Exploring the Atomic Structure of 1.8 nm Monolayer-Protected Gold Clusters with Aberration-Corrected STEM

Monolayer-protected (MP) Au clusters present attractive quantum systems with a range of potential applications e.g. in catalysis. Knowledge of the atomic structure is needed to obtain a full understanding of their intriguing physical and chemical properties. Here we employed aberration-corrected scanning transmission electron microscopy (ac-STEM), combined with multislice simulations, to make a round-robin investigation of the atomic structure of chemically synthesized clusters with nominal composition Au144(SCH2CH2Ph)60 provided by two different research groups. The MP Au clusters were “weighed” by the atom counting method, based on their integrated intensities in the high angle annular dark field (HAADF) regime and calibrated exponent of the Z dependence. For atomic structure analysis, we compared experimental images of hundreds of clusters, with atomic resolution, against a variety of structural models. Across the size range 123 to 151 atoms, only 3 percent of clusters matched the theoretically predicted


Introduction
Monolayer-protected noble metal clusters are attracting considerable interest because of the appearance of magic numbers, which account for the stability of certain sizes, and their potential applications in bio-imaging, catalysis, sensors and so on [ 1 -4 ]. Full atomic structure determination of these MP clusters is a key factor in understanding their physical and chemical properties in depth. For certain thiolated clusters, single crystal X-ray crystallography has provided reliable details of both the inner gold-core structures and the Au-S ligand units. To date the MP clusters identified by this method are Au25(SR)18 [5][6][7], Au28(SR)20 [8], Au36(SR)24 [9], Au38(SR)24 [10], Au102(SR)44 [11], Au130(SR)50 [12], and Au133(SR)52 [13,14]. In the case of clusters of nominal composition Au144(SR)60, the determination of an ordered atomic structure with single crystal X-ray diffraction has so far proved unsuccessful, even though several research groups have obtained crystals [15][16][17]. This might be because the core and/or ligand layers of the clusters are amorphous [17], and/or because the crystals contain clusters of different sizes, for there are reports of many species of closely related composition, such as Au144(SR)59 [18], Au146(SR)59 [19], Au137(SR)56 [20,21] etc. Recently, nominally Au144(SR)60 clusters doped with other metals (Ag [22][23][24][25][26], Pd [27], and Cu [28,29]) have attracted attention due to their electronic and optical properties. Understanding of the Au144(SR)60 cluster is naturally of importance for further studies of these bimetallic clusters.
A theoretical structure for Au144(SR)60 was predicted by Lopez-Acevedo et al. [30] in 2009, and was composed of a chiral icosahedral Au114 core and 30 linear Au(SR)2 units. Experimental studies by large angle X-ray diffraction (LA-XRD) [31,32] extended X-ray absorption fine structure (EXAFS) [33] and 1 H NMR [34] techniques showed indirect evidence of agreement with the proposed model. Nevertheless, the failure of single crystal X-ray experiments means that the precise structure of Au144(SR)60 remains unresolved.
Structure determination of MP Au clusters by electron microscopy is valuable, since it does not require the growth of high quality single crystals needed for single crystal X-ray crystallography.
Nanobeam electron diffraction combined with HAADF-STEM imaging was employed to determine the structure of nominally Au144(SR)60 [35] and Au130(SR)50 [36] clusters by the Yacaman and Whetten groups. The experimental diffraction patterns and HAADF images show agreement with the predicted models [30,37] in the case of specific individual clusters, but some other information, such as the cluster size distribution, percentage of clusters that fit the predicted model, and proportion of unidentified structures, is not given. Bruma et al [38] proposed a method of using scanning nanobeam diffraction to determine the structure of MP Au clusters (Au102(SR)44). They also reported the effect of electron beam damage of the clusters and found that the structure of thiolated clusters could be modified after a few seconds. Azubel et al [39] determined the structure of a MP Au68 cluster at atomic resolution by using low electron dose (~800 e − /Å 2 ) TEM together with three-dimensional reconstruction of hundreds of TEM images.
The advantage of this method, adapted from biological studies, is that the atomic structure of a MP Au cluster can in principle be revealed without any prior knowledge or fitting to model structures. However, this requires that the MP Au clusters should be homogenous both in structure and size. Statistical investigations of both cluster size and structure by HAADF-STEM have been reported by the Birmingham group, in the case of the MP Au38, MP Au40 and MP Au55 clusters [40][41][42][43]. These experiments yielded the cluster nuclearity, aspect ratio, fluxionality and, in the case of MP Au55 [43], evidence that the atomic structure matches the theoretical prediction for one isolated size fraction.
Here we report a statistical round-robin investigation of nominally Au144(SR)60 clusters, synthesized independently by two different groups. The atom counting technique is conducted to determine the range of cluster sizes in the specimen, and the atomic structures are explored by comparison with multislice simulations of different model structures. We find a very small fraction of clusters (~3%) match the theoretically predicted Au144(SR)60 structure [30], while a large proportion of the clusters are amorphous or unidentified. However, 20% of these clusters are found to exhibit ring-dot features in the atomic images, which is an indication of local icosahedral symmetry.

Experiments
MP Au clusters with (SCH2CH2Ph) ligands were synthesised at the University of Jyväskylä. The standard "MP Au144" synthesis and corresponding ESI-MS analysis are reported in ref. [20], which showed the as-prepared sample to contain Au144(SR)60, but also Au137(SR)56 clusters. The clusters were sent to Birmingham in crystal form (shown in Figure 1a), and were then dissolved in toluene and drop cast onto a TEM grid covered with an amorphous carbon film. Single gold atoms were utilised as a mass balance to "weigh" the MP clusters via their integrated HAADF intensities, i.e., by the atom counting method [44][45][46]. HAADF-STEM atomic imaging was performed in a 200 kV JEM2100F STEM (JEOL) with spherical aberration probe corrector (CEOS). HAADF images were acquired with inner and outer detector angles of 62 and 164 mrad (camera length 10 cm), and probe convergence angle of 19 mrad. The scanning time for each frame was ~2.6 seconds and the electron dose was 7.9 × 10 4 e -/Å 2 /frame. The QSTEM software package [47] was utilised to simulate the HAADF images of candidate structures for comparison, an approach successfully applied to a range of size-selected Au clusters [48][49][50]. In HAADF-STEM, the intensity of an atom is proportional to the atomic number to the power n (Z n ), where n= 1.46 ± 0.18 according to our previous calibration for the camera angle employed [51]. The HAADF intensity contributed by the nominal sixty (SCH2CH2Ph) ligands is then found equivalent to 17.9 ± 7.6 Au atoms. The Au core size (i.e. the total number of Au atoms including any in the Au(SR)2 units) is then derived on the assumption (for convenience) that every MP cluster has sixty ligands. Single Au atoms are ejected by purposely scanning of a cluster for an extended time; an example of a HAADF-STEM image containing these reference single atoms is shown in Figure S1a. The integrated HAADF intensities of Au single atoms and undamaged MP clusters are shown in Figures S1b and S1c. Figure 1b shows the resulting histogram of Au atom numbers in the MP Au clusters obtained from this atom counting approach. The corresponding diameter distribution of the MP Au clusters is shown in Figure S1d; the average diameter of the clusters is 1.8 ± 0.1 nm. The average size of the MP clusters is 137 ± 11 Au atoms, rather than 144.

Size distribution and multislice simulations
The size distribution is broad and for the purpose of illustration (only), we also show in Figure 1b a multiple Gaussian fit with three peaks at 128 ± 7, 137 ± 2 and 144 ± 9. The fit is chosen because it is compatible with the electrospray ionization (ESI) mass spectrum in ref. [20], in which both in HAADF-STEM images. Therefore, a "simulation atlas" which covers the full range of orientations was calculated for each candidate cluster structure investigated. Considering the experimental size range of the clusters in Figure 1b, models of Au102(SR)44 [11], Au130(SR)50 [37], Au133(SR)52 [14], and Au144(SR)60 [30] were chosen for comparison with the experimental images.
These cluster models have either been theoretically proposed [37,30]  Au 102 (SR) 44 Au 130 (SR) 50 Z X Y α β measured [11,14]. Figure 2 shows some examples of simulations based on the models. The simulation atlases are presented in detail in Figures S6-S11.  The comparison with the simulations was carried out for 849 experimental cluster images. This systematic investigation demonstrates that only about 3% of the clusters match the predicted Au144(SR)60 structure [30]. (Even if we restrict analysis to clusters in the size fraction 139 to 149

Atomic structure investigation
Au atoms, only 4% of the experimental images match the theoretical structure.) The fraction of images, which matches the other models, is even lower. Idealised bare Au147 (no ligands) clusters with icosahedral, decahedral and face-centred cubic structures were also used for comparison.
They are not explicit in the paper because of the low match to the simulations and the fact that they were not ligand-protected. Figures 3(a-d) show some examples of HAADF-STEM images of the MP Au clusters that are found to fit the Au144(SR)60 simulations. As we can see that the main motif in these clusters (ring-dot or straight line features) is in line with the simulations (including the location of the motif). However, the matches between experiment and simulation are not perfect at the detailed level. This deviation may arise because of instability of the clusters, at the atomic level, under the electron beam irradiation. Firstly, radiation-induced atom diffusion [53] may cause atom smearing during image acquisition. Secondly, the surface structure of thiolated clusters is expected to be sensitive to the state of the ligands [54,55], so any radiation damage to the ligands may induce cluster structure modification. Thirdly, less strongly bound Au atoms can be ejected during a normal imaging scan, as we observe individual atoms a few angstroms away from the cluster (indicated with the yellow arrow in Figure 3b). However, our previous study on MP Au55 [43] clusters found that over 40% of the cluster images matched the simulations. Thus we think that the electron beam has a limited effect on the core structure of the clusters.
Additionally, the Au144(SR)60 cluster may have several possible isomeric structures of similar energy. Tian et al [56] showed two isomers of Au38(SR)24 and the less stable isomer transformed irreversibly to the more stable isomer when heated to 50 °C. These four reasons may explain the imperfect detailed match between experimental and simulated cluster structure, even in generally well-matched cases.
Most of the cluster images were found to be amorphous (Figure 3e). However, amongst these clusters, we found that a good proportion (~20%) of clusters presented ring-dot features, as highlighted in Figure 3f found that about 7% of the cluster images were similar to the predicted Au144(SR)60 model and that an additional ~30% cluster images presented ring-dot features. These results are similar to those for the Jyväskylä sample. The appearance of this feature is consistent with the existence of local icosahedral symmetry in the clusters. A simulation atlas of an icosahedral bare Au147 is presented in Figure S12. The ring-dot feature can be seen in the simulations of Au144(SR)60 ( Figure S6-7), and Au133(SR)52 ( Figure S11) as well as Au102(SR)44 ( Figure S8) and Au130(SR)50 ( Figure S9-10). Both Au102(SR)44 and Au130(SR)50 have decahedral cores, but those are encapsulated by two five-fold symmetry caps [11,37] and a ring-dot feature occurs at certain orientations in the simulations. Thus the observed ring-dot motifs may be traceable either to an icosahedral core or, in a few cases, to a five-fold cap. Worth noting is that not all of the orientations show a ring-dot feature, so a higher percentage of clusters may contain similar structural elements. For example, in the case of Au144(SR)60and Au133(SR)52, if 20% of the images present a ring-dot feature, and assume the clusters are randomly orientated, then 40% of the clusters may contain an icosahedral core (around 50% of atlas simulations present the ring-dot feature).

Conclusion
In summary, we have employed aberration-corrected STEM to characterise the size and provide insights into the structure of nominally Au144(SR)60 clusters. We find the samples synthesised contain a range of cluster sizes, which is also indicated in the mass spectra. Image comparison with multislice simulations of model structures of all orientations shows that only about 3% of clusters fit the predicted Au144(SR)60 structure. Nevertheless, the experimental images exhibited a ring-dot feature, characteristic of local icosahedral order, in a further 20 to 30% of clusters. Based on the simulation atlases, it may be that an even higher proportion of clusters, perhaps 40%, contain icosahedral elements. Electron beam damage at the atomic scale may account for the imperfect matches between simulations and experiments, and no doubt both low doses and low beam energy would be desirable for minimising the electron beam effect on the thiolated clusters, but there is no doubt that the ability of HAADF-STEM both to size and to analyse individual clusters is a powerful tool in the structure resolution problem. In the present case, the rather broad size distribution obtained from the atom counting is consistent with the absence of wellordered X-ray diffraction patterns from the cluster crystals.