Bimetallic Sites for Catalysis: From Binuclear Metal Sites to Bimetallic Nanoclusters and Nanoparticles

1The application of bimetallic particles is widely practiced in industrial catalysis., The solid catalysts comprising two metals, either in the form of alloys or particulate mixture, can be more efficient than the corresponding monometallic catalysts. For instance, the industrial naphtha reforming catalysts based on Pt-based bimetallic particles (e.g., PtIr and PtRe) supported on Al2O3 have played a crucial role in the conversion of alkanes into aromatics. The introduction of the second metal (i.e., Ir and Re) to Pt can remarkably improve the selectivity of aromatics and prolong the catalyst’s lifetime., In another example, bimetallic AuPd catalysts are used for the production of vinyl acetate monomer by the reaction between acetic acid and ethylene in the presence of oxygen. Pd is the main functional component, while adding Au can improve Pd particles’ stability by protecting them from sintering. In another example, the Lindlar catalyst, which is made by PdPb alloys, is used for the selective hydrogenation of alkynes to alkenes. The role of Pb is generally proposed for poisoning the Pd sites to avoid the overhydrogenation of alkynes into alkanes. Moreover, the industrial catalysts used in hydrotreating processes rely on the bimetallic PdPt particles because of their superior sulfur-resistant capability over the monometallic counterparts. These widely employed bimetallic catalysts in the chemical industry demonstrate the importance of the construction of active sites through the combination of different metals.

2Although tremendous efforts have been devoted to mechanistic studies on bimetallic nanoparticles, the reasons the bimetallic particles can deliver better or different performances than the corresponding monometallic particles are still unclear. In some reactions, the bimetallic catalysts show enhanced catalytic performances (higher activity and/or selectivity), while in some others, the bimetallic catalysts are less active or show lower selectivities toward desired products. For instance, AuPd alloy nanoparticles show significantly higher activity for oxidation of primary alcohols to aldehydes and oxidation of toluene to benzyl benzoate., Moreover, AuPd bimetallic nanoparticles also show dramatically enhanced activity for hydrogenation of levulinic acid to γ-valerolactone. The addition of Pt-group metals to Au for promoting the activity for hydrogenation reactions has also been found with AuPt catalysts for chemoselective hydrogenation of nitroaromatics. However, in some cases like CO oxidation, water–gas shift, and decomposition of formic acid to CO2 and H2, AuPd bimetallic nanoparticles are less active than pure Au nanoparticles. It is well-known that the electronic structures and the geometric structures of bimetallic nanoparticles are different from the corresponding monometallic nanoparticles. However, the correlation between reactivity and structural factors is still unknown, or at least, not fully elaborated. The two metals may both participate in a catalytic reaction by accounting for different elementary steps. In another scenario, only one metal works as the functional component while the other metal is added to modify the structure of the functional component.

3The difficulty to clarify the so-called “synergistic effects” in supported bimetallic catalysts could be caused by the complexity of the structures of bimetallic nanoparticles. Depending on the atomic arrangements of the two metal elements, a bimetallic catalyst could be in the form of a random alloy, intermetallic compound, segregated structure, or mixture. Even in a single bimetallic nanoparticle (as depicted in Figure 1c), multiple types of metal sites can be formed, depending on the spatial distribution of the two elements and their catalytic functions could be quite different. Therefore, it is very difficult to rationalize the structures of the active sites within the practical bimetallic nanoparticulate catalysts.

4In recent years, the emerging works on catalysts comprising isolated metal atoms or subnanometer metal clusters as active species have brought new insights into the size-dependent catalytic properties of supported metal catalysts, which have already been extensively studied for decades based on supported metal nanoparticles with sizes above 1 nm. By comparing the catalytic performance of isolated metal atoms, subnanometer metal clusters, and nanoparticles and correlating their structural features (geometric and electronic structures) with catalytic behaviors, a deeper understanding of the size effect in supported metal catalysts has been achieved. In this direction, we have witnessed the explosion of works, in which people are striving to control the size and coordination environment of supported subnanometer metal catalysts and to understand the structures of the active sites at the molecular and atomic levels.

5By translating this concept from monometallic systems to bimetallic systems, it can be anticipated that the size of the bimetallic sites will also have a significant impact on their catalytic behaviors. For instance, it has already been found that subnanometer PtSn clusters can show different geometric and electronic structures compared to the bimetallic PtSn nanoparticles. Such structural differences are also reflected in the structural evolution behavior at high-temperature redox treatments and the catalytic performance for propane dehydrogenation.

6From a structural point of view, compared to the conventional bimetallic nanoparticles (>1 nm), the bimetallic sites are made by binuclear metal sites (two neighboring metal atoms on the solid carrier, as depicted in Figure 1a) or bimetallic nanoclusters (≤1 nm, as depicted in Figure 1b) have relatively less complicated structures. Indeed, with the advanced characterization techniques, it is possible to fully resolve the structure of bimetallic nanoclusters. In this sense, the bimetallic sites, either in the form of binuclear atoms or small clusters, can be model systems to understand the elusive “synergistic effects” in the large bimetallic nanoparticles.

7In this review, we aim to make a comprehensive overview of the recent progress on supported bimetallic catalysts, covering the range of binuclear metal sites to bimetallic nanoclusters and nanoparticles. Considering the abundant publications and reviews on bimetallic nanoparticles, we will focus on the bimetallic sites in the subnanometer regime in terms of materials synthesis and characterizations. Moreover, we will make critical comparisons of the three types of bimetallic entities in terms of their structural features and catalytic behaviors. By summarizing the results obtained in different systems, we may gain a more unified and global understanding of the working mechanism and potential of bimetallic systems in heterogeneous catalysis.

8The interaction between the bimetallic entities and substrate molecules will be greatly influenced by the geometric and electronic structures of the bimetallic entities. In this section, we will discuss and make a critical comparison in terms of the structural features of different types of bimetallic entities.

9In binuclear complexes, the two metal atoms are coordinated by ligands and the full structures can be fully resolved by proven techniques developed in molecular science. However, when the binuclear metal species are immobilized on a surface, the binding sites on the surface can partially or fully serve as “solid ligands” to coordinate with the two metal atoms. In this sense, the geometric structure of the binuclear bimetallic sites can be determined by elucidating the local structures of the binding sites on the surface and the ligand/atom directly bonding with the metal atoms. To improve the stability of the binuclear species, the bonding interaction between the metal species and the support needs to be enhanced by appropriate thermal treatment, resulting in binuclear species free of organic ligands. Taking the carbon-based support as an example, the two metal atoms are shown to be preferentially located at the vacant sites with heteroatoms (N, S, O, or P) at the binding sites, as revealed by experimental measurements and theoretical modeling.,

10It is anticipated that the distance between the two metal atoms is dependent on the structure of the vacant site for accommodating the binuclear bimetallic species. According to the models presented in the reported works, the most stable configurations of binuclear bimetallic sites are formed in the vacant sites in the carbon matrix by bonding with the carbon atoms or heteroatoms (N, O, S, etc.) in the carbon matrix. As shown in Figure 2, several typical configurations of binuclear bimetallic sites embedded in carbon matrix show distinct local structural features, resulting in different coordination environments. One critical issue related to the structure of binuclear bimetallic sites is the interaction between the two metal atoms. In binuclear bimetallic complexes, the two metal atoms can have direct metal–metal bonding or be bridged through ligands. In the cases of supported binuclear bimetallic sites, the formation of metal–metal bonding is rarely observed. Instead, the two metal atoms are usually separated by the atom (such as O, C, N, S, P, etc.) in the support, exhibiting a longer metal–metal distance than that in conventional metallic nanoparticles.

11In principle, the electronic structures can be fully elaborated if the coordination environment of the binuclear sites can be fully resolved, although as mentioned before, it is a very challenging task for a practical solid catalyst. The coordination environment of the metal atoms on the support is not as well-defined as the ligand in the molecular complex. According to the reported examples in the literature, the metal atoms in the binuclear sites are usually positively charged with interaction through the metal–heteroatom (O, C, N, S, P, etc.) interaction. The electronic structure of the metal atoms depends on their bonding with the support and the distance between the two metal atoms. For instance, in the Pt–Co binuclear sites, both metals show a chemical state of +2, and coupling between the d orbitals of both atoms may occur, according to DFT calculations.

12Active sites based on binuclear bimetallic species offer the opportunity to establish the structure–reactivity relationship in a relatively straightforward approach. For instance, it has been revealed by density functional theory (DFT) that some fundamental parameters such as electron affinity, electronegativity, and the radii of the metal atoms embedded in the carbon matrix can be used to predict the catalytic performance for the oxygen reduction reaction. In another example, the charge depletion (Mulliken charge) and d-band center are employed as a descriptor for screening catalysts of binuclear bimetallic sites stabilized in N-doped carbon matrix for electrocatalytic H2 evolution reaction. As presented in Figure 3, by varying the metal element in the binuclear sites, the NiCo bimetallic site can upshift the d-band center and thereby facilitate rapid water dissociation (lower kinetic barrier) and proton adsorption, resulting in improved electrocatalytic activity for water splitting to H2.

13Numerous works have studied the structural features of bimetallic nanoclusters via theoretical and/or experimental approaches. For instance, according to density functional theory (DFT), Au atoms in bimetallic AumAgn clusters (2 ≤ m+n ≤ 8) tend to occupy the surface sites because of the charge transfer from Ag to Au, which is consistent with the experimental observation with bimetallic AumAgn (m+n = 34) nanoclusters., The local segregation has also been observed with other bimetallic nanoclusters, and the situation is greatly related to the composition of the bimetallic clusters, even the reported works only consider the geometric structures of “naked” bimetallic nanoclusters.

14It can be expected that the real situations will be more complicated after the introduction of ligands or solid carriers. The presence of organic ligands, especially those with strong interaction with the bimetallic nanoclusters, will block part or all of the surface sites, resulting in stable nanoclusters with low reactivity. On the other hand, by proper activation means, the ligands bonded to the bimetallic nanoclusters could be removed or replaced by some weakly bonded ligands, leaving open sites for substrate molecules.

15In the case of bimetallic nanoclusters supported on a solid carrier, the three-dimensional configuration of the bimetallic nanoclusters will determine the exposed metal sites. For relatively large nanoclusters (with more than 20 atoms), only quite a few atoms will not be accessed by the reactants unless the nanoclusters show a two-dimensional configuration on the support. Indeed, two-dimensional “raft-like” structures have been observed with bimetallic PtPd nanoclusters on CeO2. By varying the composition of the support, the geometric structures of the bimetallic nanoclusters will also change accordingly. In some cases, segregation may occur due to the different metal–support affinity of the two metal elements in the bimetallic nanoclusters.

16The electronic structures of bimetallic nanoclusters also depend on many factors including the particle size, composition of the clusters, and metal–support interaction. For instance, AgnNin (n ≤ 7) clusters show size-dependent electronic structures according to the DFT calculations. Small AgnNin (n ≤ 4) clusters show Ag-like properties, while large AgnNin (5 ≤ n ≤ 7) show Ni-like properties. The impacts of tunable electronic structures are also reflected in the interaction between the bimetallic nanoclusters and substrate molecules. By replacing one Pt atom in the Pt8 cluster with one Ru atom, the activation energy for O2 will be decreased. Electron donation from Ru atom to neighboring Pt atoms occurs because of the lower electronegativity of Ru. As a result, charge polarization in the Ru1Pt7 cluster leads to a shift of the d-band center close to the Fermi level, resulting in the enhanced capability for O2 and CO activation than the Pt8 cluster. As shown in Figure 4, the profound influences of dopants in the electronic structures of bimetallic nanoclusters have also been demonstrated with the Pt19+ cluster. The doping of a single Nb atom leads to electron transfer from Nb to Pt, giving to negatively charged Pt atoms. Consequently, CO adsorption on the Nb-doped Pt cluster is weakened due to the lower capability for accepting the electrons from CO. In contrast, the electron transfer between Sn and Pt is much weaker, resulting in a minor modification of the electronic properties of the Pt19+ cluster.

17The variation in metal–molecule interaction caused by the formation of bimetallic nanoclusters could be the consequence of changes in both electronic and geometric structures. For instance, the adsorption energy of CO on Au clusters is increased after the introduction of Rh to the Au clusters, which is ascribed to the electron transfer between Au and Rh and the occupation of less coordinated sites by Au atoms. The geometric and electronic effects are usually interconnected, as found in other systems.,

18The metal–support interaction also has a profound influence on the electronic structures of bimetallic nanoclusters because of the charge transfer between the nanoclusters and the support. For example, after deposition of Pt atoms on SnOx/TiO2 surface, in which the SnOx species exist as single-layer structures, SnOx is partially transformed into metallic Sn, giving to the formation of Pt–Sn bonding. The electron transfer from Sn species to the TiO2 support causes the dipole–dipole repulsion of Sn atoms and the sintering of the PtSn clusters is therefore suppressed. When the support is changed from inorganic to organic carrier (e.g., polymers, metal–organic frameworks), the electron-transfer from the functional groups to the bimetallic nanoclusters will be largely different, although the electronic features of the polymer-encapsulated and MOF-encapsulated bimetallic nanoclusters have not been systematically compared with those supported on conventional inorganic carriers.,

19The degree of complexity for the geometric structures of bimetallic nanoparticles further increases in comparison with bimetallic nanoclusters. The bimetallic nanoclusters usually show disordered and flexible geometric structures because of the lack of enough metal–metal bonding. In the case of bimetallic nanoparticles, ordered geometric structures with crystalline features are formed, which leads to the higher stability of bimetallic nanoparticles than the nanoclusters. Of course, there are also some reports on bimetallic nanoparticles with an amorphous structure, in which the spatial arrangements of the metal atoms are not as ordered as the crystalline structure. It can be expected that the structural stability of the amorphous bimetallic nanoparticles will lie between the crystalline nanoparticles and the nanoclusters.

20As illustrated in Figure 5, the presence of more atoms in bimetallic nanoparticles also leads to the complexity in terms of the spatial distribution of the two metal elements within the nanoparticles. In an ideal scenario, the location of the two metal elements in intermetallic nanoparticles can be determined according to the crystallographic structure. However, in most of the real systems comprising bimetallic nanoparticles, the two metal elements may form various types of spatial distribution, such as random alloy, core–shell structure, and Janus-type structure. Moreover, a mixture of the above-mentioned distribution pattern may coexist in a single bimetallic nanoparticle, as observed in experimental and theoretical studies, making it a great challenge to precisely describe the spatial distribution of the metal elements.,

21To simplify the situation, one may focus on the surface or subsurface structural features of the bimetallic nanoparticles because these regions are more relevant in catalysis. In practical catalysts, the surface of a bimetallic nanoparticle is usually rough, with multiple types of surface structures such as terrace, corner, flat facet, and defective sites. The coordination number of the surface atoms has been proposed as a key factor that determines the activity, because the adsorption of reactant and intermediate on the surface atoms is influenced by the neighboring atoms. It has been revealed in the literature that, for a specific catalytic reaction, it may follow different reaction pathways when the reactant is adsorbed on different surface sites. The site-specific catalytic behavior has already been established in some monometallic system, and it can be expected that such behavior/phenomenon will be more complicated when dealing with the bimetallic system.

22Nevertheless, the combination of bimetallic nanoparticles with a solid carrier will further increase the complexity of the geometric structural features of bimetallic nanoparticles due to the formation of metal–support interface. The chemical bonding between the bimetallic nanoparticles and the support is related to the chemical properties of the two components. For instance, metal–O bonding can be present in oxide-supported bimetallic nanoparticles. When the support is changed to carbon-based materials, the metal–C/N/S bonding may serve as the linkage at the metal–support interface. Besides, due to the difference in metal–support affinity of the two metal elements, local segregation may appear at the interface between bimetallic nanoparticles and the support.

23Because of the above reasons, the theoretical modeling of a bimetallic nanoparticle is not an easy task. There are already some studies on modeling of the morphology and possible segregation of the bimetallic nanoparticles under various conditions. For instance, a genetic algorithm is employed to predict the composition of thermodynamically stable bimetallic nanoparticles comprising tens of thousands of atoms. Because it is very difficult to measure the three-dimensional atomic structures of bimetallic nanoparticles, a fast modeling approach is helpful to give an estimation of the plausible geometric structure and composition ordering (see Figure 6). Moreover, the simulated structures of bimetallic particles can be used as models to correlate with the proposed structures derived from the experimental characterizations, as will be further discussed later in this review.

24Different to bimetallic nanoclusters whose electronic structures can be varied markedly by changing one atom, the variation of the electronic structures of bimetallic nanoparticles will be less sensitive to the change of particle size because their energy diagram become continuous while the small nanoclusters are not. Usually, noticeable changes are observed with bimetallic nanoparticles comprising plasmonic metals (such as Au, Ag, and Cu) when modulating the particle size because the d-band centers of these metals are far than the Fermi energy level. Such effects are not significant for platinum group metals because their d-band centers are close to the Fermi level. The influence of the composition is, to some extent, more remarkable than particle size. As shown with AuAg nanoparticles, the peak of the adsorption band in UV–vis spectra can be modulated from 400 nm (pure Ag) to 550 nm (pure Au) when tuning the composition of the bimetallic nanoparticles. The shift of the band positions in UV–vis adsorption spectra is caused by the shift of the d-band center of the bimetallic nanoparticles, which is induced by mixing Au and Ag.

25Besides, the influence of the morphology of the bimetallic nanoparticles on electronic structures cannot be neglected, especially for plasmonic metals. The position of the plasmonic adsorption band of AuPd nanoparticles will shift when tuning the shape of the bimetallic nanoparticles from sphere to nanorods. The morphology-dependent electronic structures and related optical properties will be reflected in the catalytic applications, which will be further discussed in this review.

26To correlate the electronic structures of bimetallic nanoparticles with the metal–reactant interaction or performance in catalytic tests, some descriptors are proposed to establish a concise relationship, among which the d-band model is widely used for evaluating the formation and stabilization of intermediates on the metal surface. The center of the d-band electron energy diagram (density of states) can correlate with the adsorption of the reactant/intermediate on a specific metal surface.

27The geometric and electronic structures of the three types of bimetallic entities are greatly different from each other. Herein, we would like to emphasize the difference in terms of the number of exposed metal sites and the chemical states.

28Limited by the atomicity, the number of metal atoms for adsorption and activation of reactants in the isolated binuclear sites will be quite limited. Therefore, in many cases, it is proposed that the support will participate and play a critical role in the catalytic process. Because there are multiple metal atoms exposed in the nanoclusters and nanoparticles, the adsorption and activation of reactants can be achieved by the synergy of multiple metal atoms. Certainly, the support in the metal catalysts comprising bimetallic nanoclusters or nanoparticles can also make a significant impact, but their role is not as indispensable as that of the catalyst based on binuclear sites.

29As a result of the bonding interaction with the support for stabilization, binuclear bimetallic sites are usually positively charged due to the charge transfer from the metal atoms to the support. For instance, Pt(II) and Zn(II) are formed in the supported binuclear Pt–Zn sites prepared via a surface organometallic approach. In the case of bimetallic nanoclusters, the chemical states may vary from one atom to another within a single cluster, depending on the atom’s coordination environment. Due to the presence of metal–metal bonding, the majority of the atoms will be or close to the metallic state, while the metal atoms at the interface between the cluster and the support/ligand may become positively charged due to the electron transfer from the cluster to support/ligand. For bimetallic nanoparticles, the vast majority of the metal atoms are in the metallic state, except for those on the surface or at the metal–support interface, whose chemical states could be modified by the ligands/reactants/support.

30In addition to metal–support interactions, the structural features of bimetallic entities are also greatly influenced by the metal–reactant interactions, which will be reflected in their stability and associated structural transformations under the conditions of catalytic reactions. For instance, it has been shown in numerous systems made by bimetallic nanoparticles that segregation of the two metal elements occurs when varying the atmosphere because of the discrepancy in the interactions of the two metal elements with the reactant.− Furthermore, the high structural flexibility of subnanometer monometallic species (isolated metal atoms and clusters) have also been demonstrated in experimental and theoretical works, which show even higher sensitivity than the nanoparticulate counterparts in response to the variation of the atmosphere. The particle size and coordination environment of metal atoms and clusters can exhibit dramatic changes when altering the atmosphere/solvent/ligand.− In this sense, when discussing the structural features of bimetallic entities, the discussion should be made in the context of the support/ligand and the atmosphere/solvent. The potential structural transformations during treatments or under reaction conditions should be paid special attention to, as will also be emphasized in other sections of this Review.

31Selective generation of binuclear bimetallic sites on the surface is quite challenging because of the difficulty to control the proximity and coordination environment of the two metal atoms. In this section, we will summarize several methods that allow the formation of binuclear sites on a solid carriers.

32A straightforward strategy for generating supported binuclear metal sites is to load two metal precursors on the support simultaneously. For instance, using a defective C3N4 material as the support, Ru and Pt atoms were loaded on the C3N4 support by a photoassisted deposition process.

33Different from the preparation of supported isolated monometallic species, it is recommended to use a bimetallic metal complex as the precursor to improve the possibility of forming binuclear bimetallic sites. Fu et al. have shown the generation of Mo–Ir species on TiO2 by grafting organometallic Mo–Ir complex and subsequent calcination treatment for removal of the organic ligands. In principle, this method can also be applied to the preparation of a wide scope of binuclear species if the bimetallic organometallic complex can be well dispersed on the support and the sintering of the metal species can be avoided during the thermal treatments.

34Atomic layer deposition (ALD) is a method that allows controlling the growth of metal species on solid carried with high precision. It has been applied for the preparation of numerous supported single-atom catalysts by using an organometallic complex as the precursor and solid carrier with anchoring groups as the support. To form binuclear bimetallic sites by ALD, one straightforward strategy is to employ a bimetallic organometallic complex as the precursor of the ALD process so that the resultant material comprises binuclear sites on the support.

35Another approach is to perform consecutive deposition of the two metals, which requires the selective anchoring of the secondary metal near to the primary metal species. For example, after the deposition of isolated Pt atoms on N-doped carbon nanotubes, Ru atoms were selectively anchored to the neighboring sites of the Pt atoms by ALD, resulting in the formation of Pt–Ru binuclear sites.

36Surface organometallic chemistry has been demonstrated as an effective approach for the generation of analogue species as molecular organometallic complexes in order to transform the homogeneous catalytic reaction into a heterogeneous process. By taking dehydrated silica as the support, organometallic complexes are grafted on silica by a condensation reaction between the hydroxyl group on silica and the organometallic complex. In principle, this strategy can also be extended to bimetallic systems which can be achieved by consecutive grafting of the organometallic precursor (see Figure 7). For instance, a stepwise anchoring of organometallic complexes (W(Me)6 and Ti(Np)4, Np = neopentyl) on silica resulted in the formation of neighboring W and Ti species on silica. The presence of neighboring W and Ti species is confirmed by the 1H–1H multiple-quantum NMR spectroscopy. In another work, a bimetallic Pt–Zn catalyst based on the surface grafting of organometallic Zn and Pt complex was prepared via a combination of ALD (deposition of isolated Zn species) and surface grafting of Pt species.

37It can be expected that the consecutive grafting approach cannot guarantee the homogeneity of the formation of neighboring W and Ti species because of the random distribution of the grafted W and Ti species. To ensure the formation of binuclear metal species on the support, a bimetallic complex is employed as the precursor for the grafting process so that the final support catalyst comprises abundant proximate binuclear species. Taking bimetallic Ta–Ir complex as the precursor, Camp et al. have developed a supported Ta–Ir catalyst for H–D exchange reactions, which showed much higher activity than the homogeneous Ta–Ir complex and supported monometallic single-site Ta species.

38In addition to the immobilization and surface organometallic strategy described above, binuclear metal sites can be formed via solid-state transformation, in which the metal species are introduced and stabilized on the solid carrier during treatments such as high-temperature annealing or pyrolysis. By pyrolyzing a composite material comprising two metal precursors, it is possible to form binuclear bimetallic sites on the solid carrier (usually carbon-based materials), as demonstrated with the Co–Pt–N–C and Ni–Fe–N–C materials., As demonstrated in Figure 8, binuclear FeCo sites are formed in N-doped carbon matrix by pyrolysis treatment of a MOF structure comprising the metal precursor compounds. The decomposition of MOF structure with zinc nodes will cause the formation of defective sites in the N-doped carbon because of the evaporation of Zn atoms, which can serve as the anchoring sites for binuclear sites. Due to the difficulties to control the pyrolysis process, it is highly challenging to achieve a high homogeneity in terms of the spatial distribution of the two metal species. In other words, the direct pyrolysis strategy may result in the generation of a broad scope of species in the final materials, including binuclear bimetallic sites, bimetallic nanoclusters/nanoparticles, and monometallic nanoclusters/nanoparticles.

39The formation of binuclear bimetallic sites on carbon support can also be achieved by a subsequent anchoring strategy. The metal atoms anchored primarily on the support can serve as the binding sites for the secondary metal atoms during the high-temperature pyrolysis process through the formation of metal–metal bonding, which facilitates the selective generation of binuclear bimetallic sites. Taking Co-MOF material comprising isolated Co atoms in the MOF matrix as the starting material, binuclear Co–Fe species are formed after the introduction of Fe precursor into the Co-MOF and subsequent high-temperature pyrolysis treatment.

40In the above-mentioned approaches, there is always a considerable amount of monometallic binuclear sites generated on the carbon matrix. To improve the probability of generation of bimetallic binuclear sites in the final materials, binuclear bimetallic complexes can be employed as the precursors and loaded on the solid carrier before the high-temperature pyrolysis treatment (as illustrated in Figure 9). In this way, the intimacy of the two metal atoms will be greatly improved in comparison to the methods in which the metal species are introduced separately. In principle, this strategy can also be extended to generation of bimetallic nanoclusters on the carbon matrix by using multinuclear organometallic complex as the precursor.

41Porous materials such as zeolites and metal–organic frameworks (MOFs) with well-defined crystalline structures have been widely employed as the host for single-site metal species. In principle, by introducing extra-framework species into a porous matrix with isolated framework, metal species will lead to the formation of binuclear metal sites if the two metal atoms can have close contact. This scenario has been achieved with metal-exchanged zeolites, in which the introduced metal species have close contact with the framework Al species. Moreover, it can also occur with heteroatom zeolites such as Sn-zeolite and Ti-zeolite materials, in which the external metal species can be located at the neighboring position of the framework Sn and Ti species. However, due to the difficulty in materials synthesis and structural characterization, this type of binuclear metal site has not been carefully explored yet.

42It will be more versatile and facile to generate binuclear metal sites in MOFs because of their structural diversity. By incorporating the metal species onto the organic linkers or nodes, two metal atoms with proximity can be formed, which can work as binuclear metal sites for catalysis. For instance, The Zn–O–Zr sites are produced by the reaction between the Zr6(μ3-O)4(μ3-OH)4 nodes of MOF-808 and ZnEt2, followed by a mild thermal treatment for removal of the ligands. In another work, Rh–Cu binuclear sites are formed in Cu-MOFs by replacing one Cu atom with a Rh atom in the binuclear paddlewheel nodes of the framework through transmetalation reaction.

43Another approach for generation of binuclear metal species in crystalline porous materials is to grafting the two metal entities into the microporous cavities. For instance, Cu2+ and Ni2+ ions were introduced into MOF-808 because of the coordination interaction between the metal ions and the ligands in the MOF-808 structure. The proximate Cu and Ni atoms form binuclear metal sites in the microporous cavity of MOF-808 with flexible distance in the ranges of 0.4–0.5 nm. If can be expected that the distance of the two metal atoms can be modulated by the size of the cavity and the binding sites for the metal species, as demonstrated with the generation of bifunctional Pd–Cu catalysts in MOF structure.

44In this section, we will summarize the methodologies for synthesis of bimetallic nanoclusters by presenting the general principle of each synthesis method and discussing the structural features of some representative materials generated accordingly. Different methods will be compared in terms of a set of criteria including the versatility, scalability, homogeneity, scope of materials, and their merits for green chemistry principles (cost, waste, toxicity, etc.).

45Bimetallic nanoclusters can be generated in gas phase by size-selected method and then deposited on a substrate for subsequent studies. Their size and chemical components can be precisely controlled by the mass spectrometer and the size-selected metal clusters can be deposited on a substrate, resulting in formation of well-defined nanoclusters as model materials. It should be noted that the nanoclusters generated by size-selected method are free of capping agents, making them as ideal samples to study the intrinsic physicochemical properties of the metal clusters. For instance, Ag9Pt2 and Ag9Pt3 clusters were generated in gas phase and then deposited on an ultrathin alumina support to elucidate how Pt and Ag can work in a synergistic way for CO oxidation reaction. Although this method can be theoretically applied to all the metal elements to generate metal clusters with desired size and compositions, the yield of the target materials is quite low because of the limited flux of ion beam. Therefore, up to this date, this method can only be used for preparing model catalysts for fundamental studies.

46Size-selected monometallic nanoclusters can serve as the binding sites for the second metal, leading to the formation of bimetallic nanoclusters with precise composition. For instance, the addition of Sn to size-selected Pt clusters was achieved by a reaction between reduced Pt clusters and SnCl4. Although the resultant bimetallic nanoclusters derived from the modification may not show high uniformity as the ones from direct size selection, it can effectively broaden the scope of bimetallic nanoclusters by chemical modifications of size-selected monometallic clusters.

47Wet-chemistry synthesis is one of the most widely used and versatile methods for the production of bimetallic nanoclusters. The general principle for this method is usually to reduce the two metal precursors simultaneously in the presence of solvent and/or capping agents (polymers or ligands) to form monodispersed bimetallic nanoclusters., The growth kinetics of the two metal elements should be carefully controlled to obtain the desired materials with narrow size distribution and high uniformity in chemical compositions. By tuning the synthesis parameters (precursor, solvent, capping agent, reductant, temperature, reaction time, etc.), the particle size and the chemical composition can be controlled in a fairly wide range. Especially, by using organic ligands with functional groups (such as thiol groups) that can deliver strong metal–ligand interaction, bimetallic nanoclusters with tunable chemical components and geometric structures have been prepared and their well-defined crystallographic structures are more regular than the bimetallic nanoclusters synthesized in the presence of capping agents with less strong metal–ligand coordination interaction (such as glycol and polyvinylpyrrolidone). Nevertheless, dendrimers with well-fined molecular structures can also serve as the host for bimetallic nanoclusters. The sizes of the nanoclusters encapsulated in the dendrimers can be modulated by the synthesis conditions and the structures of the dendrimers. However, the high costs and low yields of dendrimers limits their applications for large-scale preparation of dendrimer-encapsulated bimetallic nanoclusters.

48High yields of bimetallic nanoclusters can be achieved in relatively short reaction times (from minutes to hours), and the scope of the wet-chemistry synthesis is quite wide. The resultant monodispersed bimetallic nanoclusters can be directly used for catalytic results or further deposited on solid carriers, although their well-defined structures may change after being subjected to reaction conditions, which will be discussed later in this review.

49One drawback of liquid-phase synthesis is the use of capping agents and/or ligands, which may block the surface sites of the metal clusters. This blocking effect is particularly significant, with clusters stabilized by strongly coordinated ligands, such as AuPd, AuAg, and AuCu clusters protected by thiolate groups. Although the capping agents can be removed or partially removed by postsynthesis treatments, undesired structural reconstruction (sintering or change of the spatial distribution of the chemical components) may occur. Nevertheless, for large-scale production of bimetallic nanoclusters for practical applications, the cost and the treatment of the solvent used in the synthesis should be taken into account.

50For the preparation of bimetallic nanoclusters, direct impregnation of the metal precursors on the solid carrier can work when the growth kinetics of the two metals are matched. Otherwise, it may cause the segregation of the two elements and the formation of separated monometallic clusters. By in situ IR and EXAFS spectroscopic characterization, formation of bimetallic OsRu clusters via Os–Ru bonding has been observed during the reduction of MgO-supported Os3(CO)12 and Ru3(CO)12 complex by H2. At 333–398 K, the Ru–CO bonding is dissociated due to the reduction by H2 and the formation of Ru–Os bonding is observed at temperature above 398 K. The EXAFS data indicate the formation of Ru2Os clusters after reduction treatment at 423 K. The formation of bimetallic clusters depends on the decomposition kinetics of the two metal precursors and the stability of the tiny metal clusters on the support. In the case of Rh and Os, the formation of bimetallic RhOs clusters is not observed when applying the H2 reduction treatment to MgO-supported Os3(CO)12 and Rh(C2H4)(acac) (acac = acetylacetonate) complex at temperature up to 393 K. Because of the lower stability of Rh(C2H4)(acac), the formation of Rh4–6 clusters is observed after reduction treatment at 393 K while the Os3(CO)12 species remain intact under the same conditions. Consequently, the formation of small RhOs clusters will be more difficult than the case of RuOs. Therefore, one strategy to facilitate the contact between the two metals is to use bimetallic complexes as the precursor for impregnation. After the decomposition of the bimetallic complexes, bimetallic nanoclusters with uniform chemical components are formed on the solid support. To avoid the segregation and/or sintering of the two elements, the postimpregnation treatments should be carefully proceeded to ensure the formation of bimetallic nanoclusters on the support.

51In order to facilitate the formation of bimetallic nanoclusters, a potential strategy is to employ consecutive impregnation of metal precursors on the solid carrier because the first impregnated metal can serve as the binding site for the second metal, especially when the impregnated metal species are reduced. This method has already been successfully demonstrated for the preparation of bimetallic nanoparticles with core–shell structures.,

52Because the precursors for synthesis metal nanoclusters are mostly cationic or anionic complexes, the precursors can be adsorbed by the solid carrier by strong electrostatic adsorption. As depicted in Figure 10, by properly choosing the metal precursors and controlling the support’s surface charge property, the metal precursors are adsorbed and stabilized by the support and then transformed into bimetallic nanoclusters in subsequent calcination/reduction treatments. To achieve a more homogeneous mixing of the two elements, the two metal precursors with opposite charges were loaded on the support via strong electrostatic adsorption. Consequently, the two metal precursors formed a new bimetallic complex on the support, which can be further transformed into bimetallic nanoclusters.

53By conventional liquid-phase ion exchange reactions, the metal precursors can be grafted on the solid carrier and then transformed into bimetallic species in postsynthesis treatments. Because of their high porosity and regular structure, zeolites are excellent hosts for generation and stabilization of bimetallic nanoclusters. In principle, bimetallic nanoclusters can be formed when performing the ion exchange reactions in the presence of two metal precursors if the two metals can be adsorbed and stabilized evenly by the host. However, the different exchange kinetics of the metal precursors may result in the random distribution and spatial segregation in the final metal-zeolite materials. To control the proximity of the two metals, the introduction of two metals is carried in a consecutive way. For instance, PtZn bimetallic nanoclusters are generated in MFI zeolite by introducing Pt2+ cations into a zincosilicate zeolite in order to anchor Pt species to the nearby sites surrounding framework Zn species, which are further converted to PtZn nanoclusters calcination and reduction treatments. This strategy has also been demonstrated with generation of PtSn nanoclusters in beta zeolites and should be in principle applicable for preparation of a broad scope of materials by taking metal-zeolite materials as the support.

54Ion exchange reactions can occur between monodispersed particles and molecular metal complex, as widely demonstrated with the preparation of multicomponent colloid nanoparticles. By galvanic replacement, the chemical composition of metal nanoclusters can be altered in a controllable manner. For example, the galvanic reaction between Pd147 cluster and Au3+ ions lead to the formation of isolated Au sites on Pd cluster. Of course, this method can also be applied to clusters with smaller sizes, and the doping level of the second metal into the parent metal cluster can be tuned by controlling the galvanic reaction. Furthermore, by employing ligand-protected metal clusters as the seed, the generation of bimetallic nanoclusters via a controllable galvanic replacement by varying the types of ligands, which has already demonstrated by galvanic replacement reactions with metal nanoparticles.

55The impregnation and static adsorption are versatile and effective methods for generation of bimetallic nanoclusters on solid surface. However, the resultant metal clusters may suffer severe sintering during treatments under harsh conditions. One way to address this challenge is to impose an external constraint on the bimetallic nanoclusters to restrict their mobility. Porous supports (usually with pore sizes below 10 nm) can be the ideal carrier to host bimetallic nanoclusters if the metal species can be introduced into the porous environment.

56By impregnation or ion-exchange method, part of the metal precursors can be introduced into the porous environment. For instance, small PtSn and PtZn nanoclusters are generated by introducing molecular Pt precursor into Sn-β zeolite or zincosilicate zeolite structure, respectively. The Sn and Zn species in the zeolite framework can serve as the binding sites for Pt and the rigid zeolite framework also protect the PtSn and PtZn nanoclusters from sintering into large particles., This strategy has already been demonstrated with the synthesis of PtZn and PtSn nanoclusters in MFI and beta zeolites., However, the ion-exchange method will produce a large portion of the metal species may be located on the surface or subsurface of the porous materials.

57To improve the efficiency for encapsulating the metal species by the porous scaffold, the introduction of metal species can be performed during the synthesis of the porous materials. This one-pot synthesis approach has been demonstrated effectively with the preparation of bimetallic metal clusters in microporous zeolites. For instance, bimetallic PtPd nanoclusters with average particle sizes of ∼1.5 nm in ZSM-5 zeolite (MFI-type aluminosilicate zeolite). In another work, as shown in Figure 11, bimetallic PtSn clusters with sizes of 0.5–0.6 nm are formed in the sinusoidal channels of pure silica MFI zeolite by introducing alkali metals into the synthesis mixture for stabilizing the atomically dispersed metal species during high-temperature calcination and reduction treatments. This strategy should also be applicable for the synthesis of bimetallic nanoclusters within metal–organic framework (MOF) structures because the encapsulation of Pt and Pd nanoclusters in MOF crystallites have already been realized.

58Thanks to the porous scaffold surrounding the bimetallic nanoclusters, size-selective catalysis can be achieved with those confined metal catalysts. However, the accessibility of the metal clusters is also limited to small molecules such as light alkanes, CO/CO2, and NOx, which will be further discussed in the sections on catalytic applications of confined metal catalysts in this review.

59As reported in the literature, the morphology, particle size, chemical components, and metal–support interface of metal nanoparticles can show dynamic structural transformations as responses to the change of the external environment (temperature, atmosphere, pressure etc.). For instance, the exposed facet of the PtCo bimetallic nanoparticles and the arrangements of Pt and Co within the nanoparticles evolves during the high-temperature annealing treatment. On one hand, such environment-induced structural changes should also occur with bimetallic nanoclusters, and the consequences should be carefully considered when using the as-synthesized bimetallic nanoclusters for catalytic applications, because the bimetallic nanoclusters should be more sensitive to the environment than the bimetallic nanoparticles. However, the environment-induced structural transformations have not received enough attention in the works focused on synthesis and/or catalysis with bimetallic nanoclusters. In many cases, the organic ligands capped on the surface of ligand-protected bimetallic nanoclusters need to be removed in order to liberate the metal sites for catalytic reactions. However, during the removal of the ligands, the structural configuration of the bimetallic nanoclusters could be modified, which should be carefully followed.

60On the other hand, further modifications on the structures of the bimetallic nanoclusters are also possible when appropriate postsynthesis treatments are employed. One example of this strategy is shown in the modulation of bimetallic PtSn clusters by high-temperature reduction treatment. With longer reduction time with H2 or increasing the reduction temperature, a more intimate contact between the Pt and Sn species can be achieved, resulting in the formation of Pt clusters covered by Sn species in the microporous channel of MFI zeolite. In this regard, the coverage of Sn species on Pt clusters can be tuned by the postsynthesis reduction conditions.

61In another example, PtCo bimetallic nanoclusters with tunable size and chemical compositions can be generated by reducing Pt/Co3O4 catalyst under desired conditions. Initially, the Pt species are atomically dispersed on the Co3O4 support and the Pt atoms agglomerate into Pt clusters after exposure to H2. Once the Pt clusters are formed, the reduction of the Co3O4 support will be promoted, and the reduced Co atoms can migrate into the Pt clusters as dopants, giving to the formation of PtCo bimetallic nanoclusters. This strategy could be extended for other noble metals (Pd, Rh, Ru etc.) supported on reducible solid carriers (FeOx, MnOx, NiOx, SnOx, etc.). Mild treatment conditions are recommended for generation of bimetallic nanoclusters via this approach in order to avoid the formation of large nanoparticles.

62The preparation of supported bimetallic nanoclusters with high uniformity in particle size and composition has been a long-standing goal in the field of heterogeneous catalysis. As discussed above, the following approaches have been reported in the literature for the generation of bimetallic clusters with given size and compositions: (1) size-selected clusters by physical method, (2) bimetallic clusters from organometallic precursors, and (3) ligand-protected bimetallic clusters. The bimetallic clusters generated from these strategies can be used as model systems for the fundamental understanding of the reactivities of the bimetallic clusters. However, currently, none of the above approaches can meet the requirements of preparation of supported bimetallic clusters for practical applications. In the case of the generation of bimetallic clusters from the size-selected method, the yields of the bimetallic clusters are too low for scalable production. In the case of the bimetallic clusters from organometallic precursors and ligand-protected bimetallic clusters, the decompositions of ligands are usually mandatory for liberating the active sites. However, during the removal of the organic ligands, sintering of the metal particles or segregation of the metal elements probably occur, resulting in the formation of bimetallic particles with a relatively broad distribution of particle size and composition.

63To mitigate the structural deformation of the organometallic precursors and ligand-protected clusters during the ligand removal procedure, porous materials with anchoring sites can be used as the support for hosting the bimetallic precursors and protect them from sintering/segregation. This strategy can also be applied for the precise synthesis of bimetallic nanoclusters made by a metallic component and an oxide component. By using crystalline porous materials (zeolites and metal–organic frameworks) for the accommodation of subnanometer metal clusters in the microporous channels/cavities, the oxyphilic metal species can be located in the framework materials (e.g., node positions in MOFs) and then the metal clusters are anchored at these oxyphilic metal sites, giving to the formation of well-defined noble metal–oxyphilic metal combinations.

64The synthesis of bimetallic nanoparticles has been covered by numerous comprehensive reviews.− As reflected in the literature, bimetallic nanoparticles can be readily prepared in controllable ways in terms of the particle size, chemical composition, spatial distribution of the two elements, and morphology. There are already mature synthesis methodologies reported in the literature for a wide scope of metal elements and for a broad range of bimetallic nanoparticles with various. The choice of the synthesis approach depends on the structural features of the target materials and the potential applications. In this section, we would like to highlight some representative works reported in recent years in terms of the synthesis of bimetallic nanoparticles/nanostructures with novel structural features.

65In conventional alloy nanoparticles, the two metals form a mixture and show random distribution in a bimetallic nanoparticle., The percentage of the two metals are close to each other or at the same magnitude, and for each metal atom, it will be surrounded by both elements. By decreasing the percentage of one metal to a very low level (usually below 5%), each metal atom of the minor composition will be separated by the other metal in the bimetallic particles, giving to the formation of single-atom alloy nanoparticles. Generally, the preparation of single-atom alloy nanoparticles is similar to conventional bimetallic nanoparticles. For instance, isolated metal sites can be formed by coreduction of the two metal precursors by diluting one metal in another. Besides, preparation of single-atom alloy nanoparticles can also be achieved by galvanic replacement reaction, as demonstrated with the synthesis of PdCu and PtCu single-atom alloy nanoparticles., As shown in Figure 12, Cu nanoparticles formed on mixed metal oxides can reduce H2PtCl6, resulting in the formation of isolated Pt atoms in the Cu matrix. Controlling the ratio of Pt/Cu will be critical to ensure the atomic dispersion of Pt in Cu nanoparticles.

66In terms of bimetallic nanoparticles with core–shell structures, the uneven spatial distribution of the two elements causes the lattice strain at the interface of “core” and “shell” due to the mismatch of the lattice parameters of the two metallic components. The magnitude of the lattice strain depends on the degree of the mismatch of the lattice parameters and, moreover, it can be expected that the interfacial lattice strain will strongly dependent on the thickness of the “shell”, which is confirmed by a study based on high-resolution electron microscopy study. In the case of Pd@Pt nanoparticles, the lattice strain of Pt shell will be not significant once the shell is thicker for more than 5–6 monolayers. Beyond the geometric, charge redistribution between the “core” and “shell” components will occur, resulting in the variation of the electronic properties.

67The shell thickness can be modulated by controlling the deposition rate of the “shell” composition on the “core” nanoparticles. For example, with a slow introduction rate of the Pt precursor to the synthesis mixture, the deposition of Pt is achieved on the entire surface of Pd nanocubes, resulting in the formation of uniform thin Pt shells of 1–6 monolayers. With the help of synthesis methodologies developed based on flow chemistry, it is feasible to achieve large-scale production of core–shell nanoparticles with precise control of the thickness of the shell component. Diffusion of metal atoms from the “core” to the “shell” during the deposition process may occur, leading to the formation of bimetallic shells, as observed with the preparation of Pd@AuPd nanoparticles (see Figure 13). The thickness of the “shell” can affect the stability of the core–shell structure under the reaction conditions for electrocatalytic production of H2O2. The sample with thin AuPd “shells” exhibit higher stability than the sample with thick “shells”.

68Intermetallic nanoparticles are bimetallic nanoparticles with ordered structures at atomic level and the formation of intermetallic nanoparticles usually requires high-temperature synthesis conditions., Though there are some works on wet-chemistry synthesis of intermetallic nanoparticles, the majority of the reported works involves high-temperature annealing treatments. For instance, Pt3Co intermetallic nanoparticles are formed after annealing treatment with the platinum/cobalt–carbon precursor at 900 °C, as a result of the migration of cobalt species from the carbon support to Pt nanoparticles. To avoid severe sintering of the metal particles during the high-temperature treatments, sulfur-doped carbon is employed as the support for stabilizing small bimetallic PtM intermetallic nanoparticles (<5 nm). Furthermore, core–shell Pt/PtM (M = Fe, Ni, Cu etc.) nanoparticles with intermetallic PtM core and thin Pt shell can be prepared by the combination of high-temperature annealing treatment and acid leaching of M element.

69The intermetallic nanoparticles comprising early transition metals (such as Fe, Co, Ni, etc.) and main group metals (such as Sn, Ga, etc.) have been intensively studied while the rare-earth elements are much less explored because of the difficulty to incorporate rare-earth elements into metallic particles. It has been found that PtCe intermetallic nanoparticles are formed after reducing Pt/CeO2 catalyst by H2 at above 900 °C. The resultant PtCe intermetallic nanoparticles are protected by thin layers of CeOx, as a result of strong-metal support interaction. Molten-salt synthesis methods have been developed to facilitate the formation of rare-earth intermetallic nanoparticles under relatively mild conditions (annealing at 650 °C). The choice of precursors for efficient incorporation of rare-earth elements into metallic particles is important, because the presence of H2O and Cl during the high-temperature annealing process assist the formation of bonding between noble metals and rare-earth elements.

70Another strategy to promote the stability of intermetallic nanoparticles is to confine them in a porous support. For example, PtRE (RE = Y, Ce and La) nanoparticles (<5 nm) can be synthesized after reducing the Pt and RE precursors at 700 °C and stabilized in hierarchical MFI-type zeolite. This work indicates that intermetallic nanoparticles with unconventional chemical compositions could be obtained by tuning the environment surrounding the nanoparticles.

71Tremendous efforts have been dedicated to tailoring the morphology of bimetallic nanoparticles and assembling the particles by design for the construction of novel bimetallic nanostructures. For instance, by acid leaching and annealing treatments, hollow PtNi nanoframeworks and one-dimensional PtNi nanowires with jagged surface have been prepared and used for electrocatalytic oxygen reduction reaction., These materials show enhanced activity and stability against sintering in comparison with conventional Pt/C catalyst, which are associated with the abundant undercoordinated sites on the surface and the unique morphologies. Besides, two-dimensional bimetallic nanostructures are also interesting catalytic materials., These thin materials with a few atomic-layer thickness can be considered as intermediate materials between nanoclusters and nanoparticles because they are made with the highly undercoordinated metal atoms at the edge and the coordinated metal atoms at planar positions. In section 7, we will discuss further the novel catalytic properties of the low-dimensional bimetallic nanostructures, such as ultrathin nanosheets or nanowires.

72Substituting noble metal catalysts with non-noble metal catalysts represent an important direction for achieving sustainable processes. However, for most of 3d transition metals, their high affinity to oxygen causes the low stability of their metallic nanoparticles in air due to the oxidation of 3d metal nanoparticles into oxides. By using organometallic compounds as precursor and organic ligands as stabilizer, a series of bimetallic nanoparticles based on 3d metals have been prepared, which exhibit narrow size distributions and uniform compositions.− In order to further improve the stability of non-noble metal catalysts, it is proposed to cover them with thin carbon layers to minimize their contact with air, which leads to the preparation of highly stable and active CoNi and CuFe bimetallic nanoparticles for a variety of hydrogenation reactions.− The use of thin carbon layers for protecting metal nanoparticles from degradation/deactivation/leaching has also been shown with noble metals, such as the synthesis of bimetallic PtCo nanoparticles stabilized in graphene nanopockets.

73In the above-mentioned bimetallic nanoparticulate materials, they are crystalline structures with long-range ordering of the metal atoms, although the occupation of a specific position in the crystalline structure could be either of the two metal elements as the scenario of alloy nanoparticles. Actually, bimetallic nanostructures can be formed in amorphous phase, in which the metal atoms do not show long-range ordering. For instance, the doping of B into crystalline PdRu nanoparticles will cause the crystalline-to-amorphous transition, which could be caused by the interruption of Pd–Ru bonding in the presence of B. To avoid the atomic ordering driven by thermodynamics, the synthesis of amorphous nanostructures is usually performed under mild conditions. As displayed in Figure 14, a general strategy for preparation of amorphous bimetallic nanosheets (thickness below 10 nm) by annealing mixture of metal acetylacetonates and alkali salt (KNO3 or KBr) at below 310 °C. In these structures, the metal–metal bonding is usually longer than that in the corresponding crystalline system.

74In a bulk bimetallic system, the composition of the bimetallic structures is dominated by the phase diagram. In nanoscale, the phase diagram still works, although it may differ from the conventional bulk phase diagram. To achieve the formation of thermodynamically metastable bimetallic structures, the mixing of the two metal elements need to be proceeded rapidly to avoid the segregation. For example, as shown in Figure 15, CuAg alloy nanoparticles are formed after heating the Cu and Ag precursors supported on carbon nanofibers to 1600 °C in less than 1 s via Joule heating, leading to the very fast decomposition of the precursors and the mixing of Cu and Ag atoms. Another approach to achieve the fast agglomeration of the metal atoms into metastable alloy nanoclusters/nanoclusters is using laser as the energy input source for creating local hot spots in the sample. For instance, the metal ions supported on graphene are reduced by laser propulsion and then the metal–metal bonding are formed, leading to the formation of multimetallic nanoparticles on the graphene support.

75It should be noted that the conventional phase diagrams are established under standard state conditions. By the introduction of metal–support interactions, it is possible to generate alloy nanoclusters/nanoparticles with compositions beyond the conventional phase diagram. In this regard, the use of solid carriers with rich defective sites can facilitate the capture and stabilization of metal atoms for overcoming immiscibility.

76Deactivation is a ubiquitous for solid catalysts, which is also commonly observed with bimetallic catalysts. Depending on the structural features of the bimetallic catalysts and the reaction conditions, the several types of deactivation mechanism may occur, such as the sintering of bimetallic particles, coking, leaching of metal species, segregation of the two metal elements, poisoning of the active sites, etc.− Because of the complexity in deactivation mechanism and diversity in the structural features of bimetallic catalysts, the regeneration of deactivated bimetallic catalysts should be proceeded and optimized in a case-by-case fashion. For each particular case, the regeneration procedure of the deactivated catalyst needs to be carefully investigated based on the deactivation mechanism and the physicochemical properties of the bimetallic catalysts. In principle, the lessons learned in the studies of synthesis of various types of bimetallic catalysts and the structural transformation behaviors of bimetallic catalysts can be translated into the regeneration of the bimetallic catalysts. For instance, the regeneration of bimetallic PtSn nanoparticles supported on alumina can be achieved by proper oxychlorination treatment and subsequent high-temperature reduction treatment.,

77The ultimate goal of the synthesis of bimetallic entities (from binuclear species to nanoclusters and nanoparticles) is to control the geometric structure and composition of the metal species with atomic precision. This goal is achieved in several specific examples, such as the generation of binuclear bimetallic surfaces via surface organometallic approach and ligand-coordinated bimetallic nanoclusters. However, these as-synthesized well-defined species will undergo a structural transformation in catalytic reactions. Therefore, it is critical to maintain the initial structures of the as-synthesized well-defined species or to control the reaction-induced structural transformation, resulting in solid catalysts comprising uniform bimetallic species.

78The presence of many different anchoring sites in the conventional solid carriers (high-surface-area oxides and carbon materials) hinders the precise control of the coordination environment of the bimetallic species. One promising approach is to utilize ordered porous materials such as zeolites and MOFs as the host for subnanometer bimetallic species. Although there are exciting progress in the last 10 years in this direction, the regioselective generation of highly uniform bimetallic metal species in the microporous environment is hindered by the difficulty of controlling the spatial distribution of the metal species inside the microporous crystallites and determining the location of the metal species with atomic precision. Considering that the use of an open-structure solid carrier is indispensable, the development of high-surface-area supports with uniform morphology, particle size, and porosity is also critical for the preparation of bimetallic catalysts with the well-defined metal–support interface.

79Moreover, it is very difficult to achieve a good balance between structural uniformity and metal loading. In most scenarios, for the synthesis of subnanometer bimetallic catalysts, the metal loadings have to be kept below 0.5 wt % to avoid the presence of multiple types of metal entities, limiting the practical applications of the materials developed in the fundamental studies. It has been shown that the use of MOF as the host for isolated metal atoms or clusters can achieve metal loading up to 20 wt %. This approach relies on the use of a crystalline host with an open porous structure and controllable introduction of metal species into the porous framework and adequate postsynthesis treatments.

80Fundamental understanding on the growth mechanism is vital for rational design of synthesis methodologies of bimetallic entities. Currently, there are already systematic studies on the formation and growth mechanism of monometallic and bimetallic nanoparticles, while the fundamental insights into the formation and growth mechanism of binuclear bimetallic sites and bimetallic nanoclusters on solid carriers are quite limited in the literature to date., In the future, we anticipate that more efforts dedicated to the mechanistic studies of the formation and evolution of supported subnanometer bimetallic entities will contribute to the rational synthesis of bimetallic catalysts based on binuclear sites and bimetallic nanoclusters.

81In this section, we will discuss the tools/techniques for characterizing geometric and electronic structures of bimetallic species from binuclear sites to bimetallic nanoclusters and nanoparticles. Especially, for each tool/technique, we will analyze the applicability to different types of bimetallic entities and discuss their advantages and limitations.

82The morphology is a key structural feature that should be carefully studied if one attempts to establish the structure–reactivity relationship of a bimetallic entity. The morphological properties of a bimetallic entity include the size, geometric shape, and metal–support interfacial structure. The most frequently used technique for measuring the morphology of bimetallic species is aberration-corrected scanning transmission electron microscopy (AC-STEM). By working in the high-angle annular dark-field (HAADF) mode, the bimetallic species can be easily distinguished from the support with low Z-contrast (such as alumina and carbon materials). Moreover, if the difference in the atomic numbers of the two metal elements is large enough (for instance, Pt and Ru), it is possible to differentiate the binuclear species according to their contrast difference. By extracting the contrast and location of the metal atoms from the high-resolution HAADF-STEM images, one can determine the abundance of the binuclear bimetallic sites in the whole sample, based on an automated statistical analysis of a large number of images. In the case of bimetallic nanoclusters, it is also possible to resolve the three-dimensional structure of small clusters (usually with less than 20 atoms) by quantitative analysis of the AC-STEM images, although it relies on complicated instrumental setup and image analysis.

83Compared to subnanometer bimetallic species, it is easier to describe the general morphology of bimetallic nanoparticles with nanometric precision. Features such as particle size, exposed facet, distribution of the two metal elements, and particle–support interface can be readily measured by current electron microscopy techniques. However, if one needs to know further detailed information with atomic resolution, such as the three-dimensional structure of a bimetallic nanoparticle, it will be more challenging than the case with a bimetallic nanocluster., The recently developed atomic electron tomography technique can resolve the three-dimensional structure of a PtFe nanoparticle, although its realization requires a complicated experimental setup and image processing.

84As mentioned before in this review, due to the mismatch in lattice parameters, lattice strain will be introduced into the bimetallic particles with core–shell structures. Ideally, the lattice strain in the core–shell bimetallic nanoparticles can be measured by X-ray diffraction because the lattice parameters can be directly calculated from the diffraction pattern. However, this technique requires a sample with high structural homogeneity, which is usually challenging to be achieved with bimetallic nanoparticles prepared by wet-chemistry methods. The local lattice strain can be calculated through the geometric phase analysis based on the atomic-level structural features extracted from high-resolution HAADF-STEM images, although errors may be introduced by image aberrations. As shown in Figure 16, The use of 4D-STEM can greatly improve the accuracy of strain analysis of core–shell nanoparticles because it allows quantification of the unit cell size as a function of distance from the core–shell interface due to the higher precision and absence of scan distortions that are afforded by 4D-STEM in comparison to aberration-corrected atomic-resolution STEM imaging technique. As an emerging electron microscopy technique, 4D-STEM can also be used to probe the electronic interaction between the metal particles and the support. For instance, the charge transfer from the SrTiO3 (STO) support to Au nanoparticle is directly visualized, which is supported by DFT calculations. This technique can be extended to the characterization of bimetallic nanoparticles supported on solid carriers, which may provide insights on the charge redistribution in these materials.

85It is always argued by the researchers in catalysis community that high-resolution TEM/STEM images cannot reflect the averaged information on the measured samples because the images can only provide information in very local areas. This argument is valid when the sample comprises a mixture of different types of metal entities with a broad range of particle size distribution and chemical composition. However, if the metal entities in the sample exhibit high uniformity and a large number of representative images are collected, statistical information regarding the structural features of the metal entities can be obtained by performing automated analysis of the experimental images. The recent progress in employing machine-learning methods for processing TEM/STEM images has greatly improved the efficiency and precision for analysis of the experimental images, leading to the revealing of detailed structural features in the samples.

86If the material containing bimetallic nanocluster is highly uniform, the chemical components of bimetallic clusters can be easily determined by conventional analysis such as inductively coupled plasma (ICP). Besides, due to their small sizes, the whole composition of the bimetallic clusters can also be measured by X-ray photoelectron spectroscopy (XPS), which cannot be achieved with large bimetallic nanoparticles (>5 nm) because of the limited penetration of X-ray. However, both XPS and ICP can only give an average value on the whole sample while they cannot provide the spatial distribution of the two metal elements in the solid catalyst in nanoscale or subnanometer regime.

87As complementary tool to ICP and XPS, electron energy loss spectroscopy (EELS) and energy dispersion spectroscopy (EDS) can directly measure the chemical composition of a bimetallic entity. By measuring multiple bimetallic entities, one can give a preliminary conclusion on the uniformity/homogeneity of the distribution of the two metal elements, which cannot be achieved by ICP or XPS. As described in Figure 17, by grouping the raw EDS mapping data into two-dimensional data sets, the statistical representation of local compositional distribution in bimetallic or even multimetallic nanoparticles can be obtained, which can tell the homogeneity of the compositions with spatial resolution of ∼1 nm. Nevertheless, this workflow can be automatically applied to numerous EDS maps, leading to a statistical understanding on the composition of the bimetallic and multimetallic nanoparticles.

88Furthermore, the three-dimensional composition analysis based on EELS and EDS tomography can provide three-dimensional structural information on bimetallic nanoparticles. With high-quality spectroscopic data, it is possible to reconstruct the spatial distribution of a bimetallic nanoparticle with spatial resolution less than 1 nm. The recent developments of image analysis methods and algorithms based on machine learning can greatly improves the precision of the 3D reconstruction with less input images than the conventional methodology, which will be very useful for characterization of beam-sensitive materials.

89In the case of bimetallic nanoclusters, the direct analysis of their chemical components by EELS/EDS is quite challenging because of the low signal/noise ratio of a single nanocluster. One indirect approach to resolve the structure of a bimetallic nanocluster is to perform detailed image analysis on high-quality AC-STEM image, as demonstrated with the K-means analysis with PtSn nanoclusters. By correlating the experimental and simulated HAADF-STEM images, it is possible to distinguish the Pt and Sn species and therefore to give the spatial distribution of the two elements in the zeolite support. In principle, this method can be a general approach for analyzing the structural features of supported bimetallic nanoclusters, especially when the spatial resolutions of the experimental HAADF-STEM images of challenging samples are compromised by the experimental conditions. It should be noted that, the reliability of this approach depends on the quality of the image simulation process, which should be as close to the experimental samples as possible.

90The atom-probe tomography (APT) technique has been developed to achieve a high spatial resolution in regarding the composition of bimetallic nanoparticles in three-dimensional space. For instance, as can be seen in Figure 18, the location of Ag and Au atoms in core–shell AuAg nanoparticles can be directly resolved, giving a quantitative evaluation of the exposed surface atoms in bimetallic nanoparticles. After continuous improvements, the spatial resolution of APT can be achieved at the level of ∼0.5 nm or even smaller scale, which also depends on the physicochemical properties of the materials. Although the works on employing APT for structural characterization of supported metal catalysts are reported more than 10 years ago, this technique is still not widely applied in catalysis community, which could be caused by the requirements in professional experiences in sample preparation and data analysis. We believe that the commercial APT instruments will become more and more user-friendly, and therefore this technique can be employed by the catalysis community for a broad scope of catalytic materials because it can provide indispensable information on the three-dimensional structures of the solid catalysts.

91X-ray photoelectron spectroscopy is the most frequently used technique for probing the chemical states and electronic properties of supported metal catalysts. In the case of bimetallic entities, the chemical states of each metal elements and the potential charge transfer between the metals and the metal–support interaction can be revealed by XPS. The developments of ambient-pressure XPS based on synchrotron light source allow to follow the evolution of the chemical states of metal elements in bimetallic nanoclusters and nanoparticles as well as the structural transformation driven by the reactants.

92Besides XPS, the electronic properties of bimetallic entities are critical features because the interaction between the metal entities with the reactant molecules will be greatly determined by the electronic properties. As practiced with molecular complexes, in principle, the electronic properties of binuclear metal sites should also be able to be measured by UV–vis spectroscopy. However, due to the low metal loadings and the interfere from the support, the spectroscopic features of the binuclear metal sites may be difficult to be captured.

93In the case of bimetallic nanoclusters with well-defined structures prepared by wet-chemistry synthesis, UV–vis spectroscopy is a powerful and facile tool to elucidate their electronic properties. The doping of one atom into a metal cluster can be clearly differentiate from the monometallic cluster because of the marked changes in the UV–vis spectrum, as proved with the study of bimetallic M-doped Au clusters (M = Pt, Pd). By employing gas-phase photoelectron spectroscopy, the electronic structures of bimetallic nanoclusters can be revealed with details of the orbital information, providing insights to understand the molecular-like properties of the nanoclusters with precise structures. Moreover, the changes in electronic structures of bimetallic clusters induced by adsorption of reactants can be captured by in situ UV–vis spectroscopy. The unique electronic properties of metal clusters are also reflected in their fluorescence spectra. Bimetallic nanoclusters can be usually easily differentiated from the monometallic counterparts by fluorescence spectroscopy. The electron affinities of the ligand-protected bimetallic nanoclusters can be measured by gas-phase photoelectron spectroscopy, which gives more precise information on the electronic structures of the metal clusters.

94The electronic properties of bimetallic nanoparticles are also reflected in the UV–vis spectra, although it is not straightforward to correlate the spectra with their electronic structures because of the broad bands in the spectra. Due to their plasmonic properties, bimetallic nanoparticles comprising Au, Ag, or Cu show size-dependent absorption band in UV–vis spectra. The variation of particle size and chemical components will be reflected in the UV–vis spectra.

95From a structural point of view, for metal nanoclusters of sizes around 1 nm, almost all the metal atoms can be considered as surface atoms, although capping agents or ligands may coordinate to the metal atoms (at least part of them, if not all), resulting in the separation of those atoms from the external environment. Probing the surface properties of the exposed metal atoms of the bimetallic clusters can help to understand the amount of exposed surface sites and their chemical reactivity. One commonly used approach is to employ a probe molecule to interact with the metal nanoclusters and then to measure the impacts of metal–molecule interaction by chemisorption, vibrational, and/or electronic spectroscopy.

96For instance, using CO or H2 as a probe molecule, the number of the exposed surface sites can be measured reliably by chemisorption methods, which actually has already been widely practiced with conventional nanoparticulate metal catalysts. By employing CO as probe molecule, the surface composition of bimetallic entities can be quantitative measured when only one metal can absorb CO while the other cannot. For example, the number of surface Pd sites in the dilute PdAu bimetallic nanoparticles is determined by quantified pulses of CO by assuming all the Pd sites are equal for adsorption of CO. However, it is inferred by a comparative study that multiple types of adsorption configurations could exist in the dilute single-atom PdCu alloy nanoparticles, which gives different CO/Pd ratios in the adsorption models (see Figure 19). Therefore, the precise measurements of the metal sites in bimetallic nanoparticles relies on the identification of the adsorption configuration (mono- or dicarbonyl configuration) of the probe molecule at the measurement conditions (temperature, partial pressure of the probe molecule etc.) by performing rigorous control experiments on the monometallic samples.

97It should be noted that due to the low stability of metal nanoclusters, the activation conditions used before the chemisorption measurements should be carefully selected to avoid sintering or other types of structural transformation of the nanoclusters. Nevertheless, when the probe molecules interact with binuclear bimetallic sites and bimetallic nanoclusters, the adsorption configuration of the probe molecules could be different to the classic metal nanoparticles. For example, CO may be adsorbed on small metal clusters via the bridged configuration instead of the terminal configuration. In the case of binuclear bimetallic sites, CO or H2 may not be adsorbed on those sites. Consequently, the well-established method may not be able to give quantitative results when applying the chemisorption measurement to subnanometer bimetallic sites because of the nonstoichiometric adsorption pattern, although the qualitative results can also provide useful insights. Nevertheless, the CO-induced reconstruction and/or segregation of bimetallic nanoclusters/nanoparticles should not be neglected, as a result of the different affinity of the metals to CO in the bimetallic entities. For this reason, it is recommended to collect the CO-IR spectra at different CO coverages at various temperature to evaluate the influences of the reconstruction of the bimetallic entities.

98CO-infrared (CO-IR) spectroscopy is employed to acquire further information about the surface structure of the bimetallic nanoclusters, which gives detailed information on the electronic properties and coordination environment. The shape and position of the CO adsorption bands can tell the information regarding the surface composition of the bimetallic nanoclusters and the distribution of the two metal elements. Taking PtSn nanoparticles as an example, the introduction of Sn species to Pt nanoparticles will weaken the adsorption of CO on the Pt sites, as reflected on the decrease of chemisorption capacity of CO. Nevertheless, the full width at half-maximum of the linear configuration of CO adsorbed on PtSn nanoparticles are usually narrower than that on Pt because of the preferential coverage of Sn on the highly undercoordinated Pt sites. Therefore, the combination of CO-IR and XPS is quite useful for elucidating the spatial distribution of the two metal elements in bimetallic nanoclusters and nanoparticles.

99The coordination environment of a metal entity includes the bonding between the metal atoms inside the metal entity and the bonding with the external environment (solid carrier, support, and reactant). The atomic structures of bimetallic nanoclusters can be resolved by single-crystal X-ray diffraction, as practiced with well-defined bimetallic nanoclusters. However, this approach relies on the formation of single crystals for collection of high-quality X-ray diffraction patterns, which is usually not feasible with practical heterogeneous metal catalysts. To improve the data quality of X-ray diffraction, the powder samples can be measured with synchrotron X-ray sources which give high brightness and high resolution. As shown in Figure 20, the position of binuclear Cu–Zn species confined in ZSM-5 crystallites and the organic ligands between the Cu and Zn ions has been resolved by Rietveld refinement of synchrotron X-ray diffraction. Moreover, pair distribution function (PDF) analysis is a powerful tool to elucidate the atomic structures of the solid materials at multiple length scales (from a few angstroms to several nanometers), which is based on the total scattering analysis of the long-range order measured by Bragg peaks in XRD patterns and the short-range order in the less well-defined features in the diffraction patterns. In this regard, PDF is suitable for resolving the distribution of the two metal elements and their bonding information in bimetallic nanoclusters and nanoparticles.

100X-ray absorption spectroscopy (XAS) is the most commonly used technique for charactering the coordination environment of the metal species in solid catalysts without long-range crystalline structures, which has been widely applied for elaborating the bonding information on isolated metal atoms and metal clusters on solid carriers.− In the case of subnanometer bimetallic metal entities (binuclear sites and bimetallic clusters), both the metal–metal bonding and the metal–environment bonding can be captured by XAS based on data of fairly good quality. However, in some materials (e.g., low metal loading, low uniformity), the data interpretation will be not straightforward and only qualitative speculation can be obtained, due to the low sensitivity of XAS to the disorder structures of subnanometer bimetallic species. In these cases, the combination of multiple characterization techniques will be helpful to study the local structures of the subnanometer bimetallic species.

101In principle, the coordination environment of the metal atoms in bimetallic nanoparticles can also be clarified with XAS technique, although it will be challenging to obtain as detailed information as the cases of subnanometer bimetallic species, due to the more complicated structures of bimetallic nanoparticles. As a bulk technique, EXAFS can also be a surface-sensitive technique for characterization of metal nanoparticles by careful experimental design. For instance, by subtracting the EXAFS spectra of the reduced Pt nanoparticles and the same sample after exposure to air can give the difference, which could be associated with the oxidized Pt species on the surface.

102By applying supervised machine learning methods to the analysis of the XAS data (including both the X-ray absorption near edge structure (XANES) and extended X-ray absorption fine structure (EXAFS)) and theoretical modeling, it is possible to elaborate the distribution of the metal elements in bimetallic nanoparticles., The progress in this direction is very useful for elucidating the coordination environment of the bimetallic particles. As shown in Figure 21, after consecutive oxidation and reduction treatments, the Pd2 dimers in the PdAu alloy nanoparticles (Pd/Au = 8/92) tend to become separated due to the rearrangement of the spatial distribution of Pd and Au atoms in the alloy nanoparticles, as inferred by the analysis of the XANES and EXAFS spectra based on the neural network-assisted XANES inversion method. Thanks to the advanced data analysis method, the subtle structural changes can be captured and analyzed, leading to new insights on the coordination environment of the metal species and the metal–metal bonding (average coordination number and metal–metal bond length) in the alloy nanoparticles.

103A critical issue related to the artifacts introduced in the data fitting procedure, especially for the spectra with relatively low signal-to-noise ratios. To improve the accuracy of data interpretation in analysis of EXAFS spectra, it is proposed to employ the models derived from theoretical calculations for fitting the experimental EXAFS spectra. The theoretical models can provide insights for the possible bonding information in the sample and give useful hits for setting up the fitting parameters. This suggestion is practiced with analysis of the EXAFS spectra of supported isolated metal atoms and should be employed for measurements of supported bimetallic entities, from binuclear bimetallic species to bimetallic nanoclusters and nanoparticles.,

104Another important issue associated with the characterizations of supported subnanometer metal catalysts by synchrotron-based X-ray techniques is the stability of the tiny metal species under high-flux X-ray irradiation, which is already well-documented for metal complexes grafted on solid carriers. It is found that the reduction of Rh species supported on Al2O3 is greatly enhanced by high-flux X-ray, resulting in the agglomeration of Rh species. Taking into account that long-time measurements are frequently required for supported subnanometer metal catalysts for accumulating enough signal, the beam-induced damage to the sample and the associated effects should be considered and mitigated. When performing in situ XAS measurements with supported metal catalysts, leaching of metal species from the support to the solvent/electrolyte could occur due to the high X-ray flux. In this sense, the flux of X-ray for liquid-phase reactions should be even lower than those used in gas-phase reactions to avoid the potential catalyst loss through dissolution.

105Beside the bonding information between the metal elements, the metal–ligand and metal–support bonding information is also critical, which can be acquired by solid-state NMR. For instance, by collecting the 195Pt NMR spectra of the bimetallic PtM (M = Pd, Sn) nanoparticles and a monometallic Pt reference sample, it is possible to clarify whether the Pt and the other element is well mixed in the bimetallic nanoparticles., After the introduction of sensitive spy nucleus (1H or 31P), it is possible to achieve the rapid acquisition of 195Pt magic angle spinning NMR spectra with good data quality. This new technique can characterize the coordination environment of organometallic Pt complex grafted on solid carrier and probe the metal species neighboring the Pt complex. Solid-state NMR can be considered as averaged information in the whole sample, while the TEM-based EDS mapping can only provide very local information on the spatial distribution of the two elements. Moreover, in terms of the bimetallic nanoclusters or nanoparticles with ligands on the surface, the bonding between the ligands and the metal atoms could be directly probed by 13C NMR spectra.

106Most of the above-mentioned techniques are used to probe the structural features of the bimetallic entities. However, there are still gaps between the structural features and catalytic behavior, which can be partly filled by the characterizations of surface reactivity. This approach is usually achieved by measurements based on temperature-programmed desorption (TPD) and temperature-programmed surface reaction (TPSR). For instance, the reactivities of bimetallic nanoclusters generated by physical approaches such as size-selection method are mostly measured by these techniques. For instance, the TPD data for methanol adsorbed on the surface of AuPt clusters supported on TiO2(110) surface can not only indicate the reactivity of the AuPt clusters, but also provide insights on the exposed surface sites of the bimetallic clusters according to the variation of the desorption temperature and the final products.

107As can be seen in the abundant literature, measurements of surface reactivity of bimetallic catalysts are desirable to be performed on bimetallic model surface to give molecular- and atomic-level insights on the reaction mechanism and structures of the active sites. For instance, as displayed in Figure 22, using space-quantized oxygen molecular beam as reactant, the activation on Cu(110) and Cu3Au(110) alloy surface is measured, respectively, which indicates that O2 molecule with helicopter configuration can be activated on Cu(110) surface, while the O2 activation is prohibited on Cu3Au(110) alloy surface. Another typical example is the study of geometric structure of PdAu surface for acetoxylation of ethylene to vinyl acetate, which suggests that only the Pd atoms deposited on Au(100) surface are close enough for catalyzing the coupling reaction, while the Pd atoms on Au(111) surface are not active due to their large distance., The lessons accumulated in surface science studies can be translated to practical catalysts based on bimetallic nanoclusters/nanoparticles for fundamental understanding of the structure–reactivity relationships.

108Heterogeneous catalytic reactions normally involve the dynamic structural transformation of the solid catalysts under reaction conditions, as widely observed in both surface model systems and practical catalysts based on particles., With advanced characterization techniques such as in situ TEM and in situ XPS, the dynamic structural evolution in bimetallic nanoparticles has already been revealed, including the rearrangement of the spatial distribution of the two elements, change of the exposed surface and interfacial structure, and change of chemical states and particle sizes. For instance, as revealed by in situ TEM, Ni@Au core–shell nanoparticles show reversible temperature-dependent structural transformation under CO2 hydrogenation conditions. Au shells are present at low reaction temperatures, while at high reaction temperatures, bimetallic AuNi alloy shells are formed due to the migration of Ni atoms into Au lattice. Beside the segregation under reductive atmosphere, oxidation-induced structural reconstruction is also widely observed in numerous systems. Specifically, the oxyphilic metals such as 3d metals have higher affinities than the noble metals such as Pt and Au, and therefore, upon exposure to oxidative atmosphere, surface segregation of oxyphilic metals will occur, resulting in the formation of oxide patches or layers on the surface of bimetallic nanoparticles. In some cases (see summary in Figure 23), the treatment-induced structural transformation is reversible, while in some cases, the transformations are irreversible, causing gradual segregation or sintering of the bimetallic entities, which can probably be associated with the variation of catalytic performances.,

109Regarding binuclear species and bimetallic nanoclusters, their structure evolution behavior will be more elusive than nanoparticulate counterparts due to their smaller sizes. Indeed, limited by characterization techniques, there are few reported works on structural transformation of subnanometer bimetallic species. The prior works on structural evolution of subnanometer monometallic species already show that metal entities with smaller sizes will be more sensitive to the environment than the larger ones. It can be expected that when varying the reaction conditions, the coordination environment, morphology, and chemical states of the subnanometer bimetallic species will change accordingly. To capture these subtle structural changes, characterization techniques with high spatial and temporal resolution are needed to reveal the nature of the active sites under reaction conditions. For instance, as illustrated in Figure 24, the reduction and migration of Sn species along the microporous channels of MFI zeolite to interact with Pt clusters to form bimetallic PtSn clusters has been directly followed by quasi in situ TEM and CO-IR spectroscopy.,

110The dynamic structural transformation could occur during the characterization procedure (e.g., the activation process of the sample or the high-vacuum environment inside the instrument), and the structural information obtained under those conditions may not be able to reflect the representative structures under reaction conditions. Specially, this point should be taken into account when dealing with bimetallic binuclear sites and nanoclusters, whose structures are quite flexible.

111Each characterization technique has its advantage for measuring or detecting some featured properties of the bimetallic entities but also has its limitations in clarifying the detailed structures and distinguishing one type of entity from others. The low sensitivity of EXAFS to disorder metal species, especially the oxidized metal species, may lead to controversial conclusions in comparison with the characterization results obtained with high-resolution transmission electron microscopy. Therefore, it is highly recommended to employ various characterization techniques in a concerted way in order to obtain structural information on the bimetallic entities in a global view. Furthermore, by correlating the information derived from different techniques, it is helpful to avoid the pitfalls that maybe encountered in the process of characterizing binuclear species or bimetallic nanoclusters, because some techniques are established on nanoparticles but not directly applicable to subnanometer metal species.

112Resolving the full structure (i.e., precise characterization of the particle size, composition, and three-dimensional geometric structure) of supported bimetallic clusters is highly challenging. This goal has been achieved in quite a few examples by the combination of advanced electron microscopy (aberration-corrected electron microscopy) and spectroscopy (high-resolution energy dispersive spectroscopy and electron energy loss spectroscopy) techniques. However, these techniques can only give the three-dimensional structures of bimetallic clusters in a very local region, which may not well represent the whole sample. In our opinion, there are two prerequisites for the resolution of the full structure of bimetallic clusters in a supported bimetallic catalyst: (1) a sample comprising bimetallic clusters with highly uniform structural features (particle size, composition, etc.) and (2) a combination of advanced electron microscopy and spectroscopy techniques. By fulfilling the prerequisites, the structural information derived from the characterization can represent the properties of the supported bimetallic clusters and can be further used for establishing the structure–reactivity relationship.

113The use of binuclear metal sites for catalysis has been studied by researchers in the field of inorganic and organometallic chemistry for a long time. In this section, we will briefly summarize the catalytic applications of binuclear bimetallic sites in natural enzymes and binuclear bimetallic complexes prepared by chemists and then discuss the development of supported binuclear bimetallic sites for various catalytic reactions.

114Hydrogenation reactions generally involve the activation of H2 and the transfer of the activated H species to the target molecule. In the natural [NiFe] hydrogenase, the neighboring Ni and Fe atoms will active H2 by forming a Ni···H···Fe intermediate and then reduce the substrate molecule. The idea of using binuclear bimetallic sites for hydrogenation reactions has also been practiced with organometallic complexes. For instance, a binuclear PtRh complex shows remarkable reactivity for the reduction of N2O to N2 with H2. The high reactivity of the binuclear PtRh complex relies on the synergy of Rh and Pt sites for activation of N2O and H2. More examples can be referred to some review articles in this field. A key lesson that can be learned from these enzymatic and homogeneous catalysts is that the dissociation of H2 is achieved by the involvement of two metal atoms, as proved by detailed mechanistic studies.

115By immobilizing metal complex precursors on solid carriers, binuclear bimetallic sites can be generated which can show enhanced stability than the enzymatic and homogeneous counterparts and then can be used for catalyzing hydrogenation reactions. For example, binuclear PtZn sites are formed after depositing organometallic Pt complex on Zn-modified silica by ALD and the resultant bimetallic sites show good activity for both selective hydrogenation of 1,3-butadiene and nitroaromatics. Besides, the promoting role of the metal atom to its neighboring atom has been demonstrated with a supported binuclear IrMo catalyst for selective hydrogenation of functional nitroaromatics. As shown in Figure 25, the Ir and Mo atom can serve as the active site for H2 activation and adsorption of the nitro group, respectively, and the cooperation of the two sites leads to enhanced activity and selectivity for the production of substitute aniline. In another example, binuclear PdCu sites generated on nanodiamond graphene show enhanced activity for selective hydrogenation of acetylene to ethylene than the isolated Pd and Cu sites due to optimized electronic feature of Pd atoms modified by the neighboring Cu atoms in the binuclear PdCu sites.

116Isolated metal atoms supported on solid carriers (such as Co atoms supported on N-doped carbon, Pt atoms supported on FeOx, etc.) have been reported as catalysts for selective hydrogenation reactions., In principle, the formation of binuclear bimetallic sites may further improve the catalytic performance because the H2 dissociation should be more favorable on two metal atoms than that on an isolated metal atom, and the chemoselectivity may also be modulated after the introduction of a neighboring metal atom because the adsorption configuration of the reactant can be influenced by the surrounding environment of the active sites, leading to chemoselective adsorption of the target groups. Therefore, we anticipated that more works in this direction will appear in the near future.

117As the reverse process of hydrogenation, dehydrogenation reactions also widely exist in nature. For instance, alcohol metabolism relies on the participation of alcohol dehydrogenase, in which isolated Zn(II) sites are coordinated to the active site to assist the adsorption and activation of alcohol. In terms of homogeneous systems, both mononuclear and binuclear metal complexes have been employed for various dehydrogenation reactions, and in some cases, the binuclear bimetallic complexes give remarkable reactivities. Similar to the enzymes, dehydrogenation of alcohols to aldehydes can also be catalyzed by supported isolated metal species, such as Rh and Pt., For example, it is found that isolated Rh atoms supported on carbon are much more active than Rh nanoparticles for dehydrogenation of cyclohexanol to cyclohexanone. Following the above discussion, it will be interesting to check the possibility to incorporate another metal element as the neighboring site to Rh to further enhance the activity.

118In heterogeneous systems, one of the most important dehydrogenation processes should be the dehydrogenation of light alkanes (propane, butane, and isobutane) for the production of light alkenes. Currently, supported CrOx and Pt catalysts are used in industrial processes. In the case of supported Pt catalysts, the active sites are generally considered to be small Pt clusters, while in the case of supported CrOx, atomically dispersed Cr(III) species are most likely to be the active species.,

119Besides the materials based on supported mononuclear metal species, binuclear bimetallic species have also been reported as efficient catalysts for the dehydrogenation of alkanes. For instance, as described in Figure 26, the formation of binuclear CoZr sites is markedly more active than the monometallic Co and Zr sites for propane dehydrogenation, which is ascribed to the modification of the electronic structure of Co by Zr due to the metal-to-metal charge transfer. This enhancement effect has also been found with binuclear ZnAl sites, which are more active than isolated Zn species. Taking into account that a large number of supported isolated metal species (Cr, Fe, Ga, In, Sn, etc.) have been suggested as active sites for alkane dehydrogenation,, the combination of these candidate elements could lead to the discovery of highly active catalysts comprising binuclear bimetallic sites.

120In natural enzymes, the oxidation of CO to CO2 can be realized by CO dehydrogenase, which contains MoCu bimetallic sites to perform the oxidation of CO with water (CO + H2O → CO2 + 2H+ + 2e–). The Mo and Cu atoms are bridged by S, and Mo is considered as the site for performing the redox process (between MoIV/CuI and MoVI/CuI), while Cu is responsible for the CO capture and catalytic turnover. After the oxidation of CO to CO2, the MoVI site is regenerated by extracting oxygen from water. Solid catalysts made by mixed Cu-based oxide nanoparticles have been widely employed as excellent catalysts for CO oxidation, such as Cu–Mn mixed oxides, and the active sites in these solid catalysts are proposed to be associated with the Cu–O–Mn species,, which could also be considered as motifs as binuclear bimetallic sites in CO dehydrogenase. The critical role of forming bridged M1–O–M2 species in mixed oxide catalysts for CO oxidation reaction has been validated by surface science study, in which the isolated Ni atoms deposited on CuO surface can catalyze the CO oxidation through the Eley–Rideal mechanism.

121By using N-doped carbon as the support, binuclear CoFe sites can be stabilized for mimicking the isolated CuMo sites in natural enzymes. As shown in Figure 27, CO is adsorbed on Co atoms, while O2 is activated on Fe atoms and the cooperation of the binuclear CoFe sites leads to much higher activity for CO oxidation reaction at a very low temperature (−73 °C) than the monometallic Co and Fe sites. Because low-temperature CO oxidation has also been achieved with Co3O4 nanoparticles, it will be interesting to study the similarity and dissimilarity between the Co atoms at the exposed facet of Co3O4 nanocrystals and the binuclear CoFe sites embedded in N-doped carbon matrix. Other applications of binuclear bimetallic sites embedded in N-doped carbon matrix have been demonstrated with CoCu sites for oxidative esterification of furfural to methyl 2-furoate and oxidative coupling–dehydrogenation cascade reaction of benzyl alcohol and 2′-hydroxyacetophenone for the synthesis of flavones. It is proposed that the neighboring Cu atom can promote the O2 activation on the Co atom and thus improve the yields of esters for the oxidative esterification of aldehydes.

122By comparing the atomic structures of the binuclear sites and the reaction mechanism, we can find that the models proposed in the two works are different (metallic bonding is formed in the CoFe sites, while the Co and Cu atoms are not directly bonded in the other case). This discrepancy could be related to the difference in materials preparation procedure, but also could be caused by the difficulty of precisely elucidating the atomic structures of the binuclear bimetallic sites on the solid carrier, as discussed before in this review. Nevertheless, the two atoms in CoFe sites are proposed to account for the activation of CO and O2, respectively, while the Co atoms in CoCu sites seem to play a predominant role in oxidative esterification of aldehydes.

123The activation of O2 is the critical step for oxidation reactions and the involvement of two or more atoms in the O2 activation step is necessary, as suggested by theoretical studies. However, during the past decade, isolated metal atoms supported on solid carriers have also been proposed as active sites for numerous oxidation reactions., From a mechanistic point of view, the investigations on supported binuclear bimetallic sites for oxidation reactions can help us clarify the reaction mechanism of the emerging single-atom catalysts based on isolated metal atoms. Moreover, in the reported works, the role of support in the supported binuclear bimetallic sites is not fully elucidated. The study with the model system based on Pt2 dimer supported on Fe3O4 surface indicate that the support together with the Pt2 dimer show reversible structural transformation during the active O species. These insights could be translated to the bimetallic systems, although the binding sites of O2 and the detailed mechanism could be more complicated than the monometallic systems.

124Besides the activation of O2, the selective oxidation of organic molecules may invoke other key steps during the catalytic cycle. For instance, the binuclear IrCu sites supported on In2O3 exhibit superior performance for selective oxidation of biomass-derived isoeugenol to vanillin. The presence of an adjacent Cu atom to the Ir atom can promote the activation of O2 on the Ir site, and the Cu site accounts for the subsequent oxidation of the C=C bond to the epoxy group and then facilitates the hydrolysis of the epoxy group and the oxidative cleavage of C–C bond to generate the vanillin product.

125A large portion of the literature works on the catalytic application of binuclear metal sites are related to the transformation of organic molecules, and therefore, the activation of C–H bonds will be a critical step in many of these catalytic reactions. The use of supported TaIr bimetallic sites for activation of C–H bonds in the aromatic ring has been tested by performing an isotopic exchange experiment. The TaIr bimetallic sites are quite active for catalyzing hydrogen/deuterium exchange between C6D6 and C6H5F, while monometallic Ta sites are much less active. Moreover, the TaIr sites with Ta–H and Ir–H bonding are the active sites, while the Ta and Ir fully coordinated with metal–carbon bonding are not.

126Hydroformylation of olefins is a very important process in the current chemical industry for the production of aldehydes. Currently, molecular Rh catalysts are employed in industrial processes and the addition of another metal can promote the reactivity of the mononuclear Rh catalyst because the binuclear bimetallic complex can facilitate the production of aldehyde through the binuclear elimination reaction within the RhRe complex., Following the works in the homogeneous systems, binuclear RhRe sites are supported on Al2O3 as heterogeneous binuclear bimetallic catalysts for the hydroformylation of ethylene. The introduction of atomically ReOx species in the neighboring site of isolated Rh atoms can cause the modification of the electronic structure of Rh atoms, as proved by the shift of the CO stretch frequency of the Rh(CO)2gem-dicarbonyl species. A dedicated theoretical study on the supported Rh–Re binuclear species suggests that the presence of Re at the neighboring site of Rh can weaken the Rh–CO bonding and the rate-determining step for the hydroformylation reaction is also switched from the acylation step to the requisite CO coordination before the insertion step. The mechanism for the promotion effect of Re to Rh catalyst is different in the homogeneous and heterogeneous systems, and the difference is proposed to be caused by the different reaction pathways, which should be related to the geometric and electronic differences between molecular binuclear complex and supported binuclear species. As shown in Figure 28, due to the modification of the electronic properties of Rh sites, the selectivity toward propanal is greatly improved in comparison to the monometallic Rh catalyst in hydroformylation reaction of ethylene, because the hydrogenation of ethylene is suppressed. The concept of constructing bimetallic sites for hydroformylation reaction has also been extended to Rh–WOx pair sites, which catalyzes ethylene hydroformylation reaction through a mechanism involving Rh-assisted WOx reduction, transfer of ethylene from WOx to Rh atom, and H2 activation and dissociation at the Rh–WOx interface.

127Another important reaction involving the activation of CO is the carbonylation of methanol for the production of acetic acid, which is catalyzed by a homogeneous Rh complex in the industrial process. Additionally, Ir-based catalysts have also been developed for this process and the addition of La to the Ir/C catalyst has been quite effective to enhance the performance. Mechanistic studies reveal that La can facilitate the normally rate-limiting CO insertion step in Ir-catalyzed carbonylation of methanol., Furthermore, the presence of La as the neighboring site to the Ir atom will suppress the leaching and sintering of atomically dispersed Ir species.

128The stability of binuclear bimetallic sites under the conditions for organic transformations could be a critical issue for IB group metals (such as Ag and Au) with higher reducibility than the VIII group metals, which limits their practical applications. For instance, isolated Au atoms supported on carbon suffer the sintering issue during the hydrochlorination of acetylene, due to the reduction of oxidized Au species (Au(I) and Au(III) species) into Au nanoparticles. As shown in Figure 29, the addition of isolated Pt atoms to the Au/C catalyst can effectively mitigate the sintering of Au atoms into Au nanoparticles because the migration and reduction of Au species are suppressed. Beside the addition of a noble metal, the introduction of alkali metals can also improve the stability of highly dispersed Au species supported by solid carriers. For example, the addition of cesium can also effectively enhance the stability of atomically dispersed Au species under the conditions of acetylene hydrochlorination reaction, because the Cs+ can inhibit the reduction and sintering of catalytically active Au(III) species to metallic Au. However, in both of the above examples, there is no direct experimental evidence of the formation of metal–metal bonding in the bimetallic catalysts. Instead, it is proposed that the second metal is anchored on the neighboring site of the Au(III) atoms embedded in the carbon support, which increases the energy barrier of the reduction and migration of Au atoms into Au nanoparticles on the carbon support. The distance between the two types of metal sites seems to be longer than the bimetallic binuclear sites discussed before in this section for electrocatalytic ORR reaction. Therefore, it is interesting to carry out further studies with these bimetallic catalysts to figure out the influence of the distance between the Au and the promotor metal species on the stability and catalytic reactivity, and thereby elucidate the structural features of the Au-based bimetallic sites.

129The activation of alkanes has also demonstrated the critical role of binuclear bimetallic sites. Alkane metathesis, a very challenging reaction for adjusting the chain length of the inert alkanes, is considered to proceed via several elementary steps involving C–H activation and olefin metathesis. Monometallic sites such as isolated Zr, W, and Ta species supported silica show activity for metathesis of propane., To further improve the catalytic efficiency of alkane metathesis reaction, two types of isolated metal sites (W and Ti) are grafted on the support to work synergistically for the C–H activation and subsequent β-elimination step (see Figure 30), resulting in a significant improvement in activity for propane metathesis when increasing the intimacy of the two metal sites. However, according to the synthesis method and the structural characterization of the supported organometallic W–Ti catalyst, the distance between the W and Ti sites is still unknown. The surface organometallic chemistry allows the precise control of the spatial proximity of the two types of metal sites on silica surfaces by employing different organometallic complexes as the precursor. It will be interesting to directly graft the bimetallic W–Ti complex on the silica support to generate neighboring W–Ti sites to verify whether the alkane metathesis reaction requires bifunctional sites with atomic proximity.

130Supported isolated metal atoms have been reported as motifs homogeneous Fenton reagents, which can efficiently initiate the production of radicals in solution for radical reactions. For instance, isolated Fe atoms supported on N-doped carbon have been shown as active catalysts for activating peroxymonosulfate or H2O2 to produce ·OH species, which are responsible for the degradation of organic compounds and oxidation of benzene to phenol.− The catalytic performance of the isolated Fe atoms is strongly dependent on the coordination environment. When increasing the coordination number of Fe–N bonding in the Fe–N–C catalyst, the reactivity for benzene oxidation reaction is greatly promoted due to the generation of highly active O=FeN4=O species.

131From a mechanistic point of view, the formation of radical species involves the activation of the precursor compound (e.g., H2O2, peroxymonosulfate, etc.) on the metal sites and subsequent dissociation of these compounds to form radical species. The electronic structures of the metal sites have profound impacts on the reactivity because of the electron transfer between the metal sites and the reactants. The advantages of binuclear bimetallic sites over the isolated metal sites for the generation of reactive radical species have been exhibited with MoZn and CuFe sites supported on carbon, which is ascribed to the enhanced capability for cleavage of O–O bonds in the peroxy compounds and the modification of the adsorption energy of hydroxyl.,

132Beyond the reported works, the binuclear bimetallic sites are also promising catalysts for selective oxidation of light alkanes to oxygenates. Theoretical calculations suggest that FeNi and FeCo sites give much lower activation energies for C–H activation in methane than that on the Fe2 site. Taking into account that Fe2 sites stabilized in ZSM-5 can catalyze the conversion of methane to CH3COOH, which involves the coupling of CH4, CO, and ·OH produced from H2O2. Moreover, isolated Fe sites at the node of Fe-MOF structure can catalyze the oxidation of ethane to ethanol by using N2O as oxidant, which involves the Fe(IV)-oxo species as active species. Therefore, it will be interesting to explore the oxidation of light alkanes (ethane, propane etc.) with highly active radical species on binuclear bimetallic catalysts, to improve the activity and improve the selectivity of desired products. In the reported works, oxidation of ethane with H2O2 can be catalyzed by solid catalysts and a variety of products (including acetic acid, CO2, formic acid etc.) are formed with low turnover numbers. The employment of solid catalysts with regular binuclear sites may offer the opportunity to perform the oxidation of light alkanes with high chemoselectivity and regioselectivity.

133Molecular bimetallic complexes have been reported as photocatalysts for a variety of reactions, such as water oxidation and CO2 reduction. For example, the binuclear CoZn complex gives a much higher turnover number than monometallic CoCo and ZnZn complexes for photocatalytic reduction of CO2 to CO. The advantage of the bimetallic complex over the monometallic complex is ascribed to the enhanced absorption of OH- on ZnII sites, which facilitate the proton-coupled electron transfer and subsequent cleavage of C–O in CO2 for production of CO.

134As shown in Figure 31, by immobilizing the bimetallic complexes on solid carriers, they can work directly as the functional components for photocatalytic reactions or can serve as the cocatalysts of the light absorbents (e.g., semiconductors, plasmonic metals, etc.). In the former case, binuclear ZrCo sites are generated on mesoporous silica by grafting the organometallic Zr and Co precursor consecutively and the resultant binuclear ZrCo sites show activity for photocatalytic reduction of CO2 to CO in the presence of the sacrificial agent. The charge transfer between the CoII and ZrIV atoms is responsible for the light absorption and charge separation. Under light irradiation, the metal-to-metal charge transfer from CoII to ZrIV and then to the organic molecule has been directly followed by transient spectroscopy, while the reversed charge transfer pathway is prohibited, showing a rectified charge transfer behavior. Moreover, by combining the ZrCo sites with IrO2 nanoclusters, the photocatalytic reduction of CO2 to CO can be achieved directly with H2O in the absence of a sacrificial agent, thanks to the electron transfer from IrOx nanoclusters to the binuclear ZrCo sites and the enhanced oxidation of water on IrOx. The combination of ZrCo sites with CuOx nanoclusters can lead to an active system for photocatalytic reduction of CO2 to CO, as a result of the electron donation from ZrCo sites to CuOx nanoclusters and the favored reduction of CO2 on CuOx.

135In the latter case, the binuclear bimetallic sites supported on semiconductors can be efficient cocatalysts for photocatalytic H2 evolution, as demonstrated with CoPt sites supported on TiO2 nanosheets. The TiO2 nanosheets modified with CoPt sites are more active than those modified with isolated Pt atoms and Pt nanoclusters as cocatalysts. The electronic structure of the Pt atoms is modulated by the neighboring Co atoms, giving to promoted activity for the H2 evolution reaction. As demonstrated with CuIn sites supported on polymeric carbon nitride, the synergy of Cu and In sites contribute to the photocatalytic reduction of CO2 to ethanol via performing the C–C coupling of the intermediate CO adsorbed on the neighboring Cu and In site, respectively.

136Because of their versatile structures, MOFs can serve as a class of platform materials for installing different functional components to achieve catalytic transformations involving multiple steps. For instance, by incorporating isolated metal sites through the coordination interaction with the organic linker, both photosensitizer and catalytically active sites can be constructed in MOF structure for photocatalytic transformations, such as CO2 reduction and C–N coupling reactions., Although it is proposed that the photosensitizer and the transition metal sites are located inside the same cavity of the MOF structure, the two types of functional sites are separated in the MOF framework structure. Considering that photocatalytic transformations require the electron transfer from the photosensitizer to the transition metal sites, the energy transfer efficiency should be improved when the distance between the two functional sites is minimized. One possible approach is to incorporate binuclear bimetallic sites at the nodes of the MOF structure or binuclear complex at the linker.

137As discussed in section 3 on the generation of binuclear bimetallic sites, a large portion of the reported catalysts is based on the binuclear species anchored on carbon-based materials, which are potential catalysts for electrocatalytic reactions because the carbon supports serve as the electron transport media. Developing non-noble metals as substitutes for Pt catalysts for fuel cells has attracted enormous attention among the electrocatalysis community in the last decades. According to many reported works, isolated metal atoms embedded in the N-doped carbon matrix (M–N–C species, M = Fe, Co, Ni, etc.) are considered the most promising candidate materials., Because the coordination environment of the isolated M species has profound impacts on the catalytic performance, the incorporation of a neighboring metal to the M to form binuclear bimetallic sites should also influence the reactivity of the M sites, as inferred by the theoretical calculations. Specifically, in the model constructed based on adjacent Co and Zn sites anchored on nitrogen-doped graphene, the d-band electronic structure of the Co site is influenced by the neighboring Zn site, and therefore, the bonding strength of the reaction intermediates for electrocatalytic oxygen evolution and reduction reactions are altered to enhance the catalytic performance.

138In terms of the experimental studies, binuclear CoFe sites embedded in N-doped carbon derived from the pyrolysis of a CoFe-MOF precursor are found to be more active than the mononuclear metal sites prepared in a similar procedure for electrocatalytic oxygen reduction reaction (ORR). According to the structural characterization, it is proposed that the bimetallic FeCoN5–OH sites are ∼20 times more active than the monometallic FeN4 site because the Co and Fe atoms can work synergistically to enhance the cleavage of O–O bonding in O2. In another work, the binuclear CoFe sites are also generated by pyrolysis of a CoFe-MOF precursor, and the atomic structure of the active site is proposed to be CoFe-N6 moieties, which are more active than the monometallic counterparts for both oxygen reduction and evolution reactions (see Figure 32). The performance of the optimized catalyst with CoFe binuclear sites show more superior performances than the benchmark RuO2 and Pt/C catalysts for OER and ORR, respectively. The discrepancy between the above two works in terms of the structures of the active sites could be caused by the differences in the experimental procedure for catalyst synthesis. Nevertheless, it should be emphasized again that the determination of the coordination environment of the binuclear bimetallic sites on solid carriers is still a very difficult task, limited by the characterization techniques.

139Thanks to the progress made in terms of controllable synthesis of binuclear bimetallic sites, it is possible to systematically study and compare the reactivity of different combinations of transition metals for electrocatalytic reactions. For instance, by in situ activation of a Ni–N–C catalyst comprising atomically dispersed Ni species in the electrolyte containing Fe3+, binuclear NiFe sites could be generated under the conditions for electrocatalytic reaction, as probed by in situ EXAFS. The in situ generated binuclear NiFe species are much more active than the isolated Ni and Fe atoms. It should be noted that, compared to the above-mentioned works, the NiFe sites generated under in situ conditions have more open coordination space than those embedded in the N-doped carbon matrix. By varying the metal ions in the electrolyte and the metal in the carbon matrix, the catalytic performances of various binuclear bimetallic sites have been studied, among which CoFe sites deliver the best performance for electrocatalytic oxygen evolution.

140The advantages of binuclear bimetallic sites for electrocatalytic oxygen reduction/evolution reactions have been demonstrated with other binuclear bimetallic sites embedded in N-doped carbon matrix, such as the IrCo, CoFe, NiFe, RuCo, and IrFe sites.− In most of the proposed atomic structures of the binuclear sites, the two metal atoms are coordinated by six N atoms and bridged by two N atoms, although there are differences in the synthesis methods among these works. Despite the numerous reports on application of binuclear bimetallic sites for electrocatalytic oxygen reduction/evolution reactions and the similarities in the structure of the active sites, benchmarking of the catalytic performances of various binuclear bimetallic sites under the same testing conditions is urgently needed to achieve a unified understanding on the structure–reactivity relationship in this direction.

141Although some works show the enhanced catalytic performance of binuclear bimetallic species over the isolated metal sites, the long-term stability of the binuclear bimetallic sites under electrocatalytic conditions are not fully explored and compared with the classic systems. From a mechanistic point of view, a major reason causing the deactivation of the classic M–N–C materials is related to the degradation of the carbon support and the subsequent leaching of metal species into electrolyte under electrocatalytic conditions. In this sense, the stability of the binuclear bimetallic species against leaching and sintering should be carefully examined and compared with the materials comprising isolated metal species. It is possible that, by using binuclear bimetallic sites, the degradation of the carbon support caused by the radicals from Fenton reaction could be suppressed. Despite the tremendous efforts made to clarify the nature of the active sites in non-noble-metal based electrocatalysts (most of the works are focused on the M–N–C systems), there are still some arguments on the active sites due to the complicated composition of the M–N–C materials derived from pyrolysis treatments and the role of agglomerated metal species (in the form of metallic nanoparticles or carbide nanoparticles encapsulated in thin carbon layers) in the M–N–C catalyst is still not fully unveiled.

142Moreover, the superior electrocatalytic performance of binuclear bimetallic sites have also been observed for other electrocatalytic reactions. For example, WMo bimetallic sites exhibit excellent activity for electrocatalytic H2 evolution in a large pH range. By extensive theoretical studies, it is proposed that, the W and Mo atom are located in the defective sites of N-doped graphene and stabilized by Mo–O and W–O bonding. The W–O–Mo–O–C ensembles are responsible for the high activity for H2 evolution in pH-universal electrolysis. RhFe bimetallic sites presents enhanced activity for electrocatalytic H2 evolution in acid medium than the reference materials comprising Rh or Fe nanoparticles, which is ascribed to the bonding between Rh and Fe atom. The important role of chemical bonding between the two metal atoms is also reflected in the bimetallic AuRu sites for electrocatalytic H2 evolution reaction. Because Au is more electronegative than Ru, electron donation from Ru to Au may occur, which facilitates the H+ adsorption according to the d-band center theory.

143In addition, binuclear bimetallic sites show promising performances for electrocatalytic CO2 reduction. Binuclear NiFe and NiCu sites embedded in N-doped carbon matrix have been reported as efficient catalysts for reduction of CO2 to CO, although the proposed atomic structures of the binuclear sites are different, according to the structural characterizations by AC-STEM and EXAFS., In the case of NiFe sites, it is claimed that the two metal atoms coordinate with six N atoms via forming two bridged Ni–N–Fe bonding. In the case of NiCu sites, the proposed structure consists of Ni and Cu atoms bonded with six N atoms, but in a nonbridged configuration. The different structural features of the binuclear bimetallic sites could be caused by the different electronic properties of Fe and Cu atoms, as implied by the theoretical calculations. In both works, the formation of binuclear bimetallic sites contributes the facile formation of *COOH intermediates, which is generally considered as the rate-determining step for electrocatalytic CO2 reduction. It occurs by modulating the electronic structures of the Ni atom with a neighboring Cu/Fe atom, resulting in enhanced faradic efficiency for CO production than the single-atom metal catalysts at low overpotentials, as shown in Figure 33. The remarkable advantages of binuclear bimetallic sites over the single-atom sites for electrocatalytic reductio of CO2 to CO have also been found in other combinations such as CuCo and NiIn sites, although there are some differences in the proposed structures of the binuclear sites.,

144Besides the promotion effect ascribed to the electronic modification, the synergy of two metal atoms for CO2 activation could also be a plausible explanation for the enhanced activity for electrocatalytic CO2 reduction. Inspired by the active site of CO dehydrogenase (CODH), a metalloporphyrin-based mimic has been synthesized based on the combination of Zn and Fe metalloporphyrin complex. The cooperativity between the two metal atoms was demonstrated by the enhanced performance in reduction of CO2 to CO in comparison with the mononuclear analogue. It is proposed that the activation of CO2 is proceeded through the adsorption of CO2 on both metal centers. The synergistic effects of binuclear bimetallic sites for electrocatalytic CO2 reduction have also been observed in other combinations., The studies based on well-defined molecular complexes provide fundamental understanding on the atomic structures of the active sites in binuclear bimetallic sites and could be translated into the studies based on binuclear sites embedded in carbon matrix.

145However, none of the above-mentioned binuclear bimetallic sites can achieve the production of C2+ products (such as ethylene, ethanol etc.), although these C2+ products are commonly observed in CO2 reduction reactions with Cu and Co nanoparticles., One plausible explanation could be the absence of domains with multiple metal atoms in the binuclear sites for C–C coupling reactions to form C2+ products. In this regard, the combination of binuclear bimetallic sites and Cu nanoparticles may lead to superior performance for reduction of CO2 to C2+ products via a tandem catalytic process because the CO2 to CO and CO to C2+ products can proceed separately.

146The contribution of the two competing reactions (CO2 reduction and H2 evolution reaction) at the cathode region can be tuned to give syngas with desired CO2/H2 ratio. For example, ZnLa bimetallic sites supported on N-doped carbon show tunable activity for electrocatalytic reduction of CO2 to CO and the ratio of CO/H2 in the resultant product can be modulated via tuning the number of Zn and La sites because the Zn and La sites tend to produce CO and H2, respectively.

147Limited by the availability of catalysts comprising binuclear bimetallic sites, the applications of this type of materials are not as abundant as bimetallic nanoclusters and nanoparticles, up to now. However, there are some examples showing promising catalytic performance of monometallic binuclear sites. For instance, binuclear Ir species supported on Fe2O3 show better activity than isolated Ir atoms and Ir nanoparticles for photocatalytic water oxidation reaction. Binuclear Ir species supported on SiO2 and MgO show activity for hydrogenation of ethylene. Besides, the binuclear Fe species stabilized on C3N4 exhibit remarkable activity and selectivity for epoxidation of epoxidation of trans-stilbene, while isolated Fe atom and Fe nanoparticles are barely inactive. The unique catalytic behavior of binuclear Fe species is also reflected for oxidation of CH4 in the presence of CO and H2O2, giving to the production of acetic acid. The use of monometallic binuclear sites for alkane activation reactions have been shown with binuclear Cu and Ga sites, which are stabilized in the micropores of zeolite structures., More examples can be found with the application of monometallic binuclear sites for electrocatalytic reactions, such as the binuclear Co and Fe sites embedded in N-doped carbon for oxygen evolution reaction and oxygen reduction reaction., The above-mentioned experimental works have also shown the advantages of monometallic binuclear sites over isolated metal atoms or nanoparticles. The catalytic performances could be further improved by substituting the monometallic binuclear sites by bimetallic binuclear sites and the choice of the bimetallic combination could be inferred by theoretical calculations.,− This strategy has already been practiced in many theoretical works for screening in the structural configurations of the potential binuclear bimetallic sites. Especially, by employing the calculations based on machine-learning strategies, the efficiency of theoretical screening will be greatly promoted.

148The choice of support for binuclear sites has remarkable impacts on their catalytic properties. For example, metallic Pt2 clusters can be stabilized in the microporous channels of MOF structure, and these binuclear species give remarkable activity for low-temperature water–gas shift reaction. The electronic properties of the Pt2 clusters are different to those supported on inorganic solid carriers, which are positively charged. Translating this lesson to binuclear bimetallic sites, screening suitable host for the binuclear sites and establishing the structure–reactivity relationship will be critical for developing binuclear bimetallic sites with novel catalytic properties.

149For developing novel supported binuclear bimetallic catalysts, one may achieve that goal by translating the knowledge and fundamental understandings obtained with enzymatic and homogeneous systems. This paradigm could be quite applicable when the target reactions are those dealing with organic molecules, because the rate-limiting steps of the target reaction could be quite similar to those catalyzed by enzymes or molecular metal complexes. By mimicking the composition and structural features of the corresponding enzymes or metal complexes, one could develop active and selective catalysts based on binuclear bimetallic sites. For instance, in conventional the Sonagashira couple reactions, it usually requires the use of both Pd and Cu complex for achieving high reaction efficiency, because Cu is generally considered as a promotor metal for Pd. It has been shown that bimetallic PdCu nanoparticles have shown advantages over monometallic Pd and Cu particles. However, the leaching of nanoparticulate Pd and Pd-based nanoparticles in C–C coupling reactions is widely observed, therefore, the nature of the active sites in the bimetallic system remains unknown. It is found that the use of bimetallic PdCu complex to maximize the synergy of Pd and Cu metal sites and the reaction mechanism in the elementary steps, such as the formation of aryl/acetylide species and C–C elimination. Following these works, it is interesting to develop binuclear PdCu species stabilized on solid carriers such as carbon-based materials or porous matrix for C–C coupling reactions.

150In some catalysts comprising monometallic binuclear sites, the chemical states of the two metal atoms could be different, caused by their different coordination environment. For instance, the binuclear Cu sites supported on PdTe nanowires are proposed to be paired Cu0–Cux+ species. The metallic Cu0 accounts for adsorption of CO2 while the positively charged Cux+ accounts for the adsorption of H2O, resulting in high activity and selectivity for reduction of CO2 to CO. Taking into account the possible scenarios, in which the same type of metal atoms at neighboring sites show different chemical states, the synergy of the two metal atoms in different chemical states may work like two atoms of different elements. This type of binuclear structures could commonly exist in solid catalysts comprising atomically dispersed 3d transition metals (Fe, Co, Cu, etc.), because these metals can exhibit flexible transitions between different chemical states.

151In recent years, we have witnessed the renaissance of electrocatalysis for organic synthesis, and the design of efficient solid electrocatalysts for organic reactions has been one of the frontiers in catalysis community, especially in the field of biomass conversion., Considering the use of bimetallic molecular complexes in numerous organic synthesis procedures, it is logical to expect that binuclear bimetallic sites supported on carbon could be promising catalytic materials for organic transformations.

152In most of the catalytic reactions discussed above, monometallic sites can already catalyze the transformations, although the use of binuclear bimetallic sites may perform better. More attractive examples for extending the potential of binuclear bimetallic sites will be the ones that can only be achieved with the combination of two metal atoms while a single metal atom is not sufficient. In this sense, there is much space for the development of binuclear bimetallic sites for tandem catalytic processes, in which each of the metal atom accounts for at least one elementary step. In this sense, the synergy between the two metal atoms will determine the catalytic performance of the whole process.

153The identification of mechanism of binuclear bimetallic sites is still a very challenging task, because the adsorption sites for the substrate molecules and potential dynamic structural transformations of the binuclear species are very difficult to be unveiled. As a consequence, many of the literature works rely on theoretical calculations to understand the reaction mechanism. A potential approach to acquire molecule-level insights on the reaction mechanism is to use the organometallic complexes to mimic the corresponding binuclear bimetallic species and the reaction mechanism based on the well-defined organometallic complexes, which can be studied by well-established tools and methods (NMR, X-ray diffraction etc.), can provide direct insights on the counterpart species supported on solid carrier. This paradigm has been practiced with the application of model metal complexes to study the functional mechanism of metalloenzymes, such as the employments of binuclear Fe complex for mimicking NO reductase and FeAl binuclear bimetallic complex for activation of CO2.,

154The examples discussed in this section are mostly focused on binuclear bimetallic sites stabilized on solid carriers, especially for those bimetallic sites embedded in N-doped carbon matrix. For those species, their coordination environments are largely constrained by the metal–support interaction. However, binuclear metal entities can be produced in a highly dynamic manner under reaction conditions, as demonstrated with the formation of binuclear Cu sites in CHA zeolite under the conditions for NH3–SCR reaction., The solvation effect of NH3 on Cu species drive the migration of Cu species across the cages of CHA zeolite and the dimerization of Cu–NH3 complex are responsible for catalyzing the NH3–SCR reaction. Taking into account the reactant-induced dynamic structural transformation of atomically dispersed species on solid carriers, the dynamic formation of binuclear bimetallic sites could probably be present in the supported bimetallic catalysts, which to the best of our knowledge have not been reported yet. This possibility deserves to be checked in future works related to the zeolite-supported bimetallic catalysts.

155In comparison with binuclear bimetallic sites, the high atomicity of bimetallic nanoclusters leads to distinct geometric and electronic structures, which will cause different patterns in terms of the interaction between the metal sites and the substrate molecules and thus different catalytic properties. In this section, we will discuss the catalytic applications of bimetallic nanoclusters for a variety of reactions and the structure–reactivity relationships.

156As one of the most important model reactions, CO oxidation has been frequently employed as the probe reaction to study the catalytic behavior and physicochemical properties of supported bimetallic clusters. For instance, the Ag9Pt2 and Ag9Pt3 clusters generated by the size-selected method (a physical method for generation of metal clusters with precise composition as model catalysts but with low yields) have been tested for CO oxidation reaction and the catalytic results indicate that Ag9Pt3 clusters are more active than the Ag9Pt2 clusters at 200–250 °C (see Figure 34a,b). The composition-dependent reactivity of AgPt clusters is ascribed to the favorable activation of O2 and CO at Ag and Pt atoms, respectively. The difference in reactivity could be associated with the chemical states of the Ag and Pt species in the two types of AgPt clusters, in which the more active Ag9Pt3 show a higher number of oxidized species than the Ag9Pt2 clusters (see Figure 34c–f).

157Using ligand-protected bimetallic clusters as a model catalyst, the introduction of Cu into Au25 clusters can slightly promote the activity for CO oxidation while the doping with Ag causes negative effects. Theoretical calculations indicate that the promotion of the second metal (Cu) to Au clusters could be associated with the formation of oxidized Cu species, which enhance the adsorption of CO and activation of O2. However, this observation is controversial to the promotion effect observed with bimetallic AuAg clusters supported on zeolite (pure-silica ITQ-2). This discrepancy could be related to the different reaction temperatures of the catalytic tests (<200 °C versus >200 °C). Besides, due to the higher reaction temperatures, the sintering of AuAg clusters on zeolite is also observed during the catalytic tests for CO oxidation reaction due to the removal of the thiolate ligands. This evolution behavior of bimetallic clusters has also been observed with bimetallic AuPd clusters supported on TiO2, which evolve into PdAu nanoparticles during the CO oxidation reaction.

158The advantages of bimetallic clusters over monometallic clusters for CO oxidation are shown in the CeO2-supported PtCu clusters. Generated by ALD, subnanometer Pt clusters are deposited on the (220) facets of Cu-doped CeO2 nanorods, forming interfacial PtCu sites on the surface of CeO2. The resultant bimetallic PtCu clusters are more active than the isolated Pt atoms and Pt clusters, owing to the unique interfacial structure for facilitating the adsorption and activation of CO and O2.

159In natural photosynthesis systems, the active site for water oxidation reaction is a bimetallic cluster made by Ca and Mn (CaMn4O5 cluster), which can activate H2O and produce O2 in Photosystem II. Although the detailed mechanism of the water oxidation reaction in the Photosystem II is still not fully elaborated, the extraction of electrons from H2O and the charge transfer process between the metal atoms within the CaMn4O5 cluster is the key. This lesson is also reflected in the catalytic oxidation reactions based on supported bimetallic nanoclusters.

160For instance, the catalytic activity of Cu clusters is greatly promoted by doping of Au or Ag for oxidation of cyclohexene to 2-cyclohexenyl hydroperoxide (see Figure 35). The charge transfer from Cu to metals to Au is claimed as the reason for the enhanced reactivity of CuAu bimetallic clusters, while the desorption of the product (2-cyclohexenyl hydroperoxide) is facilitated at a higher degree than that on Au and Cu, leading to the promoted activity of CuAg clusters. The profound impacts of the addition of a single atom to a metal cluster on the catalytic performance of selective oxidation reaction is shown with the superior performances of Au23Pd1 clusters over monometallic Au24 clusters for oxidation of benzyl alcohol to benzoic acid, because the Pd atom serves as the active site for hydride elimination from the α-carbon of benzyl alcohol.

161It should be noted that none of the above-mentioned reported catalysts based on bimetallic clusters surpasses the classic supported Au nanoparticles for CO oxidation reaction, while in the case of oxidation of organic molecules, metal clusters are advantageous over the conventional nanoparticulate catalysts. One possible explanation could be the difference in the rate-determining step for the two reactions. For CO oxidation, CO adsorption and O2 activation are usually considered the critical steps, while for oxidation of hydrocarbon compounds, the C–H activation is generally considered as the rate-determining step., In this sense, it will be interesting to further explore the catalytic performance of Au-based and Pd-based bimetallic clusters for a variety of oxidation reactions, such as selective oxidation of propylene to propane oxide, oxidation of aromatics to diphenyl compounds, oxidation of thiophenols to disulfide compounds, etc., because these reactions have been reported to be catalyzed by the monometallic clusters/nanoparticles.−

162In supported metal catalysts, the exposed facets of the support for hosting the metal particles will determine the structure of the metal–supporter interface, which is considered as major active sites for CO oxidation. The reactivity of bimetallic clusters is also dependent on the surface structure of the solid carrier. The PdCu clusters formed on the (100) facets of CeO2 are more active than those formed on (111) facets, which are proposed to be related to the smaller sizes of PdCu clusters generated on (100) facets and the higher number of oxygen vacancy sites. According to the study with Pd/CeO2 catalyst for CO oxidation, sintering of Pd atoms to nanoclusters or nanoparticles may occur due to the CO-induced migration of Pd atoms on the surface of CeO2. In this regard, more insights into the structural features of PdCu bimetallic nanoclusters could be obtained by performing operando characterizations.

163Indeed, the structural transformation of Pt atoms under water–gas shift reaction conditions has been observed with Pt supported on Co3O4. As shown in Figure 36a,b, isolated Pt atoms supported by Co3O4 remain to be atomically dispersed after exposure to water–gas shift reaction mixture up to 200 °C, but PtCo nanoclusters are formed at higher temperatures (≥280 °C) due to the sintering of Pt atoms and partial reduction of the support (see Figure 36c,d). More interestingly, the PtCo nanoclusters generated under reaction conditions are more active than the pristine Pt atoms (as indicated by lower apparent activation energy and higher TOFs), showing the advantages of bimetallic nanoclusters over isolated metal atoms.

164Hydrogenation reactions are one of the major applications of supported bimetallic catalysts in practical processes, which are mostly achieved with bimetallic nanoparticles in current industrial processes. By replacing nanoparticles with nanoclusters, the utilization efficiency of metals can improve significantly and the bimetallic nanoclusters may show catalytic reactivities different to nanoparticulate counterparts.

165According to theoretical calculations, a series of models of Co-based bimetallic nanoclusters with the configuration of M1-Con (M = transition metals) are constructed to study how the composition of bimetallic clusters can influence the activation of N2., The metal d orbital electrons and the intrinsic first ionization energy are used to describe the electronic features of the bimetallic clusters, which are further correlated with the adsorption energy of N2. The calculation results suggest that the PdCo4 configuration is a promising candidate for ammonia synthesis reaction because of its charge buffer capacity of Pd atom and the complementary role of Co in the activation of N≡N bond and the subsequent hydrogenation steps. In terms of the experimental works, the superior performances of RuCo bimetallic nanoclusters and nanoparticles show promising performances over other Ru-based and Co-based catalysts for ammonia synthesis., There are discrepancies between the theoretical modeling and experimental studies that could be caused by the different particle sizes of the bimetallic entities and the support.

166Taking mesoporous silica as the support, bimetallic nanoclusters can be generated and then stabilized on the solid carrier via the decomposition of the organometallic precursors, allowing the formation of bimetallic nanoclusters with narrower size distributions and more uniform compositions than the conventional coimpregnation method., For instance, Cu4Ru12C2 clusters are prepared by using (Ru6C(CO)16Cu2Cl)2(PPN)2 (PPN: bis(triphenylphosphanyl)iminium) as the precursor and mesoporous silica (MCM-41) as the porous support for hosting the clusters. The resultant bimetallic RuCu nanoclusters show good performance for selective hydrogenation of 1,5,9-cyclododecatriene to cyclododecene.

167Both the activity and selectivity of the bimetallic nanoclusters can be modulated by tuning the composition as summarized in Figure 37 and Table 1. For example, the Pd6Ru6 shows higher activity than the Ru6Sn clusters for hydrogenation of 1,5-cyclooctadiene but majorly gives the complete hydrogenation product, while the Ru6Sn gives exclusively the semihydrogenation product. Such effects have also been demonstrated in other reactions involving the hydrogenation of C=C and −COOCH3 groups. According to the lessons learned from the studies carried out with bimetallic nanoparticles, the adsorption properties of the intermediate products on nanoclusters are influenced by the electronic properties of the metal species, which are also directly associated with the composition. The catalytic consequences of hydrogenation reactions with bimetallic clusters can also be associated with the intrinsic hydrogenation capability of different elements. In the case of hydrogenation of benzoic acid, the Ru10Pt2 clusters show very high selectivity to cyclohexanecarboxylic acid, while the Cu4Ru12 give the formation of cyclohexene-1-carboxylic acid and 1,3-cyclohexadiene-1-carboxylic acid, which could be ascribed to the much higher hydrogenation capability of Pt than Cu, resulting in the formation of fully hydrogenated products on Ru10Pt2 clusters and the partially hydrogenated products on Cu4Ru12 clusters.

168In order to further clarify the detailed structural features of supported bimetallic clusters derived from organometallic complexes, the geometric and electronic features should be followed during the catalyst synthesis procedure and also tracked under reaction conditions. With the help of EXAFS, the coordination environment of the metal species can be elucidated, which suggests the rearrangement of the metal atoms in the precursor complexes due to the removal of ligands and formation of bonding between the metal atoms and the support., For instance, Pt2Ru4 clusters can be formed on Al2O3 after careful grafting and subsequent decomposition of the bimetallic PtRu complex on Al2O3, which show interesting catalytic behavior for hydrogenation of ethylene and hydrogenolysis of butane. Taking into account the limited sensitivity of EXAFS characterization, it will be necessary to combine the EXAFS characterization and the state-of-the-art STEM imaging technique to reveal the atomic structures of the bimetallic nanoclusters.

169Currently, the wet-chemistry approach for the synthesis of ligand-protected bimetallic clusters is the most practical and facile way to obtain bimetallic clusters with uniform and well-defined structures. The presence of ligands will block some of the metal sites on the surface and then hinder the catalytic reactions, although the ligands can modulate the product distributions of some hydrogenation reactions. By choosing the appropriate organic ligand, there could be some free surface sites in the ligand-protected bimetallic nanoclusters, which are available for hydrogenation reactions. By tuning the coordination environment of the bimetallic clusters, the accessibility of the metal sites can be modulated, resulting in shape-selectivity in hydrogenation reactions. As shown in Figure 38, the isolated Pd atom in PdAu9 clusters stabilized by N-heterocyclic carbene ligand can only be accessed by small alkene molecules and the ligand-protected PdAu9 clusters show interesting shape-selective behavior for hydrogenation of the C=C bonds in organic molecules. Well-defined bimetallic clusters derived from ligand-assisted synthesis are excellent model systems to identify the active sites in the bimetallic catalysts. However, as mentioned before in the section on synthesis of bimetallic nanoclusters, the stability of the ligand-protected metal clusters, especially their structural integrity during the removal of the ligands is the critical issues that limit their applications in practical applications.

170The dehydrogenation of light alkanes in industrial processes relies on the use of supported Pt particles as catalysts. By decreasing the size of Pt particles to subnanometer Pt clusters, the utilization efficiency and intrinsic activity can be greatly improved. Owing to the presence of a large number of undercoordinated Pt sites on the surface of Pt clusters, the addition of Sn can not only improve the stability of Pt clusters against sintering but also suppress the coke deposition on Pt sites due to the modulation of the bonding between Pt and alkane molecules., According to the theoretical calculations, the bonding of alkene (primary product of the alkane dehydrogenation reaction) on the Pt4 cluster shows both di-σ- and π-bonded configurations, while the Pt4Sn3 cluster binds ethylene only in the π configuration. Thus, the further dehydrogenation of alkene is inhibited on bimetallic PtSn clusters, resulting in less coke formation.

171By confining subnanometer PtSn clusters in zeolites, their stabilization against sintering is greatly improved. More importantly, the bimetallic PtSn clusters confined in the channels of pure silica MFI zeolite show remarkably higher activity than conventional PtSn nanoparticles (>2 nm) for the dehydrogenation of propane to propylene. The geometric and electronic features of PtSn clusters can be modulated by the pretreatments before the catalytic tests. As shown in Figure 39, with longer prereduction time and more intimate contact between Pt and Sn, is formed as visualized by the quasi in situ STEM measurements. Consequently, the initial selectivity to propylene is improved, and thanks to the suppressed breaking of C—C bonds the deactivation of the catalyst is also alleviated because of the less coke formation. The advantages of using zeolite as the support for stabilizing Pt-based bimetallic nanoclusters have been demonstrated with other Pt-based bimetallic nanoclusters and in other zeolites besides MFI-type structure.,,, The morphology of the zeolite support and the location of the promotor metals have significant influences on the particle size distributions and stability of the Pt species, inferring the importance to precisely control the structures of metal-zeolite catalysts, especially the location and the anchoring sites of the metal species in the microporous structure.

172Nevertheless, the chemical composition of the bimetallic nanoclusters derived from the combination of a noble metal (such as Pt, Pd, Rh, etc.) and a non-noble metal (such as Sn, Zn, In, etc.) could be further optimized for achieving high utilization efficiency of noble metals. For instance, by planting isolated Pt sites in the Zn/ZSM-5 support, it is claimed that PtZnn clusters with one Pt atom surrounded by several Zn atoms are generated in the silanol nests of ZSM-5 and the resultant PtZnn clusters can efficiently catalyze the dehydroaromatization of ethane due to the synergy between the bimetallic nanoclusters and the acid sites in ZSM-5. The concept should be extended to other combination of noble and non-noble metals for the preparation of subnanometer bimetallic clusters with low loadings of noble metals but high activities for conversion of light alkanes to olefins and aromatics.

173Introducing a small amount of oxidant into the reaction feed can shift the thermodynamic equilibrium by consuming the H2, which is called oxidative dehydrogenation of alkanes to alkenes. Bimetallic nanoclusters such as PtSn and PdCu prepared by size-selected method show promising activity for oxidative dehydrogenation of propane to propylene., It should be noted that these small clusters may suffer irreversible structural transformation under the reaction conditions. The presence of oxidant (O2) and reactant (alkane) may cause more severe sintering of the metal species in comparison to the reductive atmosphere used in the direct dehydrogenation of propane to propylene. Indeed, the sintering of small PtSn clusters normally occurs during the high-temperature reaction–regeneration cycles because the Pt species have a stronger tendency for migration on the solid carrier than in a reductive atmosphere as a consequence of the formation of volatile PtO2.,,

174To constrain the mobility of Pt species under the reaction conditions for oxidative dehydrogenation of alkanes, the interaction between Pt and the support should be promoted. The choice of CeO2 as the support for Pt-based alloy nanoparticles shows that not only the sintering of Pt species but also the coke deposition is greatly suppressed, which is attributed to the strong interaction between Pt and CeO2 and the oxygen releasing ability of the CeO2 support for the removal of the coke. Following this strategy, it will be interesting to decrease the size of the Pt-based alloy nanoparticles down to the subnanometer regime to form Pt-based bimetallic or even multimetallic nanoclusters, which could be even more active than the nanoparticulate counterparts.

175Beside light alkanes, long-chain alkanes such as C6–C8 alkanes can also be subjected to the dehydrogenation reaction, which will lead to the production of aromatics via the dehydroaromatization process. In the industrial naphtha reforming process, bimetallic Pt–Re catalyst supported on Al2O3 can effectively convert the hydrocarbon mixture of linear alkanes and cyclic alkanes into aromatics. Fundamental studies implied that the active sites are probably tiny PtRe nanoclusters with low atomicity (less than 10 atoms). Due to the very broad application of PtRe/Al2O3 catalysts in petrochemical industry, the catalyst’s structures and catalytic properties have been intensively studied decades ago, as reflected in the prior literature. However, limited by the resolution/sensitivity of the characterization tools, the atomic structures of the PtRe bimetallic species are not fully revealed yet. In our opinion, it will be of great interest to employ the newly developed electron microscopy and spectroscopy techniques to revisit the structures of the industrial PtRe/Al2O3 catalysts or even the optimized PtReSn/Al2O3, which may bring new insights/inspirations for designing advanced catalysts for hydrocarbon processing.,

176The industrial dehydroaromatization process for converting n-hexane into benzene relies on the use of Pt/KL catalyst, which comprises Pt nanoclusters in the 12MR channel of KL zeolite. The introduction of a cocatalyst metal (such as Fe, Co, etc.) into Pt/KL catalyst to form Pt-based bimetallic nanoclusters are reported to be effective to alleviate the catalyst deactivation and suppress the side reactions in dehydroaromatization of n-heptane and n-octane., Taking into account the works of zeolite-confined metal clusters for dehydrogenation of light alkanes, further exploration of other Pt-based bimetallic nanoclusters (such as PtSn, PtGe nanoclusters, etc.) and precise control of the composition and location of the nanoclusters in KL zeolite structure may lead to the generation of advanced catalysts for dehydroaromatization reactions.

177For numerous organic reactions, the combination of two different metals may show much higher performances than the monometallic catalyst due to the synergy of the two metals for activating the reactants and completing the catalytic cycle. Bimetallic nanoclusters based on organometallic complexes show interesting catalytic properties for various organic reactions. For instance, VIII group metals are widely used in carbonylation reactions. Specially, bimetallic nanoclusters based on VIII group metals (such as PdCo, RhCo clusters) are active for homologation of methanol for production of Me2O, AcOH, AcOMe, and MeCH(OMe)2., The activity and product distribution are greatly dependent on the composition of the bimetallic clusters, which could be caused by the impacts of the clusters’ structural features on the activation of methanol and the C–C coupling step.

178In addition to the bimetallic cluster catalysts derived from the organometallic approach, bimetallic nanoclusters prepared by wet-chemistry methods. For instance, AuPd nanoclusters stabilized by poly(N-vinylpyrrolidone) (PVP) are shown to be efficient catalysts for Ullmann coupling reaction while the monometallic clusters are inactive (see Figure 40). Mechanistic study based on theoretical calculations indicate that the Pd ensembles in the AuPd nanoclusters are responsible for the oxidative addition of Ar–Cl) and the role of Au is to improve the stability of the Pd ensembles. The synergy of bimetallic nanoclusters for organic reactions is observed with other reactions, such as hydrosilylation of internal alkynes. It should be noted that the AuPd nanoclusters stabilized by PVP basically show disordered structures with random distribution of Au and Pd. In order to gain further insights on the atomic structures of the active sites in bimetallic nanoclusters, the ligand-protected bimetallic nanoclusters with well-defined structures can be employed as the model catalysts for elucidating the optimum local structures of the Pd sites.

179It has been shown with a variety of organic reactions that metal clusters made by a few atoms serve as the real active sites, which are formed under reaction conditions from the metal precursors.,− The metal clusters could be formed due to the interaction between the precursors (in the form of molecular complex or metal nanoparticles) and the reactants/solvents/ligands. In this sense, beside the direct use of bimetallic nanoclusters as the catalysts for organic reactions, the bimetallic clusters may also be formed during the organic transformations catalyzed by two metals. For instance, Pd and Cu are usually employed together for Sonagashira reaction to construct C–C bonds, and the critical role of Pd clusters for C–C coupling reaction has been verified by numerous studies., Bimetallic PdCu clusters could be produced, which need to be carefully verified by electron microscopy and spectroscopy techniques.

180In the above examples, the bimetallic nanoclusters are either coordinated by ligands or polymers, which limits their stability and recyclability for practical applications. In principle, by immobilizing bimetallic nanoclusters on solid carrier should effectively improve the stability of the nanoclusters, although their catalytic performances may be affected by the support. For instance, nitrogen-rich mesoporous carbon is shown to be a superior support for hosting AuPd nanoclusters as the catalysts for Ullmann coupling reaction. In the future works, it will be interesting to explore the suitable support for bimetallic nanoclusters for the target organic reactions and develop appropriate methodologies for stabilizing the nanoclusters on the supports.

181In photocatalytic systems, bimetallic nanoclusters are usually employed as cocatalysts to facilitate the surface reactions of the photogenerated charge carriers with the substrate reactants. At the interface of metal–semiconductor interface, an electronic equilibrium will be established once the contact of metal and semiconductor is formed. Because the electronic structures of the semiconductor supports are normally fixed, the electronic structures of the metal–semiconductor junction will be largely determined by the bimetallic entities. As displayed in Figure 41, prior works on monometallic clusters show that the charge-transfer rates between ZnO nanoparticles and Au particles increase when increasing the size of Au particles from ca. 1.0 nm (Au25) to ca. 3.5 nm (Au807). Taking into account that the electronic structures of bimetallic nanoclusters depend on both the particle and chemical composition, there will be a larger variation range for the charge-transfer rates between the bimetallic nanoclusters and semiconductor support than that with the monometallic counterparts. Indeed, according to the theoretical calculations, the charge transfer between Pt13 cluster and TiO2 can be markedly improved by introducing Rh or Cu into the Pt13 clusters to form Pt7Rh6 and Pt7Cu6 clusters with bilayer structures as cocatalysts for photocatalytic H2 evolution reactions. The theoretical study infer that the first-layer metal atoms, which are located at the interface between bimetallic clusters and TiO2, are responsible for extracting the photogenerated elections from the TiO2 and then transfer the electrons to the second-layer metal atoms for H2 evolution reaction.

182In a typical experimental work, bimetallic Au24Pt and Au24Pd clusters are loaded on BaLa4Ti4O15 as the cocatalyst for photocatalytic water splitting. The catalytic tests show that the introduction of Pd into Au clusters does not improve the photocatalytic activity but Pt can promote the performance. Interestingly, structural characterizations indicate that Pt is selectively located at the interface between Au24Pt cluster and the BaLa4Ti4O15 semiconductor, while Pd is located on the external surface of Au24Pd cluster. The structural difference affects the transfer of photogenerated electrons to the metal clusters and thereby influences the efficiency for photocatalytic water splitting. Combining the above-mentioned theoretical and experimental works, we can conclude that modulating the composition of the bimetallic nanoclusters, and the spatial distribution of the metal elements is critical for tuning their photocatalytic behavior.

183Beyond working as cocatalyst, bimetallic clusters can serve as the photosensitizer for capturing photo energy and then transfer the photogenerated charge carrier (usually electrons) to the reactant for initiating the redox reaction. For example, by using Au-based bimetallic nanoclusters as the photocatalyst, intramolecular [2 + 2] cycloaddition of bisenone is achieved via the oxidative quenching cycle. By tuning the chemical composition of the bimetallic nanoclusters, the electronic properties are changed, as indicated by the photoluminescence spectra. Consequently, the activity and product distribution in photocatalytic cycloaddition reaction change remarkably with the catalyst’s composition, as summarized in Figure 42.

184To improve the charge-transfer efficiency, a promising strategy is to incorporate the bimetallic nanoclusters in a porous matrix, such as metal–organic frameworks, in order to minimize the distance between the photosensitizer and the active sites for photoredox transformations. By trapping the metal clusters in MOF structure, the stability of the ligand-protected bimetallic clusters can be greatly improved. Moreover, the active sites, which are responsible for the photoredox transformations, can be installed on the node or linker position of the MOF structure, resulting in the formation of composition materials for photocatalytic organic transformations.

185In the theoretical works on catalytic properties of bimetallic catalysts, models based on bimetallic clusters with a few atoms are usually constructed to represent the active sites in the catalysts made of bimetallic nanoclusters or nanoparticles. These theoretical studies provide rich knowledge of the influences of geometric and electronic structures on the interaction between the bimetallic nanoclusters and the reactants. For instance, taking the Co–N–C material with isolated Co sites in the N-doped carbon matrix as the support, models for different types of PtCo bimetallic sites are constructed to figure out the most active types of bimetallic sites for electrocatalytic ORR. As shown in Figure 43, the calculation results show that the activity of Pt clusters with low atomicity (with less than 8 atoms) increases dramatically after the introduction of Co (either doping in the Pt cluster or serving as the anchoring site of the Pt cluster. Guiding by the calculation results, the authors have prepared a series of bimetallic PtCo catalysts with different particle sizes by varying the Pt loading and the catalytic results indicate that Pt clusters with 2–7 Pt atoms and 2–3 Co atoms are the optimum configuration of PtCo clusters for electrocatalytic ORR. The excellent electrocatalytic activity of bimetallic PtCo clusters supported on Co–N–C support is also reflected in hydrogenation evolution reaction. The structure of the active sites is proposed to be partially oxidized Pt clusters are anchored on the isolated Co sites in the N-doped carbon matrix, due to the interaction between Pt clusters and CoN4 sites. The remarkably high activity of PtCo clusters supported on Co–N–C support has also been confirmed by the H2–O2 membrane electrode assemblies, in which the supported PtCo catalysts with low Pt loading but deliver high mass-transfer-limiting current densities and excellent power performances.

186Beside using Co–N–C as the support, bimetallic PtFe catalysts with superior activity for ORR have been prepared by loading a small amount of Pt species on the Fe–N–C support, which deliver excellent activity normalized to the mass of Pt. Detailed structural characterizations show that small bimetallic PtFe nanoparticles with sizes of ∼2 nm are stabilized on the Fe–N–C support, with greatly promoted durability. In principle, by optimizing the methods for materials synthesis, the size of the bimetallic PtFe nanoparticles may be decreased to subnanometer regime to form bimetallic PtFe nanoclusters, which may further promote the normalized Pt mass activity, or even the specific activity of Pt. Taking into account of the structural diversity of M–N-C materials (the types of M metals and the coordination environment of M metals in the N-doped carbon matrix), there are plenty room for improving the catalytic performances of PtM/M–N–C catalysts for fuel cell applications through optimizing the size of PtM clusters, the bonding between PtM clusters, and the M–N–C support.−

187In additional to carbon-supported noble metal catalysts, bimetallic nanoclusters based on non-noble metals are also promising electrocatalysts. For instance, bimetallic CoFe nanoclusters stabilized in N-doped carbon prepared by the pyrolysis of the mixture of Co and Fe molecular precursors. The kinetic study indicates that bimetallic CoFe nanoclusters show enhanced rate over the monometallic Co and Fe clusters for binding O2. Nevertheless, the bimetallic CoFe nanoclusters also show good performance for electrocatalytic OER, which is ascribed to the formation of bimetallic CoFe oxyhydroxide species under the OER condition, although the detailed structural features of the oxyhydroxide species are not clearly revealed.

188Isolated metal sites and binuclear bimetallic sites embedded in N-doped carbon matrix have been reported as active sites for a variety of electrocatalytic reactions. In principle, bimetallic nanoclusters are highly likely to form during the preparation of binuclear bimetallic sites supported on N-doped carbon materials, especially through the pyrolysis method. In this regard, it is important to carry out comparative study on the catalytic properties of binuclear bimetallic sites and bimetallic nanoclusters for a given electrocatalytic reaction. A practical approach to clarify this issue is to prepare the materials by tuning the metal loadings, which will generate a set of bimetallic catalysts with different size distributions, which are suitable for performing kinetic studies.

189As discussed in the section of hydrogenation reactions with bimetallic nanoclusters, ligand-protected metal clusters with well-defined structures are excellent model systems to establish the structure–reactivity of bimetallic nanoclusters in electrocatalytic reactions. As shown in a series of works, the addition of a single metal atom into a metal cluster to form bimetallic clusters will induce remarkable changes in electrocatalytic performances, as proved with Pt-doped Au clusters for electrocatalytic H2 evolution,, Pt-doped Cu clusters, and Pd-doped Au clusters for CO2 reduction. For metal clusters with low atomicity, the introduction of a single atom can cause marked changes in the electronic structure, resulting in modification of the electronic structures. For instance, the replacement of one Au with one Pt atom in the Au38(SCH2PhtBu)24 cluster elevates the relative energy of the highest occupied molecular orbital (HOMO) accompanied by one valence electron loss of Pt1Au37(SCH2PhtBu)24, in comparison with Au38(SCH2PhtBu)24 with 14 electrons. As a result, the electrocatalytic reduction of CO2 to CO is increased. Interestingly, if two Pt atoms are doped into the Au clusters, the resultant Pt2Au36(SCH2PhtBu)24 clusters show declined activity due to the narrowed HOMO–LUMO (lowest unoccupied molecular orbital) gap. The profound influences of the chemical compositions of bimetallic nanoclusters are found with other systems, such as the Au-based and Ag-based clusters for electrocatalytic N2 reduction and CO2 reduction.−

190In order to clarify the impacts of the chemical composition of bimetallic nanoclusters in catalytic reactions, the comparative study should be performed on a series of model catalysts derived from the same sample preparation method. Following this concept, bimetallic PtZr nanoclusters with different Pt/Zr ratios, and particle sizes are generated by controlling the pulses of arcplasma deposition and then tested for electrocatalytic HER. As shown in Figure 44, the Pt4Zr2 sample (prepared by four pluses of Pt and two pulse of Zr deposition) comprising PtZr nanoclusters of ∼0.8 nm delivers the best activity. The significant promotion effect of Zr to Pt for electrocatalytic HER can be explained by the modification of the electronic feature of Pt, which weakens the adsorption of Pt–H species. Furthermore, the arcplasma deposition method allows to prepare various types of bimetallic nanoclusters by simply change the metal precursors, enabling the generation of a library of bimetallic nanoclusters for rapid screening. As summarized in Figure 44, synergistic effect indexes of various alloy particles for the electrocatalytic HER are calculated to determine whether the combination of two metal elements in a bimetallic nanocluster is beneficial for HER. Some the results shown in the table are consistent with the literature works on bimetallic catalysts, such as the excellent activity achieved with the PtRu and WMo catalysts., However, results controversial to the prior works are also found with the bimetallic nanocluster library, which could be caused by the different structures of the bimetallic nanoclusters in comparison to the catalysts in reported works.

191From the structural point of view, bimetallic nanoclusters can be considered as transition materials between binuclear bimetallic sites and bimetallic nanoparticles. As reflected in the number of related publications, in comparison to the intensively studied field of bimetallic nanoparticles and the emerging field of binuclear bimetallic sites, the research efforts devoted to bimetallic nanoclusters are not as many as the other two fields in recent years. Taking into account the unique catalytic properties of monometallic clusters and their significant advantages over isolated metal atoms and nanoparticles in some important catalytic reactions, bimetallic nanoclusters deserve more attention from the catalysis community. On one hand, to address the distinct reactivity of bimetallic clusters, the future research efforts should be focused on the catalytic properties of subnanometer bimetallic clusters with less than 20 atoms because these entities show very different geometric and electronic structural features compared to binuclear sites and bimetallic nanoparticles. On the other hand, special attention should be paid to the systems in which bimetallic clusters may be formed under reaction conditions due to the structural evolution of binuclear bimetallic sites or bimetallic nanoparticles. In these cases, although the bimetallic nanoclusters are not present in the starting catalysts, they may serve as the real working species, which could be probably overlooked. To resolve this issue, operando studies are encouraged to be performed with the catalysts comprising binuclear bimetallic sites or bimetallic nanoparticles to identify to working active sites in those catalytic materials.

192Owing to the industrial interests of supported bimetallic catalysts, an enormous amount of work on the synthesis, characterizations and attempts to establish the structure–reactivity relationships of bimetallic nanoparticles has been carried out, and it has been reviewed and discussed in a number of review articles.,− In this section, we will highlight the progress achieved during the past decade on the catalytic applications of bimetallic nanoparticles. For each type of representative bimetallic nanostructures, we will start with discussions on their geometric and electronic structures and then the impacts of the unique structural features on their catalytic properties will be shown.

193The larger particle sizes of bimetallic nanostructures than binuclear bimetallic sites and bimetallic nanoclusters cause the higher structural diversity of bimetallic nanostructures than the subnanometer counterparts. As shown in Figure 45, depending on the spatial distribution of the two metal elements, the bimetallic nanoparticles can be generally categorized into the following types: single-atom alloy, bimetallic nanoparticle with local ensembles, bimetallic nanoparticle with random spatial distribution, and core–shell nanoparticle. Moreover, by decreasing the size of the nanoparticles to form low-dimensional bimetallic structures, a large number of undercoordinated surface sites will be exposed in the two-dimensional bimetallic structures (ultrathin nanosheets) and one-dimensional structures (ultrathin nanowires). Besides the combination of two metallic components to form bimetallic particles, the combination of a monometallic particle and a metal oxide component leads to the formation of metal-oxide interfacial structures, which may show distinct properties than the bimetallic particles due to the electronic interaction between the metallic particles and the oxide patches. The structural complexity will be further elevated when the number of involved metal elements is increased from 2 (bimetallic particles) to above 5 (high-entropy alloy particles). Because of the structural diversity, the structure–reactivity relationships of bimetallic nanostructures could be largely different from case to case. For each case, the structural characteristics need to be resolved by comprehensive characterization studies and catalytic studies based on a series of comparative catalysts are usually necessary for elucidating the active sites in the bimetallic nanostructures.

194In conventional bimetallic nanoparticles, there are regions with local segregation in the form of ensembles with a few atoms. By diluting the percentage of one metal element to a very low level (≤3 at. %), the atoms of one metal element are completely separated by the other metals, resulting in the isolation of metal atoms in a metallic matrix. As shown in Figure 46, the isolated Cu atoms in Ag matrix show a narrow d-band projected density of state (pDOS), while the pure Cu particle gives a broad one. Interestingly, the electronic features of the isolated Cu atom in Ag matrix is quite similar to the that of the free Cu atom, which is ascribed to the mismatch of the Ag 4d and Cu 3d states. The sharpness and degeneracy of the Cu 3d states in the single-atom AgCu alloy cause the different bonding with O atom in comparison with pure Cu particle because the isolated Cu atom in Ag matrix can interact with O atom through π bond, while the Cu atom in Cu particle has weaker interaction with O atom.

195In other single-atom alloy systems, the electronic structure of the isolated metal atom is greatly altered by the matrix due to the electron transfer between the two metals. For instance, the Pd atoms in the Cu matrix are more electron-rich than the Pd atoms in the monometallic Pd particle, and the isolated Pd atoms show lower adsorption energies of H species, which is beneficial for electrocatalytic H2 evolution reaction. In another example, the electronic properties of the Pt matrix are modified by isolated Ni atoms, resulting in the decrease of the hydrogen-binding energy to the nearly optimal HER activity region. Moreover, the influence of isolated Ni atoms to Pt matrix is also reflected in the adsorption energy of CO, which facilitate the conversion of CO to CO2 in electrocatalytic oxidation of alcohol.

196Bimetallic nanoparticles based on single-atom alloys have shown great potential in selective hydrogenation reactions in comparison with traditional alloy nanoparticles with local segregations. By diluting an active metal in a metallic matrix with low intrinsic activity, the selectivity for hydrogenation reactions can be greatly improved, because the side reactions caused by the segregated active metal atoms can be avoided due to the formation of single-atom alloys., For example, by introducing a very small amount of Pt or Pd in to Cu or Au nanoparticles, highly active and selective single-atom alloy catalysts have been prepared for a variety of selective hydrogenation reactions, such as hydrogenation of alkynes to alkenes, hydrogenation of 1,3-butadiene to butenes, hydrogenation of unsaturated aldehydes to unsaturated alcohols, and selective hydrogenolysis of glycerol to 1,2-propanediol.,− The formation of a single-atom alloy not only suppress the undesired side products from overhydrogenation reactions but also greatly improve the utilization efficiency of noble metals.

197The advantages of single-atom alloy catalysts have also been demonstrated in other reactions. For instance, for dehydrogenation of propane to propylene and dehydrogenation of ethanol to aldehyde,, the isolated metal sites can selectively activate the C–H bonds while the deep dehydrogenation can be suppressed due to the less favorable binding of the primary products on isolated metal sites than that on sites with local segregation. As shown in Figure 47, isolated Pd atoms dispersed in Au matrix are highly active, and selective catalysts for dehydrogenation of ethanol to acetaldehyde while the presence of Pd ensembles will cause the formation a variety of products. Reversible structural transformation may occur under reaction conditions because the exposure of single-atom PdAu alloys to CO will cause the migration of Pd atoms to form Pd ensembles, which is not favorable for selective dehydrogenation reaction. Besides, isolated Ru atoms embedded in the matrix of Cu nanoparticles exhibit excellent activity for photocatalytic dry reforming of CH4 by serving the sites for transferring the hot electrons generated in Cu nanoparticles to reactants and then facilitating the activation of C–H bonds in CH4 and C–O bonds in CO2. Once local segregation of Ru atoms is present in the bimetallic RuCu nanoparticles, the undesired deep dehydrogenation of CH4 will occur, resulting lower selectivity to syngas (CO and H2) and also faster catalyst deactivation in comparison to the single-atom RuCu alloy particles.

198From the fundamental point of view, the single-atom alloy catalysts allow description of the electronic structures of the metal centers with well-defined models. This advantage permits the establishment of the structure–reactivity correlation and the fast screening of potential catalytic materials for the target reactions, especially with the help of calculation methods based on machine learning. From the practical point of view, by diluting noble metals into a non-noble metal matrix for the generation of single-atom alloy, the normalized activity of the noble metal species can be greatly improved. However, such catalysts may still show low efficiency in the practical catalytic processes because the low metal loadings and insufficient yield of desired products on solid catalysts with limited mass or volume is the critical criteria. To overcome this limitation, more efficient synthesis methods are needed for the generation of dense single-atom alloy sites on the solid carrier.

199The ensemble effects in bimetallic catalysts proposed by Wolfgang Sachtler indicate the necessity of the formation of ensembles of metal atoms for performing the specific reaction. In other words, for such a type of a reaction, local segregation is required to form the ensemble of a few metal atoms for activation of the substrate molecules.

200The size of the metal ensembles can affect the reaction pathways for a given catalytic reaction. For instance, the formation of Au–Pt interface can greatly promote the activity for electrocatalytic oxidation HCOOH, as indicated by the two-order magnitude enhancement in activity observed with Au@Pt and Pt@Au structure in comparison to the bare Au and Pt surface. From a mechanistic point of view, the oxidation of HCOOH may involves two pathways: the dehydrogenation pathway for production of formate as the intermediate and the dehydration pathway for production of CO as intermediate. Theoretical calculations indicate that the dehydrogenation pathway can occur with isolated Pt atom in the Au matrix, while the dehydration pathway is favorable on Pt ensemble with at least three Pt atoms. These insights are consistent with the experimental works, in which the Pt/Au ratios in the bimetallic nanoparticles have significant impacts on the catalytic properties for electrocatalytic oxidation of formic acid.

201The importance of formation of Pd ensembles with multiple Pd atoms has also been suggested by theoretical study with PdCu alloy nanoparticles. Pd ensembles with Pd site on the external surface and at the subsurface region are the most active configuration of activation of H2, which is ascribed to the gradual shift of d-band centers to the Fermi level and the electron transfer between Pd ensemble and Cu matrix. As shown in Figure 48, the superior activity of metal ensembles over isolated metal sites for H2 activation is also demonstrated with bimetallic PdAu nanoparticles., With the ensembles of a few atoms, the reaction rates for dissociation of H2 will be greatly promoted. The capability for activation of small molecules such as H2 and O2 will influence their catalytic properties in selective oxidation and hydrogenation reactions. In this regard, the structural features of metal ensembles in bimetallic nanoparticles can affect the chemoselectivity. For example, bimetallic PdAg nanoparticles with diluted Pd in Ag matrix have been found as efficient catalysts for selective hydrogenation of acetylene to ethylene. In the experimental works, although Pd is highly diluted in Ag matrix, the optimum configuration of Pd site is still unclear. Theoretical calculations indicate that the isolated Pd sites and Pd dimers show low selectivity to ethylene due to the overhydrogenation reaction while the Pd trimers exhibit high selectivity to ethylene.

202The ensemble effects are not always beneficial for promoting the catalytic performances, as indicated by the catalytic behaviors of bimetallic AuPd nanoparticles, which are excellent catalyst for some reactions (such as oxidation of benzyl alcohol) but exhibit lower activities than the monometallic ones in some other reactions (such as CO oxidation and water–gas shift reaction). By rational design of a supported bimetallic Au–Pd/C and Au@Pd/C catalyst with separated Au and Pd domains, the catalytic performances for oxidation of 5-hydroxymethylfurfural in aqueous alkaline solution are enhanced because the Au and Pd sites can perform the dehydrogenation and O2 reduction steps, respectively. However, the optimum size of the metal ensembles for selective oxidation of alcohols require further investigations and the reasons why the bimetallic nanoparticles deliver lower performances than the monometallic particles for CO oxidation and water–gas shift reaction are unknown.

203In the above examples, the ensembles are already formed in the starting catalyst. Taking into account the structural transformation of bimetallic nanoparticles under reaction conditions, ensembles could be generated in bimetallic nanoparticles as the consequences of the metal-reactant/product interaction. These phenomena have been observed with numerous systems, such as the formation of Pd domains in AuPd nanoparticles driven by the CO adsorption, formation of Ni-rich surface domains in CuNi nanoparticles under CO2 hydrogenation conditions, reversible formation of RhPd and RhPt nanoparticles under NO+CO reaction conditions, Pd-rich domains in PtPd bimetallic nanoparticles for methane combustion, Pd-rich domains in PdW bimetallic nanoparticles for electrocatalytic oxygen reduction, and Sn-rich domains in PtSn bimetallic nanoparticles for electrocatalytic methanol oxidation. Under certain conditions, the migration of the metal species may occur between the metal nanoparticles. For instance, the mixture of monometallic Pt and Au nanoparticles will transform into AuPt alloy nanoparticles during continuous potential cycling in electrocatalytic oxidation of formic acid, which deliver much higher activity than the monometallic nanoparticles.

204Understanding the size and coordination environment of the ensembles in the bimetallic nanoparticles are very helpful to elaborate the structure–reactivity relationships. A key scientific issue to clarify the ensemble effects is the structural characterizations of the ensemble species in bimetallic nanoparticles (in terms of the size and geometric features), which causes difficulties to establish the structure–reactivity relationships. The progress in this direction relies on the development of advanced characterization techniques (such as high-resolution TEM imaging and analysis technique) to quantitatively resolve the number and quality of various sites in a bimetallic nanoparticle. Nevertheless, fundamental studies based on well-defined surface or crystals can also bring very helpful insights, which can be used to correlated with the results obtained with practical catalysts based on bimetallic nanoparticles.

205Liquid bimetallic nanoparticles are emerging catalytic materials because they can combine some advantages of conventional heterogeneous catalysts (such as easy separation and high stability) and homogeneous catalysts (electronic-steric control due to the formation of isolated metal sites in a liquid metal matrix). For instance, when Ga/Pd ratio is higher than 10 and the temperature is above 400 °C, liquid bimetallic alloy will be formed. Following the phase diagram, a series of Ga-rich PdGa bimetallic catalysts have been prepared by galvanic replacement reaction between liquid Ga particles and (NH4)2PdCl4, giving the formation of spherical PdGa particles (Ga/Pd atomic ratios ≥10) on porous glass. The reactivity of the Ga-rich PdGa particles has been tested for dehydrogenation of butane to butenes, in which the reactivity of PdGa samples with diluted Pd species show remarkably higher activity and stability than the monometallic Pd/SiO2 catalyst and PdGa catalyst in the form of solid alloy. The liquid PdGa alloy catalysts even show superior performance in comparison with the classic Cr/Al2O3 and Pt/Al2O3 for alkane dehydrogenation reaction. Detailed structural characterizations show that the isolated noble metal atoms (such as Pd, Pt, and Rh) diluted in liquid Ga matrix will be highly mobile. As suggested by in situ CO-IR spectroscopy and structural characterizations, the noble metal atoms can move between surface and bulk in a highly dynamic manner., The alkane molecule can be converted to alkene through dehydrogenation reaction on the surface noble metal atoms, and the noble metal atoms can subsequently migrate to bulk of Ga matrix, leading to the desorption of the alkene product. Interestingly, as shown in Figure 49, the isolated Pt atoms in the Ga matrix are resistant to CO-poisoning during propane dehydrogenation reaction while the conventional Pt/Al2O3 catalyst will be partially deactivated by CO due to the strong adsorption of CO on Pt nanoparticles. This feature infers that supported liquid alloy catalysts may have potential applications for selective hydrogenations with resistance to poison impurities.

206In the above-mentioned works related to supported liquid nanoparticles, the synthesis and catalytic applications usually involve the treatments at temperatures above 300 °C. The formation of liquid bimetallic nanoparticles is feasible at low temperature by diluting a very low amount of Pt in the liquid Ga nanoparticles. Pt species exist as isolated sites in the liquid Ga particles and show high mobility in the liquid Ga particles, which are highly active for electrocatalytic oxidation of methanol at 70 °C. One explanation for unique reactivity of the highly mobile Pt atoms in liquid Ga nanoparticles is the steady renewal of the catalytic metal centers at the interface in comparison to the fixed metal–support interface in conventional heterogeneous metal catalysts. The highly dynamic structures of the soluble metal species in the liquid metal enables the high utilization efficiency of the noble metals and the stability against deactivation under reaction conditions.

207It can be anticipated that, in order to control the migration behavior of the Pt atoms in liquid metal particles, the size and morphology of the liquid Ga nanoparticles need to be controlled. Beside the use of Ga as the “solvent” for another metal element, other metals with low melting point such as In, Sn, and GaIn alloy are also potential candidate materials for preparation of liquid bimetallic particles. In the above-mentioned works, the noble metal atoms (Pt and Pd) are proposed as isolated sites in the liquid matrix for a series of reaction. However, the presence of metal ensembles formed by a few noble metal atoms is not completely excluded according to the characterization results, and the role of the metal ensembles in the catalytic reactions is unclear.

208Although the conventional bimetallic particles usually do not show such highly dynamic and mobile features as the liquid alloy particles at ambient conditions, the mobility of the bimetallic nanoparticles could be underestimated under reaction conditions (e.g., high temperature, high pressure, and in the presence of reactants), especially when the sizes of the bimetallic nanoparticles are below 2 nm. The knowledge and insights derived from the studies with supported liquid bimetallic particles could be translated to understand the dynamic structures of bimetallic nanoparticles under reaction conditions. For instance, combination of spectroscopic characterization and multiscale theoretical modeling indicate that the isolated of Pt sites and the mobility of surface Pt and Ga atoms are key features that account for the high activity of supported PtGa nanoparticles (∼1 nm) for propane dehydrogenation reaction.

209In numerous studies based on model surface systems, the electronic structures of a monolayer metallic surface will be greatly influenced by the subsurface metal and therefore affect the catalytic properties of the surface monolayers., For instance, the d-band centers can be modulated by varying the subsurface 3d metals from Ti to Ni, leading to the variation of the binding energy of probe molecules such as ethylene, water and CO.

210By translating the knowledge accumulated in surface systems to practical systems based on nanoparticulate systems, numerous catalysts based on core–shell bimetallic nanoparticles have been designed and prepared. For example, the reactivity for electrocatalytic H2 evolution reaction is improved by decreasing the thickness of the Pt shell in Pd@Pt core–shell nanoparticles, which is ascribed to the modification of the electronic structure of the Pt overlayers caused by the charge transfer from Pd core to Pt overlayers. In the case of Pd@Pt core–shell nanoparticles formed by deposition Pt overlayers on Pd icosahedra, the structure with three Pt overlayers is found to be the most active configuration for electrocatalytic oxygen reduction reaction. The advantages of core–shell bimetallic nanoparticles over the alloy nanoparticles have also been demonstrated with Au@Co structure for the hydrolytic dehydrogenation of ammonia borane and Pd@Pt core–shell nanoparticles encapsulated in MOF matrix for CO oxidation.,

211Not only activity, but also selectivity can also be altered by the thickness of overlayers. For instance, the thickness of Pd overlayers on Au nanorods can greatly influence the selectivity for selective hydrogenation of 1,3-butadiene to butenes. With a thicker overlayer of Pd on Au nanorods, the selectivity to the butenes would be lower. Such an effect has also been observed with Pt overlayers on Au nanoparticles for selective hydrogenation of nitroaromatics. These works imply that the optimum thickness of the metal layers in core–shell nanoparticles could be dependent on the target reaction, which could be further associated with the thickness-dependent electronic features of the metal overlayers. Usually, the optimum thickness of the shell composition would be ≤3 atomic layers, because thicker overlayers will cause physicochemical properties similar to the bulk metal.

212In the above examples, the metallic core is fully covered by another metallic shell. However, if the loading of the shell metal is low, core–shell type structures with submonolayer coverage will be formed. Theoretical calculations suggest that the edge and terrace sites of the Ni patches deposited on Pt(111) surface can efficiently catalyze the ammonia decomposition reaction. This study infers the importance to control the morphology and domain size of the submonolayer structure on a metal substrate. It has been shown with Pd@Au nanoparticles that Pd dimers supported on Au particles will greatly improve the activity for electrocatalytic reduction of CO2 to CO, because Pd dimers can efficiently activate CO2 and mitigate the deactivation problem caused by strong adsorption of CO. By tip-enhanced Raman spectroscopy, the activated H species can be spilled over to the Au substrate and the distance of the H spillover can be as far as ∼20 nm. Actually, this concept has already been practiced in an early work, in which the catalytic performance of Au/TiO2 for chemoselective hydrogenation of nitroaromatics can be greatly enhanced by depositing a very small amount of Pt species (e.g., 100 ppm Pt in the total weight of the Au/TiO2 catalyst) on Au particles.

213The bimetallic nanoparticles are supported on solid carriers, but the reversible configuration, in which the metal nanoparticles are modified with another metal oxide species, has also been observed in numerous systems. Such structures are usually formed between noble metals and oxyphilic metal elements. For instance, the deposition of a monolayer of oxide/hydroxide species (such as FeOx and Ni(OH)2) on a Pt surface can generate highly active interfacial sites for CO oxidation reaction., The metallic surface and metal-oxide interface are responsible for activation of CO and O2, respectively. Controlling the chemical states of the surface oxide domains/layers is critical for maintaining high activity because overoxidation or over-reduction may cause the loss of metal-oxide interfacial sites.

214The dehydrogenation of cyclohexanone is greatly enhanced by modifying the surface of Au nanoparticles with Pd(II) species, because the Pd(II) sites can activate the C–H bonds and the metallic Au surface can facilitate the β-H elimination step (see Figure 50).,, Interestingly, the synergy of Au and Pd cannot be achieved with AuPd bimetallic nanoparticles, because the rate-limiting step, activation of C–H bonds, is not favorable on bimetallic AuPd surface.

215One difference between core–shell bimetallic structures and metal-oxide interfacial sites is the chemical states of the oxidized species which can serve as the Lewis acid sites, leading to the generation of bifunctional catalysts. This synergy is reflected with the bimetallic Ru-ReOx/SiO2 catalyst developed for production of secondary alcohols from 1,2-alkanediol by C–O hydrogenolysis. The ReOx species located on the metallic Ru particles can activate the −OH group in 1,2-alkanediol and then facilitate the breaking of C–O bonds with the help of activated H species on Ru particles. Metallic Re particles are inactivate for this reaction because of the loss of Lewis acidity. The synergy between metallic and oxide component for a variety of reactions dealing with derivatives from biomass because those catalytic transformations usually involve the hydrogenolysis or hydrodeoxygenation reactions.−

216As shown in Figure 51, the structural features of the metal-oxide interfacial structures are greatly influenced by various parameters and for different types of metal-oxide interfaces, the catalytic properties could be largely different. Some fundamental issues related to the metal-oxide bimetallic sites are still not clear in those reported systems, such as the optimal size of each component and the atomic structures of the metal-oxide interface. On one hand, researchers may employ advanced characterization techniques to address the detailed structures of the metal-oxide bimetallic sites. On the other hand, preparation of metal-oxide model catalysts with well-defined structural features will be very helpful to elucidate the nature of the active sites. The latter approach includes the construction of well-defined metal-oxide interfacial structures based on surface organometallic chemistry and construction of model system in surface science studies.,

217The strong metal–support interaction (SMSI) has profound implications in the catalytic properties of supported metal catalysts. In the classic model of SMSI, the metal particles will be partially or fully covered by the overlayers from the support under high-temperature treatments. The detailed characterizations have revealed that, beside the TiOx overlayers, PtTi alloy particles are formed in Pt/TiO2 catalyst due to the SMSI, according to the in situ TEM, EXAFS, EELS, and CO-IR results.− The morphological characterization results show that, the PtTi alloy structures are only formed on the surface of the Pt nanoparticles and probably exhibit the thickness of less than three atomic layers. The formation of TiOx overlayers is proposed as a key consequence caused by the SMSI and then has profound influences on the behavior of Pt/TiO2 for chemoselective hydrogenation reactions. In this sense, it will be important to elucidate the impacts of TiOx overlayers and the Pt–Ti alloy on the catalytic properties.

218Although the SMSI could be reversible in Pt/TiO2 catalysts, the atomic structure of Pt–TiO2 interface and the morphology of Pt nanoparticles will also change during the consecutive reduction–oxidation treatments, as shown by the in situ TEM images. As shown in Figure 52, the ultrathin PtTi alloy structures may also evolve with the treatments, which could explain the variation of the catalytic performance of Pt/TiO2 catalyst observed in some catalytic studies as well as other TiO2-supported metal catalysts., It should be noted that, the TiO2–x overlayers shown in the in situ TEM study are not fully removed by the exposure to O2 atmosphere, which is different from that observed in ex situ studies and could be related to the pressure gap between in situ TEM study and ex situ studies.,

219The formation of alloy nanoparticles induced by SMSI has also been observed in Pt/CeO2 catalyst, in which the Pt–Ce bonding is formed after exposure to CH4 at 975 °C. The formation of Pt–Ce bonding could be understood as the reduction of CeO2 support, and it requires higher temperature than the formation of Pt–Ti bonding in Pt/TiO2 catalyst, which should be related to the reducibility of the oxide support. It is shown that, SMSI can occur after reduction by H2 at 975 °C, while the reduction at 550 °C cannot trigger the SMSI in Pt/CeO2. Formation of Pt-based alloy nanoparticles is facilitated with Pt supported on two-dimensional transition metal carbides (MXenes), as demonstrated with PtNb alloy nanoparticles formed at 350 °C and PtTi alloy nanoparticles formed at 550 °C after reduction treatment by H2., The unique electronic properties of the MXene supports could be the reason accounting for the easy formation of these alloy nanoparticles. The Pt-based alloy nanoparticles supported on MXenes show promising performances for several reactions, such as water–gas shift, electrocatalytic H2 evolution, and alkane dehydrogenation reactions.

220Another typical case involving the formation of bimetallic sites is the Cu–Zn system for methanol synthesis from CO/CO2 hydrogenation. There are still many debates on the active sites as reflected in the literature, and the CuZn alloy phase is one of the proposed active sites for methanol synthesis., By tuning the grazing incidence angle of in situ X-ray photoelectron spectroscopy measurements, it is possible to detect the electronic features of the surface Cu and Zn species. The results show that the presence of CO will contribute the reduction and migration of Zn species to form Zn–Cu alloy, and the surface structure is greatly dependent on the atmosphere.

221The particle size effects of metal catalysts have been intensively studied in heterogeneous catalysis and the particle sizes of the metal species indeed have great impacts on the catalytic consequences in enormous structure-sensitive reactions. However, the majority of the studies on particle size effects are based on spherical or nearly spherical particles. In the case of low-dimensional bimetallic structures in the form of ultrathin nanosheets (2D) or nanowires (1D), the unique geometric structural features can bring novel catalytic properties different to three-dimensional nanoparticles and nanoclusters.

222One-dimensional bimetallic nanowires with average diameters below 1 nm have been prepared as efficient catalysts for electrocatalytic oxygen reduction and hydrogenation oxidation reactions., The abundant (111) facets in the ultrathin Pt-based bimetallic nanowires are proposed to be responsible for the high activities. This concept has also been shown with the two-dimensional PdMo nanosheets, whose ultrathin structures also provide abundant exposed (111) facets. The PdMo nanosheets with thickness below 1 nm show remarkably high activity and excellent stability for electrocatalytic oxygen reduction and oxygen evolution reactions. Electron transfer from Mo atoms to Pd atoms is suggested by the DFT calculations, which shifts the center of the band toward negative energy relative to the monometallic Pd nanosheet. The downshift of the d-band center relieves the overbinding of oxygen on the surface, resulting in an optimal oxygen adsorption energy. The superior electrocatalytic performance of bimetallic nanosheets has also been observed with PdFe nanosheets for reduction of N2, which is attributed to the efficient N2 activation via N2-to-Fe σ-donation on the isolated Fe sites in the Pd matrix.

223The unique catalytic properties of isolated metal atoms and subnanometer metal clusters can be ascribed to their molecule-like electronic structures. Novel catalytic properties distinct to conventional nanoparticulate materials usually appear when the sizes of metal entities are decreased to ≤0.7 nm (corresponding to the size of a M13 cluster or species with lower atomicity). In this regard, if the diameter/thickness of ultrathin one-dimensional/two-dimensional bimetallic structures can be further decreased to below 0.5 nm, which corresponds to 1–2 atomic layers, a more profound difference between these ultrathin materials and conventional small nanoparticles could be observed, and these low-dimensional bimetallic structures may show catalytic applications beyond the scope of bimetallic nanoparticles.

224Bifunctional catalysts are widely used catalysts in numerous processes, in which the two types of active sites (metal sites for redox reactions and acid/base sites for acid–base reactions) work in a synergistic way for catalytic reactions involving multiple steps., The combination of bimetallic nanoparticles and acid/basic sites can enable the reactions that a solo type of functional component is not sufficient to achieve or significantly promote its performance. The dispersion of Pt in beta and USY zeolite can be improved by addition of Pd to form bimetallic PtPd nanoparticles. Consequently, a better balance between the metallic sites and the acid sites can be achieved for isomerization and hydrocracking of heptane than the monometallic Pt-zeolite counterparts. The advantage of bifunctional catalysts has also been demonstrated in cascade process involving the formation of metastable intermediates. As presented in Figure 53a,b, by loading bimetallic AuPd nanoparticles on TS-1 zeolite, a cascade transformation from ketone to oxime can be achieved with the AuPd/TS-1 bifunctional catalyst, in which the AuPd nanoparticles are responsible for synthesis of H2O2 from H2 and O2 while the TS-1 zeolite provides acid sites for cyclohexanone ammoximation.

225Bifunctional catalysts based on bimetallic particles and acid/basic sites are also excellent catalysts for hydrogen-transfer reactions because the bimetallic nanoparticles and acid/basic sites can work in a synergistic manner for the dehydrogenation, intermediate formation, and hydrogenation elementary steps. The participation of both the metallic component and the support in hydrogen-transfer catalysis are demonstrated with monometallic catalysts supported on oxide., The use of bimetallic nanoparticles instead of monometallic nanoparticles can further promote the activities. For instance, the N-alkylation of primary amides is achieved by the cooperative catalysis based on AuPd bimetallic nanoparticles and homogeneous Lewis acid (Ba(OTf)2). The bimetallic AuPd nanoparticles supported on TiO2 and Al2O3 give excellent activity for amination of alcohols and ketones.,

226Another way to form bifunctional catalytic system is to introduce homogeneous complex to work together with the bimetallic nanoparticles by serving as the acid/basic functional component. For example, as shown in Figure 53c,d, the combination of Sc(OTf)3 and RhPt nanoparticles are efficient bifunctional catalyst for hydrogenation of arenes. The role of Sc3+ is proposed to enhance the adsorption of arene on RhPt nanoparticles and facilitate the transfer of activated H species from RhPt nanoparticles to the aromatic ring. The synergy between bimetallic nanoparticles and Lewis acid sites has also been shown with AuPd nanoparticles and Mg2+ for N-alkylation of primary amides with primary alcohol. The AuPd nanoparticles account for the hydrogenation transfer (dehydrogenation of alcohol to aldehyde and subsequent hydrogenation of imine intermediate), while the Lewis acid sites account for the condensation between primary amide and aldehyde.

227The spatial distribution of the two functional components in bifunctional catalysts has profound influence on the catalytic performances, as shown with metal-zeolite catalysts for hydrocarbon processing. In principle, there should also be an optimized spatial distribution of the two types of active sites in the integrated catalyst based on bimetallic nanoparticles and acid/basic sites. It can also be expected that the controllable preparation of such catalytic materials will be more challenging than the monometallic systems.

228Carrying out enantioselective reactions with solid catalysts is one of most desirable goals of the catalysis community. Although great efforts have been dedicated to this topic and some success has been made with ligand-modified supported metal catalysts for enantioselective hydrogenation reactions, the reaction scope and performances are still far behind the well-established homogeneous systems. One plausible reason that limits the performance of ligand-modified metal nanoparticles could be the flexible conformation of the chiral template molecules absorbed on the surface, resulting in a less-ordered microenvironment. One strategy to improve the order degree of the chiral microenvironment, bimetallic nanoparticles are embedded in chiral polymers and have shown good performances for a variety of enantioselective C–C bond-forming reactions, such as asymmetric 1,4-addition of arylboronic acids with α,β-unsaturated carbonyl compounds and cascade oxidation of alcohol to conjugated aldehyde and Micheal addition reaction. The collaborative function of bimetallic nanoparticles and chiral polymer requires the proper control of the spatial distribution of the two functional components. As shown in Figure 54, the AuPd nanoparticles need to be located on the external surface of the polymer beads while the chiral organocatalysts should be located in the inner layer, as shown in panel (ii) in Figure 54a. If the spatial distribution is reversed (shown in panel (iii) in Figure 54a), the AuPd nanoparticles will be poisoned by the amine groups in the organocatalysts, resulting catalyst deactivation.

229Using chiral molecule as template, bimetallic nanostructures with chiral features at nanometer scale can be generated via the chiral-imprinted route. For instance, mesoporous PtIr alloy structures with chiral cavities are synthesized by electrodeposition method using the combination of surfactant and chiral molecule as template. Good enantioselectivity has been achieved with the chiral PtIr alloy structure for electrocatalytic hydrogenation of acetophenone to phenylethanol. Furthermore, the chiral PtIr structure show much higher enantiomeric excess than the Pt structure, which could be associated with the higher structural stability of PtIr than Pt and therefore a more well-defined chiral structure at atomic scale.

230Another practical approach for achieving enantioselective catalysis is to employ porous framework with ordered chiral microenvironment as host for bimetallic nanoparticles. This concept has been demonstrated with metal complexes encapsulated in chiral MOF structures but not shown with bimetallic nanoparticles. Considering the suitable reactions for bimetallic nanoparticles, encapsulation of bimetallic nanoparticles in chiral MOFs could be a class of promising recyclable catalysts for enantioselective hydrogenation and oxidation reactions.

231Besides, the combination of bimetallic nanoclusters or nanoparticles with chiral zeolites can provide even better stability than the MOF structure, although it has not been achieved yet to date., Moreover, zeolite catalysts with specially modified structures can be prepared by imprinting the transition states of the target reactions by using the mimics as organic-structure directing agents. If the above two concepts are combined, it would lead to the formation of supported metal catalysts confined in well-defined chiral environment with customed surrounding environment for catalyzing enantioselective reactions.

232The above-mentioned works have demonstrated that various structural parameters can influence the catalytic properties of bimetallic nanoparticles. It can be anticipated that the complexity in terms of structure–reactivity relationship will be greatly elevated if more metal elements are introduced. Thanks to the developments in synthesis of high-entropy alloy nanoparticles, the exploration of the catalytic applications of these materials are emerging topics in catalysis community., The advantages of high-entropy alloy catalysts have been proved with various types of reactions, such as decomposition of ammonia, oxidation of ammonia to NOx, combustion of methane, and electrocatalytic oxidation of methanol and ethanol., As shown in Figure 55, the high-entropy alloy catalyst shows much better selectivity to desired products for oxidation of ammonia to NOx than the conventional phase-separated multimetal nanoparticles prepared by conventional method. The PtPdRhRuCe nanoparticles remain stable for over 30 h at 700 °C under the conditions for ammonia oxidation reaction. In comparison to the well-established PtPdRh catalysts used in the industrial process, the addition of Ru and Ce can inhabit the segregation of the noble metals and alleviate the sintering.

233The increasing number of publications on catalytic applications of high-entropy alloy nanoparticles calls for a unified understanding on the advantages of high-entropy alloy nanoparticles over the conventional monometallic and bimetallic catalysts. This goal could be achieved by translating the knowledge and experiences accumulated in previous studies with the bimetallic systems to the more complicated systems. It is well established that PtM bimetallic particles (M = Sn, Ga, In, Zn, etc.) can greatly prolong the lifetime of the Pt catalyst for dehydrogenation of light alkanes to alkenes by suppressing the coke deposition. The introduction of a third metal (such as Pb) into the bimetallic PtGa nanoparticles leads to further improvement in the catalyst’s lifetime because the Pb atoms can bond with the Pt atoms surface of PtGa alloy particles and then transform the Pt ensembles into isolated Pt sites, which are highly active and selective for transforming propane into propylene but not active for cleavage of the C–H bonds in propylene, thus suppressing the coke deposition. The catalyst’s lifetime can be further extended by employing high-entropy PtCoCuGeGaSn alloy nanoparticles supported on SiO2 as the catalysts, in which the Pt sites are claimed separated by the transition and main-group metals. As shown in Figure 56, the catalyst’s lifetime increases with the degree of Pt isolation represented by the loading of Pt in the high-entropy alloy particles. The DFT calculations and the temperature-programmed desorption experiment results suggest that the adsorption of propylene on the high-entropy alloy particles is almost completely suppressed, which contributes the exceptionally high stability.

234Some theoretical calculation studies propose that the electronic features of high-entropy alloys are greatly different to the monometallic or bimetallic particles. However, the calculation results are not convincing enough to be correlated with the experimental results, because the realistic structures of high-entropy alloy particles are too complicated to be modeled. To have a reliable description on the electronic features of high-entropy alloy particles, a practical approach is to employ machine-learning strategies to accelerate the calculation process. This strategy usually employ the use of reactivity descriptor (such as the adsorption energy of the reactant or intermediate on the catalyst’s surface) as the indicator for fast screening and the choice of the descriptor relies on the fundamental understanding on the reaction mechanism of the target reaction., In this sense, the knowledge accumulated with bimetallic systems can be translated to the high-entropy alloys to facilitate the exploration of a large number of potential candidate materials.

235In order to further explore the potentials of catalysts based on high-entropy alloys, the methodologies for materials characterizations developed for monometallic and bimetallic nanoparticles need to be adjusted or even to be reinvented. For instance, there are no well-established methods/techniques for measuring the number of exposed surface sites, the spatial distribution of the elements, and the atomic structures of the high-entropy alloys. Besides considering the theoretically infinite configurations of high-entropy alloy catalysts in terms of their composition, it will be necessary to employ high-throughput techniques to efficiently explore the synthesis, characterization, and catalytic applications of these materials in both theoretical and experimental approaches., Nevertheless, to achieve a unified understanding on the structure–reactivity relationships in high-entropy alloy catalysts, the structural features and catalytic performances should be extracted from the experimental results and classified properly in order to build a database for in-depth analysis.

236From a mechanistic point of view, there should be an optimum active site in the high-entropy alloys for the target catalytic reaction, although there are numerous types of sites in those particles. In this sense, the percentage of the useful metal atoms contributing to the formation of the real working active sites in the whole catalyst could be low. Therefore, employing advanced characterization techniques and performing rigorous kinetic studies are crucial to identify the active sites which make the major contribution, and thereafter the rational design of multimetallic catalysts would be possible.

237In this section, we have witnessed the diversity of the structural features of bimetallic nanocatalysts and the structure–reactivity relationship proposed in the literature works have greatly enriched our knowledge and fundamental understanding on the catalytic properties of the emerging bimetallic nanostructures. For a specific reaction, various types of bimetallic nanostructures may show activity for the desired transformations and some inconsistence can be found with different literature works in terms of the nature of the active sites. These discrepancies hinder the formation of a unified understanding on the structure–reactivity relationship and rational design of bimetallic and multimetallic nanocatalysts. To improve the efficiency of research activity toward understanding the nature of the active sites, we will propose a general workflow for elucidating the plausible active sites in bimetallic catalysts in the next section of this review.

238From the above discussions, it appears that heterogeneous bimetallic catalysts can offer many possibilities in the fields of catalysis, more specifically, in the field of electrocatalysis. In this section, we will present perspectives on the future developments of heterogeneous bimetallic catalysts and more broadly, the future directions of heterogeneous catalysis research.

239Because the geometric structures of the bimetallic sites have a profound influence on their catalytic performance, the precise control of the structural features determines the homogeneity of the active sites in the final catalyst. In principle, synthesis methodologies for binuclear bimetallic sites and bimetallic nanoclusters with well-defined structures have been established. However, these methods usually require the use of organic ligands for the stabilization of the tiny metal species, which would normally block the active sites. Grafting the binuclear sites or bimetallic nanoclusters on solid carriers and employing appropriate treatments may liberate the metal atoms from the blockage by the ligands. However, during the removal of the organic ligands, sintering of the binuclear sites or bimetallic nanoclusters becomes a critical issue that has not been properly solved yet in many systems. To mitigate the sintering issues, the choice of the solid carrier is critical for providing anchoring sites to limit the mobility of the bimetallic entities. Besides, synergistic effects between the bimetallic nanoclusters and the support may lead to novel catalytic properties than the unsupported nanoclusters.

240Another approach for constructing uniform binuclear bimetallic sites and bimetallic nanoclusters on solid carriers is the use of crystalline microporous materials as the host for the tiny bimetallic entities, as mentioned before in this review. We would like to emphasize here that MOF structures will be excellent supports for accommodating subnanometer bimetallic entities because the nodes and linkers provide well-defined sites for the generation of the bimetallic sites through controllable chemical transformations. In the case of taking zeolites as the support, the structural rigidity of the zeolite framework allows for the generation of highly stable bimetallic sites, but it remains a great challenge to precisely control the coordination environment of the metal species in the zeolite structure. The coordination environment of metal species located inside the pores/cavities of zeolites is not as clearly identified as the situation with the metal catalysts supported on MOFs, especially for the noble metal species stabilized in zeolites. In particular, the anchoring sites of noble metal species and the mechanism for stabilization of subnanometer metal clusters inside the zeolite structure need to be elucidated.

241In the case of bimetallic nanostructures, the control of their geometric structures with atomic-level precision is challenging due to the presence of hundreds of atoms in a single bimetallic nanoparticle. From a practical point of view, the synthesis of bimetallic nanoparticles with high uniformity in terms of the chemical composition and the spatial distribution of the metal elements in the particles is crucial for approaching the “ideal” solid catalysts with only one type of active site. In a first attempt, this goal may be achieved by preparing single-atom alloy nanoparticles in which the isolated metal atoms are selectively located in specific sites in the matrix metal. In most of the reported single-atom alloy particles, the spatial distributions of the isolated metal atoms are not precisely controlled. It can be expected that the coordination environment and electronic features of the isolated metal atoms could be dependent on their location in the matrix. To address such impacts, one may use metal nanocrystals with regularly exposed facets as the matrix and then deposit isolated metal sites on the specific facet of the metal nanocrystals for the generation of single-atom alloy structures with well-defined geometric features.

242For many bimetallic nanoparticles, the active sites are associated with the metal ensembles. However, the optimum structural features of the metal ensembles are elusive to be elucidated because of the challenges in the preparation of a series of model catalysts with different sizes of metal ensembles. Besides the size of the metal ensembles, the surrounding environment of the metal ensembles, whose impacts on the catalytic performances have barely been studied, can be a determinant. The clarification of the above-mentioned structure–reactivity relationships calls for the improvement in the precision of the synthesis methodologies for the generation of bimetallic catalysts with uniform and well-defined structures.

243Currently, state-of-the-art characterization techniques are able to directly measure the particle sizes of the bimetallic entities on solid carriers by advanced electron microscopy techniques and then provide information on the coordination environment of the bimetallic species by spectroscopy tools. However, the information from microscopy techniques is highly local, while the spectroscopy characterization results usually give an overall estimation of the whole sample. Taking into account that in the vast majority of supported bimetallic catalysts, several types of bimetallic entities are simultaneously present, a reliable structure–reactivity relationship needs to be established based on the quantification of the number of each type of bimetallic entity. However, to the best of our knowledge, there is no general characterization technique for achieving that goal, which requires a combination of electron microscopy and spectroscopy techniques., This strategy has been shown with the combination of in situ TEM and in situ X-ray absorption spectroscopy for characterizing the evolution of supported Pt catalyst, which allows giving a more reliable description of the contribution of different types of Pt species in the whole catalyst.

244The determination and quantitative measurements of the active sites will be critical for revealing the nature of the active sites and modeling them, in high-entropy alloy nanoparticles, which are emerging catalytic materials in recent years. In principle, there should be an optimized active site for the target reaction in the high-entropy alloy nanoparticles. However, due to the structural complexity, the number of the optimized sites should be low because the exposed metal sites are randomly distributed, showing highly disordered structures. In this sense, there should be plenty of room for improving the catalytic performances of high-entropy alloy nanoparticles if the catalytically relevant sites can be identified, modeled, and then generated in a rational way.

245Fundamental understandings of the nature of the active sites in supported metal catalysts will reveal the advantages and limitations of each type of metal entity. For a given reaction, the catalytic cycle usually invokes several critical elementary steps, which may require different types of active sites, as demonstrated with the broad applications of bifunctional catalysts made by metallic and acid/basic sites and in the case of supported metal catalysts based on metal entities with different particle sizes.

246The combination of different metal entities as active sites has been demonstrated to be an effective strategy for promoting the overall performances of some reactions. For instance, to mitigate the deactivation of the anode catalyst in proton-exchange membrane fuel cells by CO, an Ir/C catalyst comprising the mixture of Ir nanoparticles and isolated Ir atoms has been prepared, in which the isolated Ir atoms can facilitate the oxidation of CO adsorbed on Ir nanoparticles. As shown in Figure 57, the poisoning of Ir nanoparticles by CO is greatly suppressed and the superior performance of Ir nanoparticles for H2 oxidation reaction can be fully exhibited. According to theoretical calculations, active *OH species are formed due to the cleavage of the H–O bond in H2O at the isolated Ir atoms embedded in the N-doped carbon matrix, which can further react with the CO adsorbed on Ir nanoparticles to oxidize the CO into CO2. In this regard, the isolated Ir atoms should be located at the adjacent sites of the Ir nanoparticles in order to achieve the synergy between the two types of Ir entities. In another example, the combination of isolated Fe atoms and Fe clusters supported on N-doped carbon can boost the performance of isolated Fe atoms by placing Fe clusters at the neighboring sites, because the electronic features of the FeN4 sites are influenced by the Fe clusters, resulting in a lower energy barrier for the formation of *OH species on the FeN4 sites.

247The advantages of combining metal entities of different sizes have been shown in thermal catalytic reactions, such as the combination of isolated Rh atoms and Rh nanoparticles for dehydrogenation of cyclohexanol to phenol and the combination of isolated Pd atoms and Pd nanoparticles for hydrogenation of ketone and aldehyde compounds. In the case of Rh catalyst, the spatial intimacy of the isolated Rh atoms, and Rh nanoparticles seem not as critical as the situation in the above-mentioned supported Ir catalyst because the intermediates can transport from one site to another site for completing the catalytic transformation. In the case of Pd catalyst, the role of Pd nanoparticles is claimed to be associated with the dissociation of H2, and the activated H species will be transferred to the isolated Pd atoms, where the carbonyl groups are adsorbed, for performing the hydrogenation of the ketone and aldehyde compounds. Therefore, the isolated Pd atoms should be located at the neighboring sites of Pd nanoparticles in order to maximize the efficiency of the H transfer process. In an example of Rh/CeO2 system comprising isolated Rh atoms and Rh clusters supported on CeO2, the production of ethanol from syngas is ascribed to the synergy of Rh atoms and Rh clusters for CO hydrogenation and C–C coupling reaction, respectively.

248The above-mentioned examples are based on the combination of monometallic clusters/nanoparticles and isolated metal atoms. In principle, the synergistic effects of atomically dispersed metal species and agglomerated metal entities (clusters and nanoparticles) should also work for bimetallic systems. For instance, as illustrated in Figure 58, a mixture of bimetallic clusters/nanoparticles and isolated metal atoms (one of the metal elements in the bimetallic clusters/nanoparticles) can be constructed by loading the bimetallic entities onto a solid carrier comprising isolated metal atoms. This concept has already been demonstrated with the PtFe nanoparticles on metal–N-C support, in which the isolated metal atoms in the support can greatly promote the durability of PtFe and PtCo alloy nanoparticles by suppressing the leaching and sintering.,

249Although the size of bimetallic entities can vary greatly from the subnanometer regime (such as binuclear sites and tiny bimetallic clusters) to the nanometric regime, the active site that is directly involved in the catalytic cycle should be a highly local area. For the vast majority of the reactions catalyzed by heterogeneous metal catalysts, the substrate molecules should interact with a few atoms (probably less than 10 atoms) at the active site, including the atoms from the metal entities and from the support. In this sense, for a specific reaction, the active site in supported bimetallic catalysts comprising different bimetallic entities may share a common structural feature. However, for a given reaction, there are always arguments in the reported works in terms of the nature of the active sites, which precludes the formation of a unified description on the active sites. Some of the different conclusions may result from the discrepancies in the experimental conditions, and some of the contradictive viewpoints may originate from the mistakes in characterizing the “real” active sites.

250In the past decade, we have witnessed a very rapid growth of the number of publications in all the fields of catalysis research (spanning from homogeneous to heterogeneous and enzymatic catalysis). On one hand, a large number of new publications will certainly enrich our knowledge in catalysis and provide many information/insights on new synthesis methodologies for catalyst preparation, new techniques for catalyst characterizations, and interesting catalytic properties of new materials. On the other hand, the appearance of publications without strict quality control may cause confusion or even mislead to other researchers, which hinders the formation of a reliable and unified understanding of the active sites in solid catalysts for a target reaction. Unified understandings on the nature of the active sites rely on rigorous design and careful execution of the experiments and adequate data analysis. For instance, in order to clarify the optimum size of metal entities for a specific reaction, it is recommended to prepare a series of supported metal catalysts by a same method but modulate the particle size distributions by tuning a few synthesis parameters (the less the better). Otherwise, some unknown factors such as impurities in the final solid catalysts could affect the reliability of the comparative studies.

251As emphasized along the whole review, the dynamic structural transformations of supported metal catalysts are ubiquitous and the overlooked evolution of the metal entities under reaction conditions may lead to mistakes in identification of the active sites. Such have been frequently encountered with supported monometallic catalysts, and it can be anticipated that such mistakes will more often occur with supported bimetallic catalysts due to the higher degree of structural complexity than the monometallic system. In this regard, it becomes very important to follow the structural features of bimetallic catalysts (morphology, particle size, spatial distribution of the metal elements, etc.) under reaction conditions, or at least simulated reaction conditions. It should be noted that the data derived from the characterizations will provide multiple potential parameters/descriptors for establishing the structure–reactivity correlation, which also relies on distinguishing the critical parameters/descriptors from the spectators.

252Correlating the knowledge and lessons generated in different fields of catalysis is helpful for achieving unified understandings on the active sites. For instance, dehydrogenation of alkanes into alkenes can be catalyzed by homogeneous and heterogeneous catalysts., The homogeneous catalysts rely on the modification of the coordination environment (usually through the modification of the ligand/solvent) of the metal centers, while the heterogeneous catalysts rely on the tuning the size and composition of the metal entities, as well as the near and the next nearest neighbors. It is then of interest to compare the working mechanism of homogeneous and heterogeneous systems and then figure out the similarities and dissimilarities for the same type of reaction. This approach will be especially helpful for acquiring new insights when combining the available theoretical calculations and characterization techniques.

253Since the establishment of the fundamental concepts in catalysis in the 19th century, heterogeneous catalysis has evolved into an interdisciplinary field, in which scientists with different backgrounds spanning from chemistry to materials science and engineering are attracted to tackle the challenges in fundamental or practical perspectives. During the 20th century, the tremendous progress in heterogeneous has made great impacts on the chemical industry, which shapes our current energy landscape and human society. The vast majority of the concepts, tools, and methodologies of heterogeneous catalysis are established on the research activities in dealing with substances from fossil energy resources. Nevertheless, catalysis and the chemistry principles are the same regardless of the origin of the reactants. Therefore, catalysis will continue to serve as an indispensable role in converting renewable energy into chemicals and materials for a sustainable world. For instance, delocalized production of chemicals via electrochemical processes are considered as a promising approach to utilize the renewable electricity produced from solar and wind energy, although there are many technical problems remaining to be solved. Taking into account of the advantages of thermal catalytic processes in scaling up, the coupling of the electrochemical and thermal processes for converting feedstocks (such as CO2, N2, H2O, biomass, plastic waste, etc.) into value-added products may improve the overall efficiency and facilitate the realization of these systems. Developing coupled catalytic systems require the combination of knowledge from different disciplines. For instance, one needs to learn the principles of thermal catalysis and electrochemistry in order to develop a hybrid catalytic system based on the coupling of thermal and electrochemical processes.− For instance, the use of high-temperature solid oxide fuel cells and membrane reactor heavily depend on the integration of design of electrochemical devices and application of solid catalysts.,

254The coupling of different expertise for developing novel catalytic systems can also be achieved by introduction of other external excitations into the conventional thermal reaction systems. For instance, the combination of light irradiation and heating gives to the realization of photothermal catalytic system, and the introduction of light irradiation on the solid catalysts can lower the energy barrier of the activation of the reactants and modulate the intermediates formed on the catalyst’s surface, leading to superior performances or distinct product distributions compared to the conventional thermal systems.,, However, the contribution of light irradiation and the mechanism of the synergy of light irradiation and thermal heating in the catalytic cycles are not clearly elucidated. The situation is similar with other hybrid catalytic systems assisted by external fields, such as the electric-field-assisted reactions and magnetic-field-assisted reactions, which requires further studies to reveal the fundamental mechanisms, especially the reaction pathways in the hybrid catalytic systems.−

255Owing to the advances in the synthetic modification of enzymes, the synthesis of enzyme mimicking molecules, and genetic modification of cells to produce new enzymes, will open more opportunities for manufacturing chemicals from biomass. No doubt that enzymatic and chemo-enzymatic processes will add more potential to the catalysis field, and the combination of enzymes and solid catalysts to form hybrid systems will enable novel catalytic transformations which are not feasible via conventional routes.− As always, we expect major advances within the interphase of the different confluent fields and this progress will bring new insights and inspirations to future developments of catalysis science and technology.

256Finally, the rapid developments of software and hardware tools based on artificial intelligence and machine learning are also bring new opportunities to catalysis research. The information extraction based on natural language processing offer the possibility to summarize the useful information from the literature and further analysis based on machine learning will guide the design of experiments for materials synthesis. This strategy has already been demonstrated with the development of solid catalysts for selective oxidation reactions and the synthesis of zeolite catalysts with novel organic structure-directing agents.− Moreover, the automated high-throughput techniques can greatly promote the screening of the catalytic materials for a target reaction, and the data analysis based on artificial intelligence can dramatically decrease the time of catalyst optimization.− The combination of machine learning and theoretical calculations will greatly promote the efficiency of modeling the atomic structures of complex solid catalysts (such as zeolites and bimetallic nanoclusters/nanoparticles), which provide new insights on the elementary steps (diffusion to the catalyst’s surface, adsorption, surface reaction, and desorption) on solid catalysts.,