Dynamics of Pd Dopant Atoms inside Au Nanoclusters during Catalytic CO Oxidation

Doping gold nanoclusters with palladium has been reported to increase their catalytic activity and stability. PdAu24 nanoclusters, with the Pd dopant atom located at the center of the Au cluster core, were supported on titania and applied in catalytic CO oxidation, showing significantly higher activity than supported monometallic Au25 nanoclusters. After pretreatment, operando DRIFTS spectroscopy detected CO adsorbed on Pd during CO oxidation, indicating migration of the Pd dopant atom from the Au cluster core to the cluster surface. Increasing the number of Pd dopant atoms in the Au structure led to incorporation of Pd mostly in the S–(M–S)n protecting staples, as evidenced by in situ XAFS. A combination of oxidative and reductive thermal pretreatment resulted in the formation of isolated Pd surface sites within the Au surface. The combined analysis of in situ XAFS, operando DRIFTS, and ex situ XPS thus revealed the structural evolution of bimetallic PdAu nanoclusters, yielding a Pd single-site catalyst of 2.7 nm average particle size with improved CO oxidation activity.

S4 stirring. The reduction was then initiated by addition of 2 ml of a NaBH4 solution (0.075% w/v) in ice-cold H2O. The mixture was allowed to slowly warm to room temperature over 5 hours, accompanied by a color change to ruby red, indicating the formation of nanoparticles. The UV-Vis spectrum of the nanoparticle solution and the DRS spectrum of the TiO2 supported catalyst can be found in Figure S4.
Synthesis of Pd nanoparticles. Pd nanoparticles (Pd NPs) with an approximate particle size of 3.9 nm according to the literature were synthesized by the citrate method. 6 A solution of PdCl2 in 170 ml H2O (0.118 mmol) was prepared, to which trisodium citrate (0.4 mmol) was slowly added.
Subsequently, the reduction was initiated by addition of 11.8 ml of a NaBH4 solution in ice-cold H2O (0.075 wt%). The light grey solution was stirred at RT for 2 hours. The UV-Vis spectrum of the nanoparticle solution and the DRS spectrum of the TiO2 supported catalyst can be found in Figure S5.

Characterization Techniques
UV-Vis spectra of nanoclusters dissolved in CH2Cl2 were recorded on a Perkin Elmer Lambda 750 UV-Vis spectrometer. Diffuse Reflectance Spectroscopy (DRS) of the nanoparticle catalysts was performed using the same instrument coupled to a 60 mm integration sphere.
MALDI-RTOF mass spectra were acquired near threshold laser irradiance to obtain mass spectra of sufficient mass spectrometric resolution [3000-5000 at full width half-maximum (fwhm)]. All S5 displayed mass spectra were based on averaging 300-600 single and unselected laser pulses (λ = 337 nm at 50 Hz).
Transmission electron microscopy (TEM) and high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) was performed using a 200 kV FEI Tecnai F20 S-TWIN analytical (scanning) transmission electron microscopy [(S)TEM] instrument equipped with a Gatan GIF Tridiem filter. The energy resolution was ≤1 eV, the semiconvergence angle ∼8 mrad, the semicollection angle ∼15 mrad, and the spatial resolution on the order of 0.5 nm. Supported clusters were directly deposited on carbon-coated copper grids and plasma cleaning was applied to remove possible hydrocarbons and adsorbed water. Unsupported clusters were dissolved in dichloromethane and the solution dropcasted on carbon-coated copper grids, followed by plasma cleaning after evaporation of the solvent.
Chemical analysis with Total Reflection X-ray Fluorescence (TXRF) spectroscopy was performed with the supported cluster samples to determine the actual metal loading (%wt), using an ATOMIKA 8030C X-ray fluorescence analyzer. This spectrometer employs a total reflection geometry with an energy-dispersive Si(Li)-detector (energy resolution 160 eV). The measurements were done with monochromatized Mo-Kα excitation mode (17.48 keV) at ~70% of the critical angle for total reflection of X-rays (1.2 mrad, angle of incidence), for 100 s live time, at 50 kV and 47 mA. Samples were attached to total reflecting quartz reflectors for the TiO2 matrix by using 1 mg of sample mixed with 5 µl of 1% poly-vinyl alcohol solution (for fixation). Blank measurements of the unloaded reflectors were performed prior to each specimen measurement in order to avoid cross contamination. Results for Au and Pd were scaled to 100% mass of Ti.
Detection limits of the quantified elements (Au, Pd and Ti) are in the range of 10-100 µg/g.

UV-Vis spectra and MALDI mass spectra of the nanocluster samples in solution
Both the UV-Vis spectrum of [Au25(SC2H4Ph)18] -( Figure S1, left) and the dominant peak at m/z = ≈7394 in the MALDI mass spectrum ( Figure S1, right) are in good agreement with the reported data. 1 Figure S1. UV-Vis (left) and MALDI mass spectrum (right) of [Au25(SC2H4Ph)18] -.

Total metal loading & Pd/Au ratio of the nanocluster and nanoparticle catalysts
The total metal loadings of the catalysts (Au25/TiO2, PdAu24/TiO2, PdxAuy/TiO2, Au NPs/TiO2, Pd NPs/TiO2) were determined by TXRF. As seen from Table S1, the Pd content in PdAu24/TiO2 was too low for detection by TXRF.

(S)TEM of the TiO2-supported PdAu24, PdxAuy and Au25 nanoclusters
TEM and HAADF-STEM images were acquired for the nanocluster catalysts to determine the average particle size ( Figure S6). For TiO2-supported Au25 and PdAu24 the average diameter was 1.2 ± 0.2 nm, for PdxAuy nanoclusters it was 1.3 ± 0.2 nm. S10 Figure S6.  For PdxAuy/TiO2 catalyst, images were also taken after oxidative, reductive, and sequential treatments (pretO2-H2) and after CO oxidation (post-reaction), as shown in Figure S7. Upon pretreatment and reaction (Table S2)

Reusability of PdAu24/TiO2 in CO oxidation
Reusability test with PdAu24/TiO2 were conducted by subsequent CO oxidation reactions in the flow reactor set-up without removing the catalyst in between. After the first reaction, the sample was cooled to room temperature (under Ar). Once reached, a second CO oxidation reaction was initiated (same for the third catalytic test). As can be seen in Figure S8, activity increases in the 2 nd and 3 rd run, which indicates activation of PdAu24/TiO2 at CO oxidation conditions. Figure S8. Reusability tests with PdAu24/TiO2 in CO oxidation.

Catalytic activity of TiO2 supported Au and Pd NPs
TiO2 supported Au and Pd nanoparticles (NPs) were employed in CO oxidation to contrast their activities to that of the nanocluster catalysts. The catalysts were pretreated by pretO2-H2 as described in the main paper. As can be seen in Figure S9, Au NPs/TiO2 exhibited significantly higher conversion, which gradually increased with rising temperature. The CO conversion of the Pd NPs/TiO2 catalyst was much lower and only increased above 200 °C, due to the much stronger CO poisoning on Pd. .The low activity of PdxAuy/TiO2 indicates the presence of residual protecting ligands (or fragments thereof) even after the pretreatment steps, which are blocking the active sites. Figure S9. Comparison of the catalytic activity of Au NPs/TiO2, Pd NPs/TiO2 and PdxAuy/TiO2.
Activity was normalized to 1 wt% total metal loading and 1 mg of catalyst (as opposed to 15 mg for the measurements described in the main manuscript).

DRIFT spectra: CO dosing and TiO2 blank
CO dosing experiments were performed with the pretreated catalysts before starting the CO oxidation reaction, which DRIFT spectra are depicted in Figure S11. The CO gas phase band is clearly visible in the 1% CO atmosphere spectra with maxima at around 2170 and 2115 cm -1 .
For Au25/TiO2, no significant formation of CO-adsorbate species could be detected. This correlates with its low CO oxidation activity. For the PdAu24/TiO2 catalyst, the spectra strongly depended on the pretreatment conditions. After only pretO2, minimal formation of CO adsorbed on Au could be found, whereas after pretO2-H2, a pronounced CO-Pd band could be detected. This correlates with the better catalytic performance of the pretO2-H2 catalysts.
Blank experiments with the pure TiO2 support without any impregnated Au clusters were performed as well ( Figure S10). CO dosing experiments showed no formation of adsorbed CO species, yet still some CO2 formation could be detected in the in situ DRIFTS measurements during CO oxidation, getting stronger at higher temperature. The same trend could be observed in the kinetic tests (Figure 1b in main manuscript).  the fact that in general, the static structural disorder is higher than the temperature one for small clusters. 9 This approach assures that the coordination numbers obtained from the analysis are consistent with all the experimental information available. Finally, we fit the energy correction to S16 photoelectron reference ΔE0 and the passive electron reduction factors S0 2 , again common to all the spectra. Figure S13. EXAFS fitting.

S17
In order to evaluate if there was a significant contribution of Pd-Pd in the EXAFS, a feff9 simulation 10 was done. The results of the simulation are shown in Figure S14. From these simulations, it could be shown that in a bimetallic particle containing less than 30% of Pd, a small contribution of Pd-Pd cannot be distinguished from Pd-Pd and Pd-Au. This would only by possible if there was a stronger contribution of Pd-Pd bonds. Figure S14. EXAFS and XANES simulation to study Pd-Pd contributions in spectra of PdAu nanoparticles containing less than 30% of Pd.