Atomically precise core-shell nanoclusters have become a promising material in catalysis, biomedicine, and chemical sensing due to the unique quantum confinement effect resulting in optical properties (Jin et al., 2016; Yao et al., 2018; Xu et al., 2019; Jin R. et al., 2021; Sun et al., 2021; Zheng et al., 2021). The studies on correlation between the properties and structures of cluster compounds based on the determined crystal structures show that the core and shell structures have different effects on the performance of the cluster compounds, and modifications on the core and shell structures may induce variations on clusters properties (AbdulHalim et al., 2014; Chakraborty and Pradeep, 2017; Khatun et al., 2018; Yan et al., 2018; Jin Y. et al., 2021). The Pt core-doped nanocluster PtAu₂₄(SC₆H₁₃)₁₈ exhibits higher hydrogen production than that of Au₂₅ (Kwak et al., 2017), and the dopant AuAg₂₄ shows stronger fluorescence performance (Bootharaju et al., 2016). Surface shell dopant Au₂₄Cu₆ exhibited superior catalytic activity compared to other homometallic and Au-Cu alloy nanoclusters (Chai at al., 2019). Therefore, alloying could serve as an efficient approach to tailor the properties of nanoclusters for more applications (Ghosh et al., 2018; Jin et al., 2018a; Wang et al., 2018; Dias and Leite, 2019).

Current alloy research mainly focuses on bimetallic clusters, and there are few studies on trimetallic clusters due to factors such as synthesis, characterization, and crystallization, etc. (Kang et al., 2016; Sharma et al., 2016; Yan et al., 2016; Hossain, et al., 2018; Kang et al., 2019a; Kang et al., 2019b; Kang et al., 2020) When the third metal is doped into the bimetallic alloy clusters, what site will it occupy and what effect will it have on the overall performance?Recently, for the active metal Cu doping, several surface Cu-doped nanoclusters such as Au₁₃Cu_x (x = 2, 4, 8) (Yang et al., 2013), Cu_xAu_25-x (Yang et al., 2017), Cu₃Au₃₄ (Yang et al., 2017), Ag₂₈Cu₁₂ (Yan et al., 2016), Ag₃₀Cu₁₄ (Li at al., 2020) and Cu-internal-doped nanoclusters like Ag₆₁Cu₃₀ have been observed and well-determined by x-ray crystallography (Zou et al., 2020). Specifically, the outer Au shells always are partially alloyed by the incorporated Cu heteroatoms for Au-based nanoclusters, while core-shell alloy nanoclusters with a shell-by-shell configuration could be generated for Ag-based nanoclusters. However, for the Au-Ag alloy nanocluster, how will the copper atoms choose the sites?

Herein, we use position-determined alloy clusters [Au₉Ag₁₂(SAdm)₄(Dppm)₆Cl₆](SbF₆)₃ as templates for the doping of the third metal copper (Jin et al., 2018b). The crystallography analysis suggested that the four Cu atoms priority replace the position of the silver capped by Dppm ligand in the motif for [Au₉Ag₈Cu₄(SAdm)₄(Dppm)₆Cl₆](SbF₆)₃ (Scheme 1). And the Cu doping affected the electronicstructure, resulting in the difference of optical properties in CD spectra, DPV spectra and so on. This provides a good observation method for understanding the doping position.

The synthesized [Au₉Ag₁₂(SAdm)₄(Dppm)₆Cl₆](SbF₆)₃ based on the reported method was determined by ESI-MS and X-ray crystallography. The next is the regulation of surface structure of [Au₉Ag₁₂(SAdm)₄(Dppm)₆Cl₆](SbF₆)₃ with Cu^I(SAdm) complex precursor. As shown in Figure 1A, the [Au₉Ag₁₂(SAdm)₄(Dppm)₆Cl₆]³⁺ shows main peaks at 322, 365, 427, 480 and 670 nm, respectively, and [Au₉Ag₈Cu₄(SAdm)₄(Dppm)₆Cl₆]³⁺ shows 322, 366, 427, 482 and 710 nm, respectively. In contrast, most of the peaks for both nanoclusters have not changed significantly, except for the red shift of the 670 nm peak to 710 nm. The binding energy of Cu_2p from XPS data confirmed the Cu doping in the [Au₉Ag₈Cu₄(SAdm)₄(Dppm)₆Cl₆]³⁺ (Figure 1B), and energy level positions of other elements have basically not changed (Supplementary Figure S1). The peak at m/z 2084.85 corresponds to the 3 + charge of [Au₉Ag₁₂(SAdm)₄(Dppm)₆Cl₆] and can be perfectly assigned by the calculated result (m/z 2084.81) (Figure 1C). And peak at m/z 2026.20 corresponds to the 3 + charge of [Au₉Ag₈Cu₄(SAdm)₄(Dppm)₆Cl₆] (Cal. 2026.18) (Figure 1D). Meanwhile, the H-NMR and ¹H–¹H COSY spectra of [Au₉Ag₁₂(SAdm)₄(Dppm)₆Cl₆](SbF₆)₃ and Au₉Ag₈Cu₄(SAdm)₄(Dppm)₆Cl₆](SbF₆)₃ nanoclusters were performed, showing that the overall chemical environment is weakly affected by copper doping regulation (Supplementary Figures S2, 3).

Furthermore, in order to have a deep understanding of the regulation process, the time-dependent UV-Vis spectra and ESI mass spectra of [Au₉Ag₁₂(SAdm)₄(Dppm)₆Cl₆](SbF₆)₃ in CH₂Cl₂ after adding Cu^I(SAdm) complex precursor were performed. As shown in Figure 2A, with the increase of time of Cu^I(SAdm) complex precursor. adding, the peak centered at 427 nm always maintained, and the peak centered at 480 only 2 nm redshifts. While the 670 nm peak gradually red shift to 710 nm, with a redshift value of 40 nm. ESI mass spectra suggested the copper atoms are gradually replacing silver atoms, which leads to red shift (Figure 2B). The successful determination of [Au₉Ag₈Cu₄(SAdm)₄(Dppm)₆Cl₆]³⁺ structure allowed us to know the site of doping clearly.

As shown in Figure 3, the overall structure of [Au₉Ag₁₂(SAdm)₄(Dppm)₆Cl₆](SbF₆)₃ and [Au₉Ag₈Cu₄(SAdm)₄(Dppm)₆Cl₆](SbF₆)₃ are basically the same: firstly, five gold atoms and eight silver atoms constitute the icosahedron, then the Au₅Ag₈ icosahedron and four gold atoms constitute the Au₄@Ag₈Au₅ metallic kernel. The Au₄@Ag₈Au₅ is first capped by four Dppm ligands and two Cl ligands, forming Au₄@Ag₈Au₅(Dppm)₄Cl₂ framework. After the Au₄@Ag₈Au₅(Dppm)₄Cl₂ is further protected by two peripheral structures DppmAg₂Cl₂(SR)₂, the Au₉Ag₁₂ was obtained. By contrast, Au₅Ag₈@Au₄@Cu₄ is obtained when four copper atoms doped the position of the silver of peripheral structures DppmAg₂Cl₂(SR)₂. Meanwhile, the copper doping has little effect on the bond length and angle of the icosahedron metal core (Supplementary Figure S4). Based on the doping sites of copper atoms, we realize that the Au₄@Ag₈Au₅ will be a stable metal core. In the packing model, the difference of arrangement can be observed clearly, and it is worth mentioning that the doping can affect the arrangement of clusters in the unit cell (Figures 3C,D) from a crystal engineering point of view.

As reported, the chirality of metal clusters mainly come from chiral metalcore, the arrangement of chiral ligands and local chiral patterns on an achiral surface (Zeng and Jin, 2017). The chirality of [Au₉Ag₁₂(SAdm)₄(Dppm)₆Cl₆]³⁺ comes from the chiral Au₄@Ag₈Au₅ metallic kernel. After doping, the cluster will have a different CD spectrum compared to the parent compound. Importantly, herein, the Cu dopants also have some impacts on the chiral properties. As shown in Figure 4, the CD spectra of [Au₉Ag₁₂(SAdm)₄(Dppm)₆Cl₆]³⁺ reveal multiple CD-active peaks at 325, 363, 428 and 483 nm, respectively, and some weak peaks. While the CD spectra of [Au₉Ag₈Cu₄(SAdm)₄(Dppm)₆Cl₆]³⁺ shows peaks at 340, 373, 442, and 493 nm, respectively. The Au₃₈ cluster with Pd atoms leads to core-doped Pd₂Au₃₆(SC₂H₄Ph)₂₄. Comparison between the CD spectra of Au₃₈(SC₂H₄Ph)₂₄ and Pd₂Au₃₆(SC₂H₄Ph)₂₄ shows significant differences, revealing core-doping has strong impacts on the electronic structure of the cluster (Barrabés et al., 2014). The comparison between the CD spectra of [Au₉Ag₁₂(SAdm)₄(Dppm)₆Cl₆]³⁺ and [Au₉Ag₈Cu₄(SAdm)₄(Dppm)₆Cl₆]³⁺shows that all the peaks from [Au₉Ag₈Cu₄(SAdm)₄(Dppm)₆Cl₆]³⁺ have redshift, different from the differences between Au₃₈(SC₂H₄Ph)₂₄ and Pd₂Au₃₆(SC₂H₄Ph)₂₄. The doping location may have different impacts on the CD spectra.

In addition to the CD spectra, the electronic structures of [Au₉Ag₁₂(SAdm)₄(Dppm)₆Cl₆](SbF₆)₃ and [Au₉Ag₈Cu₄(SAdm)₄(Dppm)₆Cl₆](SbF₆)₃ are investigated by optical and electrochemical spectroscopies. Differential pulse voltammetry (DPV) of Au₉Ag₁₂ and Au₉Ag₈Cu₄ are carried out. The scan direction was detected from +1.6 to -1.6 V and then back from −1.6 to +1.6 V. As shown in Figure 5, The HOMO-LUMO gaps of Au₉Ag₁₂ and Au₉Ag₈Cu₄ are determined as 1.54 and 1.44 eV, respectively. For the differential pulse voltammetry (DPV) curves, there is a reduction peak at −1.32 V (R1) and two oxidation peaks at 0.40 V(O1) and 0.58 V (O2) for Au₉Ag₁₂, while there are two reduction peaks at −0.82 V (R1) and −1.01 (R2) and one oxidation peaks at 0.83 V (O1) for Au₉Ag₈Cu₄. So, the electrochemical energy gap is 1.72 eV for Au₉Ag₁₂ and 1.65 eV for Au₉Ag₈Cu₄. The HOMO-LUMO gaps calculated from DPV are consistent with those derived from the optical absorption spectra. So, the regulation of surface structure via Cu alloying changes the electronic structure, thereby affecting the electrochemical properties. Besides, the [Au₉Ag₁₂(Sadm)₄(Dppm)₆Cl₆](SbF₆)₃ in CH₂Cl₂ solution shows non-fluorescence, while the Cu dopant [Au₉Ag₈Cu₄(SAdm)₄(Dppm)₆Cl₆](SbF₆)₃ in CH₂Cl₂ solution shows weak fluorescence at 638 nm, once again verifying the changes in the electronic structure. (Supplementary Figure S5).

The [Au₉Ag₁₂(SAdm)₄(Dppm)₆Cl₆](SbF₆)₃ and [Au₉Ag₈Cu₄(SAdm)₄(Dppm)₆Cl₆](SbF₆)₃ show good stability in an ambient environment (Figures 6A,D) and the stability tests (i.e., under oxidizing/reducing environments) for Au₉Ag₁₂ and Au₉Ag₈Cu₄ are also performed to explore the effects of copper dopants on the stability of nanoclusters. Under the oxidizing environment (by mixing 200 μL of H₂O₂ (50%) with 6 mg of cluster in 10 ml of CH₂Cl₂), the Au₉Ag₁₂ can stabilize for several hours (Figures 6B,E), and the peaks of the UV-vis spectra are obvious. However, the Au₉Ag₈Cu₄ decompose quickly to form complexes within several mins (Figures 6C,F). This difference may be because the peripheral copper atom is easier to be oxidized. Meanwhile, the copper doping has an impact on the properties of clusters on reducing environment (by mixing the 10 ml CH₂Cl₂ solvent of 6 mg of cluster with 200 μL of EtOH solvent of 1 mg of NaBH₄). The UV-vis of Au₉Ag₁₂ changes quickly until there are no obvious peaks within 30s. And the UV-vis of Au₉Ag₈Cu₄ also changes quickly, but still some peaks can be observed within 60 min. These indicate the regulation of surface structure affects the stability of nanoclusters.

Intercluster reactions between Au₉Ag₁₂ and Au₉Ag₈Cu₄ (Abs._671nm = 0.3 for Au₉Ag₁₂ and Abs._712nm = 0.3 for Au₉Ag₈Cu₄, respectively) are performed (Zhang et al., 2016; Khatun et al., 2020; Neumaier at al., 2021). As shown in Figure 7, the reaction was completed quickly (1 min), similar to the UV-vis spectrum that prolongs the reaction for 3 h. As shown in the Figures 7A,B, intercluster reactions produce a spectrum with 428, 482 and 702 nm, respectively. Learned from the Figures 7C,D, the products are Au₉Ag₈Cu₄, Au₉Ag₉Cu₃, Au₉Ag₁₀Cu₂, Au₉Ag₁₁Cu₁,respectively. Theoretical and experimental isotopic distributions of them matched perfectly as shown in Supplementary Figures S6, 7. This indicates the copper migration between Au₉Ag₁₂ and Au₉Ag₈Cu₄ upon mixing in solution, similar to silver migration between Au₃₈(SC₂H₄Ph)₂₄ and doped Ag_xAu_38-x(SC₂H₄Ph)₂₄ nanoclusters (Zhang et al., 2016).

In conclusion, the regulation of surface structure of [Au₉Ag₁₂(SAdm)₄(Dppm)₆Cl₆](SbF₆)₃ nanocluster via alloying produced an trimetallic nanocluster formulated as [Au₉Ag₈Cu₄(SAdm)₄(Dppm)₆Cl₆](SbF₆)₃. X-ray crystallography identifies that the Cu dopants prioritily replace the position of the silver of peripheral structures DppmAg₂Cl₂(SR)₂. This controlled target metal exchange method may be extendable to other sized nanoclusters capped by multiple-ligands. Meanwhile the regulation of surface structure affected the CD spectra, DPV spectra, and stability. The [Au₉Ag₁₂(SAdm)₄(Dppm)₆Cl₆](SbF₆)₃ and [Au₉Ag₈Cu₄(SAdm)₄(Dppm)₆Cl₆](SbF₆)₃ contribute to understanding of the structure-optical property relationship deeply.