A reasonable approach for the generation of hollow icosahedral kernels in metal nanoclusters

Although the hollow icosahedral M12 kernel has been extensively observed in metal nanoclusters, its origin remains a mystery. Here we report a reasonable avenue for the generation of the hollow icosahedron: the kernel collapse from several small nano-building blocks to an integrated hollow icosahedron. On the basis of the Au alloying processes from Ag28Cu12(SR)24 to the template-maintained AuxAg28-xCu12(SR)24 and then to the template-transformed Au12CuyAg32-y(SR)30, the kernel evolution/collapse from “tetrahedral Ag4 + 4∗Ag3” to “tetrahedral Au4 + 4∗M3 (M = Au/Ag)” and then to “hollow icosahedral Au12” is mapped out. Significantly, the “kernel collapse” from small-sized nano-building blocks to large-sized nanostructures not only unveils the formation of hollow icosahedral M12 in this work, but also might be a very common approach in constructing metallic kernels of nanoclusters and nanoparticles (not limited to the M12 structure).

Of all reported nanoclusters with precise structures, the icosahedral configuration is the most typical, which is frequently observed in both metal kernels and ligand shells of nanoclusters [47][48][49][50][51] . Interestingly, except for the non-hollow icosahedral M 1 @M 12 kernel (M represents the metal), the hollow icosahedral M 12 kernel has also served as a basic nano-building block of nanoclusters (e.g., Ag 44 (SR) 30 , Au 12+n Cu 32 (SR) 30+n , Ag 50 (dppm) 6 (SR) 30 , Au 144 (SR) 60 , etc.) 10,11,[52][53][54][55][56] . Structurally, it is accepted that the non-hollow icosahedron might be more energetically favorable than the corresponding hollow one due to the extra 12 metal···metal interactions in M 1 @M 12 ; accordingly, the hollow icosahedral kernel is unlikely to arise in the initial stage of the nanocluster growth. Besides, the hollow M 12 kernel is also less likely to originate from its non-hollow counterpart because the 12 metal···metal interactions make it difficult to extract the innermost metal atom out. In this context, the origin of such hollow icosahedral kernels remains a mystery.
In this work, based on the Au-alloying-induced nanocluster transformation from M 40 (SR) 24 to M 44 (SR) 30 (M = Au/Ag/Cu), a reasonable avenue for the generation of hollow icosahedral M 12 kernels has been mapped out, i.e., the kernel collapse from several small nano-building blocks to an integrated hollow icosahedron. The proposed avenue might serve as a common approach in constructing metallic kernels of nanoclusters and nanoparticles (not limited to the M 12 structure).
Besides, along with the Au-alloying process, the thermal stability of nanoclusters was enhanced. As shown in Supplementary Fig. 12, UV-vis characteristic absorptions of the Ag 28 Cu 12 (SR) 24 nanocluster (dissolved in CH 2 Cl 2 ) gradually decreased in intensity after 1 h and completely disappeared in~4 h, indicating degradation. In contrast, the UV-vis absorptions of Au x Ag 28-x Cu 12 (SR) 24 (x = 1.32) were essentially identical in the first 2 h, and gradually decreased as time went on. Of note, the optical absorptions of Au x Ag 28-x Cu 12 (SR) 24 (x = 7.56) was almost retained within 24 h, suggesting the enhanced thermal stability of Au x Ag 28-x Cu 12 (SR) 24 (x = 7.56) over other two M 40 (SR) 24 nanoclusters. In this context, the sequence of the thermal stability of these three M 40 (SR) 24 nanoclusters was determined as Au x Ag 28-x Cu 12 (SR) 24 (x = 7.56) > Au x Ag 28-x Cu 12 (SR) 24 (x = 1.32) > Ag 28 Cu 12 (SR) 24 ; that is, increasing the Au-doping amount in nanoclusters was in favor of preparing M 40 (SR) 24 with higher thermal stability.
Kernel transformation from tetrahedron to hollow icosahedron. Figure 3 depicts the kernel collapse from the "tetrahedral Au 4 + 4*M 3 " to "hollow icosahedral Au 12 " induced by the Au alloying. Specifically, the initial Au-doping process transported the Au heteroatoms to the tetrahedral kernel, converting the Ag 4 kernel to the alloyed Au x Ag 4-x and the final Au 4 (Fig. 3a). The further Aualloying sites on M 40 -S2 predominantly located at the four M 3 triangles that adhered to a vertex-to-face relationship to the tetrahedral Au 4 kernel ( Fig. 3b and Supplementary Fig. 3c); in contrast, the other four triangles on M 40 -S2, following a face-to-face relationship to the Au 4 tetrahedral kernel, maintained unalloyed as Ag 3 ( Supplementary   Fig. 3d). For easily distinguishing these M 3 positions, we define these Au 3 positions as "stable location" (Supplementary Fig. 3c) and "unstable location" (Supplementary Fig. 3d). However, the Au doping on stable locations is simply concluded from the crystallography, and the Au positions may change throughout the crystallization process. From ESI-MS results ( Supplementary Fig. 14), a maximum of [18][19] Au heteroatoms could be doped into the M 40 cluster framework, >16 positions from the M 4 kernel and 4*M 3 stable locations; accordingly, there are other Ag positions in M 40 that could be occupied by the introduced Au. X-ray absorption fine structure spectroscopy (XAFS) measurements were then performed for grasping the in-situ Audoping process (Supplementary Figs. 29 and 30 and Supplementary  Tables 3-4). The XAFS results demonstrated that the introduced Au occupied the innermost M 4 tetrahedron first, and then substituted the Ag atoms in unstable locations, different from the crystal results wherein the unstable locations were maintained as undoped Ag throughout. We further crystallized this cluster sample and the crystal data suggested the Au heteroatoms on stable locations (i.e., Au x Ag 28x Cu 12 (SR) 24 , x = 7.76), demonstrating the intracluster Au-Ag metal exchange throughout the crystallization. In this context, we made some speculations on mass signals (Supplementary Fig. 14): the introduced Au heteroatoms occupied the innermost tetrahedron first, and then substituted Ag atoms on M 40 -S2 randomly; the mass signals i represented the dominant Au-occupation in stable locations, whereas the signals ii represented the unstable locations, resulting in two groups of signals in the 3-min mass spectrum ( Supplementary  Fig. 14). In the 3-min sample, the M 40 with Au-occupation in unstable locations might be the main product by referring XAFS results. Then, the M 40 clusters of signals ii would transform to M 44 clusters of signals iii, and then decomposed due to their instability. By comparison, the M 40 clusters of signals i were continually doped by Au and transformed to M 44 clusters of signals iv finally. In this context, the driving force for the transformation from M 40 to M 44 was determined as the Au-alloying at unstable locations, which rendered the M 40 nanoclusters unstable molecules and triggered the kernel collapse from several small nano-building blocks to an integrated hollow icosahedron.
Significantly, the further Au-alloying induced the transformation from M 40 (SR) 24 to M 44 (SR) 30 , among which process the hollow icosahedral Au 12 was generated (Fig. 3b, c). Structurally, the pre-transformed Au x Ag 28-x Cu 12 (SR) 24 possesses a "tetrahedral Au 4 + 4*M 3 " kernel (M = Au/Ag with a high Au proportion). Upon the nanocluster conversion, the M 3 triangles collapsed inward to the Au 4 tetrahedron, and finally rearranged into the hollow icosahedral Au 12 kernel in Au 12 Cu y Ag 32-y (SR) 30 (Fig. 3c). Of note, there are 16 metal atoms in the "tetrahedral M 4 + 4*M 3 " kernel while the icosahedral kernel only contains 12 metal atoms; in this context, a structural rearrangement occurred in this structural and kernel transformation (indeed, the "kernel +surface" configurations between M 40 (SR) 24 and M 44 (SR) 30 nanoclusters are different). However, due to the existence of several isoabsorption points in the UV-vis spectra, the structure transformation from M 40 (SR) 24 to M 44 (SR) 30 should follow an "intramolecular rearrangement" approach, but not an "intermolecular decomposition-recombination" approach. Accordingly, it is reasonable to conjecture the formation of icosahedral M 12 in M 44 (SR) 30 as the kernel collapse from "tetrahedral Au 4 + 4*M 3 ". Besides, all sites in the hollow icosahedron are fully occupied by Au (i.e., Au 12 ); in vivid contrast, the non-hollow M 1 @M 12 kernels of previously alloy clusters are always partially occupied by two or more types of metals. We proposed that the complete Au occupation of the hollow icosahedron resulted from the kernel collapse in which process only the collapse of Au atom to Au 4 was the most energetically favorable.
This avenue (i.e., kernel collapse) is of great importance since it maps out a reasonable avenue for the generation of the hollow icosahedral M 12 kernel in metal nanoclusters. Besides, the kernel collapse might be a very common approach in constructing metallic kernels of nanoclusters and nanoparticles (not limited to the hollow icosahedron, but also compliant to other configurations such as non-hollow icosahedron, FCC/BCC kernels, etc.), because the routine growth of several large-sized nanoclusters shell-by-shell should be not that energetically favorable. We also note that the kernel collapse should not be the unique approach for the generation of hollow icosahedra (or other structures) in metal nanoclusters and nanoparticles; other approaches may also exist and are still worth mapping out.

Discussion
In summary, on the basis of the Au-alloying-induced transformation from M 40 (SR) 24 to M 44 (SR) 30 (M = Au/Ag/Cu), a reasonable avenue-kernel collapse-for the generation of the hollow icosahedral M 12 kernel in metal nanoclusters has been mapped out. The Au alloying on Ag 28 Cu 12 (SR) 24 produced template-maintained Au x Ag 28-x Cu 12 (SR) 24 (x = 1.32), Au x Ag 28x Cu 12 (SR) 24 (x = 7.56), and template-transformed Au 12 Cu y Ag 32y (SR) 30 (y = 3.74) step by step, accompanying with which processes the cluster kernel stepwisely evolved from "tetrahedral Ag 4 + 4*Ag 3 " to "tetrahedral Au 4 + 4*Ag 3 ", then to "tetrahedral Au 4 + 4*Au 3 ", and finally to "hollow icosahedral Au 12 ". The entire process was tracked by ESI-MS, and the crystal structures of the key nodes (altogether five crystal structures) have been determined. Overall, this work presents a reasonable avenue for comprehending the generation of hollow icosahedra in metal nanoclusters, and the "structure collapse" might be a very common approach for constructing kernel structures (not limited to the hollow icosahedron) in the size growth of nanoclusters and nanoparticles.
Synthesis of [Ag 28 Cu 12 (SPhCl 2 ) 24 ] 4− . The Ag 28 Cu 12 (SPhCl 2 ) 24 was prepared by a literature method reported by the Zheng group with some modification 57 . Specifically, 60 mg of Cu(O 2 C 5 H 7 ) 2 was dissolved in 5 mL of CH 3 OH and 15 mL of CH 2 Cl 2 , to which 60 mg AgNO 3 (dissolved in 2 mL of H 2 O) was added. After stirring for 20 min, 100 μL of HSPhCl 2 was added in, and the reaction further processed for 30 min. Then, 30 mg NaBH 4 (dissolved in 2 mL of H 2 O) was added in. The reaction was allowed to proceed for 5 h. After that, the aqueous layer was removed, and the mixture in the organic phase was rotavaporated under vacuum. Then 50 mL of CH 3 OH was used to extract the Ag 28 Cu 12 (SPhCl 2 ) 24 nanocluster, to which supernatant 20 mg of (PPh 4 )Br was added in. The precipitate was then washed three times by CH 3 OH. Then the final product, i.e., [Ag 28 Cu 12 (SPhCl 2 ) 24 ] 4− (PPh 4 ) 4 , was used directly. The yield is 35% based on the Ag element (calculated from the AgNO 3 ). These three nanoclusters were prepared from parallel Au-alloying reactions (in the same condition but were stopped at different times). Specifically, 20 mg of Ag 28 Cu 12 (SPhCl 2 ) 24 was first dissolved in 20 mL of CH 2 Cl 2 and then 5 mg of Au(I)-SPhCl 2 complexes was added in. After 2 min, 100 mL of hexane was poured in to pause the reaction; the precipitate was then dissolved in 20 mL of CH 2 Cl 2 to yield the Au x Ag 28-x Cu 12 (SPhCl 2 ) 24 (x = 1.32). Expanding the reaction time from 2 min to 3 min would produce the Au x Ag 28x Cu 12 (SPhCl 2 ) 24 (x = 7.56). Expanding the reaction time from 2 min to 8 min would produce the Au 12 Cu y Ag 32-y (SPhCl 2 ) 30 .
Preparation of XAFS samples. In all, 10 mg of Ag 28 Cu 12 (SPhCl 2 ) 24 was dissolved in 10 mL of CH 2 Cl 2 and then 3 mg of Au(I)-SPhCl 2 complexes was added in. After 1 min, 200 mL of hexane was poured in to pause the reaction; the precipitate was then dissolved in 5 mL of CH 2 Cl 2 to yield the Au x Ag 28-x Cu 12 (SPhCl 2 ) 24 (Sample 1). In total, 10 mg of Ag 28 Cu 12 (SPhCl 2 ) 24 was first dissolved in 10 mL of CH 2 Cl 2 and then 3 mg of Au(I)-SPhCl 2 complexes was added in. After 2 min, 200 mL of hexane was poured in to pause the reaction; the precipitate was then dissolved in 5 mL of CH 2 Cl 2 to yield the Au x Ag 28-x Cu 12 (SPhCl 2 ) 24 (Sample 2). Single crystals of XAFS Sample 2 (i.e., Au x Ag 28-x Cu 12 (SPhCl 2 ) 24 , x = 7.76) were cultivated at room temperature by vapor diffusing the ethyl ether into the DMF solution of nanoclusters.  Supplementary Fig. 8) and 20 mg of Au x Ag 28-x Cu 12 (SPhCl 2 ) 30 (2-min sample in Supplementary Fig. 14) were dissolved in 5 mL of DMF. Single crystals of the co-crystallized nanoclusters were cultivated at room temperature by vapor diffusing the ethyl ether into the DMF solution. After 21 days, black crystals were collected, and the structure of the co-crystallized nanoclusters was determined. The CCDC number of the co-crystallized [Au 4 Ag 24 Cu 12 (SR) 24 ] 2 [Au 12 Cu y Ag 32-y (SR) 30 ] 1 (y = 3.74) is 2009377.
Time-dependent ESI-MS of the Au alloying process on Ag 28 Cu 12 (SPhCl 2 ) 24 . In total, 20 mg of Ag 28 Cu 12 (SPhCl 2 ) 24 was firstly dissolved in 20 mL of CH 2 Cl 2 and then 5 mg of Au(I)-SPhCl 2 complexes (powder) was added in. The ESI-MS measurement of the reaction was performed every minute.
X-ray absorption fine structure spectroscopy measurements. XAFS measurements at the Au L3-edge (11919 eV) were performed at the beamline BL14W1 station of the Shanghai Synchrotron Radiation Facility (SSRF), China. The storage ring of the SSRF was working at an energy of 3.5 GeV with an average electron current of 300 mA. The hard X-ray was monochromatized with a Si (311) monochromator. XAFS data were collected in the transmission mode in the energy range from 200 below to 1000 eV above the Au L3-edge. The acquired XAFS data were processed according to the standard procedures using the ARTEMIS module implemented in the IFEFFIT software packages.
X-ray crystallography. For the crystal date of Ag 28 Cu 12 (SPhCl 2 ) 24 , Au x Ag 28-x Cu 12 (SPhCl 2 ) 24 (x = 1.32), Au 12 Cu y Ag 32-y (SPhCl 2 ) 30 , Au x Ag 28-x Cu 12 (SPhCl 2 ) 24 (x = 7.76), and the co-crystallized [Au 4 Ag 24 Cu 12 (S-R) 24 ] 2 [Au 12 Cu y Ag 32-y (SR) 30 ] 1 (y = 3.74): the data collection for single-crystal X-ray diffraction was carried out on Stoe Stadivari diffractometer under nitrogen flow, using graphite-monochromatized Cu Kα radiation (λ = 1.54186 Å). For the crystal date of Au x Ag 28-x Cu 12 (SPhCl 2 ) 24 (x = 7.56), Au 12 Ag 32 (SPhCl 2 ) 30 : the data collection for single crystal X-ray diffraction was carried out on a Bruker Smart APEX II CCD diffractometer under liquid nitrogen flow, using graphite-monochromatized Mo Kα radiation (λ = 0.71069 Å). Data reductions and absorption corrections were performed using the SAINT and SADABS programs, respectively 63 . The structure was solved by direct methods and refined with full-matrix least squares on F 2 using the SHELXTL software package 64 . 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. All crystal structures were treated with PLATON SQUEEZE, and the diffuse electron densities from these residual solvent molecules were removed 65 .
Characterization. The UV-vis absorption spectra of nanoclusters were recorded using an Agilent 8453 diode array spectrometer. Electrospray ionization mass spectrometry (ESI-MS) measurements were performed by MicrOTOF-QIII highresolution mass spectrometer. The sample was directly infused into the chamber at 5 μL/min. For preparing the ESI samples, nanoclusters were dissolved in CH 2 Cl 2 (1 mg/mL) and diluted (v/v = 1:2) by CH 3 OH. Energy-dispersive X-ray spectroscopy (EDS) mapping of nanoclusters were characterized by SEM (Quanta 400 F). X-ray photoelectron spectroscopy (XPS) measurements were performed on a Thermo ESCALAB 250 configured with a monochromatized Al Kα (1486.8 eV) 150 W X-ray source, 0.5 mm circular spot size, flood gun to counter charging effects, and analysis chamber base pressure lower than 1 × 10 −9 mbar.

Data availability
The X-ray crystallographic coordinates for structures reported in this work have been deposited at the Cambridge Crystallographic Data Center (CCDC), under deposition numbers CCDC-2009375, 2009456, 2009457, 2009377, 2009378, and 2083130. These data can be obtained free of charge from the Cambridge Crystallographic Data Centre via www.ccdc.cam.ac.uk/data_request/cif, which has been mentioned in the article.