Atomically Precise Fcc-Amorphous Homometal Heterojunction with ∼1 nm Size

Abstract

The emerging of ultrasmall gold nanoparticles (nanoclusters) with atomic precision provides opportunities for precisely studying crystalline–amorphous heterostructures, despite the construction of such structures being challenging. In this work, we developed an acid-induction method and synthesized a Au₅₂(TBBT)₃₀ (TBBTH = 4-tert-butylbenzenelthiol) nanocluster with the kernel composed of two parts: the amorphous Au₂₂ part and the fcc Au₂₁ part, which represents the first construction of fcc-amorphous homometal heterojunction with ∼1 nm size. Density function theory (DFT) revealed that the HOMO–LUMO majorly distributed in the amorphous part and the HOMO–LUMO gap was dominated by the amorphous part, indicating the redox activity of the amorphous Au₂₂ part in contrast to the fcc Au₂₁ part, which was experimentally confirmed by differential pulse voltammetry, antioxidation test and anti-Galvanic reaction. But for electro-catalyzing reduction of CO₂ to CO, the crystalline surface sites were revealed to be more catalytically active than the amorphous surface sites in catalyzing the reduction of CO₂ to CO, and the most active sites were assigned to the cosurface sites of amorphous Au₂₂ and fcc Au₂₁, which is also responsible for the high performance of Au₅₂(TBBT)₃₀ relative to the pure fcc-structured Au₅₂(TBBT)₃₂ (the highest CO FE: 96.7% at −0.67 V vs 73.3% at −0.57 V; CO partial current density at the corresponding potential: −7.3 vs −2.7 mA cm^–2).

Since the discovery of amorphous structures over 60 years ago, , amorphous-forming mechanisms and practical applications have been actively studied. − Amorphous phases possess a large amount of bonds, which are randomly oriented with unsaturated electronic configurations relative to their crystalline counterparts. , Moreover, the superior elastic strain of the local structure in the amorphous phase can accelerate the charge transfer between the active center and the reaction intermediate. These advantages endow amorphous materials with good flexibility and abundant defect sites on the surface, potentially active for catalysis. , However, every coin has two sides: the inherent low conductivity of the amorphous phase does not facilitate catalysis. In contrast, the crystalline phases usually improve the catalytic performance through their contribution to the high conductivity of the materials. , Hence, crystalline–amorphous combined heterostructures have recently received extensive attention because such materials are expected to shed the separated disadvantages of crystalline and amorphous components, and have combined or synergetic merits of the two components. ,, However, it is difficult to characterize such materials in bulk or large nanosize with atomic precision, which retards the in-depth structure (composition)-property correlation. , The emergence of metal nanoclusters − in recent years provides opportunities for precise study of such materials at ∼1 nm scale despite the construction and characterization of them being challenging. Over a dozen fcc-structured metal nanoclusters − have been known, and recently an amorphous Au–Cd nanocluster was also revealed; however, metal nanoclusters composed of coexistent fcc and amorphous parts have not been reported thus far, to the best of our knowledge, which stimulates our research enthusiasm. Fortunately we synthesized such a nanocluster by developing an acid-induction method, and probed its various active locations for electrocatalytic reduction of CO₂ to CO, which will be shown below.

The synthesis details are described in the Supporting Information. Briefly, tetraoctylammonium bromide (TOAB) was first dissolved into THF in a trineck flask; then, HAuCl₄·4H₂O was added. After the mixed solution was stirred with 600 rpm for 10 min, 4-tert-butylbenzenelthiol (TBBTH) was added. When the color of the solution changed to yellow from deep red, a cold-water solution with NaBH₄ was immediately added. In rapid sequence, dilute nitric acid (50%, v/v) was added into the reaction mixture. Without the addition of an adequate amount of nitric solution, the target nanocluster cannot be obtained after trying various reaction conditions, which indicates the critical role of nitric acid in this reaction. Hydrogen ions from nitric acid exhibit at least two effects: On one hand, nitric acid can accelerate the reaction by enhancing the hydrolysis of NaBH₄; on the other hand, it can reduce the etching ability of the thiolate by interacting with thiol group, thus influencing the reaction kinetics and thermodynamics. − The growth process of gold nanoclusters was maintained for 5 h. Preparative thin-layer chromatography (PTLC) was applied to purify the crude product, which was collected and thoroughly washed with methanol. Dark crystals grew out of mixed toluene and acetonitrile under 4 °C after 1 week. The as-obtained crystals dissolved in dichloromethane were characterized with the UV/vis/NIR spectrometer, which reveals two absorption peaks at 415 and 533 nm (Figure a). Electrospray ionization mass spectrometry (ESI-MS) was employed to determine the nanocluster’s composition. − Cesium acetate (Cs­(OAc)₂) was added to assist the nanocluster ionization. As shown in Figure b, two intense peaks were observed at 5200.36 and 7733.40 M/Z, which correspond to the compositions [Au₅₂(TBBT)₃₀Cs₃]³⁺ (calculated: 5200.41 M/Z, deviation: 0.05 M/Z) and [Au₅₂(TBBT)₃₀Cs₂]²⁺ (calculated: 7733.49 M/Z, deviation: 0.09 M/Z), respectively. The experimental isotopic patterns were in accordance with the simulated isotopic pattern (see Figure c and d). The charge of the nanocluster should be neutral, since the adduct charge number is equal to the adducted Cs⁺ number, and no sound signals were found in the mass spectra without the addition of Cs­(OAC)₂. Therefore, the nanocluster should be Au₅₂(TBBT)₃₀, which was further confirmed by X-ray photoelectron spectroscopy (XPS) and single-crystal X-ray crystallography (SCXC). As shown in Figure S1, Au, S and C elements are observed in the survey spectrum and high-resolution spectra of Au₅₂(TBBT)₃₀. The peaks of Au 4f_7/2 and 4f_5/2 appear at 84.28 and 87.98 eV (Figure S1b), respectively, and the peak-spliting and fitting (Figure S2) clearly demonstrate the presence of Au (0). Note that the newly obtained Au₅₂(TBBT)₃₀ and the reported Au₅₂(PET)₃₂ (PET = phenylethanethiolate) and Au₅₂(TBBT)₃₂ are heterocomposition-homosized nanoclusters, and they have various compositions/structures, indicating the diversity in composition/structure of metal nanoclusters. However, the non-fcc Au₅₂(TBBT)₃₀ nanocluster cannot be produced from fcc Au₅₂(TBBT)₃₂ by reacting it with nitric acid (Figure S3).

1.

UV/vis/NIR absorption (a) and ESI-MS (b) spectra of Au₅₂(TBBT)₃₀; high-resolution mass spectra for [Au₅₂(TBBT)₃₀Cs₃]³⁺ (c) and [Au₅₂(TBBT)₃₀Cs₂]²⁺ (d) in constrast with the simulated one.

The crystals adopt the triclinic P1̅ space group. Figure S4 shows the total structure of Au₅₂(TBBT)₃₀ nanocluster revealed by SCXC, which is composed of a crystalline–amorphous Au₃₈ kernel and an exterior shell with four -SR-Au-SR-Au-SR- staples, six -SR-Au-SR- staples and one TBBT thiolate. As shown in Figure , the Au₃₈ kernel of Au₅₂(TBBT)₃₀ could be divided into two parts.

2.

Anatomy of the crystalline–amorphous structure of the Au₅₂(TBBT)₃₀ nanocluster: (a) archimedean antiprism capped with two half-octahedra; (b) quasi-icosahedral 12-atom unit; (c) Au₁₆ unit; (d) octahedron Au₆ unit; (e) Au₁₉ unit; (f) three amorphous atoms; (g) amorphous structure of Au₂₂ unit; (h) fcc-Au₂₁ unit; (i) crystalline–amorphous Au₃₈ kernel. All atoms are Au.

The first part is composed of 22 gold atoms (Figure ), which can be considered as an amorphous structure. To better understand this amorphous structure, more detailed analyses are given as follows. An archimedean antiprism is capped with two half-octahedra is formed by ten Au atoms (Figures a and S5), which is wrapped by a quasi-icosahedron composed of 12 atoms (Figure b) with sharing six Au atoms (Figure c). Meanwhile, an octahedron (Figure d) is connected to the Archimedean antiprism capped with two half-octahedra by sharing an Au₃ facet (Figure e). There are also three amorphous atoms (Figure f), which are connected between the quasi-icosahedral Au₁₂ and the archimedean antiprism capped with two half-octahedrons and between quasi-icosahedral Au₁₂ and octahedron (Figure g), respectively. Note that the archimedean antiprism capped with two half-octahedra, an icosahedron (quasi-icosahedron), and an octahedron which are the basic structural units of an amorphous model structure. Therefore, the Au₂₂ part is regarded as the amorphous structure. The second part consists of twenty-one gold atoms packed in the fcc mode, and its top and bottom are enclosed by two (100) faces (Figure h). The amorphous part connected the bottom (100) face and a Au₃ facet of the fcc part by sharing five Au atoms (Figure c and g). The five shared atoms (atoms in the red frame) form the novel crystalline–amorphous interface (Figure i), which was first found in gold nanoclusters.

For the protective layer, there are six monomeric staples (-RS-Au-SR-) (Figure S6b), four dimeric staples (-RS-Au-SR-Au-SR-) (Figure S6d) and six bridging thiolates (Figure S6f): one monomeric staple, one dimeric staple and five bridging thiolates adsorb on the fcc-structured Au₂₁ (Figure S6c, e, and g); two monomeric staples, two dimeric staples and one bridging thiolate are anchored to the Au₂₂ amorphous part (Figure S6e and g). Three monomeric staples and one dimeric staple bridge the fcc-structured Au₂₁ and the Au₂₂ amorphous part (Figures S6 and c). Compared with pure fcc structure, the amorphous structure needs more monomeric staples (-RS-Au-SR-) for protection compared with dimeric staples (-RS-Au-SR-Au-SR-) as shown in Table S1, probably due to the relative unstability of the amorphous structure. Thus, obtaining an amorphous structure might require the conditions where the formation of a monomeric staple is preferred. For example, acid can break the AuSR polymers and benefit the synthesis of the amorphous structure, which is evidenced by this work and the previous work.

The Au–Au bond lengths in the Au₃₈ kernel are 2.680–3.742 Å, and the average bond length is 2.920 Å, which is similar to the Au–Au distance (2.883 Å) in bulk Au. The Au–S bond lengths of the staples are 2.228–2.381 Å, and the averaged bond length of 2.301 Å is slightly shorter than the Au–S bond length of the bridging thiolates (2.348 Å). The S–Au–S bond angles of the staples are 165.96–176.55° with an averaged value of 171.28°, while the Au–S–Au angles of the staples are 89.54–105.48° with an averaged value of 98.83°.

An advantage of metal nanocluster research is that their electronic structure can be obtained by computing in the same manner as that molecular compounds are treated. − Since many physicochemical processes, especially the redox process, is related to the HOMO–LUMO, it is conceived that the HOMO–LUMO information might provide clues for searching the redox active location of gold nanoclusters. − To test the concept, theoritical calculations (see Supporting Information for the details) were conducted on the novel Au₅₂(TBBT)₃₀ nanocluster and some other existing nanoclusters. As shown in Figures and S7, the HOMO and LUMO are majorly distributed on the Au₂₂ amorphous part in the kernel of Au₅₂(SR)₃₀, which is also quantitatively supported by the Hirshfeld method in the Multiwfn program (for the average contribution of every gold atom in various locations, see Table S2). The HOMO–LUMO gap differences between two pairs of nanoclusters with close sizes but different structures (Au₃₆(SR)₂₄ vs Au₃₈(SR)_24‑T, Au₅₂(SR)₃₂ vs Au₅₂(SR)₃₀) are essentially the same, and their values are 0.76 and 0.75 eV, respectively, indicating that the non-fcc part (icosahedral Au₁₃ or amorphous part) dominates the HOMO–LUMO gap (functional group-like), which is also supported by the fact that the HOMO–LUMO gap difference between Au₄₄(SR)₂₈ and Au₄₉(2,4-DMBT)₂₇ is 0.52 eV (deviated from the above-mentioned HOMO–LUMO gap difference of 0.76 or 0.75 eV) when icosahedral Au₁₃ or the amorphous part is not contained in the dual-packed-kernelled Au₄₉(2,4-DMBT)₂₇. These facts indicate that the amorphous part is the redox active location in the kernel of Au₅₂(SR)₃₀, which is experimentally supported by the electrochemical results. As shown in Figure S8, differential pulse voltammetry (DPV) reveals that fcc-structured gold nanoclusters have larger electrochemical gaps (EG) than those of icosahedral Au₁₃-based gold nanoclusters with close sizes. For examples, fcc-structured Au₂₈ has a larger EG than that of icosahedral Au₁₃-based Au₂₅ , (1.96 vs 1.60 V), and fcc-structured Au₃₆ has a larger EG than that of bi-icosahedral Au₃₈ (2.00 vs 1.18 V). For crystalline–amorphous Au₅₂(TBBT)₃₀, the EG of DPV is 0.91 V, far smaller than the 1.61 V EG for Au₅₂(TBBT)₃₂, indicating that the Au₂₂ amorphous part dominates the EG of DPV in Au₅₂(TBBT)₃₀, which provides experimental support for the speculation that the amorphous Au₂₂ part is the active location (functional group) in Au₅₂(TBBT)₃₀ for the redox reaction. To find more evidence, we examined the reactivity of Au₅₂(TBBT)₃₀ and found that crystalline–amorphous Au₅₂(TBBT)₃₀ can be turned to pure-fcc Au₅₂(TBBT)₃₂ and Au₃₆(TBBT)₂₄ when oxidized by oxygen and Cd­(NO₃)₂, respectively (Figures S9 and S10). When 2 mg of NaBH₄ (3 equivs) is added into the Au₅₂(TBBT)₃₀ solution, Au₅₂(TBBT)₃₀ also can be transformed into the pure-fcc structured Au₃₆(TBBT)₂₄ (Figure S11). When Au₅₂(TBBT)₃₂ is exposed to air for 2 days, its structure remains unchanged (Figure S12). Similarly, no structural transformation occurs when Au₃₆(TBBT)₂₄ reacts with Cd²⁺ (Figure S13). Given the structure differences between Au₅₂(TBBT)₃₀ and Au₅₂(TBBT)₃₂ (or Au₃₆(TBBT)₂₄ ), it can be concluded that the fcc-Au₂₁ unit in the kernel of Au₅₂(TBBT)₃₀ does not change, but the amorphous Au₂₂ unit turns to a fcc structure in the transformation from Au₅₂(TBBT)₃₀ to Au₅₂(TBBT)₃₂ (or Au₃₆(TBBT)₂₄) (Figure S14), indicating the activity of the amorphous Au₂₂ unit in the kernel of Au₅₂(TBBT)₃₀.

3.

HOMO and LUMO of three groups of gold nanoclusters with comparable sizes but different structures (Au₃₆(SR)₂₄, Au₃₈(SR)₂₄-T, Au₃₈(SR)₂₄-Q; Au₄₄(SR)₂₈ and Au₄₉(SR)₂₇; Au₅₂(SR)₃₂ and Au₅₂(SR)₃₀).

To investigate the difference between the amorphous and crystalline units in catalysis, the electrocatalytic reduction of CO₂ was used as a model reaction due to its significance and popularity. ,− For comparison, the pure-fcc structured Au₅₂(TBBT)₃₂ was also investigated. As shown in the LSV curves of Figure a, both Au₅₂(TBBT)₃₀ and Au₅₂(TBBT)₃₂ nanoclusters show higher current density in CO₂-saturated 0.5 M KHCO₃ solution than that in the Ar-saturated electrolyte, illustrating the occurrence of CO₂ reduction on these two Au nanoclusters. The observed much higher current density on Au₅₂(TBBT)₃₀ than that on Au₅₂(TBBT)₃₂ in CO₂-saturated electrolyte suggests that Au₅₂(TBBT)₃₀ is more active for the reduction of CO₂ molecules than Au₅₂(TBBT)₃₂. As clearly observed in Figure b, Au₅₂(TBBT)₃₀ shows higher CO FE than Au₅₂(TBBT)₃₂ in the whole selected potential range. Au₅₂(TBBT)₃₀ exhibits the highest CO FE of 96.7% at −0.67 V, while Au₅₂(TBBT)₃₂ displays the highest CO FE of 73.3% at −0.57 V. The higher CO FE results in larger CO partial current density, as seen in Figure c. Specifically, the CO partial current density on Au₅₂(TBBT)₃₀ at −0.67 V reaches −7.3 mA cm^–2, which is much larger than −2.7 mA cm^–2 on Au₅₂(TBBT)₃₂ at −0.57 V. Besides, the mass activty of Au₅₂(TBBT)₃₀ at −0.67 V reaches −281.1 mA mg^–1, which is 2.5 times that of Au₅₂(TBBT)₃₂ (−112.9 mA mg^–1) at −0.67 V (Figure S15a). The higher CO FE and the relevant CO current density and mass activity of Au₅₂(TBBT)₃₀ compared with Au₅₂(TBBT)₃₂ indicate that the crystalline–amorphous structure can more efficiently catalyze the reduction of CO₂ to CO than the pure crystalline structure. The high catalytic selectivity of Au₅₂(TBBT)₃₀ for CO production also suppresses the competitive hydrogen evolution reaction, and the corrresponding H₂ FE of Au₅₂(TBBT)₃₀ at −0.67 V is only 3.1% (Figure S15b). Moreover, the nearly unchanged current density and CO FE on Au₅₂(TBBT)₃₀ (Figure d) during electrolysis for 20 h verify its favorable electrocatalytic durability for CO₂ reduction. Both TEM (Figure S16) and XRD (Figure S17) experiments further verify the structure stability of Au₅₂(TBBT)₃₀ during the CO₂ reduction process.

4.

Linear sweep voltammetry (LSV) curves of Au₅₂(TBBT)₃₀ and Au₅₂(TBBT)₃₂ in Ar-saturated (dotted line) and CO₂-starated (solid line) 0.5 M KHCO₃ solution with a scan rate of 10 mV s^–1 (a); CO Faradaic efficiencies (FE) of these two nanoclusters at different applied potentials (b); corresponding CO partial current density (c); electrocatalytic stability of Au₅₂(TBBT)₃₀ nanocluster for the reduction of CO₂ to CO at −0.67 V for 20 h (d).

In situ surface enhanced infrared absorption spectroscopy in attenuated total reflectance mode (ATR-SEIRAS) was employed to investigate the adsorption behavior of the crucial intermediates on theses two nanoclusters. As shown in Figure a and b, the appreciable bands centered at 1925 cm^–1 for Au₅₂(TBBT)₃₀ and centered at 1905 cm^–1 for Au₅₂(TBBT)₃₂ should be attributed to the bridge-bonded CO* on their surfaces, and the relatively positive shift of the adsorbed band for Au₅₂(TBBT)₃₀ indicates that the binding energy of CO on Au₅₂(TBBT)₃₀ is weaker than that on Au₅₂(TBBT)₃₂. , Moreover, the density functional theory (DFT) calculations were performed to survey the efficient active sites on these two nunoclusters, and to gain insight into the mechanism of the reduction of CO₂ on these sites. The given thiolates were removed to expose the active sites for the adsorption of CO₂ molecules. − For Au₅₂(TBBT)₃₀, the bridging thiolates in six -Au-SCH₃-Au- motifs on the FCC unit surface, the amorphous unit surface and the cosurface of the amorphous and fcc units (Figure S18) were removed to expose Au_52s−S1 and Au_52s−S2 (Figure S19a, d), Au_52s−S3 and Au_52s−S4 (Figure S20a, d), and Au_52s−S5 and Au_52s−S6 (Figure S21a, d) sites, respectively. For the Au₅₂(TBBT)₃₂ nanocluster, two bridging thiolates in two -Au-SCH₃-Au- motifs on the FCC unit surface (Figure S22) were removed to expose Au₅₂–S1 and Au₅₂–S2 sites (Figure S23a, d). The free energies of every elementary reaction in CO₂ reduction on Au_52s–S1-S6 and Au₅₂–S1-S2 are shown in Figures S24 and S25, respectively. It is clear that the formation of COOH*, one key intermediate usually hindered by the weak adsorption, is facile on the investigated sites of the two clusters (Figures S19b and e, S20b and e, S21b and e, S23b and e, S24, and 25). However, the desorption of CO*, another important intermediate, from these investigated sites shows an increase of free energy (Figure S19c and f, S20c and f, S21c and f, S23c and f, S24, and S25), indicating the liberation of CO should be a rate-determining step. For the Au₅₂(TBBT)₃₀ nanocluster, Au_52s−S6 shows the smallest increase of free energy for the desorption of CO* on these potential catalysis sites (Figure S24), indicating that Au_52s−S6 is the most active site on Au₅₂(TBBT)₃₀ for catalyzing the reduction of CO₂ to CO; i.e., the cosurface of the crystalline and amorphous units are more catalytically active than the single crystalline (amorphous) unit surface, which was attributed to the synergy of the crystalline and amorphous units. The increase of free energy for the desorption of CO* on Au_52s−S2 is smaller than that on Au_52s−S3 (Figure S24), indicating that the crystalline surface sites are more catalytically active than the amorphous surface sites. For Au₅₂(TBBT)₃₂, Au₅₂–S2 shows a smaller increase of free energy in the desorption of CO* than Au₅₂–S1 (Figure S25), indicating that Au₅₂–S2 is more active than Au₅₂–S1. Notably, the increase of free energy for the desorption of CO* on Au_52s−S6 is smaller than that on Au₅₂–S2 (Figures c, S21f, and S23f), accounting for the better electrocatalytic activity of Au₅₂(TBBT)₃₀ than that of Au₅₂(TBBT)₃₂. As shown in Figure e and f, Au_52s−S6 shows higher affinity than Au₅₂–S2, indicating the weaker Au-CO* bond on Au_52s−S6 than that on Au₅₂–S2, in accordance with the above-mentioned free energy results. Besides, the density of states (DOS) calculation also provides evidence: the d-band center of Au₅₂–S2 is more upshifted than that of Au_52s−S6 (Figure S26), indicating the relatively stronger interaction between the active gold centers and the adsorbates (CO*) for Au₅₂–S2, which retards the liberation of CO and thus inhibits the catalysis on Au₅₂(TBBT)₃₂, which is also consistent with the ATR-SEIRAS result. In addtion, the larger resistance for the desorption of H* on Au_52s−S6 in comparison with that on Au₅₂–S2 (Figures d and S27) further supports that Au₅₂(TBBT)₃₀ can more efficiently catalyze the reduction of CO₂ to CO, demonstrating the importance of the active site on the joint surface in metal nanoclusters.

5.

In situ electrocatalytic ATR-SEIRAS spectra for the reduction of CO₂ on Au₅₂(TBBT)₃₀ (a) and Au₅₂(TBBT)₃₂ (b) in CO₂ saturated 0.5 M KHCO₃ solution at various applied potentials (the potential interval between every two adjacent spectra is 0.1 V). Free energy diagram for the reduction of CO₂ to CO (c) and HER (d) on Au_52s−S6 of Au₅₂(TBBT)₃₀ and Au₅₂–S2 of Au₅₂(TBBT)₃₂. Charge density differences for CO* adsorption states and the corresponding charge transforming of Au₅₂–S2­(e) and Au_52s−S6 (f).

In summary, a novel crystalline–amorphous heterojunction gold nanocluster Au₅₂(TBBT)₃₀ was synthesized by developing an acid-induction method and structurally resolved by single-crystal X-ray crystallography, which reveals its unique kernel consisting of two parts: the Au₂₂ amorphous part and the fcc-structured Au₂₁ part, demonstrating the first construction of an ∼1 nm fcc-amorphous heterojunction with atomic precision. The amorphous Au₂₂ and fcc Au₂₁ parts show various reactivity, indicated by DFT calculations, which reveals that the HOMO–LUMO majorly distributes on the Au₂₂ amorphous part in the kernel and the HOMO–LUMO gap is dominated by the Au₂₂ amorphous part. The high redox activity of the amorphous Au₂₂ part compared to the fcc Au₂₁ part was experimentally supported by differential pulse voltammetry, antioxidation test, and AGR reaction. However, the crystalline surface sites are more catalytically active than the amorphous surface sites in catalyzing the reduction of CO₂ to CO, and the most active sites are found to locate on the cosurface of the amorphous Au₂₂ and fcc Au₂₁ parts which is attributed to the synergy of the two parts. This is also supported by the fact that fcc-amorphous heterojunctioned Au₅₂(TBBT)₃₀ can far more efficiently catalyze CO₂RR than the pure fcc-structured Au₅₂(TBBT)₃₂ (CO FE: 96.7% vs 73.3%; CO partial current density: 7.3 vs 3.5 mA cm^–2). Thus, our work represents an important advance in atomically precise heterojunction synthesis, characterization and structure–property correlation, having important implications for nanoparticle tuning and active location probing, and is expected to stimulate more work in related fields.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/prechem.5c00006.

• Additional experimental details, materials, and methods, including photographs of experimental setup (PDF)

• Crystallographic depiction of Au₅₂(TBBT)₃₀ (CIF)

¶.

S.Z., D.C., and P.W. contributed equally. All authors have given approval to the final version of the manuscript.

The authors declare no competing financial interest.