Insight into rare-earth-incorporated catalysts: The chance for a more efficient ammonia synthesis

Recent studies have suggested that rare earth (RE) elements in catalysts significantly influence the performance of the ammonia synthesis. The REs appear in various forms in the ammonia synthesis catalysts including supports (oxides, hydrides, and nitrides), promotors, and intermetallic. Besides the conventional RE oxide-supporting catalysts (mainly Ru/REO), some new RE-containing catalyst systems, such as electrode and nitride systems, could drive the ammonia synthesis via a benign Mars—van Krevelen mechanism or multi-active-site mode, affording high ammonia synthesis performance under mild conditions. These works demonstrate the great potential of RE-containing catalysts for more efficient ammonia synthesis. This review summarizes the contributions of different kinds of RE-based catalysts and highlights the function mechanism of incorporated REs. Finally, an overview of this area and the challenges for further investigation are provided.


Introduction 
Rare earth (RE) elements include 15 lanthanide elements from the atomic number of 57 to 71 and two d-block transition metals (TMs, scandium and yttrium) bearing similar electronic chemical properties with the lanthanide elements. The lanthanide RE elements share unique electronic configurations of [Xe] 4f n−1 5d 0−1 6s 2 (n = 1-15), which demonstrate variable valence states, low work function, and special acidity. The properties bring significant attention to the RE metals from a wide (RDS) [8]. High temperatures and high pressures are indispensable to ensure an accelerated N 2 activation and high catalytic activity of iron catalysts, which leads to high energy input, high cost for high-pressure equipment, and high operational risks. Therefore, the pursuit of efficient catalysts that work under lower temperatures and pressures never stops. The Ru-based catalysts are more efficient than the iron catalysts under milder reaction conditions (< 400 ℃, < 9 MPa) as they afford more prominent N 2 activation [9,10]. Modification of Ru with foreign components is essential to modulate its structure and intrinsic properties, and increase its ability to break the N≡N bond. Alkali/alkaline earth metal promotors are the most frequently applied promotors. They reduce the activation barrier of N 2 cleavage by reinforcing the electron injection from Ru to the lowest unoccupied molecular orbitals (LUMOs) of N 2 . However, these promoters generally induce a trade-off hydrogen poisoning effect, that is, a muchenhanced H adsorption which competes with N 2 adsorption/activation, leading to lower ammonia synthesis activity [11]. Meanwhile, the promotion of the N 2 cleavage is far from inadequate [12]. The incorporation of the RE metals was recently reported to resolve the conflict of promoted N 2 breakage with detrimental hydrogen poisoning. More interestingly, the RDS of the ammonia synthesis on the RE-containing electrides can even be altered to steps other than the N 2 activation. In different forms, the REs have demonstrated their priorities with different mechanisms. For instance, rare earth oxides (REOs) can upgrade the Ru catalysts by the rich oxygen vacancies [13]. Simultaneously, these Ru/REO catalysts can overcome the problem of methanation that confuses the traditional Ru/C catalysts. Different from the traditional alkaline earth oxide supports, RE-involving supports, such as RE nitrides [14] and oxynitride-hydrides [15], can promise a more energy-favourable N 2 activation (Mars-van Krevelen mechanism), which dramatically decreases the energy barrier for the ammonia synthesis. In some cases, the REs act as a favourable active site for the combination of N and H [16]. The combination of the RE sites with the TM sites could liberate the ammonia synthesis from the scaling relations that trouble the conventional Ru and Fe materials, allowing the ammonia synthesis to proceed via a low-barrier path. LaRuSi provides a new kind of Ru active site other than the widely accepted B 5 sites [17], which boosts the atom utilization of Ru. The summary of the functions and corresponding working mechanism of the RE-containing catalysts unquestionably would provide guidelines for screening innovative catalysts to overcome the issues hindering the efficient NH 3 synthesis under benign conditions. As far as we know, only very few reviews were presented concerning the RE-involving catalysts for the ammonia synthesis. For example, Feng et al. [18] summarized the catalytic performances and techniques to improve the activity of REOs supported catalyst. Zhang et al. [19] reviewed the RE-based alloys for the ammonia synthesis. Li et al. [20] briefly introduced the REcontaining catalysts in different forms with typical examples in a mini-review. Sato and Nagaoka [21] outlined the recently reported REOs-supporting catalysts in a review discussing the Ru-based oxide catalysts. Humphreys et al. [22] classified the recently reported ammonia synthesis catalysts via the type of active centers (e.g., Fe, Ru, Ni, and nitrides) in which some RE-based catalysts were included. A comprehensive review covering all kinds of RE-based catalysts, the history, and the insights into the roles that the REs play in different catalyst types is in great desire.
This review demonstrates the progress in the RE-involving NH 3 synthesis catalysts, highlighting the functions and working mechanisms of the RE elements. The usage of the RE elements in catalyst hosts (supports), promoters, and single-phase catalysts will be discussed. Finally, an outlook on the challenges and opportunities of the RE-incorporated NH 3 synthesis catalysts will be provided.

REs in catalyst supports
The realization that NH 3 is an ideal alternative clean energy carrier arouses more enormous ammonia demands. However, the intensified ammonia production means horrible energy input and carbon emission, as HBP is a well-known energy-intensive process. The catalysts that work under mild conditions are in urgent need, in which the Ru-based catalysts are more favorable due to their high efficiency. A well-performed Ru-based heterogeneous HBP catalyst is generally composed of three components, i.e., the active component (Ru cluster or nanoparticle), the support that helps disperse the active sites, and the promoter which is usually alkali/alkaline earth metals (Fig. 2). The support does more than simply a carrier for the active sites in a Ru-based NH 3 synthesis catalyst in many cases. It can electronically influence the active sites or act as the second active site. The RE-containing supports have been studied for the NH 3 synthesis, especially in the past decade, in different forms, including simple oxides, perovskites, electrides, hydrides, and nitrides. They can relieve hydrogen poisoning with the efficient N 2 activation on the TMs (especially Ru).

1 RE-containing simple oxides
Simple oxides are one of the most intensively used kinds of catalyst supports in industry. MgO is the classic support for the NH 3 synthesis. A series of REOs have been tested as the supports for the ammonia synthesis as well. As for the RE-containing catalyst support ( Table 1, Entries 1-21), the most famous one is CeO 2 [23][24][25][26][27]. The variable valence states and electronic structures of cerium (Ce) endow it with flexible redox properties. The surface of CeO 2 features rich defects (oxygen vacancies) due to the easy reducibility of Ce 4+ to Ce 3+ . The formed oxygen vacancies with rich electrons are stable, which provides the strong metal-support interaction and efficient electron donation to the loaded TMs [28]. The promotion effect is rewarded with high activity.
The studies on the supporting behaviour of CeO 2 for the NH 3 synthesis started in the 1990s when Niwa and Aika [23,24,29] investigated the supporting effects of the lanthanide oxides (CeO 2 , Sm 2 O 3 , and La 2 O 3 ) for the Ru catalysts. They revealed that the lanthanide oxides were very effective supports for the NH 3 synthesis, and the high-temperature reduction induced a significant increase in the catalytic activity. All the Ru/rare oxide catalysts showed a better activity comparted to Ru/MgO. Especially, Ru/CeO 2 afforded an activity of 6 times for Ru/MgO at 315 ℃, and it was characterized by the best durability among the lanthanide oxide-supporting catalysts [24]. A better resistance to the nitrogen poisoning which is caused by the strong interaction of N with Fe [30] [25][26][27]32]. Early studies mainly focus on the investigation of the effects of preparing parameters [25,32,33]. It was later realized that the reduction of CeO 2 and the strong interaction between Ru and CeO 2 are vital to the high activity of Ru/CeO 2 [26,27,34].
Recently, the promoting effect of CeO 2 in Ru/CeO 2 was clarified based on the studies on the morphology effect of CeO 2 . The content of surface oxygen vacancy was determined to be one of the most decisive factors of the activity. For example, Ma et al. [13] recently synthesized CeO 2 nanocubes, nanorods, and nanoparticles (abbreviated as c-CeO 2 , r-CeO 2 , and p-CeO 2 , respectively) with a similar size of 100-300 nm (Fig. 3) by using a modified hydrothermal method, and those morphologically different CeO 2 were used as supports to disperse 4 wt% Ru. As shown in Fig. 4(a), Ru/r-CeO 2 shows a much better activity (3830 μmol·g −1 ·h −1 ) than Ru/c-CeO 2 (1289 μmol·g −1 ·h −1 ) and Ru/p-CeO 2 (529 μmol·g −1 ·h −1 ) at 400 ℃ and 1 MPa. Ru/r-CeO 2 was characterized to show higher Ru dispersion (40.4%) than Ru/c-CeO 2 (23.1%) and Ru/p-CeO 2 (36.3%), which was attributed to the differences in surface vacancies. Besides, the activation energies (E a ) of these catalysts were 108 kJ·mol −1 for Ru/r-CeO 2 , 124 kJ·mol −1 for Ru/c-CeO 2 , and 132 kJ·mol −1 for Ru/p-CeO 2 , demonstrating that the morphology of the CeO 2 support affected not only the Ru dispersion but also the intrinsic kinetics. Both Raman and X-ray photoelectron spectroscopy (XPS) spectra indicated that the surface vacancies induced strong metal-support interaction, leading to the formation of Ru-O-Ce and Ru n+ species which ensured an efficient N 2 dissociation. As revealed in transmission electron microscopy (TEM) images, dominating that exposed planes of c-CeO 2 , r-CeO 2 , and p-CeO 2 are (100), (110), and (111) planes, respectively. Those three kinds of surfaces possess different degrees of defects (oxygen vacancies), which are confirmed by the different Ce 3+ /Ce 4+ ratios (0.59, 0.46, and 0.34 for r-CeO 2 , c-CeO 2 , and p-CeO 2 , respectively). The more defects in r-CeO 2 and Ru-O-Ce/Ru n+ on Ru/r-CeO 2 were rewarded by a higher amount of CeH 2+x which could transfer electrons to Ru efficiently [35]. The authors also discovered that alkali (Cs and K) and alkaline earth metals (Ba) were efficacious in improving the activity of Ru/CeO 2 . As indicated by the XPS investigations, Ru 0 increased significantly while the Ce 3+ /Ce 4+ ratio decreased after introducing the promoters, suggesting that those promoters reinforced the electron transferred from CeO 2 to Ru, and led to a more effective N 2 dissociation. We believe that the electron transfer from the promotors to Ru should also help. The activity of Cs-Ru/CeO 2 reached up to 33,500 μmol·g −1 ·h −1 at 400 ℃ and 3 MPa. Lin et al. [36] also studied the correlation among the CeO 2 morphology, oxygen vacancy, adsorption property, and catalytic activity of Ru/CeO 2 . Similarly, they also demonstrated r-CeO 2 to be a better support than c-CeO 2 ( Fig. 4(b)), and the NH 3 synthesis activity was closely related to the surface oxygen defects of CeO 2 . However, the authors concluded that the surface vacancy concentration was not simply determined by the Ce 3+ /Ce 4+ ratio as they observed that the Ce 3+ ratios in r-CeO 2 and c-CeO 2 resembled each other, which was different from the observations by Ma et al. [13]. They attributed the different amount of oxygen defects in Ru/r-CeO 2 to the presence of highly dispersed Ru that constitutes Ru-O-Ce species with CeO 2 . The higher activity of Ru/r-CeO 2 was due to the better dispersion of the Ru nanoparticles and the chemical state of Ru (Ru n+ instead of Ru 0 ). The Ru n+ species favoured the hydrogen adsorption via the formation of hydroxyl groups. Besides, the increased oxygen vacancies on CeO 2 were favourable for the adsorption of dinitrogen. Therefore, the combined accelerated elementary steps resulted in enhanced catalytic activity. In contrast, the Ru particles on c-CeO 2 were large and were dominated by Ru 0 , leading to inferior H 2 adsorption. The fewer oxygen vacancies resulted in the worse nitrogen adsorption. Consequently, Ru/r-CeO 2 showed a superior NH 3 synthesis activity compared to Ru/c-CeO 2 . The same group also studied the morphology effect of CeO 2 on the activity of Co/CeO 2 [37]. Polyhedral, nanorod, and hexagonal shaped CeO 2 were compared, and the polyhedral one, which showed the strongest metalsupport effect, was revealed to be the most efficient support. The high concentration of Ce 3+ and low binding energy of Co species were supposed to be the key factors determining the catalytic activity. Interestingly, Liu et al. [38] reported the reverse activity trend of Ru/r-CeO 2 and Ru/c-CeO 2 . They showed that Ru supported on cubic CeO 2 (c-CeO 2 ) afforded higher catalytic activity (27,000 μmol·g −1 ·h −1 ) than Ru supported on the microspheres (p-CeO 2 , 21,000 μmol·g −1 ·h −1 ) and nanorods (r-CeO 2 , 15,000 μmol·g −1 ·h −1 ) at 450 ℃ and 3 MPa. In contrast to the results observed by Lin et al. [36] and Ma et al. [13], their experiments revealed that c-CeO 2 , instead of r-CeO 2 , was featured with the most oxygen vacancies. We believe that it was due to the different preparation methods of Liu et al. [38] who used NH 4 HCO 3 instead of NaOH as the precipitating agent for preparing c-CeO 2 . The Ru metals on c-CeO 2 were more electron-rich than those on r-CeO 2 , which promoted the back donation of the electrons from Ru to N 2 . Furthermore, the better dispersion of Ru originated from the suitable pore size distribution of c-CeO 2 provided more the Ru active sites for the NH 3 synthesis and thus higher activity.
Wang et al. [39] developed a "yolk (carbon)-shell (CeO 2 )" support to carry Ru for the NH 3 synthesis ( Fig. 5(a)). They found that the chemical states of Ru and the metal-support interaction could be adjusted by varying the particle size of CeO 2 . By depositing CeO 2 and Ru on carbonaceous spheres with average sizes of 80, 210, and 130 nm followed by carbon removal, the Ru/CeO 2 catalysts (Ru/CeO 2 -7, Ru/CeO 2 -9, and Ru/ CeO 2 -11) with the yolk-shell structure and average diameters of 95, 225, and 150 nm were prepared. Ru/CeO 2 -9 was characterized to have the highest surface area and smallest Ru particle size. Upon the H 2 treatment, the metal-support interaction was enhanced, and more Ru-H and Ru 0 were detected on Ru/CeO 2 -9. More deeply reduced CeO 2 in Ru/CeO 2 -9 could transfer more electrons to Ru, which led to an efficient N 2 activation. The apparent E a was measured to be 90±5 kJ·mol −1 , much reduced compared to those of Ru/CeO 2 -7 (115±4 kJ·mol −1 ), Ru/CeO 2 -11 (134±4 kJ·mol −1 ), and Ru/CeO 2 (Ru/CeO 2 -CP, 140±4 kJ·mol −1 ) prepared with the traditional co-precipitation. Consequently, Ru/CeO 2 -9 showed a better activity It can be seen from the above reports that, although the morphology effect of CeO 2 is controversial, the consensus has been reached that the surface oxygen vacancy on CeO 2 plays a vital role in the catalytic performance for the NH 3 synthesis. Those defects can induce strong metal-support interaction and facilitate the N 2 activation. Li et al. [40] also drew the same conclusion by comparing three kinds of Ru/CeO 2 in which the CeO 2 samples were obtained by the incipient impregnation method using tetra-propylammonium hydroxide (TPAOH), ethylenediamine, and NaOH as the precipitant. Ru/CeO 2 -TPAOH with more oxygen vacancies and basic sites afforded the best activity.
It is well accepted that the presence of foreign elements in the CeO 2 lattice results in more defect sites [41,42]. Motivated by this understanding, Ma et al. [43] designed a series of oxygen vacancy-rich CeO 2 by doping Zr 4+ and used them as supports for the NH 3 synthesis. Raman spectra, UV/vis diffuse reflectance measurement, photoluminescence emission spectroscopy, and XPS confirmed that the Zr 4+ -doping induced the formation of more oxygen vacancies by releasing lattice oxygen atoms, and excess electrons were left behind on the Ce and oxygen vacancies. The enrichment of the oxygen vacancies surely contributed to the strong electronic metal-support interaction, leading to the upward shift of the d-band center of Ru and thus the more efficient N 2 activation. As a consequence, Ru/Ce 0.6 Zr 0.4 O 2 showed a much higher activity of 1696 μmol·g −1 ·h −1 and a smaller E a of 93 kJ·mol −1 than Ru/CeO 2 (262 μmol·g −1 ·h −1 and 121 kJ·mol −1 ). The influence of La doping has been studied by Luo et al. in 2009 [44]. They found that the presence of La could facilitate the reduction of surface CeO 2 , and those partially reduced CeO 2 created active Ru for the N 2 activation. The same group also demonstrated a similar role in Sm-doping [45]. La 0.5 Ce 0.5 O 1.75 , which can be regarded as La-doped CeO x , was evaluated by Ogura et al. [46,47] for the NH 3 synthesis as the support. They found that Ru/La 0.5 Ce 0.5 O 1.75 _650red (catalyst reduced at 650 ℃) was much more active than Ru supported on the single-RE oxides (Ru/La 2 O 3 and Ru/CeO 2 ). The catalytic activity of La 0.5 Ce 0.5 O 1.75 _650red was evaluated to be 31,300 μmol·g −1 ·h −1 while it was only 17,200 μmol·g −1 ·h −1 and 10,800 μmol·g −1 ·h −1 for Ru/ CeO 2 _650red and La 2 O 3 _650red, respectively, at 350 ℃ and 1 MPa (Figs. 6(a) and 6(b)). These REbased Ru catalysts are all superior to the benchmark Cs + /Ru/MgO catalyst with a catalytic activity of 4100 μmol·g −1 ·h −1 .
The hydrogen treatment process was of vital importance for the activation of Ru/La 0.5 Ce 0.5 O 1.75 _ 650red. A strong metal-support interaction was verified by the fact that the Ru particles were partially covered by reduced La 0.5 Ce 0.5 O 1.75 . As revealed in Fig. 6(c), Ru/La 0.5 Ce 0.5 O 1.75 _650red exhibited an E a of as low as 64 kJ·mol −1 , comparable to those of the most advanced Ru/electrides [12,48,49], TM-LiH catalysts [50,51], etc. It also demonstrated a robust immunity to the hydrogen poisoning ( Fig. 6(d)). The catalytic activity of La 0.5 Ce 0.5 O 1.75 _650red increased from 13,400 to 31,300 and 44,400 μmol·g −1 ·h −1 when the pressure was increased to 1.0 and 3.0 MPa, respectively. On the contrary, the activity of Cs + /Ru/MgO decreased when the reaction pressure increased from 0.1 to 1 MPa. Similar promotion of the La-doping was also revealed on the Ru/La 2 Ce 2 O 7 catalyst [52]. The introduction of Ba in Ru/La 0.5 Ce 0.5 O 1.75 could further enhance the catalytic activity [53]. The NH 3 synthesis rate reached 52,300 μmol·g −1 ·h −1 under mild conditions (350 ℃ and 1 MPa), which ranks among the best Ru catalysts reported so far. The catalyst evolution during the reduction process was proposed based on systematic experiments. As illustrated in Fig. 7, the Ru particles were mostly covered with Ba carbonate species when calcinated at 500 ℃ while the carbonate species were removed at 700 ℃ , forming low-crystalline nanocomposites containing Ba 2+ , Ce 3+ , and La 3+ . This process was accompanied by O 2− loss, i.e., the formation of defects. The accumulation of this nanocomposite provided a strong electron-donating system to the contacting Ru atoms and an expedited N 2 activation. Wu et al. [54] presented that the Ti-doping was also effective in tuning the physicochemical property of CeO 2 and the metal-support interaction in the Ru/CeO 2 catalyst. The presence of Ti led to variations in the morphology and dispersion of Ti and Ce, resulting in different electronic metal-support interactions. Ru/CeO 2 (Ti-Ce-S), in which Ti accumulates on the surface, showed higher activity compared to Ru/CeO 2 (Ti-Ce-E) where Ti was embedded in CeO 2 .
CeO 2 also demonstrates a distinct promoting effect for non-Ru catalysts. For instance, Lin et al. [55] reported that the pre-treatment of Co/CeO 2 in different atmospheres affects its NH 3 synthesis activity by impacting the dihydroxylation pathways. During the H 2 treatment, most hydroxyl groups reacted with hydrogen and were released as water. In contrast, hydroxyls transferred to oxides when the catalyst was subjected to an oxygen atmosphere, and they were regenerated in the H 2 reduction. The rich hydroxyls on CeO 2 retarded the exiting of redundant H atoms, which hindered the further activation of N 2 . Consequently, the oxygen-pre-treatment catalyst showed an inferior activity. Wang et al. [56] developed a dopaminemediated method to strengthen the metal-support interaction between the Co particles and the CeO 2 support, aiming to amplify the stability of Co/CeO 2 . This method can be described as: (i) a dopamine treatment to cover Co; (ii) a thermal treatment to enhance Co-CeO 2 interaction; and (iii) removal of dopamine by calcination. It was found that the dopamine treatment upgraded both catalytic activity and stability of Co/CeO 2 (Fig. 8). The E a value of the NH 3 synthesis was reduced from 107 to 71 kJ·mol −1 , while the activity increased from 3800 to 19,120 μmol·g −1 ·h −1 at 425 ℃ and 1 MPa. The optimal catalyst was featured with a constant activity over 50 h of continuous reaction. Tsuji et al. [57] demonstrated that Co-Mo/CeO 2 prepared in the presence of a strong reducing agent sodium naphthalenide (NaNaph) exhibited an activity ca. 20 times that of the classical Co 3 Mo 3 N catalyst with respect to Co and Mo weight. An excellent durability of 120 h was also achieved for Co-Mo/CeO 2 (NaNaph). Humphreys et al. [58] declared that proton-conducting oxide BaZr 0.1 Ce 0.7 Y 0.2 O 3−δ surpassed MgO-CeO 2 when used to support Ni for the NH 3 synthesis. They also proposed that upon introducing the extrinsic oxygen vacancies via doping Sm in CeO 2 or doping Y 2 O 3 in ZrO 2 , strong metal-support interaction (SMSI) may appear between Fe and these oxides, thereby enhancing the activity and stability [59].
As described above, other RE oxides, e.g., Sm 2 O 3 and La 2 O 3 , have been tested for the NH 3 synthesis as the support in the early 1990s, which displayed an inferior promoting effect to CeO 2 [23,24]. Lately, the supporting behaviour of PrO x was investigated. PrO x was declared to be a more effective support than CeO 2 for the NH 3 synthesis [60,61]. Sato et al. [61] reported that Ru/Pr 2 O 3 , in the absence of other promoters, showed higher catalytic activity than Cs-Ru/MgO (one of the most active Ru-based NH 3 synthesis catalysts), Ba-Ru/AC (widely applied industrial catalyst), and Ru/C12A7:e − (Ru/C12A7:e − is introduced in Section 2.3 under both atmospheric pressure and 1 MPa (Figs. 9(a) and 9(b)). The activity of Ru/Pr 2 O 3 reached the thermodynamic equilibrium at 400 ℃ and 0.1 MPa. It was 1.6 times that for Cs-Ru/MgO and 2.4 times that for Ba-Ru/AC. To understand the origin of the high activity of Ru/Pr 2 O 3 , the authors systematically compared it with Ru/MgO and Ru/CeO 2 . Judging from the NH 3 synthesis rate, Ru/Pr 2 O 3 transcended Ru/MgO and Ru/CeO 2 in the tested temperature range (310-390 ℃) ( Fig. 9(c)). Scanning transmission electron micrograph (STEM) characterization showed that Ru was located on Pr 2 O 3 in the form of low-crystalline Ru nano-layers instead of the nanoparticles as on MgO and CeO 2 (Figs. 9(e)-9(g)). The unique nanolayered structure exposed more step-and-terrace sites resembling the B 5 site that was effective for N 2 cleavage. Besides, Ru/Pr 2 O 3 displayed a much stronger basicity than the other two, as concluded from the CO 2 -temperatureprogrammed desorption (TPD) ( Fig. 9(d)). The unique layered Ru and strong basicity promoted N 2 dissociation synergistically, resulting in an enhanced catalytic activity. The same group then gave systematical research on the reaction kinetics of the above three catalysts [60]. For all catalysts, the RDS is confirmed to be the N 2 dissociation. The H 2 reaction order of Ru/ Pr 2 O 3 is higher than the other two, implying that hydrogen poisoning on Ru/Pr 2 O 3 is retarded. Zhang et al. [62] recently declared that Ru/Sm 2 O 3 underwent an activation process under ammonia synthesis conditions. The activity increased from 13,596 to 32,214 μmol·g cat -1 ·h -1 under 400 ℃ and 1 MPa. Mechanism studies revealed that Sm-H hydride formed on the surface of Ru/Sm 2 O 3 under the ammonia synthesis conditions. Cooperating with the Ru cluster, the surface Sm-H species could reduce the energy barrier for the nitrogen dissociation, and they directly participated in the formation of ammonia on the Ru clusters. The synergy led to the rise in the activity.
The control on the single-metal oxide is relatively limited, and the oxide with multi-metals seems to be more promising as the regulable support, as concluded by the discussion over LaCeO x [46,47]. Although we classify LaCeO x as La-doping CeO 2 , they are actually single-phase oxides with two kinds of cations. The application of these multi-metal oxides may offer more possibilities for the high-performance supporting catalysts. In this respect, perovskites have recently drawn attention from the NH 3 synthesis studies [63,64].

2 RE-incorporated perovskites
Being a special kind of oxides, ABO 3 -structured perovskite provides a good platform as a tuneable functional material for the catalysis. The electronic properties of perovskite could be regulated by varying either A or B sites, which has recently drawn attention from the field of the NH 3 synthesis (Table 1, Entries 22-32). For example, perovskites such as CaTiO 3 [63], SrTiO 3 [63], BaTiO 3 [64], and BaZrO 3 [65] have been studied to support Ru to catalyse the NH 3 synthesis. Yang et al. [66] firstly reported the RE-containing perovskite support, BaCeO 3 , for the NH 3 synthesis in 2010. Ru/BaCeO 3 was compared with Ru supported on the conventional γ-Al 2 O 3 , MgO, and CeO 2 supports. Although the Brunauer-Emmett-Teller (BET) surface area of Ru/BaCeO 3 was only 5%-10% of other catalysts, it gave the highest activity. It offered an activity of 6 times that for Ru/CeO 2 under mild conditions (300 ℃ and 3 MPa). Under the same condition, Ru/γ-Al 2 O 3 and Ru/MgO only gave an activity of 1% of Ru/BaCeO 3 . When 10% Ce was substituted by Y, the activity of Ru/BaCeO 3 was improved to 1.6 times at 425 ℃ and 3 MPa [67]. The higher activity of Ru/BaCe 0.9 Y 0.1 O 3−δ was ascribed to the enhancement of the electronic conductivity by doping with Y 3+ . The comparison of Ru/BaCe 0.9 La 0.1 O 3−δ , Ru/BaCe 0.9 Y 0.1 O 3−δ , and Ru/ BaCe 0.9 Pr 0.1 O 3−δ was conducted by Li et al. [68] They found that the Ru/BaCe 0.9 La 0.1 O 3−δ catalyst generated more oxygen vacancies than the other two catalysts, and thus more efficient electrons transferred from BaCe 0.9 La 0.1 O 3−δ to Ru. The strong metal-support interaction reduced the supported Ru size from 6.18 to 2.99 nm, ensuring more active B 5 -type sites. Moreover, the strong basicity of BaCe 0.9 La 0.1 O 3−δ provided the prominent electrons transfer from the support to Ru. As a result, 2.5% Ru/BaCe 0.9 La 0.1 O 3−δ exhibited a higher NH 3 synthesis rate (34,000 μmol·g −1 ·h −1 ) than 2.5% Ru/BaCe 0.9 Y 0.1 O 3−δ (30,000 μmol·g −1 ·h −1 ), 2.5% Ru/ BaCeO 3 (28,000 μmol·g −1 ·h −1 ), and 2.5% Ru/ BaCe 0.9 Pr 0.1 O 3−δ (28,000 μmol·g −1 ·h −1 ) at 450 ℃ and 3 MPa. Shimoda et al. [69] revealed that replacing Zr of 10 mol% with Y in BaZrO 3 slightly increased the NH 3 synthesis activity of Ru/BaZrO 3 . In contrast, doping Y into BaCeO 3 led to the activity decrease of Ru/BaCeO 3 . The BaCeO 3 supports with different basicity were prepared for the NH 3 synthesis by Li et al. [70] by calcining the precursors at 800 ℃ (BaCeO 3 -a) and 900 ℃ (BaCeO 3 -b). BaCeO 3 -a exhibited a higher basici density compared to BaCeO 3 -b, which was proportional to the NH 3 synthesis activity of Ru/ BaCeO 3 . Therefore, the basicity was considered to be a key factor in determining the catalytic performance. Smaller Ru particles and more oxygen vacancies were observed on BaCeO 3 -a, and the better activity of Ru/BaCeO 3 -a should be the result of the synergy.
Investigations on the perovskite-based catalysts are focused on replacing either A or B elements, while the modification via the substitution of O is quite scarce. Tang et al. [71] found that substituting partial oxygen with H in perovskites (SrTiO 3 , CaTiO 3 , and BaTiO 3 ) could dramatically upgrade their promoting effect for Ru, Fe, and Co in the NH 3 synthesis (Figs. 10(a) and 10(b)). For example, the activity of Co increased by a factor of 400 times when the BaTiO 3 support was replaced by BaTiO 2.73 H 0.63 . The efficient promoting effect of these perovskite-type hydrides originated from the strong electron-donating ability and the hydrogenbased Mars-van Krevelen mechanism that counteracted the hydrogen poisoning. In 2019, a novel Ce-containing perovskite oxynitride-hydride, BaCeO 3−x N y H z , was reported and used for the NH 3 synthesis [15]. The synthesis of BaCeO 3−x N y H z was performed by simply calcinating the mixture of CeO 2 and Ba(NH 2 ) 2 in NH 3 gas flow. The high temperature of over 800 ℃ for the synthesis of normal perovskites is found not indispensable for BaCeO 3 − x N y H z . Surprisingly, the TM-free BaCeO 3−x N y H z could catalyse the NH 3 synthesis with the continuous activity ( Fig. 10(c)). The formation mechanism of this special oxynitride-hydride was deduced. N and H were proved to exist in the form of N 3− and H − . As a support, BaCeO 3−x N y H z showed a much more remarkable promoting effect than BaCeO 3 . A volcanic-type correlation between the catalytic activity of Ru/BaCeO 3−x N y H z and the preparation temperature of BaCeO 3−x N y H z was observed, and 500 ℃ was the optimal temperature ( Fig. 10(d)). The activities of Ru (4.5 wt%)/BaCeO 3−x N y H z significantly exceeded those of Ru (4.3 wt%)/BaCeO 3 and Ru (5.0 wt%)/CeO 2 over the entire temperature range of 240-400 ℃. The advantage of Ru/BaCeO 3−x N y H z was www.springer.com/journal/40145 more striking below 300 ℃, where Ru/BaCeO 3−x N y H z transcended the benchmark Cs-Ru/MgO catalyst by one order of magnitude with respect to the activity.
Isotopic experiments revealed that the nitrogen isotopic exchange reaction proceeded faster than the NH x formation reaction, demonstrating that the NH x formation instead of the N 2 activation should be the RDS on the BaCeO 3−x N y H z -based catalysts. The promoting effect of BaCeO 3−x N y H z was even more prominent when bearing Co or Fe. Both Co/BaCeO 3−x N y H z and Fe/BaCeO 3−x N y H z acted as efficient NH 3 synthesis catalysts in the temperature range of 260-400 ℃ , while in sharp contrast, the Co and Fe catalysts supported by BaCeO 3 showed negligible activity. For example, the NH 3 synthesis rate at 400 ℃ and 0.9 MPa was 10,100 μmol·g −1 ·h −1 for 4.7 wt% Co/ BaCeO 3−x N y H z , which was 42 times higher than 4.6 wt% Co/BaCeO 3 (240 μmol·g −1 ·h −1 ). More remarkably, the reaction rate for 1.2 wt% Fe/BaCeO 3−x N y H z reached 6800 μmol·g −1 ·h −1 , 218 times higher than 1.1 wt% Fe/BaCeO 3 (30 μmol·g −1 ·h −1 ). Impressively, both Co/BaCeO 3 and Fe/BaCeO 3 were over 3 times more active than the benchmark Cs-Ru/MgO under the same reaction condition (Fig. 10(e)). These results suggested that the perovskite-type BaCeO 3−x N y H z possessed a significant promoting effect for both precious and non-precious metal catalysts. Moreover, the excellent durability of BaCeO 3−x N y H z -based catalysts was evidenced by the continuous reaction over 80 h without obvious activity degradation (Fig. 10(f)).
Different from the traditional Langmuir-Hinshelwood (L-H) dissociative mechanism for the TM (Ru, Fe, and Co)/BaCeO 3 ( Fig. 11(a)), two possible Mars-van Krevelen mechanisms were proposed for TM/ BaCeO 3−x N y H z : (i) as schemed in Fig. 11(b), lattice nitrogen species in BaCeO 3−x N y H z reacted with hydrogen atoms on the TM surface and/or lattice H − to form NH 3 , releasing electrons back to anion vacancies (V a ); and N-N bond cleavage was then accelerated at the interface between the TM and the support with V a . (ii) An associative mechanism (Fig. 11(c)), that is, the breakage of N-N triple bond occurred during the addition of hydrogen, in which lower energy barriers were needed to get over compared to that in the L-H mechanism; N 2 molecules could be activated on the V a sites followed by the reaction with lattice H − to form NNH (chemical formula of N2 molecule hydrogenated by one hydrogen atom) species, and stepwise the N-N bond breaking was proceeded by further hydrogenation; and after the NH 3 releasing, the electrons at the vacancy sites reacted with the adsorbed N 2 and H 2 to form lattice N 3− and H − ions, accomplishing the catalytic cycle. The authors deemed that such a unique reaction mechanism for the BaCeO 3−x N y H z -based catalysts should be responsible for the extraordinary catalytic performance.
Humphreys et al. [72] recently reported Sm-doped cerium oxynitrides with the formula Ce 1−z Sm z O 2−x N y (z ≤ 0.5) which showed a significant promotion effect for Fe. The doping of Sm 3+

3 RE-based electrides
Electrides are materials that accommodate the anionic electrons at interstitial crystal lattice sites (structural cavities or voids). Traditional organic electrides are generally unstable at room temperature, let alone under harsh catalytic reaction conditions. Therefore, the application of the electrides has long been suspended since the first report of the organic electride in the 1980s [73]. The electrides did not show their values in the catalysis until the emergence of the first inorganic electride, C12A7:e − [12,48]. C12A7:e − electride was prepared by the chemical reduction of the stable inorganic oxide, 12CaO·7Al 2 O 3 (C12A7) [74]. As shown in Figs. 12(a) and 12(b), the unit cell of C12A7 is featured with a positively charged framework structure consisting of 12 connected subnanometersized cages which share a mono-oxide layer and encapsulate two O 2− ions in two cages. The chemical reduction would inject four electrons into four cages by extracting two O 2− ions. That is to say, the concentration fraction of the anionic electrons is 1/3 electrons per unit cell (Figs. 12(a) and 12(c)). Ru-loaded C12A7:e − was demonstrated as the first Ru catalyst that altered the bottleneck (rate-determining step) of the NH 3 synthesis from the N 2 dissociation to the NH x formation [48]. The unique physicochemical property of C12A7:e − including low work function (2.4 eV) and reversible hydrogen storage-release capability made Ru/C12A7:e − with extraordinary N 2 activation ability and alleviated the hydrogen poisoning. Those effects endowed Ru/C12A7:e − with an extremely high NH 3 synthesis rate and TOF under mild reaction conditions (< 400 ℃ and < 1 MPa), surpassing that of industrially used Fe and Ru-based catalysts. This work sparkled the discovery of new electride materials (e.g., Ca 2 N:e − [49], Y 5 Si 3 [75,76], LaScSi [77], LaCu 0.67 Si 1.33 [78], Y 3 Pd 2 [79], and ZrPd 3 [80]) and their applications as catalysts/supports for diverse reactions [81], in which the NH 3 synthesis was the most focused one ( Table 1, Entries 33-37). Y 5 Si 3 was the first reported RE-containing electride [75]. It adopts a Mn 5 Si 3 -type structure and confines 0.79/f.u. (per formula unit) anionic electrons in the quasi-one-dimensional voids ( Fig. 13(a)). Like C12A7:e − , these holes can trap the hydrogen to form Y 5 Si 3 H. The work function of Y 5 Si 3 was measured to be 3.5 eV ( Fig.  13(b)), which was higher than that of C12A7:e − and much lower than that of Ru (4.7 eV). Thus, the electrons could be sufficiently donated from Y 5 Si 3 to Ru, ensuring an efficient back donation of electrons to anti-bonding π*-orbitals of N 2 . Ru/Y 5 Si 3 demonstrated similar kinetics for the NH 3 synthesis with the low E a (52 kJ·mol −1 ) in the absence of the hydrogen poisoning effect. Although the catalytic activity of Ru/Y 5 Si 3 cannot compete with that of Ru/C12A7:e − , it exhibited a TOF of 0.07 s −1 at 400 ℃ and 0.1 MPa, which was several times that of the benchmark catalysts Cs-Ru/MgO (0.008 s −1 ) and Ba-Ru/AC (0.003 s −1 ). More importantly, Y 5 Si 3 surmounted C12A7:e − and Ca2N:e − by its excellent moisture resistance capability ( Fig. 13(c)). Ru/Y 5 Si 3 did not degrade even when ~3% water was introduced to the NH 3 synthesis reactor between each run (Fig. 13(d)). As Y 5 Si 3 was synthesized by the arc-melting technique, the surface area was generally less than 2 m 2 ·g −1 . Consequently, the Ru particles could not be fully dispersed, and therefore the active B 5 sites were limited. When the Y 5 Si 3 nanoparticles prepared by an arc evaporation method were used to replace the hand-milled bulk Y 5 Si 3 , the catalytic activity of Ru/Y 5 Si 3 increased from 1450 to 4448 μmol·g −1 ·h −1 at 340 ℃ and 0.1 MPa [76].
LaScSi was reported as the first ternary intermetallic electride bearing two kinds of anionic electrons at different voids [77]. As indicated in Fig. 14(a), LaScSi is featured by a tetragonal structure with the space group I4/mmm. Two types of voids locate as V 1 in the La4 tetrahedra and V 2 in the La2Sc4 octahedra. The capability of accommodating electrons or H − was validated by density functional theory (DFT) and reversed approximation molecular orbital (raMO) [82,83] calculations. Thermal decomposition spectra demonstrate that 1.5 H atoms/f.u. LaScSi can be reversibly stored in LaScSiV 1 V 2 ( Fig. 14(b)). When Ru/LaScSi was subjected to the NH 3 synthesis, the catalytic activity reached up to 5300 μmol·g −1 ·h −1 at 400 ℃ and 0.1 MPa, which was comparable to that of Ru/C12A7:e − . The TOF of Ru/LaScSi was 12 times that of the commercial Cs-Ru/MgO at 400 ℃ and 0.1 MPa, revealing the pronounced promoting effect of LaScSi for Ru. The E a was estimated to be ~50 kJ·mol −1 , which is a typical feature of electride-based catalysts. The NH 3 synthesis rate increased monotonously as the reaction pressure increased (Fig. 14(c)), implying that the hydrogen poisoning effect on the traditional Ru catalysts was alleviated on Ru/LaScSi. In contrast, Cs-Ru/MgO showed a negative response to increasing the H 2 pressure. Furthermore, excellent vapor immunity was also observed on Ru/LaScSi (Fig. 14(d)). The presence of the RE elements seems indispensable in intermetallic electrides due to their liberality of offering the electrons to the void sites during the electride formation. Mizoguchi et al. [84] reported that LaNiSi, which has a different crystal structure from LaScSi, also displayed an electride character. The work function of LaNiSi was evaluated to be ∼ 3.2 eV, suggesting the potential of the efficient electron transferred to the supported Ru. The reversible hydrogen intercalation occurs at two-cavity sites (VNi 2 La 3 and VSi 2 La 3 ) in LaNiSi, and a maximum of 1.59 H atoms can be stored in one LaNiSi formula. Ru/LaNiSi showed a comparable activity to that of Ru/LaScSi.
The low work functions of the RE-containing electrides give the catalysts excellent electron donation ability to break the N 2 activation. The formation of ternary electrides benefits from the specific properties of the RE elements. The RE elements are featured with 4f electrons with low interaction with the nucleus, which provides the preconditions for the formation of electrides. The large size of the RE elements allows the formation of structural voids to accommodate the anionic electrons. These electrons ensure the electrides with very low work functions and thus effective promotion for supported TMs. On the other hand, most REs have a strong affinity to H atoms, inducing the fast hydrogen transfer and remitting the hydrogen poisoning effect. Therefore, the presence of RE is indispensable for the outstanding supporting manner of ternary intermetallic electrides. Very recently, we found that the REs in the electrides are more than an electron donor [16,85]. La in LaTMSi (TM = Co, Fe, Ru, or Mn) is a better active site for the NH x formation and NH 3 desorption than the Ru site, which can help break the scaling relations on the Ru/LaTMSi catalysts.

4 RE hydrides and nitrides (Table 1, Entries 38-47)
As concluded above, no matter in RE-based perovskite [71] or electride [12,49]-supporting Ru catalysts, the presence of the hydrides during the NH 3 synthesis is essential for an improved N 2 activation and the remission of the hydrogen poisoning. Actually, the perovskites such as BaTiO x H y and BaCeO 3−x N y H z are kinds of hydrides, and the electrides can be regarded as in-situ formed hydrides during the NH 3 synthesis. Gao et al. [50] and Wang et al. [51] also demonstrated that the alkali and alkaline earth metal hydrides could promote the TMs to break the scaling relation in the NH 3 synthesis, resulting in extraordinary activities under mild reaction conditions. Referring to the RE-containing hydride support, Mizoguchi et al. [86] first displayed that the hydrogenation of Ln (Ln = La, Ce, or Y) led to the formation of hydride-based electrides. As revealed in the electron density map of LaH 2 (Fig. 15(a)), conduction electrons accumulate at the void site and distribute anisotropically into the six adjacent La atoms. As revealed in Fig. 15(b), a distinct contribution from La is shown on the (110) plane. Therefore, LaH 2 V is an inorganic electride where the electrons locate at the center of the H t8 cube. The electride character was confirmed by the fact that the ionic bonding existed in the compound while there were no lobes in the cage originating from the lone pair electrons. The E a values were calculated to be 55 and 54 kJ·mol −1 for Ru/CeH 2+x and Ru/LaH 2+x , respectively, which were less than half of that for Ru-Cs/MgO (120 kJ·mol −1 ) (Fig. 15(c)). The two catalysts show positive H 2 reaction orders (Fig. 15(d)), illustrating the absence of the H 2 poisoning. Ru/CeH 2+x and Ru/LaH 2+x offered NH 3 synthesis rates of 3164 μmol·g −1 ·h −1 and 3386 μmol·g −1 ·h −1 , respectively, at 340 ℃ and 0.1 MPa, which preceded Ru/C12A7:e − (2021 μmol·g −1 ·h −1 ).
Serving as the supports, lanthanide oxyhydrides are superior to lanthanide hydrides for Ru [87]. Recently, a series of lanthanide oxyhydrides in the formula of LaH 3−2x O x developed by Ooya et al. [87] were applied as the supports for the NH 3 synthesis. As depicted in Fig. 16(a) (Fig. 16(b)). Kinetically, the RDS was determined to be the N 2 dissociation on Ru/La 2 O 3 while the NH x formation on Ru/LaH 2.5 O 0. 25 . The characterizations showed that the surface hydride-ion mobility should be the key decisive factor for the NH 3 synthesis. The working mechanism of hydrides was supposed to be consistent with the previous reports [49]. Detailly, the reversible hydrogen exchange on the Ru-hydride surface provided the effective electron donation for the N 2 activation and the inhibition of the H 2 poisoning. Although Ru/LaH 3 also possessed high hydride mobility, the nitridation of LaH 3 to LaN took place during the NH 3 synthesis (Fig. 16(c)), leading to the fast activity decay (Fig. 16(d)). In contrast, the presence of lattice oxygen restrained the transformation of LaH 2.5 O 0.25 to LaN under low temperatures (< 260 ℃ ), which sustained a continuous catalytic activity. In fact, the nitridation of LaH 2.5 O 0.25 also occurred over 300 ℃ . Therefore, Ru/LaH 2.5 O 0.25 fitted for low-temperature conditions. CeH 3−2x O x exhibited the same supporting manner as LaH 3−2x O x . Yamashita et al. [88] also demonstrated that lanthanide oxyhydrides GdHO and SmHO, loaded with Ru, showed outstanding NH 3 synthesis activities. For example, Ru/GdHO afforded an extremely high activity of 168,000 μmol·g −1 ·h −1 at 400 ℃ and 5 MPa, which was far beyond that of Cs-Ru/MgO (less than 10,000 μmol·g −1 ·h −1 ) (Fig. 16(e)).
Since the nitrides generally exhibit robust stability under the NH 3 synthesis conditions, the metal nitride catalysts have been extensively studied. For example, the application of uranium nitride as the NH 3 synthesis can date back to as early as the early 1960s [89,90]. In the past 20 years, TM-based ternary nitrides, such as Co 3 Mo 3 N [91-93] and Ni 2 Mo 3 N [94], have been systematically investigated and intensively reviewed [95][96][97][98] for the NH 3 synthesis. It has been realized that the lattice N counts for the activation of N 2 via the Mars-van Krevelen mechanism [99,100]. However, the reports concerning the use of the RE metal nitrides in the NH 3 synthesis were seldom presented until recently. The presence of LaN was even regarded as the cause of the activity attenuation on Ru/LaH 3 [87], as just discussed above in the hydrides section. Lately, a breakthrough was made by Ye et al. [14] that Ni/LaN was discovered to be a highly active catalyst under mild conditions. Ni supported on the LaN nanoparticle gave an activity of 5543 μmol·g −1 ·h −1 under a benign reaction condition of 400 ℃ and 0.1 MPa, which outdistanced all Ni-based catalysts and even surpassed many Ru catalysts (Fig. 17(a)). The combination of kinetic studies, isotope labeling characterization, and DFT calculations demonstrated that the high performance of Ni/LaN originated from a synergy of two active  www.springer.com/journal/40145 sites, including the Ni sites that dissociate H 2 and nitrogen-vacancy sites on LaN where the N 2 activation occurred following the Mars-van Krevelen mechanism ( Fig. 17(b)). This synergy also broke the scaling relations. Further, other RE nitrides (ReN = CeN, YN, and ScN) and early TM (Et = Zr, Ti, Ta, and Nb) nitrides were investigated as the supports for Ni [101]. As shown in Fig. 18(a), Ni/ReN catalysts, except ScN, catalysed the NH 3 synthesis with high activities. In contrast, Ni/EtN catalysts showed negligible activities. In particular, Ni/CeN (nanoparticle) afforded an activity of 6500 μmol·g −1 ·h −1 , reaching the thermodynamic equilibrium at 400 ℃ and 0.1 MPa. Robust durability was also achieved. Similar to Ni/LaN, Ni/CeN functioned in a dual-site mode, while the catalytic performance was determined by the nitrogen-vacancy formation energy (ENV). Ni/LaN and Ni/CeN shared a similar E a of ~50 J·mol −1 (Fig. 18(b)), while the Ni/YN showed a higher E a of 90 kJ·mol −1 . As indicated in the energy profiles ( Fig. 18(c)), the energy barrier of the N lattice hydrogenation step (TS1) was the highest among all the elementary steps for Ni/LaN (0.46 eV) and Ni/CeN (0.37 eV) catalysts. In the case of Ni/YN, the RDS was determined to be the hydrogenation of NH 2 species (TS3) with an energy barrier of 1.05 eV. This difference was attributed to the more difficult formation of the nitrogen vacancy on YN. Interestingly, non-loaded CeN, LaN, and YN also showed a certain activity toward the NH 3 synthesis. In particular, catalytic activities of 250 and 1450 μmol·g −1 ·h −1 were achieved for the CeN nanoparticle at 400 ℃ , 0.1 and 0.9 MPa, respectively ( Fig. 18(d)). The kinetic parameters (E a and reaction orders) of pure CeN nanoparticles resembled that of Ni/CeN (Fig. 18(e)), suggesting the similar reaction mechanism. Further isotope experiments coupled with the DFT calculations manifested that the nitrogen vacancies in CeN could activate both N 2 and H 2 during the reaction, resulting in the more pronounced catalytic performance than other non-loaded catalysts. It was recently found that Co/CeN worked via a concerted mechanism of associative and dissociative routes. The adsorbed N 2 could be activated on both Co metal and nitride support, which led to superior lowtemperature catalytic performance [102].
From the above discussion, it can be concluded that the RE elements show their great advantages in diverse kinds of the supports for the NH 3 synthesis. However, those supports work in different ways. For oxide-type RE-containing supports (including simple oxides and simple perovskites), the promoting effect relies more on the variable valence states of RE, which generates the surface oxygen vacancies. The electronic modification of the TMs via the strong metal-support interaction induced by the oxygen vacancies favors the N 2 activation. The reducibility of the RE-containing oxides may be responsible for the efficient hydrogen spillover that mitigates the hydrogen poisoning. For the electrides and hydrides, the low work function of the support ensures an efficient N 2 activation via the electron transfer to Ru, thereby resembling alkali-based promoters. Besides, the N-H combination proceeds following the Marsvan Krevelen mechanism in which H comes from the supports. The reversible hydrogen exchange ability of the electrides and hydrides retards the hydrogen poisoning via the efficient hydrogen transfer from Ru to the supports. In contrast, the N exchange ability of the REcontaining nitride supports (including BaCeO 3−x N y H z ) provides an efficient N 2 activation via an associative mechanism. The nitrides provide an essential active site for the NH 3 synthesis rather than act as a simple support. Undoubtedly, the RE elements play significant roles in deciding the above-mentioned functions of supports.

RE-containing promoters
Besides the supports, a promoter is another component that affects the activity dramatically. Alkali, alkali earth, and RE metals are the most commonly applied promoters in the NH 3 synthesis catalysts. The promoting effect of the RE oxides for the Fe-based catalysts has been studied since the 1950s [103]. Especially in the past 30 years, the concentration on the RE-based promoters intensified. For example, Wang and Sun [104] found that the RE oxides influence the reduction rate of the Fe 3 O 4 -based catalyst and catalytic activity. Qin et al. [105] studied the existing form of La 2 O 3 , Nd 2 O 3, and CeO 2 in the Fe-based catalysts and declared that they tend to form perovskite-type oxides. These studies revealed that the RE oxides could act as both electronic and structural promoters, and they could promote the low-temperature performance of the Fe-based catalysts.
The RE promoters are more intensively studied for the Ru catalysts. Murata et al. [106] have found that small amounts of lanthanide nitrates (La(NO 3 ) 3 , Ce(NO 3 ) 3 , and Sm(NO 3 ) 3 ) could significantly increase the activity of Ru/Al 2 O 3 in 1990. Le et al. [107,108] revealed that La(NO 3 ) 3 and Ce(NO 3 ) 3 were capable of upgrading Ru/sepiolite significantly. The effects of a series of Re (Re = La, Ce, Pr, Nd, and Sm) oxides on Ru/cordierite were investigated by Wang et al. [104] Therein, CeO 2 exhibited the most pronounced promoting effect because it uniformly covers the cordierite and modifies the support structurally and most significantly. Zhu et al. [109,110] compared the effects of Ba, K, and Sm on Ru/AC (activated carbon). Those three kinds (alkali, alkali earth, and RE metals) of the promoters have been all demonstrated to improve the activity of Ru/AC effectively, and the catalysts co-promoted by the three kinds of promoters outbalanced mono-and dual-promoted ones. Moreover, Sm showed a more pronounced suppression of methanation [111]. Combined analysis of H 2 TPD-mass spectrometry (MS) and H 2 in-situ-temperature-programmed surface reaction Fourier transform infrared spectrometer (TRSR-FTIR) on Ba-Ru-K/AC and Sm-Ba-Ru-K/AC indicated that Sm suppressed the carbon methanation by reducing the hydrogen adsorption on the catalysts. Besides, Sm was characterized to exist in the forms of Sm 2 O 3 and Sm 2 C 1−x . The formation of Sm 2 C 1−x may be ascribed to the strong interaction of Sm with the unsaturated bond of carbon, which further hindered the methanation of AC. Zhang et al. [112] also declared that Sm in Ru/Al 2 O 3 effectively inhibited the hydrogen poisoning effect, which was revealed by Kadowaki and Aika [11] and Lin et al. [113] as well.
La is another renowned promoter for the Ru-based NH 3 synthesis catalysts. La-Ba-K-Ru/AC was shown with high activity, excellent thermal stability, and strong resistance against the methanation (Fig. 19) www.springer.com/journal/40145 [114]. La was deemed to enhance the N 2 adsorption and suppress the H 2 adsorption. The highly dispersed Ru was partially covered by LaO x and BaO x , which led to the robust resistance to the methanation and particle aggregation. Mao et al. [115] demonstrated that Nd as well could alleviate the H 2 poisoning and improve the N 2 adsorption and Ru dispersion on Ru/Al 2 O 3 . The activity for Nd-Ru/Al 2 O 3 increased by a factor of 4 compared to the Nd-free sample. The effect of the RE elements on Ru/MgO was also studied. By steady-state isotopic transient analysis, Siporin and Davis [116] concluded that the Cs, Ba, and La species promoted Ru/MgO electronically while the modification of Cs was the strongest. However, La could assuage hydrogen inhibition that degraded Cs-Ru/MgO. The study of the Ru/La-Ba-MgO system revealed that the introduction of La in the preparation of Ba-MgO support not only altered the chemical and textural properties of Ba-MgO but also improved the ratio of the low-temperature phase (which tends to decompose to amorphous BaO at low temperatures of 500-800 ℃) to the high-temperature phase of BaCO 3 (which decomposes at high temperatures over 800 ℃) [117]. In the presence of H 2 , the low-temperature phase BaCO 3 decomposed more easily. Therefore, abundant amorphous BaO, the most effective Ba-containing promoter for Ru [118,119], formed during the NH 3 synthesis and led to enhanced activity.
The RE promoters generally function in the form of oxides, as concluded from the most cases. Lately, LaN has been shown as a promoter for Ru/ZrH 2 , with an excellent activity of 12,800 μmol·g −1 ·h −1 at 400 ℃ and 1 MPa [120]. As revealed in Fig. 20(a), the introduction of LaN into Ru/ZrH 2 dramatically enhances the NH 3 synthesis rate in the temperature range from 250 to 400 ℃. For instance, Ru/3LaN/ZrH 2 , which contains 3 wt% LaN, affords an NH 3 synthesis rate of 5600 μmol·g −1 ·h −1 at 350 ℃ and 1 MPa, which is 5.6 times that for Ru/ZrH 2 (1000 μmol·g −1 ·h −1 ). Normalized by the surface Ru atoms, the TOF is calculated to be 0.025 s −1 , 2.5 times that for Ru/ZrH 2 ( Fig. 20(b)). It suggests that LaN is indeed an effective promoter. However, excessive LaN up to 6 wt% on Ru/ZrH 2 results in the activity decline ( Fig. 20(c)). The pressure studies indicate the absence of the hydrogen poisoning on all the tested Ru catalysts, consistent with the conclusion drawn from the kinetic reaction order regarding H 2 (0.43) (Fig. 20(d)). As indicated by the Arrhenius plots in Fig. 20(e), the E a is determined to be 64 kJ·mol −1 , far less than that of Ru/ZrH 2 (104 kJ·mol −1 ), confirming the promoting effect of LaN on kinetics. The working mechanism of Ru/LaN/ZrH 2 was deduced based on the reaction kinetics, isothermal surface reaction, XAFS, and isotopic labelling experiments. *N 2 H 2 was detected as the main intermediates, indicating that the N 2 breakage follows an associative route. Therefore, the NH 3 synthesis for Ru/LaN/ZrH 2 proceeds via the associative and chemical looping route ( Fig. 20(f)). LaN could not only enhance the Ru dispersion but also facilitate the formation of *N 2 H 2 via the electron transfer from La to Ru, which accelerates the activation and hydrogenation of N 2 . As demonstrated by Ye et al. [14], the vacancies on LaN may be another vital site for the N 2 activation. Simultaneously, N 3− in LaN can react with H − from ZrH 2 to produce NH 3 and then replenish under the NH 3 synthesis conditions, implementing the chemical looping route.
Raróg-Pilecka group [121][122][123][124] conducted a series of works concerning the promoting effects of REs (La, Ce) on the catalytic performance of the Co-based catalysts. It is revealed that ceria in the Co/Ce/Ba catalyst hinders the sintering of Co species during the preparation and NH 3 synthesis conditions, and CeO 2 is also a stabilizer for the more active hexagonal closepacked (HCP) metallic cobalt under the NH 3 synthesis conditions, thereby leading to the enhanced activity [121]. The promoting effect of La in Co/La/Ba was highlighted by its influence on the textual properties [122].
Apart from those classified RE-containing supports, the RE metals can also serve as support modifiers. For instance, Ni et al. [125] reported layered double oxide supports modified with Y, La, and Ce. They revealed that all these three REs could remarkably enhance the NH 3 synthesis activity of Ru/MgAl-layered double oxides (LDO) while Y exhibited the most pronounced promotion. The promotion was supposed to result from the greater metal-support interaction coupled with stronger basic sites.

REs in single-phase catalysts
Compared to the supported catalysts and composited catalysts, the single-phase catalyst for the NH 3 synthesis is quite scarce because a synergistic effect of different components is essential for the effective activation of stable N ≡ N. However, the single-phase materials should possess advantages in the viewpoint of the catalyst stability because the active sites were implanted in the crystal lattice, i.e., the migration and sintering which occurs on the supported catalysts can be suppressed. Recent works have shown that the RE-containing singlephase catalysts have advantages more than merely stability.

1 Binary intermetallics
As early as 1976, Takeshita et al. [126] have tested 36 binary intermetallics involving REs (Ce, Gd, Pr, Tb, Dy, Ho, Er, and Th) in combination with the TMs (Fe, Co, and Ru). Some of these catalysts showed specific activities exceeding that of the doubly-promoted Fe catalyst. However, most catalysts decomposed during the NH 3 synthesis. Walker et al. [127] also revealed that CeRu 2 , CeCo 2 , and CeFe 2 evolved to the RE hydrides and TM particles under the NH 3 synthesis conditions, which may serve as the key active sites for the NH 3 synthesis. Zhu [128] proposed that AB 5 -type hydride LaNi 5 showed distinct activity at room temperature, and the transformation of LaNi 5 to LaNi 5 H was regarded as the key that provided the activity. In contrast, Ye et al. [129] found LaNi 5 underwent a surface phase separation to LaN and Ni during the NH 3 synthesis, which constructed a coreshell Ni-LaN structure ( Fig. 21(a)), providing high active sites for the NH 3 synthesis. The working mechanism of this catalyst resembled that of the reported Ni/LaN catalyst, which was discussed above in Section 2.4 [14,101]. Although LaNi 5 or LaNi 5 H are not single-phase catalysts for the NH 3 synthesis, the single-phase RE nitrides (LaN and CeN) are demonstrated to catalyse the NH 3 synthesis with good activity under mild conditions. Ogawa et al. [130] reported that a Laves phase binary intermetallic YRu 2 provided a higher activity than pure Ru toward the NH 3 synthesis (Fig. 21(b)). Judging from the TOF, both YRu 2 bulk (0.01918 s −1 ) and YRu 2 nanoparticles (0.01473 s −1 ) transcended pure Ru (0.00006 s −1 ) by a factor of 2 orders of magnitude. In YRu 2 , Ru was negatively charged (Fig. 21(c)), which provided a more efficient N 2 activation. Besides, the hydrogen poisoning was not detected for YRu 2 , which was benefited from its hydrogen uptake capability ( Fig. 21(d)). The combined effects were supposed to be responsible for the enhanced catalytic performance.

2 Ternary intermetallics
As discussed in Section 2.3, RTX intermetallics (R = REs, T = transition metals, and X = p block elements) LaScSi [77] and LaNiSi [84] can serve as excellent supports for the NH 3 synthesis. They upgrade the Ru catalysts by strengthening the N 2 activation and eliminating the hydrogen poisoning. However, they showed no activity without loading Ru. Recently, we found another ternary RTX intermetallic compound, LaCoSi, which adopted robust stability during the NH 3 synthesis and showed a pronounced activity under mild conditions (Table 1, Entry 48) [131]. Although possessing an extremely low surface area (1-2 m 2 ·g −1 ), LaCoSi afforded a high activity of 1250 μmol·g −1 ·h −1 at 400 ℃ and 0.1 MPa, which surpassed that of more Co-rich La-Co-Si intermetallics and other Co-based catalysts ( Fig. 22(b)). Structurally, LaCoSi adopted a tetragonal structure with the space group of P4/nmm ( Fig. 22(a)). Co was negatively charged, and hydrogen could be reversibly adsorbed in the La4 tetrahedra. The reversible hydrogen storage capability is similar to that of Y 5 Si 3 , LaScSi, and LaNiSi electrides. The catalytic kinetics of LaCoSi also resembled that for the electrides supporting the Ru catalysts. For instance, the E a was as low as 42 kJ·mol −1 for LaCoSi, and the reaction order with respect to H 2 was positive. LaCoSi was the first single-phase catalyst to shift the RDS from the N 2 activation to the NH x formation. A "hot-atom" mechanism was proposed for the N 2 activation on LaCoSi (Figs. 22(c) and 22(d)). Benefited from the Coulomb attraction with sublayer La, the adsorption of N 2 over Co was enhanced significantly with high enthalpy. The energy released and then heated adsorbed N 2 to a higher vibrational energy level, contributing to a promoted N 2 dissociation. The "hot-atom" mechanism was validated by the first-principles molecular dynamics simulations. Therefore, the high performance of LaCoSi was a combined result of electronic promotion and the special geometric structure. Further, we found that LaRuSi, sharing the identical crystal structure to LaCoSi ( Fig. 22(a)), was a typical electride [17]. To investigate the influence of electride character on the NH 3 synthesis performance, a comparison with CaRuSi having the same crystal structure but negligible anionic electrons was carried out (Table 1, Entries 49 and 50). LaRuSi showed a high activity of 1760 μmol·g −1 ·h −1 while negligible activity was observed on CaRuSi ( Fig.  22(e)). The work function and valence state of Ru are all alike for LaRuSi and CaRuSi. Consequently, the performance distinction was attributed to the electride character and the lattice hydrogen exchange under reaction temperatures. ReRuSi (Re = La, Ce, Pr, and Nd) electrides can all catalyse the NH 3 synthesis [132]. However, the traditional preparation technique (arc melting) for the RTX intermetallics generates the materials with low surface areas. As those Ru-based ternary intermetallics are robust to acid, ethylene diamine tetraacetic acid (EDTA) treatment was conducted to selectively etch surface RE and Si, which increased the amount of exposed Ru sites. For example, the activity of LaRuSi was lifted from 1810 to 5340 μmol·g −1 ·h −1 without interfering with the reaction kinetics of a typical electride catalyst (Figs. 22(f) and 22(g)). The improved activity was comparable to that of the Ru/electride catalysts. This EDTA-treatment method is universal for other ReRuSi (Fig. 22(h)).

3 Other single-phase RE-containing catalysts
Interestingly, Ru-doping LaCoO 3 perovskites were discovered as single-phase NH 3 synthesis catalysts. Wang et al. [133] demonstrated that the Ru-doping LaCoO 3 perovskite afforded much higher activity than the Ru-nanoparticle-loaded LaCoO 3 (Ru/LaCoO 3 ). With a similar Ru amount (~0.55 wt%), LaCo 0.99 Ru 0.01 O 3 displayed an activity of 8500 μmol·g −1 ·h −1 at 450 ℃ and 1 MPa, which was 1.7 times that for Ru/LaCoO 3 (4900 μmol·g −1 ·h −1 ). When the doping amount was increased to 0.93 wt%, the NH 3 synthesis rate could be improved to 10,500 μmol·g −1 ·h −1 . Superior durability for the single-phase catalysts was observed on LaCo 0.99 Ru 0.01 O 3 . In contrast, apparent activity attenuation was observed for Ru/LaCoO 3 . Sm 3 H 7 particles prepared via hydrogen plasma-metal reaction (HPMR) were reported to be capable of producing ammonia at 25 ℃ and 1 atm [134]. Actually, it worked by reacting with O 2 and N 2 to form Sm 2 O 3 and NH 3 , i.e., Sm 3 H 7 was a reactant instead of a catalyst. As discussed above, the RE nitrides could catalyse the NH 3 synthesis without the presence of the TMs [101]. Comparatively, a single-phase catalyst is very scarce as the catalytic performance is generally inferior to the supported catalyst. However, it provides an ideal platform for understanding the reaction mechanisms through experimental means and theoretical calculations.
In the single-phase catalysts, the REs contact the TMs more directly via the chemical bonding, suggesting a more pronounced electronic modification of the active centers (e.g., YRu 2 ). Furthermore, the REs impact the adsorption of reaction intermediates directly, as they do in LaCoSi via the Coulomb interaction. The atomic-level distribution of the REs in the single-phase catalysts leads to the excellent dispersion of the active components in situ formed during the reaction, providing abundant active sites. www.springer.com/journal/40145 5 Insight into specific roles of RE elements As summarized above, the RE elements emerge in various kinds of high-performance ammonia synthesis catalysts. In almost all cases, the RE-containing catalysts or supports are studied as integers while the specific roles of the RE elements are seldom discussed. For example, the oxygen vacancy, the electride character, or the reversible hydrogen/nitrogen storage capability of a catalyst/support are characterized and related to the catalytic performance. However, why it is the RE elements that can ensure the formation of high-content oxygen vacancy, help the formation of an electride, or promote the hydrogen storage, is rarely declared. In our opinion, the unique roles of the RE elements can be highlighted as follows: (i) the RE metals (mainly Ce) have variable valence states which ensure corresponding oxides with adjustable oxygen vacancies that provide the strong metal-support interaction with supported metals; (ii) the RE metals are characterized with very small electronegativities, and therefore the valence electrons are easily trapped by structural voids to form electrides or by oxygen vacancies. These electrons are loosely bounded in the RE-containing materials, i.e., the RE-containing materials have very low work functions, which act as efficient electronic promoters; (iii) the RE oxides are much more stable than the traditional alkali oxide promoters, which gives the RE-containing catalysts with high stability; (iv) the REs can remit the hydrogen poisoning effect on the traditional Ru/C or Ru/MgO catalysts due to their strong affinity to H; (v) the appropriate affinity of the RE elements with N atoms provides a critical active site for the N 2 activation via a low-energy-barrier Marsvan Krevelen mechanism, making new high-efficiency catalysts (e.g., Ni/LaN and BaCeO 3−x N y H z ) that work under mild conditions; and (vi) in the single-phase catalysts, the REs provide more straightforward electronic regulation via the chemical bonding to the active centers. They can also act on the adsorption directly or evolve to active species with high dispersion.

Challenges and outlook
Judging from the existing papers, the Ru-based catalysts dominate the well-performed ammonia synthesis catalysts under mild conditions (< 400 ℃ and < 10 MPa). However, due to the high price of Ru, the commercial use of Ru is severely limited. The market price of Ru was 1600 USD/kg in 2015 and increased to 18,500 USD/kg in 2021 [135]. To promote the revolution of the ammonia industry, it is urgent to increase the atom utilization of Ru in the catalysts. Contributions have been attached to controlling the size, dispersion, electronic state, and shape of Ru, as well as the electronic regulations via the supports and promotors. In terms of the reaction kinetics, the main challenge of the Ru catalysts lies in boosting the N 2 activation as much as possible without introducing the H 2 poisoning. In this essence, the REs have demonstrated their advantages in improving the utilization of Ru. They can enhance the activation of N 2 via pronounced electron donation as the key components in the supports, promoters, and single-phase catalysts. Besides, they can alleviate or even avoid the H 2 poisoning effect, which is benefited from their strong H affinity that shunts H from Ru sites. Furthermore, the construction of the multi-sites (e.g., the Ru site or nitrogen-vacancy site for the N 2 activation [14,16], the Ni site for the H 2 activation [14], the RE site for the NH x formation [85]) would settle the competition of different elementary steps on one specific site. Recent studies have also revealed that the ammonia synthesis via an associative mechanism is preferable to the dissociative mechanism for a high-performance catalyst. Some of the wellperformed RE-based catalysts (e.g., Ru/LaN/ZrH 2 [120] and Co/CeN [102]) have been proved to catalyse the ammonia synthesis via the associative mechanism, which will stimulate more explorations.
Although the RE-containing catalysts have provided the promise for the energy and cost reduction of massive production of ammonia, many challenges related to the RE-incorporated NH 3 synthesis catalysts have to be faced for industrial applications. Here, we discussed the challenges and perspectives: (1) Increasing stability of RE-incorporated electride (hydride and nitride) catalysts For industrial applications, stability, besides activity, is another significant factor. An industrial NH 3 synthesis catalyst would be used for 5-10 years before replacement. The RE-containing electrides, hydrides, and nitrides endowed the Ru catalysts with surprising high activity. However, the overwhelming majority of those supports are sensitive to water and soon decompose during use. The water impurity is inevitable in industrial NH 3 synthesis gas as H 2 is from techniques such as steam reforming of natural gas and water electrolysis. Therefore, increasing the stability of the electride and nitridebased catalysts is essential. Distinctively, the electride of LaRuSi is extremely immune to water. It can even be resistant to weak acids. Further understanding of the stability mechanism of LaRuSi is of great interest and crucial for the development of other stable electrides. Therefore, more studies are in demand.
Providing a dry microenvironment for the electrides (hydrides or nitrides) is a promising way to overcome the instability problem. For example, the Ru/electride catalyst can be embedded into hydrophobic porous SiO 2 . The SiO 2 layer can protect the electride from water while allowing the permeation of H 2 and N 2 , and it can also drive the desorption of the polar NH 3 molecule. Besides, the design of a composite consisting of an electride (hydride or nitride) component and an adjacent component with a high affinity to water can also be the potential to create a water-free active microdomain for the NH 3 synthesis.
(2) More insightful understanding of function mechanism of REs Most RE elements are more expensive than earthabundant components (e.g., CaO, Al 2 O 3 , and MgO). Maximizing the effect of the RE is important for cost reduction. The precondition to reducing RE dosage is obtaining unambiguous function mechanisms of the RE metals and realizing the precise fabrication.
The existence form, location, and structure of the RE need to be fully discovered by persuasive techniques, especially through the combination of in situ techniques and advanced simulations. Learning the structure evolution of the REs and their impacts on the central active sites during the NH 3 synthesis would deepen the insights into RE functions. As the NH 3 synthesis condition is a strongly reductive atmosphere, there may be significant differences between the working catalyst and the just synthesized (or spent) catalyst. Besides, although the REs are widely regarded as unable to activate N 2 directly, they may provide essential sites for adsorption and formation of intermediates. However, few of these kinds of studies are conducted. In this regard, in situ and modelling studies on the adsorption and evolution of the intermediates on the RE-based species, the interaction between the REs and the TMs, and the structure changes during the reaction are of great significance.
(3) Shaping of RE-incorporated catalysts Shaping is the first step from a research sample catalyst to an industrial NH 3 synthesis catalyst. The shaping is vital to reducing pressure drop along the catalyst bed, removing the radial heat, and avoiding internal and external mass transport resistance. However, some RE compounds are difficult to be shaped. For instance, CeO 2 powder is quite hard to be engineered into a designed macroscopic shape, which is inferior to the supports like Al 2 O 3 and carbon materials. Almost all the discussed catalysts in this review are evaluated in the powder form, and therefore future studies on the shaping of those catalysts are necessary. New techniques to settle the problems of hindering the shaping are also required for CeO 2 , and other RE-containing catalysts failed to be shaped via widely applied pelleting, granulation, or extrusion. The formation of composite catalysts, and the combination of various molding methods may be the future research direction.
(4) Ru-free catalysts with comparable or superior activity.
From the viewpoint of the cost, Ru is not the best choice for the catalysts. Besides means of reducing Ru usage by increasing mass activity, the non-noble metal catalyst is an ideal option. The recently reported Ni/LaN and Co/CeN systems work by a concerted mechanism of the associative route via LaN and dissociative route on Ni/Co for the NH 3 synthesis. They demonstrate comparable or even superior activity to the Ru catalysts under mild conditions. The REs are essential for the N 2 exchange and activation. More efficient RE-containing systems based on the combination of multiple active sites and multiple mechanisms are expected to substitute the expensive Ru-based catalysts.
Besides, for the development of innovative systems, the traditional "trial and error" method is too inefficient. The ascendent high-throughput screening methods (typically machine learning) are expected to make significant contributions.

Conclusions
The RE metals show their unique effects in the NH 3 synthesis catalysts. This review summarizes the advances of the RE-containing Haber-Bosch catalysts, which show outstanding performance under mild reaction temperatures and pressures, providing a chance for a more benign ammonia synthesis industry. The functions and working mechanism of the RE-containing supports (e.g., simple oxides, perovskites, electrides, hydrides, and nitrides), promoters, and single-phase catalysts are comprehensively summarized. We also offer an insight into the specific roles that the REs play in different kinds of catalysts. Moreover, the challenges of this field and the expected future attention from the scientific community are also proposed. We believe this review would give an integral understanding of the RE-involving catalysts for the ammonia synthesis and promote the development of the related catalysts.