The Catalytic Hydrogenation of Maleic Anhydride on CeO 2 − δ-Supported Transition Metal Catalysts

The proper selection of transition metals and support is pivotal to the design of active and selective catalysts for maleic anhydride hydrogenation (MAH). Herein, the M/CeO2−δ (M = Co, Ni, Cu, respectively) catalysts with pre-optimised metal loading of 10 wt % were prepared via a wet impregnation method and well characterized to corroborate their MAH performance with the properties of metal, support and the M/CeO2−δ catalysts. The results revealed that the metal dispersion on the catalyst declines in the order of Ni/CeO2−δ > Co/CeO2−δ > Cu/CeO2−δ, similar to the apparent activity for maleic anhydride (MA) transformation to succinic anhydride (SA). The hydrogenolysis of SA to γ-butyrolactone (GBL) occurs on Ni/CeO2−δ and Co/CeO2−δ only when the MA → SA transformation completing. The Ni/CeO2−δ displays superior activity and selectivity to Co/CeO2−δ in both MA→ SA and SA→ GBL reactions, while the Cu/CeO2−δ and CeO2−δ are both inert for SA → GBL hydrogenolysis. The MA hydrogenation to SA follows the first order kinetic law on the Ni/CeO2−δ and Co/CeO2−δ catalysts yet a more complex kinetic characteristics observed on the Cu/CeO2−δ. The distinct catalytic hydrogenation behaviours of the M/CeO2−δ catalysts are assigned to the synergism of dispersion and electronic configuration of the transition metals and oxygen vacancies.


Introduction
Catalytic hydrogenation of maleic anhydride (MA) is an essential route for the current industrial production of the succinic anhydride (SA), γ-butyrolactone (GBL) and tetrahydrofuran (THF), which are value-added intermediates for pharmaceuticals and bio-degradable polymers and highly-demanded solvents for various industrial processes [1][2][3].However, the MA hydrogenation (MAH) is a rather complex process that requires deliberately selecting metal catalyst to avoid excessive hydrogenation and concomitance of ring-open side reaction [1].
Currently, the main catalysts widely applied in the MAH processes are noble metal catalysts (Pd, Ru and Au) [4] and transition metal copper chromite (Cu-Cr) catalysts [5], though they are either costly, less selective or toxic.For developing environmentally-friendly and low-cost alternative MAH catalysts, extensive efforts have been made on: (1) exploring various transition metal Ni [1,2,6], Co [3], Cu [7][8][9] and bimetallic catalysts [10] with controlled particle sizes and surface dispersion; (2) modifying the traditional porous Al 2 O 3 [11], SiO 2 [3], zeolite [2] and SiO 2 -Al 2 O 3 composite oxide supports [8] using functional promoters (e.g., CeO 2 [12], TiO 2 [13] and ZrO 2 [14], etc.) to moderate the surface acidity and redox properties [8, 15,16]; and (3) employing new catalyst supports, for example the structure-defective TiO 2 [17] and CeO 2 [1] have emerged as excellent supports for metallic Ni catalysts and demonstrated exceptional selectivity in MAH.The previous research advances in MAH catalysts revealed that their activity and selectivity depend greatly on the proper selection of metal and support and on the corresponding metal-support interaction.However, the systematic and comparative studies of the influence of transition metals and their supports on the MAH performance have not been well documented in literature, in particular the emerging ceria and titania supports.
In our previous work, the superior activity and selectivity in MAH on the Ni/CeO 2−δ catalyst to those on the Ni/Al 2 O 3 catalyst was attributed to the synergism of the Ni-CeO 2−δ interaction, the oxygen vacancies and developed porosity of the CeO 2−δ support [1].In theory, the electron-rich O vac in CeO 2−δ may modify the geometric and electronic structure of the active metal [18,19], as well as enforce the metal-support interaction [20].It is thought that the O vac defects could synergistically enhance the catalytic performance of M/CeO 2−δ catalysts, through coordination with the oxygen-containing group in OH cleavage [21], NO 2 reduction [22] and CO 2 methanation [23].Moreover, the ionic TM (M δ+ ) species can diffuse into the CeO 2 lattice and modulate the population of O vac in mixed oxides with dependence of the ionic radius and the valence state of the M [24,25].Although M/CeO 2−δ catalysts have been examined in various catalytic reactions, there is very limited knowledge of the metal influences on the CeO 2−δ supports, particularly for the activity and selectivity in MAH.
In this paper, various TMs (Co, Ni, Cu) supported on ceria were prepared by a simple wet impregnation method and employed in liquid phase hydrogenation of MA at 210 • C and 5 MPa.Through careful characterizations, the catalyst activity and product selectivity in MAH on the M/CeO 2−δ were investigated and correlated to the properties of the materials.

Structure and Morphology
Figure 1A shows the powder X-ray diffraction (XRD) data of CeO 2 and MO x /CeO 2 (M = Co, Ni, Cu, respectively) precursors.For the CeO 2 and MO x /CeO 2 samples, the major Bragg diffraction peaks at 28.5 • , 33.1 • and 47.5 • are clearly resolved and attributed to the diffractions from the (111), (200) and (220) planes of the cubic fluorite-type ceria (JCPDS file 34-0394), respectively.The weak diffraction peaks are associated with isolated metal oxides and indexed to CuO (JCPDS file , NiO (JCPDS file 78-0643) and Co 3 O 4 (JCPDS file 43-1003) (Figure 1A).The diffraction peaks of CeO 2 in the MO x /CeO 2 catalysts slightly shift to greater 2θ angles than those of bare CeO 2 , revealing that small portions of the M +δ ions in the loaded MO x have diffused into the ceria matrix and partially substituted cerium cations (radius of Ce 4+ = 0.084 nm) [1,25,26].Indeed, the metal doping induces lattice distortion and shrinkage of the CeO 2 lattice to different levels (Table 1) due to the smaller radii of the doping cations (R(Co 2+ ) = 0.074 nm and R(Co 3+ ) = 0.063 nm, R(Ni 2+ ) = 0.072 nm and R(Cu 2+ ) = 0.073 nm), but the changes of lattice parameters of CeO 2 for the samples are insignificant because of the low doping level.
The calculated mean crystalline sizes of Co 3 O 4 , NiO, CuO and CeO 2 are listed in Table 1.It is found that NiO possesses the smallest crystallite size, while CuO the largest.The CeO 2 supports exhibit different crystallite sizes even though the MO x are loaded at similar amounts.The varying crystallite sizes of MO x and CeO 2 are due to their different MO x -CeO 2 interaction and the induced lattice distortion.After H2 reduction treatments, metal phases of Ni, Co and Cu emerge in the XRD patterns of the reduced M/CeO2−δ catalysts (Figure 1B), while the MOx diffraction peaks vanish since the crystalline MOx species are converted to corresponding metals.Compared to Co and Cu, the Ni species display relatively weaker diffraction, suggesting that Ni might be highly dispersed on the surface of CeO2−δ with a smaller crystallite size (Table 1).We suppose that the smaller crystallite sizes of metals relative to their metal oxides (Figure 2) are induced by the H2 reduction treatment because the M-CeO2−δ would inherit the strong interaction between MOx and the CeO2.The proportional reliance of metal crystallite sizes on their corresponding metal oxides is further confirmed by the crystallite size changes of Ni/CeO2−δ and NiO/CeO2 samples (Figure S1 in Supplementary Materials), where the Metal loading amount was determined by ICP-OES.D(CeO 2 ) and D(CeO 2−δ ) are the crystalline size of CeO 2 in the MO x /CeO 2 and M/CeO 2−δ ; D(MO x ) and D(M) denote the crystallite size of metal oxide and metal calculated using the Scherrer equation.
After H 2 reduction treatments, metal phases of Ni, Co and Cu emerge in the XRD patterns of the reduced M/CeO 2−δ catalysts (Figure 1B), while the MO x diffraction peaks vanish since the crystalline MO x species are converted to corresponding metals.Compared to Co and Cu, the Ni species display relatively weaker diffraction, suggesting that Ni might be highly dispersed on the surface of CeO 2−δ with a smaller crystallite size (Table 1).We suppose that the smaller crystallite sizes of metals relative to their metal oxides (Figure 2) are induced by the H 2 reduction treatment because the M-CeO 2−δ would inherit the strong interaction between MO x and the CeO 2 .The proportional reliance of metal crystallite sizes on their corresponding metal oxides is further confirmed by the crystallite size changes of Ni/CeO 2−δ and NiO/CeO 2 samples (Figure S1 in Supplementary Materials), where the strong dependence of the Ni 0 crystallite sizes on their oxide precursors showed a linear correlation with the Ni loading up to 20 wt %.
strong dependence of the Ni 0 crystallite sizes on their oxide precursors showed a linear correlation with the Ni loading up to 20 wt %.As shown in Table 1, the lattice parameters of ceria in the M/CeO2−δ catalysts appear quite similar, but become larger than those of MOx/CeO2 samples.The increasing crystallite sizes of ceria in M/CeO2−δ catalysts may be induced by comprehensive factors, including a fraction of Ce 4+ cations being reduced to Ce 3+ of a larger ionic radius (RCe(III) = 0.102 nm [27]) that expands the ceria lattice parameter, the thermal sintering of ceria and the metal exiting from ceria matrices during the reduction process.
Figure 3 shows the transition electron microscopy (TEM) images of reduced the M/CeO2−δ catalysts.It is found that most CeO2 crystallites observed in the M/CeO2−δ samples are of spherical or irregular polyhedral morphologies, independent of the deposited metals.From the TEM images, it is difficult to distinguish the metal particles because the sizes of metals are quite similar to the CeO2 support (shown in Table 1) [25].In addition, a larger CeO2−δ particle is seen in the Co/CeO2−δ catalyst compared with Ni/CeO2−δ and Cu/CeO2−δ, which is likely associated with the slight sintering of the ceria support during the reduction at higher temperature.

Reduction Behaviour of the MOx/CeO2 Catalyst Precursors
Figure 4 shows the hydrogen temperature-programmed reduction (H2-TPR) profiles of MO/CeO2 samples.The TPR profiles at the temperature below 400 °C are due to the reduction of supported MOx in the three MOx/CeO2 samples.For the CuO/CeO2 sample, the strong peak centred at 160 °C is attributed to the reduction of isolated CuO species with weak CuO-CeO2 interaction, while the shoulder peak located at 130 °C is due to the reduction of adsorbed O species on Ovac sites As shown in Table 1, the lattice parameters of ceria in the M/CeO 2−δ catalysts appear quite similar, but become larger than those of MO x /CeO 2 samples.The increasing crystallite sizes of ceria in M/CeO 2−δ catalysts may be induced by comprehensive factors, including a fraction of Ce 4+ cations being reduced to Ce 3+ of a larger ionic radius (R Ce(III) = 0.102 nm [27]) that expands the ceria lattice parameter, the thermal sintering of ceria and the metal exiting from ceria matrices during the reduction process.
Figure 3 shows the transition electron microscopy (TEM) images of reduced the M/CeO 2−δ catalysts.It is found that most CeO 2 crystallites observed in the M/CeO 2−δ samples are of spherical or irregular polyhedral morphologies, independent of the deposited metals.From the TEM images, it is difficult to distinguish the metal particles because the sizes of metals are quite similar to the CeO 2 support (shown in Table 1) [25].In addition, a larger CeO 2−δ particle is seen in the Co/CeO 2−δ catalyst compared with Ni/CeO 2−δ and Cu/CeO 2−δ , which is likely associated with the slight sintering of the ceria support during the reduction at higher temperature.
strong dependence of the Ni 0 crystallite sizes on their oxide precursors showed a linear correlation with the Ni loading up to 20 wt %.As shown in Table 1, the lattice parameters of ceria in the M/CeO2−δ catalysts appear quite similar, but become larger than those of MOx/CeO2 samples.The increasing crystallite sizes of ceria in M/CeO2−δ catalysts may be induced by comprehensive factors, including a fraction of Ce 4+ cations being reduced to Ce 3+ of a larger ionic radius (RCe(III) = 0.102 nm [27]) that expands the ceria lattice parameter, the thermal sintering of ceria and the metal exiting from ceria matrices during the reduction process.
Figure 3 shows the transition electron microscopy (TEM) images of reduced the M/CeO2−δ catalysts.It is found that most CeO2 crystallites observed in the M/CeO2−δ samples are of spherical or irregular polyhedral morphologies, independent of the deposited metals.From the TEM images, it is difficult to distinguish the metal particles because the sizes of metals are quite similar to the CeO2 support (shown in Table 1) [25].In addition, a larger CeO2−δ particle is seen in the Co/CeO2−δ catalyst compared with Ni/CeO2−δ and Cu/CeO2−δ, which is likely associated with the slight sintering of the ceria support during the reduction at higher temperature.

Reduction Behaviour of the MO x /CeO 2 Catalyst Precursors
Figure 4 shows the hydrogen temperature-programmed reduction (H 2 -TPR) profiles of MO/CeO 2 samples.The TPR profiles at the temperature below 400 • C are due to the reduction of supported MO x in the three MO x /CeO 2 samples.For the CuO/CeO 2 sample, the strong peak centred at 160 • C is attributed to the reduction of isolated CuO species with weak CuO-CeO 2 interaction, while the shoulder peak located at 130 • C is due to the reduction of adsorbed O species on O vac sites and the highly dispersed CuO with strong CuO-CeO 2 interaction [26].For the NiO/CeO 2 sample, the reduction of surface adsorbed O species starts from an even lower temperature (~100 • C) [28] with a sharp reduction peak at 195 • C, which can be related to the reduction of highly dispersed NiO on CeO 2 [28].The main TPR peak at 260 • C is assigned to the reduction of NiO clusters aggregated on the CeO 2 surface [1].On the Co 3 O 4 /CeO 2 sample, three distinct TPR peaks are observed, which could be ascribed to the reductions of Co [29,30] and isolated CoO species (380 • C), respectively [25].It is worth noting that the CeO 2 reduction occurs at a lower temperature for Co 3 O 4 /CeO 2 relative to other samples, because CoO x can enhance H 2 spillover from the reduced Co species towards CeO 2 , which decreases the CeO 2 reduction temperature [31].The different interaction between the MO x and CeO 2 , as indicated by their distinct reduction behaviours of the MO x /CeO 2 samples, would determine the dispersion of the corresponding metals on CeO 2−δ and thus affect their ability in hydrogen activation (see Section 2.4 for the H 2 -TPD analysis).
Catalysts 2017, 7, 272 5 of 15 and the highly dispersed CuO with strong CuO-CeO2 interaction [26].For the NiO/CeO2 sample, the reduction of surface adsorbed O species starts from an even lower temperature (~100 °C) [28] with a sharp reduction peak at 195 °C, which can be related to the reduction of highly dispersed NiO on CeO2 [28].The main TPR peak at 260 °C is assigned to the reduction of NiO clusters aggregated on the CeO2 surface [1].On the Co3O4/CeO2 sample, three distinct TPR peaks are observed, which could be ascribed to the reductions of Co3O4 → CoO (220 °C), CoO → Co (270 °C) [29,30] and isolated CoO species (380 °C), respectively [25].It is worth noting that the CeO2 reduction occurs at a lower temperature for Co3O4/CeO2 relative to other samples, because CoOx can enhance H2 spillover from the reduced Co species towards CeO2, which decreases the CeO2 reduction temperature [31].The different interaction between the MOx and CeO2, as indicated by their distinct reduction behaviours of the MOx/CeO2 samples, would determine the dispersion of the corresponding metals on CeO2−δ and thus affect their ability in hydrogen activation (see Section 2.4 for the H2-TPD analysis).The two broad reduction peaks appearing at 530 and 705 °C for CeO2 are attributed to the respective reductions of surface oxygen species and the bulk lattice oxygen within ceria [32].However, for the three MOx/CeO2 samples, their reduction peaks due to the removal of lattice oxygen from the CeO2 support are displayed in a similar temperature region (600~800 °C), suggesting that the involved MOx species slightly impact the reducibility of the lattice oxygen of CeO2.In contrast, the reduction of surface oxygen (~530 °C) is absent for the MOx/CeO2 since both the strong M-ceria interaction and the H2 spillover from metal onto the vicinity surface oxygen species can improve the surface oxygen removal from the CeO2 support [33].In other words, the metal on the support can catalyse the ceria reduction

Surface Structural Analysis
Raman spectroscopy is employed to investigate the surface structure of the MOx/CeO2 precursors and the M/CeO2−δ catalysts.As shown in Figure 5A, the strong vibration mode (F2g, ~466 cm −1 ), due to the symmetrical stretching vibration of Ce-O bonds, dominates the Raman spectrum of CeO2 [34].Upon metal loading, the F2g modes red shift to 454 cm −1 with peak broadening due to the strong interaction between MOx and CeO2 that gives rise to distortion of the CeO2−δ lattice and creates electron-rich Ovac for maintaining the system charge neutrality [34].Besides the F2g band, two broadening bands within the 600~700 cm −1 and 1100~1250 cm −1 regions are observed on the MOx/CeO2 samples, which can be attributed to the Frenkel defect-induced modes (D band) and the second order longitudinal optical mode (2LO), respectively [35].The D and 2LO bands are related to the Ovac that originated from reducing Ce 4+ to Ce 3+ .Compared to CeO2, MOx/CeO2 samples show broader and stronger vibrations in the D bands and 2LO modes, suggesting that the loading of MOx species facilitates the creation of Ovac thanks to metal substitution of lattice Ce, which is in good The two broad reduction peaks appearing at 530 and 705 • C for CeO 2 are attributed to the respective reductions of surface oxygen species and the bulk lattice oxygen within ceria [32].However, for the three MO x /CeO 2 samples, their reduction peaks due to the removal of lattice oxygen from the CeO 2 support are displayed in a similar temperature region (600~800 • C), suggesting that the involved MO x species slightly impact the reducibility of the lattice oxygen of CeO 2 .In contrast, the reduction of surface oxygen (~530 • C) is absent for the MO x /CeO 2 since both the strong M-ceria interaction and the H 2 spillover from metal onto the vicinity surface oxygen species can improve the surface oxygen removal from the CeO 2 support [33].In other words, the metal on the support can catalyse the ceria reduction.

Surface Structural Analysis
Raman spectroscopy is employed to investigate the surface structure of the MO x /CeO 2 precursors and the M/CeO 2−δ catalysts.As shown in Figure 5A, the strong vibration mode (F 2g , ~466 cm −1 ), due to the symmetrical stretching vibration of Ce-O bonds, dominates the Raman spectrum of CeO 2 [34].Upon metal loading, the F 2g modes red shift to 454 cm −1 with peak broadening due to the strong interaction between MO x and CeO 2 that gives rise to distortion of the CeO 2−δ lattice and creates electron-rich O vac for maintaining the system charge neutrality [34].Besides the F 2g band, two broadening bands within the 600~700 cm −1 and 1100~1250 cm −1 regions are observed on the MO x /CeO 2 samples, which can be attributed to the Frenkel defect-induced modes (D band) and the second order longitudinal optical mode (2LO), respectively [35].The D and 2LO bands are related to the O vac that originated from reducing Ce 4+ to Ce 3+ .Compared to CeO 2 , MO x /CeO 2 samples show broader and stronger vibrations in the D bands and 2LO modes, suggesting that the loading of MO x species facilitates the creation of O vac thanks to metal substitution of lattice Ce, which is in good agreement with the discussion of XRD results.The quasi density of O vac in ceria due to metal incorporation is tentatively quantified through the relative ratio of (I D + I 2LO )/I F2g , as inserted in Figure 5A.The (I D + I 2LO )/I F2g in the MO x /CeO 2 samples is more than three-fold that in bare CeO 2 , namely, the O vac density descends in the order of NiO/CeO 2 > CuO/CeO 2 > Co 3 O 4 /CeO 2 >> CeO 2 .The results clearly reflect that the O vac concentration created by the MO x loading highly depends on the metal species, likely due to their different ionic radius and the oxidation state of the dopants [24].
Catalysts 2017, 7, 272 6 of 15 agreement with the discussion of XRD results.The quasi density of Ovac in ceria due to metal incorporation is tentatively quantified through the relative ratio of (ID + I2LO)/IF2g, as inserted in Figure 5A.The (ID + I2LO)/IF2g in the MOx/CeO2 samples is more than three-fold that in bare CeO2, namely, the Ovac density descends in the order of NiO/CeO2 > CuO/CeO2 > Co3O4/CeO2 >> CeO2.The results clearly reflect that the Ovac concentration created by the MOx loading highly depends on the metal species, likely due to their different ionic radius and the oxidation state of the dopants [24].The modes of F2g, D band and 2LO can be clearly identified in the reduced M/CeO2−δ catalysts and CeO2−δ (Figure 5B) but much weaker than those of the MOx/CeO2 precursors, in particular the Ovac-related D and 2LO bands.Moreover, the F2g modes of M/CeO2−δ catalysts further red shift from 454 cm −1 to 444 cm −1 .The relative changes of Raman signals for the M/CeO2−δ catalysts to their precursors are the consequence of their decreased structure symmetry.However, the (ID + I2LO)/IF2g ratios of the M/CeO2−δ catalysts and CeO2−δ are much greater than those in their corresponding oxides (inserted in Figure 5A,B), suggesting that the H2 reduction significantly promotes Ovac density.It is worth noting that Ovac concentration of the metal-loaded catalysts is enhanced approximately nineto ten-fold that of CeO2−δ, further confirming that metal loading can dramatically improve the Ovac formation.The results are in good accordance with the TPR characterization.We also note that the Ovac density variation of the M/CeO2−δ catalysts follows the descending order of Ni/CeO2−δ > Cu/CeO2−δ > Co/CeO2−δ >> CeO2−δ, which inherits the Ovac concentration order of their precursor oxides.

Hydrogen Activation on the M/CeO2−δ Catalysts
Hydrogen temperature programmed desorption (H2-TPD) is a powerful tool to gain insights into hydrogen activation and metal-support interaction on the catalysts.As shown in Figure 6, the H2-TPD profiles of the M/CeO2−δ are deconvoluted into three peaks.The α peaks for the M/CeO2−δ catalysts appear within a very similar temperature region, which is identical to the temperature range of the sole H2-TPD peak on CeO2−δ.Such α peaks are therefore assigned to the desorption of H2 up-taking surface Ovac sites in CeO2−δ [27].The difference in the integrated areas (Table 2) reveals that the amount of Ovac of the M/CeO2−δ is dependent on the metallic species, and the Ni/CeO2−δ and Cu/CeO2−δ possess a similar amount of Ovac, but much larger than those of the Co/CeO2−δ and CeO2−δ supports.The results are consistent with the Raman and TPR analyses.The modes of F 2g , D band and 2LO can be clearly identified in the reduced M/CeO 2−δ catalysts and CeO 2−δ (Figure 5B) but much weaker than those of the MO x /CeO 2 precursors, in particular the O vac -related D and 2LO bands.Moreover, the F 2g modes of M/CeO 2−δ catalysts further red shift from 454 cm −1 to 444 cm −1 .The relative changes of Raman signals for the M/CeO 2−δ catalysts to their precursors are the consequence of their decreased structure symmetry.However, the (I D + I 2LO )/I F2g ratios of the M/CeO 2−δ catalysts and CeO 2−δ are much greater than those in their corresponding oxides (inserted in Figure 5A,B), suggesting that the H 2 reduction significantly promotes O vac density.It is worth noting that O vac concentration of the metal-loaded catalysts is enhanced approximately nine-to ten-fold that of CeO 2−δ , further confirming that metal loading can dramatically improve the O vac formation.The results are in good accordance with the TPR characterization.We also note that the O vac density variation of the M/CeO 2−δ catalysts follows the descending order of Ni/CeO 2−δ > Cu/CeO 2−δ > Co/CeO 2−δ >> CeO 2−δ , which inherits the O vac concentration order of their precursor oxides.

Hydrogen Activation on the M/CeO 2−δ Catalysts
Hydrogen temperature programmed desorption (H 2 -TPD) is a powerful tool to gain insights into hydrogen activation and metal-support interaction on the catalysts.As shown in Figure 6, the H 2 -TPD profiles of the M/CeO 2−δ are deconvoluted into three peaks.The α peaks for the M/CeO 2−δ catalysts appear within a very similar temperature region, which is identical to the temperature range of the sole H 2 -TPD peak on CeO 2−δ .Such α peaks are therefore assigned to the desorption of H 2 up-taking surface O vac sites in CeO 2−δ [27].The difference in the integrated areas (Table 2) reveals that the amount of O vac of the M/CeO 2−δ is dependent on the metallic species, and the Ni/CeO 2−δ and Cu/CeO 2−δ possess a similar amount of O vac , but much larger than those of the Co/CeO 2−δ and CeO 2−δ supports.The results are consistent with the Raman and TPR analyses.Note: DM denotes the metal dispersion, calculated via DM = (Msurf/Mtotal), where Msurf and Mtotal are the amount of surface-exposed metal and total metal loading, respectively.Assuming H/M = 1 and Msurf = 2 × (H2 desorption from metal surface), Msurf can be calculated from the amount of hydrogen (β and γ peaks) uptaken on the metal [1].
The β and γ peaks located in 100-250 °C are due to H2 desorption from the corresponding metal species.The β peaks can be ascribed to the dissociative H2 adsorbed on the highly dispersed metal species with a large density of surface defects, which can enhance the hydrogen diffusion and reduce the energy barrier for hydrogen dissociation [36].The γ peaks are associated with chemisorbed hydrogen on isolated metal clusters [37].The quantitative analyses of peak β and γ (Table 2) clearly show that Co/CeO2−δ and Ni/CeO2−δ catalysts possess comparable amounts of β-H2 and γ-H2, which are more than those of Cu/CeO2−δ.The low uptake of β and γ hydrogen on Cu/CeO2−δ may be due to the weak capability of Cu for hydrogen activation caused by its closed d 10 orbitals [38].
According to the assignment of the H2-TPD peaks, the metal dispersion (DM) can be quantified using only the metal-related hydrogen activation (sum of the β-and γ-H2), which is different from the overall H2 activation amount of the catalysts.As listed in Table 2.The DM of the M/CeO2−δ catalysts is found to be decreasing in the order of Ni/CeO2−δ > Co/CeO2−δ > Cu/CeO2−δ.

Catalytic Hydrogenation of Maleic Anhydride
Figure 7A presents the MA conversion (XMA) along the hydrogenation courses on M/CeO2−δ catalysts in a batch reactor at 210 °C and hydrogen pressure of 5.0 MPa.Among the examined catalysts, Ni/CeO2−δ is found to possess the highest activity of MA conversion, reaching ~100% MA conversion in 60 min.However, the MA conversion on Cu/CeO2−δ and Co/CeO2−δ in the same period only achieves 32% and 70%, respectively.The MA transformation completes in 3 h on Co/CeO2−δ and 4 h on Cu/CeO2−δ.Moreover, the turnover frequencies of MA transformed to SA (TOFMA → SA) over M/CeO2−δ catalysts are also estimated.The Ni/CeO2−δ catalyst records the highest TOFMA → SA of 61.7  Note: D M denotes the metal dispersion, calculated via D M = (M surf /M total ), where M surf and M total are the amount of surface-exposed metal and total metal loading, respectively.Assuming H/M = 1 and M surf = 2 × (H 2 desorption from metal surface), M surf can be calculated from the amount of hydrogen (β and γ peaks) uptaken on the metal [1].
The β and γ peaks located in 100-250 • C are due to H 2 desorption from the corresponding metal species.The β peaks can be ascribed to the dissociative H 2 adsorbed on the highly dispersed metal species with a large density of surface defects, which can enhance the hydrogen diffusion and reduce the energy barrier for hydrogen dissociation [36].The γ peaks are associated with chemisorbed hydrogen on isolated metal clusters [37].The quantitative analyses of peak β and γ (Table 2) clearly show that Co/CeO 2−δ and Ni/CeO 2−δ catalysts possess comparable amounts of β-H 2 and γ-H 2 , which are more than those of Cu/CeO 2−δ .The low uptake of β and γ hydrogen on Cu/CeO 2−δ may be due to the weak capability of Cu for hydrogen activation caused by its closed d 10 orbitals [38].
According to the assignment of the H 2 -TPD peaks, the metal dispersion (D M ) can be quantified using only the metal-related hydrogen activation (sum of the βand γ-H 2 ), which is different from the overall H 2 activation amount of the catalysts.As listed in Table 2.The D M of the M/CeO 2−δ catalysts is found to be decreasing in the order of Ni/CeO 2−δ > Co/CeO 2−δ > Cu/CeO 2−δ .

Catalytic Hydrogenation of Maleic Anhydride
Figure 7A presents the MA conversion (X MA ) along the hydrogenation courses on M/CeO 2−δ catalysts in a batch reactor at 210 • C and hydrogen pressure of 5.0 MPa.Among the examined catalysts, Ni/CeO 2−δ is found to possess the highest activity of MA conversion, reaching ~100% MA conversion in 60 min.However, the MA conversion on Cu/CeO 2−δ and Co/CeO 2−δ in the same period only achieves 32% and 70%, respectively.The MA transformation completes in 3 h on Co/CeO 2−δ and 4 h on Cu/CeO 2−δ .Moreover, the turnover frequencies of MA transformed to SA (TOF MA→SA ) over M/CeO 2−δ catalysts are also estimated.The Ni/CeO 2−δ catalyst records the highest TOF MA→SA of 61.7 min −1 , nearly twice that of the Co/CeO 2−δ catalyst (TOF MA→SA = 31.2min −1 ) and five-fold that of Cu/CeO 2−δ (TOF MA→SA = 13.2 min −1 ).The considerable difference of the MA conversion on the M/CeO 2−δ catalysts cannot only be attributed to the particle size or metal dispersion effects because the activity difference is much more significant.We speculate that there exist other factors, such as the intrinsic properties (e.g., electronic configuration) of individual metals, M-support interaction and O vac amount, to influence the catalytic performance synergistically.
Figure 7B presents the curves of −ln(1 − X MA ) versus time within the first 60 min, which are fitted according to the first order kinetic law in terms of MA conversion on the metal-based catalysts [17].The linear kinetic plots over Ni/CeO 2−δ and Co/CeO 2−δ catalysts suggest that the C=C hydrogenation follows the quasi-first order reaction in terms of MA conversion, while the nonlinear kinetic plot on Cu/CeO 2−δ reflects that its initial hydrogenation follows a complicated kinetic mechanism.The MA hydrogenation rate coefficients (k 1 ) on the Ni/CeO 2−δ and Co/CeO 2−δ catalysts are determined according to the gradients of their linear plots and summarised in Table 2.The k for Ni/CeO 2−δ is much greater than that for Co/CeO 2−δ , indicating that metallic Ni/CeO 2−δ is more reactive than Co/CeO 2−δ .
In principle, the electronic configurations of the active metals are profound in the activation of MA for the hydrogenation reaction.The previous DFT simulation demonstrated that the adsorption and activation of MA on the metal surface are primarily controlled by the back-donation of the electron from the metal onto the anti-bonding π* orbital of maleic anhydride [39].Therefore, the metallic Cu with a full-filled 3d 10 electronic configuration without the unpaired electron possesses weaker capability of electron-donation to the MA molecule.In contrast, the unpaired electrons of Co (3d 7 ) and Ni (Ni 3d 8 ) would favour the coordination and activation of MA, thus exhibit much higher activity.
Besides the intrinsic properties, the O vac is regarded as another factor to enhance the C=C hydrogenation because the O vac can enrich the electron density of active metals, which promotes the electron-donating ability and promotes H 2 dissociation [1,40].In our cases, the Ni/CeO 2−δ catalyst has richer O vac and, thus, may donate more electrons to metallic Ni in comparison to Co/CeO 2−δ .Further, the smaller particle size of Ni (XRD section) also strengthens the short-range electron transferring from O vac onto the adjacent metal clusters.Hence, the abundant O vac and the direct or indirect O vac → M electronic donation are responsible for the fact that Ni/CeO 2−δ shows higher activity in MA hydrogenation than Co/CeO 2−δ .However, it is noted that Ni/CeO 2−δ and Cu/CeO 2−δ catalysts have a similar O vac amount, but different activity in MA hydrogenation, suggesting that the intrinsic electron structure of active metal dominates the MA hydrogenation.Cu/CeO2−δ (TOFMA → SA = 13.2 min −1 ).The considerable difference of the MA conversion on the M/CeO2−δ catalysts cannot only be attributed to the particle size or metal dispersion effects because the activity difference is much more significant.We speculate that there exist other factors, such as the intrinsic properties (e.g., electronic configuration) of individual metals, M-support interaction and Ovac amount, to influence the catalytic performance synergistically.Figure 7B presents the curves of −ln(1 − XMA) versus time within the first 60 min, to the first order kinetic law in terms of MA conversion on the metal-based catalysts [17].The linear kinetic plots over Ni/CeO2−δ and Co/CeO2−δ catalysts suggest that the C=C hydrogenation follows the quasi-first order reaction in terms of MA conversion, while the nonlinear kinetic plot on Cu/CeO2−δ reflects that its initial hydrogenation follows a complicated kinetic mechanism.The MA hydrogenation rate coefficients (k1) on the Ni/CeO2−δ and Co/CeO2−δ catalysts are determined according to the gradients of their linear plots and summarised in Table 2.The k for Ni/CeO2−δ is much greater than that for Co/CeO2−δ, indicating that metallic Ni/CeO2−δ is more reactive than Co/CeO2−δ.
In principle, the electronic configurations of the active metals are profound in the activation of MA for the hydrogenation reaction.The previous DFT simulation demonstrated that the adsorption and activation of MA on the metal surface are primarily controlled by the back-donation of the electron from the metal onto the anti-bonding π* orbital of maleic anhydride [39].Therefore, the metallic Cu with a full-filled 3d 10 electronic configuration without the unpaired electron possesses weaker capability of electron-donation to the MA molecule.In contrast, the unpaired electrons of Co (3d 7 ) and Ni (Ni 3d 8 ) would favour the coordination and activation of MA, thus exhibit much higher activity.
Besides the intrinsic properties, the Ovac is regarded as another factor to enhance the C=C hydrogenation because the Ovac can enrich the electron density of active metals, which promotes the electron-donating ability and promotes H2 dissociation [1,40].In our cases, the Ni/CeO2−δ catalyst has richer Ovac and, thus, may donate more electrons to metallic Ni in comparison to Co/CeO2−δ.Further, the smaller particle size of Ni (XRD section) also strengthens the short-range electron transferring from Ovac onto the adjacent metal clusters.Hence, the abundant Ovac and the direct or indirect Ovac → M electronic donation are responsible for the fact that Ni/CeO2−δ shows higher activity in MA hydrogenation than Co/CeO2−δ.However, it is noted that Ni/CeO2−δ and Cu/CeO2−δ catalysts have a similar Ovac amount, but different activity in MA hydrogenation, suggesting that the intrinsic electron structure of active metal dominates the MA hydrogenation. Figure 8A shows the SA selectivity of the M/CeO2−δ catalysts in the prolonged MAH for 8 h.The Cu/CeO2−δ and CeO2−δ catalysts exhibit 100% selectivity to SA in the 8-h continuous MAH, without GBL observed in the products, indicating they are inert for SA hydrogenolysis to GBL.The TOFSA → GBL Figure 8A shows the SA selectivity of the M/CeO 2−δ catalysts in the prolonged MAH for 8 h.The Cu/CeO 2−δ and CeO 2−δ catalysts exhibit 100% selectivity to SA in the 8-h continuous MAH, without GBL observed in the products, indicating they are inert for SA hydrogenolysis to GBL.The TOF SA→GBL values over Ni/CeO 2−δ , Co/CeO 2−δ and Cu/CeO 2−δ are 3.08 min −1 , 0.87 min −1 and 0, respectively, indicating that Ni/CeO 2−δ shows the highest activity in the SA → GBL transformation.It is worth noting that GBL was generated only if MA fully converts to SA on the Ni/CeO 2−δ , as evidenced by the SA selectivity over Ni/CeO 2−δ , which remains 100% in the initial 40 min MAH time-on-stream before declining due to the following GBL generation.The generation of GBL on Co/CeO 2−δ also commences while the conversion of MA to SA is completing.GBL yields (Figure 8B) reach ~25% on Ni/CeO 2−δ and ~5% on Co/CeO 2−δ in 6 h, which are much slower than the MA → SA transformation.The results suggest that MAH on the M/CeO 2−δ is a multiple-step process: MAH to SA occurs first and rapidly, while GBL generates from the hydrogenolysis of SA only if no MA exists in the system.A similar phenomenon has been observed on the TiO 2 -supported Ni-based catalysts by Torres et al. [17].
Owing to the complexity of the MAH reaction, its hydrogenation mechanism has not been literature, in particular for the M/CeO 2−δ systems, although the theoretical and experimental understanding of H 2 adsorption, activation and cycling on CeO 2 and noble metals was attempted for a variety of redox reactions [38,41].Even fewer investigations of MA adsorption and hydrogenation over noble metals are openly accessible [39,42].However, on the basis of our experimental characterization data and the reported fundamental data regarding MAH reaction thermodynamics and kinetics, we may explicitly rationalize the stepwise MAH with a distinct reaction rate and selectivity on the M/CeO 2−δ catalyst systems.Owing to the complexity of the MAH reaction, its hydrogenation mechanism has not been explicated in the literature, in particular for the M/CeO2−δ systems, although the theoretical and experimental understanding of H2 adsorption, activation and cycling on CeO2 and noble metals was attempted for a variety of redox reactions [38,41].Even fewer investigations of MA adsorption and hydrogenation over noble metals are openly accessible [39,42].However, on the basis of our experimental characterization data and the reported fundamental data regarding MAH reaction thermodynamics and kinetics, we may explicitly rationalize the stepwise MAH with a distinct reaction rate and selectivity on the M/CeO2−δ catalyst systems.The preferential hydrogenation of C=C bonds rather than C=O groups in the initial MAH stage can be implicitly interpreted from kinetics and thermodynamic perspectives.The possible reaction routes for MAH to SA and GBL can be written as: Here, the free enthalpy change data of ΔHMA → SA and ΔHSA → GBL are adopted from the literature [43], while the ΔHMA → GBL of Global Reaction (3) is calculated applying Hess's law to Reactions (1) and (2).Since the molecule kinetic characteristics determine the rate and selectivity, the MA → SA transformation would be more favourable and faster than the SA → GBL and MA → GBL transformations that involve more H2 molecules than the former reaction.This interpretation is well evidenced by the fact that the GBL generation initiates only if complete MA → SA transformation on the M/CeO2−δ catalyst systems occurs, which also reveals that Reaction (3) is least viable on all of the catalysts investigated.Indeed, it was claimed that the C=C group is kinetically more active than the C=O group [44].
As for the distinct selectivity of the chosen M/CeO2−δ systems, we suppose that their different surface electron configurations of the catalysts, particularly the metal d orbitals, play essential roles, The preferential hydrogenation of C=C bonds rather than C=O groups in the initial MAH stage can be implicitly interpreted from kinetics and thermodynamic perspectives.The possible reaction routes for MAH to SA and GBL can be written as: Here, the free enthalpy change data of ∆H MA→SA and ∆H SA→GBL are adopted from the literature [43], while the ∆H MA→GBL of Global Reaction (3) is calculated applying Hess's law to Reactions (1) and (2).Since the molecule kinetic characteristics determine the rate and selectivity, the MA → SA transformation would be more favourable and faster than the SA → GBL and MA → GBL transformations that involve more H 2 molecules than the former reaction.This interpretation is well evidenced by the fact that the GBL generation initiates only if complete MA → SA transformation on the M/CeO 2−δ catalyst systems occurs, which also reveals that Reaction (3) is least viable on all of the catalysts investigated.Indeed, it was claimed that the C=C group is kinetically more active than the C=O group [44].
As for the distinct selectivity of the chosen M/CeO 2−δ systems, we suppose that their different surface electron configurations of the catalysts, particularly the metal d orbitals, play essential roles, because the overall hydrogenation reaction is usually determined by H 2 activation on the catalysts [42,45].This argument is also evidenced by our TPD data (Figure 6 and Table 2) where the Cu/CeO 2−δ displays a smaller amount of β and γ hydrogen, namely the weaker ability to activate hydrogen.For the transformation of SA → GBL, the adsorption of the C=O group was also considered to be of profound importance for the hydrogenolysis process [46].The closed d 10 shell of metal Cu is not beneficial to C=O hydrogenolysis due to its low interstitial electron density and weak energy for adsorption and activation of the C=O group [47].In contrast, metallic Ni and Co with partially filled d-orbitals are prone to coordinate with the C=O groups of organic adsorbates and thus facilitate C=O hydrogenolysis [47].
The beneficial contributions of O vac in C=O hydrogenolysis have been widely accepted [48,49].The O vac in the non-stoichiometric oxides, such as ZnO 1−x [50] and CeO 2−x [1], can polarise and activate the C=O group and thus reduce the energy barrier of C=O conversion to display reasonable hydrogenolysis selectivity.In this work, the amount of O vac of the Ni/CeO 2−δ catalyst is about 1.5-times that in the Co/CeO 2−δ catalyst, suggesting that Ni/CeO 2−δ catalyst can offer more active sites for C=O activation.However, the calculated TOF SA→GBL of Ni/CeO 2−δ is nearly triple that of Co/CeO 2−δ , which is larger than the relative O vac ratio between the two catalysts.Therefore, the electronic configuration of the active metal is the key factor to determine the catalytic hydrogenolysis of SA to GBL, while O vac can facilitate C=O conversion on Ni/CeO 2−δ due to the electron-donating and selective adsorption effects.O and citric acid (CA).purchased from the Sinopharm Chemical Reagent Co., Ltd.(Shanghai, China), were analytic grade and used as-purchased without pre-purification.

Ceria Support Preparation
Ceria supports were prepared by the sol-gel process.First, 5.00 g Ce(NO 3 ) 3 •6H 2 O were dissolved in 20 mL deionized water containing 6.56 g CA under rigorous stirring.The solution was heated in a water bath at 80 • C until a porous dry gel was obtained.During this process, the colourless solution became light yellow and finally brown.The obtained gel was then dried at 120 • C for 8 h to form a spongy material, which is subject to calcination for 3 h at 500 • C with a heating ramp of 3 • C/min.The resulting cerium oxide was used as the support to prepare the corresponding catalysts.

M/CeO 2−δ (M = Co, Ni and Cu) Catalysts Preparation
The M/CeO 2−δ (M = Co, Ni, Cu) catalysts (metal loading = 10 wt %) were prepared by an incipient wetness impregnation method as reported in our previous research paper [1].For the preparation of the Ni/CeO 2−δ catalyst, 1.01 g Ni(NO 3 ) 2 •6H 2 O were dissolved in 10 mL of the ethanol/DI water (1:1 volumetric ratio) solution, and then, 2.00 g of CeO 2 support were added into the mixed solution.The mixed solution was heated on a hotplate at 80 • C under rigorous stirring to evaporate the solvent before drying at 120 • C for 8 h in a convection oven.The obtained sample was calcined at 450 • C for 3 h in a muffle furnace with the heating ramp of 3 • C/min.The received oxide sample was reduced at 350 • C (heating ramp of 2 • C/min) in H 2 flow for 3 h to obtain the Ni/CeO 2−δ catalyst for MA hydrogenation.The Cu/CeO 2−δ and Co/CeO 2−δ catalysts were prepared using a similar procedure, but Cu/CeO 2−δ and Co/CeO 2−δ were reduced at 250 • C and 450 • C, respectively, which ensured that all of the transition metal species fully reduced at the optimum temperature as determined by H 2 -TPR.

Catalyst Characterizations
X-ray diffraction data were acquired by powder X-ray diffraction using a D8 Advance (Bruker, Billerica, MA, USA) (Cu K α1 radiation, λ = 0.15418 nm) equipped with a Vantec detector and Ni filter.Data were collected between 10 • and 80 • 2θ degrees via a continuous scan mode with a scanning rate at 4 • /min.The mean crystallite size oof the pertinent species were calculated employing the Scherrer equation: D = kλ/(βcosθ), where k is 0.9, β is the full width of the diffraction line at half of the miximum intensity and θ is the diffraction angle.The Ni loading of the catalysts was carried out by means of inductively-coupled plasma (ICP) spectroscopy with the iCAP 7400 ICP-OES (Thermo Fisher Scientific, Waltham, MA, USA) equipment.The Raman spectrum was recorded on a LabRAM HR Evolution Raman microscope with a laser wavelength of 532 nm (HORIBA, Tokyo, Japan) for surface characterization.The morphology of the materials was visualized using a JEOL JEM-2010 transmission electron microscope (TEM) (JEOL, Tokyo, Japan) operated at 120.0 kV.The samples were dispersed in ethanol assisted by the ultrasonic technique.
Hydrogen-temperature programed reduction (H 2 -TPR) experiments were performed using the Micromeritics AutoChem II 2950 system (Norcross, GA, USA) equipped with a thermal conductivity detector (TCD).Typically, 30 mg of catalyst were first degassed at 300 • C for 1 h with argon flushing.After cooling down to ambient temperature in argon, H 2 /Ar (10 vol % H 2 ) at a flow rate of 50 mL/min was passed through the catalysts, and the TPR profiles were recorded from 50 to 800 • C at a ramp of 10 • C/min.The output signal of the gas mixture was measured continuously using the TCD as a function of temperature.
Hydrogen temperature programmed desorption (H 2 -TPD) measurements were carried out on the same apparatus as TPR.In the H 2 -TPD test, 100 mg catalyst were first reduced in situ at 350 • C for 1 h in pure H 2 , followed by purging with Ar for 1 h at 360 • C to remove the excess hydrogen adsorbed on the surface and cooling down to 50 • C in Ar flow.Then, 10 vol % H 2 in Ar was injected until saturation at 50 • C and then purged with Ar until the baseline was stable, prior to carrying out TPD up to 600 • C at a heating rate of 10 • C/min.Ag 2 O was taken as the standard material to correlate the peak areas to volume data.

Catalyst Test
Liquid-phase hydrogenation of MA was carried out over M/CeO 2−δ catalysts in a batch reactor at 210 • C and 5 MPa.The reduced catalysts (0.1 g) and MA (4.9 g) were put in an autoclave (100 mL) containing THF (40 mL) solvent and purged five times with nitrogen to remove air before pressurizing up to 4.0 MPa using hydrogen.The diffusional limitation including internal and external diffusion were ruled out by performing experiments varying the particles size of the catalyst (0.425-0.250 mm) and adjusting the string rate of 500 rpm.Simultaneously, the hydrogen pressure was raised up to and maintained at 5.0 MPa.The reaction effluents were analysed using a 7890A gas chromatograph (Agilent, Santa Clara, CA, USA) equipped with an SE-52 (30 m × 0.32 mm × 1.0 µm) capillary column and a flame ionization detector (FID).The programmed temperature from 100 • C-120 • C with a ramp of 5 • C/min was used to ensure complete separation of individual components in the effluents.The temperatures of the injector and detector were set at 260 • C and 190 • C, respectively.

4 Note
: a(CeO2) and a(CeO2−δ) denote the lattice parameter of CeO2 in the MOx/CeO2 and reduced M/CeO2−δ, respectively, calculated using interplanar distance (d) and Miller indices (h k l) in the equation √ .Metal loading amount was determined by ICP-OES.D(CeO2) and D(CeO2−δ) are the crystalline size of CeO2 in the MOx/CeO2 and M/CeO2−δ; D(MOx) and D(M) denote the crystallite size of metal oxide and metal calculated using the Scherrer equation.

Figure 2 .
Figure 2. The dependence of metal crystallite sizes on those of metal oxides.

Figure 2 .
Figure 2. The dependence of metal crystallite sizes on those of metal oxides.

Figure 2 .
Figure 2. The dependence of metal crystallite sizes on those of metal oxides.

Figure 4 Figure 3 .
Figure4shows the hydrogen temperature-programmed reduction (H2-TPR) profiles of MO/CeO2 samples.The TPR profiles at the temperature below 400 °C are due to the reduction of supported MOx in the three MOx/CeO2 samples.For the CuO/CeO2 sample, the strong peak centred at 160 °C is attributed to the reduction of isolated CuO species with weak CuO-CeO2 interaction, while the shoulder peak located at 130 °C is due to the reduction of adsorbed O species on Ovac sites

Figure 5 .
Figure 5. Raman spectra of the MOx/CeO2 precursors (A) and the reduced M/CeO2−δ catalysts (B).The inserted zoom-in figures underline the existence of 2LO bands.

Figure 5 .
Figure 5. Raman spectra of the MO x /CeO 2 precursors (A) and the reduced M/CeO 2−δ catalysts (B).The inserted zoom-in figures underline the existence of 2LO bands.

Figure 7 .
Figure 7. (A) The conversion of maleic anhydride (MA) and (B) the kinetic curves of the M/CeO2−δ catalysts.

Figure 7 .
Figure 7. (A) The conversion of maleic anhydride (MA) and (B) the kinetic curves of the M/CeO 2−δ catalysts.

Table 1 .
The properties of the MOx/CeO2 precursors and corresponding M/CeO2−δ catalysts.

Table 1 .
The properties of the MO x /CeO 2 precursors and corresponding M/CeO 2−δ catalysts.

Table 2 .
H 2 uptake, metal dispersion and hydrogenation rate constants on the M/CeO 2−δ catalyst.
CeO2−δ shows the highest activity in the SA → GBL transformation.It is worth noting that GBL was generated only if MA fully converts to SA on the Ni/CeO2−δ, as evidenced by the SA selectivity over Ni/CeO2−δ, which remains 100% in the initial 40 min MAH time-on-stream before declining due to the following GBL generation.The generation of GBL on Co/CeO2−δ also commences while the conversion of MA to SA is completing.GBL yields (Figure8B) reach ~25% on Ni/CeO2−δ and ~5% on Co/CeO2−δ in 6 h, which are much slower than the MA → SA transformation.The results suggest that MAH on the M/CeO2−δ is a multiple-step process: MAH to SA occurs first and rapidly, while GBL generates from the hydrogenolysis of SA only if no MA exists in the system.A similar phenomenon has been observed on the TiO2-supported Ni-based catalysts by Torres et al.[17].