Since the introduction of the concept of catalytically active sites (CASs) by Taylor in 1925 [1], great efforts have been devoted to identifying CASs to fully understand the intrinsic mechanisms of catalytic reactions and to improve existing catalysts or design improved catalysts [2-4]. However, accurate identification of CASs remains challenging, especially for supported metal particle/cluster catalysts, owing to the so-called intrinsic "metal effects" such as electronic quantum size effect [5-7] and structure sensitivity [8, 9]. In fact, late transition metal particles with small sizes have higher lying d-band/orbital centers (which play a crucial role in determining the catalytic activity [10]) compared with those having larger sizes [9]. Metal atoms at different local environments (e.g., at steps, edges, kinks, and corners) often have different electronic states [9], thus resulting in a different catalytic activity. In particular, the identification of the CASs of supported metal particle/cluster catalysts has led to much controversy, even under similar experimental conditions such as in the same reactions and with the same catalysts [11, 12]. Thus, a complete elimination of the intrinsic metal effects is extremely difficult and even impossible for supported metal particle/cluster catalysts.

The above-mentioned issue can be solved if a uniform single-site heterogeneous catalyst with surface single metal atoms as active sites can be developed. However, it is tremendously difficult to synthesize such an atomically dispersed metal catalyst (ADMC) as metal atoms are often prone to sintering by the Ostwald ripening process during catalyst preparation and catalytic reactions [13, 14]. Despite these difficulties, it is still feasible to synthesize ADMCs, depending on the nature of the supports. Kwak et al. [15] reported that platinum atoms bind to unsaturated coordinate Al³⁺ sites on the surface of the support by a binding-driven method to achieve an ADMC. Another ADMC involving platinum atoms that are anchored at electronic defects of iron oxide surfaces has also been synthesized [16]. However, a distinct drawback of these catalysts is the aggregation of metal atoms, especially when the metal loading is over 1 wt% [15, 16], which leads to a loss of activity or selectivity.

A so-called top-down synthesis strategy is one of the most efficient methods to synthesize materials with nanometer sizes [17], even with an atomically dispersed state [18, 19]. In principle, the top-down method is feasible for the fabrication of an ADMC by controlling metal-support interactions such as the adhesion energy of metal atoms to the surfaces of the support [20, 21]. Experimentally, we have used this method to synthesize an ADMC from Ag nanoparticles (NPs) through a simple thermal diffusion process [18, 19]. In this article, we focused on the top-down synthesis process of the transformation from Ag NPs to single Ag sites with the assistance of the surface of the support. The top-down process is characterized by using in situ synchrotron X-ray diffraction (SXRD) and ex situ X-ray absorption spectroscopy. The geometric and electronic structures of the metal atoms during this process are thoroughly studied to correlate with the catalytic performance in the complete oxidation of formaldehyde (HCHO).

The synthesis of hollandite manganese oxide (HMO) NPs has previously been described [22, 23]. Briefly, two solutions were first prepared: one aqueous solution (170 mL) was prepared by dissolving KMnO₄ (11.600 g), Na₂MoO₄ (1.228 g), Na₃VO₄·12H₂O (3.840 g), and TiOSO₄·H₂SO₄·8H₂O (1.536 g) in water; the other aqueous solution (80 mL) was prepared by dissolving MnSO₄·H₂O (17.600 g) and FeSO₄·7H₂O (2.432 g) in water. The former solution was added drop-wise to the latter to obtain a brown slurry, which was subsequently refluxed at 100 ℃ for 24 h. The resulting solid was washed with de-ionized water, filtered, dried at 110 ℃ for 12 h, and calcined at 400 ℃ for 4 h.

A surface-anchored single-atom Ag catalyst (Ag₁/HMO) was prepared by the thermal surface-mediated diffusion method of the corresponding supported Ag NP catalyst (Ag_NP/HMO). In brief, an aqueous ammonia solution (25 wt%) was slowly added to an AgNO₃ solution (30 mL, containing 0.315 g of AgNO₃) under stirring until the solution became transparent. Both the transparent solution and a H₂O₂ solution (30 wt%, 90 mL) were simultaneously added to another aqueous suspension (80 mL) containing HMO (2.000 g) under stirring at 0 ℃ for 0.5 h. The final suspension was filtered, washed with de-ionized water, and dried at 80 ℃ for 12 h to obtain Ag_NP/HMO. Ag_NP/HMO was annealed at 350 or 400 ℃ (as specified in the text) for 4 h to obtain Ag₁/HMO. The referential samples (Ag/HMO-ref and HMO-ref) were synthesized according to a previous report [18].

The synchrotron X-ray diffraction (SXRD) patterns were obtained at the BL14B Beamline of the Shanghai Synchrotron Radiation Facility (SSRF) at a wavelength of 0.12398 nm. The beam was monochromatized using Si (111), and a Rh/Si mirror was used for the beam focusing to a size of approximately 0.5 mm × 0.5 mm.

High-resolution transmission electron microscopy (HRTEM) images were obtained with a JEOL JEM-2100F field-emission transmission electron microscope operating at an accelerating voltage of 200 kV and equipped with an ultra-high-resolution polepiece that provides a point-resolution better than 0.19 nm. Fine powders of the materials were dispersed in ethanol, sonicated, sprayed on a carbon coated copper grid, and then allowed to air-dry for imaging.

XAS covers the X-ray absorption near-edge structure (XANES) spectra and extended X-ray absorption fine structure (EXAFS) spectra, which were recorded at the Ag K-edge using the BL14W Beamline of the SSRF with an electron beam energy of 3.5 GeV and a ring current of 200-300 mA. Data were collected with a fixed exit monochromator using two flat Si(311) crystals. The XANES spectra were acquired at an energy step of 0.5 eV. The EXAFS spectra were collected in the transmission mode using ion chambers filled with N₂. The raw data were analyzed using the IFEFFIT 1.2.11 software package.

H₂ temperature-programmed reduction (H₂-TPR) profiles were recorded using a ChemiSorb 2720 adsorption instrument. The sample (~20 mg) was reduced in a stream of 10% H₂-90%Ar (50 mL/min) at a ramping rate of 10 ℃/min.

Prior to adsorption measurements, ~15 mg of the sample was pretreated in helium (30 mL/min) at 120 ℃ for 30 min. The sample was cooled to room temperature and CO was then pulsed from a calibrated on-line sampling loop (18.60 μL) over the catalyst until the CO adsorption reached a saturation state. The amount of adsorbed CO on the surfaces of the catalyst was calculated by a dynamic method on an AutoChem Ⅱ 2920 adsorption instrument.

The oxidation of HCHO was performed in a fixed-bed quartz reactor (i.d. = 8 mm) under atmospheric pressure. A catalyst (25 mg, 40-60 mesh) was loaded for each run. Gaseous HCHO was generated by passing a N₂ gas flow over paraformaldehyde in an incubator kept at 45 ℃, and then mixed with an O₂ flow, leading to a feed gas containing 150 ppm HCHO and 10.0 vol% O₂, balanced by N₂. The total flow rate was 100 mL/min. The effluents from the reactor were analyzed with an on-line Agilent 7890A gas chromatograph equipped with TCD and FID detectors. To compare the pre-exponential factors (Γ), Arrhenius plots were obtained according to the catalytic activity of samples at 35-60℃ with the conversion of HCHO less than 20%. The data were recorded up to the steady state for each run.

All calculations in the current work were performed using the density functional theory (DFT) method implemented in the plane-wave basis Vienna Ab-initio Simulation Package (VASP) code [24]. The projector augmented wave potentials [25] and the approximation method of generalized gradient approximation [26] were used throughout the investigation. Sampling the Brillouin zone using a (4 × 4 × 2) Monkhorst-Pack special k-point mesh yielded well-converged results. The energy cut off for the expansion of wave functions and potentials was 400.0 eV. The structures were relaxed using a conjugate gradient method with the convergence criteria for energy and forces set to 1 × 10^-5 eV per supercell and 0.1 eV/nm. A supercell of Ag₂Mn₈O₁₆ or Mn₈O₁₆ units was constructed to simulate the terminals of Ag₁/HMO or HMO. A vacuum region of 1.5 nm was used along the z direction to avoid any spurious interactions among the periodical images.

Suitable metal-atom anchoring sites of supports are important for fabricating ADMCs by a top-down synthesis. Typically, HMO surfaces consist of terminated oxygen atom arrangements with desired sizes, which can serve as ideal sites for anchoring individual metal atoms [18, 19]. HMO often grows along the [001] direction to form rods with a high aspect ratio owing to a lower surface energy of the (100) side-facets compared with the (001) top-facets (Fig. 1(a)) [27, 28]. We modified the synthetic protocol to synthesize HMO NPs to increase the number of surface metal-atom anchoring sites by decreasing the aspect ratio of the HMO rods [22]. A typical HMO NP with a nanometer size of ~4 nm is visualized in an HRTEM image (Fig. 1(b), (c)), and these HMO NPs have a higher specific surface area (157 m²/g) than HMO nanorods (46 m²/g) [18], thus increasing the number of surface anchoring sites.

A common method to achieve ADMCs is based on the fact that anchoring sites directly capture individual metal atoms from metal monomer compound precursors [15, 16]. In this work, HMO NPs were used as a support to synthesize ADMCs. To synthesize Ag₁/HMO, Ag NPs with a truncated octahedron shape were initially supported on the HMO surfaces to get Ag_NP/HMO (Fig. 1(d)). The Ag NPs with ~2 nm in height and ~4 nm in width were disintegrated into single Ag atoms which were then anchored on HMO surfaces by a thermal process (Fig. 1(e)) [18], indicating the presence of a large adhesion energy of Ag atoms to the HMO surfaces [20]. This is a typical top-down process, in which large metal particles/clusters tend to form single metal sites, unlike normal Ostwald ripening processes [14, 29].

In situ SXRD was used to investigate the transformation of supported Ag NPs to single Ag sites with the assistance of HMO NP surfaces. With an increase of the temperature, the diffraction intensity of Ag NPs became weaker and ultimately disappeared, indicating that Ag NPs gradually shrank and finally reached a highly dispersed state (Fig. 2(a)). For better visualization, a curve of the Ag(111) SXRD intensity as a function of temperature and its first derivative are shown in Fig. 2(b). When the temperature was lower than 150 ℃, the Ag(111) intensity remained unchanged, indicating that the size of Ag NPs did not change below 150 ℃. When the temperature was higher than 150 ℃, the intensity became weaker, and the diffusing rate became faster with the increase of the temperature, reaching a maximum at ~200 ℃. When the temperature was higher than ~260 ℃, the Ag(111) reflection vanished, and the diffusing rate was stable, indicating that Ag finally reached a highly dispersed state.

To study the detailed top-down process, Ag_NP/HMO was annealed at different temperatures with a temperature-programmed procedure similar to that of the in situ SXRD experiment to give a series of Ag-T samples (T = 150, 200, or 350 ℃, and here, T represents annealing temperature), and then ex situ EXAFS spectra were recorded to investigate the local environments of the Ag atoms of the samples. Fig. 3 shows Fourier transform (FT) amplitudes of the χ(R) k²-weighted EXAFS data at the Ag K-edge of the samples, and the structural parameters obtained by fitting the spectra with theoretical models are summarized in Table 1 [30]. The near neighbor Ag-Ag coordination shell appears in the FT spectrum of the Ag-150 sample with an average distance of ~0.287 nm and a coordination number (CN) of ~8, indicating the presence of metal Ag NPs, consistent with the in situ SXRD patterns (Fig. 2(a)). The FT amplitude of the EXAFS spectrum of the Ag-200 sample at 0.28-0.30 nm, which can be ascribed to the scattering between Ag atoms, is very weak, indicative of a very low average CN of Ag atoms at this shell. The results of the fitting of the EXAFS spectra, giving an Ag-Ag CN of 2.6, confirm this hypothesis (Table 1). In the FT EXAFS spectrum of the Ag-350 sample, the absence of the peak arising from Ag-Ag bonds indicates the formation of single Ag sites on the HMO surfaces [18], unlike the reported Ag₁/HMO-ref with tunnel Ag atoms (the FT XAFS spectrum of the Ag₁/HMO-ref is also shown in Fig. 3 for comparison) [18, 19]. As shown in the FT spectrum of Ag₁/HMO (Ag-350), the first shell can be attributed to two kinds of Ag-O bonds with an average bond length of ~0.233 nm and CN = 4 (Table 1), indicating that single Ag sites are on the four-fold oxygen-terminated sites of the HMO surfaces [31]. Therefore, the combination of the in situ SXRD patterns and the ex situ EXAFS spectra evidence that Ag NPs have transformed into single Ag sites through the thermal diffusion process.

According to its first derivative of the in situ SXRD patterns above, we estimated the diffusion barrier (E_d) for the Ag atoms detached from Ag NPs using E_d = ln(ν₀/γ)k_BT, where ν₀, k_B, T and γ are the usual pre-factor (often taking ν₀ = 6 × 10¹²) [32], the Boltzmann constant, the temperature in Kelvin (K) and the rate relevant to a deposition rate, respectively. According to the in situ XRD (Fig. 2) and TEM data (Fig. 1(d)) of Ag_NP/HMO, the average deposition rate of Ag NPs was ~0.003 ML/s. Therefore, we chose a typical value γ = 1 s^-1 [32]. At T = 420 K (~150 ℃), the calculated E_d was 1.07 eV for Ag atoms detached from Ag NPs. To investigate the stability of the single atom embedded on a HMO(100) surface, we carried out DFT calculations. The semi-tunnels of the HMO(100) surface provide perfect adsorption sites for single Ag atoms, that is, Ag atoms are embedded into the semi-tunnels surrounded by four oxygen atoms (the average Ag-O distance is 0.247 nm). The calculated adsorption energy (E_ad) of Ag atoms on the four-fold oxygen-terminated sites of the HMO(100) surface is -1.51 eV with respect to bulk Ag. Here, E_ad = E_{s_Ag} -E_s -E_Ag, where E_{s_Ag} is the total energy of a single Ag atom adsorbed on the HMO(100) surface, E_s is the total energy of the bare HMO surface, and E_Ag is the total energy of the Ag bulk for each Ag atom. It is clear that a single Ag atom is energetically stable when embedded in a semi-tunnel of the HMO(100) surface. We also calculated the diffusion barrier for an embedded silver atom migrating from site to site in the semi-tunnels. As shown in Fig. 4, the barrier of such diffusion inside the semi-tunnels is quite small, that is, ~0.18 eV. This indicates the facile dispersion of embedded single Ag atoms.

The stability of single metal atoms or metal NPs with relatively high melting points of the bulk metals, such as Au [33], Pt [15, 16, 34, 35], and Rh [36], has been well documented. Owing to the relatively low melting point of bulk Ag, highly dispersed Ag atoms should be more mobile and prone to aggregation or coalescence. In fact, Ag particles with a sub-10 nm size are not thermally stable and often show liquid-like behavior [37]. Even though Al₂O₃ could provide anchoring sites for stabilizing single Pt atoms [15, 34], Ag clusters, such as Ag trimers supported on Al₂O₃, are prone to sintering to form big particles, even at temperatures lower than 120 ℃ [38]. Theoretically, single Ag atoms should be energetically more favorable to aggregate into large particles than Ag NPs/clusters. However, according to the HRTEM image (Fig. 1), the in situ SXRD patterns (Fig. 2) and the ex situ EXAFS spectra (Fig. 3), single Ag atoms anchored on the surface oxygen-terminated sites are extremely stable even at > 350 ℃ under oxidizing conditions (Figs. 1-3).

Ag K-edge XANES was used to study the electronic states of Ag atoms of the Ag-T samples during the top-down process together with two referential samples, Ag foil and Ag₂O. Fig. 5(a) shows the Ag K-edge XANES spectra of the samples. The edge positions for all the Ag-T samples appear between those of the Ag foil and Ag₂O, indicating that the electronic states of Ag atoms of the Ag-T samples are at a mediate state between the Ag foil and Ag₂O.

To investigate the subtle change of the electronic states of Ag atoms during the top-down process, we focused on the 1s → 5p electron transition in the XANES spectra (the inset of Fig. 5(a)). The white-lines of the Ag-T samples were more intense than that of the Ag foil, indicating a decrease in the electron density of Ag 5p orbitals. The electron density of Ag 5p orbitals might originate from the hybridization of the Ag frontier d-sp orbitals, especially with O 2p orbitals [19], although the Ag 5p orbitals are unoccupied at the ground state of Ag atoms with a valence electron configuration of 4d¹⁰5s¹5p⁰. Similarly, the electron density of Ag 5p orbitals becomes lower during the top-down process judging from the more and more intense 1s → 5p absorption edges from Ag_NP/HMO (Ag-150) to Ag₁/HMO (Ag-350) during the top-down process.

Fig. 5(b) shows the differential X-ray absorption near-edge structure (ΔXANES) spectra, which are highly sensitive and highly accurate for exploring the unoccupied states of Ag 4d orbitals with respect to the spectrum of Ag₂O [18, 39]. A weak intensity in the ΔXANES spectrum of the Ag-T samples can be ascribed to the 1s → 4d electronic transition owing to the small d-p hybridization of the Ag frontier orbitals [39-41]. The ΔXANES spectra of the Ag-T samples show a pre-edge energy absorption shoulder in comparison with the XANES spectrum of Ag₂O with an almost filled d orbital configuration (d¹⁰) of Ag⁺ [42]. According to the absorption population at the pre-edge regime, the electron density of Ag 4d orbitals of each Ag-T sample is slightly lower than that of the Ag foil. The results are consistent with the fact that the smaller the size of Ag particles, the lower the electron density of the Ag 4d orbitals, and the stronger the Ag-Ag bonds [43]. For the Ag-T samples, the electron density of the Ag 4d orbitals gradually increases during the top-down process. Therefore, the results elucidate that when Ag NPs transform into individual surface Ag atoms anchored on the oxygen-terminated sites, Ag 5p orbitals are relaxed and lose their partial charge because all Ag atoms are exposed on the surfaces, but Ag 4d orbitals increase the electron density presumably because of the Ag-O interactions.

To investigate the change of the electron density of the Ag frontier orbitals, DFT calculations were conducted and the configurations of the surface Ag atoms and the electronic states are displayed in Fig. 6. The density of states (DOS) of single Ag atoms are remarkably different from those of bulk Ag, as explained by the interaction of Ag atoms with surface lattice oxygen atoms. According to the atom configuration of single Ag atoms, Ag orbital splitting occurs (Fig. 6), and DFT predicts the highest-lying d_z², the lowest-lying d_xy orbital, and the d_x-y²² and the doubly degenerate d_xz and d_yz orbitals. All Ag 4d orbitals are shifted up after interacting with the surface oxygen atoms, and the Ag 4d electron density obviously increases owing largely to the partial negative charge transferring from the O 2p_z orbital into the highest Ag 4d_z² orbital with partial depletion.

The above results provide strong evidence that the driving force of the top-down process is the interaction between Ag atoms and the surface oxygen atoms of the HMO. In detail, first, the surface atoms of Ag NPs spill out the 5p electrons (with the possibility of also spilling out part of the 5s electrons), which is favorable for weakening the Pauli repulsion with surface lattice oxygen atoms of HMO, allowing a strong Ag-O interaction to occur. Second, Ag atoms at the perimeter of the Ag-HMO interface strongly interact with surface oxygen atoms, allowing the electron density of the Ag 4d orbitals to increase. Because of the almost full Ag 4d shells, the increase in electrons in the Ag 4d orbitals preferentially occupy the anti-bonding orbitals, which weakens Ag-Ag bonds, leading to the low energy barrier (1.07 eV) for Ag atoms being detached from Ag NPs. Finally, the detached Ag atoms can diffuse on the HMO surface with relative ease owing to the low diffusion barrier (0.18 eV), as predicted by the DFT calculations above. Therefore, the driving force of the top-down process can be attributed to the electronic perturbation of Ag frontier orbitals induced by Ag-O interactions at the perimeter of the Ag-HMO interface [2, 3, 19].

The top-down synthesis has two important functions for catalysis research: (ⅰ) to increase the number of CASs through moving bulk metal atoms onto surfaces; (ⅱ) to provide model single-site catalysts with uniform and ideal CASs, which are favorable for investigating reaction mechanisms.

We chose the HCHO oxidation reaction as an example to interpret the functions of the top-down synthesis because the number of exposed Ag atoms directly determines the reaction rate in low-temperature HCHO oxidation, and HCHO is a typical indoor and outdoor air pollutant [44] that contributes greatly to atmospheric particulate emissions [45]. As shown in Fig. 7(a), the HCHO reaction rates at 80 ℃ increased from 2.1 × 10^-7 mol/(g·s) to 5.5 × 10^-7 mol/(g·s), as the annealing temperature of Ag_NP/HMO increased from 150 to 300 ℃, respectively, and leveled off as the annealing temperature was further increased to 400 ℃. Under the same conditions, HMO is inactive in HCHO oxidation [31].

To elucidate the results above, the apparent activation energy (E_a) and the Γ of HCHO oxidation on the Ag-T catalysts were calculated (Fig. 7(b)). The E_a of HCHO oxidation over Ag-150 (0.53 eV) was almost the same as that (0.54 eV) of Ag-350, and also similar to the reported E_a values in HCHO oxidation over other Ag-based catalysts [46]. However, the calculated Γ for Ag₁/HMO (Ag-350) was 1.4 × 10²⁶ molecules/(g·s), which is much higher than that of Ag-150 (4.5 × 10²⁵ molecules/(g·s)) and Ag-200 (9.2 × 10²⁵ molecules/(g·s)). As the TOF of the Ag-based catalysts is structure-independent and size-independent [31], Γ should be primarily determined by the number of CASs [47]. Thus, the difference in the reaction rates mainly originates from the number of CASs of Ag₁/HMO, where the number is much higher for Ag-350 than for Ag-150 and Ag-200. Therefore, the top-down strategy can increase the number of CASs, thus improving the catalytic performance.

Catalysts with single-site metal centers are beneficial for investigating reaction mechanisms because of the absence of the electronic quantum size effect [5-7] and structure sensitivity [8, 9]. Reaction mechanisms, such as the Langmuir-Hinshelwood (L-H) and Eley-Rideal (E-R) models, have often been used to describe catalytic reactions, and, interestingly, the L-H mechanism is more pronounced at low reaction temperatures for supported metal NP catalysts, but the reaction kinetics demonstrate that the L-H model cannot describe the predominant reaction mechanism for HCHO oxidation on Ag-based catalysts [46]. For Ag₁/HMO, the issue becomes simple because HCHO oxidation on single-site Ag centers cannot occur by a conventional L-H mechanism that requires two adsorption sites.

The E-R mechanism is also not suitable to describe HCHO oxidation on supported metal catalysts through a reaction of adsorbed HCHO with gaseous O₂ owing to an extremely high energy barrier for dissociation of O₂. An alternative Mars van Krevelen model is often used to describe high-temperature oxidation reactions over oxide catalysts [48], that is, adsorbed HCHO reacts with lattice oxygen to generate oxygen defects, and O₂ then dissociatively adsorbs on the active sites to replenish the oxygen vacant sites to complete the reaction cycle. With this model, there are two critical factors to determine the catalytic activity: oxidation of adsorbed formate (the stable intermediate in HCHO oxidation) and dissociation of O₂ [18, 19, 46, 49], that is, lattice oxygen and molecular oxygen activation, because the activity of lattice oxygen directly governs the oxidation ability of formate.

The activity of the lattice oxygen of the catalysts was investigated by using H₂-TPR, and the results are shown in Fig. 8(a). Obviously, the lattice oxygen of Ag₁/HMO shows a surprisingly high activity at low temperature, which is favorable for the oxidation of adsorbed formate species [18, 49]. For O₂ activation, Ag 4d orbitals with a partiallly unoccupied state have an open-shell electronic structure [38], which is favorable for improving the catalytic activity, because O₂ activation is regarded as an important rate-limiting step [49, 50]. The open-shell nature of the Ag electronic structure, caused by the d-p hybridization and the upshift of the d orbitals induced by Ag-O interactions, have been evidenced by the XANES spectra (Fig. 5), and is further confirmed by the CO pulse experiment at room temperature (Fig. 8(b)). Clearly, CO molecules chemisorb on Ag active sites rather than on the HMO support, consistent with the donating and back-donation bonding between CO and metal atoms [51-53]. With the assistance of metal atoms, the Mars van Krevelen mechanism is also feasible at low reaction temperatures for oxidation reaction. Therefore, to some extent, the top-down synthesis provides single-site catalysts as an ideal model, which is favorable for investigating complex reaction mechanisms.

In summary, the synthesis of ADMCs from Ag NPs with the assistance of an HMO support was substantiated through the thermal diffusion process. In situ SXRD, ex situ EXAFTS spectroscopy, and TEM data were used to study the formation of single Ag sites from Ag crystals during the top-down process. Ex situ X-ray absorption spectra were used to study the changes of the geometric and electronic structures of the single metal atoms during this process. The experimental results in conjunction with DFT calculations demonstrate that the driving force of the top-down process is attributed to the electronic perturbation of Ag frontier orbitals induced by Ag-O interactions at the perimeter of the Ag-HMO interface. Two possible functions of the top-down process are proposed by taking the catalytic oxidation of formaldehyde as an example. Therefore, the top-down strategy could assist the design of single-site catalysts that are favorable for studying complex reaction mechanisms of catalytic reactions.

Acknowledgments

We acknowledge the Shanghai Synchrotron Radiation Facility for the XAS and SXRD measurements.