Atomically Precise Cu Nanoclusters: Recent Advances, Challenges, and Perspectives in Synthesis and Catalytic Applications

Highlights

• Summarizing recent advances on synthesis and catalytic applications of Cu nanoclusters.

• The structure–property–functionality relationship is clearly elucidated.

• Critical analysis of the current challenges and future perspectives.

Abstract

Atomically precise metal nanoclusters are an emerging type of nanomaterial which has diverse interfacial metal–ligand coordination motifs that can significantly affect their physicochemical properties and functionalities. Among that, Cu nanoclusters have been gaining continuous increasing research attentions, thanks to the low cost, diversified structures, and superior catalytic performance for various reactions. In this review, we first summarize the recent progress regarding the synthetic methods of atomically precise Cu nanoclusters and the coordination modes between Cu and several typical ligands and then discuss the catalytic applications of these Cu nanoclusters with some explicit examples to explain the atomical-level structure–performance relationship. Finally, the current challenges and future research perspectives with some critical thoughts are elaborated. We hope this review can not only provide a whole picture of the current advances regarding the synthesis and catalytic applications of atomically precise Cu nanoclusters, but also points out some future research visions in this rapidly booming field.

Introduction

Uniqueness of Atomically Precise Cu Nanoclusters

The last decade has witnessed the great success of nanoscience and nanotechnology, yet ideal research model with well-defined composition and structure is still lacking. Specifically, most of the studied nanomaterials are quite polydisperse, that means, in most studied systems, the nano-scientists are extremely difficult to find two same nanoparticles with identical size, morphology, composition, and structure. To advance nanoscience and nanotechnology, one of the ultimate goals for the nano-scientists is to find a truly uniform system, in another word, atomically precise nanoparticles as research models [1, 2].

The emergence of atomically precise coinage metal (Au, Ag, Cu, and its alloy, etc.) nanoclusters can realize such goal [1]. Atomically precise metal nanocluster is a novel type of nanomaterial with the size in the sub-nanometer regime, normally 1–3 nm in diameter. It usually comprises 10–300 metal atoms with surface ligand as the protecting agents capping on the metal core. Note that as the nanoparticle size decreases downward to the sub-nanometer range, due to the strong quantum confinement effect, distinctly different physicochemical properties of metal nanoclusters from relatively large metal nanoparticles are observed [3]. For instance, discrete optical absorbance features can be readily identified in molecular Au nanoclusters, but such feature is absent in larger Au nanoparticle counterparts [4]. More importantly, the sub-nanometer size of the metal nanocluster is still within the resolving limitation of single crystal X-ray diffraction (SC-XRD), that allows nano-chemists to resolve their structure with atomic precision [5]. Such precise structure cannot be available for many other nanomaterials, which render metal nanoclusters unique advantages to comprehensively study the structure–property relationship in various fields, such as sensing [6, 7], assembly [8, 9], catalysis [10–12], optoelectronic [13, 14], and cancer therapy [15–17]. Furthermore, the chemical stability of these metal nanoclusters in terms of electronic structure can be explained by the “superatom” theory, where the electrons are confined in the spherical metal core of a jellium model [18]. It is believed that if the free electron number of a cluster is in good agreement of inert gas atoms (e.g., 2, 8, 18, 34, 52 electrons in the outmost orbital for He, Ne, Ar, Kr, Xe), it can be considered as a superatom having robust stability [18]. Meanwhile, the thermodynamic stability of thiolate metal nanoclusters is associated with the energy balance between the adsorption strength of the ligand shell to the metal core and the cohesive energy of the metal core [19]. In addition, for the non-magic number metal nanoclusters, the Wang group developed a superatomic orbital splitting (SOS) theory to understand the electronic configuration, where the shape of the metal core is considered in determining the order of the group orbital levels [20].

It is worth noting that, compared with noble metal nanoclusters such as Au, Ag, Pd, and Pt, Cu nanoclusters possess some unique characteristics. First, Cu is more earth abundant hence can be more cost effective for large-scale production to prepare functional nanomaterials. Secondly, Cu possesses different valence states including 0, +1, +2, which render Cu atom with different spatial arrangement in the metal core. So far, superatomic Cu nanoclusters with Cu(0) core, Cu(I) clusters, and Cu(II) complexes have been extensively reported [21, 22]. It is worth noting that different charge states impart Cu nanomaterial with drastically different physicochemical properties and reactivities, and the variation of the valence state during some catalytic reaction has also been observed [23–25]. Finally, Cu nanoclusters hold some peculiar functionalities that are not available from other noble metal nanoclusters [26]. For instance, Cu has good adsorption for CO₂ than hydrogen, and it has strong capability to construct C–C bond, so in electrochemical CO₂ reduction, using Cu nanoclusters can suppress the hydrogen evolution reaction and obtain HCO₂H, CH₃OH, CH, and even more valuable C2+ products such as C₂H₄ and C₂H₅OH [27], while only CO can be acquired when using Au or Ag nanoclusters as catalysts in most cases [28–34]. Table 1 summarizes all the Cu nanoclusters which have been discussed in this paper.

Table 1.

The formula, metal core configuration, and catalytic reaction of Cu nanoclusters discussed in this paper

Formulas	Cluster configuration	Catalytic reaction	Refs.
[Cu6(MBD)6]	Octahedron	CO2RR	[27]
[Cu29Cl4H22(Ph2phen)12]Cl	Cu13@Cu16	–	[35]
[Cu20(C≡CPh)12(OAc)6)]	Cu4@Cu16	[3 + 2] Cycloaddition	[36]
[Cu53(C≡CPhPh)9(dppp)6Cl3(NO3)9]	Layer-structure	–	[37]
[Cu26(DPPE)3(CF3CO2)8(CH3O)2(tBuC≡C)4H11]+	Cu4@Cu6@Cu16	CO2RR	[38]
[Cu32(PET)24H8Cl2](PPh4)2	Cu18@Cu4PET2@ Cu14(PET)22Cl2	Carbonylation of aniline	[39]
[Cu25H22(PPh3)12]Cl	Cu13@Cu12	–	[40]
[Cu18H17(PPh3)10]Cl	Cu8@Cu10	–	[41]
[Cu18H3(S-Adm)12(PPh3)4Cl2]	Cu10@Cu8S12P4	–	[42]
[Cu31(4-MeO-PhC≡C)21(dppe)3](ClO4)2	Cu13@Cu18	–	[43]
[Cu6(4-MeO-PhC≡C)5(dppe)3](ClO4)	Trigonal-prismatic configuration	–	[44]
[Cu50(CF3COO)12(3, 5-diMe-PhS)18(PPh3)4H2]	Cu44@Cu6	–	[45]
[Cu41Cl2(2-F-C6H4S)12(CF3COO)6(PPh3)6H19]2−	Cu29@Cu12	p-Nitrophenol reduction	[46]
[Cu41(2, 5-di-Methyl-C6H3S)12(BO3)3Cl3(PPh3)6H19]	Cu29@Cu12	p-Nitrophenol reduction	[46]
[Cu23(tBuC≡C)13(CF3COO)6]	Cu4@Cu19	–	[47]
[Cu23(tBuC≡C)13(CF3COO)6]·CHCl3	Cu4@Cu19	–	[47]
[Cu53(RCOO)10(C≡CtBu)20Cl2H18]+	Cu3@Cu10Cl2@Cu20@Cu20	–	[48]
[Cu13Na2(CZ-PrA)6(TC4A)2Cl(CH3OH)2]	Cu9@Cu4	Sulfide oxidation	[49]
[Cu13Na(CZ-PrA)6(TC4A)2(CH3OH)]·CH3OH·CH2Cl2·CH3COCH3	Cu9@Cu4Cl	Sulfide oxidation	[49]
[Cu13(S2CNnBu2)6(C≡CR)4](PF6)	Cu@Cu12 cuboctahedron	–	[50]
[Cu32H20(S2P(OiPr)2)12]	Cu14@Cu9@Cu9	CO2RR	[51]
[Cu8(H)(L1)6PF6] (L1 = 9H-carbazole-9-carbodithioate)	Cu4 double tetrahedron	CO2RR	[52]
[Cu8(tBuS)4(L1)4] (L1 = 9H-carbazole-9-carbodithioate	Cu4 double tetrahedron	CO2RR	[52]
[Cu8(tBuS)4(L2)4] (L2 = O-ethyl carbonodithiolate)	Cu4 double tetrahedron	CO2RR	[52]
[Cu13(C4B10H11)10(PPh3)2(CH3CN)2](PF6)3	–	Nitrate reduction	[53]
[Cu26(C4B10H11)16(C2B9H10C2)2(PPh3)2(CH3CN)4](PF6)4	–	Nitrate reduction	[53]
[Cu6(HL1)2(L1)4(PF6)2]	Cu6 octahedron	CO2RR	[54]
[Cu61(StBu)26S6Cl6H14]	Cu19@Cu42(StBu)26S6Cl6	–	[55]
[Cu57H20(PET)36(TPP)4]+	Cu14@Cu43S36	–	[56]
[Cu58H20PET36(PPh3)4]2+	Cu8@Cu6@Cu24@Cu12 @Cu8	[3 + 2] Cycloaddition	[57]
[Cu57H20PET36(PPh3)4]+	Cu8@Cu6@Cu24@Cu12 @Cu7	[3 + 2] Cycloaddition	[57]
[Cu28H10(C7H7S)18(TPP)3]	Cu13@Cu15S18P3	C−C coupling reaction	[58]
[Cu7(SC5H9)7(PPh3)3]	–	C−C coupling reaction	[59]
[Cu3(NHC)3(PF6)3]	–	A3 coupling and Redox-A3 coupling reaction	[60]
[Cu8(Tf-dpf)4(NO3)2](NO3)2	Two thin slices of supramolecular layers	“Aldehyde–acetylene–amine” A3 coupling reaction	[61]
[Cu8(L2)2(L3)2](L2, L3 = deprotonated pyrazole)	–	Synthesis of indolizine	[62]
Cu18H(PET)14(PPh3)6(isothiocyanate)3	Cu15@Cu3	Reduction of ferricyanide to ferrocyanide	[63]
[Cu66Cl8(PPh3)8(SC2H5)32H24](SbF6)2	Layered arrangement	Hydrogenation of cyclohexanone	[64]
[Cu20H9(Tf-dpf)10]·BF4	Cu12@2Cu4	Conjugate reduction of cinnamaldehyde	[65]
[Cu20H8(Tf-dpf)10]·(BF4)2	Cu12@2Cu4	Conjugate reduction of cinnamaldehyde	[65]
Se@Cu20(PhSe)12(PPh3)2(C6H5COO)6	Se@Cu6@Cu3@Cu3@Cu3@Cu3@Cu2	4-Nitrophenol reduction	[66]
Se@Cu20(PhSe)12(PPh3)2(CF3COO)6	Se@Cu6@Cu3@Cu3@Cu3@Cu3@Cu2	4-Nitrophenol reduction	[66]

– denotes none or not available

Cu–Ligand Coordination Modes in Cu Nanoclusters

To prevent the Cu atoms from agglomeration or aggregation, surface capping ligand is critical as the stabilizing agent to allow Cu atoms to form a certain small core but not into a bulky core or an agglomerate. Furthermore, the ligand can significantly affect the physical and chemical properties of Cu nanoclusters through binding with the surface Cu atoms [67–69]. These interfacial coordination moieties form the ligament between the metal core and the surface ligands, and such ligament governs the electron coupling behaviors. So far, various types of molecules have been employed as capping ligand to stabilize the Cu core, and four typical widely employed organic molecules are thiolate [1, 3], phosphine containing molecule [70], nitrogen containing molecule [71], and alkynyl ligand [72, 73], yet they have markedly different coordination modes to bind the Cu atoms (Scheme 1). As illustrated in Scheme 1a, one S atom can coordinate with one, two, three, and four Cu atoms [39], and one Cu atom can coordinate with one, two, three, and four S atoms [74, 75]. One Cu atom can coordinate with one or two P atoms [40, 41], and one P atom can coordinate with three Cu atoms (Scheme 1b). Interestingly, three Cu atoms and three P atoms can form a cyclic triangle coordination mode [76]. The coordination mode between N and Cu is somewhat similar to that between P and Cu, as one Cu atom can coordinate with one or two P atoms, and both N and P atoms can form the dimer mode of R–N(P)–Cu (Scheme 1c) [21, 35, 77, 78]. However, unlike P, one N atom cannot coordinate with two or more Cu atoms, but it can form the trimer mode of R–N–Cu, and a cyclic quadrangle mode of Cu₄N₄ is also available [77]. It is worth noting that, for S, N, P atoms, they coordinate with the Cu atom mainly through σ bonding, nevertheless, for alkynyl ligand, it can coordinate with Cu atom with either σ bonding or π bonding or both. As shown in Scheme 1d, one alkynyl ligand can not only coordinate with three or four Cu atoms with only σ bonding, but also can coordinate with two, three, or four Cu atoms with both σ bonding and π bonding [36, 37, 79, 80]. There are seven essential types of coordination modes: μ₂-ƞ¹, ƞ²; μ₃-ƞ¹, ƞ¹, ƞ¹; μ₃-ƞ¹, ƞ¹, ƞ²; μ₃-ƞ¹, ƞ², ƞ²; μ₄-ƞ¹, ƞ¹, ƞ¹, ƞ¹; μ₄-ƞ¹, ƞ¹, ƞ¹, ƞ²; μ₄-ƞ¹, ƞ¹, ƞ², ƞ². Such unique binding modes can impart alkynyl-protected Cu nanoclusters with some drastically different physicochemical properties and functionalities from thiolate, nitrogen, and phosphine ligand stabilized Cu nanoclusters [72, 73].

Scheme 1.

Metal–ligand coordination modes in Cu nanoclusters. a Cu–S coordination modes, b Cu–P coordination mode, c Cu–N coordination mode, d Cu–alkynyl carbon coordination mode

Synthesis of Atomically Precise Cu Nanoclusters

There are various methods to synthesize bulky metal nanoparticles. However, preparing metal nanoclusters with atomic precision is still quite challenging. This is because metal nanoclusters with molecular purity typically form under specific thermodynamic or kinetic conditions. Since the pioneering Brust method to prepare thiolate-protected Au nanoclusters reported at 1990s [81, 82], a number of methods have been developed to prepare Au nanoclusters or Ag nanoclusters protected by thiolate, alkynyl molecule, and other ligands. The typical approaches include, but are not limited to, the direct reduction of the precursor, one-pot strategy, bi-phase method, ligand exchange or etching and so on [4, 83–87]. It is worth noting that, to prepare monodisperse metal nanoclusters, there are a lot of factors influencing the output, e.g., the reaction temperature, the solvent, the reducing agent, the nature and chemical property of the ligand molecule, the stoichiometric ratio of ligand–reductant–reactant, and so on. For instance, our group developed a synchronous nucleation and passivation strategy to synthesize alkynyl-protected coinage metal nanoclusters [88], and so far, several Au [88, 89], Ag [29, 32], AuAg [90–92], AgPd [93], AgRh [94, 95] nanoclusters have been successfully fabricated by this method [73, 96], but this approach is not applicable to thiolate-protected metal nanoclusters. Meanwhile, no successful case has been achieved on Cu nanoclusters yet. It also implies that, the approaches developed for synthesizing molecular thiolate-protected metal nanoclusters or alkynyl-protected metal (and alloy but not pure Cu) nanoclusters are not applicable for preparing atomically precise Cu nanoclusters, probably due to the different metal–ligand interactions can alter the nuclei growth and surface passivation behaviors [97].

Nevertheless, several generic methods have been developed for synthesizing monodisperses or molecular Cu nanoclusters. It includes the precursor reduction method, gradient reduction strategy (GRS), one-pot synthesis (OPS), ligand-exchange-induced growth, and other ingenious methods (Fig. 1). Each method has its own advantages or disadvantages. For example, the precursor reduction method is quite straightforward, but polydisperse cluster product may be acquired; the gradient reduction strategy features two step reduction with manipulation accessibility at each step, but the overall yield might be quite low; the one-pot synthesis holds facile operation but complex product is highly possible; the other methods are more applicable to special ligand to target specific cluster molecule. The advantages and disadvantages of these methods with specific cases will be elaborated next.

Fig. 1.

Current synthetic methods for atomically precise Cu nanoclusters

Precursor Reduction Method

The simplest and most straightforward method to fabricate Cu nanoclusters is to form the precursor first and then reduce it. However, under most circumstances, polydispersed Cu nanoclusters with a wide size distribution are acquired. To improve the monodispersity, kinetic control is inevitable. One typical example is the synthesis of [Cu₁₈H₃(S-Adm)₁₂(PPh₃)₄Cl₂] molecule reported by Mandal group [42]. As shown in Scheme 2, Cu(CH₃CN)₄BF₄ was treated with auxiliary PPh₃ ligand first, and then the adamantanethiol (Adm-SH) ligand was added to form the Cu(I) complex. Upon the addition of the NaBH₄ methanol solution, the mixture changed from colorless into red, indicating the Cu₁₈ nanocluster was formed. Note that, the employment of the bulky Adm-SH ligand is critical for yielding the Cu₁₈ nanocluster here, as it acts as the main surface protecting ligand to stabilize the framework of the Cu₁₈ nanocluster [42]. The Cu(I) precursor, also called as the Cu(I) complex, can be prepared step by step. In 2020, Li et al. reported a copper hydride cluster of [Cu₃₂(PET)₂₄H₈Cl₂](PPh₄)₂ (PET = phenylethyl thiolate) [39]. For its synthesis, Cu(TMEDA)Cl (TMEDA: tetramethylethylenediamine) was first obtained by reacting CuCl with TMEDA. Upon addition of PPh₄Br and the thiol ligand (PETH), the final stage precursor of CuPET(TMEDA)Cl was obtained. The reduction of CuPET(TMEDA)Cl by sodium borohydride can yield Cu₃₂ nanoclusters [39]. In 2023, Jia et al. utilized this method to fabricate an eight-electron superatom of [Cu₃₁(4-MeO-PhC≡C)₂₁(dppe)₃](ClO₄)₂ (Cu₃₁, dppe = 1, 2-bis(diphenylphosphino)ethane) cluster [43]. Specifically, Cu(ClO₄)₂·6H₂O reacts with 4-methoxyphenylacetylene and dppe in the presence of triethylamine to form the precursor first, then Cu₃₁ nanoclusters were obtained upon the reduction of borohydride [43]. Interestingly, side product of [Cu₆(4-MeO-PhC≡C)₅(dppe)₃](ClO₄) (it is a Cu(I) cluster not a superatom) was also isolated from the obtained single crystals of the Cu₃₁ nanoclusters [44]. Recently, Fang, Wei and Shen et al. documented the fabrication of four Cu₅₀ clusters with nearly identical metal frameworks [45]. Using Cu₅₀(CF₃COO)₁₂(3, 5-diMe-PhS)₁₈(PPh₃)₄H₂ (Cu₅₀-1) as an example, Cu(CF₃COO)₂ was first prepared as the Cu source, and after it reacted with 3, 5-dimethylbenzenethiol and PPh₃, NaBH₄ was added as a reducing agent to obtain the raw cluster product [45].

Scheme 2.

Synthetic route of Cu₁₈ nanoclusters.

Reproduced with permission from Ref. [42], Copyright 2022 Royal Society of Chemistry

It is worth noting that, after forming the raw product by the precursor reduction method, adding some auxiliary ligand to etch the clusters may help to improve the yield and stability of the final product. For instance, Shen and coworkers recently reported a couple of “isostructural” Cu clusters of [Cu₄₁Cl₂(2-F-C₆H₄S)₁₂(CF₃COO)₆(PPh₃)₆H₁₉]²⁻ and [Cu₄₁(2, 5-di-Methyl-C₆H₃S)₁₂(BO₃)₃Cl₃(PPh₃)₆H₁₉] [46]. The raw product was formed by reacting Cu(CF₃COO)₂ with 2-fluorothiophenol and PPh₃ ligand, but after that, additional PPh₃Cl ligand was added to introduce Cl onto the two Cu₄₁ clusters for enhanced stability [46].

Gradient Reduction Strategy (GRS)

The gradient reduction strategy was first employed by the Zheng group to synthesize Cu₅₃ nanoclusters, but the method was not named then in 2019 [48]. Specifically, Cu(CF₃COO)₂ was first reduced by Cu powder, after the tert-butylacetylene (HC≡C^tBu) ligand was added, PhSiH₂ as the second reducing agent was introduced into the solution under vigorous stirring. It is believed that the cooperation between Cu powder and PhSiH₂ contributed together to achieve the high-nuclearity Cu(I)/Cu(0) cluster of [Cu₅₃(RCOO)₁₀(C≡C^tBu)₂₀Cl₂H₁₈]⁺ [48]. In another study, the synthesis and overall structure of Cu₅₃(C≡CPhPh)₉(dppp)₆Cl₃(NO₃)₉ nanocluster was recorded by Li and Zhang group. In a typical trial, 4-ethynylbiphenyl (BP) and 1, 3-bis(diphenylphosphino)propane (dppp) were sequentially added into an ethanol solution of Cu(NO₃)₂, then NaBH₄ was introduced, followed by aging the reaction to obtain the target cluster [37].

In 2020, the name of gradient reduction strategy was formally proposed by Sun group [47]. In this method, the valence of the Cu atom evolved from +2 to +1, then to 0 by using different reducing agents at different stages. As illustrated in Scheme 3, the comproportionation of Cu(II) and Cu powder can yield the Cu(I) complex first, then the Cu(I) complex was further reduced to generate [Cu₂₃(^tBuC≡C)₁₃(CF₃COO)₆] (SD/Cu23a) or [Cu₂₃(^tBuC≡C)₁₃(CF₃COO)₆]·CHCl₃ (SD/Cu23b). Such method showed the solvent-dependent polymorphism. It is worth noting that the Cu₂₃ nanoclusters are superatoms, not Cu(I) complex, and both of them have four valent electrons. They also contain a very rare [Cu₄]⁰ kernel surrounded by an outer Cu₁₉ shell. In addition, depending on the solvent, the Cu₂₃ nanocluster can crystallize into two polymorphs [47]. In 2022, the same group conducted the synthesis of two quasi-structurally isomeric 13-nuclei Cu nanoclusters (Cu13a and Cu13b) by the GRS method in a similar manner [49]. In this study, after the reaction of Cu(II) and CZ-PrAH (9-(prop-2-yn-1-yl)-9H-carbazole) with Cu powder to yield a yellow Cu(I) intermediate, the ligand was added into the mixture, following that a freshly prepared ethanol solution of NaBH₄ was added under vigorous stirring. The reducing agent of NaBH₄ with strong reducing capability was employed to avoid the re-oxidation of Cu(I) species [49]. Recently, Li et al. reported the comprehensive characterization and electrocatalytic CO₂ reduction of [Cu₂₆(DPPE)₃(CF₃CO₂)₈(CH₃O)₂(^tBuC≡C)₄H₁₁]⁺, which was also synthesized by this GRS approach [38].

Scheme 3.

Synthetic route of Cu23a and Cu23b nanoclusters by GRS.

Reproduced with permission from Ref. [47], Copyright 2020 American Chemical Society

One-Pot Synthesis (OPS)

Compared to the above two methods, one-pot synthesis is the most facile method to operate, as it does not require to form the Cu(I) precursor. Basically, all the reactants including the Cu salt, the ligand molecules, the reducing agent and the base are mixed to generate the Cu nanoclusters. Of course, the Cu(I) precursor is formed in the reaction system, but no isolation is necessary. It is known that such method has been employed for synthesizing quite a number of coinage metal nanoclusters, e.g., thiolate Au₂₅ nanoclusters can be easily prepared by one-pot approach [98, 99], the Tsukuda group fabricated a series of alkynyl-protected Au₂₂ clusters by this method [100], and the Wang group reported the preparation of Au₂₃(C≡CR)₁₅ through one-pot synthesis as well [101].

It is worth noting that making Cu superatom cluster is rather different from making Au and Ag superatoms, as Cu(I) is more resistant to reduction. Therefore, in the presence of hydride source, a reduction can easily yield Au or Ag superatoms [2, 102], but Cu(I)-hydride complex is usually formed. It is more difficult to observe Cu(0) clusters, mainly due to the higher stability of Cu(I)-hydride complex than Au(I)-hydride or Ag(I)-hydride complexes. However, in a ligand-deficient environment, the reduction of Cu(I)-hydride complex might generate some unstable (CuH)_x species that are amenable for cluster growth. In 2015, the Scott and Hayton group recorded the synthesis of [Cu₂₅H₂₂(PPh₃)₁₂]Cl and [Cu₁₈H₁₇(PPh₃)₁₀]Cl by this method [40]. As summarized in Scheme 4, adding 13 equiv. of Ph₂SiH₂ to a slurry containing 24 equiv. of Cu(OAc), 12 equiv. of PPh₃, and 1 equiv. of CuCl in C₆H₆ can result in a rapid color change from pale green to dark red then into deep green, concomitant with the precipitation of a dark brown solid after keeping stirring in 24 h. Two clusters of [Cu₂₅H₂₂(PPh₃)₁₂]Cl and [Cu₁₈H₁₇(PPh₃)₁₀]Cl were isolated with a yield of 23% and 14%, respectively [40]. Following this study, the same group accomplished the synthesis of [Cu₂₀(C≡CPh)₁₂(OAc)₆)] by the same approach [36].

Scheme 4.

Synthetic route of Cu₂₅ and Cu₂₈ nanoclusters.

Reproduced with permission from Ref. [40], Copyright 2015 American Chemical Society

Other Synthetic Methods

Besides the above approaches, researchers have developed several other ingenious methods to prepare atomically precise Cu nanoclusters. In 2016, the Hayton group reported a ligand-exchange-induced growth from a smaller cluster to synthesize Cu₂₉ nanoclusters [35]. As shown in Scheme 5, [Cu₂₅H₂₂(PPh₃)₁₂]Cl was first synthesized, and then the addition of 16 equiv of 4, 7-diphenyl-1, 10-phenanthroline (Ph₂phen) can cause an immediate color change from dark green to dark blue, and workup of this mixture for 15 min resulted in the isolation of [Cu₂₉Cl₄H₂₂(Ph₂phen)₁₂]Cl, a deep blue black crystalline material in a yield of 84%. Meanwhile, the reaction of [Cu₂₅H₂₂(PPh₃)₁₂]Cl with 1, 10-phenanthroline also yielded a deep blue molecule, which is probably the isostructural molecule of [Cu₂₉Cl₄H₂₂(Ph₂phen)₁₂]Cl, but no single crystal XRD data were available to confirm that [35].

Scheme 5.

Two cluster expansion synthetic routes for Cu₂₉ nanoclusters.

Reproduced with permission from Ref. [35], Copyright 2016 American Chemical Society

Note that the introduction of additional ligand does not necessarily lead to cluster growth or size expansion; instead, cluster size decrease is also possible. Chakrahari et al. reported the [Cu₁₃(S₂CNⁿBu₂)₆(C≡CR)₄](PF₆) (R=C(O)OMe, C₆H₄F) cluster which was synthesized through size transformation (Scheme 6) [103]. As the hydride in [Cu₂₈H₁₅(S₂CNⁿBu₂)₁₂]⁺ is substantially hydridic, the terminal alkyne is sufficiently acidic enough to react with it to yield [Cu₁₃(S₂CNⁿBu₂)₆(C≡CR)₄]⁺, which features a centered cuboctahedral [Cu₁₃]¹¹⁺ core with two free electrons, and [Cu₈H(S₂CNⁿBu₂)₆(C≡CR)₄]⁺ was also isolated as the side product [103].

Scheme 6.

Synthesis of [Cu₁₃(S₂CNⁿBu₂)₆(C≡CR)₄](PF₆) cluster.

Reproduced with permission from Ref. [103], Copyright 2016 Wiley VCH

Atomically Precise Cu Nanoclusters for Catalytic Applications

Thanks to the precise composition and structure of Cu nanoclusters, they can serve as ideal model catalysts for a variety of reactions [50]. Also, theoretical simulations can build precise models to reveal the reaction pathway and help to elucidate the reaction mechanism [104, 105]. In line with this, a great deal of research effort has been devoted to exploring the catalytic applications of atomically precise Cu nanoclusters, including but not limited to electrocatalytic reactions, photocatalytic reactions, organic reactions, as well as other catalytic processes (Fig. 2).

Fig. 2.

Catalytic applications of atomically precise Cu nanoclusters

Cu NCs for Electrocatalysis

The electrocatalysis technology has drawn widespread attention and found immerse values in a wide spectrum of fields especially in green energy storage and conversion, as electrocatalysis is not only highly efficient, mild condition operational, multifunctional, but also can couple with renewable intermittent energy sources such as solar, wind, and geothermal energy [106]. Therefore, electrocatalysis represents a sustainable strategy to address global environmental pollution and heavy reliance on fossil fuels [107].

In the last decade, the fast industrialization and huge demand for fossil fuels have caused a dramatic worldwide increase of CO₂ emission, which is the main culprit of global warming. However, CO₂ is an abundant source that can be used as substrate to fabricate various valuable products such formic acid, formamide, urea, and other hydrocarbon compounds [108, 109]. Electrochemical CO₂ reduction reaction (eCO₂RR) can realize such conversion with high efficiency and in an environmentally friendly manner; however, due to the chemical inertness and thermodynamic stability of linear CO₂ molecules, a prerequisite for achieving CO₂ reduction lies in finding a catalyst that can lower the chemical energy required to break the C-O bond [110, 111]. Moreover, eCO₂RR involves multiple electron transfer, and various products including C1 compounds (e.g., CO, HCO₂H, CH₃OH) and C2 compounds (e.g., CH₃CO₂H, CH₃CHO, C₂H₄, C₂H₆, etc.) can be generated. The thermodynamic potentials, various reactions, and corresponding products are summarized in Table 2 [112]. It can be noted that different reactions occur at very similar thermodynamic equilibrium potentials, plus the competing hydrogen evolution reaction (HER), it is extremely challenging to achieve highly selective CO₂ reduction product. Therefore, designing efficient, durable, and high-selectivity catalysts for eCO₂RR to acquire target product is highly desired [113].

Table 2.

Electrochemical CO₂ reduction reactions yielding various products with thermodynamic equilibrium potentials in an aqueous electrolyte at pH 7, 1 atm, and 25 °C, relative to the reversible hydrogen electrode (RHE)

Reactions	E0 (V vs. RHE)	Product
CO2 + 2H+  + 2e−  → CO (g) + H2O	− 0.11	Carbon monoxide
CO2 + 2H+  + 2e−  → HCOOH (aq)	− 0.12	Formic acid
CO2 + 6H+  + 6e−  → CH3OH (aq) + H2O	0.03	Methanol
CO2 + 8H+  + 8e−  → CH4 (g) + 2H2O	0.17	Methane
2CO2 + 10H+  + 10e−  → CH3CHO (aq) + 3H2O	0.06	Acetaldehyde
2CO2 + 12H+  + 12e−  → C2H4 (aq) + 4H2O	0.08	Ethylene
2CO2 + 12H+  + 12e−  → C2H5OH (aq) + 3H2O	0.09	Ethanol
2CO2 + 14H+  + 14e−  → C2H6 (aq) + 4H2O	0.14	Ethane
2CO2 + 16H+  + 16e−  → C2H5CHO (aq) + 5H2O	0.09	Propionaldehyde
3CO2 + 18H+  + 18e−  → C3H7OH (aq) + 5H2O	0.1	Propanol
2H+  + 2e−  → H2	0	Hydrogen evolution reaction
2H2O → O2 + 4H+  + 4e−	1.23	Oxygen evolution reaction

So far, all kinds of nanostructured materials including metal oxides, metal alloys, carbon substrates, two-dimensional nanomaterials have been investigated as catalysts for eCO₂RR [114–118]. Among a series of metal-based catalysts, Cu nanomaterials are one of the most promising catalysts that can deep electrochemically reduce CO₂ into valuable chemicals, especially hydrocarbon products [119–122]. However, for most Cu-based catalysts, the chemical nature is not uniform, e.g., no homogeneous size, shape, architecture, and structure hence it is extremely challenging to ascertain the chemical nature of the catalyst and eventually establish the structure–activity relationship [123–125]. Recently, molecular Cu nanoclusters have been drawing attention from the heterogeneous catalysis field, thanks to the high surface-to-volume ratio, strong binding capability to the key reaction intermediates, and more importantly, the atomically precise crystallographic structure and well-defined architecture [50, 90, 126].

In 2017, Tang et al. reported a structurally precise Cu–hydride nanocluster of Cu₃₂H₂₀L₁₂ (L is a dithiophosphate ligand), which offered unique selectivity for eCO₂RR at low overpotentials [51]. By density functional theory calculations, the authors first predicted that the presence of negatively charged hydride in Cu nanocluster plays a crucial role in determining the selectivity of the products, yielding HCO₂H over CO at lower overpotentials. The optimized structure of the Cu₃₂H₂₀L₁₂ cluster is shown in Fig. 3a, and the Cu₃₂ cluster has a distorted 14-Cu-atom formed hexacapped rhombohedral core sandwiched between two 9-Cu-atom formed triangular cupola fragments. Meanwhile, the 20 hydrides are divided into 12 triply coordinated H (µ₃-H), 6 tetra-coordinated H (µ₄-H), and 2 penta-coordinated H (µ₅-H). For CO₂ reduction into CO on the Cu₃₂H₂₀L₁₂ cluster, it adopts either the proton-reduction channel or lattice-hydride mechanism (Fig. 3b). It can be clearly noted that the lattice hydride pathway is more favorable, where the rate-determining step is the μ₃-H1 hydride being transferred to form the Cu₃₂H₁₉L₁₂-HCOO intermediate with a free energy change of 0.32 eV (Fig. 3b right). In contrast, the formation of Cu₃₂H₁₉L₁₂-HCOO in the proton reduction channel requires an energy gap of 1.08 eV (Fig. 3b left). Finally, the authors conducted the eCO₂RR test of the Cu₃₂H₂₀L₁₂ cluster to verify the theoretical predictions. As shown in Fig. 3c, the average current density becomes significant at the overpotential of 0.3 V and increases with the increasing of the overpotential. H₂, CO, and HCOOH are the main products, and the cumulative Faradaic efficiency (FE) is over 90%. The product selectivity is illustrated in Fig. 3d. HCO₂H is the predominant major product in the overpotential window from 0.3 to 0.4 V, and HER dominates at higher overpotentials (from 0.5 to 0.6 V) [51]. These experimental results confirmed the theoretical predictions, and this study showcases that the hydride-containing Cu nanoclusters may offer a unique product selectivity over conventional transition metal nanocatalysts for eCO₂RR.

Fig. 3.

a Total structure of Cu₃₂H₂₀L₁₂ nanocluster. Color codes: Cu (orange), S (yellow), hydride (green), H (white), P (purple). b CO₂ electroreduction on Cu₃₂H₂₀L₁₂ to form HCOOH via the proton-reduction channel (left) and the lattice-hydride channel (right). c Average current densities (black circles) and cumulative Faradaic efficiencies (stacked bars) obtained at different overpotentials. d Product selectivity for H₂, HCOOH, and CO at different overpotentials.

Reproduced with permission from Ref. [51], Copyright 2017 American Chemical Society

Furthermore, the core configuration of Cu nanoclusters can be manipulated to further mediate the activity and selectivity of eCO₂RR. In 2022, the Zang and Wang group reported three novel isomeric Cu₈-cluster with cores composed of two types of metal kernels (ditetrahedral vs. cubic) and studied the morphological kernel influence on the electrochemical eCO₂RR at the atomic level [52]. Specifically, three Cu clusters of Cu₈(H)-(L1)₆PF₆, Cu₈(^tBuS)₄(L1)₄, and Cu₈(^tBuS)₄(L2)₄ (Cu₈-1, Cu₈-2, and Cu₈-3, respectively, where L1 = 9H-carbazole-9-carbodithioate and L2 = O-ethyl carbonodithiolate) were prepared, and the total structures are illustrated in Fig. 4a–c. Cu₈-1 contains a slightly twisted cubic Cu₈⁸⁺ core, while Cu₈-2 and Cu₈-3 show the identical di-tetrahedral configurations. In eCO₂RR test, H₂, CO, and HCO₂H were the main products, and the total FE values were over 90% (Fig. 4d). Moreover, the di-tetrahedron-shaped Cu₈ clusters (Cu₈-2 and Cu₈-3) exhibited a higher FEHCOH₂ and greater selectivity than the cube-shaped Cu₈ cluster (Cu₈-1), where Cu₈-2 demonstrated remarkable HCO₂H selectivity, manifested by the FEHCO₂H values of 90% and 92% at −0.9 and −1.0 V, respectively (Fig. 4d). Meanwhile, in the long-term stability test of 8 h, only the current density from Cu₈-2 remained almost unchanged, indicating robust durability for prolonged operation (Fig. 4e). At last, the free energies of each step for Cu₈-1 and Cu₈-2 were calculated. Noteworthily, for the adsorbed *CO₂ being converted into HCOO*, the di-tetrahedron-shaped Cu₈-2 cluster (−0.91 eV) has a much lower free energy than the cube-shaped Cu₈-1 cluster (−0.23 eV). Since the rate determining step is the formation of COOH* to generate HCO₂H, the Cu₈-2 cluster is more favorable for HCO₂H production [52]. This investigation showcases the modulation of activity and selectivity toward CO₂ electroreduction by tailoring the Cu core, which can probably trigger more endeavors to design tailored Cu based catalysts for eCO₂RR.

Fig. 4.

The total structures of a Cu₈-1, b Cu₈-2, and c Cu₈-3. Color codes: Cu (brown), S (yellow), C (gray), O (red), N (blue). Electrocatalytic performances of d FE_HCOOH of Cu₈-1, Cu₈-2, and Cu₈-3 at different applied potentials. e Stability tests of the catalysts for eCO₂RR (Cu₈-1 and Cu₈-2 at -0.9 V, Cu₈-3 at −1.0 V).

Reproduced with permission from Ref. [52], Copyright 2022 Wiley VCH

It is worth noting that for most reported atomically precise Cu nanoclusters in the eCO₂RR, C1 products of CO and HCO₂H with high selectivity are normally obtained. This is probably due to the Cu active sites are coordination symmetric (typically CuS₃). Compared to the C1 products of CO/HCO₂H, the high energy density hydrocarbon products such as CH₄ and C₂H₄ are highly sought as potential fuels [127–130]. To switch the C1 products to more valuable hydrocarbons, Wu et al. recently reported a catalyst based Cu₆ nanocluster with symmetry-broken CuS₂N₁ active sites [27]. The total structure of the as-prepared Cu₆(MBD)₆ nanocluster (MBD = 2-mercaptobenzimidazole) is shown in Fig. 5a. Each MBD ligand has the tridentate sites to coordinate three Cu atoms, where the thiolate S atom binds with two Cu atoms and the N atom binds to the other Cu atom (Fig. 5b). Meanwhile, each Cu atom is coordinated by two S atoms and one N atom (Fig. 5c). It is symmetry-broken distorted CuS₂N₁ geometry, as the bonding length values are all different (Cu₁−S₁ = 2.2118 A, Cu₁−S_1a = 2.2924 A, Cu₁−N₁ = 1.9757 A). In addition, six Cu atoms formed a distorted octahedron with the Cu–Cu bonding ranging from 3.083 to 3.254 Å, which might be favorable for C–C coupling (Fig. 5d). There is sufficient space above the distorted CuS₂N₁ geometry, which can be accessible for CO₂ adsorption (Fig. 5e). When using Cu₆(MBD)₆ nanocluster as the catalyst, the product distribution is shown in Fig. 5f. When the potential goes from −0.7 to −1.4 V, the FE_CO and FE_H2 gradually decreased; meanwhile, the FEC₂H₄ and FEC₂H₄ gradually increased (Fig. 5f). The highest FE_CH4 reached 42.5% at −1.4 V with a large partial current density of −119.0 mA cm⁻², with the FEC₂H₄ of 23.0% and the partial current density of −64.4 mA cm⁻² (Fig. 5g). That is, in a wide potential window, CH₄ and C₂H₄ are the dominant products rather than CO. To gain more insights into the advantages of symmetry-breaking at the catalytic sites, in situ Cu K-edge XANES measurement of Cu₆(MBD)₆ was conducted, revealing that there is clear electron transfer from Cu₆(MBD)₆ to the CO₂ molecule. Thus, the Cu₆(MBD)₆ cluster with asymmetric CuS₂N₁ sites has a better CO₂ activation capacity than the Cu₈(^tBuS)₄(L2)₄ cluster with the symmetric CuS₃ sites. The integrated projected density of states (IPDOS) of Cu sites in Cu₆(MBD)₆ and Cu₈(^tBuS)₄(L2)₄ were then calculated. For Cu₆(MBD)₆, the d_x2-y2 orbital of Cu-S₂N₁ site has the highest number of electrons close to the Fermi energy level (Fig. 5h), whereas the highest d orbital is Cu d_xz in CuS₃ sites of Cu₈(^tBuS)₄(L2)₄ (Fig. 5i). It suggests that the coordination-symmetry breaking significantly affects the highest occupied d-orbital, which plays a critical role in regulating the coordination mode of the reactant and intermediate. When the C atom of CO₂ is adsorbed onto the Cu₆(MBD)₆, the highest occupied d_x2-y2 orbital can energy-match with the lowest occupied π* orbital of CO₂ to form a π-complex rather than the conventional σ-complex, where the π-complex can lower the energy barrier of CO₂ activation. If the O atom of CO₂ is adsorbed onto the Cu₆(MBD)₆, a σ-complex is formed, and for Cu₈(^tBuS)₄(L2)₄, a π-complex is formed only when the O atom in CO₂ is the adsorbed atom. Consequently, due to superiority of the C atom adsorbing on Cu₆(MBD)₆, the hydrogenation on the less hindered O atom to form the key *COOH intermediate rather than the *OCHO intermediate is more favorable, making the reaction proceed to yield the CH₄/C₂H₄ products [27]. This study highlights the importance of regulating the coordination mode of Cu nanoclusters, which might be a promising direction for designing metal nanoclusters as highly efficient CO₂ reduction electrocatalysts toward valuable hydrocarbon production.

Fig. 5.

a–e Structural analysis of Cu₆(MBD)₆ nanocluster. Color codes: pink, Cu; blue, N; yellow, S; deep gray, C; light gray, H. f FEs of CH₄, C₂H₄, CO and H₂ at different potential on Cu₆(MBD)₆. g Partial current densities of CH₄, C₂H₄, and CO on Cu₆(MBD)₆. IPDOS of 3d orbitals: d-density of states vs. the Fermi level projected onto Cu atoms in (h) Cu₆(MBD)₆ and (i) Cu₈(^tBuS)₄(L2)₄.

Reproduced with permission from Ref. [27], Copyright 2023 Wiley VCH

The large surface area, good catalytic activity and strong capability to suppress HER also enable Cu nanoclusters as potent catalysts for electrochemical nitrate reduction reaction (eNO₃RR) [131, 132], a process that can turn nitrate contaminant into valuable NH₃ product. Specifically, NH₃ is an important feedstock for preparing nitrogen-containing fertilizer, pharmaceuticals, chemicals, and agricultural products, and it is also a valuable green energy carrier [133, 134]. However, the current global production of NH₃ is largely relying on the energy-intensive and large CO₂-emissive Haber–Bosch process. The electrochemical N₂ reduction powered by renewable energy offers a promising approach but it is significantly restricted by the low solubility of N₂ in water and the extremely high energy required to break the N≡N bond (941 kJ mol⁻¹). In stark contrast, nitrate is one of the major contaminants in wastewater and also has much weaker N–O bonding energy (250 kJ mol⁻¹). Therefore, developing efficient and stable electrocatalysts for NO₃RR is imperative for offering a sustainable strategy to produce NH₃ [135–138]. Recently, atomically precise metal nanoclusters have demonstrated great potential for catalyzing eNO₃RR, mainly thanks to the strong catalytic capability, and more importantly, the atomic precise structure can provide an ideal model to elucidate the complicated reaction pathway and establish the structure–activity correlation [131, 132, 139]. For instance, our group found that Ag₃₀Pd₄(C₆H₉)₂₆](BPh₄)₂ nanocluster can achieve the Faradaic efficiency of NH₃ over 90% in eNO₃RR, where the Ag site is responsible for converting NO₃⁻ into NO₂⁻ and the Pd site makes the major contribution to catalyze NO₂⁻ into NH₃ hence the whole reaction adopted a tandem catalytic mechanism [93].

Recently, Wang and Zang et al. utilized bulky carborane-alkynyl ligand to prepare atom-precise monomer Cu₁₃·3PF₆ and bridged dimer Cu₂₆·4PF₆ clusters, and both clusters exhibited remarkable catalytic activity and selectivity in eNO₃RR [53]. As illustrated in Fig. 6a, Cu₁₃·3PF₆ has a metal skeleton, which could be viewed as two pentagonal bipyramids merged by sharing one equatorial edge but with one vertex of the pentagonal bipyramid being lost. Cu₂₆·4PF₆ is a dimer of Cu₁₃·3PF₆. To be specific, in Cu₂₆·4PF₆, each Cu₁₃ monomer loses a carbocycloalkynyl ligand and the adjacent PPh₃, and the vacated space is occupied by the nido-carboranealkynyl ligands unit of the other monomer via cyclopentadienyl anionic coordination (Fig. 6b). Interestingly, both clusters have fair stability and accessible open metal sites for eNO₃RR, where eNO₃RR was carried out in an H-type cell containing 0.1 M KNO₃ in 0.5 M K₂SO₄ medium (Fig. 6c). When the applied potential goes more negatively, the FENH₃ exhibited a volcano-shape change, and both clusters reached the maximal value at -0.85 V (vs. RHE). The maximal FENH₃ of Cu₂₆·4PF₆ is 85.1%, much higher than that of Cu₁₃·3PF₆. Meanwhile, at each measured potential, the FE and yield rate of NH₃ for Cu₂₆·4PF₆ surpass that of Cu₁₃·3PF₆ (Fig. 6d, e), despite the NH₃ yield rate increasing with the increasing of the applied potential (Fig. 6e). Finally, the authors elucidated the reaction pathway and mechanism with the aid of theoretical simulations and in-situ FTIR spectroscopic study. The free energy diagram in Fig. 6f suggests that Cu₁₃·3PF₆ and Cu₂₆·4PF₆ share the identical reaction pathway (*NO₃ → *NO₂ → *NO → *NHO → *NH₂O → *NH₂OH → *NH₂ → *NH₃). For both clusters, the rate determining step (RDS) is identified as *NHO-to-*NH₂O, and the energy consumption for Cu₂₆·4PF₆ is much lower than that of Cu₁₃·3PF₆ (0.27 vs. 0.58 eV), indicating Cu₂₆·4PF₆ is more energetically favorable for eNO₃RR. Moreover, the subsequent *NH₂O-to-*NH₃ reduction step is also a downhill process, demonstrating a strong interaction between the key intermediates and the active sites, which can promote the electron transfer and the following hydrogenation steps to accelerate the eNO₃RR process. This study not only establishes a platform to disclose the structure–activity relationship of Cu nanoclusters for eNO₃RR, but also provides a feasible strategy to access desirable Cu nanoclusters as efficient eNO₃RR catalysts [53].

Fig. 6.

Structure of a Cu₁₃·3PF₆ and b Cu₂₆·4PF₆·c Schematic illustration of NO₃RR. d Comparison of FE_NH3 at various potentials. e Potential dependent NH₃ yield rate for Cu₁₃·3PF₆ and Cu₂₆·4PF₆. f Corresponding adsorption configurations of the reaction intermediates and Gibbs free energy in eNO₃RR. Cu, orange; F, yellow; P, violet; B, green; C, gray; O, red; N, blue; H, white. Reproduced with permission with Ref [53].

Copyright 2024 American Association for the Advancement of Science

Cu NCs for Photocatalysis

Solar energy is an inexhaustible natural energy source, and photocatalytic technology can utilize solar energy to realize environmental decontamination and energy conversion, particularly, photocatalytic technology can reduce CO₂ into valuable hydrocarbon fuels [140, 141]. The principle of photocatalytic reduction of CO₂ is to use photoexcited semiconductor photocatalysts to produce photogenerated electrons and holes, and the carriers take redox reactions on the surface of the catalyst, reducing CO₂ into useful chemical raw materials [142, 143]. Compared with other methods, the photocatalytic reduction approach has some unique merits. On one hand, photocatalytic method directly uses solar energy for CO₂ reduction and transformation, which is safe and pollution-free. On the other hand, the device for photocatalytic reduction is quite simple, and it can operate at very mild conditions, so the whole process is economically feasible [144–147].

Among a variety of photocatalysts, the artificial photosynthesis system using inexhaustible solar energy to simultaneously reduce CO₂ and oxidize H₂O to produce valuable chemicals has been attracting more and more research attention [148, 149]. The current widely reported photocatalytic system usually comprises noble-metal-containing photosensitizers and/or organic dyes as electron donors, which are sophisticated and synthetically challenging to be available [150]. Therefore, it is highly desired to prepare a single component photocatalyst with suitable optical band and catalytic active center, that is similar to natural photocatalytic system, which can directly convert CO₂ and H₂O into value-added chemicals.

Metal nanoclusters, an emerging type of organic–inorganic hybrid material, composed of an organic layer and metal core, are promising candidates as a single component catalyst to realize the above goal [151, 152]. The ultrasmall size, well-defined configuration, and most importantly, the atomically precise structure can provide a perfect platform to probe the atomical level structure–performance understanding regarding the photocatalysis mechanism [153, 154].

Recently, the Zang group reported a stable Cu–S–N cluster photocatalyst with local protonated N–H groups, and such cluster can achieve ~ 100% selectivity for CO evolution under visible light. Specifically, two clusters of Cu₆-NH and Cu₆-N were prepared [54]. As shown in Fig. 7a, b, Cu₆-NH cluster possesses a distorted Cu octahedron, and one Cu atom coordinates with one N atom and two S atoms from three ligands. Such configuration is quite identical with the previously reported Cu₆-N cluster (Fig. 7c, d), except that the ligands in Cu₆-N are fully deprotonated, despite the two clusters having the same metal kernel (Fig. 7e). Both clusters have considerable absorbance capacities in the wavelength region of 400–550 nm, and comparable band gap energies (2.36 and 2.39 eV for Cu₆-N and Cu₆-NH, respectively). However, their visible-light-driven photoreduction capability of CO₂ showed some significant difference. Both clusters exhibited exceptionally high selectivity of CO with over 99%, but the CO evolution rate is much higher for Cu₆-NH than that of Cu₆-N (Fig. 7f). Furthermore, the Cu₆-NH photocatalyst demonstrated superior catalytic stability, as no detectable change is observed in the XRD patterns before and after the photocatalytic test (Fig. 7g). Finally, they conducted DFT calculations to elucidate the underlying physical origin, and found that, in the rate-determining step of forming *COOH, the Cu₆-NH cluster has a much lower energy barrier. This is due to the structural difference, that is, the presence of deprotonated pyrimidine N in Cu₆-N and protonated pyrimidine N in Cu₆-NH (Fig. 7h). The protonated pyrimidine N atoms in Cu₆-NH acted as a proton relay station to provide a local proton, hence facilitating the proton coupling process, leading to enhanced photocatalytic efficiency [54]. This study highlights the great promise of using Cu nanoclusters as photosensitive semiconductors for photoreduction of CO₂.

Fig. 7.

Overall structures of the a Cu₆-NH and d Cu₆-N clusters. Coordination modes in b Cu₆–NH and c Cu₆–N. e Cu₆ framework; The molecule packing diagrams of f CO₂–CO photoreduction performances of Cu₆–NH and Cu₆–N. g PXRD patterns of Cu₆–NH before and after 24 h photocatalytic reaction. h Different angles showing the Cu₆ NC structures and the coordination environment around the Cu sites for Cu₆-N and Cu₆-NH. Color codes: brown, Cu; yellow, S; gray, C; white, H; blue, N.

Reproduced with permission from Ref. [54], Copyright 2023 Wiley–VCH

The precise structure of Cu nanoclusters allows the in-depth study of structure–property relationships in photocatalysis at the atomical level. Moreover, isomeric Cu nanoclusters can make the comparative structure–property relationship study feasible, as they have minimal structural difference. Isomeric Cu nanoclusters usually originate from the chiral isomerism, the structural isomerism, but the quasi-structural isomerism has been ignored for a long time. The quasi-structural isomeric Cu nanocluster usually displays comparative structural features, e.g., core geometric configuration, surface spatial arrangement.

In 2022, the Sun group reported two quasi-structurally isomeric Cu₁₃ nanoclusters with highly similar kernel and different spatial arrangements of peripheral ligands (Fig. 8a) [49]. Both Cu13a and Cu13b clusters have a highly similar Cu₁₃ kernel but different degrees of distortion (Fig. 8b, c). The formation of these two Cu cluster isomers is governed by the exotic chlorine ion, and also due to the charge transfer from Cl⁻ to Cu core, Cu13a presented lower superoxide radical (O₂^•−) yield and higher singlet oxygen (¹O₂) compared to that of Cu13b. The conduction band minimum (CBM) is −1.20 and −1.26 V for Cu13a and Cu13b, respectively, so through the band gap value, the valence band maximum (VBM) is calculated as 0.74 and 0.66 V for Cu13a and Cu13b, respectively (Fig. 8d). More negative CBM value drives Cu13b more energetically feasible for O₂^•−, while for Cu13a, the larger band gap favors the exchange of an electron with ³O₂, also more efficient for ¹O₂ generation (Fig. 8e). Consequently, the two clusters demonstrated different performance particularly in selectivity in sulfides oxidation into sulfoxides. As summarized in Fig. 8f, despite methyl sulfide can be transformed into dimethyl sulfoxide by two clusters with nearly identical high conversion rates (near unity), Cu13a showed a certainly higher degree of enhancement than that of Cu13b [49]. This work provides a new pathway for boosting the photocatalytic selectivity of Cu nanoclusters.

Fig. 8.

a Schematic demonstration of the quasi-isomerization. b, c Side views of the molecular structures of Cu13a and Cu13b. d Schematic diagram for the band structures of Cu13a and Cu13b. e Mechanism of Cu13 cluster and ³O₂ for ¹O₂ and O₂^•− photogeneration. f Oxidation of sulfides catalyzed by Cu₁₃ nanoclusters. Color labels: red and purple, Cu; cyan, Na; gray, C; red, O; yellow, S; blue, N; green, Cl.

Reproduced with permission from Ref. [49], Copyright 2022 American Chemical Society

In modern synthesis, bond-forming reactions are critical for generating new products, particularly, bond-forming reactions at mild conditions can find versatile applications in fine chemical synthesis, new drug discovery, and sustainable production of the industrial precursor. To that end, visible light photocatalysis is more advantageous as compared to thermal catalysis, however, developing high-efficiency, stable catalyst is the key [155]. Cu-based photocatalysis has shown great promise as an inexpensive and attractive method, compared to the expensive metal-complex and less stable organic dyes [156, 157]. Nevertheless, the cross-coupling reactions of aryl bromides and iodides are relatively easy to realize, but the cross-coupling reactions of aryl chloride remained extremely challenging.

In 2022, the Bakr and Rueping team reported a [Cu₆₁(S^tBu)₂₆S₆Cl₆H₁₄] nanocluster (Cu₆₁NC)-based catalyst that can enable C-N bond-forming reaction of aryl chlorides under visible-light irradiation at ambient conditions [55]. As shown in Fig. 9a, the pronounced molecular ion is observed, and the experimental pattern agrees well with the simulated pattern. Figure 9b illustrates that it has two characteristic bands, and the crystal structure can be found inset. In the presence of base, at room temperature, the Cu₆₁ NC cluster can catalyze the Ullman reaction for het(aryl) very well (Fig. 9c). The authors also proposed a plausible mechanism (Fig. 9d). Specifically, Cu NC reacts with the base forming CuNC-amine complex A first, then blue LED irradiation leads to the formation of complex B. After a SET process to give an electron to aryl chloride, the oxidized CuNC-Nu complex, a halide anion, and an arene radical were generated, evidenced by the fluorescence and quenching lifetime tests. After that, the radical attacks the CuNC-Nu complex, delivering the C–N bond formation and recovering the Cu₆₁ clusters [55]. This study introduces atomically precise Cu clusters as a new family of photocatalysts for bond-forming reactions under mild conditions.

Fig. 9.

ESI MS spectrum of Cu₆₁ NC. The molecular ion peak is at m/z of 6613.5. Insets: a Comparison of experimental and simulated isotopic patterns. b UV–Vis absorbance spectrum showing two bands at 355 and 440 nm. c Cu₆₁ NC catalyzed Ullmann C–N Coupling. d Proposed reaction mechanism.

Reproduced with permission from Ref. [55], Copyright 2022 American Chemical Society

Cu NCs for Organic Reactions

Click Reaction

The perfect monodisperse and atomically precise nature of molecular metal nanoclusters can facilitate the development of detailed structure–activity relationships [158–160]. Using these atomically precise metal nanoclusters as catalysts can uncover new insights, but some of them may not be suited for this purpose. Specifically, a vast majority of these metal nanoclusters are protected by a shell of thiolate ligands [1, 152]. These ligands can impart the clusters significant thermal and chemical stability, but they can also block some active sites and hence must be partially removed before catalysis can occur [161]. Meanwhile, the strong bonding strength of the metal-sulfur bond makes it hard to remove completely [161]. However, the harsh reaction conditions often lead to the significant structure or nuclearity change of the metal nanoclusters, making the structure–activity relationship unreliable. To address this issue, the researchers switched to other ligands, which do not require harsh pretreatment for activation. To that end, alkynyl molecules, phosphine ligand, halogen, hydride, and their mixed combination have been widely utilized to prepare atomically precise metal nanoclusters including Cu nanoclusters [56, 72, 73, 162].

In 2018, the Hayton group reported the synthesis of [Cu₂₀(C≡CPh)₁₂(OAc)₆] cluster and its catalytic application in Click reaction [36]. Figure 10a shows the crystal structure of the Cu₂₀ cluster, which has 4 THF molecules incorporated into the crystal lattice as solvates. It has a tetrahedral [Cu₄]²⁺ core (Fig. 10b), which is encapsulated by a [Cu₁₆(C≡CPh)₁₂(OAc)₆]²⁻ shell (Fig. 10c). There are 12 acetylide ligands in the [Cu₁₆(C≡CPh)₁₂(OAc)₆]²⁻ shell distributed by 4 [cyclo-Cu(C≡CPh)]₃ units, which are located at the vertices of a tetrahedron, meanwhile, there are six acetate ligands situated at the edges of the tetrahedron (Fig. 10c). Then the Cu₂₀ cluster was immobilized on dry, partially dehydroxylated silica, and the supported cluster catalyst displayed excellent performance toward the Click reaction of cycloadditions between benzyl azide and terminal alkynes. As demonstrated in Fig. 10d, the catalyst is effective for various terminal alkynes without harsh pre-treatment and even with a rather low cluster loading of 0.05 mol%. More importantly, X-ray absorption near edge structure (XANES) test after the catalytic test shows that the cluster undergoes no significant nuclearity changes under click reaction conditions [36]. It also indicates that the Cu species mainly kept as Cu(0) and Cu(I) in the cluster during the reaction process, where the Cu(I) species is well known to be the active site for click reaction. Great promises of using atomically precise Cu nanoclusters to examine the structure/activity relationship have been demonstrated in this study, especially for the non-thiolate protected Cu nanoclusters.

Fig. 10.

a [Cu₂₀(C≡CPh)₁₂(OAc)₆]·C₄H₈O. b Tetrahedral Cu₄²⁺ core (blue) and face-capping Cu atoms (green). c Ligand binding modes. d Cycloadditions catalyzed by Cu₂₀ nanoclusters. Color legend: Cu = blue, green; C = gray; O = red.

Reproduced with permission from Ref. [36], Copyright 2018 American Chemical Society

In nanocatalytic regime, elucidating single-atom effect of the nanocatalyst is interesting yet extremely challenging, as one single-atom modification can cause appreciable change to the overall particle’s structure, not to mention that, the size polydispersity and ambiguous surface structure of the metal nanoparticle [163]. Nevertheless, atomically precise metal nanoclusters have created great opportunities to tailor the chemical structure at the atomic level.

Recently, the Bakr and Rueping team reported the synthesis of [Cu₅₈H₂₀PET₃₆(PPh₃)₄]²⁺ (Cu₅₈) and its analogue of [Cu₅₇H₂₀PET₃₆(PPh₃)₄]⁺ (Cu₅₇) with one surface Cu atom removed, and Cu₅₇ showed much higher catalytic activity toward [3 + 2] azide-alkyne cycloaddition (AAC, click reaction) than that of Cu₅₈ [57]. As illustrated in Fig. 11a, Cu₅₈ has a multishell architecture, and the metal skeleton possesses five concentric shells of Cu₈@Cu₆@Cu₂₄@Cu₁₂@Cu₈. Meanwhile, the ligand layer in Cu₅₈ consists of a P₄ tetrahedron, an S₁₂ icosahedron, and an S₂₄ truncated cube. Cu₅₇ has nearly the identical metal kernel structure except one Cu atom was removed on shell-4. Subsequently, the two nanoclusters were employed as catalyst for the click reaction. With the 0.025 mol% of Cu₅₇ or Cu₅₈ stoichiometric ratio, and in the presence of 1.5 h light irradiation, the product yields of 97% and 77% were achieved for Cu₅₇ and Cu₅₈, respectively (Fig. 11b). In addition, both clusters showed good catalytic stability, as no obvious product yield decline was observed after three catalytic cycles (Fig. 11c). Overall, Cu₅₇ performed more efficiently than Cu₅₈ in the click reaction, and one single atom removal made great difference in the nanoclusters’ catalytic activity [57]. This work may open a new avenue for the nanoparticles’ catalytic design by targeted isostructural single-atom manipulation.

Fig. 11.

a Crystal structure and dissection of Cu₅₈ and Cu₅₇. b Cycloaddition between phenylacetylene and benzyl azide using Cu₅₈ and Cu₅₇ as catalyst under visible light. c Time-dependent product yields of Cu₅₈-catalyzed AAC under visible light. Brown and blue: Cu; yellow: S; pink: P.

Reproduced with permission from Ref. [57], Copyright 2023 Wiley VCH

C–C Coupling Reaction

The C–C bond formation is a widely employed transformation in natural product synthesis, pharmaceutical synthesis, medicinal chemistry, and preparation of functional materials. So far, various catalytic methods have been developed for this important cross-coupling reaction. Among that, Sonogashira cross-coupling reaction is one of the most important and widespread used sp²–sp² carbon–carbon bond formation reactions in organic synthesis, which is usually catalyzed by Pd and other transition metals [157, 164]. It can realize the coupling of aryl or vinyl halides with terminal acetylenes, but most of the catalysts suffer from that the undesired alkyne homocoupling side reaction is extremely difficult to suppress.

In 2023, the Rueping and Bakr group reported the [Cu₂₈H₁₀(C₇H₇S)₁₈(TPP)₃] cluster (Cu₂₈; C₇H₇S: o-thiocresol; TPP: triphenylphosphine) with a defined defect for Sonogashira C–C coupling reaction. The skeleton structure of Cu₂₈ is shown in Fig. 12a, and it features a centered Cu₁₃ anti-cuboctahedron core, like the Cu₂₅, Ag₁₉, and Ag₂₅ nanoclusters, and Cu₂₈ also has a unique cage-like Cu₁₅S₁₈P₃ shell. Interestingly, Cu₂₈ is a defective nanocluster which contains a surface vacancy, as one Cu atom is missing in one tetrahedron’s vertex, and this vertex is expected to connect with a phosphine ligand. Subsequently, Cu₂₈ was examined as a catalyst for Sonogashira C–C coupling reaction. Upon UV-LEDs irradiation, the desired C–C coupling product with a high yield of 82% was acquired without the byproduct of homocoupling reaction (Fig. 12b) [58]. The authors then conducted the mechanistic control studies and proposed the tentative reaction mechanism (Fig. 12c). Upon blue-light irradiation, Cu₂₈ was photoexcited (A), and it can induce a single electron transfer (SET) process with aryl iodides to yield the Ar-radical (C) and oxidized Cu₂₈⁺ (B). The Ar-radical then attacks the alkyne C–C triple bond to generate vinylic type C-radical intermediate (D). Another single electron transfer process to B can regenerate the Cu₂₈ photocatalyst. Note that the C–C triple bond can be activated by Cu₂₈ through forming a Cu NC-π-alkynyl type complex hence easily being attacked by the Ar-radical. The authors detected a little stilbene-type byproduct in GC–MS measurement generated by a hydrogen atom transfer (HAT) process. Finally, the base-induced deprotonation of vinyl cation E leads to the desired Sonogashira C–C coupled product [58]. Upon the photoexcitation of blue-LEDs, the oxidation state of the Cu₂₈ cluster undergoes a transient transition from 0 to I, form [Cu₂₈]⁺ NC, while after the second single electron transfer (SET) process, it returns to the ground state of Cu₂₈.The results demonstrate that Cu₂₈ is an active photocatalyst enables the SET process with aryl iodides to produce Ar-radical which attacks the C–C triple bond, totally preventing the formation of homocoupling product. This investigation paves a way to acquire Cu NCs with defined surface defects as active sites, and more importantly, these defective Cu NCs can serve as model systems to provide profound understanding on defect effect for heterogeneous catalysis.

Fig. 12.

a Structure of Cu₂₈ nanocluster. b Sonogashira C–C coupling reaction. c Proposed reaction mechanism.

Reproduced with permission from Ref. [58], Copyright 2023 Wiley VCH

Recently, Biswas et al. reported a hydride-free [Cu₇(SC₅H₉)₇(PPh₃)₃] (Cu₇) nanocluster (NC), which demonstrated remarkable specificity in a photoinduced C–C coupling reaction [59]. The Cu₇ NC has a metal core with 4 Cu atoms forming a tetrahedron (Fig. 13a), and three Cu atoms are connected with the bottom face of the tetrahedron with another three Cu atoms (Fig. 13b). The metal–ligand binding configurations are shown in Fig. 13c, d, where the thiolate ligands mainly adopt the bridging modes, and three P atoms connected with three Cu atoms directly. Interestingly, the Cu₇ NC exhibits intriguing photoluminescence (PL). Upon exciting the cluster at 395 nm, a robust PL emission is observed at room temperature with an emission peak occurring at 436 nm. Such outstanding photophysical properties inspired the authors to explore the Cu₇ NC as catalyst for photoinduced C–C bond formation reaction. Under purple light irradiation from LEDs, the Cu₇ NC (5 mol% Cu) was able to catalyze the C_sp2-C_sp coupling reaction between phenyl iodide and phenyl acetylene to generate the desired product with high yield (Fig. 13e). With the aid of control experiments and DFT calculations, the authors also proposed tentative reaction mechanism (Fig. 13f). Initially, the Cu active site in Cu₇ NC binds to alkyne to form an intermediate of Int-I in the presence of MeONa. Subsequently, Int-I is photoexcited to generate Int-I*, which engages the oxidative addition step to yield the intermediate of Int-II. The following reductive elimination of Int-II produces the final product, along with the release of the Cu₇ NC. In this mechanistic rationale cycle, the oxidative addition step is identified as the rate-determining step [59]. This work provides some mechanistic insights of using atomically precise Cu nanoclusters for C–C coupling reactions, and underscore the great potential of hydride-free Cu nanoclusters for catalysis.

Fig. 13.

a Cu₄ head. b Three Cu atoms connected with the Cu₄ head. c S-Cu binding configuration and d P-Cu binding configuration in the Cu₇ core. e Reaction optimization of C_sp2–C_sp cross-coupling in the presence of Cu₇ NC and f proposed reaction mechanism. Cu, brown; S, yellow; and P, magenta. Reproduced with permission [59], Copyright 2024 American Chemical Society

A³ Coupling Reaction

To forge new C–C and C–N bonds, three-component dehydrogenative coupling reaction represents a new practical methodology. Yet, to realize all-in-one three-component dehydrogenative coupling in a single catalytic system remains very challenging [165]. For instance, to fulfill the efficient synthesis of propargylamines including C1-propargylamines, Cu compounds are the most widely recognized catalysts to implement the A³ coupling reactions or redox-A³ coupling reactions, but the reaction conditions are rather harsh [166]. Therefore, a Cu-based catalytic system with high efficiency and regio-selectivity with broad substrate scope working under mild conditions is highly desirable.

Recently, the Zang group reported a tridentate N-heterocyclic carbene ligand stabilized [Cu₃(NHC)₃(PF₆)₃] (Cu₃NC) cluster for both A³ coupling reactions and redox-A³ coupling reactions [60]. The crystal structure of the NHC ligand and Cu₃NC are shown in Fig. 14a, b, respectively, where the NHC ligand endows the Cu₃NC with dual attributes of flexibility and rigidity. On one hand, the stable Cu–C and Cu–N bonds between the NHC ligand shell and the metal core favors the stability of the Cu₃NC. On the other hand, the pyridine of the N‑heterocyclic carbene has somewhat dynamic balance between the aliphatic amines and the pyridine to protect the catalytic centers and prevent the Cu₃NC deactivation. Such dynamic balance endows Cu₃NC with flexible features. Consequently, high activity and high regioselectivity can be achieved in all-in-one A³ coupling reactions with inert substrates at room temperature (Fig. 14c), as manifested by 71 examples of A³ coupling reactions and 14 examples of redox A³ coupling reaction, both up to 99% yield [60]. The following mechanistic and control experimental tests demonstrated that the remarkable catalytic performance originates from the flexible and rigid dual attributes of the Cu₃NC.

Fig. 14.

a Structure of the N-heterocyclic carbene (NHC) ligand. b Total structure of the Cu₃NC. c Cu₃NC catalyzed both the A³ coupling reaction and the redox-A³ coupling reaction. R₁, R₂, R₃ and R represent kinds of functional group.

Reproduced with permission from Ref. [160], Copyright 2023 Nature Publishing Group

It is worth noting that most ligand protected Cu nanoclusters feature a core–shell framework hence the active sites can be blocked by metal–ligand shell or surface passivation units. Therefore, a strategy that can largely expose the metal active sites is long pursued for catalytic coupling reaction. In 2023, Xu et al. reported a novel approach to construct array-based ([Cu₈(Tf-dpf)₄(NO₃)₂](NO₃)₂ clusters [61]. Compared with the ([Cu₈(RS)₆(PPh₃)₄(MeCN)₄H]⁺ with the cubic core, the above array based Cu₈ cluster displays largely uncoordinated metal sites, hence exhibited significantly enhanced catalytic activity in the “aldehyde-acetylene-amine” A³-coupling reaction for synthesizing propargylamines. It shows the great potential to fabricate array-based Cu nanoclusters for A³ coupling reaction and beyond [61].

Synthesis for Indolizines

To realize the practical catalytic application of Cu nanoclusters, the long-term stability remains a long-standing challenge. In the regime of coinage metal nanocluster, compared with Au, Ag, and Au/Ag alloy clusters, the study of Cu nanoclusters lags far behind, mainly due to the high reactivity of Cu and low standard reduction potential (E⁰_Cu2+/Cu+  = 0.34 V, E⁰_Cu+/Cu = 0.52 V) [22]. The surface organic ligand such as thiolate molecules, alkynes, halides, hydride can improve the stability due to the formation of strong chemical bonds, but the long-term stability for some specific organic reactions are still unsatisfactory. These ligands are considered as soft base, and Cu²⁺ or Cu⁺ ions belong to soft acid; therefore, the ligand–metal interaction fits the Pearson’s hard/soft acid/base theory [167].

In a recent study conducted by the Li and Ni team, pyrazolate (denote as Pz) ligands were employed to synthesize Cu₈ nanoclusters [62]. Pz is a soft base, and thanks to the high pK_a value of 19.8 for deprotonated pyrazole [168], Pz protected Cu clusters are expected to exhibit exceptional alkali-resistant capability. As determined from SCXRD (Fig. 15a), the flexible bis-pyrazole ligands (H2L1) can transform into two novel deprotonated bis-pyrazole ligands (L2 and L3), and all these L2 and L3 ligands can assemble with Cu⁺ to yield Cu₈ nanoclusters. The crystal of Cu₈ clusters can keep intact in various organic solvents, highly concentrated acid, saturated alkali, oxidant, and boiling water, and still are fine for SCXRD test. More impressively, the introduction of organic acid or base (100 eq. HOAc, or 400 eq. dibutylamine) cannot impose structural decomposition, neither, confirming the remarkable stability never documented for Cu nanoclusters (Fig. 15b). Subsequently, the synthesis for indolizines from 2-pyridinecarboxaldehyde derivatives, terminal alkynes and secondary amines was tested using Cu₈ nanocluster as a catalyst. Interestingly, the reaction cannot proceed without Cu₈ cluster as the catalyst (Fig. 15c). The optimized condition is 0.005% mol of substrates. Other Cu nanoclusters such as Cu₁₀ or Cu₁₈ had a much lower yield, as they are not stable in basic conditions (Fig. 15d). Consequently, the turnover frequency (TOF) of Cu₈ nanoclusters can reach 3880 h⁻¹, about 3 orders of magnitude larger than the reported catalysts toward the synthesis of indolizines [62]. This study provides a new strategy to synthesize pyrazolate-protected Cu nanoclusters with ultrahigh chemical stability for practical applications.

Reduction of Ferricyanide to Ferrocyanide

Ferricyanide ([Fe(CN)₆]³⁻) is an environmentally unfriendly material that readily reacts with acids to produce toxic hydrogen cyanide gas. The catalytic reduction of [Fe(CN)₆]³⁻ to ferrocyanide ([Fe(CN)₆]⁴⁻) is one effective strategy to limit the toxicity of [Fe(CN)₆]³⁻ [169].

In a recent study, the Luo and Sun team reported the Cu₁₈H(PET)₁₄(PPh₃)₆(isothiocyanate)₃ (Cu₁₈H in short) nanocluster [63], which exhibits excellent catalytic performance in the reduction of ferricyanide. Cu₁₈H comprises a pseudo D3-symmetrical triple-stranded helical Cu₁₅ kernel, and can be structurally described as layer-by-layer combination of multiple chiral Cu nanoclusters linked through copper-thiolate bonds. Remarkably, the aggregate state of the Cu₁₈H nanoclusters can catalyze electron transfer reactions efficiently. To assess its catalytic capability, ultraviolet/visible absorption spectroscopy was used to monitor the reduction process. When Cu₁₈H and NaBH₄ were added simultaneously, [Fe(CN)₆]³⁻ can be completely converted into [Fe(CN)₆]⁴⁻ very rapidly, and the solution turned into colorless within 14 s. In a control test, the reduction reaction of [Fe(CN)₆]³⁻ without Cu₁₈H required 16 min. It indicates that Cu₁₈H has good catalytic reactivity for the reduction of [Fe(CN)₆]³⁻ [63]. This study makes an interesting model for investigating and elucidating the aggregated state of copper nanoclusters.

Hydrogenation of Cyclohexanone

The chemical hydrogenation of cyclohexanone is among the most important and prevalent transformations in industrial organic synthesis, thanks to its effectiveness and economic viability [170, 171]. A previous report suggested that copper hydride nanoclusters are very promising for catalyzing the hydrogenation of carbonyl compounds, despite that there are challenges to control the product selectivity [172].

In a recent work done by Sun et al., a cluster of [Cu₆₆Cl₈(PPh₃)₈(SC₂H₅)₃₂H₂₄](SbF₆)₂ (Cu₆₆ in short) was synthesized and its catalytic performance toward the hydrogenation of cyclohexanone was examined [64]. Cu₆₆ contains an orderly of 16 Cu₄ squares, and the ligands coordinate to the surface of the cluster in a regiospecific manner, displaying square pattern as well. To enhance the monodispersity and stability of the cluster catalyst, the Cu₆₆ nanoclusters were immobilized on carbon black (XC-72) to afford the Cu₆₆/XC-72 catalyst, and it demonstrated an unprecedented performance toward the hydrogenation of cyclohexanone. The selectivity of cyclohexanol is nearly 100%. Under specific conditions, complete conversion of cyclohexanone to the desired product was achieved within 40 h. The authors also tested all the other Cu catalysts, including (Cu(CF₃CO₂)₂, CuCl, Cu₅₄Cl₁₂(NO₃)₁₂(SC₄H₉)₂₀S (Cu₅₄), Cu₅₀(PhCOO)₁₀(4-F-PhS)₂₀(PPh₃)H₂ (Cu₅₀), SeCu₂₀(PhSe)₁₂(PPh₃)₂(C₆H₅COO)₆ (Cu₂₀), and [Cu₂₅(SPhCl₂)₁₈H₁₀]³⁻ (Cu₂₅), but they showed significantly lower yields for the corresponding product. Moreover, the Cu₆₆/XC-72 catalyst exhibited no significant decay in cyclohexanol product selectivity, highlighting the robustness of the Cu₆₆/XC-72 catalyst [64]. This study showcases Cu nanoclusters have high stability and exceptional catalytic activity in the hydrogenation of carbonyl compounds, envisioning a bright future in both fundamental research and practical applications.

Cu NCs for Other Reactions

Atomically precise Cu nanoclusters can also find versatile catalytic applications toward other chemical reactions. One typical reaction is the catalytic hydrogenation of nitroarene. It is worth noting that the catalytic hydrogenation of toxic nitroaromatics into non-toxic amino molecules is a quite efficient and economical method, as amino molecules are valuable intermediates for pharmaceuticals and dyes [169, 173, 174]. Coinage metal nanoclusters have been emerging as a new type of important hydrogenation catalysts, and the catalytic hydrogenation of 4-nitrophenol into 4-aminophenol in the presence of BH₄⁻ reducing agent has been employed as a model reaction in several previous studies [32, 92, 95, 175].

In 2023, Luo et al. reported an intrinsically chiral Cu hydride-rich nanocluster of [Cu₅₇H₂₀(PET)₃₆(TPP)₄]⁺ (Cu₅₇H₂₀) and its catalytic activity toward reduction of 4-nitrophenol [56]. Single crystal X-ray diffraction measurement shows that the unit cell of the Cu₅₇H₂₀ cluster contains a pair of enantiomers, C-Cu₅₇H₂₀ and A-Cu₅₇H₂₀ (C, clockwise; A, anti-clockwise) (Fig. 16a). The hydride positions were further validated by DFT optimization of the model cluster of (Cu₅₇H₂₀)_opt. Interestingly, the DFT analysis also suggests that the interfacial μ₃-H in (Cu₅₇H₂₀)_opt can be transformed into an interstitial μ5-H species in the perfect cluster of (Cu₅₇H₂₀)_opt as long as the vacant position is recovered by the 58th Cu atom (Fig. 16b). The Cu₅₇H₂₀ cluster was then employed as the catalyst to catalyze the reduction of 4-nitrophenol. The reaction process was monitored by UV–vis absorption spectra (Fig. 16c), where strong absorbance peak at 400 nm from 4-nitrophenol decreased rapidly in 20 min and the absorbance peak at 300 nm from 4-aminophenol increased. The bright yellow solution also turned colorless. Using the absorbance change at 400 nm as metric, the reduction adopts a pseudo-first-order reaction kinetics, and the reaction rate was calculated as 18 min⁻¹ (Fig. 16d). A control experiment in the absence of Cu₅₇H₂₀ was also conducted, and no reduction occurs on 4-nitrophenol when adding BH₄⁻, probably due to the high activation barrier between the repelling 4-nitrophenol anion and BH₄⁻ [176]. Finally, the authors performed a 4-nitrophenol reduction test by using BD₄⁻ to check the accessibility of the hydride in Cu₅₇H₂₀. After the catalytic test, the cluster was examined by the ESI–MS measurement (Fig. 16e). It shows well defined peak feature of [Cu₅₇H₁₉D(PET)₃₆(TPP)₄ + Cu]²⁺, indicating one H⁻ is indeed replaced by one D⁻ of BD₄⁻, that means, the exposed interfacial μ₃-H might be involved in the catalytic reaction [56]. By using defective Cu₅₇H₂₀ cluster as model catalyst, this work offers atom-precise insights into the vacant defect role in catalysis.

Fig. 16.

a Total structures of the C–Cu₅₇H₂₀ and A-Cu₅₇H₂₀ isomers. b Transformation of a μ₃-H in (Cu₅₈H₂₀)_opt into a μ₅-H in (Cu₅₈H₂₀)opt. c Absorption spectra versus time. d Plot of ln(C_t/C₀) versus time for 4-nitrophenol reduction. e ESI–MS spectrum after catalytic reaction with NaBD₄. Color labels: Cu, turquoise; C, gray; P, orange; S, yellow; red, H; yellow, S.

Reproduced with permission from Ref. [56], Copyright 2023 Wiley–VCH

Fig. 15.

a Schematic illustration for preparing pyrazolate-protected Cu₈ nanocluster. b Radar chart of various factors for the guidance and assessment of chemical stability of Cu₈ and other Cu-NCs. c TOF comparison for the indolizine synthesis based on Cu₈ and other catalysts. d Indolizines synthesis catalyzed by Cu₈ nanoclusters.

Reproduced with permission from Ref. [62], Copyright 2023 Wiley VC

In addition, a very small change in composition can make a big difference in structure and catalytic performance for some specific catalytic reactions. In a recent report, the Wang and Jiang team disclosed the dramatic difference between Cu₂₀H₈ and Cu₂₀H₉ clusters in catalysis [65]. Specifically, two copper hydride clusters, [Cu₂₀H₉(Tf-dpf)₁₀]·BF₄ (Cu₂₀H₉) and [Cu₂₀H₈(Tf-dpf)₁₀]·(BF₄)₂ (Cu₂₀H₈) (Tf-dpf = N, N′-di(5-trifluoromethyl-2-pyridyl)formamidinate) were fabricated, and both have twenty Cu atoms and ten amidinate ligands but with one hydride difference in the ligand shell. Such differences led to drastically different geometric and electronic structures, resulting in different catalytic properties. Cu₂₀H₈ showed 25 times higher catalytic activity than Cu₂₀H₉ in the conjugate reduction of cinnamaldehyde, mainly due to the easier dissociation process of a Tf-dpf⁻ ligand in Cu₂₀H₈ [65]. This work highlights the sensitivity of structure and composition toward catalytic reactions by using atomically precise copper hydride nanoclusters as catalysts.

Recently, the Shen group found that the open metal site plays a critical role in the catalytic reduction of 4-nitrophenol by using two Cu₄₁ nanoclusters as model catalysts [46]. Specifically, two isostructural Cu nanoclusters of [Cu₄₁Cl₂(2-FC₆H₄S)₁₂(CF₃COO)₆(PPh₃)₆H₁₉]²⁻ (1H) and [Cu₄₁(2,5-di-Methyl-C₆H₃S)₁₂(BO₃)₃Cl₃(PPh₃)₆H₁₉] (2H) were prepared and fully characterized, and the main structural difference of them lies in the absence or presence of two additional chlorides on the surface, which in turn heavily governs the exposure of metal sites. The 2H cluster with more open active sites exhibited ~ sixfold increase of rate constant in the catalytic reduction of 4-nitrophenol than 1H cluster [46]. This work manifests that atomically precise Cu nanoclusters can serve as catalyst models to directly visualize the active sites that drive the chemical transformation. Meanwhile, the same group also discovered that the small distinctions in two carboxylate-protected Cu₂₀ clusters can cause distinct catalytic performance toward 4-nitrophenol reduction [66]. Two clusters Se@Cu₂₀(PhSe)₁₂(PPh₃)₂(C₆H₅COO)₆ (Cu₂₀-1) and Se@Cu₂₀(PhSe)₁₂(PPh₃)₂(CF₃COO)₆ (Cu₂₀-2) share with the identical metal skeleton and similar ligand distributions, and the main difference is the carboxylate ligand: C₆H₅COO⁻ for Cu₂₀-1 while CF₃COO⁻ for Cu₂₀-2. Consequently, Cu₂₀-1 has a catalytic activity enhancement of 16-fold than Cu₂₀-2. Such catalytic performance distinction is attributed to the carboxylate effect, that is, the functional group (-C₆H₅CO₂ or -CF₃CO₂) altered the electronic structure of the Cu nanocluster [66].

Challenges and Perspectives

The last decade, especially the recent five years, have witnessed great success in atomically precise Cu nanoclusters, including synthesis, structural analysis, property exploration, and various catalytic applications. Despite the significant progress, there are still some important obstacles for nourishing the atomically precise Cu nanocluster field, especially in both synthetic regime and catalytic field. Some possible challenges and future perspectives are discussed below, aiming to motivate or advance this fast-growing field:

1. High-efficiency synthetic method to obtain atomically precise Cu nanoclusters is still highly desired. Compared with the fruitful achievements in Au and Ag nanoclusters, the synthesis and total structure determination of Cu nanoclusters have been lagging behind, probably due to the variability of Cu valence states. It is known that the subtle reaction conditions may affect the output of the final product, e.g., the solvent, the reaction temperature, the ligand (discuss below), the reducing agent, the pH value in the reaction system, the stoichiometric ratio of all the reactants, etc. For example, by using CH₂Cl₂ as the solvent, the Cl atoms can coordinate with Ag and become the capping ligand in alkynyl-protected Ag₁₁₂ nanoclusters [177], and it might also be applicable for preparing Cu nanoclusters. More importantly, the current synthetic methods of atomically precise Cu nanoclusters are heavily dependent on the “the-trial-and-error” efforts, which are quite tedious, expensive, and low efficient. High throughput synthesis with the aid of artificial intelligence or machine learning technique represents an embryonic avenue yet still have a long way to go [178, 179].

2. The role of ligand should be carefully considered. The choosing of ligand molecule not only affects the synthesis, but also plays a critical role in the catalytic process. Thiolate, alkynyl molecule, hydride, halogen, phosphine, N-heterocyclic carbene, inorganic anions (e.g., BF₄^-, CH₃COO^-) are most widely employed ligand molecules for stabilizing Cu nanoclusters, but they possess different binding strength to the Cu core. It is commonly believed, the surface ligand can block some active sites for catalysis [161], as they can inhibit the reactants for accessing the metal sites. However, more and more cases have shown that, the ligand can facilitate the catalytic process, e.g., accelerate the transfer of the reactants or intermediates via weak interactions. Weak surface ligand is more easily stripped off or exchanged by the reactant to create catalytic sites [10]. Moreover, the type of ligand molecule affects the catalytic activity drastically, e.g., Tsukuda group reported that alkynyl-protected Au₂₅ clusters exhibited markedly higher HER activity than thiolate-protected Au₂₅ molecules [180], while our group discovered that, the Faradaic efficiency of CO for alkynyl-protected Ag₃₂ clusters is much higher than the thiolate and phosphine ligand co-protected Ag₃₂ clusters in eCO₂RR [32]. Such catalytic property difference is attributed to the electronic structure change owing to the electronic perturbation of the π-conjugated units [72]. Even with the same type of molecule in ligand, the steric hindrance and electronic structure can affect the catalytic properties, e.g., using bulky ligand may create some low coordinated or undercoordinated metal sites for catalysis [159, 181]. The Xie group found that, in the oxygen evolution reaction catalyzed by three thiolates protected Au₂₅ nanoclusters, p-mercaptobenzoic acid stabilized Au₂₅ nanoclusters exhibited markedly superior catalytic performance, simply because the ligand’s stronger electron-withdrawing ability can create more partial positive charges on Au(I) as active site for facilitating feasible adsorption of OH^- in alkaline media [182]. Yoo et al. found that the locally induced hydrophobicity by bulky alkyl functionality near the surface of the Ag₂₅(SR)₁₈ cluster dramatically enhanced the eCO₂RR activity, where the hydrophobic Ag₂₅ cluster exhibited remarkable selectivity for CO (FE_CO > 90%) and achieved a high current density of up to -240 mA cm^-2 with excellent durability lasting for over 120 h [183]. Nevertheless, the ligand role of atomically precise Cu nanoclusters during the synthetic process and catalytic duration remains largely to be explored.

3. The metal core tailoring. The metal core configuration and atomic spatial arrangement can be critical for exposing available catalytic active sites for Cu nanoclusters. So far, there are core-shell, rod-like, array-like core structure and other core configurations for Cu nanoclusters, but a trade-off between activity, selectivity and stability must be taken into account for specific catalytic reactions [69]. More profoundly, more dedicated core structure engineering might be necessary for Cu nanoclusters. Vacancy engineering has demonstrated great potential, which is expected to catch more future research attention [58]. Another important direction is alloying other metals to form atomically precise Cu–alloy nanoclusters. For catalytic reactions, the size, composition, configuration, electronic structure are the main factors that influence the catalytic performance, and introducing another metal to form Cu–alloy can modulate all the above factors. Cu-based bimetallic nanoclusters have demonstrated unusual properties and extraordinary catalytic capabilities [90, 184], yet controlling the number and the exact position of a foreign metal atom remains challenging [163, 185]. Particularly, single-metal-atom doping have been gaining tremendous efforts in catalysis study but increasing the production yield for generating single-atom doping during the synthetic process, understanding the geometric and electronic structure change in the on-working catalytic process is essential for future high-performance catalyst design [186–189].

4. In situ reaction mechanism study. Understanding the reaction mechanism in on-working status is quite valuable but still challenging. The in situ and operando spectroscopic and synchronous techniques such as surface enhanced Raman scattering, infrared spectroscopy, high-energy X-ray diffraction (HE-XRD), extended X-ray absorption fine structure (EXAFS), X-ray absorption near edge structure (XANES), and small-angle neutron scattering (SANS) can provide the possibility of acquiring the structure of the catalytic intermediates, which can enhance the evidence for proposed reaction mechanism [190–193]. For instance, the in situ infrared spectroscopy was employed to capture the reaction intermediates during the electrocatalytic nitrate reduction catalyzed by Ag₃₀Pd₄ nanoclusters; hence, the tandem catalytic mechanism was successfully revealed by our group [93]. Nevertheless, the consecutive picture of the whole reaction process including adsorption, transfer, desorption of the reaction intermediates on Cu nanoclusters are quite limited in current studies, and more advanced in situ/operando study and the necessary instrumental modification for some specific in situ/operando investigations are expected.

5. The more precise theoretical simulations. Theoretical simulations can calculate the energy barrier of each step, acquire the optimal configuration between the cluster and the reactant, the intermediate, or the product, hence plays a critical role in speculating the active sites and deciding the rate-determining step [104]. However, the accuracy of these calculations depends on the authenticity and the extent of the structure proximity between the real crystal structure and the theoretical model. The fully ligand capped Cu nanoclusters are too complicated to build a real model, so the “ligand simplification” strategy is widely used. For instance, Au₂₅(SCH₃)₁₈ has been used to replace Au₂₅(SC₂H₄Ph)₁₈, Au₂₅(SC₆H₁₃)₁₈ and Au₂₅(SC₁₂H₂₅)₁₈ in several cases for building theoretical models [194–196]. Authentic models can help to understand the true reaction process and gain more realistic reaction mechanism but given the complexity of the cluster structures and the reaction conditions, more precise and realistic theoretical simulations still have a long way to go.

Author Contributions

MC contributed to the investigation and original draft writing. CG, LQ, LW, LQ, and KC were involved in the literature search and formal analysis. ZT performed the funding acquisition and draft finalization.

Declarations

Conflict of interest

The authors declare no conflict of interest. They have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.