Size and Shape of Nanoclusters: Single-Shot Imaging Approach

A method of single-shot imaging via aberration-corrected scanning transmission electron microscopy equipped with high angle annular dark-field detector (STEM-HAADF) has been applied to size-selected gold model catalysts (Au(25) and Au(39) ) on hydroxyapatite. Through quantitative intensity analysis, the size, in terms of number of atoms as well as 3D shape of the clusters are obtained.

Gold nanoparticles on solid supports are of particular interest for their high catalytic activity for low-temperature CO oxidation. [ 1 ] Although a number of factors infl uencing the catalytic properties have been recognized, including method of synthesis, size, shape, and interaction with support, the origin of the reactivity is still much of a debate. [ 2 ] By way of example, Lopez et al. pointed out that, among these factors, the availability of a high concentration of low-coordination sites on small Au particles, particularly those at corners, is the most important. [ 2b ] In contrast, Valden et al. showed that the catalytic rate of Au nanoparticles decreases with decreasing particle size when they are smaller than 3 nm, despite an increase in the amount of low-coordination atoms. [ 3 ] In the later articles, the same group demonstrated that the catalytic activity of the bilayer Au fi lms is the highest as compared with those of the bilayer and hemispherical Au particles. [ 4 ] This further suggested that low-coordination corner or edge Au sites may not be so crucial as they were previously considered. More recently, based on the scanning transmission electron microscopy (STEM) analysis, a high catalytic activity for carbon monoxide oxidation is associated with the presence of bilayer Au nanoparticles of 0.5 nm on iron oxide. [ 5 ] However, this is not supported in a subsequent article, in which the absolute necessity of monolayer or bilayer structure for high catalytic performance is ruled out. [ 6 ] These distinct reports indicate that more in-depth research is needed in order to establish a clear link among these identifi ed factors and any combination of them should be treated on a case-by-case basis.
Measuring size and shape of nanometer catalytic clusters is a challenging task. In particular, real catalyst samples are often characterized by a distribution of size and shape. Although the capability of STEM, with or without an aberration corrector, as a structural probe has been demonstrated, [7][8][9][10][11] there have been few reports on viability of the technique as an effective tool for obtaining size and shape Size and Shape of Nanoclusters: Single-Shot Imaging Approach Y. Han , D. S. He , Y. Liu , S. Xie , T. Tsukuda , and Z. Y. Li * information of real nanocatalysts. Here, we apply this technique to size-selected model catalysts, Au 25 and Au 39 (clusters containing nominal 25 and 39 gold atoms, respectively). We show, by optimizing imaging conditions of a STEM equipped with a high-angle annular dark-fi eld (HAADF) detector, that the size and shape of the clusters can be obtained from the STEM-HAADF intensity profi le through single-shot imaging. The quantitative analysis of STEM-HAADF intensity allows us to derive that the deposited and then calcined Au clusters maintained discrete sizes being multiples of either Au 25 or Au 39 . The overall shape of the clusters can be modelled as hemispherical-like, suggesting a strong cluster-support interaction. Figure 1 a,b show two high-resolution images taken from samples of Au 25 and Au 29 supported on hydroxyapatite (HAP), respectively. Each cluster was analysed with the background subtracted with reference to its surrounding areas. The histograms of the integrated intensities for the samples Au 25 and Au 39 are shown in Figure 2 a,b, respectively. To determine the number of peaks (k) and the associated peak positions in each case, the Integrated Complete Likelihood (ICL) criterion was employed based on Gaussian distributions. [ 12 ] The intensities larger than 120 × 10 4 for Au 25 and 95 × 10 4 for Au 39 were not included in the ICL calculation due to a small number of data points. The best fi ts were achieved when k = 8 for Au 25 and k = 4 for Au 39 , corresponding the existence of 8 and 4 mixed Gaussian functions, respectively, as shown in Figure 2 .
There are several features of interest in Figure 2 that relate to the nature of the size-controlled Au clusters. The fi tted Gaussian functions are evenly spaced in both cases. The spacing between the neighboring peaks for the Au 25 sample is (136 ± 11) × 10 2 , with that for the Au 39 sample being (208 ± 22) × 10 2 . The ratio of these two values is 1.53, being very close to the expected ratio of numbers of atoms in Au 39 and Au 25 monomers, i.e., 39/25 = 1.56. Therefore, the fi rst peaks in each case are attributed to the corresponding monomers, with the rest of peaks being the multiples. The multiples of the Au 25 and Au 39 monomers are labeled in Figure 1 according to the above analysis.
Here, the size-controlled Au 25 and Au 39 nanoclusters were acting as mutual mass calibrators. In the quantitative analysis of the cluster intensities, a number of factors can introduce errors, including, for example, the emission current, the aberrations of the magnetic lenses, the position of the objective aperture to the Ronchigram, the position of the electron beam with regards to the ADF detector and the offset and gain values of the ADF preamplifi er. In particular, to assess the microscopy conditions on quantifi cation of mass of a cluster, two specimens, both containing single Au atoms supported on carbon fi lms (images not shown in this paper), were used as an independent check between two different runs. The difference in the mean integrated intensity for a single atom is ∼ 5%, far smaller than the integrated intensity difference, ∼ 56%, between the Au 25 and Au 39 cluster samples. Whenever possible, both Au 25 and Au 39 samples were always imaged in the same session to compare.
The results in Figure 2 show that a large fraction of the clusters in both samples had aggregated, consisting several clusters of initial sizes. The clear integer multiples in the intensity histograms suggest that the aggregation is not via an Ostwald ripening process, rather it may be through cluster fusion, in which clusters retained their original identities. This is consistent with the broad cluster diameter distribution (full-width-half-maximum) in each case, 1.25 nm for Au 25 and 1.5 nm for Au 39 , with the mean diameter of around 1.7-1.9 nm (See Figure 3 ).
The above result should be compared with the previous published data for the same set of samples when they were imaged immediately after the synthesis, in which both types of the clusters had been shown to be 1.1 ± 0.5 nm in diameter. [ 13 ] The large discrepancy in diameter distribution between the two studies cannot be completely attributed to the limited resolution and low sensitivity of the STEM without a probe corrector used in the previous study. One possibility is that the aggregation had occurred during the time span over two months from the sample preparation in Japan to the STEM observations in the UK. Further investigation is planned to see if these aggregated clusters, which have retained the memory of their initial sizes, still maintain their size-dependent catalytic performance.
To gain insight of the bonding strength between clusters and the HAP support, we plot in Figure 4 the integrated HAADF intensity of each cluster as a function of its diameter, together with the relationship generated from the idealized spherical and hemispherical models (shown with solid and dashed lines). In the geometric models, the Wigner-Seitz radius of gold, r ws = 0.165 nm, is used. [ 10 , 14 ] It is clear from Figure 4 that majority data points, though scattered, follow the trend of the hemispherical model, when the diameter is less than ∼ 2 nm. The result suggests that the clusters are fl attened on the support, likely as a result of the strong interaction between the Au and PO 4 3 − moieties of HAP. [ 15 ] In summary, Au 25 and Au 39 nanoclusters supported on hydroxyapatite (HAP) were investigated using aberration corrected scanning transmission electron microscopy. We demonstrate that the quantitative analysis of the HAADF intensities of the nanoclusters provide an invaluable tool in obtaining direct information of cluster size in terms of the number of atoms within. We show that the well sizecontrolled clusters, as in this work, can act as ideal mass calibrators, similar as those prepared in the gas phase and mass selected. [ 9 , 10 , 16 ] We have associated the evenly spaced peaks in the histograms with multiples of single clusters of either Au 25    or Au 39 in each case, from which Oswald ripening mechanism for the observed cluster aggregation can be ruled out. There is a clear sign indicating that the aggregated clusters retained their original identity. This feature together with the strong interaction with the hydroxyapatite support based on the 3D shape analysis suggests new opportunities in the development of stable Au catalysts with size-dependent reactivity.

Experimental Section
The synthetic method of glutatoneate-protected Au (Au:SG) clusters has been reported previously. [ 13 , 17 ] Briefl y, chilled (0 ° C) aqueous solution of NaBH 4 was rapidly injected into the methanol solution of glutathione (GSH) and HAuCl 4 kept at 0 ° C under vigorous stirring. The precipitates were collected and dried in vacuo to obtain crude mixture of Au:SG. The Au 25 (SG) 18 and Au 39 (SG) 24 clusters were isolated using polyacrylamide gel electrophoresis; the compositions and purities were confi rmed by electrospray ionization mass spectrometry. The Au n clusters ( n = 25, 39) supported on hydroxyapatite (HAP), Ca 10 (PO 4 ) 6 (OH) 2 , were prepared via two steps. [ 13 , 18 ] Firstly, the Au 25 (SG) 18 or Au 39 (SG) 24 clusters were adsorbed on HAP (0.2 wt%) by mixing them with HAP in basic water. Secondly, the Au n (SG) m /HAP composites collected by fi ltering were calcined at 300 ° C for 2 hours in vacuo. The GS ligands were confi rmed to have been removed in this step, as evidenced by inductively coupled plasma analysis.
The Au 25 and Au 39 clusters supported on HAP were dispersed in methanol and loaded onto holey carbon fi lms on copper grids directly. They were characterized using a JEOL 2100F transmission electron microscope, equipped with a CEOS probe aberration corrector, operated at 200 kV. The semi-convergence angle for the incident electron beam was ∼ 20 mrad and the inner and an outer collection angles for the JEOL ADF detector were 55 mrad and 148 mrad, respectively. The same values for the gain and offset of the preamplifi er were used throughout the observations of the two samples. The emission current was monitored to be stable. In order to minimize the effect of the electron beam to both the clusters and the HAP substrate, the experimental conditions were carefully tuned, with a 20 μ m condenser aperture, a medium magnifi cation ( × 10M) and a short exposure time (20 μ s/pixel) employed. Researchers (GR-003). Note: The caption for Figure 3 in the Early View pdf of this article was incorrect. The caption has now been corrected.