Introduction

Nanoparticles play a central role in the rapidly growing nanoscience and nanotechnology fields, with a wide range of applications being developed including nanocatalysis, sensing, optical, and biology1,2,3,4,5. Atomic level understanding of nanoparticle structure is of great importance in order to establish definitive structure—property relationships6, thereby facilitating systematic tailoring of material properties and developing of various applications of nanoparticles1,2,3. In this regard, atomically precise gold nanoparticles have attracted great interest in recent years for both fundamental research and technological applications1. Recent success in the synthesis of atomically well-defined nanoparticles6,7,8,9,10 has offered exciting opportunities to pursue fundamental understanding of the stability11,12,13, isomerism14, optical15,16,17,18, chiroptical3, catalytic2, 19, 20, and magnetic21,22,23 properties of Au nanoparticles.

Single-atom doping has gained significant interest for its potential to design novel bi-functional heterogeneous catalysts with superior or new properties compared to the homo-gold counterparts1, 24. For example, it has been demonstrated that a single atom of Pd, Pt, Cd, and Hg can be successfully doped into gold nanoparticles not only to enhance the stability of the nanoparticle but also to tune the catalytic and optical properties of the nanoparticle25,26,27,28,29. It is worth noting that the reported single-atom doped/alloyed gold nanoparticles are mainly limited to heterometals from a different group of elements rather than in the same group as gold (i.e., Cu, Ag). For example, a work done by Copley et al.10 shows that reaction between [Au11(PMePh2)10]+ and [MCl(PMePh2)] (M = Ag or Cu) results in the formation of a nanocluster with multiple heterometals, i.e., [Au9M4Cl4(PMePh2)8]+. This reaction is believed to occur through intermediate cations containing different numbers of metal dopants, i.e., [Au11M2Cl2(PMePh2)10]3+ and [Au10M3Cl3(PMePh2)9]2+ 10. Another interesting finding by Bakr and co-workers30 is that the single Pd atom in the Pd1Ag24 nanocluster could be replaced by a gold atom, resulting in single gold atom-doped Au1Ag24. Despite many efforts, preparation of gold nanoparticles doped with a single Cu or Ag atom still remains challenging due to the similar configuration of outmost electrons (d 10 s 1) of Cu and Ag as that of Au. This similarity leads to easy formation of a distribution of Cu or Ag dopants in the alloy nanoparticles31,32,33,34,35.

Although the similarity in electronic structure of Ag and Cu with Au (d 10 s 1) poses a major challenge for single-atom doping of gold nanoparticles, we rationalize that a single atom of Ag or Cu should easily fill into a vacancy if the latter is pre-formed within the gold nanoparticle. This method may be able to circumvent the limitation from the similar electron configuration of the same group metals. In terms of hollow gold nanoparticles, Das et al.36 reported a centrally hollow [Au24(PPh3)10(SC2H4Ph)5Cl2]+ nanoparticle formed by reaction of non-hollow [Au25(PPh3)10(SC2H4Ph)5Cl2]2+ with excess triphenylphosphine (PPh3). Single crystal X-ray diffraction analysis shows that the nanoparticle consists of two incomplete icosahedral Au12 units linked by five thiolate linkages36. In comparison to the vertex-sharing biicosahedral [Au25(PPh3)10(SC2H4Ph)5X2]2+, the Au24 nanoparticle lacks the central Au atom (i.e., the shared vertex atom in the biicosahedral Au25), which exerts a major influence on the optical properties of the nanoparticle36, 37. This hollow structure opens up the possibility of re-filling the central vacancy of the Au24 nanoparticle by another atom from the same group as gold. Since there is only one vacancy in the Au24 nanoparticle, we expect that single-atom doping can be realized by using hollow Au24 as a template. Furthermore, by subsequently hollowing the resultant single-atom alloyed nanoparticles and then re-filling with a heterometal atom, one may achieve atom-by-atom doping in a highly controlled fashion.

Herein, we report the shuttling of single metal atom(s) of Au, Ag, and Cu using the hollow Au24 nanoparticle as a model system. Surprisingly, we discover intriguing pathways of shuttling for different metals. Instead of simple filling of the central vacancy, we find that the incoming atom squeezes the pre-existent gold atom of the nanoparticle into the hollow site to produce M1Au24 nanoparticles (M = Au/Ag/Cu). The obtained non-hollow M1Au24 nanoparticles can be further converted to M2Au23 nanoparticles by the hollowing-refilling strategy. The determination of the atomic structures of Cu1Au24 and Ag1Au24 nanoparticles by X-ray crystallography, together with density functional theory (DFT) simulations, provides a clear map on how the single-atom shuttling occurs in the atomically precise nanoparticles.

Results

Shuttling a metal atom into a hollow nanoparticle

The hollow [Au24(PPh3)10(SC2H4Ph)5Cl2]+ nanoparticle (Fig. 1a, abbreviated as Au24 hereafter) is chosen as a model to demonstrate the filling of the central vacancy and dislodging of an atom out of the resultant 25-atom nanoparticle (Fig. 1b, abbreviated as Au25 hereafter). The hollow Au24 nanoparticle was made by the reaction of [Au25(PPh3)10(SR)5Cl2]2+ with excess PPh3 36.

Fig. 1
figure 1

X-ray structures and UV–Vis spectra of Au24 and Au25 nanoclusters. X-ray structures of the hollow Au24 rod with the central atom dislodged (a), and the Au25 rod (b)36, 37, the central gold atom in the Au25 rod is shown using space-filling model for clarity. Color code: Au, yellow; P, orange; S, red; Cl, green. C and H atoms are not shown for clarity; UV–Vis spectra of the hollow Au24 rod and the Au25 rod are shown in c and d, respectively

Full size image

In the present work, we have discovered that reaction of the Au24 (dissolved in CH2Cl2, Fig. 1c) with Au(I)Cl readily restores Au25 within a few seconds, evidenced by ESI-MS analysis of the final product (Fig. 2b, black line) with a major peak of 2+ charge at m/z = 4151.6 Da (expected m/z = 4151.6 Da), also evidenced by the UV–Vis spectrum (Fig. 2a) being identical to that of Au25 (Fig. 1d)37. In order to obtain single-atom doping with Ag and Cu, we further tested the Au24 with Cu(I)Cl and Ag(I)Cl salts. Results show that addition of CuCl or AgCl to a dichloromethane solution of the Au24 leads to a rapid (~4 s) change of the solution color from red to green, indicating the possible formation of new products doped with Cu or Ag. The UV–Vis spectrum of the Cu-doped nanoparticle is found to be similar to that of the Au25 nanoparticle (Fig. 2a, blue line), while the Ag-doped nanoparticle exhibits a slight red shift by ~11 nm (Fig. 2a, red line). ESI-MS analysis of the doped clusters (Fig. 2b) shows that the major mass peak for Cu doping is located at m/z = 4085.1 (Fig. 2b, blue line), assigned to [Cu1Au24(PPh3)10(PET)5Cl2]2+ (theoretical m/z = 4085.1 Da), and for Ag doping, the peak at m/z = 4107.0 (Fig. 2b, red line) corresponds to [Ag1Au24(PPh3)10(PET)5Cl2]2+ (theoretical m/z = 4107.1 Da).

Fig. 2
figure 2

UV–Vis and ESI-MS spectra of homo-gold and doped MAu24 nanoclusters. a UV–Vis spectra of homo-gold [Au25(PPh3)10(PET)5Cl2]2+ (black line), single Ag doped [Ag1Au24(PPh3)10(PET)5Cl2]2+ (red line), and single Cu doped [Cu1Au24(PPh3)10(PET)5Cl2]2+ (blue line); b Positive mode ESI-MS spectra of homo-gold [Au25(PPh3)10(PET)5Cl2]2+ (black line), single Ag doped [Ag1Au24(PPh3)10(PET)5Cl2]2+ (red line), and single Cu doped [Cu1Au24(PPh3)10(PET)5Cl2]2+ (blue line)

Full size image

We further crystallized the products and performed X-ray crystallography to determine the sites occupied by the incoming Cu and Ag atoms in the structure of the doped nanoparticles (for details see Supplementary Figs. 1–3 and Supplementary Tables 1 and 2). Since the atomic numbers of Cu (Z = 29) and Ag (Z = 47) are considerably less than that of Au (Z = 79), they can be readily differentiated in the X-ray crystallographic analysis. Partial occupancy analysis was employed to find the location of Cu and Ag atoms (details are given in the Supplementary Note 2). Results show that Cu can occupy either the apex or waist positions of the rod-shaped nanoparticle (Fig. 3, right), while Ag was only found at the apex of the nanoparticle (Fig. 3, left). Interestingly, the central position of the nanoparticle is 100% occupied by gold atom in both products, rather than a Cu or Ag atom, as one would expect since the central vacancy is ready for filling.

Fig. 3
figure 3

Shuttling one Ag or Cu atom into the 24-atom hollow gold nanoparticle: pathways of single Ag/Cu entering the hollow Au24 nanoparticle. Note, Ag1Au24 and Cu1Au24 are presented using X-ray cryptographic data of this work, and Au24 structure is adopted from ref. 36. Color codes: Au, yellow; Ag, blue; Cu, magenta. Other non-metal atoms are not shown for clarity

Full size image

The results of Au24 reaction with AuCl, AgCl, and CuCl clearly demonstrate the success in single-atom doping into the gold nanoparticle. In the case of reaction with AuCl, the pathway of how the central vacancy is filled cannot be revealed, but the reactions of Au24 with AgCl and CuCl clearly show that the copper or silver atom does not directly take the central empty position as one would initially expect, instead the Cu or Ag dopant should squeeze one surface gold atom into the central vacancy. To map out the mechanistic details, we further carried out DFT simulations on the formation of hollow Au24 from the Au25 nanoparticle and the back filling of Au24 to form MAu24 (M = Cu or Ag).

On the shuttling-out mechanism for the formation of hollow Au24

Experimentally we found that excess phosphine ligands play a key role in the formation of hollow Au24 nanoparticle from its parent Au25 nanoparticle, in agreement with the previous study36. DFT calculations were performed using [Au25(PH3)10(SH)5Cl2]2+ as a model of the experimental nanoparticle by simplifying PPh3 to PH3 and SC2H4Ph to SH. Results show that adsorption of a PH3 onto a gold atom located at the waist position (Au1, Fig. 4, green ball) of the nanoparticle is the most likely mechanism to initiate the reaction. A PH3 of the rod via a migration process (Reac → Int1, Supplementary Movie 1) may form a bond with the Au1. Of note, the Au–PPh3 bond is flexible, which allows rapid exchange between the free and bound PPh3 38.

Fig. 4
figure 4

Mechanisms for the formation of hollow Au24 cluster proposed by DFT calculations. The values of interatomic distances are a = b = c = 2.97, a′ = 3.23, b′ = 2.90, and c′ = 2.86 Å. DFT results show ΔE 1 = 25.9, ΔE 2 = 7.1, ΔE 3 = −5.2, ΔE 4 = 34.4, and ΔE 5 = −30.7 kcal/mol. Color codes: Au1, green; Au2, gray; other Au, yellow; S, red; P, magenta. Other atoms and bonds are not shown for clarity

Full size image

Upon the formation of Au–PH3 bond and subsequent Au–S bond breaking (Int1, Fig. 4), the gold atom at the center of the nanoparticle (Au2, Fig. 4, gray ball) dislocates toward the surface of the nanoparticle, evidenced by changes in the Au–Au atomic distances (Fig. 4). The Au1–Au2 bond distance becomes 2.90 Å (b′ in Fig. 4) which is considerably less than the bond distance between Au2 and gold atoms located at the lower side of the waist position (a′ = 3.23 Å, Fig. 4). Next, the Au2 is completely pulled up to the surface of the nanoparticle (Int1 → Int2 and Supplementary Movie 2). In turn, this exposes the Au1 to PH3 ligands in the reaction medium to form Au(PH3)2 on the surface of the nanoparticle (Int2 → Int3). The Au(PH3)2 + moiety eventually detaches from the nanoparticle to result in the [Au24(PPh3)9(SR)5Cl2]+ nanoparticle (Int3 → Int4). The generation of Au(PPh3)2 + ion is indeed experimentally confirmed by ESI-MS (Supplementary Fig. 4). Finally, the as-formed [Au24(PPh3)9(SR)5Cl2]+ nanoparticle reacts with a PH3 to result in the hollow [Au24(PPh3)10(SR)5Cl2]+ nanoparticle (Int4 → Pr).

On the shuttling-in mechanism to form CuAu24 and AgAu24

The MAu24 nanoparticle has five non-equivalent types of metal positions (P1–P5 as indicated in Fig. 5a). Geometry optimizations of MAu24 with M located at the different positions show that both Cu and Ag energetically disfavor to occupy positions of P2 and P4 (Supplementary Tables 3 and 4), in good agreement with the X-ray crystallography analysis. However, relative energetics of the nanoparticles with M at positions P1, P3, and P5 are not in line with the experimental results. DFT results show that Cu prefers to occupy the sites in the order of P1 ≈ P5 > P3, while for the case of Ag, the order is P5 > P3 > P1 (Supplementary Tables 3 and 4). For completeness, the Grimme-D239 and the exchange hole dipole moment (XDM)40, 41 methods were used to incorporate the van der Waals (vdW) interactions into the systems. As Supplementary Tables 3 and 4 show, DFT-D2 and DFT-XDM calculations yield nearly the same results as DFT does. The X-ray crystallography analysis indicates that Ag prefers to locate at site P1 and Cu at P3 (Fig. 3); therefore, our calculations reveal that in addition to the relative stability of the nanoparticles based on their energetics, other factors such as reaction kinetics and entropy effects (10 P3 sites vs two P1 sites) also play significant roles in the formation of MAu24.

Fig. 5
figure 5

Mechanisms for the formation of doped MAu24 clusters based on DFT calculations. a Designation of sites P1–P5 in the Au25 structure. DFT-calculated mechanisms for MAu24 (M = Cu or Ag) formation with M at b waist and c apex positions. Color code: Au, yellow; S, red; Cl, green; Cu, cyan; Ag, gray. Note, in b, a gold atom at the waist position that is pushed into the vacancy to form CuAu24 is shown in orange. In c, to show two different pathways, i.e., paths 1 and 2, corresponding gold atoms are either presented in dark blue or magenta. Only one Cl and S of the nanoparticle is presented. Other atoms and bonds are not shown for clarity

Full size image

We next considered whether the location of incoming M (Cu or Ag) is dictated by the initial interaction of M+ with the capping ligands of the nanoparticle (-SR and Cl–Au, Supplementary Fig. 5). Interaction energy of Ag+ and Cu+ with Cl–Au is found to be 9.6 and 8.0 kcal mol−1, respectively, more favorable than those with—SR. These results show that the interaction of Cl–Au with Ag is stronger than with Cu. This may indicate the single-atom transfer and its possible location is determined by the interaction of the M (Cu or Ag) with capping ligands of the nanoparticle, in agreement with our experimental trend. In addition, compared to Cu, the larger vdW radius of Ag also prevents the silver atoms from interacting efficiently with -S- due to steric hindrance. Of note, the protecting ligands of the nanoparticle make the particle surface considerably packed, which causes high spatial hindrance for Ag+ to pass through and approach the -SR group (Supplementary Fig. 6). However, Cu has a smaller vdW radius and can interact with surface thiolate ligand, which eventually pushes a gold atom at the waist position into the vacancy at the center of Au24 to form CuAu24 (Fig. 5b).

Further, we consider possible mechanisms to form AgAu24 with Ag located at the apical site of the nanoparticle. An Ag+ may interact with the Cl atom at the apex of the nanoparticle (Int6, Fig. 5c), which eventually moves to interact with three gold atoms located at the apical as well as the end positions of the nanoparticle (Int6 → Int7 → Int8, Fig. 5c). There are two possible pathways for the Ag at this position to locate at the apical site by squeezing a gold atom at sites of icosahedral center (Fig. 5c, Path 1, shown by blue arrows, Supplementary Movie 3) and waist (Fig. 5c, Path 2, shown by red arrows, Supplementary Movie 4). Our calculations using the nudged elastic band (NEB) approach42 show barrier energy of pathway 2 is 19.8 kcal mol−1 lower than that for pathway 1. This result indicates metal mobility is most likely to happen through the surface of the nanoparticle rather than the core of the icosahedron, in agreement with the mechanism for the Au24 formation.

Shuttling a second heteroatom into the nanoparticle

To shuttle a second heteroatom into the nanoparticle, the Cu1Au24 and Ag1Au24 nanoparticles were, respectively, used as the starting material. Reaction of the starting material with PPh3 at 40 °C produced hollow nanoparticles. As shown in Supplementary Fig. 7, the complete disappearance of the 700 nm peak indicates that all the M25 nanoparticles have been converted to hollow M24. The second step is to fill the hollow structure with heterometal atom by adding CuCl or AgCl salts to the solution. The color of the solution changed immediately from red to green. As shown in Fig. 6a, compared with Au25, the copper-doped product has a similar UV–Vis spectrum as that of Au24, however, the silver-doped nanoparticle shifted from ~685 nm (Ag1Au24) to ~712 nm. In the ESI-MS spectra (Fig. 6b), the Ag2Au23 nanoparticle with +2 charge was found (m/z = 4062.8 Da, theoretical m/z = 4062.6 Da), which implies a step-by-step doping of silver to the two apex sites (Fig. 7). For Cu doping, the product comprises Cu1Au24 (major, m/z = 4085.1, theoretical m/z = 4085.1) and Cu2Au23 (less, m/z = 4018.1 Da, theoretical m/z = 4018.1 Da). To explain why Cu1Au24 is the major product, we note that the starting Cu1Au24 material is a mixture of apex- and waist-doped nanoparticles, and the strong binding of PPh3 to Cu should cause the dislodging of waist Cu atom in the Cu1Au24 (70% population, see Fig. 3 above) to produce the hollow Au24 nanoparticle, and then reaction of Au24 with CuCl produces Cu1Au24, while the apex-doped Cu1Au24 (30% population, Fig. 3) produces Cu1Au23 and its reaction with CuCl gives rise to Cu2Au23, hence a minor component in the product.

Fig. 6
figure 6

UV–Vis and ESI–TOF–MS spectra of the secondary shuttling products. a UV–Vis spectra of [Cu x Au25−x (PPh3)10(PET)5Cl2]2+ (x = 1,2; blue line) and [Ag2Au23(PPh3)10(PET)5Cl2]2+ (red line), and; b Positive mode ESI-MS spectra of [Cu x Au25−x (PPh3)10(PET)5Cl2]2+ (x = 1,2; blue line) and [Ag2Au23(PPh3)10(PET)5Cl2]2+ (red line)

Full size image
Fig. 7
figure 7

Proposed mechanism of injecting two Ag atoms into the nanoparticle via the hollowing-refilling sequence: step 1, using PPh3 to make a hole in the solid Au25 nanocluster; step 2, using AgCl to refill this hole and produce Ag1Au24; step 3, continue using PPh3 to make a hole in the Ag1Au24 nanocluster and form the hollow Ag1Au23 nanocluster; step 4, using AgCl to refill the hole and yield the Ag2Au23 nanocluster. Color code: Au, yellow; Ag, magenta

Full size image

Discussion

In summary, we have successfully implemented the single-metal atom shuttling into an atomically precise metal nanoparticle and mapped out the mechanism of the conversion between the Au24 and the Au25 nanoparticles. Our results provide a clear map of how single metal atom transfer occurs between two atomically precise nanoparticles. Based on the experimental and theoretical results, the driven force of single-atom transfer is caused by the ligand, i.e., the free PPh3 for the shuttling-out process, and the surface -Cl and -SR ligands for the shuttling-in process. The stronger binding between Ag and -Cl compared with Ag–SR leads to the exclusive Ag atom doping at the apex of the nanoparticle, while the similar energy of Cu–Cl and Cu–SR leads to the Cu atom doping into both the apex and waist positions. This work provides fundamental understanding of how to shuttle a single atom in and out of metal nanoparticles by a chemical method. The ligand-induced single-atom shuttling process also provides a strategy for controlling the doping position and the doping number of heteroatoms in alloy nanoparticles.

Methods

Materials

Unless specified, reagents were purchased from ACROS Organics or Sigma-Aldrich and used without further purification. Tetrachloroauric(III) acid (HAuCl4·3H2O, >99.99% metals basis), CuCl (99%), AgCl (99%), AuCl (99%), NaSbF6 (>99%), PPh3 (>99%), and NaBH4 (>98%) were received from ACROS Organic. Ethanol (HPLC grade, ≥99.9%), methanol (HPLC grade, ≥99.9%), and methylene chloride (HPLC grade, ≥99.9%) were from Sigma-Aldrich. UV–Vis absorption spectra were obtained using an Agilent 8453 instrument, and solution samples were prepared using DCM as the solvent. ESI-MS was recorded using a Waters Q-TOF mass spectrometer equipped with Z-spray source. The source temperature was kept at 70 °C. The sample was directly infused into the chamber at 5 μL min−1. The spray voltage was kept at 2.20 kV and the cone voltage at 60 V.

[Au24(PPh3)10(SR)5Cl2]+ and [Au25(PPh3)10(SR)5Cl2]2+ were synthesized according to the literature method34, 35 (for details see Supplementary Note 1).

[M1Au24(PPh3)(SR)5Cl2](SbF6)2 (M = Au/Ag/Cu)

The [Au24(PPh3)(SR)5Cl2]Cl nanocluster (~2 mg) was dissolved in CH2Cl2, then ~1 mg MCl salt (M = Au/Ag/Cu) was added into the solution, respectively. After shaking for a few seconds, the solution color rapidly changed from red to green. Then, the solution was centrifuged to remove the exceed salt (solid), and the solution was then dried under N2. To exchange for the anion, the obtained nanoclusters were dissolved in EtOH, then a right amount of NaSbF6 was added into the solution. The precipitate was collected after centrifugation, followed by crystallization in dichloromethane/pentane.

[M2Au23(PPh3)(SR)5Cl2](SbF6)2 (M = Ag/Cu)

Approximately 5 mg of [M1Au24(PPh3)(SR)5Cl2]Cl2 nanoclusters (M = Ag/Cu) was dissolved in 2 mL CH2Cl2 solution, followed by adding 1 g of PPh3. The reaction was allowed to proceed overnight at 40 °C. Then, 10 mL of hexane was added to remove the excess PPh3. Then, 1 mg of MCl salt (M = Au/Ag/Cu) was added into the solution, respectively. After shaking for a few seconds, the solution color changed from red to green. Then, the solution was centrifuged to remove the excess salt, and the solution was dried under N2.

X-ray crystallography

The data collections for single crystal X-ray diffraction was carried out on a Bruker Smart APEX II CCD diffractometer, using a Cu–Kα radiation (λ = 1.54178 Å). Data reductions and absorption corrections were performed using the SAINT and SADABS programs43, respectively. The structure was solved by direct methods and refined with full-matrix least squares on F 2 using the SHELXTL software package44. All non-hydrogen atoms were refined anisotropically, and all the hydrogen atoms were set in geometrically calculated positions and refined isotropically using a riding model. X-ray diffraction data refinement involving partial occupancy was used to locate the heteroatom atom.

Computational details

DFT calculations were carried out using the Quantum Espresso package45. The Projector Augmented-Wave (PAW) method was applied to describe the interaction between the electrons and nuclei46. The Perdew–Burke–Ernzerhof (PBE) form of the generalized gradient approximation was employed for electron exchange and correlation47. The gold cluster was placed at the center of a cubic box of 30.0 Å × 30.0 Å × 30.0 Å. The kinetic energy cutoff was chosen to be 450 eV and integration in the reciprocal space was carried out at the Γ k-point of the Brillouin zone. The NEB approach was used to find minimum energy path of transitions42.

Data availability

The X-ray crystallographic coordinates for structures reported in this work (see Supplementary Tables 1, 2, and Supplementary Note 2) have been deposited at the Cambridge Crystallographic Data Centre (CCDC), under deposition numbers CCDC 1562010 and CCDC 1561987. These data can be obtained free of charge from The Cambridge Crystallographic Data Centre via www.ccdc.cam.ac.uk/data_request/cif.