Resolving the puzzle of single-atom silver dispersion on nanosized γ-Al2O3 surface for high catalytic performance

Ag/γ-Al2O3 is widely used for catalyzing various reactions, and its performance depends on the valence state, morphology and dispersion of Ag species. However, detailed anchoring mechanism of Ag species on γ-Al2O3 remains largely unknown. Herein, we reveal that the terminal hydroxyls on γ-Al2O3 are responsible for anchoring Ag species. The abundant terminal hydroxyls existed on nanosized γ-Al2O3 can lead to single-atom silver dispersion, thereby resulting in markedly enhanced performance than the Ag cluster on microsized γ-Al2O3. Density-functional-theory calculations confirm that Ag atom is mainly anchored by the terminal hydroxyls on (100) surface, forming a staple-like local structure with each Ag atom bonded with two or three terminal hydroxyls. Our finding resolves the puzzle on why the single-atom silver dispersion can be spontaneously achieved only on nanosized γ-Al2O3, but not on microsized γ-Al2O3. The obtained insight into the Ag species dispersion will benefit future design of more efficient supported Ag catalysts.

A lumina-supported silver (Ag/Al 2 O 3 ) is a widely used catalyst in a variety of industrial applications, such as NO x emission control via selective catalytic reduction (SCR) with using hydrocarbons (HC-SCR) [1][2][3][4][5][6] or NH 3 (NH 3 -SCR) 7,8 , soot oxidation 9,10 , selective catalytic oxidation of ammonia (NH 3 -SCO) [11][12][13] , oxidation of volatile organic compounds 14,15 , ethylene epoxidation 16,17 , among others. The valence state, dispersion, and morphology of Ag species have been found to play important roles to the activities of Ag-based catalysts as these factors are all closely related to how Ag species are anchored on the Al 2 O 3 surface. Hence, it is of both fundamental and technical importance to uncover the anchoring mechanism of Ag species on Al 2 O 3 for the design of more efficient Ag/Al 2 O 3 catalysts.
The strong metal-support interaction has been often identified as a key factor to affecting metal anchoring on oxide surfaces [18][19][20] . Both previous experiments 21 and DFT calculations 22 have shown that the presence of electronic defects on reducible oxides (e.g., CeO 2 and TiO 2 ) is the direct cause for the strong interaction between metal and support 19,23 . Due to the lack of electronic defects on nonreducible oxides such as Al 2 O 3 and SiO 2 , some other factors must be responsible for the dispersion and thermal stability of active metals. Kwak et al. 23 reported that unsaturated pentacoordinate Al 3+ (Al 3+ penta ) centers on the γ-Al 2 O 3 surface, created by dehydration and dehydroxylation at 573 K, are the anchoring sites for Pt to entail strong interactions with γ-Al 2 O 3 through oxygen bridges. They proposed that the coordinative saturation of Al 3+ penta sites is the driving force for Pt anchoring on γ-Al 2 O 3 . However, the source of the oxygen bridges was not explained specifically. Surface OH groups have also been shown to influence the dispersion of active metals on support (η-Al 2 O 3 and SiO 2 ) 24,25 , while the interaction between metals and OH groups has not been elucidated at the atomic level. Despite Ag/γ-Al 2 O 3 catalysts have been extensively investigated, particularly on their valence state, dispersion, or morphology of Ag species, detailed anchoring mechanism of the Ag species on γ-Al 2 O 3 surface is still unclear to date. Hence, it is timely to explore whether the Al 3+ penta sites, OH groups, or other sites are responsible for anchoring Ag species on γ-Al 2 O 3 , as well as the associated chemical mechanism for singleatom Ag dispersion.
Here, we select two kinds of γ-Al 2 O 3 , nanosized γ-Al 2 O 3 and microsized γ-Al 2 O 3 (details given in Supplementary Information Methods), as the support for the preparation of Ag/Al 2 O 3 catalysts. Most surprisingly, we observe that Ag can be atomically dispersed on the nanosized γ-Al 2 O 3 , whereas Ag mainly exits in the form of nanoclusters on the microsized γ-Al 2 O 3 . By using a series of surface-science measurements and Density-functional-theory (DFT) computation, we reveal that the terminal hydroxyl groups on γ-Al 2 O 3 surface are responsible for anchoring the Ag species on γ-Al 2 O 3 . More importantly, we find that the presence of abundant terminal hydroxyl groups on the nanosized γ-Al 2 O 3 leads to the single-atom dispersion of Ag species on the γ-Al 2 O 3 surface. As expected, the Ag/Al 2 O 3 with single-atom dispersion of Ag species demonstrates markedly higher performance than cluster-dispersion of Ag species for the SCR of NO (SCR-NO) reaction.

Results
Ag dispersion on nanosized γ-Al 2 O 3 and microsized γ-Al 2 O 3 . Nanosized γ-Al 2 O 3 (nano-Al 2 O 3 , average size: 10 nm) and microsized γ-Al 2 O 3 (micro-Al 2 O 3 , average size: 5 μm) were employed as the supports for the preparation of Ag/Al 2 O 3 catalysts. HAADF-STEM images were taken to reveal the Ag morphology and dispersion on Ag/nano-Al 2 O 3 with different Ag loadings (1, 2, 4%) and 1% Ag/micro-Al 2 O 3 (see Fig. 1). As shown in (Fig. 1a,  atomically dispersed Ag species were observed for the 1% Ag/nano-Al 2 O 3 sample, and Ag species still maintained atomic dispersion while a very few sub-nanometer Ag cluster species were also present on 2% Ag/nano-Al 2 O 3 . When the Ag loading was increased up to 4% (Fig. 1c), more Ag clusters appeared alongside Ag atoms, and the lattice spacing of Ag clusters was 0.235 nm, corresponding to the (111) planes of cubic metallic Ag (JCPDS file no. 01-071-3762) [26][27][28] . In contrast, as presented in Fig. 1d, single-atom Ag was rarely observed on 1% Ag/micro-Al 2 O 3 , while Ag nanoparticles of 2 nm average size (with lattice spacing = 0.235 nm) comprised the main morphology for Ag species. These results indicate that 1% Ag can be atomically dispersed on nano-Al 2 O 3 , whereas the Ag species agglomerates into nanoparticles on micro-Al 2 O 3 .
HC-SCR activity of Ag/Al 2 O 3 with atomically dispersed Ag. Ag/Al 2 O 3 catalysts have been extensively explored for the SCR of NO reaction, where both isolated Ag + species and Ag n δ+ clusters are generally considered as the active centers 1,29-31 . We tested the 1% Ag/nano-Al 2 O 3 and 1% Ag/micro-Al 2 O 3 catalysts for C 3 H 6 -H 2 -SCR and ethanol-SCR of NO to investigate the effect of Ag dispersion on catalytic activity. The results are shown in Fig. 2a, b, respectively. The 1% Ag/nano-Al 2 O 3 presented much higher activity than 1% Ag/micro-Al 2 O 3 for both C 3 H 6 -H 2 -SCR and ethanol-SCR, demonstrating that the atomically dispersed Ag species on Ag/nano-Al 2 O 3 are markedly superior to the Ag nanoparticles on Ag/micro-Al 2 O 3 . These results suggest that isolated Ag + species are more active than Ag n δ+ clusters or Ag nanoparticles for NO x reduction. After HC-SCR of NO testing, the morphology of single-atom dispersed Ag on 1% Ag/nano-Al 2 O 3 catalyst was investigated again by using HAADF-STEM, and the images are shown in Fig. 2c, d. There was no sign of agglomeration of single-atom Ag species on spent Ag/nano-Al 2 O 3 sample, indicating the single-atom dispersed Ag on the nano-Al 2 O 3 catalyst was very stable.
Charge state of Ag species and its connection with γ-Al 2 O 3 . To solve the puzzle regarding the distinct dispersion behavior of Ag species on nano-Al 2 O 3 and micro-Al 2 O 3 sample, we first attempted to identify the anchoring site and the associated chemical mechanism for Ag species on γ-Al 2 O 3 . We noted that Kwak et al. 23 reported that Al 3+ penta centers on γ-Al 2 O 3 , created by dehydration and dehydroxylation at 573 K (see the diagram in Fig. 3d), are the anchoring sites for Pt, and that the coordinate saturation is the driving force due to the changes in NMR signal of Al 3+ penta centers after Pt loading. Hence, first, we examined the possibility of Ag anchoring on Al 3+ penta centers at room temperature by using 27 Fig. 1) show typical diffraction peaks of γ-Al 2 O 3 (JCPDS 02-1420), while no Ag-containing phases appear until the Ag loading reaches 4 wt%, indicating that Ag species are in high dispersion at low loadings. Previously, we measured the UV-vis profiles of Ag/nano-Al 2 O 3 and Ag/micro-Al 2 O 3 with various Ag loadings 12 . The UV-vis spectra of 1% Ag/nano-Al 2 O 3 and 1% Ag/ micro-Al 2 O 3 after the Kubelka-Munk transformation are presented in Supplementary Fig. 2a. The spectra were deconvoluted into Gaussian subbands on the basis of the assignments. Next, the percentages of Ag species were calculated based on an analysis of the integrated peak areas. As shown in Supplementary Fig. 2b, the percentage of isolated Ag + was more than 95% on nano-Al 2 O 3 , while about 60% of Ag species existed in the Ag n δ+ state on micro-Al 2 O 3 . These UV-vis results reveal that the Ag species on nano-Al 2 O 3 are mainly the isolated Ag + ions at low Ag loading (1-2 wt%) and form the metallic Ag n clusters at relatively high Ag loading (4-8 wt%) 6,30,32-37 , whereas even 1% Ag on micro-Al 2 O 3 tends to form Ag n δ+ clusters, consistent with the HAADF-STEM images. XAFS measurements were then conducted to study the valence and coordination condition of Ag species on the nano-Al 2 O 3 surface. Ag-K edge X-ray adsorption spectra of Ag/ nano-Al 2 O 3 with different Ag loadings (2, 4, and 8 wt%) were measured, together with those of Ag foil, Ag 2 O, and AgNO 3 as references. The normalized near-edge structure (XANES) of the samples (Fig. 3b) further shows that the Ag was in the oxidized state at low Ag loadings (2 wt%) and closer to the metallic state at high Ag loadings (4 and 8 wt%) 38  Ag loadings (2, 4, and 8 wt%) are shown in Fig. 3c. The peak at about 1.7 Å was assigned to Ag-O interactions 40,41 , and it exhibits similar peak intensities in the spectra of the three Ag/nano-Al 2 O 3 samples. An intense peak at 2.67 Å also appeared when Ag loading was increased to 4 or 8%. This peak is similar to that of Ag foil, and was therefore attributed to Ag-Ag metallic bonding, indicating the formation of Ag clusters in the metallic state (Ag n 0 ) 40,41 . In addition, no Ag-O-Ag shell was detected in any of the Ag/nano-Al 2 O 3 samples.
These results above demonstrate that Ag species initially connect to Al 2 O 3 through oxygen bridges during Ag loading, and then aggregate to form metallic Ag when the oxygen bridges are depleted with increasing the Ag loading. Hence, it is critical to find the origin of oxygen bridges in order to understand the Ag anchoring mechanism. All of Ag species are in ionic state during the impregnation step, therefore the oxygen bridge should be only originated from the γ-Al 2 O 3 substrate. Surface oxygen atom generally exits in a form of hydroxyl groups, and it is known that γ-Al 2 O 3 surface entails abundant and diverse surface hydroxyl groups. Since Ag + has high electron affinity 42 , we speculated that Ag species are mainly anchored by the Al-OH sites through interaction between Ag + and O atoms. If so, the oxygen bridges would remain unchanged after the Ag loading (as schematically shown by the diagram in Fig. 3e).
Correlation between Ag loading and hydroxyl content. The relationship between hydroxyl content and Ag loading was investigated through the 1 H MAS NMR method and the results are displayed in Fig. 4a. The overlapping spectra were deconvoluted, and three peaks centered at −0.2, 1.3 and 4.0 ppm chemical shifts were fitted for the nano-Al 2 O 3 support (zero Ag loading, 0% Ag). According to the literatures [43][44][45][46] , the bands at −0.2 and 1.3 ppm were assigned to HO-μ 1 (Al IV ) and HO-μ 2 (Al v , Al n ), respectively, and the broader band at 4.0 ppm was assigned to the HO-μ 3 (Al n ) and residual water [here, μ 1 , μ 2 , μ 3 represent terminal (type I), doubly bridging (type II), triply bridging (type III) hydroxyls]. Loading Ag onto the nano-Al 2 O 3 induced a substantial reduction in the peak intensity of the terminal hydroxyl groups (−0.2 ppm), evidencing the sharp decrease in the number of hydroxyl groups with 1 or 2% Ag loading. In contrast, no obvious decrease in peak intensities was observed for doubly and triply bridging hydroxyls. These results show that the terminal hydroxyl groups on Al 2 O 3 surface are responsible for anchoring the Ag species. In addition, a new peak at around 0.2 ppm arisen with the Ag-loaded samples, thereby being relevant to the Ag species. Thus, the new peak was tentatively assigned to a new type of doubly bridging hydroxyls [type II', HO-μ 2 (Ag, Al n )], owing to the anchoring of Ag on terminal hydroxyl groups.
Surface OH groups on Al 2 O 3 can act as acid sites, and NH 3 adsorption could take place on these sites. Therefore, monitoring the changes in OH-related peaks by FTIR during NH 3 adsorption can be used to characterize the types, strengths, and concentrations of surface OH groups. If Ag species occupy the OH sites when being loaded on γ-Al 2 O 3 , the intensities of negative OH peaks would decrease since fewer OH sites are available for NH 3 Fig. 4 1 H NMR and in situ DRIFTS spectra of Ag/γ-Al 2 O 3 . a Normalized 1 H MAS NMR of Ag/nano-Al 2 O 3 with different silver loadings (0, 1, 2, 4, 8%) (Samples were dehydrated at 473 K for 12 h before NMR measurements), b peak resolving of OH consumption peaks after in situ DRIFTS of NH 3 adsorption over Ag/nano-Al 2 O 3 with different silver loadings (0, 1, 2, 4, 8%), c relationship between Ag content and peak areas of hydroxyl groups, d peak resolving of OH consumption peaks after in situ DRIFTS of NH 3 adsorption over nano-Al 2 O 3 and micro-Al 2 O 3 .
Ag anchoring and surface hydroxyl groups (see Supplementary  Fig. 3). The bands at 1689, 1478, and 1395 cm −1 were attributed to the vibrations of NH 4 + chemisorbed on Brønsted acid sites [47][48][49] . The bands at 1614 and 1239 cm −1 were assigned to NH 3 on Lewis acid sites [49][50][51] . It is well established that γ-A1 2 O 3 surfaces contain various types of OH groups, characterized by their O-H stretching frequencies. The negative bands at 3775, 3729, and 3668 cm −1 were assigned to the occupation of isolated hydroxyl group of type I, II, and III 52-60 , respectively. The peak intensities for the surface OH groups 52,53 clearly decreased after Ag species were loaded, indicating that some OH sites were occupied by Ag species.
The overlapping peaks related to OH occupation over Ag/ nano-Al 2 O 3 with different silver loadings (0, 1, 2, 4, 8%) were deconvoluted (see Fig. 4b, d). Four subpeaks were assigned to three types of OH groups (type I, type II, type III) and the newly formed OH groups (type II′), respectively. The peak areas were calculated by integration, and the correlations between Ag content and hydroxyl peak areas were next analyzed. As shown in Fig. 4c, the peak area of terminal OH groups (type I) was sharply dropped at Ag loading of 1 and 2%, but further increasing the Ag content to 4 and 8% induced no obvious change in peak area. Meanwhile, a new type of doubly bridging OH groups (type II′) appeared at 3750 cm −1 after Ag was loaded. In contrast, there was no correlation between Ag loading and the other two types of OH groups (type II, III). These results strongly support that Ag species are anchored on the Al 2 O 3 surface through interaction between Ag and terminal hydroxyl groups (type I), while producing a new type of hydroxyl group.
We also measured the concentration of surface OH groups on microsize Al 2 O 3 using in situ DRIFTS spectra of NH 3 adsorption, and compared with those on nano-Al 2 O 3 . As shown in Fig. 4d, the content of terminal hydroxyl groups in micro-Al 2 O 3 was only 6.3 %, about ten times lower than that in nano-Al 2 O 3 (63.8%). Since the nano-Al 2 O 3 contains abundant terminal hydroxyls for Ag anchoring, 1% Ag can be dispersed in single-atom Ag fashion on nano-Al 2 O 3 . In contrast, there are insufficient number of terminal hydroxyl groups on micro-Al 2 O 3 for anchoring 1% Ag. As such, Ag species aggregate into Ag cluster on micro-Al 2 O 3 .
Anchoring mechanism of Ag on terminal hydroxyl groups. DFT calculations were carried out to investigate the interaction between Ag species and hydroxyl groups. Previous studies have shown that the (110) surface dominates on γ-Al 2 O 3 , which occupies about 70% of the total area, followed by the (100) surface (∼20% of the total area) 57,61 . The relaxed structures of the dehydrated (100) and (110) surfaces of γ-Al 2 O 3 were shown in Supplementary Fig. 4. A total of 16 Al atoms and 24 O atoms were exposed on (110) surfaces of 2 × 2 γ-Al 2 O 3 , where Al atoms were classified as Al trip and Al tetra , and O atoms were O two and O trip , respectively. A total of 16 Al atoms and 24 O atoms were also exposed on the (100) surfaces of 2 × 2 γ-Al 2 O 3 . However, Al atoms were Al tetra and Al penta , and O atoms were only O trip , respectively. Since our catalysts were prepared by using the impregnation method in DI-water, the (100) and (110) surfaces of γ-Al 2 O 3 in the calculation were considered to be 100% hydroxylated. The hydroxyl surface coverage of γ-Al 2 O 3 (100) and (110) facet in this work is 16.5 OH nm −2 and 14.7 OH nm −2 , respectively. These results are within the range of OH surface coverage reported by Digne et al. 61 . We further calculated the stretching frequencies of the OH species on the (100) and (110) surfaces. The results are given in Supplementary Table 1. It is shown that the calculated frequencies are in good agreement with the data previously reported 61 , and also in line with the assignment of the experimental values (Fig. 4b).
The hydroxylated (110) and (100) surfaces of γ-Al 2 O 3 were then relaxed (see Fig. 5a). For the (110) surface, there are three types of hydroxyl groups (types I-III), and the fraction of terminal hydroxyl groups (type I) is~30%. For the (100) surface, surface reconstruction occurs after hydroxylation, and all Al atoms become six-coordinated. So, only two types of hydroxyl groups (type I and type III) are present on the (100) surface, and the fraction of terminal hydroxyl groups (type I) is 52%. These results indicate that terminal hydroxyl groups (type I) are more easily formed on (100) surfaces of γ-Al 2 O 3 . XRD profiles ( Supplementary Fig. 5) show that nano-γ-Al 2 O 3 exhibited more (100) surfaces than micro-γ-Al 2 O 3 , supporting that nano-γ-Al 2 O 3 entails much more terminal hydroxyl groups (type I) (see Fig. 4).
We then calculated the interaction between Ag atom and different types of hydroxyls on the (110)  hydroxyls (type II), and one type of triply bridging hydroxyls (type III). The relaxed structure of Ag binding on the hydroxylated (110) surfaces of γ-Al 2 O 3 are presented in Supplementary Fig. 6. It is shown that Ag species cannot be stably bonded to any sites, demonstrating that the (110) surface of γ-Al 2 O 3 cannot anchor Ag species.
For the (100) surface, only terminal hydroxyls (type I) are on the surface. A stable dumb-bell-like double diamond structure with six terminal hydroxyl groups was formed on the (100) surface (Fig. 5a, blue dotted lines). The relaxed structure of Ag binding on the (100) surfaces of γ-Al 2 O 3 is shown in Fig. 5b. Despite the Ag atom is adjacent to O atom of terminal hydroxyls (type I) or triply bridging hydroxyls (type III) in the initial model, the optimized models show that Ag can be only stably bonded to terminal hydroxyls (type I), and a staple-like structure was formed by consuming two or three terminal hydroxyls (type I), producing a new type of doubly bridging hydroxyls (type II′) (consistent with NMR and DRIFTS results shown in Fig. 4). The adsorption energies of Ag in the model I and II are −5.69 and −3.85 eV, respectively, indicating that Ag species are strongly anchored on terminal hydroxyls (type I).
The relaxed structures of second Ag atom binding on the (100) surfaces of γ-Al 2 O 3 are shown in Fig. 5c. The second Ag atom was still bonded to terminal hydroxyls (type I), and the adsorption energies of Ag in the model III and model IV are −1.69 and −2.99 eV, respectively. These results indicate that when Ag loading is low and sufficient terminal hydroxyl groups are available, Ag species are mainly anchored on the terminal hydroxyl group sites and easily form a stable single-atom Ag dispersion. However, when Ag loading is high, all terminal hydroxyls are consumed, Ag no longer can be anchored at hydroxyl sites, and it then easily agglomerates and forms metallic Ag cluster during high temperature calcination.
Kwak et al. have suggested that Al 3+ (Al 3+ penta ) centers are the anchoring sites for Pt species on γ-Al 2 O 3 23 . Based on the present results, we propose that Pt anchoring on the γ-Al 2 O 3 might be also closely related to the terminal hydroxyl groups rather than Al 3+ penta . We measured and compared the 27 Al MAS NMR signals of nano-Al 2 O 3 pretreated with different conditions. The results are presented in Supplementary Fig. 7. When γ-Al 2 O 3 was dehydroxylated at 673 K for 12 h, the NMR signal of Al 3+ penta was clearly detected. However, after the dehydroxylated nano-Al 2 O 3 was exposed to ambient air for 12 h, or impregnated in DIwater for 2 h followed by dryness at 378 K for 6 h, the NMR signal of Al 3+ penta almost disappeared, indicating that Al 3+ penta sites can be easily coordinatively saturated by interaction with H 2 O. These results demonstrate that Al 3+ penta sites are unlikely available on γ-Al 2 O 3 for Ag/Pt anchoring during the catalyst preparation by using impregnation method (in DI-water) whether γ-Al 2 O 3 was pretreated or not. We also calculated the adsorption of Ag/Pt atoms on the Al 3+ penta centers. As shown in Supplementary  Figs. 8 and 9, both Ag and Pt atoms can only coordinate with two O atoms adjacent to Al 3+ penta but not directly with Al 3+ penta sites. The calculated adsorption energies of Ag/Pt on O atoms adjacent to Al 3+ penta were also much weaker than the adsorption energies of Ag/Pt on the terminal hydroxyl groups, indicating that both Ag and Pt atoms tend to interact with terminal hydroxyl groups even if the Al 3+ penta center is available on γ-Al 2 O 3 . Thus, we believe that the terminal hydroxyl groups are also responsible for the anchoring of Pt on γ-Al 2 O 3 .
In summary, this comprehensive study reveals that, other than electronic defects and Al 3+ penta centers, the hydroxyl group is the key site for metal-support interactions that affects the valence state, morphology, and dispersion of Ag species on γ-Al 2 O 3 . Based on the measurements of MAS NMR, in situ DRIFTS, HAADF-STEM, we show that the Ag species are mainly anchored by the terminal hydroxyl groups (type I) on γ-Al 2 O 3 through interaction between Ag and the O atom in Al-OH groups. Nanosized γ-Al 2 O 3 entails abundant terminal hydroxyl groups so that Ag could be atomically dispersed on nano-Al 2 O 3 , while Ag tends to agglomerate to form the Ag clusters on micro-Al 2 O 3 since the number of terminal hydroxyl groups are insufficient for Ag binding. Single-atom Ag dispersion on Ag/nano-Al 2 O 3 gives rise to markedly higher catalytic performance in HC-SCR of NO, compared with Ag-cluster dispersion. DFT calculations confirm that it is the (100) surfaces of γ-Al 2 O 3 that can accommodate much more terminal hydroxyl groups than the (110) surfaces. As such, single Ag atom can be only anchored by the terminal hydroxyls on the (100) surfaces through consuming two or three terminal hydroxyls and forming new doubly bridging hydroxyls. Our study resolves the puzzle on why the single-atom Ag dispersion can be spontaneously achieved only on surface of nanosized γ-Al 2 O 3 since nanosized γ-Al 2 O 3 entails predominantly the (100) surfaces. Activity test. The activity test for HC-SCR was performed at steady state in a fixed-bed flow reactor with the gas hourly space velocity (GHSV) of 100,000 h −1 . The mixture gas of C 3 H 6 -H 2 -SCR is NO 800 ppm, C 3 H 6 1565 ppm, H 2 1%, O 2 10%, and N 2 balance. The mixture gas of ethanol-SCR is NO 800 ppm, C 2 H 5 OH 1565 ppm, O 2 10%, H 2 O 5%, and N 2 balance. An online FTIR spectrometer was used to continuously analyze the concentrations of NO x .
MAS NMR. 27 Al and 1 H NMR spectra were recorded at 11.7T on a Bruker-Advance III 500 spectrometer equipped with a 4 mm double-resonance probe. The corresponding resonance frequency and magic angle spinning rate were 500.57 MHz and 10 kHz respectively. 27 Al NMR measurements were conducted with no pretreatment on the samples. For 1 H NMR measurements, samples were predehydrated on a vacuum line. The temperature was gradually increased at a rate of 1 K min −1 and the sample was kept at a final temperature of 473 K and at a pressure below 10 −3 Pa for 12 h. The average weight of samples is 0.0513 g and the deviation for samples is lower than 2.7%, all NMR spectra were normalized to the sample weight.
XAFS. Ag K-edge XAFS spectra were recorded in transmission mode at room temperature, using the BL14W1 XAFS beam line at the Shanghai Synchrotron Radiation Facility. The XANES and EXAFS data reduction and analysis were performed using the Athena program that is part of the IFFEFIT software package 62,63 . The filtered k 2 weighted χ(k) was Fourier-transformed into R space (k range: 2.9-14.3 Å −1 for Ag-K EXAFS).
In situ DRIFTS. The in situ diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) was conducted on a Nexus 670 (Thermo Nicolet) FTIR equipped with an MCT/A detector. All spectra were recorded in the range 4000-800 cm −1 by accumulating 100 scans with a resolution of 4 cm −1 . A background spectrum was subtracted from each spectrum. Samples were pretreated at 473 K for 30 min in a flow of N 2 + O 2 (20%) before in situ DRIFTS measurements.
DFT calculations. Geometries and energies were calculated using the DFT method in the formalism of Perdew-Burke-Ernzerhof (PBE) functional 64 with van der Waals correction proposed by Becke-Jonson (i.e., DFT-D3 method) 65 , as implemented in the Vienna ab initio simulation package (VASP 5.4.4) 66 . The projector augmented wave method was used to describe the interaction between the ions and the electrons 67 . Convergence tests were performed for all initial parameters. The vacuum gap was 20 Å so that inter-slab interactions were negligible in the periodic systems. DFT calculations were carried out with a plane-wave energy cutoff of 400 eV. Based on our previous reports 68, 69 , the dehydrated (110) and (100) surfaces of γ-Al 2 O 3 were modeled using (2 × 2) supercells. According to our preliminary convergence test, Monkhorst−Pack k-point sets of (1 × 1 × 1) were used for the Al 2 O 3 surfaces. The top two layers and the adsorbents were fully relaxed, while the bottom layers were fixed to mimic the bulk region. The adsorption energies of Ag/ Pt on the γ-Al 2 O 3 surface were calculated as follows: E ad = E adsorbate + surface -(E surface + E adsorbate ), where E adsorbate+surface and E surface are the total energies of the adsorbed system and alumina slab, respectively; and E ad reflects the stability of the adsorbates on the γ-Al 2 O 3 surface. Negative E ad values mean that the adsorbed state is energetically favorable. The DFT-predicted adsorption energies of Ag/Pt atom are with respect to the energy of a gas-phase Ag atom. All the DFT computations were spin polarized. We compute the adsorption energy of Ag on the γ-Al 2 O 3 (110) surface, with (−5.76 eV) and without dipole correction (−5.69 eV). The difference in the adsorption energy of Ag is less than 1.2%. Consequently, all calculation results do not include the dipole correction.