Recent progress of thermocatalytic ammonia synthesis via an associative mechanism

Abstract

Ammonia is an important chemical and a carrier of hydrogen. Artificial nitrogen fixation mainly relies on the Haber−Bosch process driven by thermo-catalysis, where the activation of N₂ usually follows a dissociative route. Nonetheless, an associative route is common in electrocatalytic or biological nitrogen fixation under ambient conditions. Recently, it was reported that N₂ can be activated via an associative mechanism in thermocatalytic ammonia synthesis over some special catalytic materials or active sites, which exhibits enormous potential to realize efficient ammonia synthesis under relatively mild conditions (≤ 400 °C and ≤ 1 MPa). In this review, we focus on the recent progress of ammonia synthesis that follows an associative mechanism. Besides the hydride and nitride materials, the catalysts with highly dispersed active sites are included. We discuss the key factors that drive N₂ activation via an associative route and examine their correlation with catalytic performance. Finally, we delve into the perspectives of catalyst design tailored to facilitate an associative mechanism.

1. Introduction

The gigantic growth of global population and economy has led to huge increase of energy usage [1]. The enormous consumption of fossil fuels brings about greenhouse gas emission and drastic climate change [2]. There is a great push towards building energy systems that are efficient and green for the sustainability of human civilization. The target requires a transform of dependence from fossil to renewable energy (e.g., wind, tidal, and solar energy), aiming to construct a low-carbon or zero-carbon society. However, the task is formidable because many major industries are energy intensive. It is especially so for the ammonia synthesis industry, which consumes ca. 2% of global energy supply and releases more than 400 Mt of CO₂ yearly [3,4]. Therefore, it is urgent to realize energy saving and emission reduction in ammonia synthesis process.

Currently, ammonia is largely produced in industrial scale by the Haber−Bosch (HB) process from dihydrogen and dinitrogen [5]:

$${\mathrm{N}}_2+3{\mathrm{H}}_2\iff 2\mathrm{N}{\mathrm{H}}_3$$ (1)

Although the reaction is thermodynamically favorable because ∆G⁰ = −16.48 kJ mol⁻¹, it is not spontaneous. To propel the reaction, the HB process requires heavy energy input to maintain the high pressures (15−30 MPa) and temperatures (450−500 ℃). The main hurdle in ammonia synthesis is that N₂ is highly inactive due to the steady N N triple bond with a bonding energy of 945 kJ mol⁻¹ [6]. The energy gap between the highest occupied molecular orbital (HOMO) and the lowest unoccupied molecular orbital (LUMO) of N₂ molecule is 10.82 eV, which hinders electron donation and acceptance for N₂ adsorption and activation. Moreover, the molecular properties of N₂ with respect to the low proton affinity and high ionization potential as well as non-polarity determine that it is hard to activate N₂ molecules under relatively mild conditions without the application of an electric or photic field. Therefore, developing advanced catalysts that could activate N₂ molecules under mild conditions (≤ 400 °C and ≤ 1 MPa) remains a huge challenge.

Clarifying the catalytic reaction mechanism is of great significance for the construction of active centers and the design of new catalysts. In the past decades, considerable efforts were devoted to investigating the mechanism of NH₃ synthesis by in situ spectroscopy, isotope labeling, kinetic analysis, and theoretical calculation [7], [8], [9], [10], [11], [12]. In 2007, G. Etrl was awarded the Nobel Prize in recognition of his contributions in establishing a method for studying surface chemistry. By means of surface science, Etrl studied the adsorption behavior of N₂, H₂ and NH₃ on various Fe single-crystal surfaces, and found that the low sticking coefficient (typically in the order of 10⁻⁶) was the main obstacle to N₂ activation [13]. Furthermore, the Auger spectroscopic analysis revealed that the concentration of adsorbed N atoms (N_ad) on Fe(111) surface decreased with the increase of H₂ pressure, which concluded that dissociative nitrogen chemisorption is the rate-determining step (RDS). On this basis, the reaction mechanism of NH₃ synthesis was justified experimentally, and the elementary reaction steps were summarized as depicted in Fig. 1a [14]. It was suggested that the adsorbed N₂ first dissociates into two N_ad species, as the RDS, and then the N_ad is stepwise hydrogenated to NH₃. Such a sequence of elementary steps involved in NH₃ synthesis is called “dissociative mechanism”. Based on the idea of “dissociative mechanism”, Stoltze and Nørskov established a microscopic model using the kinetic results of Fe single-crystal surface (e.g., surface adsorption energy and sticking coefficient) as input parameters (Fig. 1b) [15]. It was found that the reaction rates predicted by the theoretical model were surprisingly in agreement with the experimental yields of Topsøe industrial plants, providing strong support for the “dissociative mechanism” concept.

Fig. 1.

(a) Mechanism and potential energy diagram of ammonia synthesis on iron [14]. (b) Comparison of calculated and measured NH₃ mole fractions at the reactor outlet for the industrial catalyst Topsøe KMIR [15]. (c) Associative reaction pathways for reduction of N₂ to NH₃. (1) Distal pathway and (2) alternate pathway.

Although the “dissociative mechanism” elucidates NH₃ synthesis in the HB process at high temperatures and pressures, it is not applicable to low-temperature NH₃ synthesis, such as nitrogen fixation by nitrogenase. This is because under ambient conditions, dinitrogen in enzyme catalysis does not undergo prior dissociation but is directly hydrogenated to give the final NH₃. Deviated from the conventional “dissociative mechanism”, the process concerning associative hydrogenation of dinitrogen to NH₃ is referred to as “associative mechanism”. The detection of hydrazine (N₂H₄) as a key intermediate during nitrogen fixation by nitrogenase provides substantial evidence supporting this mechanism [16,17]. Inspired by nitrogenase, nitrogen fixation at ambient temperature and pressure was achieved for the first time using a Mo-based complex (HIPTMo) containing tetradentate triamidoamine ligands in 2003 [18]. It opened up a new route of homogeneous catalysis for artificial nitrogen fixation. Arising from FeMo-nitrogenase, several homogeneous catalysts based on Mo or Fe as active components were developed. Similar to enzyme catalysis, the dinitrogen was bridged by complexes without broken triple bond and the terminal nitrogen atom was protonated via an associative pathway. Matthew et al. used electron paramagnetic resonance (EPR) spectroscopy to directly detect Fe−NNH species, revealing that P₃BFe complexes catalyze N₂ reduction by associative mechanism [19].

As depicted in Fig. 1c, the dinitrogen is non-dissociatively adsorbed on active sites, and then undergoes stepwise hydrogenation to generate N₂H₄ species, followed by the cleavage of the N−N bond to form NH₃ species. The hydrogenation process of N₂ could involve distal or alternate pathways. For distal hydrogenation, protons are bonded to one N atom to form *N−NH₂ or *N−NH₃ species, whereas for alternate hydrogenation, the two N atoms are hydrogenated by protons to form *NH−NH or *NH₂−NH₂ species. Inspired by the “associative mechanism”, researchers attempted to design heterogeneous catalysts for the effective synthesis of NH₃ via an associative route under mild conditions.

In this review, we summarize the recent progress of NH₃ synthesis over the catalysts that can facilitate the associative mechanism, including the catalysts that are endowed with highly dispersed active sites, hydrides, nitrides, and other kinds of catalysts (Scheme 1). The key factors responsible for the realization of associative mechanism and the resulting performance are discussed and the perspectives of catalyst design tailored to facilitate an associative mechanism are presented. This review affords guidance for the design and preparation of advanced catalysts for NH₃ synthesis via the associative mechanism under mild conditions.

Scheme 1.

Representative catalysts for NH₃synthesis based on the associative mechanism.

2. Catalytic system for ammonia synthesis via associative mechanism

2.1. Catalysts with highly dispersed active sites

The synthesis of ammonia over heterogeneous catalysts is a structure-sensitive reaction, and the metal size of catalysts has immediate implications on catalytic performance, which can be reflected by the turnover frequency (TOF). Boudart and coworkers evidenced that for a Fe/MgO catalyst, the TOF value of small Fe particles was much smaller than that of large ones [20]. They claimed that the high TOF of the large particles was due to the higher concentration of C₇ sites (a surface Fe atom with seven nearest neighbors, regarded as the most active site for dissociative chemisorption of N₂) relative to that of the small particles [21], which was proved by Somorjai and co-workers [22,23]. In the case of supported Ru catalysts, Jacobsen et al. reported that the increase of B₅ site (a three-fold hollow site with a bridge site nearby) concentration led to an increase in catalytic activity, and they related the catalytic activity with the number of B₅ sites [24]. Based on the “B₅” theory, Nørskov and co-workers calculated the number of B₅ sites on Ru particle with different sizes using density functional theory (DFT) [10]. They proposed that the supported Ru catalyst with an average Ru particle diameter of 2.0 nm has the maximum number of B₅ sites while that with an average diameter less than 2.0 nm lacks B₅ sites. As a result, the supported Ru catalyst with small nanoparticles was predicted to show poor catalytic activity. These results seem to suggest that particles less than 2 nm in size are not favorable for NH₃ synthesis due to the lack of B₅ sites. Nonetheless, recent researches disclosed that the catalysts with highly dispersed active sites can exhibit superior NH₃ synthesis performance, and the reaction mechanism is quite different from that of C₇ and B₅ sites.

2.1.1. Atomically dispersed catalysts

With the rapid advance of surface science and characterization techniques, researchers are able to precisely control and characterize the size and shape of particles in the sub-nanometer scale. In NH₃ synthesis, the atomically dispersed catalysts were superior to the particle counterparts, and the synthesis followed an associative mechanism. Qiu et al. reported a siliceous zeolite-supported Ru single-atom catalyst (Ru SAs/S-1) for NH₃ synthesis, and the presence of atomically dispersed Ru atoms was confirmed by aberration-corrected high-angle annular dark-field scanning transmission electron microscopy (AC—HAADF-STEM) [25]. At equal Ru loading, the catalytic activity of Ru SAs/S-1 was higher than that of conventional Cs-Ru/MgO (Fig. 2a). DFT calculations suggested that the N₂ linearly adsorbed on a Ru single-atom site was distally hydrogenated by adsorbed hydrogen, and with the final cleavage of N–N bond, NH₃ was formed via a distal pathway of associative mechanism. However, there was no direct experimental evidence to support this conclusion. Later, by near-edge X-ray absorption fine structure (NEXAFS) and in situ diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) analyses, Wang et al. detected the key intermediates on Ru/HZ single-atom catalyst (SAC) during NH₃ synthesis [26]. The results (Fig. 2b) revealed that the species containing =N− groups were the primary intermediates. DRIFTS results showed the appearance of N₂H₄ intermediates, suggesting that the distal pathway is more favorable for N₂ activation on Ru/HZ SAC. Combined with NEXAFS and DRIFTS results, it strongly confirmed the associative mechanism on Ru SAC, which is quite distinct from the typical nanoparticle catalysts that follow the dissociative mechanism.

Fig. 2.

(a) Temperature dependence of NH₃ synthesis rate over Ru SAs/S-1 and Cs-Ru/MgO (reaction conditions: pressure, 0.1 MPa; H₂:N₂ = 45:15 mL min⁻¹, with a WHSV of 18,000 mL g⁻¹ h⁻¹) [25]. (b) N K-edge NEXAFS spectra of Ru/HZ SAC after exposure to 25% N₂/75% H₂ at different temperatures and 1 MPa for 24 h [26]. (c) NH₃ synthesis rate of different Co-based catalysts at 350 °C under 1 MPa [11]. (d) NH₃ synthesis rate based on Co content over as-synthesized catalysts at 1 MPa [27]. (e) Three moieties of the Fe-N NH₂ intermediate [28]. (f) NH₃ synthesis rate of selected Ru and/or Co catalysts for NH₃ synthesis at 400 °C and 1 MPa [29].

The pore structure of zeolite, in which metal single atoms were anchored, constrained the metal loading (0.36% for Ru SAC/S-1 or 0.20% for Ru/HZ SAC), leading to unsatisfactory apparent catalytic performance. With consideration given to metal loading, Wang et al. anchored Co single atoms on N-doped carbon, realizing a Co loading of 3.73 wt.% [11]. The experimental results showed that Co₁–N_3.5 as active sites enabled a direct hydrogenation of N₂ to form NH₃ following an associative mechanism, exhibiting a much higher NH₃ synthesis rate than that of Co nanoparticle catalyst (Fig. 2c). By means of controlling pyrolysis temperature, a series of Co SACs with different Co₁–N_x (x = 2∼4) coordination numbers were synthesized, and it was found that Co₁–N₂ was the most active among the Co SACs, outperforming the Co nanoparticle catalyst (Fig. 2d) [27]. This metal–nitrogen complex as active site is available for other catalysts system. Li et al. proposed a FeN₃-embedded graphene SAC as a DFT model [28]. The results (Fig. 2e) showed that the high-spin polarization of the FeN₃ active center was favorable for N−N bond cleavage via either the distal or alternate pathway of associative mechanism. Besides SACs, Peng et al. investigated Ru−Co dual single-atom catalysts (RuCo DSAC) for NH₃ synthesis [29]. The results (Fig. 2f) demonstrated that the NH₃ synthesis rate of RuCo DSAC was 8.2-fold and 7.5-fold that of Co SAC and Ru SAC at 400 °C, respectively. The experimental and DFT studies revealed that the cooperation of Ru and Co centers afforded efficient NH₃ synthesis via the associative pathway.

2.1.2. Atomic clusters and subnanometric clusters catalysts

Another fascinating field in atomic cluster catalysts was also investigated in depth. Unlike SACs where single atoms are highly dispersed on the support in isolation, a number of atoms are assembled in some individual clusters to form atomic cluster catalyst, which results in unique geometric and electronic structures [30]. When the atom number of active site increases from one to three or more, the coordination between atoms cannot be neglected, which has a significant influence on catalytic performance, and this category of catalysts can be defined as atomic clusters and/or subnanometric clusters catalysts. Based on first-principles theoretical studies, Li et al. proposed that Fe₃ clusters anchored on θ-Al₂O₃ (Fig. 3a) should have a TOF two orders of magnitude higher than that of Fe C₇ sites, because Fe₃ clusters are favorable for efficient N₂ activation via the alternate pathway of associative mechanism rather than the direct N₂ dissociation [31]. For the first time, the feasibility of atomic cluster catalysts facilitating thermocatalytic NH₃ production via an associative mechanism was theoretically disclosed. Moreover, the research group also found that the preferred NH₃ synthesis reaction pathway on Rh₁Co₃ single-cluster catalyst is associative mechanism (Fig. 3b) [32]. Due to the unsaturated coordination of Rh atom, N₂ and H₂ molecules preferred to co-adsorb on the same Rh center of Rh₁Co₃ site. Therefore, the charge buffer capacity of Rh is decisive, which directly underlies the catalytic performance for NH₃ synthesis. As for Ru-based catalysts, the DFT calculations of Li's group (Fig. 3c) proved that the associative mechanism was also preferred over the Ru atomic cluster catalyst Ru₃/N_v-g-C₃N₄ [33].

Fig. 3.

(a) TOFs of ammonia synthesis over the three catalysts as a function of N₂ partial pressure at 700 K and 100 bar [31]. (b) The model of preferred associative mechanism over Rh₁Co₃ ACCs [32]. (c) TOFs and contributions from dissociative and associative mechanisms as a function of pressure under constant temperature of 623 K [33]. (d) Turnover frequencies (TOF_{Co total}) at 400 °C and 1 MPa [34]. (e) NH₃ synthesis rate over the as-synthesized Ru catalysts [12]. (f) UV–vis absorbance spectra of the products over 0.5Ru-SNCs catalysts at different temperatures during NH₃ synthesis [35]. (g) NH₃ synthesis pathway over Ru catalysts with different sizes [12]. (h) Relationship of N₂ dissociation and N₂ hydrogenation energy barriers with the corresponding *N₂ Bader charge [37].

In experimental studies, some atomic cluster catalysts were successfully prepared and employed to catalyze NH₃ synthesis. Peng et al. prepared Co, Mn, and Fe atomic cluster catalysts using complexation-pyrolysis strategy, and Co₂ atomic cluster catalyst (ACC, Fig. 3d) showed remarkable NH₃ synthesis performance compared with Co nanoparticle catalyst and Co SAC [34]. The high NH₃ synthesis rate of Co₂ ACC was originated from the strong interclustering and metal-support interactions, which accelerated electron transfer between Co and support, facilitating N₂ activation and hydrogenation to NH₃. Meanwhile, Ru₃ ACC was successfully prepared by the same method, and displayed an unprecedentedly high NH₃ synthesis rate through the associative pathway (Fig. 3e) [12]. In summary, in the field of single atom and atomic cluster, the associative mechanism is preferred on the catalysts due to the unique geometric and electronic structures of single atom and single atomic cluster.

When the size of active metal increases from atomic cluster to subnanometric cluster, the number of Ru atoms that involved in N₂ activation becomes larger, but the N₂ activation pathway remains unchanged owing to the lack of B₅ sites (< 2 nm). Through controlling the impregnation temperature below 0 °C, the particle size of Ru metal could be limited to subnanometer scale (0.8 nm), effectively promoting NH₃ synthesis to surpass those Ru nanoparticle catalyst prepared by conventional method [35]. By virtue of ultraviolet-visible (UV–vis) absorption spectroscopy (Fig. 3f), the absorption band at 468 nm corresponding to the key intermediate species, viz. N₂H₄, could be detected over the Ru subnanometric scale but not over the nanoparticle catalyst. Meanwhile, in situ DRIFTS experiment confirmed the presence of *N₂D_x intermediates, which suggested that the Ru subnanometer catalyst preferably followed an alternate pathway of associative mechanism for N₂ activation. Zhang et al. prepared subnano Ru clusters on Sm₂O₃, which afforded high ammonia yields at 400 °C [36]. The theoretical studies showed that the distal pathway of associative mechanism is favoured in the process of NH₃ synthesis over subnano Ru clusters catalyst, which is more energy efficient than the dissociative pathway over B₅ sites.

As mentioned above, the particle size of metal greatly affects the N₂ activation pathway, where small-size metal catalysts involving single atoms, atomic clusters, and subnanometric clusters activate N₂ via an associative mechanism, while the nanoparticles activate N₂ via a dissociative mechanism. The boundary of metal size that distinguishes the dominant reaction mechanism is ambiguous without systematic experiments, because the chemical environment of active metal and performance test conditions could be regulated at will. For this purpose, Li et al. reported multiscale Ru catalysts with sizes ranging from single atoms, atomic clusters, subnanometric clusters, to nanoparticles, and systematically summarized the size effect of Ru entities on reaction mechanism [12]. Fig. 3g pointed out that with the diminution of Ru entity size from particle to subnanometer cluster, there is a change of dominant pathway for N₂ activation from dissociation to association. In addition, Peng et al. reported the theoretical boundary of Fe size effect on reaction mechanism [37]. As shown in Fig. 3h, the energy barrier of NN–H formation is much lower than that of NN bond cleavage over Fe₂ cluster catalyst, suggesting that the dominance of associative mechanism on the Fe₂ cluster catalyst. With the increase of Fe coordination numbers in Fe_n clusters (i.e., Fe₃ and Fe₄), the energy barrier of NN–H formation increases while that of NN bond cleavage significantly reduces. There is a crossover over the Fe₄ cluster catalyst, where the energy barrier of NN bond cleavage (1.34 eV) is close to that of NN–H formation (1.32 eV), indicating that the dissociative mechanism is also active.

2.1.3. Single-atom alloy catalysts

Among metallic catalysts, the incorporation of a heterometal to form alloy is a feasible strategy to improve NH₃ synthesis performance due to synergistic effect. When one of the two metals in the alloy was lowered to atomic level, there is the formation of single-atom alloys (SAA), which also have unique electronic and geometric structures alike those of single-atom sites [38]. Moreover, within the structure of SAA catalyst, there is strong metallic interaction (i.e. alloy bonding) between the guest (single atoms) and the host (metal supports) [39]. Since the properties of SAA are similar to those of SAC, the associative mechanism is also preferred for N₂ activation over SAA. Using wet-chemistry method, Zhang et al. prepared Ru-Co SAA, and found that the Ru-Co SAA catalysts exhibited NH₃ synthesis performance superior to that of traditional Ru-Co nanoparticle alloy catalysts [40], [41], [42]. It was proved that unlike the Ru-Co nanoparticle alloy catalysts that follow the dissociative pathway for N₂ activation, the Ru-Co SAA catalysts adhere to the associative mechanism. The NH₃ synthesis performances of mentioned catalysts are summarized in Table 1.

Table 1.

NH₃synthesis performance of selected catalysts with highly dispersed active sites.

	Catalysts	T (°C)	P (MPa)	WHSV (mL g−1 h−1)	Rate (mmol g−1 h−1)	Ref.
Atomically dispersed catalysts	Ru SAs/S-1	375	1.0	18 000	6.10	[25]
	Ru/HZ SAC	300	1.0	60 000	12.60	[26]
	Co-N-C	350	1.0	60 000	4.34	[11]
	Co-N2	300	1.0	60 000	2.73	[27]
Atomic clusters and subnanometric clusters catalysts	RuCo DSAC	350	1.0	60 000	3.80	[29]
	Co2 ACCs	400	1.0	60 000	8.54	[34]
	Ru ACCs	400	1.0	60 000	7.42	[12]
	0.5Ru-SNCs	400	1.0	60 000	11.70	[35]
	Ru/Sm2O3	400	1.0	24 000	32.21	[36]
Single-atom alloy catalysts	Co1Ru TCs	360	1.0	60 000	4.70	[40]
	Ba/Co1Ru SAA	400	1.0	60 000	12.00	[41]
	Ru1.7Co1 SAA	400	3.0	60 000	10.60	[42]

In short, it is feasible to prepare subnanometer-scale catalysts that can achieve efficient NH₃ synthesis through the associative pathway, which is a promising strategy to bypass the bottleneck of dinitrogen dissociation. In virtue of surface spectroscopy and isotope labeling techniques, the key intermediates in NH₃ synthesis reaction could be identified, which eventually defined the reaction mechanism. However, it still remains debatable in the determination of an accurate boundary between associative and dissociative mechanisms, or in the parallel reaction mechanism as indicated by Li's calculation [33]. More effort is required to achieve precise control of metal size to establish a set of model catalysts to investigate the NH₃ synthesis mechanism.

2.2. Metal hydrides

Hydride is a kind of compounds containing negatively charged hydrogen anion (H⁻) [43]. With the development of synthesis technology, a class of hydrogenous compounds, where hydrogen is bonded to electropositive element or group (metal or nonmetal in ionic, metallic or covalent manner), were fabricated and referred to as hydrides [44,45]. The general formula of hydride is MH_n or XMH_n, where M is a metal cation and X is a metal or nonmetal cation that binds with H. Hydrides can be classified as binary metal hydrides (LiH, BaH₂, TiH₂) [46], [47], [48], complex hydrides (NaAlH₄, LiBH₄, Mg₂FeH₆) [49], [50], [51], and mixed anion hydrides (BaTiO_2.5H_0.5, LnHO, SrCrO₂H) [52], [53], [54]. Since hydride is hydrogen-rich and decomposes easily to release H₂, efforts were put in to use hydrides as hydrogen carriers [55,56]. In recent years, there has been increasing interest in employing hydrides as hydrogen providers in NH₃ synthesis.

It is important to note that electrons (e⁻) and protons (H⁺) are needed in the process of NH₃ synthesis [57]. In conventional catalyst systems for NH₃ synthesis, Brønsted acids and active metals could provide protons and electrons, respectively. In the cases of hydrides, it could synchronously supply electrons and protons through the conversion of H⁻ to H⁰ and further to H⁺, respectively [58]. As a matter of fact, the hydrogen species in hydrides are directly involved in NH₃ formation via a hydride-mediated Mars van Krevelen (MvK) mechanism, and with the supply of gaseous H₂, lattice hydrogen could be regenerated, hence accomplishing a NH₃ synthesis cycle. A possible mechanism is that N₂ chemisorbs on positively charged site without dissociation due to strong kinetics inhibition. Then the non-dissociative N₂ is stepwise hydrogenated by lattice hydrogen to form NH₃. The hydrogen vacancies are replenished by the H₂ from the gaseous phase. The process is a combination of associative mechanism and MvK mechanism. Benefiting from the strong reducing property and ample supplement of hydrogen species, hydride exhibits outstanding NH₃ synthesis performance as a catalyst or a support. To visualize the performance comparison of hydrides, the NH₃ synthesis rates of recent hydride-based catalysts are summarized in Table 2.

Table 2.

NH₃synthesis performance of selected hydride-based catalysts.

	Catalysts	T (°C)	P (MPa)	WHSV (mL g−1 h−1)	Rate (mmol g−1 h−1)	Ref.
Binary metal hydrides	Co-LiH	300	1.0	60 000	4.70	[59]
	Pd−20LiH-w	300	0.1	60 000	0.75	[60]
	BaH2—Co/CNTs	300	1.0	60 000	4.80	[61]
	Ru/CaFH	125	0.1	36 000	0.19	[62]
	Ru/BaO—CaH2	340	0.1	36 000	10.50	[63]
	TiH2	400	5.0	66 000	2.20	[64]
	VH0.39	400	5.0	66 000	3.20
	Ru/3LaN/ZrH2	350	1.0	60 000	5.60	[65]
	Ba-Ru/ZrH2	400	1.0	60 000	27.50	[66]
	Ru/3TiCN/ZrH2	350	1.0	60 000	10.30	[67]
	Ru/LaH2+x	340	0.1	36 000	3.39	[68]
Complex hydrides	Li4FeH6	300	1.0	60 000	5.00	[69]
	Li4RuH6/MgO	300	1.0	60 000	22.00	[70]
	Ba2RuH6/MgO	300	0.1	60 000	13.00	[71]
Mixed anion hydrides	BaTiO2.5H0.5	400	5.0	66 000	1.40	[72]
	BaCeO3–xNyHz	400	0.9	36 000	10.10	[73]
	[BaCrHN]	300	1.0	60 000	6.86	[74]
	h-Ca3CrN3H	400	5.0	66 000	3.80	[75]
	TiO2–xHy/Fe	500	0.1	2400	560 (ppm)	[76]
	Ru/SrBaTiO3−xHx	400	5.0	66 000	15.00	[77]

2.2.1. Binary metal hydrides

Binary metal hydrides MH_n have the simplest composite among hydrides. Given that many metals are highly reactive to hydrogen [78], binary metal hydrides could be synthesized by heating metals in a hydrogen atmosphere, where the metals can directly or reversibly react with hydrogen. As an example, LiH can be formed by exposing molten lithium to hydrogen in the temperature range of 427 to 727 °C [79]. With high hydrogen content, LiH was employed in NH₃ synthesis by Chen's group in 2017 [59]. They found that LiH triggered the catalytic activity of transition metals (V, Cr, Mn, E, Co, and Ni), which increased by four orders of magnitude at 300 °C (Fig. 4a). Also, LiH-modified transition metals showed superior catalytic activities over Ru/MgO by 12–20 times at 250 °C (Fig. 4b). As a matter of fact, the early transition metals are usually trapped in over-strong binding of N-intermediates, resulting in poor catalytic performance. As a reducing agent, the H⁻ species of LiH effectively attacked N-intermediates and weakened the binding of N-intermediates on transition metal, and thus the LiH-modified transition metals exhibited unprecedented catalytic activities. More importantly, LiH as a second catalytic site also provided plenty of protons to get involved in NH₃ synthesis. In addition, further enhancement by LiH could be achieved through the addition of Pd or Co–Mg–O solid solution that is large in specific surface area [60,80].

Fig. 4.

(a) Activity comparison of 3d TMs (from V to Ni) with or without LiH composition at 300 °C. (b) NH₃ synthesis rate as a function of temperature. The inset is the enlarged portion of the rates in the temperature range of 147−227 °C. Reaction conditions: N₂:H₂=1:3 with a WHSV of 60 000 mL g⁻¹ h⁻¹; pressure, 10 bar [59]. (c) Temperature dependence of catalytic performance over BaH_2—Co/CNTs and BaO—Co/CNTs catalysts [61]. (d) Rates of ammonia formation for Ru/CaFH, Ru/BaO–BaH₂, and Ru/CaH₂ at 50, 75, 100, and 125  °C [62]. (e) Comparison of catalytic activities over various catalysts for NH₃ synthesis at 400  °C, 5 MPa [64]. (f) NH₃ synthesis performance of various catalysts at 400 °C under 0.2 or 1 MPa [67].

Other binary metal hydrides also exhibited unexpected catalytic performances for NH₃ synthesis. Chen et al. (Fig. 4c) encapsulated BaH₂ nanoparticles in Co/CNT, and found that the NH₃ synthesis rate of BaH_2—Co/CNT was 2 orders of magnitude higher than that of BaO—Co/CNT [61]. The strong reducing property of BaH₂ enhanced the detachment of activated N species from the Co surface, leading to excellent catalytic activity. Hosono et al. reported a Ru supported on CaFH solid solution that showed a NH₃ synthesis rate of 0.19 mmol g⁻¹ h⁻¹ even at 125 °C (Fig. 4d), whereas Ru/CaH₂ and Ru/BaH₂ both displayed no activity for NH₃ synthesis [62]. The insertion of F⁻ in CaFH weakened the Ca−H bonds and released more H⁻, which facilitated NH₃ synthesis at low temperatures. In view of the inferior performance of Ru/CaH₂ in NH₃ synthesis, BaO was added as a structural promoter or an electronic promoter [63]. The results showed that part of the BaO was reduced by CaH₂ to BaH₂ and the formation of BaO-BaH₂ composites was beneficial for hydrogen release and storage.

In addition, Kobayashi's group conducted NH₃ synthesis tests on a series of hydrides at 400  °C under 5 MPa, including TiH₂, VH_0.39, NbH_0.6, and ZrH₂ [64]. As shown in Fig. 4e, VH_0.39 and TiH₂ exhibited an NH₃ synthesis rate of 3.2 mmol g⁻¹ h⁻¹ and 2.2 mmol g⁻¹ h⁻¹, respectively, while NbH_0.6 and ZrH₂ were inactive for NH₃ synthesis. Although ZrH₂ by itself was poor in NH₃ synthesis performance, the ZrH₂ loaded with Ru nanoparticles showed an excellent NH₃ synthesis activity (9.0 mmol g⁻¹ h⁻¹), obviously superior to Ru/ZrO₂ and Ru/ZrN [66]. The moderate interaction of Ru−H bond can inhibit the aggregation of Ru particles and facilitate the effect of hydrogen spillover, avoiding hydrogen poisoning that occurs on Ru/ZrO₂. The catalytic performance of Ru/ZrH₂ can be further enhanced by the introduction of a second catalytic site. For example, nitride and carbonitrides (LaN, TiN, and TiCN) were commonly composited with Ru/ZrH₂ to obtain considerable NH₃ synthesis rates (Fig. 4f) [65,67]. Moreover, Mizoguchi et al. reported a series of Ru/LnH_2+x (Ln = La, Ce, or Y) that showed NH₃ synthesis activities comparable to that of benchmark Ru–Cs/MgO at 340 °C [68].

2.2.2. Complex hydrides

Different from a simple binary metal hydride, a complex hydride contains another metal cation (usually transition metal, such as Ru, Re, or Ni) that is inserted in its crystalline structure [81]. Some complex hydrides can be directly synthesized from the combination of binary metal hydride and transition metal powder by ball milling (Mg₂FeH₆) [82] or by combustion synthesis (Mg₂NiH₄) [83] in hydrogen at high pressures (> 16 MPa) and high temperatures (∼750 °C). Unlike transition metal (TM) supported on binary metal hydride, the TM element in complex hydride is nonmetallic and bonds with H⁻ to form “complex” unit [TMH_x]^δ−, including tetrahedral [NiH₄]⁴⁻, square-pyramidal [CoH₅]⁴⁻ and so on, resulting in different electronic, compositional and structural properties when compared to transition metal surfaces and isolated transition metal complexes (Fig. 5a).

Fig. 5.

(a) Li₄RuH₆ surface and the local coordination of Ru in Li₄RuH₆ (Ru, red; H, yellow; Li, white), Ru(0001) surface (the B₅ site is highlighted in bright red) and the model of molecular Ru complex (L, ligand) [70]. (b) Catalytic activities of select catalysts for NH₃ synthesis at 400 °C, 5 MPa [72]. (c) The possible mechanisms for NH₃ formation over TM/BaCeO_3–xN_yH_z through an anion-vacancy (V_a)-mediated Mars–van Krevelen mechanism [73]. (d) Comparison of NH₃ production rates of Cr-based catalysts as well as reference Ru/MgO and Cs-Ru/MgO catalysts under the conditions of 10 bar and 573 K [74]. (e) Activity comparison of ammonia synthesis over TiO_2–xH_y/Fe-NL, pure Fe-NL, and the commercial benchmark wüstite-based Fe catalyst at 300−500 °C and 1 atm [76].

Containing electron-rich [TMH_x]^δ− unit, complex hydrides have huge potential in NH₃ production. Chen's group found the [Li₄FeH₆]⁻ and [Li₅FeH₆]⁻ clusters (formed in the exposure of Fe-LiH composite to a flow of hydrogen) were the genuine active centers for N₂ activation rather than the Fe C₇ sites [69]. As a result, Fe-5LiH/MgO catalyst exhibited a higher NH₃ synthesis rate than that of Cs-Ru/MgO catalyst. Inspired by the role of Li₄FeH₆ in Fe-LiH composite, Chen and coworkers synthesized Li₄RuH₆ and Ba₂RuH₆ by ball milling of ruthenium metal powder with LiH or BaH₂ [70,71]. Impressively, Li₄RuH₆ and Ba₂RuH₆ both exhibited ultrahigh NH₃ synthesis activities, reaching 22 mmol g⁻¹ h⁻¹ and 34 mmol g⁻¹ h⁻¹, respectively, at 300 °C and 1 MPa. The study demonstrated that N₂ preferred to adsorb on [RuH₆]⁴⁻ site to form [RuH₆]−NN species, which underwent non-dissociative reductive protonation by H⁻, forming [RuH₅]−NHN species. The subsequent hydrogenation to [RuH₅]−NHNH and [RuH₅]−NHNH₂ species were realized by neighboring lattice H⁻. The detection of NHNH and NHNH₂ species confirms that the reaction pathway of NH₃ synthesis on Li₄RuH₆ followed an alternate pathway of associative mechanism. The same mechanism occurred on Ba₂RuH₆, but the significant difference was that N₂ was no longer adsorbed on the Ru site but on the Ba site [71]. The mechanism mentioned above illustrated the key role of complex hydrides in electron and proton transfer.

2.2.3. Mixed anion hydrides

There is another series of hydrides containing anions such as O²⁻ and N³⁻ called mixed anion hydrides. Kobayashi et al. synthesized BaTiO_2.5H_0.5 through the reduction of BaCeO₃ by CaH₂ and investigated its NH₃ synthesis activity, reporting 1.4 mmol g⁻¹ h⁻¹ at 400 °C and 5 MPa (Fig. 5b) [72]. Similarly, Kitano et al. applied BaCeO_3–xN_yH_z (Fig. 5c) with high density of H⁻ and N³⁻ to efficiently catalyze NH₃ production [73]. The lattice H⁻ and N³⁻ species played a critical role in the enhancement of NH₃ synthesis activity. Recently, Cr-based mixed anion hydrides were employed for NH₃ synthesis, such as [BaCrHN] from Guo's group [74] and h-Ca₃CrN₃H from Kobayashi's group [75]. In the case of [BaCrHN] (Fig. 5d), the presence of lattice H⁻ and N³⁻ enabled Cr to be active for NH₃ synthesis, showing a NH₃ synthesis rate ca. four times that of benchmark Cs-promoted Ru/MgO catalyst. For h-Ca₃CrN₃H, the NH₃ synthesis activity was higher than those of other hydrides such as TiH₂, BaTiO_2.5H_0.5, and VH_0.39. These results verified that the mixed anion hydrides are active for NH₃ synthesis, and they could function as support to enhance the catalytic activity of Ru, Co, and Fe. Zhang et al. reported a Fe supported on TiO_2−xH_y catalyst (Fig. 5e), which exhibited an NH₃ synthesis activity superior to those of commercial benchmark Fe catalysts, and found that O_V−H in TiO_2−xH_y facilitated nitrogen hydrogenation to form N–H bond [76]. Kobayashi et al. investigated the NH₃ synthesis performance of Ru, Fe, and Co particles supported on BaTiO_3−xH_x as well as that of Ru particles supported on (Ca, Sr, Ba)TiO_3−xH_x [77]. The studies demonstrated that Ru/BaTiO_3−xH_x was the most active catalyst among the Ru/ATiO_3−xH_x (A = Ca, Sr, Ba).

In general, following either distal or alternate pathway of the associative mechanism, hydride-based catalysts exhibit outstanding NH₃ synthesis performance at low temperatures and pressures, especially in terms of hydride-supported TM catalysts. However, because the hydrides are sensitive to air and moisture, their industrial application is still a challenge. Moreover, the large-scale preparation of hydride is tough as the preparation requires a flow of high-pressure H₂ at elevated temperature. More efforts are needed to develop advanced hydride catalysts that are safe and readily available.

2.3. Metal nitrides

In recent years, metal nitrides have become promising materials for ammonia synthesis under mild conditions [84,85]. Similar to the MvK process over metal oxides in oxidation reactions, lattice nitrogen in metal nitrides can react with hydrogen to generate ammonia together with the formation of nitrogen vacancies [85], [86], [87]. The nitrogen vacancies can serve as active sites for the activation of N₂. Different from the dissociative route for N₂ activation over most metal sites, N₂ molecules can be adsorbed and activated over nitrogen vacancies in an end-on configuration without dissociation [88]. Then, the activated N₂ can stepwise react with dissociated hydrogen to form ammonia via an associative route. It was reported that the activity of metal nitrides in ammonia synthesis is closely related to the formation energy of nitrogen vacancy (E_NV) [89]. As shown in Fig. 6a, the metal nitrides with low E_NV value, such as CeN and LaN, exhibited high ammonia synthesis rates, while the metal nitrides with high E_NV value, such as ZrN or TiN, showed low NH₃ synthesis rates [89]. The NH₃ synthesis rates of nitride-based catalysts are summarized in Table 3.

Fig. 6.

(a) Catalytic activity for NH₃ synthesis over various metal nitrides at 400 °C [89]. (b) Schematics of the formation of nitrogen vacancies (V_N) and the related electron transfer pathways. The V_N formation energies (E_NV) were determined by DFT calculations [88]. (c) DFT studies of the reaction path over Ni/LaN for NH₃ synthesis [88]. (d) Schematics of NH₃ synthesis pathway over Ru/LaN/ZrH₂[65].

Table 3.

NH₃synthesis performance of selected nitride-based catalysts and other catalytic system.

Catalysts		T (°C)	P (MPa)	WHSV (mL g−1 h−1)	Rate (mmol g−1 h−1)	Ref.
Nitrides	Ni/LaN NPs	400	0.1	36 000	5.54	[88]
	Ni/CeN NPs	400	0.1	36 000	6.50	[89]
	Ru/3LaN/ZrH2	350	1.0	60 000	5.60	[65]
	Co/CeN NPs	400	0.9	36 000	27.1	[90]
	Ru/3TiCN/ZrH2	400	1.0	60 000	25.6	[67]
Other catalytic system	Ba-Ru-Li/AC	400	1.0	60 000	19.6	[91]
	RuLa/HZ	400	0.2	60 000	9.6	[92]

The combination of metal nitrides with transition metals is an effective strategy to improve NH₃ synthesis rate. Ye et al. [88] reported that the nickel-loaded lanthanum nitride (Ni/LaN) enabled stable and highly efficient ammonia synthesis, with the credit given to a dual-site mechanism that avoided the commonly encountered scaling relations. As shown in Fig. 6b, the nickel metal can not only dissociate H₂ molecules but also lower the E_NV of LaN, which decreased from 1.9 eV of LaN to 1.6 eV of Ni/LaN. The facile formation of nitrogen vacancies over metal nitrides effectively promoted the adsorption and activation of N₂ molecules. Theoretical calculations indicated that the direct dissociation of N N was associated with a large energy barrier of at least 2.5 eV, which was unlikely to occur. By comparison, the hydrogenation process for N N bond breaking via a distal pathway of associative mechanism was more favorable to proceed (Fig. 6c). The use of distinct sites for activating the two reactants, and the synergy between them, resulted in the Ni/LaN catalyst exhibiting an activity that far exceeded that of most conventional cobalt- and nickel-based catalysts. Besides, Li et al. [65] reported that the LaN-promoted Ru/ZrH₂ catalyst exhibited a high NH₃ synthesis rate under mild conditions (Fig. 6d). It was reckoned that the N₂ hydrogenation followed an associative pathway via a favorable chemical looping route under mild conditions.

Furthermore, the integration of associative and dissociative routes over metal-nitride-based catalysts was developed to enhance ammonia synthesis performance. Ye et al. reported that the Co-loaded CeN catalyst can efficiently realize a high ammonia synthesis rate under mild conditions [90]. As shown in Fig. 7a, the ammonia synthesis rate over Co/CeN can reach up to 27.1 mmol g⁻¹ h⁻¹, which was significantly higher than that over Co/C12A7:e⁻ and Co-Mo/CeO₂ catalysts. It was proposed that the formation of nitrogen vacancies in CeN resulted in a low work function (2.6 eV), which accounted for the strong electron donation ability that facilitated efficient N₂ cleavage over Co sites via a dissociative route. The strong nitrogen affinity of surface nitrogen vacancies in CeN led to the activation of N₂ at the sites via an associative route. Due to this unique mechanism, the ammonia synthesis activity of Co/CeN was much higher than those of other reported Co-based catalysts (Fig. 7b). At the same period, Zhou et al. [67] reported that Ru-based catalyst promoted by titanium carbonitride (TiCN) can achieve an outstanding ammonia synthesis rate via the combination of dissociative and associative routes. More experimental evidences about the co-involvement of the dissociative and associative mechanisms were provided. As shown in Fig. 7c, the results of AC-STEM showed that TiCN can reduce the aggregation of Ru metal particles and enhance the exposure of Ru B₅ step sites for N₂ dissociation. The formation of ¹⁴N¹⁵N signal (m/z = 29) in ¹⁵N₂ isotopic exchange experiment definitely demonstrated that ¹⁵N₂ can be dissociated over Ru sites and then reacted with lattice ¹⁴N in TiCN to generate ¹⁴N¹⁵N species (Fig. 7d). In addition, the presence of Ru metal can facilitate the formation of nitrogen vacancies on the TiCN surface for the hydrogenation of N₂. The combination of NEXAFS (Fig. 7e), in situ DRIFTS, and X-ray photoelectron spectroscopy (XPS) results confirmed the presence of intermediate N₂H_x species, providing evidence for the participation of associative route over the Ru/TiCN/ZrH₂ catalyst. The synergistic effect of Ru and TiCN on N₂ activation via integrating the dissociative and associative mechanisms accounted for the superior NH₃ synthesis rate of Ru/TiCN/ZrH₂ under mild conditions (Fig. 7f).

Fig. 7.

(a) Catalytic activity for NH₃ synthesis over various Co catalysts at 400 °C [90]. (b) Proposed co-involvement dissociative and associative mechanisms of NH₃ synthesis over Co/CeN [90]. (c) AC-STEM images of the used Ru/TiCN/ZrH₂ and the model of Ru nanoparticle with B₅ sites [67]. (d) ¹⁵N₂ isotopic exchange experiment over Ru/3TiCN/ZrH₂ at 400 °C [67]. (e) Nitrogen K-edge NEXAFS spectra of Ru/3TiCN/ZrH₂ after being treated in the reaction atmosphere for 24 h [67]. (f) The proposed dissociative and associative mechanisms of NH₃ synthesis over Ru/3TiCN/ZrH₂[67].

In general, the active sites of nitrogen vacancies over metal nitrides play an essential role in driving N₂ activation via an associative route. The E_NV value of metal nitride is a key factor in determining NH₃ synthesis performance. Despite the metal nitrides with low E_NV are favorable for realizing high ammonia performance, they are extremely sensitive to air and moisture and thus result in the difficulty of preparation and undesired instability of catalysts. Therefore, the development of highly active and stable metal nitrides is a long-term goal for ammonia synthesis.

2.4. Other catalytic systems

As mentioned above, the regulation of Ru metal size would change the reaction route of ammonia synthesis. Besides, the addition of promoter into Ru catalysts or alloying Ru with rare earth elements can also change the activation route of N₂ molecules. Zheng et al. [91] reported that Li-promoted Ru catalyst exhibited a significantly higher ammonia production rate than previously reported Ru-based catalysts at comparably mild conditions. In comparison with other alkali dopants (Fig. 8a), such as Na, K, and Cs, the promotion effect of Li on ammonia synthesis is obvious despite it being the poorest in electron-donating ability. The results of high-resolution transmission electron microscopy (HRTEM) and atomic electron energy loss spectroscopy (EELS) mapping (Fig. 8b) showed that the Li⁺ preferentially stayed at the step sites presumably due to its small and matching size, which blocked the B₅ step sites for N₂ dissociation. Meanwhile, Li can significantly polarize and stabilize adsorbed dinitrogen intermediates on terrace sites, which facilitated the activation of N₂ via an alternate pathway of associative mechanism for ammonia synthesis. Zhang et al. reported that the Ru-M (M = La or Y) alloy catalysts can maximize the electronic effect of rare earth elements, leading to the formation of a new type of active site [92]. The ammonia synthesis and TOF value of RuLa/HZ catalyst were significantly higher than these of Ru/HZ (Fig. 8c). It was proposed that the formation of Ru-La alloy enhanced the sintering resistance of Ru NPs and exposed more Ru terrace active sites for N₂ hydrogenation to form N₂H_x (x = 1∼4) via an alternate route of associative mechanism. Besides, the Ru-La interaction promoted electron transfer from La to Ru centers and then accelerated N₂ activation on the Ru-La sites via the dissociative mechanism (Fig. 8d). The co-involvement of dissociative and associative mechanisms for ammonia synthesis over RuLa/HZ accounted for the outstanding catalytic performance at mild conditions.

Fig. 8.

(a) A systematic study of promoting Ba-Ru/AC with alkali metals in NH₃ synthesis with respect to total Ru content [91]. (b) An atomic EELS mapping (inset) of a STEM image of a Ru nanoparticle (blue), showing pre-concentration of Li⁺ (green) in edged/stepped site regions [91]. (c) TOF_{Ru total} and TOF_{Ru sur} of the as-synthesized catalysts at 400 °C and 1 MPa [92]. (d) Reaction schematic for NH₃ synthesis through collaborative deployment of the associative (Route 1) and dissociative routes (Route 2) [92]. (e) Work function versus N₂ adsorption energy (ΔE_N2*) for Ti/(n, n) BNNT as well as bulk Ti/h-BN heterostructure [93]. (f) Energy diagram of N₂-to-NH₃ conversion on Ti/(8, 8) BNNT, Ti/(9, 9) BNNT, and bulk Ti/h-BN heterostructure [93].

Apart from the metal sites for N₂ activation, theoretical studies showed that N₂ molecules can be activated over non-metal sites via an associative route. Based on the results of first principles calculations and microkinetic modeling, Zhou et al. reported that boron nitride nanotubes (BNNTs) with the nanowires of early transition metals encapsulated can serve as an efficient catalyst for ammonia synthesis [93]. The results showed that the filling of BNNTs with early transition metals, such as Ti, Zr, V, Cr, and Sc, can destruct the π conjugation of BN network and liberate the p_z orbitals of boron atoms for N₂ fixation. The binding strength of reaction molecules and intermediates on the boron sites can be tuned by the type of metal filler and nanotube diameter through the engineering of the BNNTs work function (Fig. 8e). With the boron atoms serving as reaction centers, N₂ molecules were preferentially hydrogenated via an alternate pathway of associative mechanism (Fig. 8f). This work provides a new idea to develop non-metal sites for NH₃ synthesis under mild conditions.

3. Summary and perspectives

So far, efforts have been devoted to developing efficient catalysts that follow the associative mechanism for ammonia synthesis. The discrimination between dissociative and associative mechanisms in experiments is still a standing challenge, and many efforts were put into finding solid evidence, such as the detection of key intermediates. The process of associative mechanism involving N₂H_x intermediates, which could be detected in virtue of spectroscopy and surface science, differs from that of dissociative mechanism. Herein, we have summarized some means or techniques to detect N₂H_x intermediates from the following aspects: (i) UV–vis spectrum developed by Chrisp and Watt [94] is one of the most direct methods in the detection of intermediates. The hydrazine (N₂H₄) in outlet gas could be collected by p-dimethylaminobenzalde solution, and the color of solution correlated with N₂H₄ concentration is determined by UV–vis spectrum. In UV–vis profiles, a peak at ∼462 nm is attributable to the N₂H₄, revealing the formation of N₂H₄ intermediates [27]. However, the application of this method is limited owing to the difficult collection of gaseous N₂H₄, especially at high-temperature conditions; (ii) Due to the difficult collection of N₂H₄ intermediates, the characterization of N₂H_x that adsorbed on surface is a substitute of UV–vis spectrum, such as in situ DRIFTS. When the catalyst is exposed to an atmosphere of 25%N₂–75%D₂, it is found that the IR band at ∼2372 cm^–1 is assigned to N₂D₂ fragment, while the band at ∼1595 cm^–1 belongs to NN stretching in N₂D_x species [12]. These in situ observations indicate that the N₂H_x species are formed during the NH₃ synthesis, which can support the existence of associative mechanism; (iii) Moreover, the adsorbed N₂H_x could also be detected by ex situ N K-edge NEXAFS spectra. After pre-treatment of catalyst in feed gas, the ex situ N K-edge NEXAFS spectra could be collected. The N₂H_x species is reflected by an absorption peak at 398.5 eV [66]. The combination of UV–vis, in situ DRIFTS, and N K-edge NEXAFS characterizations can solidly support the associative mechanism in thermocatalytic ammonia synthesis

In comparison with the direct dissociation of the strong N N triple bond, the stepwise breaking of the N N bond upon hydrogenation requires much lower activation energy [26,27]. The NH₃ synthesis via an associative mechanism is promising to operate under mild conditions, analogous to biological nitrogen fixation at ambient conditions. The NH₃ synthesis with relatively low temperatures and pressures can not only lower the quality standard of production equipment but also is conducive to energy conservation and CO₂ emission control, which is meaningful for the magnificent blueprint of carbon-neutrality. It has been reported that the catalysts with highly dispersed active sites, metal hydride, and metal nitride materials can realize thermocatalytic NH₃ synthesis via an associative route. With the decreased size of active metal from nanoparticles to subnanometric clusters and single atoms, the number of active sites for N₂ dissociation, such as Ru B₅ sites and Fe C₇ sites, gradually diminishes and disappears. Meanwhile, the strong intra-cluster interaction of atomic clusters and unique configuration enable the strong interactions between the d-orbitals of Ru and the s and p orbitals of N₂ molecules, thus driving NH₃ synthesis via an associative route [12,35]. Given the facile sintering of active metal below nanoscale, the construction of uniform active sites with high stability is essential to realize high-performance NH₃ synthesis. Over metal nitrides, the creation of unique nitrogen vacancy sites is the key to direct NH₃ synthesis via an associative route. The construction of metal nitrides with low E_NV value and high stability is highly desired to realize efficient NH₃ synthesis performance under mild conditions.

It is worth noting that the exploitation of catalysts following the associative mechanism for NH₃ synthesis has gained more and more recent attention. There are research fields that are promising for the development of efficient catalysts, and a deep fundamental understanding of how site structure and reaction mechanism are related is inevitable.

1. The precise construction of active sites with a specific number of atoms is a promising way to develop highly efficient catalysts for NH₃ synthesis via an associative route. Although the atomic cluster catalysts were reported to be active for NH₃ synthesis, the optimal number of atoms for a cluster to realize the maximum NH₃ synthesis rate is still uncertain. With the assistance of advanced preparation techniques, such as atomic layer deposition (ALD), it is possible to synthesize metal clusters with atomic precision [95,96]. It provides platforms to investigate the size boundary between dissociative route and associative route over atomic clusters. The aim is to achieve high performance at precise active sites and to maximize metal utilization.

2. Defect engineering over metal nitrides is a potential route to enhance NH₃ synthesis performance. In view of the activation of N₂ over nitrogen vacancies, the metal nitrides with high E_NV values are not suitable for NH₃ synthesis due to the rarity of nitrogen-vacancy sites. The construction of abundant vacancy sites over metal nitrides is conducive to performance improvement. In general, the loading of transition metal and the doping of various ions into metal nitrides, as well as the mixing of metal nitrides are effective strategies to lower the E_NV value and promote the formation of defect sites over metal nitrides.

3. To realize efficient NH₃ synthesis via an associative route, the design of multicomponent catalysts is promising. Multicomponent catalysts, such as high-entropy alloys and intermetallic compounds, can be composed of active metals as well as electronic and structural auxiliaries [97,98]. For instance, ternary ruthenium complex hydrides including Li₄RuH₆ and Ba₂RuH₆ were reported to follow an associative mechanism for NH₃ synthesis. In these catalyst systems, electrons and protons are transported by hydridic hydrogen, and N_xH_y (x = 0–2, y = 0–3) intermediates are stabilized by Li/Ba cations. With the synergistic effect of different compositions, high performance of NH₃ synthesis was realized via an associative route. Besides, given that the biological fixation of nitrogen occurs at ambient conditions via an associative route [99,100], the construction of active sites imitating the composition of nitrogenase enzyme is a prospective approach to realize efficient NH₃ synthesis under mild conditions.

4. The introduction of artificial intelligence into the design of catalysts provides an opportunity to develop advanced catalysts for NH₃ synthesis via an associative route. The machine learning approach has shown great potential in the design and prediction of high-performance catalysts, especially combined with DFT calculations [101]. Aiming to direct the activation of N₂ via an associative route, a rewarding approach is to select and predict special materials and/or active sites for the hydrogenation of N₂ molecules through massive screening. Meanwhile, high throughput experimentation is an efficient solution to fleetly optimize the composition of catalysts. The combination of machine learning methodology and high-throughput experimentation would achieve rapid composition optimization and predictive discovery of catalysts for NH₃ synthesis at mild conditions.

Since the first NH₃ synthesis device started to produce NH₃ in 1913, the catalysts of NH₃ synthesis have been developed for over 100 years. The reaction mechanism of industrial Fe-based catalysts via a dissociative mechanism has been verified by G. Ertl [13]. In recent years, with the development of preparation and characterization technology, more advanced materials, such as metal hydrides and metal nitrides, have been reported to obey an associative mechanism for NH₃ synthesis. However, the industrial application of NH₃ synthesis via an associative mechanism still faces the long-term challenges, mainly focusing on the accessibility and durability of catalysts. Specifically, the industrial application of catalysts with highly dispersed active sites is trapped in the metal agglomeration, i.e. the growth of metal size from subnanometer to nanometer in the long-period operation. As a result, the NH₃ synthesis mechanism is changed from associative route to dissociative route during the reaction process. Therefore, the means of confining the metal growth is crucial, such as coordination by heteroatoms or spatial confinement. In terms of metal hydrides or metal nitrides, they are subjected to the structural instability because of their sensitivity toward air and moisture to form oxides or hydroxides. Moreover, the large-scale preparation of metal hydrides or metal nitrides in industry is still tough, because the preparation requires a flow of high-pressure H₂ or NH₃ at elevated temperature (e.g. ≥ 700 °C), which severely threatens the safety of personnel and equipment. More efforts are needed to develop advanced metal hydrides and metal nitrides that are safe and readily available.

Declaration of competing interest

The authors declare that they have no conflicts of interest in this work.

Biographies

Yangyu Zhang graduated from Jinan University in 2017. He is currently pursuing a PhD degree at National Engineering Research Center of Chemical Fertilizer Catalysts, Fuzhou University. The focus of the PhD project is heterogeneous catalytic ammonia synthesis.

Jiaxin Li, graduated from Zhongyuan Institute of Technology in 2022, is currently studying at the College of Chemical Engineering, Fuzhou University, and her current main research field is Ru-based catalysts for ammonia synthesis.

Yanliang Zhou, Associate Professor, Master Supervisor at National Engineering Research Center of Chemical Fertilizer Catalysts, Fuzhou University, is a high-level introduced talent of Fujian Province. His research topic is the development of advanced catalysts for “green ammonia” synthesis.

Xiuyun Wang, Professor, is a doctoral supervisor at National Engineering Research Center of Chemical Fertilizer Catalysts, Fuzhou University. Her research focuses on the development of new ammonia synthesis catalyst at low temperature and low pressure.

Lilong Jiang, Professor, Doctoral Supervisor, is a dean of College of Chemical Engineering, Fuzhou University, and leader of National Engineering Research Center for Chemical Fertilizer Catalyst. He is focusing on the catalyst engineering of industrial-scale ammonia synthesis and high-value utilization of NH₃.