C-H activation is a key step in organic transformations and is considered very useful in petrochemical, pharmaceutical, and fine chemical industries [1-4]. However, due to the high bond energies of C-H bonds, the catalytic and selective activation of C-H bonds (particularly in alkanes) to produce high value-added compounds is a significant challenge [5]. The key to solve this problem lies in the successful development of efficient catalysts and the comprehensive understanding of the reaction mechanism.

Among the various transition metals, Pd has long been recognized as an efficient component to activate C-H bonds in alkanes [6-8], alcohols [9-14], and aromatic compounds [3, 15, 16]. During the last few years, several efficient Pd catalysts including single-site Pd catalysts [17] and Pd-based bimetallic nanocatalysts [18-22] have been developed to maximize the utilization of Pd. In particular, researchers have systematically investigated the catalytic performances of AuPd nanomaterials toward C-H activation of alcohols [10], toluene [15], and more recently, methane [7]. They have proposed that Au can electronically influence the catalytic properties of Pd. Besides, the low-coordination number corner and edge positions are implicated as active sites for C-H activation. It is clear that the microstructure and electronic structure of Pd sites are crucial to the realization of C-H activation [23, 24]. However, due to the lack of an in-depth understanding of the relationship among Pd location, electronic structure, and catalytic performance, the rational control of the electronic structure of Pd in bimetallic nanocrystals to regulate the catalytic performance is considered highly challenging. The key challenges include: (1) site-specific deposition of Pd atoms on metal nanocrystals, (2) continuous regulation of the surface electronic structure of Pd, and (3) precise determination of the location and electronic structure of Pd at low Pd contents.

Here, we report a systematic study of AuPd bimetallic catalysts performed with the aim of addressing each of the above issues. First, using the high-sensitivity low-energy ion scattering technique to precisely determine the surface Pd-to-Au atomic ratio, we prove that stepwise photodeposition enables site-specific deposition of Pd atoms on Au nanoparticles (NPs). The control of Pd location on Au NPs enables the continuous regulation of the surface electronic structure, which provides an ideal model for the study of the structure-property relationship. Moreover, it is revealed that electron-rich Pd atoms promote C-H activation, thereby significantly reducing the activation energy of benzyl alcohol oxidation. A maximum turnover frequency (TOF; ~500000 h^-1) is achieved by the synthesized AuPd/TiO₂ catalyst whose Pd 3d X-ray photoelectron spectroscopy (XPS) peak downshifts by 1.0 eV as compared to net Pd atoms.

AuPd_x/TiO₂ catalysts were synthesized by a stepwise photodeposition method, with a fixed Au loading of 1 wt% [25, 26]. Typically, 10 mmol L^-1 HAuCl₄ (0.51 mL) and TiO₂ (100 mg) were dispersed in 10 mL of methanol solution in a Pyrex quartz reactor. Before the irradiation process, gaseous Ar was bubbled through the mixture for 30 min to ensure that the solution had no dissolved O₂. The mixture was then subjected to UV light irradiation for 1 h using a high-pressure Xe lamp (300 W) as the light source. During the photodeposition process, the color of the mixture gradually changed, indicating the reduction of the metal precursors. Subsequently, a certain amount (depending on the Pd-to-Au ratio) of 10 mmol L^-1 H₂PdCl₄ was added into the mixture, and the resulting mixture was irradiated for 1 h to obtain AuPd_x/TiO₂. The resulting powder was collected by centrifugation, washed twice with ethanol, and then dried at room temperature in a vacuum oven. Au/TiO₂ and Pd/TiO₂ samples were synthesized by similar procedures.

The oxidation of benzyl alcohol was conducted in a stirred autoclave reactor (30 mL, DHA-M630, China). The reaction condition including oxygen pressure was optimized before the systematic investigation of the synthesized AuPd_x catalysts. Typically, the vessel was first charged with alcohol (4 mL) and the catalyst (10 mg). The autoclave was then purged three times, with oxygen leaving the vessel at 1.0 MPa gauge. Subsequently, the reaction mixture was placed in the heating system for 20 min for the temperature to reach the required value and stirred at 700 rpm. An oxygen cylinder was connected to maintain the pressure. The products were collected and analyzed by gas chromatography mass spectrometry and gas chromatography with a flame ionization detector. The products were identified by comparing with known standards. To quantify the amounts of reactants consumed and products generated, the external calibration method was used.

Transmission electron microscopy (TEM) images were recorded using a Hitachi HT7700 microscope operated at 120 kV after drop casting the nanocrystal dispersions onto carbon-coated Cu grids and drying under ambient conditions. High-resolution TEM (HR-TEM) images, and elemental mapping and line-scanning profiles were obtained using a JEOL JEM2100F electron microscope. Powder X-ray diffraction patterns were recorded on a Rigaku Ultima IV diffractometer with the Cu KR radiation. The metal contents were determined by inductively coupled plasma optical emission spectrometry (ICP-OES, Thermo iCAP6300). The high-sensitivity low-energy ion scattering (HS-LEIS) spectra were obtained using an Ion-TOF Qtac100 instrument. To minimize the surface damage, a neon ion source was used; the kinetic energy was 3 keV, ion flux was 6000 pA m^-2, and spot size was 2 mm × 2 mm.

AuPd_x/TiO₂ (x is the Pd-to-Au molar ratio; Au mass loading was fixed at 1 wt% for all samples) catalysts were synthesized by a stepwise photodeposition method [25, 26], as schematically illustrated in Fig. 1. Firstly, AuCl₄⁻ ions were reduced to Au NPs on TiO₂ by the photogenerated electrons. Owing to the high work function, the as-obtained Au NPs serve as a sink of photoinduced electrons, which subsequently reduce the Pd²⁺ ions on the surface of Au NPs, yielding Au@Pd core-shell structures [18]. According to the literature [27], the hot electrons generated when Au/TiO₂ is subjected to UV irradiation concentrate along the corners and edges of the Au NPs. Accordingly, photodeposition of Pd initially takes place at the corners and edges of Au NPs, and then, on the terraces, forming mono- and multilayers. The location of Pd on Au NPs can therefore be controlled by tuning the Pd-to-Au feeding ratio. The ICP-OES results (Table S1) suggest that the actual loading and Pd-to-Au ratio is accordant with normal value.

For AuPd_x/TiO₂ catalysts with large Pd-to-Au ratios, the Au@Pd core-shell structure could be determined by TEM. Figure 2a shows a typical TEM image of AuPd_1.0/TiO₂. Clearly, the AuPd NPs have a spherical morphology and sizes of around 8 nm. As observed from the line-scanning profile of a representative particle (Fig. 2b), strong Au and Pd signals are detected from the particle center and particle edges, respectively. The elemental mapping images reveal that the diameter of Au distribution is smaller than that of Pd distribution (Fig. 2c). These results clearly indicate that the photodeposited AuPd_x particles have typical Au@Pd core-shell structures. Besides, the AuPd/TiO₂ and Au/TiO₂ samples exhibit similar Au(200) diffraction peaks without visible shifts (Fig. S1), further confirming the formation of a core-shell structure rather than an alloy.

For AuPd_x/TiO₂ catalysts with low Pd contents, because the contrast between Pd and Au was poor, it was difficult to identify the exact location of Pd atoms in the AuPd NPs by traditional approaches including elemental mapping and line-scanning analysis. To address this problem, HS-LEIS, a technique with surface sensitivity and specificity usually used to analyze the outermost atomic layer [28, 29], was employed to determine the surface composition of the photodeposited AuPd_x samples. As shown in Fig. 3a, a variation in the Pd-to-Au ratio leads to changes in the relative intensities of Pd and Au. The surface atomic ratios of Pd were quantitatively analyzed from the Pd-to-Au relative intensity ratio determined by HS-LEIS. For comparison, the theoretical values were calculated by site-specific deposition of Pd on icosahedral Au (~8 nm). Pd atoms were assumed to deposit first on the corners and edges of Au NPs, and then, on the terraces, forming mono- and multilayers. Specifically, a monolayer coverage of Pd on icosahedral Au requires a minimal Pd-to-Au surface ratio of 0.20, while the occupation of all the corners and edges of icosahedral Au by Pd atoms requires a minimal surface Pd-to-Au ratio of 0.001 and 0.05, respectively (see the Supporting Information for details). It is interesting to note that the experimental results roughly fit the theoretical results (Fig. 3b), suggesting that the photodeposition of AuPd proceeded as expected. For a Pd-to-Au ratio lower than 0.2 (red region), the Pd/(Au + Pd) surface atomic ratio linearly increases with the Pd-to-Au ratio, corresponding to the process of Pd monolayer formation on Au [30]. Once Au NPs are fully covered by Pd, a further increase in the Pd-to-Au ratio hardly influences the Pd/(Au + Pd) atomic ratio on the outmost surface.

The electronic structures of photodeposited AuPd_x were investigated by XPS. Due to the overlap of Pd 3d_5/2 and Au 4d_5/2 peaks, Pd 3d_3/2 (Fig. 3c and S2) and Au 4f_7/2 peaks (Fig. S3) were used to analyze the electronic structures of Pd and Au, respectively. The constant Ti and O XPS signals (Fig. S4) indicate that the electronic structure of the TiO₂ support is barely affected by the deposited AuPd particles. Interestingly, as the Pd-to-Au ratio decreases from 1.0 to 0.006, the Au 4f_7/2 peaks slightly upshift to higher binding energies, while the Pd 3d_5/2 peaks downshift to lower binding energies (Fig. 3d). The opposite trends suggest that electrons are transferred from Au to Pd. As compared to those in the Pd monometallic catalyst, Pd atoms in the photodeposited AuPd catalysts are electron rich (denoted as Pd^δ-). Besides, no visible Pd²⁺ peaks are detected in samples with Pd-to-Au ratios below 0.2. Similar results have been reported in the literature [31, 32]. A sharp change in the Pd^δ- binding energy is observed during Pd monolayer formation (Pd-to-Au ratio less than 0.2), suggesting a significant variation of the electronic structure of the Pd atom (Fig. 3d). A charge compensation model in which Au donates d electrons to Pd and accepts sp electrons was proposed to interpret the interaction between Au and Pd [31, 32].

More importantly, the variation of Pd 3d_5/2 and Au 4f_7/2 peak positions with the Pd-to-Au ratio implies that the electronic structure of Pd atoms in AuPd_x can be regulated by changing the Pd atomic locations. Pd atoms located at the corners of AuPd_0.006 are the most electron rich, featuring a 1.0 eV downshift in the binding energy as compared to that of the net Pd. Pd atoms in the Pd mono- and multilayers are less electron rich, resembling those of monometallic Pd catalysts. Interestingly, once AuPd_0.05 is calcined at 350 ℃ (denoted as AuPd_0.05-calcined), Pd atoms tend to diffuse into the bulk of Au NPs or aggregate into large particles. Subsequently, the Pd/(Pd + Au) surface atomic ratio declines from 0.23 to 0.01 (Fig. 3b). The rearrangement of the surface Au and Pd atoms eventually changes the electron-rich structure of Pd atoms (Fig. S5).

Figure 4a shows the relationship between Pd-to-Au ratios and TOF values of AuPd catalysts for solvent-free oxidation of benzyl alcohol; the structural models of the catalysts are also shown. The number of Pd atoms needed for covering the corners, edges, and facets are calculated using an icosahedral model (see the Supporting Information for details); the corresponding conversion and selectivity data for benzyl alcohol oxidation over AuPd_x are listed in Table S2. Basically, all AuPd_x catalysts show good selectivity toward benzaldehyde (> 80%). The main by-products include toluene, benzoic acid, and benzyl benzoate. Interestingly, Au/TiO₂ hardly shows any activity for benzyl alcohol oxidation at the given reaction conditions, excluding Au atoms as the active sites. Similar results have been reported in the literature [41]. Accordingly, we considered only the surface Pd atoms as the active sites and defined the TOF value as the number of benzyl alcohol molecules reacted per hour per surface Pd atom. It is important to highlight that AuPd_0.001 exhibits the maximum TOF of ~500000 h^-1, which is, to the best of our knowledge, the highest value ever reported for solvent-free oxidation of benzyl alcohol at similar reaction conditions (Table 1). As the Pd-to-Au ratio increases to 0.03 and 0.05, Pd atoms begin to deposit on the edges of Au NPs and become less electron rich. As a result, the TOF declines to ~200000 h^-1. By plotting the TOF as a function of the Pd binding energy (Fig. 4b), we can see that the TOF decreases almost linearly as the binding energy increases from 339 eV (AuPd_0.006) to 339.7 eV (AuPd_0.2). These results clearly demonstrate that electron-rich Pd atoms are more active in benzyl alcohol oxidation. As the Pd-to-Au ratio further increases, forming Pd monolayer and multilayers, the electron density of Pd decreases and approaches that of bulk Pd. Correspondingly, the TOF of AuPd declines, and finally, becomes the same as that of Pd.

To clearly elucidate the mechanism behind the electronic structure dependence, kinetic studies were conducted. As shown in Fig. 4c, the activation energy (E_a) of AuPd catalysts decreases from 127.0 kJ mol^-1 for AuPd_0.60 to 30.5 kJ mol^-1 for AuPd_0.006, indicating that electron-rich Pd atoms significantly reduce the activation energy of the reaction, accelerating the reaction. It is generally accepted that C-H and O-H cleavage are the key steps in alcohol oxidation [42-44]. To examine the reaction rate-determining step, kinetic isotope effect (KIE) was investigated using benzyl-d2 alcohol (C₆H₅CD₂OH) and benzyl aldehyde-OD (C₆H₅CH₂OD) as substrates over AuPd_0.05 [45]. As shown in Fig. 4d, a significant KIE (k_H/k_D = 11.34) is observed for benzyl-d2 alcohol; however, the replacement of benzyl alcohol by benzyl aldehyde-OD barely changes the TOF (k_H/k_D = 1.11). These results indicate that the cleavage of C-H instead of O-H is the rate-determining step for benzyl alcohol oxidation over AuPd catalysts. Moreover, we investigated the substituent effect with 4-methoxybenzyl alcohol and 4-nitrobenzyl alcohol as reactants [44]. As shown in Fig. 4d, the presence of the methoxy group (electron-donating group) at 4-position remarkably increases the reaction rate, while the 4-nitro group (electron-withdrawing group) decreases the reaction rate. These results suggest that a rich electronic environment is crucial for C-H activation in aerobic alcohol oxidation [46, 47]. It is interesting to note that after the calcination of AuPd_0.05 at 350 ℃ to demolish the electron-rich Pd atoms (Fig. S5), the TOF rapidly decreases from 212000 to 5100 h^-1. Combining all these results together, we can conclude that electron-rich Pd atoms improve the reaction rate by promoting the C-H bond activation, decreasing the activation energy.

In this study, we have systematically investigated the relationship among the Pd location, surface electronic structure, and catalytic performance. Electron-rich Pd atoms in AuPd bimetallic NPs are revealed to facilitate C-H bond activation. Considering the recent success of Au-Pd catalyst application in methane selective oxidation [48], electron-rich Pd atoms could be a future hot spot for alkane activation. Furthermore, designing relative position of two metals to control the active sites in catalyst shines a light on the development of high-activity and cost-effective noble metal catalysts.

Chemicals, XRD patterns, additional XPS spectra. This material is available free of charge via the Internet.