catalysts Nickel Supported on AlCeO 3 as a Highly Selective and Stable Catalyst for Hydrogen Production via the Glycerol Steam Reforming Reaction

: In this study, a critical comparison between two low metal (Ni) loading catalysts is presented, namely Ni / Al 2 O 3 and Ni / AlCeO 3 for the glycerol steam reforming (GSR) reaction. The surface and bulk properties of the catalysts were evaluated using a plethora of techniques, such as N 2 adsorption / desorption, Inductively Coupled Plasma Atomic Emission Spectroscopy (ICP–AES), X-ray Di ﬀ raction (XRD), X-ray Photoelectron Spectroscopy (XPS), Scanning Electron Microscopy / Energy Dispersive X-Ray Spectroscopy (SEM / EDX, Transmission Electron Microscopy (TEM), CO 2 and NH 3 – Temperature Programmed Desorption (TPD), and Temperature Programmed Reduction (H 2 –TPR). Carbon deposited on the catalyst’s surfaces was probed using Temperature Programmed Oxidation (TPO), SEM, and TEM. It is demonstrated that Ce-modiﬁcation of Al 2 O 3 induces an increase of the surface basicity and Ni dispersion. These features lead to a higher conversion of glycerol to gaseous products (60% to 80%), particularly H 2 and CO 2 , enhancement of WGS reaction, and a higher resistance to coke deposition. Allyl alcohol was found to be the main liquid product for the Ni / AlCeO 3 catalyst, the production of which ceases over 700 ◦ C. It is also highly signiﬁcant that the Ni / AlCeO 3 catalyst demonstrated stable values for H 2 yield (2.9–2.3) and selectivity (89–81%), in addition to CO 2 (75–67%) and CO (23–29%) selectivity during a (20 h) long time-on-stream study. Following the reaction, SEM / EDX and TEM analysis showed heavy coke deposition over the Ni / Al 2 O 3 catalyst, whereas for the Ni / AlCeO 3 catalyst TPO studies showed the formation of more defective coke, the latter being more easily oxidized.


Introduction
Biofuels are expected to play an important role in meeting the combined challenge of providing adequate energy supplies, while simultaneously combating the threat posed by climate change, with projections estimating that their use will grow from 1.3 million barrels of oil equivalent (BOE) in 2012 to 4.6 BOE in 2040 [1]. Although biodiesel had captured approximately 25% of total biofuel output in 2015 [2], the high costs associated with its production means that the sector remains uncompetitive, with favourable government policies underpinning much of its growth [3]. The industry is also facing the challenge of dealing, in a sustainable way, with the co-production of crude glycerol (C 3 H 8 O 3 ), which constitutes its major by-product and amounts to approximately 10 wt % of the oil undergoing transesterification [4].
However, glycerol can be converted into hydrogen, a fuel that finds a variety of present-day applications in transport, building and other industry sectors, while at the same time is expected to have a leading role in a future carbon-free energy mix [5]. Although different thermochemical processes may be used for the conversion of C 3 H 8 O 3 into H 2-e.g., aqueous phase reforming (APR), autothermal reforming (ATR) and super critical water reforming (SCWR)-the process that appears most promising is that of the steam reforming of glycerol (GSR) [6][7][8]. This is because GSR has a high H 2 production capacity per mol of C 3 H 8 O 3 reformed (Equation (1)) and is a mature industrial technology unlikely to require major technical adjustments in switching feedstocks [9,10]. The GSR (Equation (1)), a strongly endothermic reaction (∆H Θ = 123 kJ/mole), combines the decomposition of glycerol (Equation (2)) and the water-gas shift reaction (WGS, Equation (3)), but a number of parallel reactions that include methanation (Equation (4)), methane dry reforming (Equation (5)) and carbon formation reactions (Equations (6)- (9)) also take place [11,12]. According to the thermodynamic studies undertaken, the GSR should be undertaken at high temperature (>630 • C), high water to glycerol feed ratio (WGFR < 9:1, molar) and at atmospheric pressure [13,14].
CO + H 2 O ↔ CO 2 + H 2 , 2CO ↔ CO 2 + C, To be effective, the catalysts for use in the GSR should not promote C-O cleavage and CO or CO 2 hydrogenation, but they should favour the cleavage of the C-H, C-C and O-H bonds [15,16]. As Ni-based systems are known to be highly active in reforming reactions, most efforts currently found in the GSR related literature have focused on the development of such catalysts, using a variety of metal oxides (e.g., Al 2 O 3 , ZrO 2 , SiO 2 ) as supports [17][18][19][20]. The most commonly used support is Al 2 O 3 as it possesses a number of desirable properties, such as chemical and mechanical resistance under reaction conditions and a high specific surface area, which favours the dispersion of the active phase on the support [20,21]. However, there are two main disadvantages associated with the use of alumina; firstly, the Lewis type acid sites that it possesses promote acid-base-catalyzed reactions, which in turn favour the formation of a filamentous type of carbon, and secondly, it fosters the sintering of metallic particles [22,23]. For example, Chiodo et al. [24], using a low steam to glycerol ratio (3 mol/mol) and testing the stability at 800 • C for 20 h, reported rapid deactivation for a Ni/Al catalyst (30 wt % Ni). Cheng et al. [25], reported significant carbon deposition for a Ni/Al (15 wt % Ni) at 550 • C tested for approximately 4 h, even under excess steam conditions. Previous work by our group [15] showed severe deactivation for a Ni/Al (8 wt % Ni) during 20 h time-on-stream at 600 • C (WGFR of 9/1, molar), which was attributed to the high degree of crystallinity of the carbon deposits. Thus, much effort has been spent on overcoming the disadvantages associated with the use of alumina by inducing support-mediated promotional effects using different alkaline earth metals (e.g., Mg, Ca), lanthanide metals (e.g., Ce, La), and/or transition metals (e.g., Zr, Cu) as promoters for Ni/Al catalysts [26][27][28].
As is well understood, ceria can affect the surface acidity of alumina due to its higher point of zero charge (PZC) [29]. It also has the ability to act as an O 2 buffer, storing/releasing O 2 via the Ce 3+ /Ce 4+ redox couple in CeO 2 [30], and thus promote water dissociation and the water gas shift reaction [31]. Recent research also suggests that a [O δ− , δ + ] dipolar layer is created on the surface of Ni particles that can protect them against thermal sintering [32][33][34]. For the GSR, a handful of works exist where ceria was used as promoter to alumina for Ni-based catalysts [35][36][37][38]. The general conclusion from these papers is that the addition of CeO 2 as alumina modifier enhances the activity of the non-promoted catalyst [35][36][37]. This is mainly related to the ability of ceria to stabilize the nickel particles and to promote steam reforming of the oxygenated hydrocarbons intermediates, leading to a reduction in coke deposition. It has also been suggested that ceria hinders secondary dehydration reactions (favoured by the presence of acid sites in the support), and lead to the formation of hydrocarbons that are coke precursors (and thus, responsible for catalyst deactivation) [38]. It has also been suggested that the formation of a CeAlO 3 perovskite structure can suppress the interaction between Ni and Al 2 O 3 and increase the number of active Brønsted acid sites. This, in turn, improves the bifunctional metal-acid properties of Ni/CeO 2 -Al 2 O 3 and favours the hydrogenolysis and dehydrogenation-dehydration of condensable intermediate products that produce more H 2 [39].
In the present investigation, low metal loading (8 wt %) Ni catalysts supported on γ-Al 2 O 3 and γ-Al 2 O 3 modified with CeO 2 were investigated in terms of catalytic activity and time-on-stream stability during the GSR. Different characterization techniques, i.e., N 2 -physisorption, ICP, XRD, temperature programmed reduction (H 2 -TPR), CO 2 -TPD, NH 3 -TPD, XPS, TPO, SEM and TEM were used in an effort to identify the catalyst surface and bulk properties which affect the reaction and its products. The performance of the catalysts was investigated with the aim of identifying the effect of temperature on the total conversion of glycerol and the conversion of glycerol to gaseous products, the selectivity towards H 2 , CO 2 , CO, CH 4 and the liquid effluents produced during the reaction and for the determination of the H 2 /CO and CO/CO 2 molar ratio in the gaseous products mixture. Time-on-stream experiments were conducted for 20 h under harsh reaction conditions in order to induce carbon deposition and the main liquid effluents were quantified.

Chemical Analysis and Textural Properties
The physicochemical, structural and textural properties of the calcined and reduced catalysts are summarized in Table 1. As a first observation, both catalysts have a similar Ni metal loading, which was measured for the calcined samples at 7.14 and 7.69 wt % for Ni/Al 2 O 3 and Ni/AlCeO 3 , respectively. Regarding the properties of the calcined supports used herein, N 2 -physisorption measurements showed that Al 2 O 3 has a much higher specific surface area (SSA) than AlCeO 3 (195 m 2 g −1 compared to 48 m 2 g −1 ) and pore volume (0.65 cm g −1 compared to 0.24 cm g −1 ). Following Ni impregnation and then catalyst calcination, the SSA of the Ni/Al 2 O 3 catalyst dropped to 158 m 2 g −1 and then further to 136 m 2 g −1 after the reduction procedure. For Ni/AlCeO 3 , this drop was less pronounced after calcination, but somewhat sharper after reduction (43 m 2 g −1 after calcination and 26 m 2 g −1 after reduction). Also, for both catalysts, the pore volume remained almost unchanged following calcination, but dropped substantially after reduction. However, the doping of Al 2 O 3 with CeO 2 resulted in enhanced basicity, as shown from the potentiometric titration curves obtained for the Al 2 O 3 and AlCeO 3 suspensions, with the PZC values recorded at 6.8 and 8.2, respectively ( Figure 1). It is noted that the PZC is defined as the pH value where the basic and acidic sites on the surface are in equilibrium [40].  Also, for both catalysts, the pore volume remained almost unchanged following calcination, but dropped substantially after reduction. However, the doping of Al2O3 with CeO2 resulted in enhanced basicity, as shown from the potentiometric titration curves obtained for the Al2O3 and AlCeO3 suspensions, with the PZC values recorded at 6.8 and 8.2, respectively ( Figure 1). It is noted that the PZC is defined as the pH value where the basic and acidic sites on the surface are in equilibrium [40]. The N2 adsorption-desorption isotherms of both catalysts, displayed as an inset in Figure 2, are of Type IV with an H4-type hysteresis loop [41], and show major adsorption at high pressures (0.7 < P/P0 < 1.0) and only minor adsorption at low pressure (P/P0 < 0.05). This is indicative of mainly mesoporous material with some macroporosity [42] and is corroborated by the corresponding pore size distributions.  Figure 3 presents the XRD patterns for the supports (after calcination) and catalytic samples (after calcination and after reduction). Regarding the Al2O3 support and Ni/Al2O3 catalyst, the The N 2 adsorption-desorption isotherms of both catalysts, displayed as an inset in Figure 2, are of Type IV with an H4-type hysteresis loop [41], and show major adsorption at high pressures (0.7 < P/P0 < 1.0) and only minor adsorption at low pressure (P/P 0 < 0.05). This is indicative of mainly mesoporous material with some macroporosity [42] and is corroborated by the corresponding pore size distributions.  Also, for both catalysts, the pore volume remained almost unchanged following calcination, but dropped substantially after reduction. However, the doping of Al2O3 with CeO2 resulted in enhanced basicity, as shown from the potentiometric titration curves obtained for the Al2O3 and AlCeO3 suspensions, with the PZC values recorded at 6.8 and 8.2, respectively ( Figure 1). It is noted that the PZC is defined as the pH value where the basic and acidic sites on the surface are in equilibrium [40]. The N2 adsorption-desorption isotherms of both catalysts, displayed as an inset in Figure 2, are of Type IV with an H4-type hysteresis loop [41], and show major adsorption at high pressures (0.7 < P/P0 < 1.0) and only minor adsorption at low pressure (P/P0 < 0.05). This is indicative of mainly mesoporous material with some macroporosity [42] and is corroborated by the corresponding pore size distributions.  Figure 3 presents the XRD patterns for the supports (after calcination) and catalytic samples (after calcination and after reduction). Regarding the Al2O3 support and Ni/Al2O3 catalyst, the   (Figure 3b) [44]. The XRD patterns of the Ni/AlCeO 3 catalyst after calcination and after reduction (Figure 3b) are nearly identical to the pattern of the calcined AlCeO 3 support with the only differences being two small peaks corresponding to NiAl 2 O 4 detected on the calcined catalyst (2θ = 32.0 • and 37.0 • ), and two peaks, corresponding to Ni 0 , detected at 2θ = 43.5 o and 50.8 • on the reduced sample. It is noted that the reaction 4CeAlO 3 + O 2 ↔ 4CeO 2 + 2Al 2 O 3 is reversible, i.e., CeAlO 3 can be oxidised to CeO 2 by heating in air and CeO 2 can be reduced by heating in H 2 flow [45]. The Ni species mean particle size (Table 1), for the reduced samples, was determined from the XRD spectra using Scherrer analysis and was found at 16 [43]. However, it is also clear from the diffractograms that the Al2O3 and NiAl2O4 peaks are not as intense on the reduced sample in comparison to the calcined one. Moreover, on the reduced catalyst, two small peaks corresponding to metallic nickel (Ni 0 ), at 2θ = 44.0° and 51.2° [43], can be observed. For the calcined ΑlCeO3 support, γ-Al2O3 was detected only at 2θ = 67.2° and the crystal phase of CeO2 dominated with peaks at 2θ = 28.5°, 33.0°, 47.5°, 56.5° and 59.0° (Figure 3b) [44]. The XRD patterns of the Ni/AlCeO3 catalyst after calcination and after reduction ( Figure 3b) are nearly identical to the pattern of the calcined AlCeO3 support with the only differences being two small peaks corresponding to NiAl2O4 detected on the calcined catalyst (2θ = 32.0° and 37.0°), and two peaks, corresponding to Ni 0 , detected at 2θ = 43.5 ο and 50.8° on the reduced sample. It is noted that the reaction 4CeAlO3 + O2 ↔ 4CeO2 + 2Al2O3 is reversible, i.e., CeAlO3 can be oxidised to CeO2 by heating in air and CeO2 can be reduced by heating in H2 flow [45]. The Ni species mean particle size (Table 1), for the reduced samples, was determined from the XRD spectra using Scherrer analysis and was found at 16.8 nm for the Ni/Al2O3 and at 14.2 nm for the Ni/AlCeO3 catalyst.  Figure 4a presents the CO2-TPD profiles obtained over the Ni/Al2O3 and Ni/AlCeO3 catalysts. The Ni/Al2O3 catalyst presents mainly two types of basic sites of weak and medium strength. This is reflected by the peaks that appeared in the temperature regimes of 30 °C < Tmax < 220 °C and 220 °C < Tmax < 450 °C. The strong basic site population is very limited. On the other hand, it seems that the modification of Al2O3 support with Ce leads to a catalyst with a limited population of weak basic sites and a wider distribution of medium strength basic sites, whereas the population of strong basic sites is enhanced significantly. This result is in good agreement with the PZC studies, discussed earlier.  Figure 4a presents the CO 2 -TPD profiles obtained over the Ni/Al 2 O 3 and Ni/AlCeO 3 catalysts. The Ni/Al 2 O 3 catalyst presents mainly two types of basic sites of weak and medium strength. This is reflected by the peaks that appeared in the temperature regimes of 30 • C < T max < 220 • C and 220 • C < T max < 450 • C. The strong basic site population is very limited. On the other hand, it seems that the modification of Al 2 O 3 support with Ce leads to a catalyst with a limited population of weak basic sites and a wider distribution of medium strength basic sites, whereas the population of strong basic sites is enhanced significantly. This result is in good agreement with the PZC studies, discussed earlier.  Figure 4b presents the NH3-TPD profiles obtained over the Ni/Al2O3 and Ni/AlCeO3 catalysts. The Ni/Al2O3 presents a rather rich surface in terms of acid sites of weak, medium and high strength. Incorporation of Ce into the catalyst support leads to a drastic re-distribution of the acid sites of the catalyst, with only the weak sites dominating the surface. It is commonly accepted that on the surface of pure Al2O3 only the Lewis-type acid centres are observed. Miranda et al. [46] attributed desorption peaks at 77-277 °C and at 277-627 °C to the presence of two kinds of acid sites having different strengths. According to the authors, the lower desorption peak is linked to desorption of NH3 bridge species which are bonded to penta-coordinated aluminum (Lewis weak sites), while the higher desorption peak to terminally bonded NH3 on tri-coordinated aluminum (Lewis strong sites) [47]. Thus, it can be concluded that the high temperature acid sites correspond to the Brønsted acid sites of the solid surface [48] and the strong Lewis acid sites are associated with Al2O3 [47,49], whereas the low temperature acid sites are weak acid sites and mainly associated with CeO2 [50] due to its generally weak basicity [39,51].

Ni species Reducibility
Metal-support interactions (electronic or geometrical) are crucial in defining the catalytic properties. In general, NiO interaction with the support, as well as the size of NiO, regulates the extent of its reduction. Weak interaction with the support is considered to give rise to reduction peaks at low temperatures, whereas peaks at high temperatures are due to a strong NiO-support interaction. In the H2-TPR profile of the Ni/Al2O3 catalyst ( Figure 5), two predominant peaks at 420 °C and 690 °C are apparent ( Figure 5). The peak at 690 °C is due to the reduction of nickel aluminate (NiAl2O4), while the peak at 420 °C is due to small NiO crystallites interacting with the support (supported NiO). The small hump at 390 °C for the Ni/Al2O3 catalyst can be assigned to the reduction of bulk NiO [52]. The addition of Ce causes a redistribution of species and most likely changes the support size thus affecting the NiO-support contact area, with the latter being increased. This causes a reduction profile with peaks at lower temperatures compared with the Ni/Al2O3 catalyst.   Incorporation of Ce into the catalyst support leads to a drastic re-distribution of the acid sites of the catalyst, with only the weak sites dominating the surface. It is commonly accepted that on the surface of pure Al 2 O 3 only the Lewis-type acid centres are observed. Miranda et al. [46] attributed desorption peaks at 77-277 • C and at 277-627 • C to the presence of two kinds of acid sites having different strengths. According to the authors, the lower desorption peak is linked to desorption of NH 3 bridge species which are bonded to penta-coordinated aluminum (Lewis weak sites), while the higher desorption peak to terminally bonded NH 3 on tri-coordinated aluminum (Lewis strong sites) [47]. Thus, it can be concluded that the high temperature acid sites correspond to the Brønsted acid sites of the solid surface [48] and the strong Lewis acid sites are associated with Al 2 O 3 [47,49], whereas the low temperature acid sites are weak acid sites and mainly associated with CeO 2 [50] due to its generally weak basicity [39,51].

Ni species Reducibility
Metal-support interactions (electronic or geometrical) are crucial in defining the catalytic properties. In general, NiO interaction with the support, as well as the size of NiO, regulates the extent of its reduction. Weak interaction with the support is considered to give rise to reduction peaks at low temperatures, whereas peaks at high temperatures are due to a strong NiO-support interaction. In the H 2 -TPR profile of the Ni/Al 2 O 3 catalyst ( Figure 5), two predominant peaks at 420 • C and 690 • C are apparent ( Figure 5). The peak at 690 • C is due to the reduction of nickel aluminate (NiAl 2 O 4 ), while the peak at 420 • C is due to small NiO crystallites interacting with the support (supported NiO). The small hump at 390 • C for the Ni/Al 2 O 3 catalyst can be assigned to the reduction of bulk NiO [52]. The addition of Ce causes a redistribution of species and most likely changes the support size thus affecting the NiO-support contact area, with the latter being increased. This causes a reduction profile with peaks at lower temperatures compared with the Ni/Al 2 O 3 catalyst.   Figure 4b presents the NH3-TPD profiles obtained over the Ni/Al2O3 and Ni/AlCeO3 catalysts. The Ni/Al2O3 presents a rather rich surface in terms of acid sites of weak, medium and high strength. Incorporation of Ce into the catalyst support leads to a drastic re-distribution of the acid sites of the catalyst, with only the weak sites dominating the surface. It is commonly accepted that on the surface of pure Al2O3 only the Lewis-type acid centres are observed. Miranda et al. [46] attributed desorption peaks at 77-277 °C and at 277-627 °C to the presence of two kinds of acid sites having different strengths. According to the authors, the lower desorption peak is linked to desorption of NH3 bridge species which are bonded to penta-coordinated aluminum (Lewis weak sites), while the higher desorption peak to terminally bonded NH3 on tri-coordinated aluminum (Lewis strong sites) [47]. Thus, it can be concluded that the high temperature acid sites correspond to the Brønsted acid sites of the solid surface [48] and the strong Lewis acid sites are associated with Al2O3 [47,49], whereas the low temperature acid sites are weak acid sites and mainly associated with CeO2 [50] due to its generally weak basicity [39,51].

Ni species Reducibility
Metal-support interactions (electronic or geometrical) are crucial in defining the catalytic properties. In general, NiO interaction with the support, as well as the size of NiO, regulates the extent of its reduction. Weak interaction with the support is considered to give rise to reduction peaks at low temperatures, whereas peaks at high temperatures are due to a strong NiO-support interaction. In the H2-TPR profile of the Ni/Al2O3 catalyst ( Figure 5), two predominant peaks at 420 °C and 690 °C are apparent ( Figure 5). The peak at 690 °C is due to the reduction of nickel aluminate (NiAl2O4), while the peak at 420 °C is due to small NiO crystallites interacting with the support (supported NiO). The small hump at 390 °C for the Ni/Al2O3 catalyst can be assigned to the reduction of bulk NiO [52]. The addition of Ce causes a redistribution of species and most likely changes the support size thus affecting the NiO-support contact area, with the latter being increased. This causes a reduction profile with peaks at lower temperatures compared with the Ni/Al2O3 catalyst.  These TPR reduction profiles are in agreement with the ones presented by Yang et al. [53] where Ni/Al 2 O 3 reduction takes place between 400-900 • C, whereas the addition of Ce gave rise to another reduction peak in the 200-400 • C region. This is due to the excellent redox properties of the Ce-related phases formed [54]. Thus, it can be concluded that the nickel species on the AlCeO 3 support retained a higher reducibility than Al 2 O 3 ; the higher reducibility being a result of the incorporation of ceria into the alumina lattice [39]. Also, the complex profile of NiO/AlCeO 3 catalyst can be due to the fact that the nucleation and growth of the NiO over the Ce-modified Al 2 O 3 is different compared to that of Al 2 O 3 . This could be due to the presence of new sites (e.g., Ce 3+ , Al-O-Ce site, defects), which can act as anchoring sites for the growth of NiO, which affect the kinetics of the growth and ultimately the NiO size (distribution of NiO sizes). Furthermore, a support that exhibits strong metal-support interactions is anticipated to show less sintering during oxidation/reduction. These TPR reduction profiles are in agreement with the ones presented by Yang et al. [53] where Ni/Al2O3 reduction takes place between 400-900 °C, whereas the addition of Ce gave rise to another reduction peak in the 200-400 °C region. This is due to the excellent redox properties of the Ce-related phases formed [54]. Thus, it can be concluded that the nickel species on the AlCeO3 support retained a higher reducibility than Al2O3; the higher reducibility being a result of the incorporation of ceria into the alumina lattice [39]. Also, the complex profile of NiO/AlCeO3 catalyst can be due to the fact that the nucleation and growth of the NiO over the Ce-modified Al2O3 is different compared to that of Al2O3. This could be due to the presence of new sites (e.g., Ce 3+ , Al-O-Ce site, defects), which can act as anchoring sites for the growth of NiO, which affect the kinetics of the growth and ultimately the NiO size (distribution of NiO sizes). Furthermore, a support that exhibits strong metal-support interactions is anticipated to show less sintering during oxidation/reduction.

Surface Analysis
The XPS Ni 2p and Ce 3d peaks for both the calcined and reduced Ni/Al2O3 and Ni/AlCeO3 catalysts are shown in Figure 6a-f. The Ni 2p3/2 peak for the calcined Ni/Al2O3 sample shows a peak at a binding energy of 856.1 eV, associated with the presence of NiAl2O4 (Figure 6a). The reduced catalyst also shows the presence of a main Ni 2p3/2 peak at a binding energy of 856.1 eV due to NiAl2O4/Ni2O3 but also a shoulder at lower binding energies at around 853 eV, due to the presence of metallic Ni (Figure 6b). For the Ni/AlCeO3 catalysts, only the Ni 2p3/2 peak and satellite is shown in the Figures 6c,d as the 2p1/2 peak and satellite overlap with the Ce 3d peaks.  The calcined Ni/AlCeO 3 catalyst exhibits a Ni 2p 3/2 peak at a binding energy of 856.2 eV due to NiAl 2 O 4 . The reduced Ni/AlCeO 3 catalyst shows a more complex peak shape with 3 components at binding energies of 856.2, 854.1 and around 852.6 eV. The complexity of the Ni 2p spectra in terms of their photoelectron and satellite features (particularly Ni oxide) precludes a sensible peak fit of the Ni 2p envelope [55]. The different components can be tentatively assigned to NiAl 2 O 4 /Ni 2 O 3 (856.2 eV), NiO (854.1 eV) and Ni metal (852.6 eV). XPS Ce 3d photoelectron spectra for CeO 2 (Ce 4+ ) and Ce 2 O 3 (Ce 3+ ) (Figure 6e,f) are also well known for their complex photoelectron and satellite peak structure [56][57][58]. Furthermore, for the catalyst surfaces examined here, overlap with the main Ni 2p 1/2 satellite further complicates the spectral shape. Nevertheless, using standard spectra for CeO 2 and Ce 2 O 3 recorded on this XPS instrument as a reference, and with the assistance of other XPS work on cerium oxides [56][57][58] the Ce 3d spectra have been peak fitted with a reasonable degree of certainty. The photoelectron and satellite peaks for the calcined Ni/AlCeO 3 catalyst are considered first (Figure 6e). The calcined catalyst surface shows peaks generally representative of Ce 4+ (CeO 2 ). Using the nomenclature employed in previous XPS work on cerium oxide [56,57], the peaks are labelled v and u, corresponding to transitions associated with the Ce 3d 5/2 and Ce 3d 3/2 respectively. The peaks observed are those corresponding to the v, v", v"', u, u" and u"', corresponding to the Ce 4+ species, hence CeO 2 is clearly the predominant oxide present. However, the peak fit shows the presence of weak peaks associated with the Ce 3+ species (v 0 , v', u 0 and u') and approximate fractions of the two oxide forms is 85% CeO 2 and 15% Ce 2 O 3 , corresponding to a stoichiometry of CeO 1.94 . The peak shape agrees well with that given by Henderson et al., for a reported stoichiometry of CeO 1.95 [58]. For the reduced catalyst (Figure 6f), the Ce 3d spectrum is very different, with clear peaks present for both the Ce 4+ (v, v"', u, u" and u"') and Ce 3+ (v' and u') species. The peak fit gives a ratio of 41% CeO 2 and 59% Ce 2 O 3 , corresponding to a stoichiometry of CeO 1.80 . The peak shape recorded for the reduced Ni/AlCeO 3 catalyst agrees very well with the Ce 3d peak shapes shown for samples with reported stoichiometries of CeO 1.78 [56] and CeO 1.82 [58], having relative peak intensities in-between those observed for CeO 1.78 and CeO 1.82 , as would be expected for a stoichiometry of CeO 1.80 .
As is widely accepted, CeO 2 increases the dispersion of the metallic phase [59] and this is also observed for the catalysts tested herein with the results shown in Table 2. It has also been suggested that a high dispersion causes a specific interaction between the nickel and ceria species, were the Ni particles close to CeO 2 activate the dissociation of H 2 and by spillover favour the reduction of the ceria surface [60].  Figure 7a presents the results obtained following the catalytic tests, performed under experimental protocol #1, in terms of glycerol total conversion (X C3H8O3 , %) and glycerol conversion to gaseous products (X C3H8O3 gaseous , %) in relation to reaction temperature. It is clear that a very high conversion of glycerol was achieved by both catalysts, but also for the AlCeO 3 support, over the whole temperature range (from ≈85 at 400 • C to > 95% at 750 • C). In contrast, the total conversion of glycerol was relatively low for the bare alumina support, ranging from approximately 70% at 400 • C to 80% at 750 • C. The high conversion values recorded are due not only to the reforming of glycerol but also to its thermal decomposition, which can take place simultaneously during the GSR process [61][62][63].
The endothermic nature of the GSR reaction can be clearly deduced from the conversion of glycerol to gaseous products, which increases sharply with increasing temperature (Figure 7a). However, it is clear that the Ni/AlCeO3 catalyst produces a far greater number of gaseous products in comparison to Ni/Al2O3, especially for temperatures lower than 600 °C. Interestingly, the values of Xgaseous for the Ni/AlCeO3 sample ranged from 60% to 80% for the whole temperature range. Furthermore, the AlCeO3 supporting material seems to produce more or less the same number of gaseous products as the Ni/Al2O3 catalytic sample. Previously published works [39] argued that the formation of CeAlO3 perovskite structures suppresses any strong interaction between the nickel species and the Al2O3 support, thereby increasing the number of active sites and the reducibility of the active nickel species. From the results presented herein, the superior behaviour of the Ni/AlCeO3 can be related to its higher dispersion (XPS), i.e., smaller Ni crystallites, and also to higher electron density and accessibility of the active sites caused by the close contact between nickel and the cerium species. However, it is clear that the Ni/AlCeO 3 catalyst produces a far greater number of gaseous products in comparison to Ni/Al 2 O 3 , especially for temperatures lower than 600 • C. Interestingly, the values of X gaseous for the Ni/AlCeO 3 sample ranged from 60% to 80% for the whole temperature range. Furthermore, the AlCeO 3 supporting material seems to produce more or less the same number of gaseous products as the Ni/Al 2 O 3 catalytic sample. Previously published works [39] argued that the formation of CeAlO 3 perovskite structures suppresses any strong interaction between the nickel species and the Al 2 O 3 support, thereby increasing the number of active sites and the reducibility of the active nickel species. From the results presented herein, the superior behaviour of the Ni/AlCeO 3 can be related to its higher dispersion (XPS), i.e., smaller Ni crystallites, and also to higher electron density and accessibility of the active sites caused by the close contact between nickel and the cerium species.

Gaseous Products' Selectivity
The influence of reaction temperature on H 2 selectivity (S H2 ) and yield (Y H2 ), following experimental protocol #1, for both catalysts and supports is presented in Figure 7b. As a first observation, the H 2 production of the Ni/AlCeO 3 catalyst is markedly higher in comparison to the Ni/Al 2 O 3 sample for the whole temperature range, as its S H2 and Y H2 take the values of 80-90% and 4-6 mol/mol (6 mol/mol is very close to the value predicted by thermodynamics). This improved performance can also be observed for the AlCeO 3 support in comparison to the pure alumina. Thus, it can be suggested that the Ni active sites combined with the Brønsted acid sites promote the hydrogenolysis and dehydrogenation-dehydration of the bifunctional metal-acid Ni/AlCeO 3 catalyst, thereby converting more condensable intermediates into gaseous products, as well as leading to higher glycerol conversion and H 2 yield.
The influence of reaction temperature on the selectivity to CO 2 and CO (S CO2 and S CO ) for all samples is presented in Figure 7c. It is obvious that the Ni/AlCeO 3 sample is more selective towards CO 2 and less selective towards CO for the whole temperature range. This is in contrast to the trend observed for the Ni/Al 2 O 3 catalyst and supporting materials, as these materials appear more selective towards CO and less selective towards CO 2 . Thus, it seems that the ability to transform OHCs into H 2 and CO 2 is significantly higher for the Ni/AlCeO 3 catalyst probably due to its enhanced basic character (higher PZC value), which was introduced by the addition of CeO 2 to the alumina. Moreover, from Figure 7d it can be seen that both catalytic samples exhibit very low S CH4 for the whole temperature range; in contrast, S CH4 increases for the supports with increasing temperature. This was rather expected as according to the literature [36] at high water to glycerol feed ratios and at temperatures higher than 650 • C, the formation of CH 4 is strongly inhibited due to the methane steam reforming reaction being catalysed by the metallic active phase. Keeping in mind that glycerol decomposition to CH 4 during the reforming process is highly favoured [64], it appears that both catalysts have the capacity to reform the produced CH 4 into H 2 and CO.
Finally, the influence of reaction temperature on the H 2 /CO and the CO/CO 2 molar ratios in the gaseous products' mixture is presented in Figure 7e. For the Ni/AlCeO 3 catalyst, the CO/CO 2 molar ratio is close to zero for the whole temperature range, while the H 2 /CO molar ratio value decreases with increasing temperature from a value of about 17 (450 • C) to a value of about 10 (550-750 • C). In contrast, for the Ni/Al 2 O 3 catalyst, both ratios are equal to 2-3 and appear relatively stable over the whole temperature range under investigation. The H 2 /CO molar ratio value is more or less negligible for the supporting materials, whereas the CO/CO 2 molar ratio values range from about 2 between 400 and 500 • C, increases from 500 to 600 • C and decreases again for 650 • C < T < 750 • C (its maximum value was~5.5 for the Al and~4 for the AlCeO 3 sample). It is known that in the reaction process, dehydrogenation of the adsorbed glycerol molecule first takes place on the metal surface of catalyst to give adsorbed intermediates for the cleavage of C-C or C-O bonds.
From the results presented above it is obvious that the presence of ceria in the alumina supporting material has an important effect on the conversion and the gaseous product distribution, mainly by increasing H 2 and CO 2 production to the detriment of CH 4 and CO formation. It is likely that for the Ni/AlCeO 3 catalyst there is a first reaction step that involves rapid C-C breaking and CO formation, which is then followed by the WGS reaction, enhancing H 2 and CO 2 production. Sad et al. [65] has argued that acidic supports (such as alumina) favour the production of oxygenates (such as acrolein, acetol, acetaldehyde and acetic acid) via dehydration and dehydrogenation reactions, and that H 2 production is favoured by the use of non-acidic supports. However, the GSR process is also influenced by pyrolysis phenomena, given that the glycerol molecule is not thermally stable, which means that the intermediates formed by glycerol cracking are reformed on the catalyst surface [66].

Liquid Products' Selectivity
In accordance with previous research (e.g., [10,12,31]), a variety of liquid products were identified for the GSR however, only the main ones, i.e., acetone, acrolein, acetaldehyde, acetic acid, allyl alcohol, and acetol were quantified. The influence of temperature on liquid product selectivity for the Ni/Al 2 O 3 and Ni/AlCeO 3 catalysts and the corresponding supports is shown in Figure 8. For the Ni/AlCeO 3 catalyst and the AlCeO 3 support, production of liquid effluents ceases over 700 • C, 50 • C higher compared to the Ni/Al 2 O 3 sample. In contrast, pure alumina produces liquid effluents even at temperatures as high as 750 • C. Another important observation is the differences in the liquid product distribution between modified and unmodified samples. Specifically, for the Ni/AlCeO 3 and AlCeO 3 , allyl alcohol seems to be the main product (at least for high temperatures), with acetaldehyde and acetone the secondary products. On the other hand, acetone is the main product at the high temperature range for the Ni/Al 2 O 3 sample and acetic acid and acetol for pure alumina. the Ni/AlCeO3 catalyst and the AlCeO3 support, production of liquid effluents ceases over 700 °C, 50 °C higher compared to the Ni/Al2O3 sample. In contrast, pure alumina produces liquid effluents even at temperatures as high as 750 °C. Another important observation is the differences in the liquid product distribution between modified and unmodified samples. Specifically, for the Ni/AlCeO3 and AlCeO3, allyl alcohol seems to be the main product (at least for high temperatures), with acetaldehyde and acetone the secondary products. On the other hand, acetone is the main product at the high temperature range for the Ni/Al2O3 sample and acetic acid and acetol for pure alumina. According to the literature, there are four main reaction routes of glycerol consisting of three types of reactions, i.e., dehydrogenation, hydrogenolysis, and dehydration. These reactions represent the effects of the bifunctional metal-acid properties of the catalysts [47]. The first and the second routes of glycerol dehydrogenation and hydrogenolysis generate a common product, such as ethylene glycol, which is converted to acetaldehyde, glycolaldehyde, and ethanol. The hydrogenolysis of glycerol also produces methanol. The third route is glycerol dehydration to hydroxyacetone (acetol), which can be converted simultaneously in three ways to form a variety of products such as acetaldehyde, acetone, ethanol, acetic acid, propionic acid, formic acid, phenol, 1,2-propanediol, propanal, 1-propanol, and 2-cyclo-pentanone. This pathway is caused by the activation of the terminal OH of glycerol on the Lewis acid sites, which is represented by the strong acid sites of the catalysts. The fourth pathway is another glycerol dehydration reaction producing 3-hydroxypropanal, which is a starting reactant for the production of several chemicals. The species observed and confirming this pathway include allyl alcohol, acetaldehyde, methanol, formic acid, According to the literature, there are four main reaction routes of glycerol consisting of three types of reactions, i.e., dehydrogenation, hydrogenolysis, and dehydration. These reactions represent the effects of the bifunctional metal-acid properties of the catalysts [47]. The first and the second routes of glycerol dehydrogenation and hydrogenolysis generate a common product, such as ethylene glycol, which is converted to acetaldehyde, glycolaldehyde, and ethanol. The hydrogenolysis of glycerol also produces methanol. The third route is glycerol dehydration to hydroxyacetone (acetol), which can be converted simultaneously in three ways to form a variety of products such as acetaldehyde, acetone, ethanol, acetic acid, propionic acid, formic acid, phenol, 1,2-propanediol, propanal, 1-propanol, and 2-cyclo-pentanone. This pathway is caused by the activation of the terminal OH of glycerol on the Lewis acid sites, which is represented by the strong acid sites of the catalysts. The fourth pathway is another glycerol dehydration reaction producing 3-hydroxypropanal, which is a starting reactant for the production of several chemicals. The species observed and confirming this pathway include allyl alcohol, acetaldehyde, methanol, formic acid, C 2 H 4 , and C 2 H 6 . This pathway involves the protonation of the secondary OH of glycerol to form acrolein, which favours the Brønsted acid sites mainly represented by the moderate acid sites [67].

Catalytic Stability
The results for catalytic stability are presented in Figure 9 and Table 3. It is noted that these experiments were carried out under more severe conditions (experimental protocol #2) in order to provoke carbon deposition and catalyst deactivation. From the results presented herein it is clear that although the Ni/AlCeO 3 catalyst suffers a slight decrease concerning glycerol total conversion (94-77%) and glycerol conversion to gaseous products (48-37%), it maintains remarkably stable values for H 2 yield (2.9-2.3) and selectivity (89-81%), as well as CO 2 (75-67%) and CO (23-29%) selectivity. As a result, the H 2 /CO and CO/CO 2 molar ratios remain reasonably constant for the duration of the experiment, with values ranging between 9 to 6.5 and 0.3 to 0.4, respectively. The catalytic performance observed indicated that the effect of the WGS reaction do not weaken with time. It is accepted that the addition of basic modifiers, such as CeO 2 to Ni/Al 2 O 3 catalysts can prevent carbon formation by favouring both the adsorption of H 2 O, O 2 , CO 2 or-OH fragments and the spillover of such fragments from the support to the metal particles [67], and thus facilitate carbon gasification.

Catalytic Stability
The results for catalytic stability are presented in Figure 9 and Table 3. It is noted that these experiments were carried out under more severe conditions (experimental protocol #2) in order to provoke carbon deposition and catalyst deactivation. From the results presented herein it is clear that although the Ni/AlCeO3 catalyst suffers a slight decrease concerning glycerol total conversion (94-77%) and glycerol conversion to gaseous products (48-37%), it maintains remarkably stable values for H2 yield (2.9-2.3) and selectivity (89-81%), as well as CO2 (75-67%) and CO (23-29%) selectivity. As a result, the H2/CO and CO/CO2 molar ratios remain reasonably constant for the duration of the experiment, with values ranging between 9 to 6.5 and 0.3 to 0.4, respectively. The catalytic performance observed indicated that the effect of the WGS reaction do not weaken with time. It is accepted that the addition of basic modifiers, such as CeO2 to Ni/Al2O3 catalysts can prevent carbon formation by favouring both the adsorption of H2O, O2, CO2 or-OH fragments and the spillover of such fragments from the support to the metal particles [67], and thus facilitate carbon gasification. In contrast, although the decrease in total XC3H8O3 and XC3H8O3 gaseous products for the Ni/Al2O3 catalyst was similar to the Ni/AlCeO3, SH2, YH2 and SCO2 decreased substantially; this decrease was accompanied by an increase in the SCO values. As a result, for the Ni/Al2O3 catalyst, the H2/CO molar ratio decreased from 9.5 to 0.6 and the CO/CO2 ratio increased from 0.3 to 8.0 with time on stream; a significant increase in methane production was also observed. The deactivation of Ni catalysts  In contrast, although the decrease in total X C3H8O3 and X C3H8O3 gaseous products for the Ni/Al 2 O 3 catalyst was similar to the Ni/AlCeO 3 , S H2 , Y H2 and S CO2 decreased substantially; this decrease was accompanied by an increase in the S CO values. As a result, for the Ni/Al 2 O 3 catalyst, the H 2 /CO molar ratio decreased from 9.5 to 0.6 and the CO/CO 2 ratio increased from 0.3 to 8.0 with time on stream; a significant increase in methane production was also observed. The deactivation of Ni catalysts supported on pure alumina during the GSR reaction has also been reported by other research groups and has been attributed to carbon formation and sintering [68][69][70].
The difference in the catalytic behavior of the samples can be strongly affected by the nature of the carrier, suggesting that the activity and the rate of deactivation is likely to be related to the different extent of electronic interaction between supported metal and support, influencing the bonding and reactivity of the chemisorbed species. The presence of Ni 2+ /Ni 0 and Ce 4+ /Ce 3+ couples means that different species can participate in the activation of the glycerol molecules. According to Bazin et al. [71] an increase of the local electron density is expected with increased metal dispersion, as has been proven for the Ni/AlCeO 3 catalyst; it exhibits a great number of active sites accessible to the reactant molecules compared to that observed for the Ni/Al 2 O 3 catalyst.

Characterization of Used Catalysts
The nature of the carbonaceous deposits formed on to the spent catalytic samples following the time-on-stream experiments (experimental protocol #2) was examined using TPO, SEM, and TEM.
The TPO results for both catalysts are presented in Figure 10. As a first observation, significantly more carbon was deposited on to the Ni/Al 2 O 3 catalyst (0.41 g coke /g catalyst ) in comparison to the Ni/AlCeO 3 sample (0.14 g coke /g catalyst ). The Ni/Al 2 O 3 catalyst shows a weak peak at 470 • C and two broad peaks at 550 • C and 650 • C, which indicate the existence of different co-existing carbon allotropes. For the Ni/AlCeO 3 catalyst, there is a clear peak at~375 • C, whereas a larger peak at 490 • C is observed; this points to the existence of coke that contains a higher fraction of defective carbon, which is known to be more easily oxidized. In addition, the high temperature decomposition peak appears as one single thermal event centred at 625 • C, i.e., slightly lower than the corresponding peak of the Ni/Al 2 O 3 catalyst. It has been reported that the thermal decomposition of physisorbed carbonaceous products, termed as soft (or amorphous) coke, occurs at temperatures between 200 and 500 • C [39], while the gasification of hard coke (comprising of bulky carbonaceous species and often referred to as filamentous carbon) [72][73][74] takes place between 500 and 800 • C. The hard coke species are 500-600 °C and can be associated with filamentous coke (carbon nanotubes, CNTs, and nanofibers, CNFs, that contain defects), whilst the coke gasified at over 600 °C can be associated with graphitic carbon species. This graphitic carbon is an inert coke that does not easily react with oxygen or steam [75]. Thus, not only was a higher quantity of coke deposited onto the spent Ni/Al2O3 catalyst, but also the fractions of graphitic carbon were greater and more difficult to oxidize than on the Ni/AlCeO3 catalyst. These results explain the excellent stability observed for the Ni/AlCeO3 catalyst. The different coke species on the catalyst surface can be formed by the Boudouard, methane decomposition and polymerization reactions [72,73]. Thus, the redox properties of CeO2 and a higher dispersion of the active phase for the Ni/AlCeO3 catalyst has resulted in markedly lower coke yields.
The morphology of the carbon deposits was initially examined using SEM/EDX; representative SEM micrographs and the carbon EDX images of the corresponding areas are shown in Figure 11. Although it is not straightforward to determine the precise nature of the coke deposits from this initial examination (i.e., whether these are nano-fibers, micro-fibers or carbon nanotubes), entangled tubular arrangements (filaments) can be discerned from the images. Also, the orientation of these filaments is quite random, which makes it difficult to estimate their length with any degree of accuracy. However, the EDX carbon images (red) show heavy coke deposition, particularly on the Ni/Al2O3 spent catalyst, in good agreement with the TPO results discussed above.  The hard coke species are 500-600 • C and can be associated with filamentous coke (carbon nanotubes, CNTs, and nanofibers, CNFs, that contain defects), whilst the coke gasified at over 600 • C can be associated with graphitic carbon species. This graphitic carbon is an inert coke that does not easily react with oxygen or steam [75]. Thus, not only was a higher quantity of coke deposited onto the spent Ni/Al 2 O 3 catalyst, but also the fractions of graphitic carbon were greater and more difficult to oxidize than on the Ni/AlCeO 3 catalyst. These results explain the excellent stability observed for the Ni/AlCeO 3 catalyst. The different coke species on the catalyst surface can be formed by the Boudouard, methane decomposition and polymerization reactions [72,73]. Thus, the redox properties of CeO 2 and a higher dispersion of the active phase for the Ni/AlCeO 3 catalyst has resulted in markedly lower coke yields.
The morphology of the carbon deposits was initially examined using SEM/EDX; representative SEM micrographs and the carbon EDX images of the corresponding areas are shown in Figure 11. Although it is not straightforward to determine the precise nature of the coke deposits from this initial examination (i.e., whether these are nano-fibers, micro-fibers or carbon nanotubes), entangled tubular arrangements (filaments) can be discerned from the images. Also, the orientation of these filaments is quite random, which makes it difficult to estimate their length with any degree of accuracy. However, the EDX carbon images (red) show heavy coke deposition, particularly on the Ni/Al 2 O 3 spent catalyst, in good agreement with the TPO results discussed above. The hard coke species are 500-600 °C and can be associated with filamentous coke (carbon nanotubes, CNTs, and nanofibers, CNFs, that contain defects), whilst the coke gasified at over 600 °C can be associated with graphitic carbon species. This graphitic carbon is an inert coke that does not easily react with oxygen or steam [75]. Thus, not only was a higher quantity of coke deposited onto the spent Ni/Al2O3 catalyst, but also the fractions of graphitic carbon were greater and more difficult to oxidize than on the Ni/AlCeO3 catalyst. These results explain the excellent stability observed for the Ni/AlCeO3 catalyst. The different coke species on the catalyst surface can be formed by the Boudouard, methane decomposition and polymerization reactions [72,73]. Thus, the redox properties of CeO2 and a higher dispersion of the active phase for the Ni/AlCeO3 catalyst has resulted in markedly lower coke yields.
The morphology of the carbon deposits was initially examined using SEM/EDX; representative SEM micrographs and the carbon EDX images of the corresponding areas are shown in Figure 11. Although it is not straightforward to determine the precise nature of the coke deposits from this initial examination (i.e., whether these are nano-fibers, micro-fibers or carbon nanotubes), entangled tubular arrangements (filaments) can be discerned from the images. Also, the orientation of these filaments is quite random, which makes it difficult to estimate their length with any degree of accuracy. However, the EDX carbon images (red) show heavy coke deposition, particularly on the Ni/Al2O3 spent catalyst, in good agreement with the TPO results discussed above.   Figure 12 presents representative TEM images of the two spent catalytic samples and the corresponding particle size distribution histograms. The images confirm the formation of different carbon allotropes on the catalytic surface (indicated by black dashed circles), but also show heavy filament formation on the Ni/Al 2 O 3 system, in excellent agreement with the discussion above. Encapsulating carbon can also be observed in the lower image for this system (examples of Ni particles are shown by red dashed circles), which is known to lead to a loss of activity as the Ni particle is no longer available during the reaction. Moreover, the particle size distribution histograms show that the mean Ni size of the Ni/AlCeO 3 catalyst is substantially smaller than that of the Ni/Al 2 O 3 , helping to explain the improved activity and stability characteristics of the former sample. Figure 12 presents representative TEM images of the two spent catalytic samples and the corresponding particle size distribution histograms. The images confirm the formation of different carbon allotropes on the catalytic surface (indicated by black dashed circles), but also show heavy filament formation on the Ni/Al2O3 system, in excellent agreement with the discussion above. Encapsulating carbon can also be observed in the lower image for this system (examples of Ni particles are shown by red dashed circles), which is known to lead to a loss of activity as the Ni particle is no longer available during the reaction. Moreover, the particle size distribution histograms show that the mean Ni size of the Ni/AlCeO3 catalyst is substantially smaller than that of the Ni/Al2O3, helping to explain the improved activity and stability characteristics of the former sample.

Catalyst Preparation
The alumina support (pellets) was acquired from Akzo, while the AlCeO3 (powder) was sourced from Sigma Aldrich. The pelletized support was crushed and sieved to particle sizes in the range 350-500 μm, while the powder was first pelletized and then crushed to the same size. The catalysts were produced via the wet impregnation technique using an aqueous solution of Ni(NO3)2 6H2O, obtained from Sigma Aldrich, with a concentration of 0.17 M to obtain catalysts with metal content of 8 wt %. The water contained in the slurries was evaporated under continuous stirring at 75 °C over 5 h. After, the suspensions were dried for 12 h at 120 °C and the calcination of the catalysts was carried out at 800 °C for 4 h.

Catalyst Preparation
The alumina support (pellets) was acquired from Akzo, while the AlCeO 3 (powder) was sourced from Sigma Aldrich. The pelletized support was crushed and sieved to particle sizes in the range 350-500 µm, while the powder was first pelletized and then crushed to the same size. The catalysts were produced via the wet impregnation technique using an aqueous solution of Ni(NO 3 ) 2 6H 2 O, obtained from Sigma Aldrich, with a concentration of 0.17 M to obtain catalysts with metal content of 8 wt %. The water contained in the slurries was evaporated under continuous stirring at 75 • C over 5 h. After, the suspensions were dried for 12 h at 120 • C and the calcination of the catalysts was carried out at 800 • C for 4 h.

Catalyst Characterization
For the determination of the catalysts' surface and bulk properties, different characterization techniques were employed. These included the calculation/determination of: (a) the Point of Zero Charge (PZC) for the calcined supports via Potentiometric Mass Titrations (PMTs), (b) the total specific surface area (SSA) of the catalytic materials and corresponding supports via the Brunauer-Emmet-Teller (BET) method, (c) the pore size distribution (PSD) of the catalytic materials and corresponding supports using the Barrett, Joyner, and Halenda (BJH) method, (d) the total metal loading (wt %) of the calcined catalysts using Inductively Coupled Plasma Atomic Emission Spectroscopy (ICP-AES), (e) the crystalline structure of the calcined supports, as well as the calcined and reduced catalysts by X-ray Diffraction (XRD) analysis, (f) the degree of Ni species reducibility using Temperature Programmed Reduction (H 2 -TPR), (g) the acid and basic properties of the catalysts via CO 2 -and NH 3 -TPD experiments, (h) the nature of the various surface species and their oxidation states, on the reduced catalysts, using X-ray Photoelectron Spectroscopy (XPS) analysis, (i) the morphological characteristics of the spent catalysts using Scanning Electron Microscopy (SEM) and Transmission Electron Microscopy (TEM), and (j) the amount and nature of the carbonaceous deposits on the spent catalysts using Temperature Programmed Oxidation (TPO). The methodology and equipment used has been described in detail in previous publications by our group, and in particular: PZC in [26], XRD in [11], ICP at [76], and N 2 adsorption-desorption, H 2 -TPR, CO 2 -TPD, NH 3 -TPD, XPS, SEM, TEM and TPO in [17,23].

Catalytic Tests
For the catalytic testing a continuous flow fixed-bed reactor was employed; the exact system and procedure has been described in detail in previous publications (e.g., [17,23]). Succinctly, using glycerol with 99.5% purity (Sigma-Aldrich, St. Louis, MO, USA) and a Weight Hourly Space Velocity (WHSV) of 50,000 mL g −1 h −1 , two different experimental protocols were employed. For the first protocol, the gas feed at the inlet of the reactor was a gas mixture of 73% H 2 O, 4% C 3 H 8 O 3 and 23% He (i.e., 20 v.v. % of C 3 H 8 O 3 diluted in H 2 O) and the catalytic performance was investigated between 400-750 • C. The second protocol was used for the investigation of catalytic stability during time-on-stream and was carried out for 20 h at 600 • C. The conditions chosen were more severe and the gas mixture at the reactor's inlet consisted of 63% H 2 O, 7% C 3 H 8 O 3 , 30% He (31 • of C 3 H 8 O 3 diluted in H 2 O). Before commencing the experiments, catalytic activation was undertaken in situ using a flow (100 mL min −1 ) of high purity H 2 (5.0) at 800 • C for 1h. The system was then purged with He and the temperature lowered according to the protocol that was to be followed. The liquid products were analysed via a combination of Gas Chromatography (Agilent, Santa Clara, CA, USA, 7890A) and Mass Spectroscopy (Agilent 5975C). The gaseous products were determined via an Agilent gas chromatographer (7890A). Detailed information regarding the analysis of liquid and gaseous products can be found in Reference [11].

Reaction Metrics
The investigation of catalytic performance necessitated the calculation of total glycerol conversion, conversion of glycerol into gaseous products, and determination of the H 2 yield, H 2 , CH 4 , CO 2 and CO selectivity. For the calculations Equations (10) and (14) shown below were used. The selectivity of acetone [(CH 3 ) 2 CO], acetaldehyde (C 2 H 4 O), acetol (C 3 H 6 O 2 ), allyl alcohol (CH 2 =CHCH 2 OH), acrolein (C 3 H 4 O) and acetic acid (C 2 H 4 O) was calculated based on Equation (15): %glycerol conversion (total conversion) = Glycerol in − Glycerol out Glycerol in × 100 (10) %glycerol conversion (gaseous products) = C atoms in the gas products total C atoms in the f eedstock × 100 (11) H 2 yield = H 2 mol produced mol o f glycerol in the f eedstock (12) %H 2 selectivity = H 2 mol produced C atoms produced in the gas phase × 1 RR × 100 (13) where, RR is the reforming ratio (7/3), defined as the ratio of moles of H 2 to CO 2 formed.
% selectivity o f i = C atoms in species i C atoms produced in the gas phase × 100 (14) where, species i refers to CO, CO 2 and CH 4 .
% selectivity o f i = C atoms in species i C atoms produced in the liquid phase × 100 (15) where, species i refers to acetol, acetone, allyl alcohol, acetaldehyde, acrolein and acetic acid.

Conclusions
A critical assessment of the effect of the Ce-modification of Al 2 O 3 on the catalytic performance of Ni/Al 2 O 3 and Ni/AlCeO 3 catalysts towards glycerol steam reforming reaction and H 2 production has been performed. A thorough comparison of the AlCeO 3 , Al 2 O 3 (supports) as well as Ni/Al 2 O 3 and Ni/AlCeO 3 (catalysts) has been presented.
The study has shown that Ce-modification of Al 2 O 3 leads to a Ni catalyst with increased basicity (PZC and CO 2 -TPD studies) and a higher Ni dispersion (H 2 -TPR, XPS studies). The Ni/AlCeO 3 sample was more selective towards CO 2 and less selective towards CO. The opposite trend was observed for the Ni/Al 2 O 3 catalyst. Importantly, both Ni/Al 2 O 3 and Ni/AlCeO 3 had low CH 4 selectivity. For the Ni/AlCeO 3 catalyst the CO/CO 2 molar ratio was almost zero, while the H 2 /CO molar ratio value decreases (from 17 to 10) with increasing temperature (450 • C to 750 • C). Regarding the liquid products for the Ni/AlCeO 3 and AlCeO 3 , allyl alcohol was found to be the main product (at least for high temperatures), with acetaldehyde and acetone the secondary ones. On the other hand, acetone was the main product at the high T range for the Ni/Al 2 O 3 sample and acetic acid and acetol for pure alumina. Stability studies performed for over than 20 h on stream showed that the Ni/AlCeO 3 catalyst experienced a slight decrease on glycerol total conversion (94-77%) and glycerol conversion to gaseous products (48-37%), but it maintained remarkably stable values for H 2 yield (2.9-2.3) and selectivity (89-81%), as well as for the CO 2 (75-67%) and CO (23-29%) selectivity. Characterization of the exhausted catalysts was performed in order to understand the nature of the deposited coke. It was found that Ni/AlCeO 3 was far more resistant to coke, while at the same time, it favored the formation of more defective carbon compared to the coke deposited onto Ni/Al 2 O 3 .