NiAl₂O₄ was prepared via a co-precipitation method with the Ni/Al mole ratio at 1:2 [38]. In a typical procedure, 40 mmol nickel nitrate hexahydrate (Ni(NO₃)₂·6H₂O, Adamas-beta) and 80 mmol aluminum nitrate nonahydrate (Al(NO₃)₃·9H₂O, General Reagent) were dissolved in 200 mL ultra-pure water (18.2 mΩ) under vigorous stirring. Ammonium hydroxide (NH₃·H₂O, General Reagent) was used as the precipitant until the pH value reached 9. After stirring for another one hour, the suspension was settled overnight for aging. Then the powder was separated by vacuum filtering and washed with deionized water until the pH value of the filtrate was around 7, followed by the overnight drying in a 100°C oven. Finally, the powder was calcined at 800 °C under N₂ atmosphere for 8 h, 5°C·min⁻¹and used as support. γ-Al₂O₃ support was prepared via the calcination of commercial boehmite, which was purchased from Aluminum Corporation of China, at 500°C for 4 h with a heating rate of 5°C·min⁻¹.

Ru/NiAl₂O₄ (denoted as Ru/NAO-x, where x represents the reduction temperature) and Ru/γ-Al₂O₃ (denoted as Ru/AO) catalysts were synthesized by incipient wetness impregnation method using 10wt% RuCl₃ aqueous solution as the precursor. Before each activity test, the catalysts were all activated at different temperatures (100°C, 200°C, 300°C and 400°C) in 10% H₂/Ar for 3 h.

X-ray power diffraction (XRD) patterns were recorded on a D8 Focus Diffractometer (CuKα1 radiation, k = 1.5406 Å) at 40 kV and 40 mA, with the scattering angles between 10-80 ^o. High angle annular dark field (HAADF)-STEM images were collected on a FEI Talos F200X G2 electron microscopy at 200 kV and recorded with a convergence semi angle of 11 mrad, and inner- and outer collection angles of 59 and 200 mrad, respectively. X-ray photoelectron spectra (XPS) were collected on a Thermofisher Scientific Escalab 250 Xi equipment with monochromatic Al Kα radiation, and all the results were calibrated using C 1 s at 284.8 eV. XAS measurements were conducted at the Canadian Light Source for the Ru L₂-edge on the SXRMB/CLS beamline and Ru foil was used for energy calibration. X-ray absorption near-edge structure (XANES) data analysis was performed using Athena software.

CO chemisorption tests were operated on Huasi DAS-7200 automatic chemisorption instrument. Typically, 0.1 g catalyst was injected to the quartz sample tube. Before each run, the catalyst sample was in-situ reduced under 10% H₂/Ar atmosphere (30 mL·min⁻¹) for 1 h, and cooled down to ambient temperature with continuous He sweeping (30 mL·min⁻¹). The CO chemisorption results were calculated based on the peak areas of CO pulse injection (200 μL per pulse, filled with 50% CO/He), which were monitored by a TCD detector (45 °C, 80 mA).

The temperature-programmed surface reaction (TPSR) and in-situ diffuse reflectance infrared Fourier transform (DRIFT) spectroscopy were both collected in a Nicolet iS50 FTIR (Thermofisher) equipped with a reaction chamber (Harreck) that was connected to a Mass Spectroscopy instrument. Before each test, the catalyst samples (50-70 mg) were all in-situ reduced under 5% H₂/Ar for 1 h, swept in pure Ar for another 1 h, and finally cooled down to 30 °C. Then the substrate was introduced into the chamber by Ar bubbles (60 mL·min⁻¹ as flow rate) with a quartz bubbler maintained at 150 °C in an oil bath. The temperature of the tubes, which was connected the bubbler to the reaction chamber, was constantly maintained at 90 °C via a heater. Before the collection of DRIFT spectra and fragment trends in MS, the physisorbed molecules were removed by pure Ar with a flow rate of 30 mL·min⁻¹. For TPSR runs, the reaction chamber was heated from 30 °C to 440 °C, with a rate of 5 oC·min⁻¹ and a 5%H₂/Ar flow rate of 10 mL·min⁻¹.

Raman spectra were collected on a HORIBA LabRAM HR system exposed under the excitation line of 514 nm. The tests of the catalyst were divided into two parts, i.e., in-situ reduction from ambient conditions to the desired temperature, and surface HDO reaction using anisole as the probe molecule. First, the catalyst sample was dehydrated in testing cell under Ar flow at 100 °C for 10 min, then switched to 5%H₂/Ar (10 mL·min⁻¹), and the reduction started from 100 °C to 400 °C and kept for another 30 min. After reduction, the reaction cell was cooled down to room temperature again under the protection of Ar flow. Anisole was introduced into the cell at room temperature and swept by Ar (10 mL·min⁻¹) for 20 min to remove the physiosorbed molecules. The Raman spectra were collected from 100 °C to 250 °C with a step of 50 °C. When approaching 200 °C, the reaction atmosphere was switched from Ar to H₂/Ar (10 mL·min⁻¹) for demethoxylation reaction.

The catalyst performance was evaluated in a 50-mL stainless batch reactor with a magnetic stirrer. For each run, the reactor was loaded according to the expected dosage of the substrate, catalyst and solvent. After that, the reactor was sealed with graphite ring, purged with pure N₂ for 3 times to expel the residue air, followed by the injection of 0.1 MPa pure H₂, and kept stabilized by pure N₂ at 2 MPa as the initial pressure. After the end of each reaction, the batch reactor was quickly removed from the heater and quenched in ice water. The liquid product was carefully extracted using 5 g ethyl-acetate from the aqueous reaction residue. The qualitative analysis of the extraction was carried out in a GC-MS system equipped with a HP-5 column, Agilent 7890A-5975C, and quantitatively analyzed by Agilent 7890B equipped with a HP-5 column and an FID detector, using tridecane as the internal standard.

Kinetic measurements were performed using 4-methylanisole as the reactant. Before testing, the diffusion and thermodynamic limitations were excluded by changing the stirring rate and by altering the catalyst-to-reactant ratio, and all the kinetic results were collected by keeping the conversion below 10%. The reaction rate was calculated according to the generation rate of deoxygenated product, toluene. The kinetic models were established according to the Langmuir-Hinshelwood model, which was illustrated in supporting information in detail.

XRD patterns of the catalyst samples are presented in Fig. 1a, whereas no obvious diffraction peaks corresponding to Ru species can be observed over all these three catalysts, indicating the well-dispersion of Ru nanoparticles on NiAl₂O₄ and γ-Al₂O₃ supports. The diffraction peaks at 19.1°, 31.4 ^o, 37.0 ^o, 44.9 ^o, 59.7 ^o, and 65.5 ^o of Ru/NAO-200 (400) catalysts are assigned to (111), (220), (311), (400), (511), and (440) facets of NiAl₂O₄ spinel (JCPDS #10-0339). In contrast, only three diffraction peaks were observed for 2%Ru/AO-400 catalyst, which are at 37.4 ^o, 42.8 ^o, and 67.3 ^o, relating to (311), (321), and (522) facets of γ-Al₂O₃ (JCPDS #04-0880).

Ru L₂-edge XANES spectra for three reduced catalysts as well as Ru foil and RuO₂ are present in Fig. 1b. In general, a large ‘white line’ refers to the more oxidized Ru, whereas a small one indicates the reduced-state of Ru species [36]. The reduction of the whiteline intensity indicates that the chemical state of Ru in Ru/NAO-400 catalysts was more electron-enriched than those in Ru/NAO-200 and Ru/AO [36,39]. In Fig. 1c, two Ru species were obtained in the XPS spectra of Ru 3d_5/2, in which the fitting peaks at around 280.2 eV and 281.0 eV were assigned to the metallic Ru⁰ and RuO_x species, respectively [40,41,42]. According to the XPS analysis and fitting results in Fig. 1c and Table 1, we calculated the Ru⁰ proportions in three catalysts, which were 80.9%, 54.5% and 38.3% in Ru/NAO-400, Ru/NAO-200 and Ru/AO-400, respectively.

HAADF-STEM was carried out to count the Ru particle size distributions in three catalysts. As is illustrated in Fig. 1d-f, the average sizes of Ru particles are 1.2 nm, 1.2 nm and 1.8 nm in Ru/NAO-400, Ru/NAO-200 and Ru/AO-400, respectively. Figure S1 displays the EDS mapping images of Ru/NAO-400 sample, in which it can be seen Ru was well dispersed. Besides, these three catalysts also exhibited similar dispersion of Ru sites according to the CO pulse titration results (Table 1).

Table 1 summarized some other physical and chemical properties of these three catalysts and their corresponding catalytic activities in demethoxylation of 4-methylanisole. According to the demethoxylation rates over three different catalysts in Table 1, it can be preliminarily concluded that the Ru⁰/Ru^δ+ ratio was crucial to the demethoxylation performance of the catalyst.

Figure 2 shows the in-situ Raman spectra of anisole demethoxylation over Ru/NAO-400 and NiAl₂O₄ support, which were performed to investigate the real sites responsible for the activation of C-O bonds during reaction. Under ambient conditions, we first focused on two intense peaks at 314 cm⁻¹ and 475 cm⁻¹ (Figure S2a) over Ru/NAO-400 corresponding to the Ni-O species and Ru-O species, respectively [43,44,45,46,47]. These peaks can also be observed in the in-situ Raman spectra of NiAl₂O₄ spinel, NiO/Al₂O₃ hybrid and RuO₂ (shown in Figure S3). After dehydration at 100 °C for 10 min, all intensities reduced significantly. Further reduction at 400 °C in 5%H₂/Ar atmosphere led to the disappearance of the signal at 475 cm⁻¹, together with the appearance of a new peak at 356 cm⁻¹. Compared with the in-situ Raman spectrum of RuO₂ (Figure S3d), we can propose that the signal at 356 cm⁻¹ can be ascribed to the partially reduced RuO_x species. Interestingly, we also found the pair peaks, related to the RuO_x species, at 375 cm⁻¹ and 471 cm⁻¹ over Ru/AO-400 (Figure S2b), in which the peak shifts could be attributed to the different electronic states of Ru over different supports [48]. In addition, a new band appeared at around 553 cm⁻¹ over Ru/NAO-400, which could be assigned to the NiO phase with nickel incorporated into the subsurface of the alumina support [45,49], emerged and kept its existence even after reduction.

In the Raman experiments, anisole was injected into the reaction cell at ambient temperature, and was further swept in Ar for 10 min. Before the adsorption of anisole, both samples were reduced and then cooled down to room temperature, whereas the procedures were in-situ operated in the testing cell. When the temperature raised to 100 °C, the peak at 470 cm⁻¹ recovered simultaneously with the disappearance of the band at 356 cm⁻¹, and kept steady even when the temperature reached 150 °C (Fig. 2a). This result indicates that the oxygen in anisole was adsorbed by RuO_x species, and the interaction between Ru and adsorbed oxygen species contributed to the recovery of the Ru-O bending at 470 cm⁻¹. Right after the introduction of hydrogen at 200 °C, the intensity of the newly formed Ru-O bond went down again with the increase of the temperature. Simultaneously, the reappearance of the peak at 356 cm⁻¹ after reaction further indicates that the demethoxylation of anisole underwent the redox pathway over active RuO_x species in Ru/NAO-400. In contrast, only two bands at 320 cm⁻¹ and 555 cm⁻¹ without obvious changes were observed during the reaction conditions over pure NiAl₂O₄ support was shown in Fig. 2b. According to the in-situ Raman spectra, we can conclude that RuO_x species worked as the active sites that interacted with the O atom in methoxy during demethoxylation.

TPSR-MS of anisole was further carried out to investigate the ad/de-sorption behavior of the reactant, intermediate species and products, which could help unravel the possible reaction pathway. The starting points of demethoxylation over three catalysts could also be obtained from the uplifting temperatures of the generation of methanol and benzene. To be noted, the starting points of demethoxylation hereby were all determined by the uplift trend of methanol (m/z = 31), due to the fragment overlap between anisole and benzene at m/z = 78 (Figure S5).

Figure 3a and b show that the generation temperatures of methanol were quite close over two NiAl₂O₄-based catalysts, which were 103 °C and 107 °C over Ru/NAO-400 and Ru/NAO-200, respectively, suggesting they might possess similar activation ability for C-O bond cleavage. This speculation could also be confirmed via the in-situ DRIFT spectra of anisole adsorption. The bands corresponded to methoxy C-O bond at 1240 cm⁻¹ and 1238 cm⁻¹ in Fig. 3d were both red-shifted compared with gaseous anisole, indicating the existence of strong interaction between the substrate and the catalyst surface. As a contrast in Fig. 3c, the demethoxylation over Ru/AO-400 initiated from 115 oC, whereas only a slight redshift of 5 cm⁻¹ in DRIFT was observed, indicating that the activation of C-O bond was less effective over Ru/AO-400. More interestingly, the adsorption of anisole on Ru/AO-400 was weak since the desorption peak of anisole was at 98 °C, right before the starting point of demethoxylation. Besides Ru/AO-400, the desorption peaks of anisole also vary between the other two catalysts, which were 210 oC and 168 oC over Ru/NAO-400 and Ru/NAO-200, respectively, which could be ascribed to the fact that the chemical compositions of Ru, namely the proportion of Ru⁰ species, play an important role in the adsorption of the aromatic ring in anisole, which was also mentioned by the previous work [50]. However, a higher concentration of Ru⁰ species can enable the stronger adsorption of the reactant, but may also make it more difficult for the desorption of produced aromatics, which will hinder the recovery of Ru⁰ species. These influences can be seen from the second desorption peaks of fragment at m/z = 78 at, 327 °C, 267 °C and 232 °C over Ru/NAO-400、Ru/NAO-200和Ru/AO-400, respectively. Hence, combined with the Raman results, an appropriate Ru⁰/Ru^δ+ ratio played vital role in demethoxylation, and Ru/NAO-200 catalyst likely possessed a good ratio of Ru⁰/Ru^δ+. This analysis is in good agreement with catalytic performance summarized in Fig. 2. In addition, we also observed the formation of H₂ (m/z = 2) and formaldehyde (m/z = 30) when we raised the temperature, suggesting the decomposition of generated methanol, or methoxy, that later initiated the WGS according to the uplift of CO₂ (m/z = 44) [34,38,51]. Interestingly, the generation temperature of CO₂ was at 300 °C over Ru/NAO-200 in Fig. 3b, higher than 154°C and 118°C over Ru/NAO-400 and Ru/AO-400, respectively, which could be ascribed to the low WGS rate over Ru/NAO-200. The difference between Ru/AO-400 and Ru/NAO-400 was also thought to be related to the acidity of the support, and the more acidic Ru/AO-400 had lower desorption temperature. Combining the activity evaluation, in-situ Raman, DRIFT and TPSR-MS, we can conclude that the synergy between Ru⁰ and Ru^δ+ is crucial for the demethoxylation reaction, i.e., Ru⁰ is in charge of the dissociation of H₂ and suitable adsorption of the aromatics, while Ru^δ+ is responsible for the activation of C-O bond.

According to the TPSR experiments, we found that the hydrogenolysis of methoxy underwent complicated reaction pathways, including both demethoxylation and methoxy decomposition, followed by the CO-WGS. To further investigate the influence of Ru electronic states in these reactions, we carried out the kinetic studies over these three catalysts, Ru/NAO-400, Ru/NAO-200 and Ru/AO-400. H₂ and 4-methylanisole were selected as reactants in this reaction. Based on the in-situ Raman results, RuO_x species, namely Ru^δ+, were responsible for the activation of C-O bond, while the adsorption of aromatic ring and the dissociation of hydrogen were generally believed to occur on metallic M⁰ sites [50,52]. Hence, the kinetic modeling was established on a dual active sites model, i.e., Ru⁰ for activation of hydrogen, and the synergy between Ru⁰ and Ru^δ+ to dissociate the C-O bond in 4-methylanisole. As is described in SI, the demethoxylation of 4-methylanisole was divided into 9 elementary steps, which were i) dissociative adsorption of H₂; ii) dissociative adsorption of 4-methylanisole; iii) surface reaction between adsorbed H and aromatic species; iv) desorption of generated toluene; v) decomposition of adsorbed methoxy into formaldehyde; vi) decomposition of adsorbed formaldehyde into CO; vii) adsorption of H₂O molecules; viii) surface reaction of CO-WGS; and ix) desorption of generated CO₂. The detailed derivation of the kinetic modeling is attached in the supporting information.

Table 2 shows the finest fitted rate expressions and regression forms with respect to hydrogen partial pressure (P_H2) or 4-methylanisole concentration (C_An). As is illustrated in Table 2 and Figures S5-S7, the best fitness results indicate that the RDS would be the desorption of produced toluene (Step iv) over Ru/NAO-400, which had more Ru⁰ species; while the surface reaction of CO-WGS (Step viii) was the RDS over Ru/NAO-200 which possessed relatively more Ru^δ+ species than Ru/NAO-400 (Table 1). This phenomenon was also in accordance with the TPSR-MS results on Ru/NAO-200 (Fig. 4b) that the temperature of CO₂ generation was much higher over Ru/NAO-200 (after 300°C) than that over both Ru/NAO-400 and Ru/AO-400, which were 154 °C and 118 °C, respectively. Interestingly, the dissociation of H₂ (Step i) is the rate-determining step for Ru/AO-400 catalyst as a result of its high ratio of RuO_x species, with its lowest Ru⁰ concentration limiting the activation of H₂ molecules. This is also suggested by the H₂-TPD profiles of these three catalysts in Figure S8 that the adsorption amount of H₂ was much lower over Ru/AO-400 than those over the other two catalysts.

Moreover, the demethoxylation rates over Ru/NAO catalysts with different reduction temperatures followed a volcanic distribution (Fig. 4a), in which Ru/NAO-200 exhibited the highest demethoxylation rate, approaching 20.6 mmol·h⁻¹·gcat⁻¹. In contrast, the rate was extremely low over NAO support, also suggesting that Ru was essential for this reaction. Thus, we can conclude that demethoxylation of 4-methylanisole was an electronic structure sensitive reaction, in which the synergy of different Ru species is sensitive in influencing the catalytic performances. The apparent activation energies for demethoxylation were subsequently measured over three catalysts, which was 39 ± 3 kJ·mol⁻¹, 20 ± 1 kJ·mol⁻¹ and 59 ± 6 kJ·mol⁻¹ over Ru/NAO-400, Ru/NAO-200 and Ru/AO-400, respectively (Fig. 4b), further manifesting that an appropriate Ru⁰/Ru^δ+ ratio could lower the energy barrier, thus promoting the demethoxylation reaction over Ru/NAO-200 catalyst.

Since the surface reaction of CO-WGS after methanol (or methoxy) decomposition was the RDS over Ru/NAO-200, which was the catalyst with the best demethoxylation activity. In-situ DRIFT study of methanol were carried out to investigate the decomposition of methoxy species over different catalysts. We chose Ru/NAO-400 and Ru/NAO-200 as two models that allowed the comparisons on the electronic effect of Ru on the methoxy decomposition. Figures 5a and b show the in-situ DRIFT spectra of methanol over Ru/NAO-400 and Ru/NAO-200. The peak sets of ~ 2950 cm⁻¹ ν_as(C-H), ~ 2840 cm⁻¹ (ν_s(C-H)), 1450-1350 cm⁻¹ (δ(C-H)) and ~ 1030 cm⁻¹ (ν(C-O)) represent a bridged methoxy over the catalyst surface while the band intensity of ν(C-O) decreased simultaneously with the emerging of the band at ~ 1100 cm⁻¹ after adsorption, indicating the transformation of bridged mode to on-top adsorbed methoxys over both Ru/NAO-400 and Ru/NAO-200 catalysts [53,54].

Methoxy species immediately decomposed under the ambient conditions over both catalysts, as we also observed the bands of carbonyl species, which were 1649 cm⁻¹ over Ru/NAO-400, and 1654 cm⁻¹ over Ru/NAO-200, in which the band strengths were probably due to the difference in the interaction between the carbonyl species and Ru species with different electronic states [55]. More interestingly, adsorbed CO molecule (bands in the range of 2050 ~ 2070 cm⁻¹) was even detected in the collected spectra, suggesting a deep decomposition of methoxy to CO. However, the residence time of CO was longer over Ru/NAO-200, and CO molecules kept its existence even after 20 min of Ar sweeping (Fig. 5b). In contrast, the peak intensity of CO was relatively weak at the initial stage, then disappeared with the prolonging sweeping time over Ru/NAO-400 (Fig. 5a). This can be ascribed to the possibility that CO was rapidly consumed via water-gas shift (WGS), once it was generated on the surface [56], in accordance with the anisole-TPSR results in Fig. 3a and b that the generation of CO₂ was easier over Ru/NAO-400 than over Ru/NAO-200.

To verify the capability of CO conversion over these two catalysts, we further performed in-situ DRIFT-MS studies on WGS. In Fig. 6a and b, we can observe three bands of CO species on Ru catalysts at ~ 2135 cm⁻¹, ~ 2070 cm⁻¹ and ~ 2005 cm⁻¹ which are assigned to multi-carbonyl species on Ru^δ+ (Ru^δ+-CO_n), linearly adsorbed CO on Ru^δ+ (Ru^δ+-CO), and linearly adsorbed CO on Ru⁰ (Ru⁰-CO), respectively [57,58]. The bands corresponding to Ru^δ+-CO and Ru⁰-CO were located at relatively higher frequencies over Ru/NAO-200, which can be ascribed to its positively charged Ru particles [59]. With the simultaneous increase in temperature in a continuous 10%CO/He flow with water, the intensity ratios of Ru⁰-CO and Ru^δ+-CO first decreased (Figs. 6c and d), indicating that the linearly adsorbed CO molecules were transferred from Ru⁰ to Ru^δ+ species. After 250 °C, the ratio curve of Ru⁰/Ru^δ+ shot up steeply with the production of CO₂ gas (m/z = 44), suggesting the reduction of some Ru^δ+ species. In addition, the production of H₂ (m/z = 2) over both catalysts existed at higher temperatures than those of CO₂. All these phenomena indicate that CO molecules helped remove the active oxygen species on these two catalysts, leading to the formation of active sites for the dissociation of water. Thus, H₂ was generated via water splitting at higher temperature ranges (both after 310 oC) following the redox mechanism [38,60]. Moreover, the starting point of CO₂ production was much higher over Ru/NAO-200, 196°C in Fig. 6d, compared to 139 °C over Ru/NAO-400 (Fig. 6c), suggesting that it was much more difficult to convert CO over Ru/NAO-200 during the WGS. More interestingly, we also observe that there was no obvious difference (Δ) in Ru⁰/Ru^δ+ ratios over Ru/NAO-200 catalyst, comparing their initial and final states. However, some Ru^δ+ species were completely reduced over Ru/NAO-400 after the test, in which the difference reached ~ 0.3 in Fig. 6c. All these phenomena suggested that the reduction of surface Ru^δ+ species was much more difficult in Ru/NAO-200, contributing to a relatively longer residence of CO on the catalyst surface. As a result, the surface reaction of WGS is the rate-determining step over Ru/NAO-200 catalyst, which is in accordance with the conclusions from the kinetic measurement.