Tuning Particle Sizes and Active Sites of Ni/CeO2 Catalysts and Their Influence on Maleic Anhydride Hydrogenation

Supported metal catalysts are widely used in industrial processes, and the particle size of the active metal plays a key role in determining the catalytic activity. Herein, CeO2-supported Ni catalysts with different Ni loading and particle size were prepared by the impregnation method, and the hydrogenation performance of maleic anhydride (MA) over the Ni/CeO2 catalysts was investigated deeply. It was found that changes in Ni loading causes changes in metal particle size and active sites, which significantly affected the conversion and selectivity of MAH reaction. The conversion of MA reached the maximum at about 17.5 Ni loading compared with other contents of Ni loading because of its proper particle size and active sites. In addition, the effects of Ni grain size, surface oxygen vacancy, and Ni–CeO2 interaction on MAH were investigated in detail, and the possible mechanism for MAH over Ni/CeO2 catalysts was deduced. This work greatly deepens the fundamental understanding of Ni loading and size regimes over Ni/CeO2 catalysts for the hydrogenation of MA and provides a theoretical and experimental basis for the preparation of high-activity catalysts for MAH.


Introduction
Maleic anhydride (MA) is an important C4 fundamental material in the chemical industry that can be obtained by oxidation of coking benzene, butane, or biomass platform compounds. MA is a multifunctional and five-membered ring compound composed of one C=C, double C=O bonds, and one C-O-C functional group. A series of high-value-added fine chemicals such as succinic anhydride (SA), γ-butyrolactone (GBL), and tetrahydrofuran (THF) can be synthesized by maleic anhydride hydrogenation (MAH). These solvents and intermediates are widely used in the military, textile, pharmaceutical, and food industries [1][2][3]. The hydrogenation of MA involves C=C and C=O hydrogenation, and the investigation of hydrogenation mechanism for C=C and C=O bonds has been a hot topic. Until now, the catalysts used in the MAH have mainly been supported Ni-based catalysts, and the supports have mainly been metal or nonmetal oxides such as Al 2 O 3 , SiO 2 , TiO 2 , and CeO 2 [4][5][6][7].
For supported catalysts, the type of active metals, the acid and base properties of the surface, defect sites, and metal-support interactions have important effects on the adsorption and activation forms, hydrogenation path, and product selectivity of MAH. Among these factors, the particle size of the active metal plays a crucial role in the catalytic performance of catalysts [8][9][10]. The geometrical structure, electronic structure, and dispersion of metal particles change dynamically with changes in particle size, and these changes lead to variation in the active sites on the catalyst surface, which significantly affects the catalytic activity of the catalyst [11][12][13][14]. Zhao et al. [15] discovered that for the Ni/SiO 2 catalytic system, when Ni species were fine clusters, the product GBL was obtained from the hydrogenation of MA because of the strong interaction between Ni and the support. However, succinic anhydride (SA) was obtained when the Ni species was in a crystalline state and had weak interaction with the support. Li et al. [16] found that the selectivity of MAH was closely related to the grain size of the active metal, Ni. For a Ni/HY-Al 2 O 3 catalyst, smaller sizes of Ni nanoparticles were favorable for the formation of SA, while as the Ni loading amount increased, the particle size of Ni and the selectivity of GBL increased. Meyer et al. [17] observed that NiO had a stronger interaction with the support when the Ni loading was lower (less than 8 wt%) and that the Ni nanoparticles were conducive to the generation of SA. However, when the Ni loading was gradually increased, NiO particles tended to aggregate on the surface of the support, which reduced the interaction between NiO and the support until more GBL products were finally obtained. Bertone et al. [3] found that compared with a Ni/SiO 2 catalyst, a Ni/SiO 2 -Al 2 O 3 catalyst had smaller grain size of Ni on the surface and showed higher GBL selectivity. They speculated that the Lewis acid on the surface of the SiO 2 -Al 2 O 3 support promoted the formation of GBL. Ma et al. [18] prepared Pd/CeO 2 catalysts with different Pd particle sizes on a CeO 2 carrier and found that the CeO 2 -supported Pd single atomic catalyst showed the best activity for CO oxidation reaction. In addition, in recent works [4,19,20], we synthesized a series of Ni/CeO 2 catalysts under different conditions and investigated deeply the important role of CeO 2 in MA hydrogenation. These works will be very helpful for investigating the effect of the particle size and active sites of metal on MAH. On the basis of regulating the particle size and active sites of metal on CeO 2 support, they provided a new opportunity to comprehensively understand the interaction between the active metal and support and systematically study the change in the active sites of catalysts in heterogeneous catalysis.
Based on the above discussion, in this paper, Ni-supported catalysts with different Ni loading were prepared by the impregnation method using CeO 2 as support, and the hydrogenation performance of the catalysts was investigated carefully. It was found that changes in Ni loading caused changes in the metal particle size and active sites, which significantly affected the conversion and selectivity of MAH reaction. In this work, the effects of Ni grain size, dispersion, surface oxygen vacancy, and Ni-CeO 2 interaction on the hydrogenation of MA were investigated in detail, and the synthesis process of metalsupported catalysts was optimized. This paper provides a theoretical and experimental basis for the preparation of MAH catalysts with higher activity and selectivity.

Catalysts Preparation
The chemicals, including Ce(NO 3 ) 3 ·6H 2 O, Ni(NO 3 ) 2 ·6H 2 O, and NaOH, were purchased from the Sinopharm Chemical Reagent Co., Ltd. (Shanghai, China) and used without any purification. A CeO 2 support was prepared by the sol-gel method. First, 5.00 g Ce(NO 3 ) 3 ·6H 2 O was dissolved into 20 mL distilled water, and then, 6.56 g citric acid (CA) was added and stirred. After the cerium salt and citric acid were completely dissolved, the solution was heated in a water bath at 80 • C until the dry sol was formed. After drying at 120 • C for 8 h in the oven, the dry sol formed a spongy material. It was then moved to a muffle oven and calcined at 500 • C for 3 h to finally obtain the CeO 2 support. xNi/CeO 2 (x: mass content of Ni) catalysts with different loading contents were prepared by citric acid assisted over-volume impregnation method. For the 5Ni/CeO 2 catalyst, 0.505 g Ni(NO 3 ) 2 ·6H 2 O was dissolved in a mixture of 10 mL ethanol and deionized water (volume ratio 1:1), followed by 2.00 g solid CeO 2 and 0.139 g citric acid (mole ratio 1:1). After stirring at room temperature for 30 min, the mixture was placed in a water bath at 80 • C to volatilize the solvent. After the solvent was completely volatilized, the sample was transferred to a drying oven at 120 • C for 8 h. The obtained samples were calcinated at 450 • C for 3 h (heating rate of 3 • C/min) and then reduced for 3 h at 350 • C with H 2 at a flow rate of 50 mL/min to prepare the catalyst, which was used for subsequent characterization and evaluation. According to different content of Ni, the catalysts were labeled as 5Ni/CeO 2 , 10Ni/CeO 2 , 15Ni/CeO 2 , 17.5Ni/CeO 2 , 20Ni/CeO 2 , and 30Ni/CeO 2 .

Catalyst Characterizations and Tests
X-ray diffraction (XRD) was performed using a Bruker D8 Advanced X-ray diffractometer (Billerica, MA, USA). The instrument used Cu Kα1 radiation (λ = 0.15418 nm) as an X-ray source and was supplied with a Ni filter and Vantec detector. The scanning range was 10~80 • , and the scanning rate was 30 • /min. The average crystallite size was calculated by the Scherrer formula, D = Kλ/βcosθ, where K is Scherrer's constant (0.89). The characterizations of H 2 -TPR (hydrogen temperature programed reduction) and H 2 -TPD (hydrogen temperature programmed desorption) were determined using a Micromeritics AutoChem II 2950 chemisorption apparatus. Raman spectroscopy (Raman) was performed on a Raman spectrometer with a laser wavelength of 532 nm (HORIBA, Tokyo, Japan). X-ray photoelectron spectroscopy (XPS) was recorded using a SCIENTIFIC ESCALAB 250 X-ray photoelectron spectrometer (Thermo Company, Waltham, MA, USA) with a standard Al-Kα (h = 1486.6 eV). The spectra were calibrated according to standard C 1s (284.6 eV).
The catalytic performance of the catalyst was evaluated in a 100 mL stainless steel autoclave. First, 0.1 g catalyst was added together with 4.9 g MA and 40 mL THF into the reactor. The N 2 was passed through to replace the air in the reactor 5 times, and then H 2 was passed through 5 times to replace N 2 . Then, the reaction system was heated to 210 • C with stirring at 500 rpm, and the pressure was kept at 5.0 MPa. The product was analyzed using an Agilent 7890A gas chromatograph. To verify precise separation of each component in the products, the programmed temperature was selected. The primary temperature of the oven was increased to 120 • C from 100 • C at a ramp of 5 • C min −1 , and the temperatures of the detector and injector were 190 • C and 260 • C, respectively. The conversion and selectivity of MA to the product were calculated according to the following equations [20]: where C MA , C SA , and C GBL represent the percent content of the reactant and the two products in the reaction, respectively, and X MA and S SA represent the conversion of MA and selectivity of SA. Figure 1 shows the XRD patterns of xNiO/CeO 2 samples with different Ni contents. As shown in Figure 1A, after metal Ni loading, the CeO 2 support still maintained the crystal structure of fluorite cubic phase (JCPDS File 34-0394), similarly to pure CeO 2 [21]. The enlarged pattern (shown in Figure 1B) revealed that the diffraction peaks of the CeO 2 (111) crystal plane in xNiO/CeO 2 samples moved to a higher angle, indicating that the crystal cell parameters of CeO 2 shrank after Ni loading. This may have been due to the Ni 2+ , with its smaller ionic radius (R = 0.72 nm), replacing Ce 4+ (R = 0.81 nm) in the CeO 2 lattice, which resulted in reductions in the cell parameters of CeO 2 [22]. The average crystal sizes of NiO and CeO 2 in the xNiO/CeO 2 samples were calculated by the Scherrer formula, and the results are listed in Table 1. Compared with pure CeO 2 support, the grain size of CeO 2 increased after Ni loading, which may have been caused by sintering during the thermal calcination or the lattice distortion of CeO 2 caused by Ni species [23]. diffraction peak gradually increased, indicating that NiO particles aggregated on the surface of the catalyst and the grain size gradually grew. The particle sizes of NiO are also listed in Table 1, revealing that as Ni loading increased, the particle size of NiO increased from about 10.1 nm to 35.1 nm. The change in NiO grain size led to a change in the interaction between NiO and CeO2, which may have affected the reduction behavior of NiO and the structural difference of the catalyst surface.   Figure 2 shows the XRD patterns of xNi/CeO2 catalysts after reduction at 350 °C . As shown in Figure 2, the CeO2 support maintained a fluorite cubic structure after reduction, and the characteristic peak of NiO disappeared, while the characteristic diffraction peak of the Ni (111) plane appeared at 44.6° (JCPDS 01-1258), indicating that NiO was reduced to metallic Ni. However, for the 5Ni/CeO2 catalyst, the diffraction peak of the metal Ni was not observed, which may have been due to the high dispersion of amorphous Ni species on the catalyst surface or the smaller particle size of Ni (<4 nm). The crystal sizes of Ni in xNi/CeO2 catalysts with different loading content were calculated by the Scherrer formula and are listed in Table 1. As the Ni loading content increased, the metal Ni aggregates on the surface of the catalyst increased, and the average grain size increased gradually from about 10.7 nm to 36.6 nm.  As shown in Figure 1A, the XRD diffraction peaks at 37.0 • , 43.0 • , and 62.9 • corresponded to the characteristic diffraction peaks of NiO's (111), (200), and (220) crystal planes (JCPDS 47-1049), respectively. As Ni loading increased, the intensity of the NiO diffraction peak gradually increased, indicating that NiO particles aggregated on the surface of the catalyst and the grain size gradually grew. The particle sizes of NiO are also listed in Table 1, revealing that as Ni loading increased, the particle size of NiO increased from about 10.1 nm to 35.1 nm. The change in NiO grain size led to a change in the interaction between NiO and CeO 2 , which may have affected the reduction behavior of NiO and the structural difference of the catalyst surface. Figure 2 shows the XRD patterns of xNi/CeO 2 catalysts after reduction at 350 • C. As shown in Figure 2, the CeO 2 support maintained a fluorite cubic structure after reduction, and the characteristic peak of NiO disappeared, while the characteristic diffraction peak of the Ni (111) plane appeared at 44.6 • (JCPDS 01-1258), indicating that NiO was reduced to metallic Ni. However, for the 5Ni/CeO 2 catalyst, the diffraction peak of the metal Ni was not observed, which may have been due to the high dispersion of amorphous Ni species on the catalyst surface or the smaller particle size of Ni (<4 nm). The crystal sizes of Ni in xNi/CeO 2 catalysts with different loading content were calculated by the Scherrer formula and are listed in Table 1. As the Ni loading content increased, the metal Ni aggregates on the surface of the catalyst increased, and the average grain size increased gradually from about 10.7 nm to 36.6 nm.  Figure 3 shows the H2-TPR spectrum of xNiO/CeO2 samples. The peaks of H2 consumption, named α, β, γ, and δ, were well fitted through a Gauss-type function for these samples. The α peak showed lower intensity and a broader shape at about 150 °C , which was attributed to the reduction of oxygen species adsorbed on the surface of CeO2 [24,25]. It has been reported that parts of Ni 2+ species could enter the CeO2 lattice to replace Ce 4+ , which resulted in the distortion of CeO2 lattice and produced oxygen vacancies to balance charges [24]. Raman results also confirmed that the loaded NiO species promoted the formation of oxygen vacancies on the CeO2 surface ( Figure 4A). These oxygen vacancies could adsorb some small oxygen-containing molecules and generate reactive oxygen species, which can easily react with hydrogen [25]. The sharp β peak of H2 consumption at about 200 °C could be attributed to the H2 depletion caused by the dissociation and adsorption of H2 onto the oxygen vacancies or the Ni-Ce interface and the formation of OH groups on the surface. A similar result was found in Ni-Ce solid solution [26]. As shown in Figure 3, as Ni loading increased, the β peak gradually moves towards higher temperatures, and the peak intensity decreased, indicating that the increase in Ni content inhibited the dissociation and adsorption of H2 on the oxygen vacancies or the Ni-Ce interface, which may have been caused by the excessive Ni species masking the oxygen vacancies on the surface.   Figure 3 shows the H 2 -TPR spectrum of xNiO/CeO 2 samples. The peaks of H 2 consumption, named α, β, γ, and δ, were well fitted through a Gauss-type function for these samples. The α peak showed lower intensity and a broader shape at about 150 • C, which was attributed to the reduction of oxygen species adsorbed on the surface of CeO 2 [24,25]. It has been reported that parts of Ni 2+ species could enter the CeO 2 lattice to replace Ce 4+ , which resulted in the distortion of CeO 2 lattice and produced oxygen vacancies to balance charges [24]. Raman results also confirmed that the loaded NiO species promoted the formation of oxygen vacancies on the CeO 2 surface ( Figure 4A). These oxygen vacancies could adsorb some small oxygen-containing molecules and generate reactive oxygen species, which can easily react with hydrogen [25]. The sharp β peak of H 2 consumption at about 200 • C could be attributed to the H 2 depletion caused by the dissociation and adsorption of H 2 onto the oxygen vacancies or the Ni-Ce interface and the formation of OH groups on the surface. A similar result was found in Ni-Ce solid solution [26]. As shown in Figure 3, as Ni loading increased, the β peak gradually moves towards higher temperatures, and the peak intensity decreased, indicating that the increase in Ni content inhibited the dissociation and adsorption of H 2 on the oxygen vacancies or the Ni-Ce interface, which may have been caused by the excessive Ni species masking the oxygen vacancies on the surface.  Figure 3 shows the H2-TPR spectrum of xNiO/CeO2 samples. The peaks of H2 consumption, named α, β, γ, and δ, were well fitted through a Gauss-type function for these samples. The α peak showed lower intensity and a broader shape at about 150 °C , which was attributed to the reduction of oxygen species adsorbed on the surface of CeO2 [24,25]. It has been reported that parts of Ni 2+ species could enter the CeO2 lattice to replace Ce 4+ , which resulted in the distortion of CeO2 lattice and produced oxygen vacancies to balance charges [24]. Raman results also confirmed that the loaded NiO species promoted the formation of oxygen vacancies on the CeO2 surface ( Figure 4A). These oxygen vacancies could adsorb some small oxygen-containing molecules and generate reactive oxygen species, which can easily react with hydrogen [25]. The sharp β peak of H2 consumption at about 200 °C could be attributed to the H2 depletion caused by the dissociation and adsorption of H2 onto the oxygen vacancies or the Ni-Ce interface and the formation of OH groups on the surface. A similar result was found in Ni-Ce solid solution [26]. As shown in Figure 3, as Ni loading increased, the β peak gradually moves towards higher temperatures, and the peak intensity decreased, indicating that the increase in Ni content inhibited the dissociation and adsorption of H2 on the oxygen vacancies or the Ni-Ce interface, which may have been caused by the excessive Ni species masking the oxygen vacancies on the surface.    Figure 4. Raman spectra (A) xNiO/CeO2 precursors and (B) reduced xNi/CeO2 catalysts (a)

Catalyst Characterization
In general, the reduction of NiO species occurs in the temperature range of 20 °C . The asymmetric reduction peaks of NiO were deconvolved into two peaks f consumption, which are labeled γ and δ, respectively. The γ peak at 240 °C w tributed to the reduction of highly dispersed NiO species closely linked to the support. The stronger metal-support interaction promoted the reduction of NiO at temperatures [21]. The δ peak at high temperature (about 275 °C ) was ascribed t reduction of bulk NiO species aggregated on the CeO2 surface. From Figure 3, the r tion temperature of NiO species on the CeO2 surface was lower than that of bulk This was mainly because loaded NiO, with smaller size and larger surface area, more easily contact with H2, which resulted in the lower reduction temperature. M ver, oxygen vacancies and preferential reduced Ni species on the surface of CeO2 su (at 240 °C ) promoted the dissociation and activation of H2, and the overflow of H a to NiO with large particle size was favorable to the reduction of NiO at low tempera It should be noted that as Ni loading increased, the δ peak moved towards high tem atures. A possible reason for this is that the activation and migration of H2 may have inhibited because of the increase in NiO particle size and the decrease in oxygen vac thus retarding the reduction of NiO at low temperatures.
In order to study the effect of Ni loading on the surface structure of CeO2, R characterizations for xNiO/CeO2 samples were conducted, and the results are show Figure 4. The Raman peak intensity of CeO2 in the figure was 0.6 times that of the or peak intensity in order to facilitate comparison of results. For CeO2 support, a s Raman vibration peak was observed at 466 cm −1 , corresponding to the F2g vibr mode for the Ce-O bond in the cubic fluorite structure of CeO2 [27]. After the loadi NiO on the surface of CeO2, the F2g peak intensity of CeO2 decreased, the peak s widened, and the peak position moved towards low wavelengths. This was becaus strong interaction between NiO and CeO2 led to lattice distortion of CeO2, which red the symmetry of the Ce-O bond [25]. Besides the F2g vibration peak, the Raman vibr peak at 600 cm −1 was attributed to the vibration (D band) caused by defect sites o CeO2 surface [25]. Compared with that of the pure CeO2 support, the peak intens the D band of the xNiO/CeO2 sample increased significantly, indicating that the exis of NiO promoted the formation of oxygen vacancies on the CeO2 surface. Howeve vibration peak of NiO at 520 cm −1 could not be observed and may be covered by th vibration peak of CeO2 [26]. Raman spectrum results for the xNiO/CeO2 catalyst reduction are shown in Figure 4B. Similarly to the xNiO/CeO2 precursor, two R characteristic peaks were observed at 466 cm −1 and 600 cm −1 , corresponding to th vibration of Ce-O bond for cubic fluorite CeO2 and the D-band vibration induce surface defects, respectively [25]. In general, the reduction of NiO species occurs in the temperature range of 200-300 • C. The asymmetric reduction peaks of NiO were deconvolved into two peaks for H 2 consumption, which are labeled γ and δ, respectively. The γ peak at 240 • C was attributed to the reduction of highly dispersed NiO species closely linked to the CeO 2 support. The stronger metal-support interaction promoted the reduction of NiO at lower temperatures [21]. The δ peak at high temperature (about 275 • C) was ascribed to the reduction of bulk NiO species aggregated on the CeO 2 surface. From Figure 3, the reduction temperature of NiO species on the CeO 2 surface was lower than that of bulk NiO. This was mainly because loaded NiO, with smaller size and larger surface area, could more easily contact with H 2 , which resulted in the lower reduction temperature. Moreover, oxygen vacancies and preferential reduced Ni species on the surface of CeO 2 support (at 240 • C) promoted the dissociation and activation of H 2 , and the overflow of H atoms to NiO with large particle size was favorable to the reduction of NiO at low temperature. It should be noted that as Ni loading increased, the δ peak moved towards high temperatures. A possible reason for this is that the activation and migration of H 2 may have been inhibited because of the increase in NiO particle size and the decrease in oxygen vacancy, thus retarding the reduction of NiO at low temperatures.
In order to study the effect of Ni loading on the surface structure of CeO 2 , Raman characterizations for xNiO/CeO 2 samples were conducted, and the results are shown in Figure 4. The Raman peak intensity of CeO 2 in the figure was 0.6 times that of the original peak intensity in order to facilitate comparison of results. For CeO 2 support, a strong Raman vibration peak was observed at 466 cm −1 , corresponding to the F2g vibration mode for the Ce-O bond in the cubic fluorite structure of CeO 2 [27]. After the loading of NiO on the surface of CeO 2 , the F2g peak intensity of CeO 2 decreased, the peak shape widened, and the peak position moved towards low wavelengths. This was because the strong interaction between NiO and CeO 2 led to lattice distortion of CeO 2 , which reduced the symmetry of the Ce-O bond [25]. Besides the F2g vibration peak, the Raman vibration peak at 600 cm −1 was attributed to the vibration (D band) caused by defect sites on the CeO 2 surface [25]. Compared with that of the pure CeO2 support, the peak intensity of the D band of the xNiO/CeO 2 sample increased significantly, indicating that the existence of NiO promoted the formation of oxygen vacancies on the CeO 2 surface. However, the vibration peak of NiO at 520 cm −1 could not be observed and may be covered by the F2g vibration peak of CeO 2 [26]. Raman spectrum results for the xNiO/CeO 2 catalyst after reduction are shown in Figure 4B. Similarly to the xNiO/CeO 2 precursor, two Raman characteristic peaks were observed at 466 cm −1 and 600 cm −1 , corresponding to the F2g vibration of Ce-O bond for cubic fluorite CeO 2 and the D-band vibration induced by surface defects, respectively [25]. Figure 5 shows the variation trend of the I D /I F2g ratio with Ni content before and after reduction, which reflects the influence of Ni loading on the oxygen vacancy concentration on the catalyst surface [20]. As shown in Figure 5, the oxygen vacancy concentrations of all xNiO/CeO 2 samples loaded with Ni were higher than that of the CeO 2 support without Ni, which indicates that the addition of Ni was beneficial to the formation of oxygen vacancies on the surface of CeO 2 . Among these NiO/CeO 2 samples, the I D /I F2g ratio of the 5NiO/CeO 2 sample is the highest, and then the I D /I F2g ratio decreased gradually as the Ni content increased, which means that the oxygen vacancy decreased as the Ni content increased. A possible reason for this is that the aggregation of NiO and the growth in particle size on the surface of CeO 2 weakened the interaction of NiO and CeO 2 and covered part of the oxygen vacancies on the surface, which resulted in a decrease in oxygen vacancies.
Nanomaterials 2022, 12, x FOR PEER REVIEW Figure 5 shows the variation trend of the ID/IF2g ratio with Ni content before a reduction, which reflects the influence of Ni loading on the oxygen vacancy con tion on the catalyst surface [20]. As shown in Figure 5, the oxygen vacancy concen of all xNiO/CeO2 samples loaded with Ni were higher than that of the CeO2 without Ni, which indicates that the addition of Ni was beneficial to the forma oxygen vacancies on the surface of CeO2. Among these NiO/CeO2 samples, the ID tio of the 5NiO/CeO2 sample is the highest, and then the ID/IF2g ratio decreased gr as the Ni content increased, which means that the oxygen vacancy decreased as content increased. A possible reason for this is that the aggregation of NiO a growth in particle size on the surface of CeO2 weakened the interaction of NiO an and covered part of the oxygen vacancies on the surface, which resulted in a dec oxygen vacancies. As shown in Figure 5, compared with the xNiO/CeO2 samples, the ID/IF2g ra the xNi/CeO2 catalysts increased significantly after reduction, indicating that the vacancy concentrations on the surface of the xNi/CeO2 catalysts increased obvious H2 reduction. The oxygen vacancy increments of 5Ni/CeO2 and 10Ni/CeO2 were cantly larger than those of other catalysts with higher Ni loading, which sugge lower Ni content was beneficial to the formation of oxygen vacancies on the surfac catalyst. When the loaded content of Ni was low, Ni species and CeO2 were contact and interacted strongly each other, which could have promoted the redu the CeO2 surface and facilitated the formation of oxygen vacancies on the surface ever, as Ni loading increased, the active Ni species began to aggregate and co surface of CeO2, which weakened the Ni-CeO2 interaction and inhibited the redu CeO2 surface.
In order to further study the effect of Ni content on the surface species of N five samples of CeO2, 5Ni/CeO2, 10Ni/CeO2, 17.5Ni/CeO2, and 30Ni/CeO2 were terized by the XPS technique. Figure 6A shows the Ce 3d XPS spectra of the catal peak of Ce is deconvolved into five groups of characteristic peaks according to t ature [28,29]. The three characteristic peaks labeled u and v, u" and v", and u‴ belong to the XPS peaks of 3d1/2 and 3d5/2 of Ce 4+ 3d, while the two characteristic p u' and v' and u0 and v0 belong to the 3d1/2 and 3d5/2 of Ce 3+ 3d. Compared with pur the XPS peak of Ce 4+ in the 5Ni/CeO2 catalyst moved slightly towards the high direction, indicating that the strong interaction between Ni and CeO2 changed t As shown in Figure 5, compared with the xNiO/CeO 2 samples, the I D /I F2g ratios for the xNi/CeO 2 catalysts increased significantly after reduction, indicating that the oxygen vacancy concentrations on the surface of the xNi/CeO 2 catalysts increased obviously after H 2 reduction. The oxygen vacancy increments of 5Ni/CeO 2 and 10Ni/CeO 2 were significantly larger than those of other catalysts with higher Ni loading, which suggests that lower Ni content was beneficial to the formation of oxygen vacancies on the surface of the catalyst. When the loaded content of Ni was low, Ni species and CeO 2 were in close contact and interacted strongly each other, which could have promoted the reduction of the CeO 2 surface and facilitated the formation of oxygen vacancies on the surface. However, as Ni loading increased, the active Ni species began to aggregate and cover the surface of CeO 2 , which weakened the Ni-CeO 2 interaction and inhibited the reduction of CeO 2 surface.
In order to further study the effect of Ni content on the surface species of Ni/CeO 2 , five samples of CeO 2 , 5Ni/CeO 2 , 10Ni/CeO 2 , 17.5Ni/CeO 2 , and 30Ni/CeO 2 were characterized by the XPS technique. Figure 6A shows the Ce 3d XPS spectra of the catalyst. The peak of Ce is deconvolved into five groups of characteristic peaks according to the literature [28,29]. The three characteristic peaks labeled u and v, u" and v", and u"' and v"' belong to the XPS peaks of 3d 1/2 and 3d 5/2 of Ce 4+ 3d, while the two characteristic peaks of u' and v' and u 0 and v 0 belong to the 3d 1/2 and 3d 5/2 of Ce 3+ 3d. Compared with pure CeO 2 , the XPS peak of Ce 4+ in the 5Ni/CeO 2 catalyst moved slightly towards the high-energy direction, indicating that the strong interaction between Ni and CeO 2 changed the electronic configuration of Ce on the surface. Similar phenomena were observed in Pt/CeO 2 and Cu/CeO 2 catalysts, and the peak shift of Ce 4+ should be caused by electron transfer from metal to CeO 2 [30,31]. According to the XPS peaks of Ce 3+ and Ce 4+ , the concentration of Ce 3+ on the catalyst surface was estimated, and the results are listed in Table 2. Per Table 2, the amount of Ce 3+ on the surface of the 5Ni/CeO 2 catalyst was the highest among these samples. As the Ni loading amount increased, the amount of Ce 3+ on the surface gradually decreased and was even lower than that of pure CeO 2 after reduction for 17.5Ni/CeO 2 and 30Ni/CeO 2 . This may have been caused by excessive Ni covering the Ce 3+ on the surface of the catalyst. from metal to CeO2 [30,31]. According to the XPS peaks of Ce 3+ and Ce 4+ , the concentration of Ce 3+ on the catalyst surface was estimated, and the results are listed in Table 2. Per Table 2, the amount of Ce 3+ on the surface of the 5Ni/CeO2 catalyst was the highest among these samples. As the Ni loading amount increased, the amount of Ce 3+ on the surface gradually decreased and was even lower than that of pure CeO2 after reduction for 17.5Ni/CeO2 and 30Ni/CeO2. This may have been caused by excessive Ni covering the Ce 3+ on the surface of the catalyst.   Figure 6B shows the O 1s XPS spectra of the reduced xNi/CeO2 catalysts. After deconvolution, three groups of XPS peaks of O were observed, representing three types of O species. OI and OII represented O species with different coordination in the CeO2 lattice, and OIII represented oxygen species adsorbed at defect sites on the catalyst surface. The OI peak at 528.8 eV was the oxygen species coordinated with Ce 3+ in the CeO2 lattice, while the OII peak with slightly higher binding energy (529.4 eV) represented the oxygen species coordinated with Ce 4+ [32]. The concentration of oxygen vacancies on the surface of the catalyst can be estimated by the ratio OIII/(OI+OII+OIII), and the results are listed in Table 2. Per Table 2, as the Ni content increased, the concentration of oxygen vacancies gradually decreased but was higher than that of the pure CeO2, indicating that the introduction of Ni promotes the formation of oxygen vacancies on CeO2 surface, which was consistent with the Raman results. Figure 7 shows the XPS peaks of Ni 2p3/2 for all catalysts. In addition to the satellite shake-up peak of Ni at about 861.0 eV, three fitting peaks represented three kinds of Ni species with different chemical states, namely α, β, and γ, which were assigned to Ni 0 (~852.4 eV), Ni 2+ (~854.7 eV), and Ni 3+ (~856.8 eV), respectively. Three kinds of Ni species coexisted on the surface of the Ni/CeO2 catalysts. According to previous research [30,33,34], highly dispersed Ni clusters can interact with CeO2 support to generate the Ni-O-Ce structure, in which case the outer electrons of Ni would transfer to the 4f orbital of Ce through the Ni-O-Ce bond, which would result in the formation of Ni 2+ or Ni δ+ . Ni 3+ ions should come from Ni species that enter into the CeO2 lattice and form a NixCe1-xO2-y   Table 2. Per Table 2, as the Ni content increased, the concentration of oxygen vacancies gradually decreased but was higher than that of the pure CeO 2 , indicating that the introduction of Ni promotes the formation of oxygen vacancies on CeO 2 surface, which was consistent with the Raman results. Figure 7 shows the XPS peaks of Ni 2 p3/2 for all catalysts. In addition to the satellite shake-up peak of Ni at about 861.0 eV, three fitting peaks represented three kinds of Ni species with different chemical states, namely α, β, and γ, which were assigned to Ni 0 (~852.4 eV), Ni 2+ (~854.7 eV), and Ni 3+ (~856.8 eV), respectively. Three kinds of Ni species coexisted on the surface of the Ni/CeO 2 catalysts. According to previous research [30,33,34], highly dispersed Ni clusters can interact with CeO 2 support to generate the Ni-O-Ce structure, in which case the outer electrons of Ni would transfer to the 4f orbital of Ce through the Ni-O-Ce bond, which would result in the formation of Ni 2+ or Ni δ+ . Ni 3+ ions should come from Ni species that enter into the CeO 2 lattice and form a Ni x Ce 1−x O 2−y solid solution with CeO 2 [35]. According to the peak area of different Ni species, the proportionate relationship among different Ni species was estimated, and the results are listed in Table 2. From Table 2, as Ni loading increased, the content of Ni 0 gradually increased, while the content of Ni 2+ gradually decreased. This was due to the fact that when the content of loaded Ni was low, the Ni particles with smaller size were highly dispersed on the surface of the catalyst and had stronger interaction with CeO 2 support, which made the outer electrons of Ni easily transfer to CeO 2 , thus forming more Ni 2+ . However, the increase in Ni loading led to the growth of the Ni particle size, which weakened the electron induction effect of CeO 2 on Ni and led to the decrease in Ni 2+ content. In addition, the relative content of Ni 3+ was relatively low for all xNi/CeO 2 catalysts, which means that only a small amount of Ni formed a Ni x Ce 1−x O 2−y solid solution with the CeO 2 support because of the limitation of the loading method (the impregnation method).
Nanomaterials 2022, 12, x FOR PEER REVIEW solid solution with CeO2 [35]. According to the peak area of different Ni speci proportionate relationship among different Ni species was estimated, and the resu listed in Table 2. From Table 2, as Ni loading increased, the content of Ni 0 gradua creased, while the content of Ni 2+ gradually decreased. This was due to the fact tha the content of loaded Ni was low, the Ni particles with smaller size were high persed on the surface of the catalyst and had stronger interaction with CeO2 su which made the outer electrons of Ni easily transfer to CeO2, thus forming mor However, the increase in Ni loading led to the growth of the Ni particle size, weakened the electron induction effect of CeO2 on Ni and led to the decrease content. In addition, the relative content of Ni 3+ was relatively low for all xNi/CeO lysts, which means that only a small amount of Ni formed a NixCe1-xO2-y solid so with the CeO2 support because of the limitation of the loading method (the impreg method).  The H2-TPD characterization results of the CeO2 support and each catalyst are in Figure 8. As shown in Figure 8A, the CeO2 support had the ability to activate a sorb H2 before and after reduction at 350 °C , and the oxygen vacancies had a gr fluence on the form of existence for the adsorption of H2 [36]. In order to further stu hydrogen species on CeO2 surface, H2-TPD combined with a mass spectrometer (M used to detect the desorbed H2 species. Figure 9A shows that H2 was desorbed form of H2O in the range of 150-400 °C on the surface of unreduced CeO2 (labelled indicating that the adsorption of H2 on the surface was irreversible, and OH group generated on the surface of the support. In contrast, as shown in Figure 9B, on the s of reduced CeO2 (labelled CeO2-350), the adsorbed atomic H was desorbed from th surface in the form of H2 at about 80 °C , meaning that the oxygen vacancies on the s of reduced CeO2 were favorable for the reversible adsorption of H2, which was con with literature reports [37]. The H 2 -TPD characterization results of the CeO 2 support and each catalyst are listed in Figure 8. As shown in Figure 8A, the CeO 2 support had the ability to activate and adsorb H 2 before and after reduction at 350 • C, and the oxygen vacancies had a great influence on the form of existence for the adsorption of H 2 [36]. In order to further study the hydrogen species on CeO 2 surface, H 2 -TPD combined with a mass spectrometer (MS) was used to detect the desorbed H 2 species. Figure 9A shows that H 2 was desorbed in the form of H 2 O in the range of 150-400 • C on the surface of unreduced CeO 2 (labelled CeO 2 ), indicating that the adsorption of H 2 on the surface was irreversible, and OH groups were generated on the surface of the support. In contrast, as shown in Figure 9B, on the surface of reduced CeO 2 (labelled CeO 2 -350), the adsorbed atomic H was desorbed from the CeO 2 surface in the form of H 2 at about 80 • C, meaning that the oxygen vacancies on the surface of reduced CeO 2 were favorable for the reversible adsorption of H 2 , which was consistent with literature reports [37]. As shown in Figure 8B, xNi/CeO2 catalysts with different Ni loadings had simil H2-TPD spectra. The desorption peak was fitted into three desorption peaks by t Gaussian method. The desorption peak (α peak) at about 80 °C was similar to the desorption peak of CeO2 after reduction and was attributed to the desorption of H2 fro the support surface. The desorption peaks β and γ were attributed to the desorption of adsorbed on different Ni species. The β peak could be assigned to the H2 desorption of species at the Ni-CeO2 interface. The strong interaction between Ni and CeO2 suppo weakened the binding ability of Ni species to H2 and then lowered the energy barrier H2 desorption. The desorption peak γ (at 178 °C ) was assigned to the desorption of h drogen species adsorbed on the surface of Ni in bulk phase, which was similar to the desorption on the Ni surface in Ni/Al2O3 and Ni/SiO2 systems and indicated that t support had little influence on the H2 adsorption capacity on Ni species here [38]. It c be concluded that different Ni species on the xNi/CeO2 surface had different adsorpti and activation abilities for H2.
Based on H2-TPD results, H2 adsorption volumes at different active sites were es As shown in Figure 8B, xNi/CeO2 catalysts with different Ni loadings had simi H2-TPD spectra. The desorption peak was fitted into three desorption peaks by t Gaussian method. The desorption peak (α peak) at about 80 °C was similar to the desorption peak of CeO2 after reduction and was attributed to the desorption of H2 fro the support surface. The desorption peaks β and γ were attributed to the desorption of adsorbed on different Ni species. The β peak could be assigned to the H2 desorption of species at the Ni-CeO2 interface. The strong interaction between Ni and CeO2 suppo weakened the binding ability of Ni species to H2 and then lowered the energy barrier H2 desorption. The desorption peak γ (at 178 °C ) was assigned to the desorption of h drogen species adsorbed on the surface of Ni in bulk phase, which was similar to the desorption on the Ni surface in Ni/Al2O3 and Ni/SiO2 systems and indicated that t support had little influence on the H2 adsorption capacity on Ni species here [38]. It c be concluded that different Ni species on the xNi/CeO2 surface had different adsorpti and activation abilities for H2.
Based on H2-TPD results, H2 adsorption volumes at different active sites were es mated and correlated with Ni loading. Figure 10A shows that compared with reduc As shown in Figure 8B, xNi/CeO 2 catalysts with different Ni loadings had similar H 2 -TPD spectra. The desorption peak was fitted into three desorption peaks by the Gaussian method. The desorption peak (α peak) at about 80 • C was similar to the H 2 desorption peak of CeO 2 after reduction and was attributed to the desorption of H 2 from the support surface. The desorption peaks β and γ were attributed to the desorption of H 2 adsorbed on different Ni species. The β peak could be assigned to the H 2 desorption of Ni species at the Ni-CeO 2 interface. The strong interaction between Ni and CeO 2 support weakened the binding ability of Ni species to H 2 and then lowered the energy barrier of H 2 desorption. The desorption peak γ (at 178 • C) was assigned to the desorption of hydrogen species adsorbed on the surface of Ni in bulk phase, which was similar to the H 2 desorption on the Ni surface in Ni/Al 2 O 3 and Ni/SiO 2 systems and indicated that the support had little influence on the H 2 adsorption capacity on Ni species here [38]. It can be concluded that different Ni species on the xNi/CeO 2 surface had different adsorption and activation abilities for H 2 .
Based on H 2 -TPD results, H 2 adsorption volumes at different active sites were estimated and correlated with Ni loading. Figure 10A shows that compared with reduced CeO 2 , the H 2 adsorption capacity of xNi/CeO 2 catalysts greatly increased, confirming that Ni was the center of adsorption and activation of H 2 . For all xNi/CeO 2 catalysts, as the Ni loading increased, the amount of adsorbed H 2 on the catalyst first increased and then decreased gradually. A possible reason for this is that as the Ni loading increased, the Ni particles aggregated on the surface of CeO 2 , and the grain size became larger, which may have reduced the number of active sites for H 2 adsorption. Figure 10B shows that the peak area of α desorption for 5Ni/CeO 2 was the largest among the samples. The peak area of the other samples decreased as the Ni loading increased, which indicates that the 5Ni/CeO 2 catalyst possessed the highest concentration of oxygen vacancies for H 2 adsorption. According to Raman and XPS results, excessive Ni was not conducive to the formation of oxygen vacancies on the surface and inhibited the ability of oxygen vacancies to activate hydrogen [39]. In addition, as the Ni loading increased, the peak areas of β and γ increased gradually in the beginning and then decreased obviously after 17.5Ni/CeO 2 . The results showed that a proper amount of Ni loading was helpful to increase the number of active sites on the support surface, while an excessive amount of Ni loading may have led to the aggregation and growth of Ni species, which could reduce the surface area of Ni particles and the number of active sites on the surface. have reduced the number of active sites for H2 adsorption. Figure 10B shows that the peak area of α desorption for 5Ni/CeO2 was the largest among the samples. The peak area of the other samples decreased as the Ni loading increased, which indicates that the 5Ni/CeO2 catalyst possessed the highest concentration of oxygen vacancies for H2 adsorption. According to Raman and XPS results, excessive Ni was not conducive to the formation of oxygen vacancies on the surface and inhibited the ability of oxygen vacancies to activate hydrogen [39]. In addition, as the Ni loading increased, the peak areas of β and γ increased gradually in the beginning and then decreased obviously after 17.5Ni/CeO2. The results showed that a proper amount of Ni loading was helpful to increase the number of active sites on the support surface, while an excessive amount of Ni loading may have led to the aggregation and growth of Ni species, which could reduce the surface area of Ni particles and the number of active sites on the surface.  Figure 11 shows the conversion curves of maleic anhydride (MA) over xNi/CeO2 catalysts and reduced CeO2 support at 210 °C and 5 MPa. After a reaction time of 1 h, the conversion of MA for all xNi/CeO2 catalysts was close to 100%, and the main product was succinic anhydride (SA), indicating that all xNi/CeO2 catalysts showed high hydrogenation activity for the C=C bond. It is noteworthy that the reduced CeO2 carrier also had a certain ability of MAH and that the conversion of MA was about 30% after 1 h under the same conditions. When Ni species were loaded on the surface of CeO2, the activity of MA hydrogenation increased sharply, indicating that Ni was the main active site for the MAH reaction. For all xNi/CeO2 catalysts, in the initial time, the catalytic activity for MAH increased gradually as Ni content increased until 17.5 wt%, and then the conversion of MA decreased slightly until 1 h.  Figure 11 shows the conversion curves of maleic anhydride (MA) over xNi/CeO 2 catalysts and reduced CeO 2 support at 210 • C and 5 MPa. After a reaction time of 1 h, the conversion of MA for all xNi/CeO 2 catalysts was close to 100%, and the main product was succinic anhydride (SA), indicating that all xNi/CeO 2 catalysts showed high hydrogenation activity for the C=C bond. It is noteworthy that the reduced CeO 2 carrier also had a certain ability of MAH and that the conversion of MA was about 30% after 1 h under the same conditions. When Ni species were loaded on the surface of CeO 2 , the activity of MA hydrogenation increased sharply, indicating that Ni was the main active site for the MAH reaction. For all xNi/CeO 2 catalysts, in the initial time, the catalytic activity for MAH increased gradually as Ni content increased until 17.5 wt%, and then the conversion of MA decreased slightly until 1 h. In order to further investigate the C=C hydrogenation performance of xNi/C catalysts, the turnover frequency values for MA to SA (TOFMA→SA) over the active Ni w calculated and correlated with the oxygen vacancies, Ni species, and Ni loading cont As shown in Figure 12, the TOFMA→SA of the xNi/CeO2 catalysts decreased as the Ni c tent increased, which was consistent with the change trend of oxygen vacancies on surface, indicating that the oxygen vacancies of the catalyst also played an important in the C=C hydrogenation of MA. According to H2-TPR and H2-TPD results, oxygen cancies not only improved the dissociation and adsorption capacity of H2 on the cata but promoted the diffusion of active H on the catalyst surface, providing more activ species for the hydrogenation reaction [37]. Moreover, according to theoretical calc tions, oxygen vacancies with rich electron structure can provide electrons to the ac metal and enhance the electron-giving ability of the active metal, thus improving the C hydrogenation performance of the metal [40]. For the xNi/CeO2 catalytic system, it can speculated that the synergistic effect between active metal Ni and oxygen vacan (Ovac) could have improved the C=C hydrogenation performance of Ni.  Figure 13A shows the trend of SA selectivity with reaction time on different c In order to further investigate the C=C hydrogenation performance of xNi/CeO 2 catalysts, the turnover frequency values for MA to SA (TOF MA→SA ) over the active Ni were calculated and correlated with the oxygen vacancies, Ni species, and Ni loading content. As shown in Figure 12, the TOF MA→SA of the xNi/CeO 2 catalysts decreased as the Ni content increased, which was consistent with the change trend of oxygen vacancies on the surface, indicating that the oxygen vacancies of the catalyst also played an important role in the C=C hydrogenation of MA. According to H 2 -TPR and H 2 -TPD results, oxygen vacancies not only improved the dissociation and adsorption capacity of H 2 on the catalyst but promoted the diffusion of active H on the catalyst surface, providing more active H species for the hydrogenation reaction [37]. Moreover, according to theoretical calculations, oxygen vacancies with rich electron structure can provide electrons to the active metal and enhance the electron-giving ability of the active metal, thus improving the C=C hydrogenation performance of the metal [40]. For the xNi/CeO 2 catalytic system, it can be speculated that the synergistic effect between active metal Ni and oxygen vacancies (Ovac) could have improved the C=C hydrogenation performance of Ni. In order to further investigate the C=C hydrogenation performance of xNi/CeO catalysts, the turnover frequency values for MA to SA (TOFMA→SA) over the active Ni were calculated and correlated with the oxygen vacancies, Ni species, and Ni loading content As shown in Figure 12, the TOFMA→SA of the xNi/CeO2 catalysts decreased as the Ni con tent increased, which was consistent with the change trend of oxygen vacancies on the surface, indicating that the oxygen vacancies of the catalyst also played an important role in the C=C hydrogenation of MA. According to H2-TPR and H2-TPD results, oxygen va cancies not only improved the dissociation and adsorption capacity of H2 on the catalys but promoted the diffusion of active H on the catalyst surface, providing more active H species for the hydrogenation reaction [37]. Moreover, according to theoretical calcula tions, oxygen vacancies with rich electron structure can provide electrons to the active metal and enhance the electron-giving ability of the active metal, thus improving the C=C hydrogenation performance of the metal [40]. For the xNi/CeO2 catalytic system, it can be speculated that the synergistic effect between active metal Ni and oxygen vacancies (Ovac) could have improved the C=C hydrogenation performance of Ni.  Figure 13A shows the trend of SA selectivity with reaction time on different cata lysts. From Figure 13A and B, the selectivity of SA for all xNi/CeO2 catalysts was around 100% at the initial reaction time of 40 min and then decreased gradually while the selec tivity of γ-butyrolactone (GBL) increased gradually. The selectivity of GBL on the  Figure 13A shows the trend of SA selectivity with reaction time on different catalysts. From Figure 13A and B, the selectivity of SA for all xNi/CeO 2 catalysts was around 100% at the initial reaction time of 40 min and then decreased gradually while the selectivity of γ-butyrolactone (GBL) increased gradually. The selectivity of GBL on the 17.5Ni/CeO 2 catalyst was the highest (about 35.7%) after 8 h compared with the other xNi/CeO 2 catalysts. In addition, the selectivity of SA on the CeO 2 support remained at 100% within 8 h of the reaction, indicating that the CeO 2 support had almost no hydrogenation activity for the C=O bond. The above results identify that the metal Ni was the active center for the hydrogenation of SA to GBL and that the content of Ni loading significantly affected the C=O hydrogenation over the catalyst. As for the stability of the xNi/CeO 2 catalysts, it should be noted that all samples showed good stability in the hydrogenation process. After a reaction time of 1 h, the conversion of MA for all xNi/CeO 2 catalysts was close to 100%, and the catalysts kept their high catalytic performance. Furthermore, after five cycles of use, all the catalysts kept their high activity and selectivity, and there was no obvious decrease in either. In addition, the stability of the 17.5Ni/CeO 2 catalyst had no obvious change compared with other catalysts in the MAH process.  It has been reported that the metal Ni with certain grain size is the active center of hydrogenation of SA to GBL. Meyer et al. [17] studied the effect of Ni loading on the hydrogenation of MA and found that Ni/SiO2-Al2O3 catalyst had hydrogenation activity for C=O only when Ni loading was more than 8 wt%. They concluded that a certain size of Ni grain was the active center for the hydrogenation of SA to GBL. In this work, the selectivity of GBL also showed a strong dependence on the particle size of Ni. However, when the particle size of Ni exceeded a certain amount (17.5 nm), the hydrogenation activity of C=O started to decrease. For example, though the average sizes of Ni particles on the 20Ni/CeO2 (18.9 nm) and 30Ni/CeO2 catalysts (36.6 nm) were larger than that of the 17.5Ni/CeO2 catalyst (17.5 nm), the selectivities of GBL were lower than that of 17.5Ni/CeO2 (as shown in Figure 13B).

Catalytic Performance
In order to understand deeply the influence of catalysts on the hydrogenation activity of C=O, the values of TOFSA→GBL over different catalysts were calculated. As shown in Figure 14, as the Ni loading increased, the value of TOFSA→GBL gradually increased. When the content of Ni was 17.5 wt%, the value of TOFSA→GBL reached the maximum. It then rapidly decreased as the Ni loading increased further. At the same time, the H2 concentration adsorbed on Ovac decreased monotonously as the Ni loading increased. The volcanic curve for TOFSA→GBL showed that the hydrogenation of SA to GBL was structure sensitive, which is quite different from the trend of TOFMA→SA in Figure 12A. It has been reported that the metal Ni with certain grain size is the active center of hydrogenation of SA to GBL. Meyer et al. [17] studied the effect of Ni loading on the hydrogenation of MA and found that Ni/SiO 2 -Al 2 O 3 catalyst had hydrogenation activity for C=O only when Ni loading was more than 8 wt%. They concluded that a certain size of Ni grain was the active center for the hydrogenation of SA to GBL. In this work, the selectivity of GBL also showed a strong dependence on the particle size of Ni. However, when the particle size of Ni exceeded a certain amount (17.5 nm), the hydrogenation activity of C=O started to decrease. For example, though the average sizes of Ni particles on the 20Ni/CeO 2 (18.9 nm) and 30Ni/CeO 2 catalysts (36.6 nm) were larger than that of the 17.5Ni/CeO 2 catalyst (17.5 nm), the selectivities of GBL were lower than that of 17.5Ni/CeO 2 (as shown in Figure 13B).
In order to understand deeply the influence of catalysts on the hydrogenation activity of C=O, the values of TOF SA→GBL over different catalysts were calculated. As shown in Figure 14, as the Ni loading increased, the value of TOF SA→GBL gradually increased. When the content of Ni was 17.5 wt%, the value of TOF SA→GBL reached the maximum. It then rapidly decreased as the Ni loading increased further. At the same time, the H 2 concentration adsorbed on Ovac decreased monotonously as the Ni loading increased. The volcanic curve for TOF SA→GBL showed that the hydrogenation of SA to GBL was structure sensitive, which is quite different from the trend of TOF MA→SA in Figure 12A. According to previous research [41][42][43], for reducible supports, such as TiO2 a CeO2, the metal-support interface is considered to be the active site of C=O adsorpti activation. It was found that the C=O functional groups could be adsorbed and polariz at the interfaces of Pt-TiOx and Ni-TiOx and that the catalytic activity of hydrogenati for crotonaldehyde to crotonyl alcohol was significantly improved [41]. Because of stro interaction with the carrier, the electronic configuration for most of the metal particles the interface was in an ionic state (such as Ni 2+ at the interface of Ni-TiO2). These io metal particles could play the role of a Lewis acid and participate in the adsorption a activation of C=O functional groups [42]. In addition, in the hydrogenation reaction citral, a small amount of Ni 2+ at the Ni-TiO2 interface promoted the adsorption and ac vation of C=O in the citral molecule and finally improved the selectivity of the hyd genation of citral to citric alcohol [41,43].
As shown in Figure 14B, the change trend of TOFSA→GBL was consistent with t change in the total amount of H2 adsorbed on the active sites of interface Ni and bulk which indicates that both the interfacial Ni and bulk Ni 0 could catalyze the hydrogen tion reaction of the C=O bond. According to the characterization results of H2-TP H2-TPD, and XPS, the Ni species at the interface showed a valence state of Ni δ+ because the strong interaction with the CeO2 support [34]. Therefore, it can be inferred that Ni δ the interface could also promote the adsorption of C=O on the catalyst surface as t Lewis acid site. Based on the catalytic effect of metal Ni on the adsorption and activati of C atoms in C=O and subsequent C-O bond breaking [33], we propose the possi mechanism of the Ni δ+ -Ni 0 synergistic effect on the hydrogenation reaction of C=O. shown in Figure 15, first, the metal Ni 0 adsorbs and activates C atoms in the C=O fu tional group, and Ni δ+ at the interface acts as a Lewis acid to synergistically activate atoms. Second, the synergistic effect of Ni δ+ and Ni 0 promotes the adsorption and acti tion of C=O, and the activated C=O group reacts with highly active hydrogen atoms the surface of metal Ni, which results in the C=O bond hydrogenation and subseque C-O fracture. According to this mechanism, if the particle size of Ni becomes larger, t distance between the top Ni 0 and the bottom Ni δ+ increases, which weakens the syn gistic activation for C=O by Ni δ+ -Ni 0 . This constitutes a good explanation for the ph nomenon in which the selectivity of GBL decreased as the average particle size of Ni creased beyond 17.5 nm. According to previous research [41][42][43], for reducible supports, such as TiO 2 and CeO 2 , the metal-support interface is considered to be the active site of C=O adsorption activation. It was found that the C=O functional groups could be adsorbed and polarized at the interfaces of Pt-TiOx and Ni-TiOx and that the catalytic activity of hydrogenation for crotonaldehyde to crotonyl alcohol was significantly improved [41]. Because of strong interaction with the carrier, the electronic configuration for most of the metal particles at the interface was in an ionic state (such as Ni 2+ at the interface of Ni-TiO 2 ). These ionic metal particles could play the role of a Lewis acid and participate in the adsorption and activation of C=O functional groups [42]. In addition, in the hydrogenation reaction of citral, a small amount of Ni 2+ at the Ni-TiO 2 interface promoted the adsorption and activation of C=O in the citral molecule and finally improved the selectivity of the hydrogenation of citral to citric alcohol [41,43].
As shown in Figure 14B, the change trend of TOF SA→GBL was consistent with the change in the total amount of H 2 adsorbed on the active sites of interface Ni and bulk Ni, which indicates that both the interfacial Ni and bulk Ni 0 could catalyze the hydrogenation reaction of the C=O bond. According to the characterization results of H 2 -TPR, H 2 -TPD, and XPS, the Ni species at the interface showed a valence state of Ni δ+ because of the strong interaction with the CeO 2 support [34]. Therefore, it can be inferred that Ni δ+ at the interface could also promote the adsorption of C=O on the catalyst surface as the Lewis acid site. Based on the catalytic effect of metal Ni on the adsorption and activation of C atoms in C=O and subsequent C-O bond breaking [33], we propose the possible mechanism of the Ni δ+ -Ni 0 synergistic effect on the hydrogenation reaction of C=O. As shown in Figure 15, first, the metal Ni 0 adsorbs and activates C atoms in the C=O functional group, and Ni δ+ at the interface acts as a Lewis acid to synergistically activate O atoms. Second, the synergistic effect of Ni δ+ and Ni 0 promotes the adsorption and activation of C=O, and the activated C=O group reacts with highly active hydrogen atoms on the surface of metal Ni, which results in the C=O bond hydrogenation and subsequent C-O fracture. According to this mechanism, if the particle size of Ni becomes larger, the distance between the top Ni 0 and the bottom Ni δ+ increases, which weakens the synergistic activation for C=O by Ni δ+ -Ni 0 . This constitutes a good explanation for the phenomenon in which the selectivity of GBL decreased as the average particle size of Ni increased beyond 17.5 nm.

Conclusions
In this work, Ni/CeO2 catalysts were synthesized by the impregnation method, series of xNi/CeO2 catalysts with different particle sizes and active sites were succes prepared by changing the Ni loading. The effects of particle size and active si Ni/CeO2 on the hydrogenation of MA were systematically studied. It was found th catalytic activity of the xNi/CeO2 catalysts was size dependent for MAH and th metal Ni was the active center for the catalytic hydrogenation of C=C from MA to S of C=O from SA to GBL. In the beginning of the reaction, the hydrogenation activ the catalyst increased as the Ni loading increased until 17.5Ni/CeO2 and then decr gradually as the Ni loading increased further. The oxygen vacancies on the surf Ni/CeO2 could promote the adsorption and activation of H2, and the synergistic ef active metal Ni and oxygen vacancies could improve the hydrogenation of the C=C The synergistic effect of Ni δ+ species obtained from the strong electronic attraction CeO2 support and Ni 0 promoted the adsorption and activation of C=O in MAH current results confirmed that the particle size and catalytic ability of Ni/CeO2 cat could be modulated through changing the Ni loading on the CeO2 support. This not only provides a deep understanding of MA hydrogenation over Ni/CeO2 ca but highlights the potential of size-dependent catalysts in heterogeneous catalysis.   Figure 15. The synergy of Ni δ+ -Ni 0 in C=O hydrogenolysis over Ni/CeO 2 catalyst.

Conclusions
In this work, Ni/CeO 2 catalysts were synthesized by the impregnation method, and a series of xNi/CeO 2 catalysts with different particle sizes and active sites were successfully prepared by changing the Ni loading. The effects of particle size and active sites of Ni/CeO 2 on the hydrogenation of MA were systematically studied. It was found that the catalytic activity of the xNi/CeO 2 catalysts was size dependent for MAH and that the metal Ni was the active center for the catalytic hydrogenation of C=C from MA to SA and of C=O from SA to GBL. In the beginning of the reaction, the hydrogenation activity of the catalyst increased as the Ni loading increased until 17.5Ni/CeO 2 and then decreased gradually as the Ni loading increased further. The oxygen vacancies on the surface of Ni/CeO 2 could promote the adsorption and activation of H 2 , and the synergistic effect of active metal Ni and oxygen vacancies could improve the hydrogenation of the C=C bond. The synergistic effect of Ni δ+ species obtained from the strong electronic attraction of the CeO 2 support and Ni 0 promoted the adsorption and activation of C=O in MAH. The current results confirmed that the particle size and catalytic ability of Ni/CeO 2 catalysts could be modulated through changing the Ni loading on the CeO 2 support. This work not only provides a deep understanding of MA hydrogenation over Ni/CeO 2 catalysts but highlights the potential of size-dependent catalysts in heterogeneous catalysis.