Single-gold etching at the hypercarbon atom of C-centred hexagold(I) clusters protected by chiral N-heterocyclic carbenes

Etching is a top-down method to downsize the structures at the atomic level and modify the chemical and physical properties of a wide range of nanomaterials such as nanocrystals and colloidal nanoparticles^1,2,3,4,5, and nanoclusters^{6,7,8,9,10,11} for a variety of applications. For example, chemical etching methods that involve ligand engineering have made great advances, such as thiolate etching of phosphine- or thiolate-protected nanogold clusters^6,7,8, reverse etching of thiolated-Au₂₅ with phosphine (exchange of strong donor ligands for weaker donor ligands)⁹, and phosphine exchanged by N-heterocyclic carbene (NHC) to give NHC-containing Au₁₁^10,11. Rapid developments in X-ray crystallography have revealed that chemical etching alters the nanocluster structures of the metal core¹² and ligand surface^{13,14,15,16,17}. However, the control and understanding of chemical etching at the atomic level have only just begun^18,19. Recently, Cao et al. used real-time electrospray ionisation mass spectrometry to reveal degradation and anomalous recombination processes in the chemical etching of Au₂₅ nanoclusters¹⁹. It also remains controversial whether the ligand-exchange mechanisms in nanogold regions containing Au^I and Au⁰ atoms is S_N2-like bimolecular nucleophilic substitution or S_N1-like type unimolecular nucleophilic substitution^20,21,22. Despite the promise of chemical etching as a general technique to downsize metal clusters at the atomic level, little attention has been paid to ligand-protected Au^I clusters.

Among the Au^I cluster family, clusters radially coordinated to main-group elements^{23,24,25,26,27,28,29} such as O-centre²³, N-centre^24,25, C-centre²⁶, and S-centre^27,28,29 are attractive due to the polyhedral structures similar to nanogold clusters, Au^I⋅⋅⋅Au^I interactions^{30,31,32,33,34} and structure-dependent photophysical properties. In particular, the hypercoordinated carbon (hypercarbon)³⁵-centred hexagold(I) (CAu^I₆) cluster [C(Au^I-L)₆]X₂ (L = ligand; X = counterion)²⁶ that bridges nano-sized metal clusters and organic molecules^36,37 has attracted attention and significant advances have been made in this area. Many of these CAu^I₆ clusters exhibit structure-specific luminescence^{38,39,40,41,42,43,44} and can be used as bio-labels^39,44 by using highly bottom-up designable shell ligands based on phosphines^38,39,40 and NHCs^{41,42,43,44,45}. However, there is only one isolated example of a C-centred pentagold(I) cluster [C(Au^I-L)₅]X (L = triphenylphosphine, TPP) produced by the bottom-up synthesis of aurating CH₂[B(OCH₃)₂]₂ with a gold(I) complex, and if the reaction time is extended, the TPP-protected CAu^I₆ cluster is the main product⁴⁶. Thus, it is still difficult to control the number of Au^I atoms bound to the hypercarbon centre. In particular, from the viewpoint of the effects of reducing the number of Au^I atoms on photophysical properties and reactivity, the development of a highly generalised single-gold etching method for gold(I) clusters is an important research topic.

Here, we discovered that a chiral NHC-protected C-centred hexagold(I) cluster can be etched with a bisphosphine ligand to generate a chiral NHC-protected CAu^I₅ cluster. This was achieved by controlling the number of gold atoms centred at the hypercarbon at the atomic level (Fig. 1). Furthermore, this etching method is also useful for the synthesis of TPP-protected CAu^I₅ analogues. In general, smaller gold clusters show more blue-shifted emission than larger clusters, but the ligand-protected CAu^I₅ clusters show unusually red-shifted signals in both absorption and emission spectra compared to the CAu^I₆ counterparts, which was rationalised by theoretical calculations. Further experimental and theoretical studies suggest that a tandem dissociation-association-elimination pathway is involved in the etching mechanism. The NHC-protected CAu^I₆ clusters are generally chemically stable, and the active site had to be placed on the hypercarbon to confer reactivity. In fact, CAu^I₅ synthesised by this method was highly reactive with AuCl, producing a CAu^I₆ cluster with a different ligand. Thus, the chemical etching method is expected to be a way not only to reduce the size of metal ion clusters and significantly change their electronic structure, but also to asymmetrise the metal ion cluster structure and provide active sites.

Etching-induced elimination of [LAu^I] at the hypercarbon centre of the ligand-protected C-centred hexagold(I) (CAu^I₆) cluster results in the formation of the ligand-protected C-centred pentagold(I) (CAu^I₅) cluster, wherein the ligand can be an N-heterocyclic carbene or phosphine.

Full size image

Synthesis of CAu^I ₅ by etching CAu^I ₆ with bisphosphine

We previously reported enantiopure NHC-protected asymmetrically twisted CAu^I₆ clusters: [(C)(Au^I-SS-NHC)₆](BF₄)₂ and [(C)(Au^I-RR-NHC)₆](BF₄)₂ ( = SS- and RR-1^NHC, SS-NHC = N,N’-bis[(S)-α-methylbenzyl]-benzimidazol-2-ylidene, RR-NHC = N,N’-bis[(R)-α-methylbenzyl]-benzimidazol-2-ylidene)⁴³. Although NHCs are generally thought to bind strongly to coinage metal^47,48, we investigated whether etching occurs when bisphosphine is added to NHC-protected CAu^I₆ clusters. For example, when 2.5 equiv. of cis-1,2-bis(diphenylphosphino)ethene (cis-depe) were added to a dichloromethane solution of SS-1^NHC, the original pale-yellow solution immediately turned orange. Subsequent crystallisation from diethyl ether/dichloromethane at 4 °C gave orange-red blocky crystals of [(C)(Au^I-SS-NHC)₅](BF₄) (SS-2^NHC) in 80% yield on a hypercarbon basis. Its enantiomer [(C)(Au^I-RR-NHC)₅](BF₄) (RR-2^NHC) was also synthesised. Another product [(cis-depe)₂Au^I](BF₄) (3) was obtained as colourless blocky crystals by prolonged recrystallisation. They were characterised by ESI-MS spectrometry, NMR spectroscopy and elemental analysis (Suppl. Figs. 2–12). The ESI-MS spectrum found the CAu^I₅ cluster SS-2^NHC at m/z 2628.85 corresponding to [(C)(Au^I-SS-NHC)₅]⁺ (calcd. 2628.73). In the ¹H NMR spectrum of the CAu^I₅ cluster SS-2^NHC in d₆-acetone, the signals attributed to the NHC ligand showed a significant downshift compared to those of the CAu^I₆ cluster SS-1^NHC (Suppl. Fig. 6). This is attributed to the magnetic environment, which is deshielded from the shell ligands with less steric hindrance. Similarly in the ¹³C NMR spectra, the signal at 198.8 ppm attributed to the NHC carbon-donors of SS-2^NHC was downshifted from those of SS-1^NHC at 190.0 ppm (Suppl. Fig. 7), suggesting a marked influence on resonances from different gold nuclearities.

Moreover, a phosphine analogue [(C)(Au^I-TPP)₅](BF₄) (2^TPP) was also obtained in 90% yield by etching [(C)(Au^I-TPP)₆](BF₄)₂ (1^TPP)²⁶ instead. The ³¹P NMR spectrum of 2^TPP in d₆-acetone showed a singlet signal at 32.6 ppm, which was downshifted from that of 1^TPP (Suppl. Fig. 8)⁴⁶. Furthermore, etching SS-1^NHC and 1^TPP with excess cis-depe (50 equiv.) yielded corresponding pentagold(I) clusters in both cases, with no detectable smaller gold species such as tetragold(I) or trigold(I) clusters. Etching SS-1^NHC and 1^TPP with 1,2-bis(diphenylphosphino)benzene also yielded the corresponding CAu^I₅ clusters. These data indicate that etching the CAu^I₆ cluster with cis-depe provides high selectivity for the CAu^I₅ cluster.

Single-crystal structures and computational bonding analysis

The single-crystal X-ray diffraction (SCXRD) determined structures in Fig. 2 show the overall structure of SS- and RR-2^NHC including a hypercarbon, five gold(I) ions, five ligands and a BF₄^– counterion. SS- and RR-2^NHC are crystallised in the I₄ space group with low flack parameters of 0.010(9) and –0.027(11), respectively. Their flack parameters are very low (Suppl. Table 1), suggesting that optically pure molecules are packed. Take the example of SS-2^NHC as shown in Table 1, the Au^I⋅⋅⋅Au^I distances (2.8667(10)–3.3141(15) Å) and the C_NHC–Au^I bonds (2.03(2)–2.08(4) Å) are similar to those in SS-1^NHC, but the C_centre–Au^I bonds (2.03(2)–2.075(8) Å) of SS-2^NHC are slightly shorter than those of SS-1^NHC (2.100(14)–2.126(12) Å)⁴³, suggesting that the endohedral five C_centre–Au^I bonds in the CAu^I₅ cluster are more favourable. The [(C)(Au^I-L)₅]⁺ cation in SS-2^NHC can be regarded as eliminating one [LAu^I] moiety from the [(C)(Au^I-L)₆]²⁺ cation in SS-1^NHC. As a result, the NHC ligands of SS-2^NHC rearrange themselves to minimise mutual steric hindrance (in Fig. 2b, three grey-coloured benzimidazolylidene moieties on the same plane and two orange-coloured benzimidazolylidene moieties on two planes with a 63° crossing angle), forming a monocationic CAu^I₅ cluster with C₂-symmetry (Suppl. Fig. 15). It should be noted that the hypercarbon of SS-2^NHC is close to the centroid of the four gold atoms at the bottom of the square pyramid (0.46(3) Å), which could be an important coordinating site for post-functionalisation (vide infra). Meanwhile, the surface vacancy found in this SS-2^NHC molecule is well shielded in its packing structure by intermolecular interactions with a ligand on the gold(I) at the apex of another cluster molecule (Suppl. Fig. 16), thus maintaining high chemical stability in the solid state. In contrast, the phosphine-protected analogue 2^TPP crystallised in the P2₁/n space group (Fig. 2c)⁴² and exhibited Au^I⋅⋅⋅Au^I interactions (2.85528(18)–3.21332(19) Å) and P–Au^I bonds (2.2546(8)–2.2735(8) Å). The C_centre–Au^I bonds (2.064(3)–2.082(3) Å) of 2^TPP are slightly shorter than those of CAu^I₆ counterpart (average 2.12 Å)²⁶, similar to the shorter C_centre–Au^I bonds of SS-2^NHC compared to SS-1^NHC.

a Etching syntheses of CAu^I₆ clusters^26,43 to CAu^I₅ clusters using cis-1,2-bis(diphenylphosphino)ethene (cis-depe) (Suppl. Fig. 1). b Single-crystal X-ray diffraction (SCXRD) structures of the cations [(C)(Au^I-L)₅]⁺ (L = SS- and RR-NHC) and the CAu^I₅ cores of SS- and RR-2^NHC with optically active NHC ligands. c SCXRD structures of the cation [(C)(Au^I-TPP)₅]^+ 46 and the core of 2^TPP with TPP. d Microscopy images of crystals of SS-2^NHC (orange crystals) and 2^TPP (yellow crystals) under ambient light. e Explanation and examples of geometry index τ values⁵¹, which is used for showing the geometric differences of CAu^I₅ cores with NHCs and TPP (Table 1). The τ value represents the geometric difference between the regular square pyramidal (τ = 0) and trigonal bipyramidal (τ = 1), τ = (β – α)/60°, where α and β are the two largest basal angles. Colour code: Au, yellow; C, grey; N, blue; P, orange; H, white. BF₄^– counterions and solvent molecules are omitted for clarity.

Full size image

Moreover, the bonding characters of CAu^I_n (n = 5, 6) clusters were computationally studied based on the above crystallography data, and the bond distances of the crystal structures were well reproduced by the density functional theory (DFT) calculations. The calculated C_centre–Au^I and Au^I⋅⋅⋅Au^I distances of CAu^I_n (n = 5, 6) cores as well as Wiberg bond orders (WBO) suggested very interesting structural dependencies (Suppl. Table 3). The C_centre–Au^I and Au^I⋅⋅⋅Au^I distances of the N,N’-diisopropylimidazolidene (IiPr)-protected CAu^I₆ cluster⁴¹ are 2.19 and 3.10 Å, respectively. Their bond orders are 0.41 and 0.16, respectively, indicating that the C_centre–Au^I bond is stronger than each Au^I⋅⋅⋅Au^I interaction. In the CAu^I₆ cores of SS-1^NHC and 1^TPP, the binding characteristics obtained are largely unchanged. Of note, the C_centre–Au^I bonding in the CAu^I₅ cores of SS-2^NHC and 2^TPP is slightly stronger compared to the corresponding CAu^I₆ clusters. This is demonstrated by the shorter bond lengths (2.09–2.16 Å) and larger WBO values (0.50–0.57). Regarding the aurophilic interactions in the CAu^I₅ cores, the Au^I⋅⋅⋅Au^I distances and bond orders are nearly the same as those of CAu^I₆ cores. Therefore, missing one Au atom in the CAu^I₅ cores may result in stronger C_centre–Au^I bonds, which may be important for stabilising the CAu^I₅ clusters. Moreover, the orbital interactions of the CAu^I₆ cluster were previously discussed in detail^49,50: the SS-1^NHC and SS-2^NHC clusters have [CAu₆]²⁺ and [CAu₅]⁺ cores, respectively, and the C–Au^I bond orders exhibit an unusual C–Au^I bond hypervalence.

In addition, to better understand the geometric differences of CAu^I₅ cores with NHCs and TPP, we introduced the index parameter τ (Fig. 2e, Table 1), wherein τ is 0 for perfect square pyramidal and 1 for perfect trigonal bipyramidal⁵¹. This evaluation method suggests that the CAu^I₅ cores of SS- and RR-2^NHC have a distorted square pyramidal geometry (τ = 0.32), while the CAu^I₅ core of 2^TPP is much closer to a trigonal bipyramidal geometry (τ = 0.68), indicating that NHCs and TPP exert different ligand effects on the CAu^I₅ core.

Absorption, emission profiles, and theoretical calculations

The CAu^I₆ clusters are known to be efficient emitters with intriguing structure-dependent properties^{38,39,40,41,42,43,44,45}, while the photophysical properties of the CAu^I₅ clusters remain unknown. In general, reducing the metal core size is known to induce a blue shift in absorption and emission⁵². However, the UV-vis spectra of SS-2^NHC and 2^TPP in dichloromethane, in contrast, showed their maximum absorption wavelengths at 420 nm and 382 nm, respectively, and were significantly more red-shifted than those of CAu^I₆ clusters, SS-1^NHC (λ^max = 373 nm) and 1^TPP (λ^max = 365 nm) (Suppl. Fig. 17). Similarly, photoluminescence of the CAu^I₅ clusters showed a bathochromic shift signal in contrast to the CAu^I₆ clusters (Fig. 3a, Suppl. Fig. 18). The solid-state SS-2^NHC exhibited orange-red emission (λ_em^max = 676 nm), which is 151 nm more red-shift than SS-1^NHC. The acetone solution of SS-2^NHC was also red-emissive, with no apparent solvation effects (Suppl. Fig. 19). On the other hand, the emission of 2^TPP in the solid state is 59 nm more red-shifted than 1^TPP and emits yellow at 365 nm excitation (λ_em^max = 584 nm). Neither 1^TPP nor 2^TPP emits light in solution. This is because the terminal coordination of phosphine to the gold(I) atom may facilitate nonradiative relaxation pathways³⁸. Comparing the absolute quantum yields (Φ), SS-2^NHC in the solid state showed the strongest emission. Here, the Φ values of SS-2^NHC, SS-1^NHC, 2^TPP, and 1^TPP were 0.61, 0.02, 0.29, and 0.19, respectively (Suppl. Fig. 20). They exhibit microsecond-level lifetimes (Suppl. Fig. 21), suggesting phosphorescence properties in the solid state.

a Emission spectra: SS-1^NHC (blue line, excited by 266 nm), SS-2^NHC (orange line, excited by 510 nm), 1^TPP (blue line, excited by 356 nm), 2^TPP (yellow line, excited by 365 nm), insets: photographs (size: 12 mm ×12 mm) of crystals under 365 nm UV-light irradiation. b The highest occupied molecular orbital (HOMO) and the lowest unoccupied molecular orbital (LUMO) distributions of SS-2^NHC. c HOMO–LUMO gap values based on theoretical calculations: SS-1^NHC (3.99 eV), SS-2^NHC (3.30 eV), 1^TPP (4.50 eV), and 2^TPP (3.98 eV). Source data are provided as a Source Data file.

Full size image

To gain insight into the optical properties specific to their electronic structures, we performed DFT and time-dependent (TD)-DFT calculations based on the SCXRD structures (Suppl. Figs. 22–27). The UV-vis spectrum of SS-2^NHC calculated by TD-DFT (Suppl. Fig. 22) reproduces well the maximum band in the range from 393 to 415 nm, which is mainly due to the transitions of HOMO-1 → LUMO, HOMO-2 → LUMO and HOMO → LUMO + 1 (Suppl. Table 4). Orbital composition analysis by Mulliken partition (Suppl. Table 6) reveals that the occupied orbitals of HOMO-n (n = 0–2) in SS-2^NHC are mainly derived from gold(I) ions (52.6–55.4%) with an increased contribution from the hypercarbon (26.9–32.5%) compared with that in SS-1^NHC 43. The unoccupied orbitals of LUMO+n (n = 0, 1) in SS-2^NHC are localised mainly at NHCs (64.7–74.8%), with small fractions of gold(I) ions (25.0–35.2%). Thus, the dominant metal-to-ligand charge-transfer (MLCT) mixed with a slight metal-centred (MC) charge-transfer was responsible for the low-energy absorption bands⁴². On the other hand, the compositions of the frontier orbitals of the TPP-protected CAu^I_n (n = 5, 6) clusters are also comparable (Suppl. Tables 7, 8, Suppl. Figs. 26, 27). These theoretical results suggest that the C-centred gold(I) core and ligands are essentially enrolled in their electronic structures, and thus different optical features can be explained by altering the gold nuclearity and ligands primarily via the MLCT transition⁴².

Importantly, the HOMO–LUMO gaps calculated for these CAu^I_n (n = 5, 6) clusters (Fig. 3c) show that the gaps of CAu^I₅ clusters are clearly smaller than those of the CAu^I₆ clusters: 3.30 eV (SS-2^NHC) < 3.99 eV (SS-1^NHC), 3.98 eV (2^TPP) < 4.50 eV (1^TPP). This is in good agreement with the fact that the absorptions of CAu^I₅ clusters are more red-shifted than those of the CAu^I₆ clusters. The calculated phosphorescence energies well reproduce the smaller phosphorescence energy of SS-2^NHC (1.83 eV) than that of SS-1^NHC (2.36 eV) (Suppl. Table 9) observed in experiments, which confirms that the emission of SS-2^NHC is red-shifted and reveals that the C-centred gold(I) clusters have a pronounced size-dependent effect on the photoluminescence properties.

Probing the atomic-level etching process

To understand the mechanism of this efficient etching method to reshape C-centred gold(I) clusters, we investigated this process using UV-vis absorption spectroscopy, NMR spectroscopy and ESI-MS spectrometry. First, the UV-vis spectra of SS-1^NHC etched with cis-depe showed a rapid change (Fig. 4a,b). When cis-depe was added to a dichloromethane solution of SS-1^NHC (c = 5 × 10^–5 M, 293 K), its characteristic peak (λ^max = 345 nm) gradually decreased and new peaks appeared at 385 nm and 421 nm derived from SS-2^NHC. Accordingly, the original pale-yellow solution turned bright yellow. In contrast, the time-course UV-vis spectra of 1^TPP etched with cis-depe (Suppl. Fig. 28) showed that the peak at 382 nm appeared even more instantaneously for 2^TPP. The colourless solution turned yellow within 5 s, a much faster change than etching SS-1^NHC. This can be reasonably explained by the weaker binding of triphenylphosphine to gold than NHC⁴⁸, consistent with the shorter C_NHC–Au^I bonds than P–Au^I bonds in the single-crystal structures.

a Time-course UV-vis absorption spectra of the etching reaction of SS-1^NHC (c = 5 × 10^–5 M, 293 K) using cis-depe in dichloromethane: stage A, 1 s intervals; stage B, 5 min intervals; insets: photographs taken at the beginning of the reaction and after 70 min under ambient light, and a colour scale corresponding to the absorbance changes of the reaction. b The changes in absorbance at 345 nm and 421 nm as a function of time (corresponding to the spectra in a), suggesting that SS-2^NHC was rapidly formed. c ¹H NMR spectrum (d₆-acetone, 300 K), showing signals of SS-1^NHC (purple) and internal standard (IS, 1,3,5-trimethoxybenzene, blue-labelled). d Signals from SS-2^NHC (pink-labelled, 53% yield) measured immediately after adding cis-depe. e After 0.5 h, SS-2^NHC was formed in 98% yield. f A proposed etching mechanism with two intermediates Int1^L and Int2^L (L = NHC, TPP), with Int1^TPP and Int2^TPP detected by ESI-MS spectrometry (Suppl. Fig. 32). Colour code: Au, yellow; C, grey; L, cyan; P, orange. Source data are provided as a Source Data file.

Full size image

Next, the ¹H NMR spectra of SS-1^NHC etched with cis-depe in d₆-acetone were measured over time (Fig. 4c–e). The results showed that SS-2^NHC was formed in 98 % yield 0.5 h after the addition of cis-depe (1,3,5-trimethoxybenzene as the internal standard), with the detection of another product [(cis-depe)₂Au^I](BF₄) (3) (Suppl. Fig. 29). On the other hand, when 1^TPP was etched with cis-depe, the time-course of the ¹H NMR spectra in d₆-acetone showed that 2^TPP was formed in 88% yield after 18 min (Suppl. Fig. 30). The ³¹P NMR spectrum in d₆-acetone after the reaction showed signals for complex 3 (δ 21.3 ppm) and free triphenylphosphine (δ –4.2 ppm). Furthermore, we turned to ESI-MS spectrometry to obtain more molecular information. As a result, a signal corresponding to SS-2^NHC was observed at m/z 2628.71 (calcd. 2628.73 for [(C)(Au^I-SS-NHC)₅]⁺) immediately after adding cis-depe, and the very weak signals of two intermediates Int1^NHC and Int2^NHC were found (Suppl. Fig. 31). Interestingly, in the process of etching 1^TPP, two di-cationic mass peaks, Int1^TPP [(C)(Au^I-TPP)₆]²⁺ (m/z found 1252.08, calcd. 1252.12) and Int2^TPP ([(C)(Au^I-TPP)₅(Au^I-cis-depe)]²⁺ (m/z found 1450.62, calcd. 1450.86) were detected (Suppl. Fig. 32). However, no information was available for the association adduct [(C)(Au^I-TPP)₆(cis-depe)]²⁺ by binding cis-depe to 1^TPP. This would suggest that the initial stage of etching phosphine-protected CAu^I₆ cluster with cis-depe is a dissociation process. Similarly, the formation of the association adduct [(C)(Au^I-SS-NHC)₆(cis-depe)]²⁺ would be difficult due to the steric hindrance from NHC ligands in SS-1^NHC. Overall, as shown in Fig. 4f, the initial dissociation process similar to the ligand-exchange S_N1-like mechanism²² generating the first intermediates Int1^L (L = TPP or NHC) would occur when etching the ligand-protected CAu^I₆ clusters. Subsequent association with cis-depe would form the second intermediates Int2^L (L = TPP or NHC), and then the elimination of Au^I with cis-depe finally produce the corresponding CAu^I₅ clusters.

To illustrate the proposed etching mechanism as shown in Fig. 4f, we computed the energy profiles of the etching process in dichloromethane (Suppl. Fig. 33). In the first dissociation stage, to break one of the six Au^I–L bonds from the original ligand-protected CAu^I₆ cluster, for example when L = TPP, Int1^TPP formed in relatively high energy (30.2 kcal mol^–1), suggesting the high stability of 1^TPP. In the second association stage, when cis-depe coordinated to Int1^TPP, the resulting Int2^TPP was largely stabilised with a dramatically decreased energy of 10.9 kcal mol^–1. Third, followed by the elimination of [cis-depeAu]⁺ to form a highly stable complex 3, 2^TPP was finally formed with an energy of 16.4 kcal mol^–1 via breaking the C_centre–Au^I bond and four Au^I⋅⋅⋅Au^I contacts of Int2^TPP. The energy profiles were similarly illustrated when L = NHC (Suppl. Fig. 33b). Therefore, the theoretical data supported the tandem dissociation-association-elimination pathway in this etching process.

Stability study of CAu^I ₅ clusters

SS-2^NHC and 2^TPP in the solid state are stable for more than a year under ambient conditions, but they are reactive in solution, in contrast to the more stable CAu^I₆ clusters⁴³. The changes over time of ¹H NMR spectra indicate that the solution of SS-2^NHC in d₆-acetone is stable for at least one week (Suppl. Fig. 34). On the other hand, once dissolved in CDCl₃, both SS-2^NHC and 2^TPP partially reverted to the CAu^I₆ cluster (Suppl. Figs. 35, 36), and the original yellowish solutions gradually faded. These results suggest that the CAu^I₅ clusters are more reactive in solution than the corresponding CAu^I₆ clusters, which is consistent with the higher energies of the CAu^I₅ clusters in the calculated energy profiles described above.

Reactivity of the CAu^I ₅ cluster

The metal clusters with exposed surfaces are of growing interest⁵³. The more open-spaced coordination site at the bottom of the distorted square pyramidal structures of SS- and RR-2^NHC were assumed to be the site where the sixth Au^I species is most accessible to the hypercarbon. Therefore, we added an acetone solution of (tht)Au^ICl (1 equiv.) to an acetone solution of SS-2^NHC at room temperature. The reaction was conducted in an ultrasonication bath for 20 min, and the crystallisation yielded a heteroleptic Cl-coordinated CAu^I₆ cluster [(C)(Au^I-SS-NHC)₅(Au^ICl)](BF₄) (SS-4^NHC) (Fig. 5, see characterisation data in Suppl. Figs. 37–41, Suppl. Table 10). Its enantiomer RR-4^NHC was similarly obtained. In the overall SCXRD structures of SS- and RR-4^NHC, in particular, the structures corresponding to the five NHC ligand parts of the CAu^I₅ clusters, SS- and RR-2^NHC, were found to be largely intact (Suppl. Fig. 37). The sixth gold(I) atom was coordinated to the Cl-anion with bond distances of 2.308(6) Å and 2.310(7) Å in SS- and RR-4^NHC, respectively. However, the introduction of AuCl into 2^TPP was not successful, probably due to its low stability, and only the original CAu^I₆ cluster 1^TPP was finally isolated. Given the NHC-protected monogold chloride complexes are extensively used as active catalysts^54,55,56, this cluster-based analogue Cl-coordinated CAu^I₆ cluster protected by the NHC ligands will be a milestone in the development of highly reactive hexagold(I) clusters.

a Reaction of SS-(or RR-)2^NHC with (tht)Au^ICl (tht = tetrahydrothiophene) yielded Cl-coordinated CAu^I₆ clusters: SS- and RR-4^NHC [(C)(Au^I-L)₅(Au^ICl)](BF₄) (L = SS- and RR-NHC). b The SCXRD structure of the [(C)(Au^I-SS-NHC)₅(Au^ICl)]⁺ cation and the Cl-coordinated CAu^I₆ core in SS-4^NHC. The arrangement of three grey-coloured benzimidazolylidene moieties on the same plane and two orange-coloured benzimidazolylidene moieties on two planes in SS-4^NHC is similar to that in SS-2^NHC (Suppl. Fig. 37). Colour code: Au, yellow and orange; C, grey; N, blue; Cl, green. Hydrogen atoms, BF₄^– counterion, and solvent molecules are omitted for clarity.

Full size image

In addition, the Cl-coordinated CAu^I₆ clusters SS- and RR-4^NHC restored very weak green emission in the solid state and no emission was observed in the solution at room temperature. In particular, the circular dichroism spectra of SS- and RR-2^NHC and SS- and RR-4^NHC in dichloromethane (Suppl. Fig. 42) showed similar chiroptical signals with the strongest signal at 250 nm contributed mainly by the chiral ligands, which can be explained by the similar arrangement of the ligand shell and symmetric metal cores. In a word, the etch-produced CAu^I₅ clusters (SS- and RR-2^NHC) and the post-functionalised Cl-coordinated CAu^I₆ clusters (SS- and RR-4^NHC) exhibit different properties from the original CAu^I₆ clusters (SS- and RR-1^NHC), thus revealing intriguing structure-property relationships by using etching as atomically precise “surgery” at the hypercarbon atom.

In summary, we have shown that etching of the NHC-protected CAu^I₆ clusters allows size-selective synthesis of the corresponding CAu^I₅ clusters. The peculiar red-shift signals in the absorption and emission of CAu^I₅ clusters can be explained by theoretical calculations. A tandem dissociation-association-elimination pathway for the atomical-level etching was proposed based on experimental and theoretical studies. The envisaged coordination ability of the hypercarbon atom in CAu^I₅ clusters was confirmed by adding Au^ICl, leading to the novel heteroleptic Cl-coordinated CAu^I₆ clusters. These results of the single-gold etching of the CAu^I₆ clusters at the atomic level indicate a unique and highly generalised method using phosphine ligands for etching of NHC-protected gold clusters. This study not only elucidated the unusual photophysical properties of metal clusters containing fewer metal nuclei, but also provided opportunities to explore post-functionalisation and reactivities in surface-exposed metal ion clusters. This result shows that establishing a synthesis method using precision etching of CAu^I_n (n < 6) clusters is important for elucidating the chemical and physical properties and reactivity of unsymmetric clusters. Therefore, the chemical etching method is a way to reduce the size of metal ion clusters and will be developed to control the electronic structure, asymmetrisation of the metal ion cluster structure, catalytic reactions, and metal ion exchange.

NMR spectra

¹H, ¹³C and 2D NMR spectra were measured on a Bruker AVANCE III-500 (500 MHz) spectrometer. The residual solvent signal was used to calibrate the ¹H (7.26 ppm), ¹³C NMR (77.16 ppm) measurements when CDCl₃ was used. The residual solvent signal was used to calibrate the ¹H (2.05 ppm), ¹³C NMR (29.84 ppm) measurements when d₆-acetone was used.

ESI-MS analysis

ESI-TOF-MS data were measured on a Micromass LCT Premier XE mass spectrometer. Unless otherwise noted, the experimental conditions were as follows (ion mode, positive; capillary voltage, 2400 V; sample cone voltage, 30 V; desolvation temperature, 150 °C; source temperature, 80 °C).

Elemental analysis

Elemental analyses (C, N, H) were conducted in the microanalytical laboratory, Department of Chemistry, School of Science, the University of Tokyo, using a Vario MICRO Cube elemental analyser with MgO added.

Single-crystal X-ray diffraction analysis

X-ray crystallographic analysis was performed using a Rigaku XtaLAB PRO MM007DW PILATUS diffractometer with MoKα and CuKα radiation (93 K), and the obtained data were calculated using the Crystal Structure crystallographic software package. The refinement was performed using an OLEX2 software⁵⁷ with SHELXT⁵⁸. All hydrogen atoms were geometrically placed and refined using the riding model.

Photophysical analysis

The UV-vis absorption spectra were measured on a JASCO V-770 UV-vis spectrophotometer, wherein the temperature is set at 293 K unless otherwise mentioned. The emission and excitation spectra were measured using a Jasco FP-8300 fluorometer. The absolute quantum yields and lifetime measurements in the solid state were measured using Quantaurus-QY (Hamamatsu C9920-02G) and Hamamatsu C11367-02, respectively. CD spectra were measured on a JASCO J-820 circular dichroism spectrometer. The experimental conditions of CD analysis were as follows: bandwidth, 1 nm; response, 0.5 s; data acquisition interval, 0.5 nm; scanning rate, 100 nm min^–1.

DFT and TD-DFT calculations

We applied the B3LYP functional⁵⁹ for geometry optimizations and TD-DFT calculations. The relativistic effective core potential LANL2DZ⁶⁰ was used for the Au atoms, and the basis set for the other atoms was 6-31G*⁶¹. Since MLCTs in the Au clusters etched in this study do not show long-distance charge transfer and essentially correspond to charge reorganisation, global hybrid functionals such as B3LYP adequately describe these electronic transitions. For simulating absorption spectra, 200 excited states were solved to cover the spectrum in the energy range up to about 220 nm was calculated in the velocity form. TD-DFT calculation was conducted with the polarizable continuum model (PCM) and the non-equilibrium linear response scheme⁶², including the solvent effect of CH₂Cl₂. All calculations were conducted using the Gaussian 16 suite of programs⁶³. The orbital composition was analysed using the Multiwfn program⁶⁴.

Handling

All the syntheses were conducted under air unless otherwise mentioned.

Chemical reagents

Unless otherwise noted, all the solvents were purchased from WAKO Pure Chemical Industries Ltd. and used without further purification. The >96% (NMR) pure cis-1,2-bis(diethylphosphino)ethene (cis-depe) was purchased from WAKO Pure Chemical Industries Ltd. and the >98% (GC) pure1,2-bis(diethylphosphino)benzene was purchased from TCI Co., Ltd., and used without further purification. The starting materials of carbon-centred hexagold(I) clusters [(C)(Au^I-SS-NHC)₆](BF₄)₂ (SS-1^NHC) [(C)(Au^I-RR-NHC)₆](BF₄)₂ (RR-1^NHC)⁴³ and [(C)(Au^I-TPP)₆](BF₄)₂ (1^TPP)²⁶ were synthesised according to the reported procedures.

Synthesis of [(C)(Au^I-SS-NHC)₅](BF₄) (SS-2^NHC) and [(C)(Au^I-RR-NHC)₅](BF₄) (RR-2^NHC)

To the solution of SS-1^NHC [(C)(Au^I-SS-NHC)₆](BF₄)₂ (6.7 mg, 2 μmol) in CH₂Cl₂ (1 mL), a solution of 2.5 equiv. of cis-depe (5 μmol, 2.0 mg) in CH₂Cl₂ (1 mL) was added dropwise at room temperature. The original pale-yellow solution turned orange immediately. Next, the resulting reaction mixture was concentrated to 0.3 mL using an evaporator and then filtered into a tube through the cotton, finally layered with 3 mL Et₂O for slow diffusion and stored in a refrigerator at 4 °C. After one day, orange-red blocky crystals of SS-2^NHC [(C)(Au^I-SS-NHC)₅](BF₄) were formed and isolated (4.3 mg, 80% yield, based on the hypercarbon). The RR-2^NHC [(C)(Au^I-RR-NHC)₅](BF₄) (4.6 mg, 85% yield, based on the hypercarbon) is obtained similarly by using RR-1^NHC [(C)(Au^I-RR-NHC)₆](BF₄)₂ (6.7 mg, 2 μmol) as starting material. Anal. calcd. for [C₁₁₆H₁₁₀Au₅BF₄N₁₀](CH₂Cl₂)₂: C, 49.11; H, 3.98; N, 4.85. Found: C, 49.20; H, 4.23; N, 4.96. ESI-MS (CH₂Cl₂, positive): [SS-2^NHC]⁺ [C₁₁₆H₁₁₀N₁₀Au₅]⁺, m/z 2628.85 (calcd. 2628.73). ESI-MS (CH₂Cl₂, positive): [RR-2^NHC]⁺ [C₁₁₆H₁₁₀N₁₀Au₅]⁺, m/z 2628.92 (calcd. 2628.73). ¹H NMR (500 MHz, 300 K, CDCl₃): δ 7.66–7.60 (m, 4H), 7.38 (q, J = 7.4 Hz, 2H), 7.16–7.12 (m, 2H), 7.03–7.01 (m, 2H), 7.00 (d, J = 1.3 Hz, 1H), 6.95 (t, J = 7.4 Hz, 3H), 1.73 (d, J = 7.3 Hz, 5H). ¹³C NMR (126 MHz, 300 K, CDCl₃): δ 206.1 (C_NHC), 198.8, 140.１, 133.2, 129.3, 128.6, 128.1, 123.8, 114.1, 59.0 (–CH–), 17.5 (–CH₃). In the ¹³C NMR spectrum, the signal of the hypercarbon atom was not detected even after a long-time accumulation.

Synthesis of [(C)(Au^I-TPP)₅](BF₄) (2^TPP)

The synthesis of 2^TPP is similar to that of SS-2^NHC by using 1^TPP [(C)(Au^I-TPP)₆](BF₄)₂ as the starting material. After adding a solution of cis-depe (25 μmol, 9.9 mg, 2.5 equiv.) in CH₂Cl₂ (1 mL) to a solution of 1^TPP (29.4 mg, 10 μmol) in CH₂Cl₂ (1 mL), the resulting mixture turned from colourless to yellow. Next, the resulting reaction mixture was concentrated to 0.3 mL using an evaporator, and then filtered into a tube through the cotton, finally layered with 3 mL Et₂O for slow diffusion and stored in a refrigerator at 4 °C. After several days, the yellow blocky crystals of 2^TPP were isolated (22.3 mg, yield 93%, based on the hypercarbon). Anal. calcd. for [C₉₁H₇₅Au₅BF₄P₅]: C, 45.63; H, 3.16; N, 0. Found: C, 45.61; H, 3.24; N, 0.22. ESI-MS (CH₂Cl₂, positive): [2^TPP]⁺ [C₉₁H₇₅P₅Au₆]⁺, m/z 2307.21 (calcd. 2307.29). ¹H NMR (500 MHz, 300 K, CDCl₃): δ 7.61–7.53 (m, 2H), 7.46–7.37 (m, 2H), 7.19 (td, J = 7.9, 1.8 Hz, 3H). ³¹P NMR (202 MHz, 300 K, CDCl₃): δ 32.59 (s); ¹³C NMR (126 MHz, 300 K, CDCl₃): δ 134.9 (d, J = 14.9 Hz), 133.0 (d, J = 50.3 Hz), 131.9, 129.9 (d, J = 11.4 Hz). In the ¹³C NMR spectrum, the signal of the hypercarbon atom was not detected even after a long-time accumulation.

Synthesis of complex [(cis-depe)₂Au^I](BF₄) (3)

In the above-mentioned synthesis of SS-2^NHC, after isolating the desired crystals of SS-2^NHC. The residue was used for recrystallisation, and several colourless crystals of [(cis-depe)₂Au^I](BF₄) (3) (yield 50% by ¹H NMR) were formed after one week. Similarly, after isolating crystals of 2^TPP from its crystallisation tube, wherein the residue was recrystallised to give 3 (yield 99%, by ¹H NMR). Its single crystal structure was determined by SCXRD (Suppl. Fig. 14). Its ¹H NMR spectrum is consistent with literature⁶⁵.

Synthesis of [(C)(Au^I-SS-NHC)₅(AuCl)](BF₄) (SS-4^NHC) and [(C)(Au^I-RR-NHC)₅(AuCl)](BF₄) (RR-4^NHC)

To a solution of SS-2^NHC (16 mg, 6 µmol) in 10 mL acetone, 1.2 equiv. of (tht)AuCl (2.3 mg, 7 µmol) in acetone (1 mL) was added dropwise (note: fast mixing caused decomposition and the formation of black precipitates) with continuous ultrasonic oscillation for 20 min. On completion, the resulting mixture was concentrated to 0.5 mL, and then filtered into a tube through the cotton, finally layered with 3 mL Et₂O for slow diffusion and stored in a refrigerator at 4 °C. After several days, dark blocky crystals of [(C)(Au^I-SS-NHC)₅(AuCl)](BF₄) SS-4^NHC was isolated (10.0 mg, 57% yield, based on the hypercarbon). Anal. calcd. for [C₁₁₆H₁₁₀Au₆BClF₄N₁₀](CH₃COCH₃)(H₂O): C, 47.26; H, 3.93; N, 4.63. Found: C, 46.90; H, 4.34; N, 4.86. ESI-MS (CH₂Cl₂, positive): [SS-4^NHC]⁺ [C₁₁₆H₁₁₀N₁₀ClAu₆]⁺, m/z 2860.71 (calcd. 2860.66). ESI-MS (CH₂Cl₂, positive): [RR-4^NHC]⁺ [C₁₁₆H₁₁₀N₁₀ClAu₆]⁺, m/z 2860.70 (calcd. 2860.66). ¹H NMR (500 MHz, 300 K, CDCl₃): δ 7.64 (d, J = 7.6 Hz, 4H), 7.38 (q, J = 7.3 Hz, 2H), 7.23–7.22 (m, 2H), 7.11–7.02 (m, 4H), 6.96 (t, J = 7.6 Hz, 4H), 1.62 (d, J = 7.3 Hz, 6H). ¹³C NMR (126 MHz, 300 K, CDCl₃): δ 189.93 (C_NHC), 139.8, 132.8, 129.4, 128.8, 128.0, 124.4, 114.6, 59.8 (–CH–), 17.3 (–CH₃). In the ¹³C NMR spectrum, the signal of the central carbon atom was not detected even after a long-time accumulation.

Kinetic studies of the etching process

Time-course experiments monitored by UV-vis spectroscopy

Etching [(C)(Au^I-SS-NHC)₆](BF₄)₂ (SS-1^NHC) with cis-depe. A solution of SS-1^NHC (5.0 ×10^–5 M) in CH₂Cl₂ was prepared by dissolving SS-1^NHC (1.7 mg, 0.5 µmol) in 10 mL CH₂Cl₂. We first measured this dichloromethane solution of SS-1^NHC (5.0 ×10^–5 M, 3 mL) by UV-vis spectroscopy at 293 K. Then, once 30 µL of cis-depe (2 equiv., 0.01 M) in CH₂Cl₂ was added (meanwhile, the solution was charged with a small magnetic stir and stirring at a rate of 60 rpm), the UV-vis spectra (Fig. 4a in the main text) of the resulting reaction were immediately measured at intervals of 1 s (stage A), and then at intervals of 5 min (stage B).

Etching [(C)(Au^I-TPP)₆](BF₄)₂ (1^TPP) with cis-depe. A solution of 1^TPP (5.0 ×10^–5 M) in CH₂Cl₂ was prepared by dissolving 1^TPP (1.5 mg, 0.5 µmol) in 10 mL CH₂Cl₂. We first measured this dichloromethane solution of 1^TPP (5.0 ×10^–5 M, 3 mL) by UV-vis spectroscopy at 293 K. Then, once 30 µL of cis-depe (2 equiv., 0.01 M) in CH₂Cl₂ was added (meanwhile, the solution was charged with a small magnetic stir and stirring at a rate of 60 rpm), the UV-vis spectra of the resulting reaction were immediately measured at intervals of 1 s (Suppl. Fig. 28).

Time-course experiments monitored by ¹H NMR spectroscopy

Etching [(C)(Au^I-SS-NHC)₆](BF₄)₂ (SS-1^NHC) with cis-depe. A solution of SS-1^NHC (2.0 ×10^–3 M) in d₆-acetone was prepared by dissolving SS-1^NHC (3.3 mg, 1 µmol) in d₆-acetone (0.5 mL) in the presence of an internal standard (IS, 1,3,5-trimethoxybenzene, 6 equiv., 6 µmol, 1.0 mg). We firstly measured the solution of SS-1^NHC in d₆-acetone by ¹H NMR spectroscopy at 300 K. Then, once the solution of cis-depe (2.5 equiv., 3.3 ×10^–2 M, 75 µL) in d₆-acetone was added, the ¹H NMR spectra of the resulting mixed sample were immediately measured and then measured continuously at intervals of approx. 2 min. The ¹H NMR spectra of this reaction were measured for 30 min (Suppl. Fig. 29).

Etching [(C)(Au^I-TPP)₆](BF₄)₂ (1^TPP) with cis-depe. A solution of 1^TPP (2.0 ×10^–3 M) in d₆-acetone was prepared by dissolving 1^TPP (2.9 mg, 1 µmol) in 0.5 mL d₆-acetone in the presence of an internal standard (IS, 1,3,5-trimethoxybenzene, 6 equiv., 6 µmol, 1.0 mg). We firstly measured the solution of 1^TPP in d₆-acetone ¹H NMR spectroscopy at 300 K. Then, once the solution of cis-depe (2.5 equiv., 3.3 ×10^–2 M, 75 µL) in d₆-acetone was added, the ¹H NMR spectra of the resulting mixed sample were immediately measured and then measured continuously at intervals of approx. 2 min. The ¹H NMR spectra of this reaction were measured for 18 min (Suppl. Fig. 30).

Time-course experiments monitored by ESI-MS spectrometry

Etching [(C)(Au^I-SS-NHC)₆](BF₄)₂ (SS-1^NHC) with cis-depe. A solution of SS-1^NHC (5.0 × 10^–5 M) in acetone was prepared by dissolving SS-1^NHC (1.7 mg, 0.5 µmol) in 10 mL acetone. We first measured the original solution of SS-1^NHC (5.0 × 10^–5 M, 1 mL) in acetone by ESI-MS spectrometry. Then, once 10 µL of cis-depe (2 equiv., 0.01 M) in acetone was added (meanwhile the solution was charged with a small magnetic stir and stirring), the ESI-MS spectra of the resulting reaction were measured as shown in Suppl. Fig. 31.

Etching [(C)(Au^I-TPP)₆](BF₄)₂ (1^TPP) with cis-depe. A solution of 1^TPP (5.0 × 10^–5 M) in acetone was prepared by dissolving 1^TPP (1.5 mg, 5.0 × 10^–4 mmol) in acetone. We first measured the original solution of 1^TPP (5.0 × 10^–5 M, 1 mL) in acetone by ESI-MS spectrometry. Then, once 10 µL of cis-depe (2 equiv., 0.01 M) in acetone was added (meanwhile the solution was charged with a small magnetic stir and stirring), the ESI-MS spectra of the resulting reaction were measured. The intermediate species was observed as shown in Suppl. Fig. 32.

Theoretical calculation details for etching mechanism

The proposed intermediates including Int1^NHC, Int2^NHC, Int1^TPP, and Int2^TPP were simulated and optimised using the Gaussian 16 suite of programs⁶³. Optimisation was performed using the B3LYP functional⁵⁹ combined with basis sets of 6-31G* (for C, N, P, H)⁶¹ and LANL2DZ (for Au)⁶⁰. The solvent effects were evaluated by single-point calculations in the optimised structure using the polarisable continuum model (PCM). All chemical species involved were optimised in the singlet state. The calculated energy profiles were illustrated in Suppl. Fig. 33.