Structural evolution after oxidative pretreatment and CO oxidation of Au nanoclusters with different ligand shell composition: a view on the Au core

The reactivity of supported monolayer protected Au nanoclusters is directly affected by their structural dynamics under pretreatment and reaction conditions. The effect of different types of ligands of Au clusters supported on CeO2 on their core structure evolution, under oxidative pretreatment and CO oxidation reaction, was investigated. X-ray absorption and X-ray photoelectron spectroscopy studies revealed that the clusters evolve to a similar core structure above 250 °C in all the cases, indicating the active role of the ligand–support interaction in the reaction.

In situ transmission IR studies suggested that this might be related to the formation of ligand residues on the cluster-support interface, which hindered the catalytic reaction. Furthermore, (scanning) transmission electron microscopy ((S)TEM) images indicated cluster growth, but this was not considered to be the main cause of the activity differences. 7 However, a thorough investigation focused on the evolution of the Au core was not conducted at this time. Nevertheless, previous investigations have shown that the removal of ligands often directly causes changes to the metallic core. 5,9,[19][20][21] Thus, both aspects can affect the catalytic activity and should therefore also be considered in structure-activity relationship investigations.
Our previous studies of the Au 38 (SC 2 H 4 Ph) 24 /CeO 2 system by extended X-ray absorption fine structure (EXAFS) measurements have shown that the process of ligand removal leads to more drastic changes than originally expected: the ligands not only migrated onto the support, but also caused a partial migration of Au atoms away from the cluster core. This completely exposed the Au surface and interface at 250 1C, resulting in significantly increased activity in CO oxidation. 5 Jin and coworkers also intensively investigated how the thiolate desorption influences the CO oxidation activity of Au 25 (SR) 18 /CeO 2 catalysts and found that different gold species are formed upon ligand removal, of which the Au d+ (d: between 0 and 1) sites were found to be the major active sites in the catalytic reaction. 9 Our previous study also noticed formation of both Au + and Au d+ sites through the oxidative pretreatment at 250 1C, the former of which mostly converted toward Au d+ sites in the subsequent CO oxidation. 7 For a series of CeO 2 supported phosphine-protected clusters, the core structures appeared stable after thermal treatment at 120 1C while the ligands were shown to desorb from the clusters and interact with the support instead. 12 A similar migration effect was assumed to take place in our previous study as well, however, a definitive proof by probing the state of the Au core was missing. Therefore, in addition to the previous kinetic tests and infrared studies, 7 X-ray absorption spectroscopy (XAS) and X-ray photoelectron spectroscopy (XPS) measurements were performed with the same set of Au 11 / CeO 2 , Biico Au 25 /CeO 2 and Au 25 /CeO 2 catalysts. EXAFS at Au L 3edge indeed showed that the contribution of the ligands is negligible after 200 1C while they could still be detected in the catalyst system at this temperature previously, 7 which confirms that they relocate to the support surface instead. Moreover, it appears that the difference in reactivity at 250 1C might be attributed to the ligand and its interaction with the support, since the final state of the Au cluster core after 250 1C is similar for the three samples. However, after 300 1C, a redispersion of both Au 11 /CeO 2 and Biico Au 25 /CeO 2 toward slightly smaller species was observed, which correlates with increased CO conversion.
Extended X-ray absorption fine structure measurements (EXAFS) at Au L 3 -edge were obtained at the NOTOS beamline of ALBA synchrotron (Spain). The supported cluster samples were grinded and pressed into pellets, which were mounted on a sample holder using Kapton tape. The unsupported clusters were dissolved in dichloromethane and dropcasted on Kapton tape for measuring. Consecutively, the samples were measured at room temperature in fluorescence mode by a Silicon Drift Detector. Data treatment and fitting was performed with the Artemis software package. 22,23 The extracted EXAFS signal after background correction at pre-and post-edge was subsequently k 2 weighted. A model with two shells (with Au-P/S and Au-Au) was used to fit the data, based on the resolved crystal structures of the clusters. [24][25][26][27] Due to the similar scattering paths of Au-P and Au-S, 20 those could not be resolved for Biico Au 25 /CeO 2 and thus only a single path (Au-S) was used for fitting. One Debye Waller factor per different path (Au-Au, Au-P and Au-S) was used. Both an energy correction to the theoretical energy and an amplitude correction were applied. See Table S1 in the ESI † for details. Note that the path contributions are only above the noise level estimated as the intensity of the Fourier transformed EXAFS spectra at R = 10 Å if CN Z 0.3 (see also exemplary illustration in Fig. S5, ESI †).
X-Ray photoelectron spectroscopy (XPS) measurements of the clusters as synthesized, after oxidative pretreatment at 250 1C and after CO oxidation were done on a in situ near ambient pressure (NAP)-XPS system, equipped with a Phoibos 150 NAP hemispherical analyser and a XR 50 MF X-ray source (microfocus), all SPECS GmbH. Monochromatic Al K a radiation was used to acquire the data, which was analysed with the CasaXPS software. Peaks were fitted after linear background subtraction with Gauss-Lorentz sum functions. The spectra were referenced to the Fermi edge and the C 1s signal (284.5 eV). Peak positions and full width at half-maximum (FWHM) were left unconstrained. Due to differential charging, the following samples were measured in 1 mbar N 2 atmosphere instead of UHV conditions: Au 11 as prepared, Au 25 after reaction and Biico Au 25 after oxidation.

Results and discussion
The effects of different pretreatments and reaction temperatures were examined for a set of Au nanocluster catalyst in CO oxidation, as described and discussed in our previous publication. 7 The most striking differences were observed for those samples pretreated and after reaction at 250 1C. Whereas the Au 25 /CeO 2 catalyst showed a sudden increase in activity when pretreated at 250 1C, only about E20% conversion was achieved at that temperature with Au 11 /CeO 2 and Biico Au 25 / CeO 2 . After pretreatment at 300 1C, no such differences were observed and all clusters exhibited high activity. Fig. 1 gives an overview of the conversion of the individual catalysts at the respective maximum temperatures; a more detailed discussion can be found elsewhere. 7 This phenomenon was assumed to be related to the respective ligand environments: while both Au 11 /CeO 2 and Biico Au 25 /CeO 2 were partially protected by phosphine ligands, Au 25 /CeO 2 is a purely thiolate-protected cluster. Indeed, in situ transmission IR studies and temperature programmed oxidation (TPO) measurements further affirmed the assumption that ligand residues were responsible for the different catalytic behaviour since they seemed to block the active sites. However, besides probing the particle size by (S)TEM and detecting an increase in size for all cluster catalysts, no further investigations of the dynamics of the Au core were carried out. 7 Thus, to confirm that the dominant factor for the activity differences noticed was indeed the different original ligand shell, further characterization of the catalysts with focus on the state of the Au core were carried out by XAS and XPS. Fig. 2 shows the Fourier transform of the EXAFS spectra (R-space) and Table 1  The X-ray absorption near edge structure spectra (XANES) are shown in Fig. S1 (ESI †). The unsupported Au nanoclusters (dotted black line) clearly showed peaks corresponding to Au-S/ P and Au-Au distances. For the unsupported Au 11 clusters, the estimated Au-Au and Au-P bond lengths were within the same region as those of similarly sized phosphine protected Au nanoclusters. 12,28,29 Moreover, while a Au-Au coordination number (CN) of 2.3 seems quite reasonable for a cluster of that size, 28 the Au-P CN was with 2.6 much higher than the expected one of 0.63-0.73. The origin of this difference could not be resolved, however, a comparable overestimation was also observed for the CN Au-P/S in Biico Au 25 /CeO 2 . Au 11 /CeO 2 exhibited a decrease in the CN Au-P and a slight increase in the CN Au-Au , indicating signs of ligand loss already during the supporting process. After pretreatment and reaction at 150 1C, the R-space and XANES spectra changed significantly ( Fig. 2 and Fig. S1, ESI †), consistent with previous work showing that partial removal of ligands hinders the stabilization of the cluster. 30,31 This suggests further ligand loss and increasing size of the remaining bare Au core during the reaction. No significant differences are observed at 200 1C with respect to 150 1C, while increasing to 250 1C leads to an increase in CN Au-Au . In this case, there are no representative changes between pretreatment, reaction and three reaction cycles (rep), indicating the stabilization of the clusters. The increase in bond distance R Au-Au from 2.81 to 2.85 Å can be related to the cluster growth, as already observed in previous work. 4,5,30 However, the CN Au-Au is still smaller than fcc gold (CN D 12), and so are the bond lengths (R 4 2.9). 32 For the mixed ligand shell cluster Biico Au 25 , the average CNs for Au-P/S and Au-Au were found to be 1.2 and 2.0, respectively. This presents a deviation from published values for similar icosahedral core structures. 20,33 However, it has been reported previously that the fitted CN AuÀAu values of ligand-protected clusters may be too small owing to differences in the bond lengths within the cluster structure. [34][35][36][37][38][39] The average bond lengths of 2.3 Å (Au-P/S) and 2.8 (Au-Au) are nevertheless in good agreement with literature. 20 Similar to Au 11 /CeO 2 , a decrease of the CN AuÀP/S and an increase of the CN AuÀAu were noticed upon supporting, indicating a first structural change, presumably initiated by a first detachment of ligands.
This trend then further continued with rising temperatures; however, Biico Au 25 /CeO 2 clearly exhibited higher thermal stability than Au 11 /CeO 2 : in this case, the major increase in the Au-Au CN appeared at 200 1C instead of 150 1C. Afterwards, the structure of the Au core did not seem to change significantly besides a slight reduction in the CN Au-Au at 300 1C.
Finally, the evolution of the Au 25 /CeO 2 catalysts also follows a quite similar trend. The unsupported cluster has a fitted CN AuÀS of 1.1 and a CN AuÀAu of 0.8. While the former is in agreement with literature 20 -as are both bond lengths 20,35again a significant underestimation is found for the CN AuÀAu . Upon immobilizing the clusters on the ceria support, only marginal alterations were observed for the Au-S bonds, indicating that the ligand sphere should be mostly intact at this point. However, at the same time, the CN AuÀAu increased to 3.3 and the bond lengths to 2.83, suggesting structural changes due to Each graph shows the spectrum of the unsupported clusters (black dotted), the supported catalysts (fresh; dark blue), the clusters after pretreatment and reaction at different temperatures (COoxXXX 1C; blue or red solid lines), the catalyst after pretreatment at 250 1C but before reaction (pret250 1C; grey) and after 3 consecutive CO oxidation runs (COox250 1C rep; red dashed).
contact with CeO 2 . The on-set of ligand desorption was noticed at 150 1C through a decreasing CN AuÀS and an increasing CN AuÀAu and seemed completed at 200 1C.
This general trend of ligand removal and subsequent slight growth of Au clusters during oxidative pretreatment and reaction has already been observed using EXAFS by several groups. 5,8,10,12,19,20,31 This same trend can also be observed in the XANES spectra of all three catalysts (see Fig. S1, ESI †). For the unsupported clusters, a more intense white line feature (E11924 eV) was detected for the more positively charged P-protected clusters. 12,20 Upon supporting and thermal oxidative treatment, the white line feature was reduced, indicating the Au core becoming more metallic. 12 For all cluster catalysts, the spectrum converged toward the one metallic gold, denoting the removal of partially charged gold related to the Au-ligand bonds, as well as due to an increase in cluster size. It should be noted that the CN values for Au-Au after reaction at 250 1C suggest that the average particles sizes were somewhat larger 20,40 than the 2-3 nm estimated by (S)TEM for Biico Au 25 /CeO 2 and Au 25 /CeO 2 and more toward the 4-7 nm estimated for Au 11 /CeO 2 . 7 However, the average bond lengths of 2.84-2.85 Å are in good agreement with the 2-3 nm estimated from microscopy. 32 Moreover, both the CN AuÀAu and the Au-Au bond lengths of all three samples after reaction at 250 1C showed very similar values, which indicates that they are reasonably close in size.
Based on XAS, it seems that the process of ligand detachment from Au is completed at 200 1C (COox200 1C) for all three clusters. None of the samples showed significant Au-ligand contributions after oxidative pretreatment at 250 1C (pret250 1C) either. In good agreement, CO adsorbed on Au was observed by in situ transmission IR spectroscopy for all samples after pretreatment at 250 1C, which indicated that accessible, unprotected Au surface was available at this point. 7 It thus seems likely that the ligands migrated to the ceria surface already at temperatures below 200 1C, but were not completely decomposed to CO 2 and subsequently removed from the system unless higher temperatures were applied, as evidenced by our previous experiments. 7 It also further affirms that their presence on the support, most likely adjacent to the Au clusters, is a crucial factor for the catalytic activity of these cluster catalysts in CO oxidation, since no striking difference in the Au-Au bond lengths or coordination numbers was observed among the investigated catalysts after reaction at 250 1C. Similar migration of either phosphine 12,21,41 or thiolate 5,31 ligands has also been reported for other systems before.
Interestingly, after reaction at 300 1C, the CN AuÀAu values for Au 11 /CeO 2 and Biico Au 25 /CeO 2 (slightly) decreased with respect to the values obtained after 250 1C, even after reaction cycles (rep), while it remained almost constant for Au 25 /CeO 2 (within error range). The rather pronounced decrease for the Au 11 /CeO 2 samples is also corroborated by changes in the XANES region (Fig. S1, ESI †), which shows an increase in the white line feature. This points toward the formation of more oxidized (smaller) particles.
Previous works have noted that the detachment of the ligand motifs (-S-Au-S-Au-S-) lead to SO x 5,31 species and single or small Au clusters near the main Au cluster core. Therefore, one hypothesis could be that the decrease in the CN AuÀAu observed at 300 1C could be related to the formation of small Au clusters stabilized on the surface around the main cluster core structure once the temperature rises to 300 1C. Previous work on Au 38 clusters supported on CeO 2 has found that this effect seemed to be mediated by the sulphur residues, which appeared to migrate from the cluster surface to the support at pretreatment temperatures of 250 1C, drawing parts of the Au atoms with them. 5 It is possible that a similar effect might result in the formation of smaller particles also in the case of the (partially) phosphine-protected clusters. It is worth noting that this decrease in CN Au-Au appears at the same temperature at which the main ligand decomposition/ removal process seems to take place according to our previous study. 7 It thus might be possible that the P-and S-ligand residues could contribute to the formation of smaller gold particles. This would also explain the apparent changes in the Au 11 /CeO 2 XANES spectrum at 300 1C. Considering that these spectroscopic techniques represent the average signal, the differences in the values could be explained. This change also correlates with a strong increase in catalytic CO oxidation activity of both Au 11 /CeO 2 and Biico Au 25 /CeO 2 (Fig. 1).
In order to complete the study on the oxidation state after pretreatment and reaction at 250 1C, XPS analysis was performed and is displayed in Fig. 3. Au 11 /CeO 2 shows a binding energy of 85.2 eV for the Au 4f 7/2 transition, close to reported values in literature 21,42 with the same cluster but using different support. However, it is significantly larger than the binding energy of bulk Au with around 84.0 eV. It is known that the binding energies of Au in such small clusters can increase due to initial and final state effects. 43 These binding energy values suggest charged Au that can be related to the Au atoms bonded to the ligands. Upon pretreatment and removal of the ligands, a higher proportion of metallic Au is denoted by the decrease in the binding energy value from 85.2 eV to 83.4 eV for the 4f 7/2 transition. However, a second component related to charged gold is still present in the spectrum (Au 4f 7/2 transition at 85.0 eV), which could be denoted to Au atoms bonded to P or strongly interacting with the CeO 2 surface. This is in agreement with reported studies by XPS with similar systems. 21 Furthermore, it also connects to our previous results with CO adsorption experiments follow by in situ IR, showing two different Au species after pretreatment. 7 The comparison of the respective XPS peak areas showed that around 68% of the Au species were in metallic state. After the reaction, only one component was present with a binding energy of 84.0 eV. Due to the relatively low binding energy, the Au species appeared to be in metallic state, in agreement with the XAFS data.
In the case of BiicoAu 25 /CeO 2 after supporting (as prepared), there is good agreement with literature values for the unsupported clusters, 26 indicating that the structure is largely preserved at this step, which can also be seen in the EXAFS and XANES spectra (Fig. 2 and Fig. S1, ESI †). After the oxidative pretreatment, the binding energy of the 4f 7/2 transition increased from 84.5 eV to 85.8 eV. Similar to Au 11 /CeO 2 , a second component is present after oxidative pretreatment (Fig. 3), which has a rather low 4f 7/2 binding energy of 82.8 eV. This most likely metallic component made up about 17% of the analyzed gold. After CO oxidation, the binding energy decreased to 83.8 eV due to the Au core becoming more metallic during the reaction.
For Au 25 /CeO 2 , a similar trend as with Au 11 /CeO 2 can be observed, though not as extreme. The initial binding energy of the Au 4f 7/2 peak is 84.9 eV, which is slightly higher compared to the literature. 17,30 The oxidative pretreatment leads to a decrease in binding energy, indicating formation of metallic gold, presumably due to ligand removal. This has previously also been observed for similar catalyst systems and pretreatments. 5,17 After the reaction, the Au is present in two forms. A metallic component with a binding energy of 83.4 eV for the Au 4f 7/2 transition and an oxidic component with a binding energy of 87.1 eV for the 4f 7/2 transition. It should be noted that the peak for this transition is overlapping with the 4f 5/2 peak for the metallic component, hence only three peaks are present in the spectrum. The metallic component makes up around 45% of the Au present in the sample. However, due to the peaks overlapping this value is not very precise. Furthermore, the still oxidic component is shifted to an extraordinary high binding energy indicating a very electronegative binding partner. However, in situ IR studies of CO adsorption on the same catalyst did not find any presence of Au + species, 7 and, similarly, XANES data also did not reveal Au + . This discrepancy might thus be due to the poorer data quality in this measurement (in 1 mbar N 2 instead of UHV).
Summarizing, a general trend toward more metallic gold species is observed in the XPS measurements for all catalysts. For Au 11 /CeO 2 and Biico Au 25 /CeO 2 , oxidized species can still be detected after oxidative pretreatment. This could indicate that their ligand removal process proceeds differently to the one of the fully thiolate-protected cluster. After reaction, a shift toward lower binding energies has occurred for all clusters (aside from the oxidized component in the Au 25 /CeO 2 spectrum). This is in good agreement with the process of ligand removal and consequently formation of slightly sintered Au particles.

Conclusions
The state of the Au core of Au 11 , Biico Au 25 and Au 25 clusters on ceria used as catalysts for CO oxidation was probed at different stages of the process by ex situ EXAFS and XPS. While previously a strong activity difference at 250 1C was noticed, 7 the clusters showed similar Au-Au coordination numbers and bond lengths after reaction at this temperature. Moreover, most of the Au surface appeared to be free of ligands and thus accessible already from 200 1C onwards for all clusters. Furthermore, significant changes in the Au oxidation state were noticed after oxidative pretreatment at 250 1C by XPS, which shows that the treatment indeed affected the Au core (i.e. the ligand shell is not intact anymore at this point). After reaction, formation of mostly metallic gold was observed for all samples. This affirms the previous assumption that blocking of active sites by ligand residues might be an important factor for the difference in catalytic activity at 250 1C. 7 After reaction at 300 1C, a decrease in the CN Au-Au was noted for both Au 11 / CeO 2 and Biico Au 25 /CeO 2 by EXAFS, which could be related to formation of smaller particles. This structural change appears at the same temperature for which an significant increase in CO oxidation activity was noticed, as well as for which the main removal of the ligand framework from the catalytic system takes place. 7 Therefore, a synergistic effect between the nature of the support, the cluster structure, and the ligand shell seems to determine the surface configuration of the gold nanocluster catalysts after pretreatment and during the reaction, which in turn determines the catalytic behaviour. These results emphasize once again the complexity of the structural changes of such supported nanoclusters during pretreatment and catalytic reaction, and show that they are driven by different effects. A combination of different (in situ/operando) techniques is therefore crucial to understand the evolution of the system.

Conflicts of interest
There are no conflicts to declare.