an iridium micro-electrode and a (111)-orientated 9.5 mol % doped yttria
stabilized zirconia (YSZ) singe crystal at 450 C and 10 5 Pa by means of in situ XPS
(Figure 13).92 A broad N 1s XPS peak between 397 and 398 eV was identified under
cathodic voltages more negative than 1.5 V, which were deconvoluted into three
components. The first peak with a binding energy (BE) at 397.2 eV was assigned
to an N3– ion and the other two components at BEs of 397.7 and 398.6 eV corre-
sponded to (di)nitrogen ions in more positive oxidation states. The N 1s peak was
reversible, indicating that the nitrogen species reduced at the electrode possibly
accumulated only in the first top monolayers and easily desorbed from the surface
when the voltage was switched off. Only at potentials more negative than 2.0 V,
did reduction of the solid electrolyte take place, which could be confirmed by the
appearance of new components of Zr 3d and Y 3d signals with a lower formal charge.

Advances of spectroelectrochemical cells also make it feasible for in situ time-

resolved studies of structural changes of electrode surface accompanying the NRR
by using surface X-ray scattering technique. In particular, the dynamic motions of mi-
crostructures such as kinks and steps can be tracked, providing insight into active re-
action sites. To probe the microstructural and morphological changes of catalysts at

Chem 7, 1708–1754, July 8, 2021 1731

 ll
Review

Figure 13. Scheme of the cell arrangement for electrochemical and XPS experiments
Reprinted with permission from Valov et al.92 Copyright 2011 Royal Society of Chemistry.

the atomic scale, in situ electrochemical transmission electron microscopy (TEM)

provides a unique possibility. We expect that this powerful emerging technique
will be developed for use in NRR in the near future after some technical advances.

HETEROGENEOUS CATALYSTS FOR NRR

Experimentally explored catalysts
Materials that have been examined as catalysts for the NRR may be divided into two cat-
egories: (1) metal-containing electrocatalysts, such as pure metals, oxides, chalcogen-
ides, nitrides, carbides, sulfides, phosphides, single-metal atoms, and metal-doped car-
bons; (2) metal-free electrocatalysts including carbon and its derivatives, boron and its
compounds, polymeric carbon nitride, and black phosphorus. A summary of electroca-
talytic studies on the reduction of N2 with these materials is given in Table S1.

Metals, oxides, nitrides, carbides, chalcogenides, phosphides, bismuth

oxyhalides, hybrids, single-atom, and cluster catalysts
The empty orbitals in metals can accept the lone-pair electrons of N2 molecules to
reduce the electron density of their HOMOs, and the separated electrons in metal
atoms are donated into the antibonding p orbitals of adsorbed N2 molecules to
enhance the electron density of their LUMOs, thereby facilitating electron transfer
to weaken the NhN bond. Noble metals such as Ru,24,57 were the first cathode
demonstrated to activate N2 for hydrogenation thanks to their strongly localized
d-electrons, favorable electron affinity, and exceptional thermal and chemical sta-
bilities. Afterward, a range of other precious metals such as Au, Pd, Ag, Pt, bime-
tallic AuAg, AuCo, Au/Ni, AuCu, AuBi, PtAu, PdCu, PdZn, PdRu, PdAg, PdNi,
PdPb, RuCu, PtMo, Pt3Fe, and noble-metal-free metals including Bi, Mo, Sn, Zn,
Sb, Cu, bimetallic CoFe, MoCo, and Rh2Sb have been reported to be active for
NRR (Table S1). Thus far, an unprecedented FE reported for NH3 formation via
electrochemical NRR was achieved over couples of Ni and Au NPs (Au6/Ni), ap-
proaching 67.8% at 0.14 V (versus RHE) in 0.05 M H2SO4 electrolytes,93 although
the maximum NH3 production rate was not that high (7.4 mg h 1 mgcat. 1). How-
ever, the NH3 yield rate needs to be further confirmed by using 15N2 labeling
via 1H NMR. Despite exhibiting a lower FE of NH3 (54.96%), bimetallic Au1Cu1 af-
forded a substantially higher NH3 yield rate (154.91 mg h 1 mgcat. 1) at 0.2 V
(versus RHE) in similar acidic electrolytes.94 The outstanding performance was
ascribed to the synergy of the Au and Cu components via modulating the elec-
tronic structure and further changing the binding affinity of adsorbed N atoms

1732 Chem 7, 1708–1754, July 8, 2021

 ll
Review

Figure 14. ZIF-modified Pt/Au for enhanced NRR

(A) The functions of ZIF: (1) inducing electron-deficient sites on Pt via electron density migration
from Pt to ZIF and (2) lowering the d-band of Pt/Au.
(B) Evaluating NRR performance using 60 nm Pt/Au@ZIF at various electrochemical potentials.
(C) Proposed NRR mechanism at the Pt/Au@ZIF interfacial cavities.
Reprinted with permission from Sim et al.96 Copyright 2019 Wiley-VCH.

on the catalyst. However, the purity of both N2 and Ar used in this work was not
provided. Additionally, the origin and quantity of NH3 formed were not investi-
gated by using 15N2 isotope labeling. The NH3 yield rate was further improved
up to 228.85 G 12.96 mg h 1 mgRh 1 on high-index facets of Rh2Sb nanorods im-
mobilized on commercial carbon at 0.45 V (versus RHE) in 0.5 M Na2SO4.95 A
zeolitic imidazole framework (ZIF)-modified Pt/Au (Figure 14A) was recently shown
to drive NRR with an FE of >44.0% and an NH3 yield rate of >161.0 mg h 1 mgcat. 1
at 2.9 V (versus RHE) using dry tetrahydrofuran (THF) solution containing lithium
trifluoromethanesulfonate as an electrolyte (Figure 14B).96 Coating with ZIF was
supposed to lower the d-band position of the electrocatalyst leading to weakening
of H adsorption, and concurrently creating electron-deficient sites to enhance
catalyst-N2 interaction. The hydrophobic ZIF layer also effectively suppressed
HER. The metal-Nimidazole interaction was further posited to be crucial for kinetic
promotion of the NRR (Figure 14C). The good thing is that the authors subtracted
all FE and NH3 yield rate against control Ar experiments, which is important to
avoid the risk of misinterpretation due to background interferences. However,
the effect of contaminants in the supplied N2 was not clarified. Furthermore,
whether nitrogen in the catalytic system could have contributed to the formation
of NH3 was not confirmed by 15N-isotope labeling.

Non-noble metals, especially transition metals, have attracted extensive interest

because of their earth-abundance, low cost, and d-orbital electrons available for p
back-donation to activate N2. Apart from transition metals, the p-block metal Bi
recently gained attention in NRR owing to its poor binding with H adatoms but
strong interaction between the Bi 6p band and the N 2p orbitals.17 Defect-rich
Bi(110) nanoplates,97 and Bi-MOF derived NPs90 were all shown to catalyze NRR
with an NH3 FE exceeding 11.0% at –0.6 V (versus RHE) in either neutral or acidic so-
lutions. Quasi-spherical Bi polyhedrons with an average size of 7.4 nm afforded an
even unexpectedly high FE of up to 66.0% and an NH3 yield rate of 200 mmol g–1
h–1 (0.052 mmol cm–2 h–1) in H2SO4/K2SO4 solutions (pH 3.5).17 Potassium cations
were hypothesized to lower the free energy change for the PDS and regulate the
proton diffusion process thereby promoting selective NRR. However, N2 adsorption
on the Bi surfaces was calculated to be thermodynamically unfavorable with an en-
ergy barrier of more than 2.7 eV.22

Chem 7, 1708–1754, July 8, 2021 1733

 ll
Review

Figure 15. V14NO for electrocatalytic NRR

(A) Schematic illustration of the isotopic exchange experiments on V 14 NO to determine the density
of initial and steady-state active sites in the NRR.
(B) Isotopic composition of active surface N sites as a function of time on stream in the NRR.
(C) Amounts of produced 14 NH4 + and 15 NH4 + during isotopic exchange experiments at various time
points during the NRR.
Reprinted with permission from Yang et al.32 Copyright 2019 Wiley-VCH.

Metal oxides comprising Ti4+, Nbx+ (NbO2, Nb2O5), Fex+ (Fe2O3, Fe3O4), Cox+
(Co3O4), Mnx+ (MnO, MnO2, Mn3O4), Mo3+, W6+, Bi3+, Ta5+, Sn4+, Cr3+, and Cex+,
V4+, Zr4+, and La4+ have been evaluated for NRR (Table S1). Among these oxides,
oxygen-vacancy-rich TiO2 NPs in situ grown on conductive Ti3C2Tx nanosheets ex-
hibited an impressive NH3 yield of 32.17 mg h 1 mgcat. 1 at 0.55 V (versus RHE)
and an FE of 16.1% at 0.45 V (versus RHE) in 0.1 M HCl.98 Despite this promising
efficiency in NRR, the purity of N2 employed in this work was not reported. In addi-
tion, 15N-isotope labeling was not conducted to verify the actual reduction of N2 to
yield NH3. Nb4+ with one 4d electron was demonstrated to be more prone to back-
donation than Nb5+ without 4d electrons.99 Hence, superior NRR activity was at-
tained over NbO2 with an NH3 FE approaching 32.0% at 0.60 V (versus RHE) in
0.05 M H2SO4 solution. It is worth noting that reduction of metal oxides and removal
of surface oxygen vacancies during long-term NRR may be an issue that merits
attention.

Transition metal nitrides, such as VN, MoN, W2N3, LaN, ZrN, NbN, and CrN, have
been reported to catalyze electrochemical ammonia formation through the Mars-
van Krevelen mechanism (Table S1).100 Only those surface N sites in oxygen-modi-
fied VN catalyst involved in the NRR were claimed to be active at the steady state
(Figure 15A). The fraction of active sites could be quantified using an isotopic ex-
change method (Figures 15B and 15C).32 Activation of adsorbed N2 on the N va-
cancy of oxygen-modified VN (VNO) to adsorbed N2H was predicted to be the
rate-limiting step in the NRR. However, a recent work from Qiao et al. showed
that N-vacancy engineered 2D W2N3 nanosheets followed a distal mechanism,
and the last step of releasing the second NH3 was the PDS.101 The nitrogen va-
cancies lowered the thermodynamic limiting potential and boosted the NRR reac-
tion with an average NH3 formation rate of 11.66 G 0.98 mg h 1 mgcat. 1 (3.80 G
0.32 3 10 11 mol cm 2 s 1) and FE of 11.67% G 0.93% at 0.2 V (versus RHE) in
0.1 M KOH solution. Only 15NH4 was detected in NMR when using 15N2 as a supplied
gas, ruling out the possibility that the NH3 was generated from the decomposition of
W2N3 or other contaminants. However, quantitative isotope-labeling experiments
were not presented. Note that the chemical stability of nitrides during NRR remains

1734 Chem 7, 1708–1754, July 8, 2021

 ll
Review

an issue. For example, Mo2N was observed to undergo rapid decomposition in

aqueous electrolytes and exhibited marginal catalytic activity for the NRR.102

2D d2-d4 M3C2 MXenes without terminal groups were shown to exhibit good activity
toward the NRR.103 M2C was inferred to be more effective due to its least number of
atomic layers relative to M3C2 and M4C3 MXenes. In particular, 4d4-Mo2C was calcu-
lated to possess the lowest free energy barrier among the synthesized M2C
MXenes.104 Other metal carbides such as MoC6, MoC, Cr3C2, Fe3C, and V8C7
were also demonstrated to display reasonable activity for NRR at moderate overpo-
tentials (Table S1). It was predicted that introducing carbon vacancies to increase the
ratio of surface metal to carbon atoms could reduce the adsorption of hydrogen
atoms and improve the adsorption capacity of nitrogen molecules.105 Further inte-
gration of carbon vacancies and mesoporous structure brings about more coordina-
tively unsaturated sites to adsorb and activate N2 molecules over vanadium carbide
(V8C7), promoting the NRR to NH3 with an FE of 18.3% and an NH3 yield rate of
23.2 mg h 1 mgcat. 1 at 0.1 V (versus RHE) in 0.1 M HCl.106

Perhaps inspired by nature’s nitrogenase enzymes, MoS2 has sparked research inter-
est for NRR.84 The Mo-edge was suggested to have a relatively low barrier in the
PDS.107 Other sulfides, such as Fe3S4, FeS2, NbS2, ReS2, Sb2S3, VS2, and Re2MnS6,
have also been reported to catalyze electrochemical ammonia formation. Notably,
the Li+ interactions with S-edge sites of MoS2 were found to suppress HER by
reducing H* adsorption free energy while increasing N2 adsorption free energy,
thus, decreasing the activation energy barrier of the NRR controlling step.108 An
NH3 yield rate of 43.4 mg h 1 mgMoS2 1 and FE of 9.81% were observed in the pres-
ence of strong Li–S interactions, more than eight and eighteen times compared with
the same electrocatalyst in the absence of Li–S interactions, respectively. FeS2 can
prohibit HER. Further, due to possibility of FeS2 inhibiting the HER, doping of
FeS2 with Mo(IV) ions could favor adsorption and activation of N2, permitting selec-
tive NRR to NH3 at mild overpotentials.109 Additionally, selenides of ReSe2110 and
NbSe2111 have been explored for electrocatalytic NRR. In particular, selenium va-
cancy-rich ReSe2 nanofibers sandwiched by hydrophobic and porous carbonized
bacterial cellulose layers could effectively drive NRR with an FE of 42.5% and an
NH3 yield rate of 28.3 mg h 1 cm 2 at 0.25 V (versus RHE) in 0.1 M Na2SO4 electro-
lytes.110 However, 15N2 isotopic labeling to quantify the amount of NH3 was not
provided.

The negatively charged phosphorus can induce electron cloud density deviation of
metals and modulate the surface charge state of metal phosphides, thereby
improving their intrinsic catalytic properties. Indeed, phosphides, such as MoP,
PdP2, Ru2P, RhPx, FeP, FeP2, Ni2P, CoP, CoP3, and Cu3P, have been demonstrated
to be capable of catalyzing ambient electrochemical NH3 synthesis (Table S1).

Bismuth oxyhalides (BiOX, X = Cl, Br, I) possess a unique, layered structure, tunable
energy band gap, and strong light response to boost solar energy conversion,
showing potential in photocatalysis.112 The interlayer can induce a built-in field to
separate the charges and make the electrons be assembled in the Bi-O layer for acti-
vating N2 molecules. The edge of BiOX is constructed by O atoms and oxygen va-
cancies can easily form at the surface of BiOX under mild conditions. However, appli-
cation of BiOX in electrocatalytic NRR has been rarely reported. Only Bi4O5I2
modified with oxygen vacancies and hydroxyl groups on the surface was demon-
strated to electrocatalyze the conversion of N2 to NH3 with a prominent FE of
32.4%, outperforming most p-block-element-based catalysts.113

Chem 7, 1708–1754, July 8, 2021 1735

 ll
Review

Increasing effort is concentrated on creation of hybrids to promote nitrogen fixation

by taking advantage of a synergistic effect between different components. Amor-
phous Bi4V2O11/CeO2114 and phosphorus-doped carbon/Sb/SbPO4115 both sub-
stantially exceeded respective individual counterparts in NRR with the latter exhib-
iting an impressive NH3 FE over 30.0% in either acidic or neutral electrolytes at low
overpotentials. 3D graphdiyne–cobalt nitride (GDY/Co2N) has been applied for
NRR.116 It is noteworthy that a respectable NH3 yield rate (219.72 mg h 1 mgcat. 1)
and FE (58.6%) were achieved at atmospheric pressure and room temperature in
acidic conditions. However, despite this high efficiency, the stability of Co2N during
prolonged electrolysis needs to be resolved. The NH3 yield rate needs to be further
confirmed by quantitative 15N2 labeling measurements.

The size of a metal catalyst is a key factor in determining the catalytic performance
since the specific activity per metal atom increases with downsizing metal particle.
The ultimate specific activity can be achieved by using single-atom catalyst (SAC),
which contains atomically dispersed metal atoms in support materials. In addition
to ultimate specific activity, SAC shows distinct catalytic activity due to its size ef-
fects, such as low-coordination environment of metal center, quantum size effects,
and metal-support effects. Many SACs, such as N-doped carbon or metal-oxide-
supported single-metal atoms, have been reported in literature for various applica-
tions. In recent years, applications of SACs for NRR also have notably increased. In
2018, Choi et al. screened 120 N-doped or defective graphene-based SACs for
NRR by DFT calculations and suggested several promising SACs including Ti@N4
and [email protected] The key finding of this work was that SACs can suppress HER during
NRR compared with metal surfaces owing to the lack of atomic ensemble in the
active sites. These results indicate that SACs can be a promising platform for
increasing faradic efficiency of NRR. Several SACs have indeed been experimentally
proven to be very active for NRR in the literature (Table S1). For example, Sun and
coworkers prepared Ru single atoms embedded at ZrO2 and N-doped carbon
(Ru@ZrO2/NC), and the Ru@ZrO2/NC exhibited high NH3 yield rate (3.665 mg h 1
mgRu 1) and FE (21.0%) at 0.21 V (versus RHE) (Figure 16) in 0.1 M HCl solution.118
These data need to be further confirmed by quantitative NMR measurements. DFT
calculations revealed that Ru single atoms anchored at surface O-vacancies in ZrO2
and N-doped graphene showed low free energy for NRR and improved *N2/*H
adsorption selectivity. In addition to Ru SAC, SACs consisting of Mo, Fe, Y, and Pt
have also been experimentally demonstrated to be efficient for NRR (Table S1).
Xin and coworkers demonstrated that atomically dispersed Mo at N-doped porous
carbon (SA-Mo/NPC) could catalyze NRR with high efficiency.61 The SA-Mo/NPC
showed improved NH3 yield rate (34.0 mg h 1 mgcat. 1) and FE (14.6%) at room tem-
perature. Various types of Fe single atoms embedded at N-doped carbon have dis-
played improved catalytic activity toward NRR.119,120 Zheng and coworkers revealed
that Fe-N/C-carbon nanotube (Fe-N/C-CNT) containing Fe-N3 active sites achieved
34.83 mg h 1 mgcat. 1 of NH3 yield rate and 9.28% of FE in 0.1 M KOH solutions.120
Further DFT calculations established that Fe–N3 sites could strongly bind N2 and
reduce N2 to NH3 with 0.84 eV of free energy. FeN4 moiety embedded carbon cat-
alysts have also been reported for NRR in both neutral and alkaline media.121 A co-
ordination of Fe–(O–C2)4 as active sites supported on nitrogen-free lignocellulose-
derived carbon provided even higher NRR performance.122 The NH3 production
rate reached 307.7 mg h 1 mgcat. 1 with an FE of 51.0% at 0.1 V (versus RHE)
over Fe–(O–C2)4 immobilized on a glassy carbon electrode in 0.1 M KOH electro-
lytes. Li and coworkers demonstrated that Au single-atom-embedded C3N4 (Au1/
C3N4) exhibited outstanding NH3 formation (1,305 mg h 1 mgAu 1) and FE (11.1%)
at 0.1 V (versus RHE).123 Theoretical calculations suggested that Au1/C3N4

1736 Chem 7, 1708–1754, July 8, 2021

 ll
Review

Figure 16. NRR evaluation and structure characterization of Ru@ZrO2/NC

(A and B) (A) FEs and (B) yield rates of NH 3 over Ru@ZrO 2 /NC, Ru@ZrO 2 /C, Ru/NC, RU/C, and NC at
various applied potentials.
(C) Low-magnification HAADF-STEM image of Ru@ZrO2 /NC after six cycle tests of NRR.
(D–H) EDS maps of C (D), N (E), Ru (F), Zr (G), and O (H) atoms in the region shown in (C),
respectively.
(I) HAADF-STEM image showing the presence of single Ru sites annotated with yellow dotted
circles.
Reprinted with permission from Tao et al. 118 Copyright 2019 Elsevier.

required lower free energy for NRR than Au(211) surface. However, whether the ni-
trogen in C3N4 had contributed to the evolution of NH3 remained unclear. The for-
mation of NH3 from NRR needs to be explored by 15N2-isotope labeling qualitatively
and quantitatively.

Metal-free carbon and its derivatives, boron and its compounds, polymeric carbon
nitride, black phosphorus, and red phosphorus
A number of carbon materials, such as heteroatom (N, S, F, Te, or Se)-doped carbon,
P-doped carbon nanotube, Cl-doped graphdiyne, defective graphene, (reduced)
graphene oxide, heteroatom (P, N, B, S, and F)-doped graphene, B-doped nano-
structured diamond, B- and N-codoped carbon, and N, P-codoped carbon, were
shown to exhibit comparable and even superior NRR activity than many metal-based
electrocatalysts (Table S1). A nanohybrid of S dots and graphene was also revealed
to be active for N2 fixation.124 Strikingly, an NH3 production rate of 19.1 mg h 1 cm 2
and FE as high as 21.2% were attained over boron-doped nanostructured diamond/
Ti electrodes in 0.05 M H2SO4 solution containing 0.2 M Li2SO4.125 Of interest is that
the electrode retained stability for over 8 days without obvious decay in activity.

Chem 7, 1708–1754, July 8, 2021 1737

 ll
Review

Boron doping was supposed to induce NRR active sites in diamond lattices and
buildup of surface charges, leading to enhanced catalytic performance.

In addition to carbon-based materials, electron-deficient boron126 and its com-

pounds such as boron nitride, C-doped boron nitride, boron carbide, and boron
phosphide have emerged as appealing metal-free catalysts for the NRR. A ‘‘p
back-donation’’ process was proposed on boron (borylene) with empty sp2 orbital
and occupied p orbital for effective N2 activation.127 Compared with conventional
transition metal surfaces that were shown to require up to 1 eV for NRR by DFT cal-
culations, the (021) and (104) surfaces of boron nanosheets demand much less free
energy to drive the NRR (0.21 and 0.39 eV, respectively).126 Indeed, exposed icosa-
hedron boron in the (021) and (104) surfaces was observed to effectively catalyze the
NRR.

Polymeric carbon nitride (PCN) provides plentiful and homogeneously distributed

nitrogen vacancies (NVs).128 By heating in an atmosphere of argon, the contents
of NVs in PCN obtained by polycondensation of melamine were readily tuned.
This allows for modulation of the p-electron delocalization in the conjugated system,
resulting in strong N2 activation.128

Zigzag and diff-zigzag edges of exfoliated black phosphorous (BP) nanosheets were
recently shown to be active centers for N2 adsorption and NhN triple bond activa-
tion, which enable NRR via an alternating hydrogenation pathway, delivering a high
NH3 production rate of 31.37 mg h–1 mgcat.–1 albeit with a moderate FE of 5.07% at
0.6 V (versus RHE) in 0.1 M HCl solution.129 The FE for NH3 formation was stated to
be improved by N and O doping of BP surfaces to impart hydrophobicity and hinder
the competitive HER.130 However, BP suffers from oxidative degradation especially
in the presence of water,131 thus, limiting its practical application in NRR. Nanorib-
bons of red phosphorus, a more stable allotrope of phosphorus, were shown to facil-
itate NRR, providing an FE of 9.4% and an NH3 yield of 15.4 mg h–1 mgcat.–1 at –0.5 V
(versus RHE) in 0.1 M Na2SO4.132 The NRR performance was credited to the
apparent anisotropy and the quasi-one dimensional morphology of nanoribbons,
endowing red phosphorus with efficient charge transfer and abundant exposed-
edge sites.

Potential alternative catalysts for N2 conversion

To date, various types of catalysts have been suggested for NRR. Here, we briefly
introduce promising catalysts for NRR suggested in the literatures, such as boron-
based catalysts and metal sulfides, but not experimentally realized. Previous studies
have shown that boron can reduce N2 both in homogeneous and heterogeneous
catalysts. It has been shown that boron atom can activate N2 with p back-donation
which is the mechanism of N2 activation on transition metals (Figure 17A).127 Exper-
iments combined with DFT calculations demonstrate that boron 2D nanosheets can
facilitate NRR with low overpotential.126 Also, boron-doped or boron-based cata-
lysts, such as B-doped graphene36 and B4C,133 have been proven efficient for
NRR experimentally. In addition to these catalysts, single-boron-doped catalysts
and transition metal borides have been suggested for NRR by DFT calculations (Fig-
ure 17B). Liu et al.134 explored boron single site doped in various 2D materials (gra-
phene, BN, BS, BP, s-triazine, tri-s-triazine, h-MoS2, and t-MoS2) and showed that
negatively charged boron sites are promising for low overpotential and high selec-
tivity for NRR. Among various boron sites, boron-doped graphene and h-MoS2 were
identified as promising boron sites for NRR. In addition to doped boron sites, tran-
sition metal borides (MBx) have been suggested for NRR.135 Surface boron atoms in

1738 Chem 7, 1708–1754, July 8, 2021

 ll
Review

Figure 17. Volcano plot for NRR activity on transition metal sulfides
(A) Simplified schematic of the end-on bound transition metal-N 2 complexes (left) and borylene-N2
complex (right). Reprinted with permission from Légaré et al. 127 Copyright 2018 American
Association for the Advancement of Science.
(B) Different boron doping sites in various 2D materials and maximum free energy required for NRR.
Reprinted with permission from Liu et al. 134 Copyright 2019 American Chemical Society.
(C) Volcano plot for limiting potential of NRR on various transition metal sulfides via associative
mechanism (left) and dissociative mechanism (right). Reprinted with permission from Abghoui
et al. 136 Copyright 2019 Wiley-VCH.

Mo2B, Fe2B, and Co2B are suggested for NRR with ( 0.75 ~ 0.84 V versus RHE) be-
ing the optimal limiting potential.135

As described above, ammonia in nature is synthesized at ambient conditions by en-

zymes whose active site, for example, is a transition-metal sulfide cluster (Mo-
Fe7S9N). Skúlason and coworkers studied the possibility of catalyzing ammonia for-
mation on the surface of transition metal sulfides that, owing to presence of the

Chem 7, 1708–1754, July 8, 2021 1739

 ll
Review

transition metal and sulfur atoms, offer some similarity to the structure of active sites
of the enzyme.136 NRR on various low-index surfaces of transition metal sulfides such
as rocksalt(100), NiAs-type(111), pyrite(100), and pyrite(111) via associative and
dissociative mechanisms were investigated. They obtained a volcano plot for NRR
activity on transition metal sulfides (Figure 17C). RuS2 was predicted to have low
overpotential (0.3 V), and the other promising candidates (CrS, NbS, VS, and TiS)
showed 0.7~1.1 V of overpotential.

ML models for NRR

Recently, machine learning (ML) approaches have become popular and proven use-
ful in the design of catalysts as the advancement in computational power led to big
data and development in ML methods.137 In the NRR field, ML has been used to
screen promising catalysts138,139 and optimize morphology of catalysts.140 Here,
we briefly discuss several ML models that were specifically used to develop NRR cat-
alysts, and refer readers to the recent comprehensive review papers for general as-
pects of ML in catalysis.137,141

Kim et al.138 screened various transition metal alloys for NRR by developing a slab
graph convolutional neural network (SGCNN), which is an extended model of the ex-
isting crystal graph CNN (CGCNN).142 The CGCNN has proven its accuracy in pre-
dicting bulk properties but not yet tested for surface-related properties at that time.
To incorporate surface effects, surface graph (S1 or S2) and bulk graph (B) are con-
structed (Figure 18A). Total 3,040 binding energies, which consist of 465 catalysts
with different crystal spaces (fcc, bcc, hcp, L10, B2, and core-shell), and 5 adsorbate
types (H, N2, N2H, NH, and NH2) and sites were used for the SGCNN training. For
input features, a combination of 5 low-level input features (group number [GR], Paul-
ing-scale electronegativity [EN], atomic volume [AV], electron affinity [EA], and
atomic weight [AW]) was used since it shows minimum mean absolute error (MAE)
of 0.23 eV (Figure 18B). The SGCNN model trained by 465 catalysts was used for
screening 870 candidates consisting of (111) facet of A3B structure (Figure 18C).
Among 870 candidates, 10 catalysts (e.g., V3Ir, Tc3Hf, V3Ni, and Tc3Ta) were
included in the screening criteria (Figure 18D).

Various structures of single transition-metal-atom-embedded B-doped graphene

were explored for NRR by using machine learning-based methods (Figure 18E).139
The Coulomb matrix was used as a descriptor for representing atomic geometric
structures. The eligible catalysts which could facilitate *N2, *N2H, and *NH3 forma-
tion were classified by using the deep neural network (DNN). The reaction free en-
ergies were predicted by ML model trained by the light gradient boosting machine
(LGBM) model (RMSE = 0.11 eV). The most important feature was the coordination
number of the metal and the number of hydrogen atoms for the DNN and LGBM
models, respectively. Promising catalysts, such as CrB3C1 (UL = 0.29 V), HfB1C2
(UL = 0.42 V), and TcB3C1 (UL = 0.44 V), were finally screened.

ML accelerates not only the catalysts screening process but also optimization of the
material morphology. Liu and coworkers explored how the electrode’s morphology
contributes to NRR activity and provided the means to quickly optimize wire-array
morphology.140 About 200,000 finite element methods (FEM)143 simulations were
conducted to yield values of total current density (itotal) and FE of NRR as a function
of feature variable vector x, which is described in Figure 18F. The variable vector x
consists of physical features of morphologies such as diameter (d), length (l), and
array periodicity in a square lattice (p) and kinetics information such as exchange cur-
rent density of the Volmer step of HER (i1,0,HER), the Heyrovsky step of HER (i2,0,HER),

1740 Chem 7, 1708–1754, July 8, 2021

 ll
Review

Figure 18. Studies of ML on screening promising NRR catalysts and optimizing morphologies of
catalysts
(A) Schematic of the main architecture of the SGCNN model. B and S 1 (or S 2 ) represent the bulk
graph and surface graph with the top layer (or top two layers), respectively.
(B) MAE values as a function of various feature combinations.
(C) Bulk and (111) surface of the L1 2 ordered phase of A 3 B (left) and screening process to search for
stable and efficient NRR catalysts, which potentially outperform Mo(110) (right).
(D) UL and F values predicted by the SGCNN model for 10 screened promising catalysts. Reprinted
with permission from Kim et al. 138 Copyright 2020 American Chemical Society.
(E) Structure of B-doped graphene with single-metal atoms considered for screening. Reprinted
with permission from Zafari et al. 139 Copyright 2020 Royal Society of Chemistry.
(F) ANN (10 neurons in each hidden layer) architecture and criteria for eligible NRR catalysts.
(G) Selection of hypothetical promising catalysts for NRR among over 4,000,000 scenarios
generated by trained MLPNNs. Reprinted with permission from Hoar et al. 140 Copyright 2020
American Chemical Society.

Chem 7, 1708–1754, July 8, 2021 1741

 ll
Review

the first electron transfer step of NRR (i1,0,NRR) and the first proton transfer step of
NRR (i2,0,NRR), and overpotential (h). The two feed-forward144 multilayer perceptron
neural networks (MLPNNs) were used for training the FEM datasets. The MLPNN
prediction shows an accuracy of 2.810%2 and 6.76 3 10 4 mA2/cm4 of mean squared
error (MSE) for FE of NRR and total current density, respectively. In total, 4,000,000
scenarios were generated by trained MLPNNs and over 200 scenarios, which meet |
itotal| > 2.0 mA/cm2 and FE > 90.0% were shown in Figure 18G. Top 4 scenarios which
are denoted in Figure 18G, converge to a similar range of parameter values, h z
0.7 V versus RHE, d z 8 mm, l z 40 mm, p z 10 mm, i1,0,NRR z 10 3, i2,0,NRR z
10 3, i1,0,HER z 10 8, and i2,0,HER z 10 8 mA/cm2, corresponding to the reasonable
kinetic reactivity and geometry for optimal NRR catalyst.

All of these ML approaches efficiently predict NRR activity. For example, 870 cata-
lysts were screened by the SGCNN model trained by data of 465 catalysts (53% of
screening data)138 and 4,000,000 scenarios were investigated by using the MLPNNs
trained by 200,000 FEM data (5%).144 In other words, noticeable amounts of poten-
tially unnecessary simulations (from 47% to 95%) were eliminated.

Challenges and outlooks of theoretical approaches for understanding NRR

As aforementioned, theoretical studies have played an important role in elucidating
the reaction mechanism of NRR, understanding the NRR activity trend, and suggest-
ing promising catalysts (Figures 9, 10, and 17). Furthermore, ML has been emerged
as a sufficient method for accelerating the development NRR catalysts (Figure 18).
However, current result of DFT calculations sometimes does not describe the exper-
imental results. The most dominant reaction during NRR is the HER; however, the
competition between the NRR and HER has not been well discussed theoretically.
Here, we discuss the challenges and outlooks of theoretical approaches for further
understanding of NRR.

One of the most significant problems in current DFT calculations is that the reported
theoretical limiting potentials disagree with the experimental electrode potential at
maximum NRR activity in most cases. The theoretical limiting potentials (UL), where
the overall electrochemical reaction becomes thermodynamically exothermic, on
various fcc transition metal surfaces are at least –1 V (versus RHE) theoretically (Fig-
ure 10). However, maximum NRR activities are often observed at –0.05 ~ –0.60 V
(versus RHE), highly more positive than theoretical UL (Table S1). For example,
maximum NH3 yield rates were obtained at –0.1 V (versus RHE) on Ru nanopar-
ticles145 and –0.2 V (versus RHE) on Au nanorods,146 while their theoretical UL are
around at –1.0 V (versus RHE) and –3.0 V (versus RHE), respectively. The origin for
such large discrepancy has not been clearly clarified; however, a possible reason
has been suggested, namely dominant H coverages at negative electrode poten-
tials. The H adsorption via * + (H+ + e ) / *H (due to its electrochemical nature
of the reaction) increases faster than the N2 adsorption (* + N2 / *N2, non-electro-
chemical reaction) with more negative electrode potential.27 Thus, dominant H cov-
erages at negative electrode potentials block the N2 adsorption after which NRR ac-
tivity decreases although the electrochemical reduction rates of NRR themselves
might increase with more negative potential. However, this possibility has not
been quantitatively established or validated since the in situ measurements of N2
and H coverage during NRR is still highly challenging. Thus, the in situ investigation,
either experimentally or theoretically, of N2 versus H coverage changes and NRR ac-
tivity with electrode potential would provide invaluable insights to fully understand
the origin of the decreasing NRR activity with electrode potential.

1742 Chem 7, 1708–1754, July 8, 2021

 ll
Review

Most of theoretical studies on NRR neglected the kinetics and focused on the ther-
modynamic reaction energies.147 Only a few studies considered activation energy
during the proton-electron transfer reactions in NRR.148,149 To fully understand
the NRR, theoretical studies should consider the potential-dependent competition
between NRR and HER including the N2 versus H binding and kinetics.

Electrolytes as well as the active sites are important in enhancing the NRR activity. By
tuning the electrolytes (e.g., ILs and non-aqueous electrolytes), enhancement in N2
solubility and FE of NRR can be achieved.15,74 Several theoretical studies on the ef-
fect of electrolyte have been reported. For example, Zhang et al. demonstrated that
HER can be suppressed during NRR by utilizing a non-aqueous bulky proton
donor150 and Ortuño et al. showed stabilization of reaction intermediates of NRR
by ILs.76 We expect that further theoretical studies on electrolytes at atomic level
will be helpful for designing electrolytes.

So far, ML techniques have been rarely used in the NRR field, while they have been
widely utilized for various chemical reactions. We believe that ML approaches devel-
oped in other catalytic reactions would be also helpful in investigating the NRR. For
example, Ulissi and coworkers identified the most likely reaction pathways of syngas
on Rh(111) among more than 2,000 potential pathways by using the Gaussian pro-
cess with group additivity descriptors.151 Similar ML approaches will be applied
for investigating the reaction mechanism and reaction kinetics of NRR. While various
catalysts have been theoretically suggested, synthesizability as well as NRR activity is
also quite important. Recently, Jang et al. proposed data-driven metric of the crys-
tal-likeness score, which estimates the synthesizability of crystalline material.152 This
ML approach will be employed to estimate the synthesizability of proposed catalysts
before its experimental validation.

STRATEGIES FOR IMPROVING N2 ELECTROCATALYTIC ACTIVITY

Strategies to promote the electrocatalytic performance of materials for NRR should
take into account the following: (1) enhancing adsorption and activation of N2 at low
overpotentials, (2) suppressing side reactions such as HER, and (3) increasing long-
term stability, as illustrated in Figure 19. This section focuses on basic information
and understanding for improving the activity.

Surface engineering
Surface modification
Surface functionalization with chemical moieties offers an effective way to tune elec-
trocatalysis through modulating the electronic properties of catalyst surfaces and
the resultant adsorption energies of reaction intermediates.12 Surface selenation
was recently demonstrated to greatly boost NRR performance (Figure 20).153 An up-
shift of the d-band center was observed for surface selenated Rh nanocrystals (NCs),
which was hypothesized to strengthen N2 adsorption. The Se sites may also play a
role in promoting electron transfer between N-species and nearby Rh-sites. Alterna-
tively, introducing –O or –OH species was shown to facilitate improved NRR. Oxygen
functional groups in oxidized carbon nanotubes or graphene oxide alter the electron
distribution in the surrounding carbon atoms, resulting in optimized binding
strengths of N2 and NRR reaction intermediates on carbon to enhance NRR.154 Sur-
face hydroxyl modification of Ti3C2 MXene either by alkalization intercalation and
agitation155 or via ethylene glycol solvothermal treatment156 was manifested to in-
crease the electronic density of states at the Fermi level and lower resistance, as
compared with oxygen termination and original Ti3C2. This favors electron transfer
to the 2p orbitals of N2 and promotes its chemisorption and activation. By contrast,

Chem 7, 1708–1754, July 8, 2021 1743

 ll
Review

Figure 19. Summary of strategies for improving the performance of electrocatalysts for NRR

Ti3C2 with –F groups was predicted to exhibit poor N2 adsorption capacity with
higher energy barrier, unfavorable for NRR. Li+ modification was shown to enable
suppression of the undesired HER.62 It may also afford positively charged active sites
that favor N2 capture and reduction.108 Enlightened by the metal-sulfur (M–S) link-
ages in the nitrogenase enzyme, surface tethering of Ru with aliphatic thiols was
applied to modify its electronic structure, thus, promoting adsorption and polariza-
tion of N2.157 Concomitantly, the organic and hydrophobic nature of the thiols
impeded proton transfer while allowing flow of N2 to the surface of the catalyst.
This synergistically boosted the rate of NH3 generation.

Heteroatom doping can profoundly adjust the electronic structures of materials.

Doping of nanocarbons with single or multiple heteroatoms (p-block elements
such as B, N, O, S, Se, Te, P, F, Cl, Br, and I) was revealed to induce charge and
spin densities on adjacent carbon atoms.158 Charge accumulation favors N2 adsorp-
tion on carbon atoms, while spin polarization facilitates the PDS of the first proton-
ation to yield *NNH. These are expected to synergistically regulate the competition
between NRR and HER. Boron doping can create Lewis acid sites with empty orbitals
to bind N2 forming B–N bonds. Among the boron structures, BC3, BC2O, and BCO2,
BC3 was computed to display the lowest reaction energy barrier, which probably
functioned as the major electrocatalytic centers in NRR.36 However, the overpoten-
tial for HER may decrease with the increase of the boron content.125 There exists an
optimal boron content to improve the overall NRR efficiency and selectivity. The pyr-
idinic N3 configuration embedded in a graphitic layer, containing one protonated

1744 Chem 7, 1708–1754, July 8, 2021

 ll
Review

Figure 20. Schematic illustration of selenization process and comparison of Rh, Pt, Pd, and Ru for
NRR before and after surface selenization
(A) Schematic illustrating selenization of Rh NCs.
(B) Schematic showing the proposed mechanism for NRR on the catalytic surface of the Rh NCs/C
and Rh-Se NCs/C.
(C) NH3 yield at 0.10 V (versus RHE) of various catalysts.
(D) FE for NH3 formation at 0.10 V (versus RHE) on various catalysts.
Reprinted with permission from Yang et al.153 Copyright 2020 Wiley-VCH.

pyridinic nitrogen and one neighboring atomic vacancy site, was predicted to pro-
mote N2 chemisorption and reduction with relatively lower free energy require-
ments. The content of pyridinic-N in carbons can be further improved by addition
of a secondary dopant (B or F), resulting in higher NRR activity.87 In contrast to pris-
tine and O-doped carbon atoms that barely adsorb N2, S-, Se-, and Te-substituted C
exhibited substantially increased N2 adsorption, benefiting NRR.159 Apart from the
above-mentioned effects, doping of elements (such as B, F, and low-valence transi-
tion metal dopants)160 into metal oxides or oxide-hydroxides may enrich surface ox-
ygen vacancies by replacing the O atom rather than the metal atom.161 Alternatively,
doping of TiO2 with stable Zr4+ induced tensile and compressive strains, hence,
generating oxygen vacancies associated with adjoining bi-Ti3+ sites.161 The oxygen
vacancies promoted the back-donation of electrons to N2 and accelerated the con-
version of *NH2NH2 to *NH2 and *NH3. The dopant and oxygen vacancies jointly
contribute to efficient conversion of N2. Incorporation of Fe into reducible CeO2
also facilitated formation of active Ce3+–Ce3+ pairs to improve the selectivity of
NRR with respect to HER.162 Additionally, morphology evolution from CeO2 nano-
particles to nanosheets with a larger surface area was induced upon Fe doping.
These coupling impacts led to enhanced NH3 yield. A recent work from Chen
et al.85 showed that M4+-positioned Re in ReS2 nanosheets could be replaced by
metal species (e.g., Fe, Co, Ni, Cu, and Zn), leading to intensified charge density

Chem 7, 1708–1754, July 8, 2021 1745

 ll
Review

around the dopant cations. The low-valent metal domain was revealed to offer more
active sites for N2 chemisorption and weaken the NhN bonds for easier dissociation
through proton coupling.

To further boost NRR activities, there are continuous efforts to modify the bulk and
surface electronic properties and physical characteristics of electrocatalysts by
tailoring the types and location of dopant atoms. Elucidation of active sites and opti-
mization of activity are critically required as well. This is likely to be addressed by
combining knowledge-guided controllable synthesis complimented by improved
in situ and operando characterization techniques. Noteworthy, electrocatalysis of
the NRR on N-doped materials should be studied with scrutiny to confirm the origin
of evolved NH3.

Surface structure tuning

Modulation of the surface structure of electrocatalysts can be achieved mainly by
tailoring several important aspects: (1) surface amorphism; (2) exposed crystal fac-
ets; (3) lattice strain; (4) surface active areas; (5) surface porosity; (6) edges; (7) surface
vacancies; (8) surface phase; and (9) single-atom sites.

Amorphization of an electrocatalyst enables a large increase of unsaturated coordi-

nation sites (resulting from dangling bonds), often giving rise to better catalytic ac-
tivity than the crystalline counterpart.163 In addition to abundant active sites, local-
ized electrons in metal oxides could be augmented for p-back-donation, favoring N2
activation and NhN cleavage.114 Despite these advantages, amorphous materials
suffer from poor chemical and structure stability, hindering practical applications
in NRR.

Crystal facets play a pivotal role in determining activity and selectivity of reactions.
Surface facets differ in Lewis acidity and polarizing power and can therefore affect
the adsorption and activation of N2. Exposed facets may also influence adsorption
of electrolytes on the catalyst surface, thus, affecting the catalytic performance. Fac-
ets that can selectively stabilize *N2H or destabilize *NH2 are favorable to reduce N2
to NH3. Guided by these findings, Mo(110) nanofilms,164 Au(730) nanorods,146 and
RhSb nanorods with high-index facets95 were designed for effective N2 electrore-
duction. For most metals except Re, the (211) step was modeled to be more reactive
with lower theoretical overpotential than the (111) terrace.38

Strain (by changing the lattice spacing) can increase surface reactivity and selectivity
of electrocatalysts by breaking the linear scaling relationship. In the d-band model,
lattice expansion mitigates orbital overlap in transition metal atoms, thus, narrowing
d-band along with shift to a higher energy level. This leads to a stronger interaction
with adsorbates. In the eigenstress model, eigenstress takes place where adsorbates
evoke strain at the catalyst surface. The binding energies of different intermediates
can vary independently according to sizes and binding interactions with the strained
surface. It is worth pointing out that the explicit use of strain to enhance the NRR is
rarely reported.

Design of catalysts with high active surface areas can enrich the accessibility of cat-
alytic sites for reactions.165 This allows large physicochemical contact between the
catalyst surface and N2, which is essential for ambient N2 conversion. The interac-
tions between electrode and electrolyte are also strengthened, promoting interfa-
cial adsorption-desorption dynamics and mass transport. Likewise, creation of a hi-
erarchical architecture with high porosity and interconnected micro-, meso-, and

1746 Chem 7, 1708–1754, July 8, 2021

 ll
Review

macroporous texture offers a way to enhance the adsorption and capture of N2,
potentially enhancing N2 conversion rate and efficiency.165 Macropores (>50 nm)
accelerate transport and diffusion of reactants and products over electrodes,
whereas mesopores (2–50 nm) and micropores (<2 nm) facilitate a high exposure
of active sites. Further tailoring the diameter, volume, structure, and hydrophobicity
of pores or channels allows for effective regulation of activity and selectivity.

Edges have coordination unsaturation and dangling bonds, providing useful cata-
lytic sites. Increase of edge defects was demonstrated to promote NRR electrocatal-
ysis. N2 molecules were observed to preferably interact with the middle Ti atoms at
the edge plane of Ti3C2 MXenes with low activation barriers.166 As opposed to the
basal plane that is inert for NRR, the edges of MoS2 are capable of polarizing and
converting N2 to NH3.167 The positively charged Mo-edge was revealed to aid the
formation of Mo N bonds and reduce energy barriers for the PDS of *N2 proton-
ation to form *NNH.

Engineering of surface vacancies can regulate the electronic structure of catalysts,

enabling enhanced charge transfer and optimal adsorption energetics of intermedi-
ates for electrocatalytic reactions. Oxygen vacancies in metal oxides,21,35 NVs in
polymeric carbon nitride128 and metal nitrides,2,32,101 and chalcogen vacancies in
metal chalcogenides110 have been claimed to efficiently bind and activate N2. The
underlying role of these vacancies is to modulate electron delocalization (inducing
electron deficiency), boosting back-donation of electrons into the p N–N antibond-
ing to weaken the NhN triple bond. Additionally, carbon vacancies in MoC were
predicted to diminish proton accumulation and increase the surface affinity of the
metal carbide for N2 and the NRR reaction intermediates.168 Analogously, unsatu-
rated carbons with single vacancy or double vacancy in reduced graphene oxide
were found to bind N2 strongly but suppress the binding of H, thereby improving
the NH3 selectivity.154 Apart from the vacancies mentioned above, cation vacancies
such as Fe3+,169 Mo3+,170 and Bi3+97 also enable a stronger binding of N2 and its
reduction intermediates, in effect lowering the overpotential required for the NRR.
However, despite tremendous attempts to tune vacancies for efficient NRR, the
decisive factors that vacancies play to determine catalytic behaviors remain to be
answered. Creating a synergism between different vacancy sites may be promising
to further enhance NRR. Equally importantly, the evolution of vacancies during NRR
needs to be clarified. The surface vacancies may be deactivated by refilling corre-
sponding atoms, restricting long-term N2 fixation. To ameliorate this issue, partially
filling these vacancies with an appropriate nonmetal dopant was shown to be
useful.171

Surface active phases can be tailored by alloying or construction of bimetallic struc-

tures. This can induce surface rearrangement as well as spill-over effects and alter
adsorption energies of reactants and intermediates, ultimately affecting reaction ki-
netics and selectivity. By incorporation of a suitable second metal, the d-band posi-
tion of a metal can be lowered, weakening H adsorption. Concurrently, electron-
deficient sites are created, affording strong N2 affinity to kinetically drive NRR.96
Additionally, effective d–d coupling between two metal sites may occur, which
can bridge the electron transfer Coulomb gap toward rapid NRR.172

Single-atom sites display distinct catalytic activities due to low-coordination envi-

ronment of the metal center, quantum size effects, and metal-support interactions.
Isolated metal atoms have shown efficiency in catalyzing NRR at low overpotentials.
The single-atom sites that are highly spin-polarized were supposed to activate *N2,

Chem 7, 1708–1754, July 8, 2021 1747

 ll
Review

stabilize *N2H, and destabilize *NH2 species. Meanwhile, adsorption of H on single

atoms (with only top adsorption sites) may be hampered as a result of an ensemble
effect, thus, restraining the HER.117 Besides, single atoms supported on a curvature-
rich surface may trigger interfacial polarization, which drives electron injection into
the antibonding orbitals of N2 to accelerate NRR.173 However, more efforts are
needed to unveil the exact coordination states of single sites for NRR. Finely tuning
the interface between single-metal sites and support118 as well as two individual dis-
similar atoms may further improve NRR.

Construction of heterostructures
Heterostructure formation can enhance charge transfer and induce combined elec-
tronic and geometric effects to facilitate N2 conversion. Relevant parameters such as
interfacial compositions and facets, active surface areas, interfacial defects, and elec-
tronic coupling can be tuned to reach an optimum for NRR. For example, optimizing
the interaction between entities at the interface permits creation of remarkably stable
active sites for N2 activation.116 Hydrogen evolution is also impeded. In particular,
the synergy arising from the interaction of different components can be attuned to
maximize the orbital overlap with N2 for optimal adsorption and reduction. Alterna-
tively, a synergistic interplay between one phase as N2 binding and activation sites
and nearby phases as centers for hydrogenation could facilitate NH3 formation. Such
concept has been underscored by introducing a CoOx layer to modulate the local elec-
tronic structure of Au NPs with positive valence sites.174 The Au1+ active sites on Au NPs
facilitate break of the NhN bond and formation of N=N– intermediates via NRR, while
Au0 likely binds N–N– intermediates for subsequent hydrogenation reaction.

CONCLUSIONS AND FUTURE PERSPECTIVE

Over the past 5 years, significant and encouraging advances have been achieved for
ambient electrochemical NRR to synthesize NH3. However, there remain key challenges
that need to be addressed in realizing the ‘‘green ammonia economy.’’ Despite being
thermodynamically feasible, high overpotentials are usually demanded to overcome
the kinetic barrier of NhN cleavage. Both yield rate and FE toward NH3 formation
remain low, which are typically less than 10–8 mol cm–2 s–1 and 10% (with the HER as
the dominant reaction) respectively, far below the rate of 9.3 3 10 7 mol cm 2 s 1
that is desired for a commercially viable project.175 Large overpotential and low selec-
tivity intensify energy efficiency problem. Much more research efforts are thus required
to make the electrochemical route competitive with the current NH3 industry. The
longevity of the NRR process is another critical concern. The normally reported stability
tests in the literature are very commonly under 20 h, which is a too short duration for
industrial implementation where thousands of hours of stable operation at high current
density are expected. Therefore, future protocols for catalyst durability need to target
much longer test durations. Endeavors also need to be put on illumination of catalyst
degradation mechanisms. To promote the electrochemical synthesis, catalyst and elec-
trolyte should always be taken into account and optimized together. Hence, we provide
the following perspectives:

(1) Coupling different strategies for design of electrocatalysts. Compared with

other electrochemical processes, the variety of materials reported for NRR
is rather small. Investigation of new promising electrocatalysts continues to
be an exciting arena. A combination of various design schemes enables
one to jointly/collaboratively optimize the binding energies of the reactant
and key intermediates, facilitating preferential NRR electrocatalysis.
Mimicking an enzyme system allows for creation of multi-model active sites

1748 Chem 7, 1708–1754, July 8, 2021

 ll
Review

to improve catalyst performance. Meanwhile, manipulation of the interplay

between single atoms with different compositions with HER-inactive supports
may maximize the exposure of active sites while concurrently suppressing the
parasitic HER.
(2) Tuning electrocatalytic interface for active and selective NRR. At the electrode-
electrolyte interface, the NRR and HER contest for protons from the electrolyte,
electrons transferred from the catalyst, and surface active sites. The competition
from the HER drastically affects the selectivity of NRR and governs the kinetics of
NRR, such that the NRR activity drops at relatively more negative potentials. The
low adsorption of N2 molecules on the catalyst surface is another hurdle that
limits the NRR. Design of a solid-liquid-gas tri-phase interface in a flow cell shows
the potential to mitigate these issues. In this system, an airbag structure is
formed on the surface of a superhydrophobic substrate. N2 can be rapidly and
continuously delivered to the catalyst layer, ensuring a high local concentration
of the reactant. The contact area and time between the catalyst and N2 are also
greatly increased, leading to a high N2 reduction rate.
(3) Combining in situ characterizations, theoretical calculations, and predictions
(using a ML model). Development of in situ/operando techniques will aid in
building up more valid theoretical models and deepen understanding of re-
action pathways, degradation modes as well as catalytic structure-property
relationships specific to given materials, further guiding new direction for
catalyst design. Greater advances are expected through the collective knowl-
edge and insights to be gained from fundamental research that integrates ex-
periments and theory.
(4) Establishing standards and protocols for accurate, reliable, and repeatable
NRR measurements. The impurity composition of N2 supply should be clearly
specified. Highly pure N2 (R99.999%) and Ar (R99.999%) are suggested for
adoption in NRR. Prior to NRR, the N2 feed gas, electrolyte salts, electrocata-
lysts, and PEMs need to be purified to rule out artifacts from exogenous nitro-
gen contaminants (NH3, nitrate/nitrite, and nitrous oxide). In particular, for
catalysts that either contain structural/lattice nitrogen or are prepared from
nitrates or ammonium precursors, 15N2 isotopic labeling should be conduct-
ed to confirm the actual reduction of N2. To avoid inadvertent NH3 contami-
nation and guarantee a true NRR, multiple control experiments at different
applied potentials are necessary. It is also urgently required to develop
more selective, precise, and in situ methods for NH3 determination.

SUPPLEMENTAL INFORMATION
Supplemental Information can be found online at https://doi.org/10.1016/j.chempr.
2021.01.009.

ACKNOWLEDGMENTS
This work was supported by the National Natural Science Foundation of China
(21972010); Beijing Natural Science Foundation (2192039); the State Key Laboratory
of Organic-Inorganic Composites (oic-201901001); Beijing University of Chemical
Technology (XK180301, XK1804-2); and National Research Foundation of Korea
(NRF-2019M3D1A1079303 and NRF-2019M3D3A1A01069099).

AUTHOR CONTRIBUTIONS
Conceptualization, Z.S.; writing – original draft, Z.S., H.S., X.L., C.C., and Y.J.;
writing – review & editing, Z.S., J.M., and J.Q.; funding acquisition, Z.S. and Y.J.; su-
pervision, Z.S.

Chem 7, 1708–1754, July 8, 2021 1749

 ll
Review

DECLARATION OF INTERESTS
The authors declare no competing interests.

REFERENCES
 c, V., Yang, S., Schwalbe,
1. Andersen, S.Z., Coli 15. Zhou, F., Azofra, L.M., Ali, M., Kar, M., 26. van der Ham, C.J.M., Koper, M.T.M., and
J.A., Nielander, A.C., McEnaney, J.M., Simonov, A.N., McDonnell-Worth, C., Sun, C., Hetterscheid, D.G.H. (2014). Challenges in
Enemark-Rasmussen, K., Baker, J.G., Singh, Zhang, X., and MacFarlane, D.R. (2017). reduction of dinitrogen by proton and
A.R., Rohr, B.A., et al. (2019). A rigorous Electro-synthesis of ammonia from nitrogen electron transfer. Chem. Soc. Rev. 43, 5183–
electrochemical ammonia synthesis protocol at ambient temperature and pressure in ionic 5191.
with quantitative isotope measurements. liquids. Energy Environ. Sci. 10, 2516–2520.
Nature 570, 504–508. 27. Hu, L., Xing, Z., and Feng, X. (2020).
16. Chalkley, M.J., Del Castillo, T.J., Matson, B.D., Understanding the electrocatalytic interface
2. Ye, T.N., Park, S.W., Lu, Y., Li, J., Sasase, M., and Peters, J.C. (2018). Fe-mediated nitrogen for ambient ammonia synthesis. ACS Energy
Kitano, M., Tada, T., and Hosono, H. (2020). fixation with a metallocene mediator: Lett. 5, 430–436.
Vacancy-enabled N2 activation for ammonia exploring pKa effects and demonstrating
synthesis on an Ni-loaded catalyst. Nature electrocatalysis. J. Am. Chem. Soc. 140, 6122– 28. Sippel, D., Rohde, M., Netzer, J., Trncik, C.,
583, 391–395. 6129. Gies, J., Grunau, K., Djurdjevic, I., Decamps,
L., Andrade, S.L.A., and Einsle, O. (2018). A
3. Soloveichik, G. (2019). Electrochemical 17. Hao, Y.C., Guo, Y., Chen, L.W., Shu, M., Wang, bound reaction intermediate sheds light on
synthesis of ammonia as a potential X.Y., Bu, T.A., Gao, W.Y., Zhang, N., Su, X., the mechanism of nitrogenase. Science 359,
alternative to the Haber–Bosch process. Nat. Feng, X., et al. (2019). Promoting nitrogen 1484–1489.
Catal. 2, 377–380. electroreduction to ammonia with bismuth
nanocrystals and potassium cations in water. 29. Skúlason, E., Bligaard, T., Gudmundsdóttir,
4. Liu, T., Gau, M.R., and Tomson, N.C. (2020). Nat. Catal. 2, 448–456. S., Studt, F., Rossmeisl, J., Abild-Pedersen, F.,
Mimicking the constrained geometry of a Vegge, T., Jónsson, H., and Nørskov, J.K.
nitrogen-fixation intermediate. J. Am. Chem. 18. Ma, J.L., Bao, D., Shi, M.M., Yan, J.M., and (2012). A theoretical evaluation of possible
Soc. 142, 8142–8146. Zhang, X.B. (2017). Reversible nitrogen transition metal electro-catalysts for N2
fixation based on a rechargeable lithium- reduction. Phys. Chem. Chem. Phys. 14, 1235–
5. Foster, S.L., Bakovic, S.I.P., Duda, R.D., nitrogen battery for energy storage. Chem 2, 1245.
Maheshwari, S., Milton, R.D., Minteer, S.D., 525–532.
Janik, M.J., Renner, J.N., and Greenlee, L.F. 30. Pool, J.A., Lobkovsky, E., and Chirik, P.J.
(2018). Catalysts for nitrogen reduction to 19. Koh, C.S.L., Lee, H.K., Fan Sim, H.Y., Han, X., (2004). Hydrogenation and cleavage of
ammonia. Nat. Catal. 1, 490–500. Phan-Quang, G.C., and Ling, X.Y. (2020). dinitrogen to ammonia with a zirconium
Turning water from a hindrance to the complex. Nature 427, 527–530.
6. Lancaster, K.M., Roemelt, M., Ettenhuber, P., promotor of preferential electrochemical
Hu, Y., Ribbe, M.W., Neese, F., Bergmann, U., 31. Iwamoto, M., Akiyama, M., Aihara, K., and
nitrogen reduction. Chem. Mater. 32, 1674–
and DeBeer, S. (2011). X-ray emission Deguchi, T. (2017). Ammonia synthesis on
1683.
spectroscopy evidences a central carbon in wool-like Au, Pt, Pd, Ag, or Cu electrode
the nitrogenase iron-molybdenum cofactor. 20. Wang, J., Yu, L., Hu, L., Chen, G., Xin, H., and catalysts in nonthermal atmospheric-pressure
Science 334, 974–977. Feng, X. (2018). Ambient ammonia synthesis plasma of N2 and H2. ACS Catal. 7, 6924–
via palladium-catalyzed 6929.
7. Davy, H. (1807). I. The Bakerian Lecture, on electrohydrogenation of dinitrogen at low
some chemical agencies of electricity. Phil. 32. Yang, X., Kattel, S., Nash, J., Chang, X., Lee,
overpotential. Nat. Commun. 9, 1795. J.H., Yan, Y., Chen, J.G., and Xu, B. (2019).
Trans. R. Soc. 97, 1–56.
Quantification of active sites and elucidation
21. Han, Z., Choi, C., Hong, S., Wu, T.S., Soo, Y.L.,
8. Fichter, F., and Suter, R. (1922). Zur Frage der of the reaction mechanism of the
Jung, Y., Qiu, J., and Sun, Z. (2019). Activated
kathodischen Reduktion Des elementaren electrochemical nitrogen reduction reaction
TiO2 with tuned vacancy for efficient
Stickstoffs. Helv. Chim. Acta 5, 246–255. on vanadium nitride. Angew. Chem. Int. Ed.
electrochemical nitrogen reduction. Appl.
Engl. 58, 13768–13772.
9. MacLaughlin, C. (2019). Role for Catal. B 257, 117896.
standardization in electrocatalytic ammonia 33. Abghoui, Y., Garden, A.L., Hlynsson, V.F.,
22. Jaecheol, C., Hoang-Long, D., Manjunath, C.,
synthesis: a conversation with leo liu, Lauren Björgvinsdóttir, S., Ólafsdóttir, H., and
Bryan, H.R., Alexandr, S., and Douglas, M.
Greenlee, and Douglas macfarlane. ACS Skúlason, E. (2015). Enabling electrochemical
(2020). Promoting nitrogen electroreduction
Energy Lett. 4, 1432–1436. reduction of nitrogen to ammonia at ambient
to ammonia with bismuth nanocrystals and
conditions through rational catalyst design.
potassium cations in water. https://chemrxiv.
10. Marnellos, G., and Stoukides, M. (1998). Phys. Chem. Chem. Phys. 17, 4909–4918.
Ammonia synthesis at atmospheric pressure. org/articles/preprint/Promoting_Nitrogen_
Science 282, 98–100. Electroreduction_to_Ammonia_with_ 34. Comer, B.M., Liu, Y.H., Dixit, M.B., Hatzell,
Bismuth_Nanocrystals_and_Potassium_ K.B., Ye, Y., Crumlin, E.J., Hatzell, M.C., and
11. Rod, T.H., Logadottir, A., and Nørskov, J.K. Cations_in_Water/11768814/1. Medford, A.J. (2018). The role of adventitious
(2000). Ammonia synthesis at low carbon in photo-catalytic nitrogen fixation by
temperatures. J. Chem. Phys. 112, 5343–5347. 23. Lazouski, N., Chung, M., Williams, K., Gala, titania. J. Am. Chem. Soc. 140, 15157–15160.
M.L., and Manthiram, K. (2020). Non-aqueous
12. Sun, Z., Ma, T., Tao, H., Fan, Q., and Han, B. gas diffusion electrodes for rapid ammonia 35. Sun, Z., Huo, R., Choi, C., Hong, S., Wu, T.S.,
(2017). Fundamentals and challenges of synthesis from nitrogen and water-splitting- Qiu, J., Yan, C., Han, Z., Liu, Y., Soo, Y.-L., and
electrochemical CO2 reduction using two- derived hydrogen. Nat. Catal. 3, 463–469. Joong, Y. (2019). Oxygen vacancy enables
dimensional materials. Chem 3, 560–587. electrochemical N2 fixation over WO3 with
24. Kitano, M., Inoue, Y., Yamazaki, Y., Hayashi, F., tailored structure. Nano Energy 62, 869–875.
13. Singh, A.R., Rohr, B.A., Statt, M.J., Schwalbe, Kanbara, S., Matsuishi, S., Yokoyama, T., Kim,
J.A., Cargnello, M., and Nørskov, J.K. (2019). S.W., Hara, M., and Hosono, H. (2012). 36. Yu, X., Han, P., Wei, Z., Huang, L., Gu, Z.,
Strategies toward selective electrochemical Ammonia synthesis using a stable electride as Peng, S., Ma, J., and Zheng, G. (2018). Boron-
ammonia synthesis. ACS Catal. 9, 8316–8324. an electron donor and reversible hydrogen doped graphene for electrocatalytic N2
store. Nat. Chem. 4, 934–940. reduction. Joule 2, 1610–1622.
14. Singh, A.R., Rohr, B.A., Schwalbe, J.A.,
Cargnello, M., Chan, K., Jaramillo, T.F., 25. Cui, X., Tang, C., and Zhang, Q. (2018). A 37. Medford, A.J., Vojvodic, A., Hummelshøj, J.S.,
Chorkendorff, I., and Nørskov, J.K. (2017). review of electrocatalytic reduction of Voss, J., Abild-Pedersen, F., Studt, F.,
Electrochemical ammonia synthesis—the dinitrogen to ammonia under ambient Bligaard, T., Nilsson, A., and Nørskov, J.K.
selectivity challenge. ACS Catal. 7, 706–709. conditions. Adv. Energy Mater. 8, 1800369. (2015). From the Sabatier principle to a

1750 Chem 7, 1708–1754, July 8, 2021

 ll
Review

predictive theory of transition-metal nitrite and ammonia ions) in environmental 65. Battino, R., Rettich, T.R., and Tominaga, T.
heterogeneous catalysis. J. Catal. 328, 36–42. samples by ion chromatography. Pol. J. (1984). The solubility of nitrogen and air in
Environ. Stud. 15, 5–18. liquids. J. Phys. Chem. Ref. Data 13, 563–600.
38. Montoya, J.H., Tsai, C., Vojvodic, A., and
Nørskov, J.K. (2015). The challenge of 52. LeDuy, A., and Samson, R. (1982). Testing of 66. Kang, C.S.M., Zhang, X., and MacFarlane,
electrochemical ammonia synthesis: a new an ammonia ion selective electrode for D.R. (2019). High nitrogen gas solubility and
perspective on the role of nitrogen scaling ammonia nitrogen measurement in the physicochemical properties of [C4mpyr]
relations. ChemSusChem 8, 2180–2186. methanogenic sludge. Biotechnol. Lett. 4, [eFAP]–fluorinated solvent mixtures. J. Phys.
303–306. Chem. C 123, 21376–21385.
39. Suryanto, B.H.R., Du, H.L., Wang, D., Chen, J.,
Simonov, A.N., and MacFarlane, D.R. (2019). 67. Tsuneto, A., Kudo, A., and Sakata, T. (1994).
53. Cohn, V.H., and Lyle, J. (1966). A fluorometric
Challenges and prospects in the catalysis of Lithium-mediated electrochemical reduction
assay for glutathione. Anal. Biochem. 14,
electroreduction of nitrogen to ammonia. of high pressure N2 to NH3. J. Electroanal.
434–440.
Nat. Catal. 2, 290–296. Chem. 367, 183–188.

40. Halmann, M. (1984). Electrochemical 54. Liu, C., Sakimoto, K.K., Colón, B.C., Silver, 68. Köleli, F., and Röpke, T. (2006).
reduction of molecular nitrogen to ammonia P.A., and Nocera, D.G. (2017). Ambient Electrochemical hydrogenation of dinitrogen
in aqueous alakli: a re-examination. nitrogen reduction cycle using a hybrid to ammonia on a polyaniline electrode. Appl.
J. Electroanal. Chem. Interfacial Electrochem. inorganic-biological system. Proc. Natl. Acad. Catal. B 62, 306–310.
181, 307–308. Sci. USA 114, 6450–6455.
69. Kim, K., Lee, N., Yoo, C.-Y., Kim, J.-N., Yoon,
41. Tang, C., and Qiao, S.Z. (2019). How to 55. Hodgetts, R.Y., Kiryutin, A.S., Nichols, P., Du, H.C., and Han, J.-I. (2016). Communication
explore ambient electrocatalytic nitrogen H.L., Bakker, J.M., Macfarlane, D.R., and —electrochemical reduction of nitrogen to
reduction reliably and insightfully. Chem. Soc. Simonov, A.N. (2020). Refining universal ammonia in 2-propanol under ambient
Rev. 48, 3166–3180. procedures for ammonium quantification via temperature and pressure. J. Electrochem.
rapid 1H NMR analysis for dinitrogen Soc. 163, F610–F612.
42. Chen, Y., Liu, H., Ha, N., Licht, S., Gu, S., and reduction studies. ACS Energy Lett. 5,
Li, W. (2020). Revealing nitrogen-containing 736–741. 70. Kim, K., Yoo, C.Y., Kim, J.N., Yoon, H.C., and
species in commercial catalysts used for Han, J.I. (2016). Electrochemical synthesis of
ammonia electrosynthesis. Nat. Catal. 3, 56. Yu, W., Lewis, N.S., Gray, H.B., and Dalleska, ammonia from water and nitrogen in
1055–1061. N.F. (2020). Isotopically selective ethylenediamine under ambient temperature
quantification by UPLC-MS of aqueous and pressure. J. Electrochem. Soc. 163,
43. Chen, G.F., Ren, S., Zhang, L., Cheng, H., Luo, ammonia at submicromolar concentrations F1523–F1526.
Y., Zhu, K., Ding, L.X., and Wang, H. (2019). using dansyl chloride derivatization. ACS
Advances in electrocatalytic N2 reduction— Energy Lett. 5, 1532–1536. 71. Lee, H.K., Koh, C.S.L., Lee, Y.H., Liu, C.,
strategies to tackle the selectivity challenge. Phang, I.Y., Han, X., Tsung, C.K., and Ling,
Small Methods 3, 1800337. 57. Yao, Y., Wang, H., Yuan, X.Z., Li, H., and Shao, X.Y. (2018). Favoring the unfavored: selective
M. (2019). Electrochemical nitrogen reduction electrochemical nitrogen fixation using a
44. Choi, J., Du, H.L., Nguyen, C.K., Suryanto,
reaction on ruthenium. ACS Energy Lett. 4, reticular chemistry approach. Sci. Adv. 4,
B.H.R., Simonov, A.N., and MacFarlane, D.R. eaar3208.
1336–1341.
(2020). Electroreduction of nitrates, nitrites,
and gaseous nitrogen oxides: a potential 72. Lazouski, N., Schiffer, Z.J., Williams, K., and
58. Yao, Y., Zhu, S., Wang, H., Li, H., and Shao, M. Manthiram, K. (2019). Understanding
source of ammonia in dinitrogen reduction
(2020). A spectroscopic study of
studies. ACS Energy Lett. 5, 2095–2097. continuous lithium-mediated electrochemical
electrochemical nitrogen and nitrate nitrogen reduction. Joule 3, 1127–1139.
45. Dabundo, R., Lehmann, M.F., Treibergs, L., reduction on rhodium surfaces. Angew.
Tobias, C.R., Altabet, M.A., Moisander, P.H., Chem. Int. Ed. Engl. 59, 10479–10483. 73. Kang, C.S.M., Zhang, X., and MacFarlane,
and Granger, J. (2014). The contamination of D.R. (2018). Synthesis and physicochemical
commercial 15N2 gas stocks with 15N-labeled 59. Watt, G.W., and Chrisp, J.D. (1952). properties of fluorinated ionic liquids with
nitrate and ammonium and consequences for Spectrophotometric method for high nitrogen gas solubility. J. Phys. Chem. C
nitrogen fixation measurements. PLoS One 9, determination of hydrazine. Anal. Chem. 24, 122, 24550–24558.
e110335. 2006–2008.
74. Suryanto, B.H.R., Kang, C.S.M., Wang, D.,
46. Shan, W., Liu, F., He, H., Shi, X., and Zhang, C. 60. Giddey, S., Badwal, S.P.S., and Kulkarni, A. Xiao, C., Zhou, F., Azofra, L.M., Cavallo, L.,
(2011). The remarkable improvement of a Ce– (2013). Review of electrochemical ammonia Zhang, X., and MacFarlane, D.R. (2018).
Ti based catalyst for NO x abatement, production technologies and materials. Int. J. Rational electrode–electrolyte design for
prepared by a homogeneous precipitation Hydrogen Energy 38, 14576–14594. efficient ammonia electrosynthesis under
method. ChemCatChem 3, 1286–1289. ambient conditions. ACS Energy Lett. 3,
61. Han, L., Liu, X., Chen, J., Lin, R., Liu, H., Lü, F., 1219–1224.
47. Wang, Z., Wang, Z., Ye, Y., Chen, N., and Li, H. Bak, S., Liang, Z., Zhao, S., Stavitski, E., et al.
(2016). Study on the removal of nitric oxide (2019). Atomically dispersed molybdenum 75. Yan, R., Antonietti, M., and Oschatz, M. (2018).
(NO) by dual oxidant (H2O2/S2O82 ) system. catalysts for efficient ambient nitrogen Toward the experimental understanding of
Chem. Eng. Sci. 145, 133–140. fixation. Angew. Chem. Int. Ed. Engl. 58, the energy storage mechanism and ion
2321–2325. dynamics in ionic liquid based
48. Ivancic, I., and Degobbis, D. (1984). An supercapacitors. Adv. Energy Mater. 8,
optimal manual procedure for ammonia 62. Chen, G.F., Cao, X., Wu, S., Zeng, X., Ding, 1800026.
analysis in natural waters by the indophenol L.X., Zhu, M., and Wang, H. (2017). Ammonia
blue method. Water Res. 18, 1143–1147. 76. Ortuño, M.A., Hollóczki, O., Kirchner, B., and
electrosynthesis with high selectivity under
López, N. (2019). Selective electrochemical
49. Yuen, S.H., and Pollard, A.G. (1954). ambient conditions via a Li+ incorporation
nitrogen reduction driven by hydrogen bond
Determination of nitrogen in agricultural strategy. J. Am. Chem. Soc. 139, 9771–9774.
interactions at metal–ionic liquid interfaces.
materials by the nessler reagent. II.—micro- J. Phys. Chem. Lett. 10, 513–517.
determinations in plant tissue and in soil 63. Köleli, F., and Kayan, D.B. (2010). Low
extracts. J. Sci. Food Agric. 5, 364–369. overpotential reduction of dinitrogen to 77. Li, X.F., Li, Q.K., Cheng, J., Liu, L., Yan, Q., Wu,
ammonia in aqueous media. J. Electroanal. Y., Zhang, X.H., Wang, Z.Y., Qiu, Q., and Luo,
50. Bower, C.E., and Holm-Hansen, T. (1980). A Chem. 638, 119–122. Y. (2016). Conversion of dinitrogen to
salicylate–hypochlorite method for ammonia by FeN3-embedded graphene.
determining ammonia in seawater. Can. J. 64. Zhang, Q., Liu, B., Yu, L., Bei, Y., and Tang, B. J. Am. Chem. Soc. 138, 8706–8709.
Fish. Aquat. Sci. 37, 794–798. (2020). Synergistic promotion of the
electrochemical reduction of nitrogen to 78. Einsle, O., Tezcan, F.A., Andrade, S.L.A.,
51. Michalski, R., and Kurzyca, I. (2006). ammonia by phosphorus and potassium. Schmid, B., Yoshida, M., Howard, J.B., and
Determination of nitrogen species (nitrate, ChemCatChem 12, 334–341. Rees, D.C. (2002). Nitrogenase MoFe-protein

Chem 7, 1708–1754, July 8, 2021 1751

 ll
Review

at 1.16 Å resolution: a central ligand in the under ambient conditions. Angew. Chem. Int. 104. Wang, S., Li, B., Li, L., Tian, Z., Zhang, Q.,
FeMo-cofactor. Science 297, 1696–1700. Ed. Engl. 59, 3511–3516. Chen, L., and Zeng, X.C. (2020). Highly
efficient N2 fixation catalysts: transition-metal
79. Yandulov, D.V., and Schrock, R.R. (2003). 92. Valov, I., Luerssen, B., Mutoro, E., Gregoratti, carbides M2C (MXenes). Nanoscale 12,
Catalytic reduction of dinitrogen to ammonia L., De Souza, R.A., Bredow, T., Günther, S., 538–547.
at a single molybdenum center. Science 301, Barinov, A., Dudin, P., Martin, M., and Janek,
76–78. J. (2011). Electrochemical activation of 105. Matanovic, I., and Garzon, F.H. (2018).
molecular nitrogen at the Ir/YSZ interface. Nitrogen electroreduction and hydrogen
80. Back, S., and Jung, Y. (2016). On the Phys. Chem. Chem. Phys. 13, 3394–3410. evolution on cubic molybdenum carbide: a
mechanism of electrochemical ammonia density functional study. Phys. Chem. Chem.
synthesis on the Ru catalyst. Phys. Chem. 93. Xue, Z.H., Zhang, S.N., Lin, Y.X., Su, H., Zhai, Phys. 20, 14679–14687.
Chem. Phys. 18, 9161–9166. G.Y., Han, J.T., Yu, Q.Y., Li, X.H., Antonietti,
M., and Chen, J.S. (2019). Electrochemical 106. Zhang, C., Wang, D., Wan, Y., Lv, R., Li, S., Li,
81. Ling, C., Zhang, Y., Li, Q., Bai, X., Shi, L., and reduction of N2 into NH3 by donor-acceptor B., Zou, X., and Yang, S. (2020). Vanadium
Wang, J. (2019). New mechanism for N2 couples of Ni and Au nanoparticles with a carbide with periodic anionic vacancies for
reduction: the essential role of surface 67.8% faradaic efficiency. J. Am. Chem. Soc. effective electrocatalytic nitrogen reduction.
hydrogenation. J. Am. Chem. Soc. 141, 141, 14976–14980. Mater. Today 40, 18–25.
18264–18270.
94. Liu, Y., Huang, L., Zhu, X., Fang, Y., and Dong, 107. Li, X., Li, T., Ma, Y., Wei, Q., Qiu, W., Guo, H.,
82. Yang, X., Nash, J., Anibal, J., Dunwell, M., S. (2020). Coupling Cu with Au for enhanced Shi, X., Zhang, P., Asiri, A.M., Chen, L., et al.
Kattel, S., Stavitski, E., Attenkofer, K., Chen, electrocatalytic activity of nitrogen reduction (2018). Boosted electrocatalytic N2 reduction
J.G., Yan, Y., and Xu, B. (2018). Mechanistic reaction. Nanoscale 12, 1811–1816. to NH3 by defect-rich MoS2 nanoflower. Adv.
insights into electrochemical nitrogen Energy Mater. 8, 1801357.
reduction reaction on vanadium nitride 95. Zhang, N., Li, L., Wang, J., Hu, Z., Shao, Q.,
nanoparticles. J. Am. Chem. Soc. 140, 13387– Xiao, X., and Huang, X. (2020). Surface- 108. Liu, Y., Han, M., Xiong, Q., Zhang, S., Zhao, C.,
13391. regulated rhodium–antimony nanorods for Gong, W., Wang, G., Zhang, H., and Zhao, H.
nitrogen fixation. Angew. Chem. Int. Ed. Engl. (2019). Dramatically enhanced ambient
83. Yao, Y., Zhu, S., Wang, H., Li, H., and Shao, M. 59, 8066–8071. ammonia electrosynthesis performance by in-
(2018). A spectroscopic study on the nitrogen operando created Li–S interactions on MoS2
96. Sim, H.Y.F., Chen, J.R.T., Koh, C.S.L., Lee, electrocatalyst. Adv. Energy Mater. 9,
electrochemical reduction reaction on gold
H.K., Han, X., Phan-Quang, G.C., Pang, J.Y., 1803935.
and platinum surfaces. J. Am. Chem. Soc. 140,
Lay, C.L., Pedireddy, S., Phang, I.Y., et al.
1496–1501.
(2020). ZIF-induced d-band modification in a 109. Wang, H.-B., Wang, J.Q., Zhang, R., Cheng,
84. Zeng, L., Li, X., Chen, S., Wen, J., Huang, W., bimetallic nanocatalyst: achieving over 44 % C.-Q., Qiu, K.-W., Yang, Y.-J., Mao, J., Liu, H.,
and Chen, A. (2020). Unique hollow Ni– efficiency in the ambient nitrogen reduction Du, M., Dong, C.-K., et al. (2020). Bionic
Fe@MoS2 nanocubes with boosted reaction. Angew. Chem. Int. Ed. Engl. 59, design of a Mo(IV)-doped FeS2 catalyst for
electrocatalytic activity for N2 reduction to 16997–17003. electroreduction of dinitrogen to ammonia.
NH3. J. Mater. Chem. A 8, 7339–7349. ACS Catal. 10, 4914–4921.
97. Wang, Y., Shi, M.-M., Bao, D., Meng, F.-L.,
85. Lai, F., Chen, N., Ye, X., He, G., Zong, W., Holt, Zhang, Q., Zhou, Y.T., Liu, K.-H., Zhang, Y., 110. Lai, F., Zong, W., He, G., Xu, Y., Huang, H.,
K.B., Pan, B., Parkin, I.P., Liu, T., and Chen, R. Wang, J.-Z., Chen, Z.-W., et al. (2019). Weng, B., Rao, D., Martens, J.A., Hofkens, J.,
(2020). Refining energy levels in ReS2 Generating defect-rich bismuth for enhancing Parkin, I.P., and Liu, T. (2020). N2
nanosheets by low-valent transition-metal the rate of nitrogen electroreduction to electroreduction to NH3 by selenium vacancy-
doping for dual-boosted electrochemical ammonia. Angew. Chem. Int. Ed. Engl. 58, rich ReSe2 catalysis at an abrupt interface.
ammonia/hydrogen production. Adv. Funct. 9464–9469. Angew. Chem. Int. Ed. Engl. 59, 13320–13327.
Mater. 30, 1907376. 98. Fang, Y., Liu, Z., Han, J., Jin, Z., Han, Y., Wang, 111. Wang, Y., Chen, A., Lai, S., Peng, X., Zhao, S.,
F., Niu, Y., Wu, Y., and Xu, Y. (2019). High- Hu, G., Qiu, Y., Ren, J., Liu, X., and Luo, J.
86. Lv, X.W., Liu, Y., Hao, R., Tian, W., and Yuan,
performance electrocatalytic conversion of N2 (2020). Self-supported NbSe2 nanosheet
Z.Y. (2020). Urchin-like Al-doped Co3O4
to NH3 using oxygen-vacancy-rich TiO2 in situ arrays for highly efficient ammonia
nanospheres rich in surface oxygen vacancies
grown on Ti3C2Tx MXene. Adv. Energy Mater. electrosynthesis under ambient conditions.
enable efficient ammonia electrosynthesis.
9, 1803406. J. Catal. 381, 78–83.
ACS Appl. Mater. Interfaces 12, 17502–17508.
99. Huang, L., Wu, J., Han, P., Al-Enizi, A.M., 112. Sun, Z., Talreja, N., Tao, H., Texter, J., Muhler,
87. Ren, J.T., Wan, C.Y., Pei, T.Y., Lv, X.W., and Almutairi, T.M., Zhang, L., and Zheng, G. M., Strunk, J., and Chen, J. (2018). Catalysis of
Yuan, Z.Y. (2020). Promotion of (2019). NbO2 electrocatalyst toward 32% carbon dioxide photoreduction on
electrocatalytic nitrogen reduction reaction faradaic efficiency for N2 fixation. Small nanosheets: fundamentals and challenges.
on N-doped porous carbon with secondary Methods 3, 1800386. Angew. Chem. Int. Ed. Engl. 57, 7610–7627.
heteroatoms. Appl. Catal. B 266, 118633.
100. Abghoui, Y., and Skúlason, E. (2017). 113. Lv, C., Zhong, L., Yao, Y., Liu, D., Kong, Y., Jin,
88. Yuan, L.P., Wu, Z.Y., Jiang, W.J., Tang, T., Niu, Electrochemical synthesis of ammonia via X., Fang, Z., Xu, W., Yan, C., Dinh, K.N., et al.
S., and Hu, J.S. (2020). Phosphorus-doping Mars-van Krevelen mechanism on the (111) (2020). Boosting electrocatalytic ammonia
activates carbon nanotubes for efficient facets of group III–VII transition metal production through mimicking ‘‘p back-
electroreduction of nitrogen to ammonia. mononitrides. Catal. Today 286, 78–84. donation’’. Chem 6, 2690–2702.
Nano Res. 13, 1376–1382.
101. Jin, H., Li, L., Liu, X., Tang, C., Xu, W., Chen, S., 114. Lv, C., Yan, C., Chen, G., Ding, Y., Sun, J.,
89. Fu, Y., Li, T., Zhou, G., Guo, J., Ao, Y., Hu, Y., Song, L., Zheng, Y., and Qiao, S.Z. (2019). Zhou, Y., and Yu, G. (2018). An amorphous
Shen, J., Liu, L., and Wu, X. (2020). Dual-metal- Nitrogen vacancies on 2D layered W2N3: a noble-metal-free electrocatalyst that enables
driven selective pathway of nitrogen stable and efficient active site for nitrogen nitrogen fixation under ambient conditions.
reduction in orderly atomic-hybridized reduction reaction. Adv. Mater. 31, 1902709. Angew. Chem. Int. Ed. Engl. 57, 6073–6076.
Re2MnS6 ultrathin nanosheets. Nano Lett. 20,
4960–4967. 102. Hu, B., Hu, M., Seefeldt, L., and Liu, T.L. (2019). 115. Liu, X., Jang, H., Li, P., Wang, J., Qin, Q., Kim,
Electrochemical dinitrogen reduction to M.G., Li, G., and Cho, J. (2019). Antimony-
90. Yao, D., Tang, C., Li, L., Xia, B., Vasileff, A., Jin, ammonia by Mo2N: catalysis or based composites loaded on phosphorus-
H., Zhang, Y., and Qiao, S.Z. (2020). In situ decomposition? ACS Energy Lett. 4, 1053– doped carbon for boosting faradaic efficiency
fragmented bismuth nanoparticles for 1054. of the electrochemical nitrogen reduction
electrocatalytic nitrogen reduction. Adv. reaction. Angew. Chem. Int. Ed. Engl. 58,
Energy Mater. 10, 2001289. 103. Peng, J., Chen, X., Ong, W.J., Zhao, X., and Li, 13329–13334.
N. (2019). Surface and heterointerface
91. Xu, W., Fan, G., Chen, J., Li, J., Zhang, L., Zhu, engineering of 2D MXenes and their 116. Fang, Y., Xue, Y., Li, Y., Yu, H., Hui, L., Liu, Y.,
S., et al. (2020). Nanoporous palladium nanocomposites: insights into electro- and Xing, C., Zhang, C., Zhang, D., Wang, Z., et al.
hydride for electrocatalytic N2 reduction photocatalysis. Chem 5, 18–50. (2020). Graphdiyne interface engineering:

1752 Chem 7, 1708–1754, July 8, 2021

 ll
Review

highly active and selective ammonia 129. Zhang, L., Ding, L.X., Chen, G.F., Yang, X., and material properties. Phys. Rev. Lett. 120,
synthesis. Angew. Chem. Int. Ed. Engl. 59, Wang, H. (2019). Ammonia synthesis under 145301.
13021–13027. ambient conditions: selective
electroreduction of dinitrogen to ammonia on 143. Logan, D.L. (2011). A First Course in the Finite
117. Choi, C., Back, S., Kim, N.Y., Lim, J., Kim, Y.H., black phosphorus nanosheets. Angew. Chem. Element Method (Cengage Learning).
and Jung, Y. (2018). Suppression of hydrogen Int. Ed. Engl. 58, 2612–2616.
evolution reaction in electrochemical N2 144. Goodfellow, I., Bengio, Y., and Courville, A.
reduction using single-atom catalysts: a 130. Xu, G., Li, H., Bati, A.S.R., Bat-Erdene, M., (2016). Deep Learning (MIT Press).
computational guideline. ACS Catal. 8, 7517– Nine, M.J., Losic, D., Chen, Y., Shapter, J.G.,
7525. Batmunkh, M., and Ma, T. (2020). Nitrogen- 145. Wang, D., Azofra, L.M., Harb, M., Cavallo, L.,
doped phosphorene for electrocatalytic Zhang, X., Suryanto, B.H.R., and MacFarlane,
118. Tao, H., Choi, C., Ding, L.X., Jiang, Z., Han, Z., ammonia synthesis. J. Mater. Chem. A 8, D.R. (2018). Energy-efficient nitrogen
Jia, M., Fan, Q., Gao, Y., Wang, H., Robertson, 15875–15883. reduction to ammonia at low overpotential in
A.W., et al. (2019). Nitrogen fixation by Ru aqueous electrolyte under ambient
single-atom electrocatalytic reduction. Chem 131. Zhang, Y., Dong, N., Tao, H., Yan, C., Huang, conditions. ChemSusChem 11, 3416–3422.
5, 204–214. J., Liu, T., Robertson, A.W., Texter, J., Wang,
J., and Sun, Z. (2017). Exfoliation of stable 2D 146. Bao, D., Zhang, Q., Meng, F.L., Zhong, H.X.,
119. Lü, F., Zhao, S., Guo, R., He, J., Peng, X., Bao, black phosphorus for device fabrication. Shi, M.M., Zhang, Y., Yan, J.M., Jiang, Q., and
H., Fu, J., Han, L., Qi, G., Luo, J., et al. (2019). Chem. Mater. 29, 6445–6456. Zhang, X.B. (2017). Electrochemical reduction
Nitrogen-coordinated single Fe sites for of N2 under ambient conditions for artificial
efficient electrocatalytic N2 fixation in neutral 132. Liu, Q., Zhang, X., Wang, J., Zhang, Y., Bian, N2 fixation and renewable energy storage
media. Nano Energy 61, 420–427. S., Cheng, Z., Kang, N., Huang, H., Gu, S., using N2/NH3 cycle. Adv. Mater. 29, 1604799.
Wang, Y., et al. (2020). Crystalline red
120. Wang, Y., Cui, X., Zhao, J., Jia, G., Gu, L., 147. Choi, J., Suryanto, B.H.R., Wang, D., Du, H.L.,
phosphorus nanoribbons: large-scale
Zhang, Q., Meng, L., Shi, Z., Zheng, L., Wang, Hodgetts, R.Y., Ferrero Vallana, F.M.,
synthesis and electrochemical nitrogen
C., et al. (2019). Rational design of Fe–N/C MacFarlane, D.R., and Simonov, A.N. (2020).
fixation. Angew. Chem. Int. Ed. Engl. 59,
hybrid for enhanced nitrogen reduction Identification and elimination of false
14383–14387.
electrocatalysis under ambient conditions in positives in electrochemical nitrogen
aqueous solution. ACS Catal. 9, 336–344. 133. Qiu, W., Xie, X.Y., Qiu, J., Fang, W.H., Liang, reduction studies. Nat. Commun. 11, 5546.
121. Wang, M., Liu, S., Qian, T., Liu, J., Zhou, J., Ji, R., Ren, X., Ji, X., Cui, G., Asiri, A.M., Cui, G.,
et al. (2018). High-performance artificial 148. Tayyebi, E., Abghoui, Y., and Skúlason, E.
H., Xiong, J., Zhong, J., and Yan, C. (2019). (2019). Elucidating the mechanism of
Over 56.55% faradaic efficiency of ambient nitrogen fixation at ambient conditions using
a metal-free electrocatalyst. Nat. Commun. 9, electrochemical N2 reduction at the Ru(0001)
ammonia synthesis enabled by positively electrode. ACS Catal. 9, 11137–11145.
shifting the reaction potential. Nat. Commun. 3485.
10, 341. 149. Chen, L.Y., Kuo, T.C., Hong, Z.S., Cheng, M.J.,
134. Liu, C., Li, Q., Wu, C., Zhang, J., Jin, Y.,
MacFarlane, D.R., and Sun, C. (2019). Single- and Goddard, W.A. (2019). Mechanism and
122. Zhang, S., Jin, M., Shi, T., Han, M., Sun, Q., Lin, kinetics for both thermal and electrochemical
Y., Ding, Z., Zheng, L.R., Wang, G., Zhang, Y., boron catalysts for nitrogen reduction
reaction. J. Am. Chem. Soc. 141, 2884–2888. reduction of N2 catalysed by Ru(0001) based
et al. (2020). Electrocatalytically active Fe-(O- on quantum mechanics. Phys. Chem. Chem.
C2)4 single-atom sites for efficient reduction of Phys. 21, 17605–17612.
nitrogen to ammonia. Angew. Chem. Int. Ed. 135. Liu, X., Jiao, Y., Zheng, Y., and Qiao, S.Z.
Engl. 59, 13423–13429. (2020). Isolated boron sites for
electroreduction of dinitrogen to ammonia. 150. Zhang, L., Mallikarjun Sharada, S., Singh, A.R.,
ACS Catal. 10, 1847–1854. Rohr, B.A., Su, Y., Qiao, L., and Nørskov, J.K.
123. Wang, X., Wang, W., Qiao, M., Wu, G., Chen,
(2018). A theoretical study of the effect of a
W., Yuan, T., Xu, Q., Chen, M., Zhang, Y.,
136. Abghoui, Y., Sigtryggsson, S.B., and Skúlason, non-aqueous proton donor on
Wang, X., et al. (2018). Atomically dispersed
E. (2019). Biomimetic nitrogen fixation electrochemical ammonia synthesis. Phys.
Au1 catalyst towards efficient electrochemical
catalyzed by transition metal sulfide surfaces Chem. Chem. Phys. 20, 4982–4989.
synthesis of ammonia. Sci. Bull. 63, 1246–
1253. in an electrolytic cell. ChemSusChem 12,
4265–4273. 151. Ulissi, Z.W., Medford, A.J., Bligaard, T., and
124. Chen, H., Zhu, X., Huang, H., Wang, H., Wang, Nørskov, J.K. (2017). To address surface
T., Zhao, R., Zheng, H., Asiri, A.M., Luo, Y., and 137. Gu, G.H., Noh, J., Kim, I., and Jung, Y. (2019). reaction network complexity using scaling
Sun, X. (2019). Sulfur dots–graphene Machine learning for renewable energy relations machine learning and DFT
nanohybrid: a metal-free electrocatalyst for materials. J. Mater. Chem. A 7, 17096–17117. calculations. Nat. Commun. 8, 14621.
efficient N2-to-NH3 fixation under ambient
138. Kim, M., Yeo, B.C., Park, Y., Lee, H.M., Han, 152. Jang, J., Gu, G.H., Noh, J., Kim, J., and Jung,
conditions. Chem. Commun. 55, 3152–3155.
S.S., and Kim, D. (2020). Artificial intelligence Y. (2020). Structure-based synthesizability
125. Liu, B., Zheng, Y., Peng, H.Q., Ji, B., Yang, Y., to accelerate the discovery of N2 prediction of crystals using partially
Tang, Y., Lee, C.S., and Zhang, W. (2020). electroreduction catalysts. Chem. Mater. 32, supervised learning. J. Am. Chem. Soc. 142,
Nanostructured and boron-doped diamond 709–720. 18836–18843.
as an electrocatalyst for nitrogen fixation. ACS
Energy Lett. 5, 2590–2596. 139. Zafari, M., Kumar, D., Umer, M., and Kim, K.S. 153. Yang, C., Huang, B., Bai, S., Feng, Y., Shao, Q.,
(2020). Machine learning-based high and Huang, X. (2020). A generalized surface
126. Fan, Q., Choi, C., Yan, C., Liu, Y., Qiu, J., throughput screening for nitrogen fixation on chalcogenation strategy for boosting the
Hong, S., Jung, Y., and Sun, Z. (2019). High- boron-doped single atom catalysts. J. Mater. electrochemical N2 fixation of metal
yield production of few-layer boron Chem. A 8, 5209–5216. nanocrystals. Adv. Mater. 32, e2001267.
nanosheets for efficient electrocatalytic N2
reduction. Chem. Commun. 55, 4246–4249. 140. Hoar, B.B., Lu, S., and Liu, C. (2020). Machine- 154. Zhang, M., Choi, C., Huo, R., Gu, G.H., Hong,
learning-enabled exploration of morphology S., Yan, C., Xu, S., Robertson, A.W., Qiu, J.,
127. Légaré, M.A., Bélanger-Chabot, G., influence on wire-array electrodes for Jung, Y., and Sun, Z. (2020). Reduced
Dewhurst, R.D., Welz, E., Krummenacher, I., electrochemical nitrogen fixation. J. Phys. graphene oxides with engineered defects
Engels, B., and Braunschweig, H. (2018). Chem. Lett. 11, 4625–4630. enable efficient electrochemical reduction of
Nitrogen fixation and reduction at boron. dinitrogen to ammonia in wide pH range.
Science 359, 896–900. 141. Noh, J., Gu, G.H., Kim, S., and Jung, Y. (2020). Nano Energy 68, 104323.
Machine-enabled inverse design of inorganic
128. Lv, C., Qian, Y., Yan, C., Ding, Y., Liu, Y., Chen, solid materials: promises and challenges. 155. Jin, Z., Liu, C., Liu, Z., Han, J., Fang, Y., Han, Y.,
G., and Yu, G. (2018). Defect engineering Chem. Sci. 11, 4871–4881. Niu, Y., Wu, Y., Sun, C., and Xu, Y. (2020).
metal-free polymeric carbon nitride Rational design of hydroxyl-rich Ti3C2Tx
electrocatalyst for effective nitrogen fixation 142. Xie, T., and Grossman, J.C. (2018). Crystal MXene quantum dots for high-performance
under ambient conditions. Angew. Chem. Int. graph convolutional neural networks for an electrochemical N2 reduction. Adv. Energy
Ed. Engl. 57, 10246–10250. accurate and interpretable prediction of Mater. 10, 2000797.

Chem 7, 1708–1754, July 8, 2021 1753

 ll
Review

156. Xia, J., Yang, S.Z., Wang, B., Wu, P., Popovs, I., electrocatalytic nitrogen fixation. J. Mater. on carbon fabric for electrochemical N2
Li, H., Irle, S., Dai, S., and Zhu, H. (2020). Chem. A 8, 5865–5873. fixation. ACS Sustain. Chem. Eng. 7, 8853–
Boosting electrosynthesis of ammonia on 8859.
surface-engineered MXene Ti3C2. Nano 163. Shi, M.M., Bao, D., Li, S.J., Wulan, B.R., Yan,
Energy 72, 104681. J.M., and Jiang, Q. (2018). Anchoring PdCu 170. Yang, X., Ling, F., Su, J., Zi, X., Zhang, H.,
amorphous nanocluster on graphene for Zhang, H., Li, J., Zhou, M., and Wang, Y.
157. Ahmed, M.I., Liu, C., Zhao, Y., Ren, W., Chen, electrochemical reduction of N2 to NH3 under (2020). Insights into the role of cation vacancy
X., Chen, S., and Zhao, C. (2020). Metal–sulfur ambient conditions in aqueous solution. Adv. for significantly enhanced electrochemical
linkages achieved by organic tethering of Energy Mater. 8, 1800124. nitrogen reduction. Appl. Catal. B 264,
ruthenium nanocrystals for enhanced 118477.
electrochemical nitrogen reduction. Angew. 164. Yang, D., Chen, T., and Wang, Z. (2017).
Chem. Int. Ed. Engl. 59, 21465–21469. Electrochemical reduction of aqueous
nitrogen (N2) at a low overpotential on (110)- 171. Chu, K., Li, Q.Q., Liu, Y.P., Wang, J., and
158. Zou, H., Rong, W., Long, B., Ji, Y., and Duan, L. oriented Mo nanofilm. J. Mater. Chem. A 5, Cheng, Y.H. (2020). Filling the nitrogen
(2019). Corrosion-induced Cl-doped ultrathin 18967–18971. vacancies with sulphur dopants in graphitic
graphdiyne toward electrocatalytic nitrogen C3N4 for efficient and robust electrocatalytic
reduction at ambient conditions. ACS Catal. 165. Nazemi, M., Panikkanvalappil, S.R., and El- nitrogen reduction. Appl. Catal. B 267,
9, 10649–10655. Sayed, M.A. (2018). Enhancing the rate of 118693.
electrochemical nitrogen reduction reaction
159. Yang, Y., Zhang, L., Hu, Z., Zheng, Y., Tang, C., for ammonia synthesis under ambient 172. Tong, W., Huang, B., Wang, P., Li, L., Shao, Q.,
Chen, P., Wang, R., Qiu, K., Mao, J., Ling, T., conditions using hollow gold nanocages. and Huang, X. (2020). Crystal-phase-
and Qiao, S.-Z. (2020). The crucial role of Nano Energy 49, 316–323. engineered PdCu electrocatalyst for
charge accumulation and spin polarization in enhanced ammonia synthesis. Angew. Chem.
activating carbon-based catalysts for 166. Huang, L., Gu, X., and Zheng, G. (2019). Int. Ed. Engl. 59, 2649–2653.
electrocatalytic nitrogen reduction. Angew. Tuning active sites of MXene for efficient
Chem. Int. Ed. Engl. 59, 4525–4531. electrocatalytic N2 fixation. Chem 5, 15–17. 173. Li, J., Chen, S., Quan, F., Zhan, G., Jia,
F., Ai, Z., and Zhang, L. (2020).
160. Yang, M., Huo, R., Shen, H., Xia, Q., Qiu, J., 167. Zhang, L., Ji, X., Ren, X., Ma, Y., Shi, X., Tian, Z., Accelerated dinitrogen electroreduction
Robertson, A.W., Li, X., and Sun, Z. (2020). Asiri, A.M., Chen, L., Tang, B., and Sun, X. to ammonia via interfacial polarization
Metal-tuned W18O49 for efficient (2018). Electrochemical ammonia synthesis via triggered by single-atom protrusions.
electrocatalytic N2 reduction. ACS nitrogen reduction reaction on a MoS2 Chem 6, 885–901.
Sustainable Chem. Eng. 8, 2957–2963. catalyst: theoretical and experimental studies.
Adv. Mater. 30, e1800191.
161. Wu, T., Zhu, X., Xing, Z., Mou, S., Li, C., Qiao, 174. Zheng, J., Lyu, Y., Qiao, M., Veder, J.P.,
Y., Liu, Q., Luo, Y., Shi, X., Zhang, Y., and Sun, 168. Guo, X., and Huang, S. (2018). Tuning Marco, R.D., Bradley, J., Wang, R., Li, Y.,
X. (2019). Greatly improving electrochemical nitrogen reduction reaction activity via Huang, A., Jiang, S.P., and Wang, S. (2019).
N2 reduction over TiO2 nanoparticles by iron controllable Fe magnetic moment: A Tuning the electron localization of gold
doping. Angew. Chem. Int. Ed. Engl. 58, computational study of single Fe atom enables the control of nitrogen-to-ammonia
18449–18453. supported on defective graphene. fixation. Angew. Chem. Int. Ed. Engl. 58,
Electrochim. Acta 284, 392–399. 18604–18609.
162. Chu, K., Cheng, Y.H., Li, Q.Q., Liu, Y.P., and
Tian, Y. (2020). Fe-doping induced 169. Li, Y., Kong, Y., Hou, Y., Yang, B., Li, Z., Lei, L., 175. McPherson, I., and Zhang, J. (2020). Can
morphological changes, oxygen vacancies and Wen, Z. (2019). In situ growth of nitrogen- electrification of ammonia synthesis decrease
and Ce3+–Ce3+ pairs in CeO2 for promoting doped carbon-coated g-Fe2O3 nanoparticles its carbon footprint? Joule 4, 12–14.

1754 Chem 7, 1708–1754, July 8, 2021