Performance of Preformed Au/Cu Nanoclusters Deposited on MgO Powders in the Catalytic Reduction of 4-Nitrophenol in Solution

The deposition of preformed nanocluster beams onto suitable supports represents a new paradigm for the precise preparation of heterogeneous catalysts. The performance of the new materials must be validated in model catalytic reactions. We show that Au/Cu nanoalloy clusters (nanoparticles) of variable composition, created by sputtering and gas phase condensation before deposition onto magnesium oxide (MgO) powders, are highly active for the catalytic reduction of 4-nitrophenol in solution at room temperature. Au/Cu bimetallic clusters offer decreased catalyst cost compared with pure Au and the prospect of beneficial synergistic effects. Energy-dispersive X-ray spectroscopy (EDS) coupled with aberration-corrected STEM imaging confirms that the Au/Cu bimetallic clusters have an alloy structure with Au and Cu atoms randomly located. Reaction rate analysis shows that


Introduction
Cluster beam deposition is a new method to prepare heterogeneous catalysts for research and development, in which atomic clusters (nanoparticles) of tunable size typically below 10 nm are preassembled into a beam and then deposited in a vacuum chamber onto the catalyst support. [1][2][3][4] Potential advantages of the approach include the absence of solvent and effluent in the catalyst synthesis; control of cluster size, composition and morphology; and the absence of ligands compared with colloidal routes. [5][6][7][8] However, the technique is at an early stage, [9] most especially where catalytic behaviour under realistic reaction conditions is concerned. Thus there is an urgent need to validate the performance of this new class of nanomaterials in a series of model chemical transformations, and compare their behaviour with catalysts prepared by more traditional and well established routes. In this work, we report a first investigation of a solution phase transformation performed by nanoalloy catalysts prepared by cluster beam deposition.
The discovery of the catalytic activity of gold (Au) nanoparticles for low temperature oxidation of CO provoked an explosion of interest in gold catalysis. [10,11] Au clusters can catalyse a range of reactions, e.g., the water-gas shift reaction [12][13][14] and selective oxidation of carbon-carbon double bonds [15,16] and carbon-oxygen bonds. [17,18] The 4-nitrophenol reduction by sodium borohydride (NaBH4) is considered to be a standard model catalytic reaction [19,20] to evaluate nano catalyst activity; precise optical measurement of the amount of 4-nitrophenol at very low concentrations is feasible. It is commonly believed that this reaction follows a Langmuir-Hinshelwood (LH) mechanism (Figure   1), [21,22] in which 4-nitrophenol is adsorbed on the surface of Au particles and reacts with activated hydrogen on the surface formed by decomposition of NaBH4. It is thought that adding to Au a second metal element, M, which has a larger adsorption energy for nitrophenol than Au, can enhance the reaction rate. [23] The O-N bonds from the nitro group become weaker due to the electron delocalization from the O to the metal atoms, which directly correlates to the reaction rate. Based on this understanding, much effort has been made to produce Au/M bimetallic clusters on different supports and investigate their catalytic performance. For example, Pretzer et al. [24] synthesized Au nanoparticles decorated with palladium (Pd) and found that the activity of the Pd/Au catalyst is 5.5 (13) times that of pure Au (Pd) nanoparticles. Similar evidence of synergistic effects in nitrophenol and nitrothiophenol reduction over Au/Ni catalysts is reported. [25] The use of copper (Cu) as an additive to improve the activity of noble metal clusters has become a "hot topic" because it can decrease the cost of catalysts and improve catalytic properties simultaneously. [26] It has been reported that adding Cu to Au nanoparticles can significantly improve their catalytic activity in many reactions, e.g., CO oxidation. [27][28][29][30] However, for nitrophenol reduction catalysed by Au/Cu bimetallic clusters, limited reports can be found and the catalytic mechanism is unconfirmed. It is commonly believed that Cu has a stronger interaction with the nitro group of nitrophenol than Au, which increases the adsorption energy of nitrophenol on the catalyst surface.
Deposition of Au on the surface of Cu nanoparticles by a chemical method led to catalytic activity for 4-nitrophenol reduction that was enhanced by one order of magnitude in comparison with pure Au catalyst. [31] However, there was no clear evidence that the enhanced activity results from a synergetic effect between Au and Cu, since no Au/Cu alloy structure was observed. Au/Cu bimetallic clusters produced on planar sapphire substrates were active for 4-nitrophenol reduction and the activity could be enhanced by visible light (due to the localized surface plasmon resonance). [32] But the activity without excitation of the light was not explored. Thus, it is important to clarify the active sites in the Au/Cu system due to their significance in catalyst design. Traditionally, Au/Cu bimetallic clusters have been efficiently prepared by various chemical methods, for example, coimpregnation, [33] deposition-precipitation [34] and colloidal methods [35] etc. However, a common disadvantage of these chemical methods is the presence of residual impurities, coming from the anion group of the metal salts not fully burned off in the calcination process or from capping ligands, as used purposely to reduce cluster aggregation. Some researchers report that capping ligands can either decrease the cluster activity by hindering reagent access to the catalyst [36,37] or can increase the activity via electron donation. [38] The existence of these impurities complicates the explanation of the original catalytic activity of the cluster, and can sometimes lead to misinterpretations. Moreover, another major challenge for chemical synthesis of binary nanoparticles is making sure that all the Au and Cu atoms are alloyed inside the clusters. A well-controlled way to make naked Au/Cu nanoparticle catalysts are desirable. In this work, we employed a dual-magnetron sputtering gas condensation cluster source to produce Au/Cu bimetallic clusters deposited onto an inert powder support, magnesium oxide (MgO). By changing the power applied to the two targets (Au, Cu), the average Au/Cu ratio can be tuned.
Aberration-Corrected Scanning Transmission Electron Microscopy (STEM) in high-angle annular dark-field (HAADF) mode, coupled with Energy-Dispersive X-ray Spectroscopy (EDS), reveals that the Au/Cu bimetallic clusters formed have a random alloy structure. We find that these bimetallic clusters are much more active in selective nitrophenol hydrogenation than pure Au or Cu clusters, and 25 times for active than reference Au/Cu binary nanoparticles produced by impregnation. The interplay between surface Au and Cu sites is deduced to create the most active site for reaction. Our study thus validates the performance of the new nanoalloy materials in a solution phase reaction, and provides an insight into nanocatalyst design of bimetallic systems at the atomic scale.

Analysis of magnetron sputtering deposited Au/Cu bimetallic clusters
Au/Cu clusters with three different Au/Cu ratios -that we term Au-rich, Au/Cu-equal (approximately) and Cu-rich -were produced by applying sputtering powers to the Au and Cu targets of 6 W : 3 W, 6 W : 6 W and 3 W : 6 W, respectively. The cluster size was characterized by two methods, time-offlight mass filtering which gives cluster size (mass) information before depositing onto the supports, and STEM which provides cluster size (diameter) information after landing on the supports. Figure   2 shows mass spectra obtained during the deposition as well as STEM cluster diameter distribution histograms for the following samples: Au-rich (a), (d); Au/Cu-equal (b), (e); and Cu-rich (c), (f). The insets show typical low magnification HAADF images of the kind used to acquire the cluster diameter histograms. From the mass spectra, the Au-rich and Cu-rich clusters have peaks at similar masses of ~ 40,000 amu, which is equivalent to the size of Au200, whereas the Au/Cu-equal clusters peak at a larger value of ~ 65,000 amu, equivalent to the size of Au300. The cluster diameter histograms based on the HAADF images also show the same trend, i.e. that Au-rich clusters and Cu-rich clusters have a smaller cluster diameter ~ 3.6 nm, than Au/Cu-equal clusters, ~ 4.6 nm. To clarify if agglomeration takes place after clusters land on the support, the diameter of the Au/Cu-equal clusters at the peak position in the mass spectrum is estimated based on the cluster volume assuming a quasi-spherical shape. If the cluster composition is assumed to be Au1Cu1 and the cluster has a spherical shape on the support, the diameter of a cluster with a mass of 65,000 amu should be ~ 2.4 nm (~ 3.0 nm for 7 hemispherical clusters), which is much smaller than the average diameter, 4.6 nm, obtained from HAADF images. This indicates that the Au/Cu-equal clusters are larger because they aggregate on the support surface, or possibly lose their spherical shape after landing on the support. In the HAADF images some large clusters formed by coalescence can easily be observed. It should be noted that the surface agglomeration which happens during the deposition process occurs because the powder supports in the vibration cup cannot be agitated entirely evenly, especially when they are charged by the cluster ion beam, leading to uneven cluster coverage. This behavior occurs for all three samples.
Such surface agglomeration has also been reported for deposition of Fe-Co nanoalloy clusters onto carbon supports by dual plasma guns. [39] Table 1 shows the ICP results as a function of sputtering power applied to each target, yielding the Au-rich, Au/Cu-equal (approximately) and Cu-rich nanoalloy samples. It can be seen from the Table   that the total metal loadings of these three samples are different, but the atomic ratios of Au and Cu in Au-rich, Au/Cu-equal and Cu-rich samples are 3.87:1, 1.16:1 and 0.45:1, respectively, which indicates the overall catalyst composition is strongly dependent on the power applied to each target.
To identify the distributions of Au and Cu atoms inside the clusters, compositional line profiles across individual clusters in each kind of sample were obtained by EDS line scanning (Figure 3). Within the clusters, both Au and Cu signals are detected in each of these three samples and no obvious valleys or peaks are observed, which suggests Au atoms and Cu atoms are mixed together instead of forming core-shell or Janus structures. Theoretically, mixing of Au and Cu is energetically-favored in Au/Cu compounds compared with separated phases. [40][41][42][43] In addition, a steady increase of Cu signal observed from the Au-rich sample to the Cu-rich sample confirms the increasing Cu content inside the clusters.  by red lines correspond to (22 ̅ 0) planes with an interplanar spacing of 1.39 Å, which lies between the interplanar spacing of (22 ̅ 0) planes for pure Au (1.44 Å) and pure Cu (1.28 Å). The same trend is also found in an Au/Cu-equal cluster (Figure 4 (b)) and a Cu-rich cluster (Figure 4 (c)); again the measured interplanar spacings of the (002) planes are between those of pure Au and pure Cu. This result further supports the Au/Cu alloy structure, according to Vegard's law. [44] In addition, Yin et al. distinguished the Au/Cu core/shell structure through observing the intensity contrast of HAADF STEM images. [45] A trough or a peak was observed in the intensity line profile across an Au/Cu cluster with core/shell structure. For our Au/Cu bimetallic clusters, it is apparent that no clear core-shell structure exists, which is consistent with the EDS line scanning results. This alloy structure is also observed in clusters of other orientations and amorphous clusters from all the three samples by distinguishing the HAADF intensity of STEM images. The alloy structure we find has also been reported in chemically prepared Au/Cu bimetallic clusters by other groups. [46] For the bulk Au/Cu alloy, chemically ordered structures are energetically favored and chemically ordered Au3Cu, Au1Cu1 and Au1Cu3 phases are found at low temperature. Here, Cu atoms and Au atoms are arranged periodically. In electron diffraction patterns, chemically ordered structures can lead to the appearance of super-structure diffraction spots along certain directions. [47,48] For example, in Au1Cu3 nanorods, a [110] oriented chemical ordering was observed by monitoring the appearance of (110) super-structure diffraction spots in the electron diffraction pattern. [49] To confirm the absence of chemical ordered structures in our Au/Cu bimetallic clusters, fast Fourier transforms (FFT) have been calculated for the STEM image of clusters shown in Figure 4 and are displayed as inset. If the ordered structures -Au3Cu1, Au1Cu1 or Au1Cu3 -exist in the clusters, the super-structure diffraction spots from (110) or (100) planes should appear in the FFT images. However, in the insets to Figure 4, no super-structure diffraction spots are observed at all, which indicates that Au and Cu atoms are randomly located in the clusters and there is no chemical 11 ordering. Finally, the intensity of columns in the HAADF images is related to the Z number of atoms. [50] The heavier the atom, the higher the intensity. [51] Considering Cu is much lighter than Au, it is reasonable to assume that lower column intensity corresponds to more Cu atoms in a column. In   Figures 4 (a) and (b), some atom columns are significantly darker than the adjacent columns, highlighted by white arrows, which suggests that more Cu atoms are located in these darker columns than in the neighbors. Again, this behavior is consistent with a random alloy structure.  Au-rich, (b) Au/Cu-equal, (c) Cu-rich. The absorbance peak at 400 nm corresponds to 4-nitrophenol and the peak intensity relates to the concentration.

Evaluation of catalytic activity for 4-nitrophenol reduction by NaBH4
After adding NaBH4 to the nitrophenol solution, 4-nitrophenolate ions are formed, which show a strong visible absorbance at 400 nm. [52] The intensity of the absorbance peak, proportional to the concentration of 4-nitrophenolate ions, is used to evaluate the progress of the reaction. Figure 5 shows typical UV-VIS absorbance spectra with time intervals of 5 min after adding of Au-rich, Au/Cu-equal and Cu-rich cluster catalysts supported on MgO. It is obvious that in the case of the Au/Cu-equal cluster catalyst, on adding the catalyst into the reaction mixture, the absorbance peak at 400 nm decreases immediately and continuously. However, in the case of the Au-rich and Cu-rich cluster catalysts, the peak intensity decreases much more slowly, which indicates that these catalysts are less active than the Au/Cu-equal catalysts.
The reduction of 4-nitrophenol catalysed by metal particles is considered pseudo-first order with respect to the concentration of 4-nitrophenol in the presence of a large excess of NaBH4. [53,54] The apparent reaction rate constant kapp, which can be used to compare the activity of the catalysts, is defined through the Equation (1): Here Ct represents the concentration of nitrophenol at time t. Because the optical absorbance at time t (At) is proportional to Ct, a plot of Ct/C0 and thus -ln(Ct/C0) versus t can be acquired easily from the absorbance, as shown in Figure 6. C0 is the initial concentration of 4-nitrophenol. In accordance with the pseudo-first order kinetic assumption, the relationships for Ct/C0 and -ln(Ct/C0) versus reaction time are fitted by an exponential decay and linear growth, respectively. The apparent rate constant is obtained from the linear slope in Figure 6. The kapp obtained for the Au/Cu-equal cluster catalysed reaction is 17.8 × 10 -3 min -1 , which is almost 20 (10) times higher than that obtained for the Au-rich (Cu-rich) clusters. Considering the difference of metal loading in these three samples, the intrinsic reaction rate constant knor is calculated by normalizing the kapp values by the number of moles of metal as summarized in Table 2. The normalised reaction rate constant knor for Au/Cu-equal clusters is 3.49×10 4 min -1 mol -1 , which is 8.9 (6.6) times that of Au-rich (Cu-rich) clusters. It should also be noted that, compared with the Au-rich catalyst, the Cu-rich catalyst has a lower metal loading but higher normalized activity, indicating the Cu-rich clusters are more active than Au-rich clusters. We will discuss this behavior below. In addition, we also compared the catalytic activity of the Au/Cu-equal cluster on MgO sample with the reference catalyst (Au1Cu1 on MgO powder) produced by the traditional impregnation method. The knor for the reference sample is 0.14×10 4 min -1 mole of metal -1 , which is 25 times smaller than that for the Au/Cu-equal cluster sample (see table 2 and supporting information Figure S2(a)).
A list of recent studies on the 4-nitrophenol reduction over chemically produced Au and Cu based catalysts is summarized in Table S2 together with our Au/Cu-equal cluster catalyst. Although the intrinsic reaction rate constant obtained for the new cluster catalyst is not yet competitive with the best Au catalysts synthesized chemically, this method still provides a way to establish the original catalytic activity of "naked" clusters.

Discussion of the catalytic activity for 4-nitrophenol reduction over Au/Cu nanoalloy cluster catalysts and model calculations
The principal topics for discussion arising from the experimental results concern: (i) why the random nanoalloy Au/Cu-equal cluster catalysts supported on MgO are so much more active than the Au-rich or Cu-rich clusters and (ii) why the Au/Cu-equal cluster catalysts are significantly more active than the reference Au/Cu catalysts generated by impregnation. In addressing these questions, we also bear in mind the results that Cu-rich clusters are somewhat more active than Au-rich clusters. A central issue in understanding the enhancement of catalytic activity must be the nature of the binding of the principal reactant (4-nitrophenol) and product (4-aminophenol) to the MgO-supported Au/Cu nanoparticles. The results from the model calculations, namely that Cu has a stronger interaction than Au with both the reagent, 4-nitrophenol, and the product, 4-aminophenol, provide an appealing basis for explaining why the Au/Cu-equal clusters are the most active for the reduction of the nitrophenol.
Assuming the reaction follows an LH mechanism, the reduction takes place on the cluster surface.
For the pure Au cluster, the catalytic activity is limited by the weak adsorption of nitrophenol, whereas the pure Cu cluster has an overly strong interaction with the aminophenol, which hinders the desorption of the product and again leads to limited activity. The combination of Au with Cu increases the adsorption energy of the nitrophenol compared with gold and reduces the adsorption energy of aminophenol compared with copper. We have shown that the optimized adsorption configuration for this reaction involves the nitro group bonding with adjacent Au and Cu sites through the two O atoms; then the abundance of Au/Cu sites on the cluster surface will regulate the catalytic activity. Given the random arrangement of Au and Cu atoms inside the clusters, more Au/Cu sites exist on the Au/Cuequal cluster surface than Au-rich and Cu-rich clusters, which gives a reason why the Au/Cu-equal cluster catalyst exhibits the highest catalytic activity. Moreover, we can associate the higher catalytic activity of the Cu-rich clusters than the Au-rich clusters with the fact that the Au/Cu atomic ratio obtained is closer to 1:1 in the Cu-rich clusters (0.45:1) compared with the Au-rich clusters (3.87:1), which allows for more of the proposed Au-Cu active sites on the surface of the Cu-rich clusters.
The reason why the Au/Cu-equal cluster catalyst is much more active than the Au/Cu reference sample may be associated with the diameter distribution and composition of the clusters in the reference sample, see Figure S2(b) and

Conclusion
The

Au/Cu cluster deposition
Our clusters were produced in a dual-target magnetron sputtering gas condensation cluster source (at Teer Coatings). We deposited Au/Cu clusters with three different Au/Cu ratios on agitated MgO powder supports. A schematic diagram of the system is shown in the supporting information, Figure   S1. Detailed information can be found in an earlier report. [5] In the magnetron sputtering chamber, two magnetrons (copper and gold) are mounted in parallel with a condensation length (the distance between these targets and the exit nozzle of the condensation chamber) of 24 cm. The sputtering power applied to each magnetron is separately controlled in order to tune the material ratio in the resulting binary clusters. Au and Cu atoms are sputtered out from the targets and condensed in cold Ar/He gas to form Au/Cu clusters of various sizes. A pressure of ~ 0.21 mbar was maintained in the condensation chamber, with 100 standard cubic centimeters per minute (sccm) Ar flow and 20 sccm He flow. Positively charged clusters are extracted, accelerated and guided by a series of ion optical lenses. The ultimate flight direction of the cluster beam is controlled by an "octosphere" deflector that can transmit the cluster beam forward into a lateral time-of-flight mass filter [55] to measure the clusters' size distribution or bend the beam through 90 o to propagate vertically towards a lower chamber for powder deposition. In this deposition chamber, the powder supports (0.8 g MgO powders) were loaded into a cup which is agitated continuously to maximize exposure of all the powder to the cluster beam. In addition, during the deposition process, the cup was biased with a potential of -1 kV to accelerate and thus help immobilize clusters on the powder surface.
The Au/Cu reference sample was made by a traditional impregnation method [27] on the same support. Detailed information can be found in the supporting information section. The support used in this experiment was MgO powder obtained from Alfa Aesar with particle size between 100 and 200 nm. The powder itself has been tested and confirmed to be catalytically inert for the nitrophenol reduction reaction.

Characterization of Au/Cu nanoalloy clusters on MgO powder supports.
The elemental composition of the cluster catalysts was characterized by EDS in the STEM and inductively-coupled plasma mass spectrometry (ICP-MS) following digestion in aqua regia. The cluster size and atomic structure were characterized by a JEOL JEM-2100F scanning transmission electron microscope equipped with a Cs probe corrector (CEOS) at a convergence angle of 20 mrad and a HAADF detector operating with inner angle of 62 mrad and outer angle of 164 mrad at 200 kV.
STEM samples were prepared by dispersing the MgO powder decorated with clusters in deionized water, sonicating for several minutes and drop casting onto a nickel grid coated with an amorphouscarbon film.

4-nitrophenol reduction measurement
The reduction of 4-nitrophenol was carried out in aqueous solution at room temperature with NaBH4 acting as the reductant. The reaction scheme is shown in Scheme 1. 4-nitrophenol (1.67 mg, Sigma-Aldrich) and NaBH4 (18.92 mg, Sigma-Aldrich) were added in deionized water (200 ml) sequentially, which gave a 4-nitrophenol concentration of 0.06 mM and NaBH4 concentration of 2.5 mM. After shaking the solution for 2 min, the color became yellow, which indicated 4-nitrophenol conversion to 4-nitrophenolate. [56] For each test, 30 mg catalyst was added to 50 ml of this solution and continuously magnetically stirred. To monitor the reaction process, the optical absorbance of the reaction solution was recorded by a UV-VIS spectrophotometer (Agilent Technologies Cary Series) at intervals of 5 min. For each measurement, 2 ml of the analyte solution was filtered by a syringe filter (pore size: 0.2 μm) to remove the catalyst and poured into a cuvette. Supporting Information is available from the Wiley Online Library or from the author.