Support Effects on the Activity of Ni Catalysts for the Propane Steam Reforming Reaction

The catalytic performance of supported Ni catalysts for the propane steam reforming reaction was investigated with respect to the nature of the support. It was found that Ni is much more active when supported on ZrO2 or YSZ compared to TiO2, whereas Al2O3− and CeO2-supported catalysts exhibit intermediate performance. The turnover frequency (TOF) of C3H8 conversion increases by more than one order of magnitude in the order Ni/TiO2 < Ni/CeO2 < Ni/Al2O3 < Ni/YSZ < Ni/ZrO2, accompanied by a parallel increase of the selectivity toward the intermediate methane produced. In situ FTIR experiments indicate that CHx species produced via the dissociative adsorption of propane are the key reaction intermediates, with their hydrogenation to CH4 and/or conversion to formates and, eventually, to CO, being favored over the most active Ni/ZrO2 catalyst. Long term stability test showed that Ni/ZrO2 exhibits excellent stability for more than 30 h on stream and thus, it can be considered as a suitable catalyst for the production of H2 via propane steam reforming.


Introduction
The high energy efficiency of fuel cells has drawn considerable attention toward the development of hydrogen production technologies. Hydrogen can be produced from fossil fuels either via hydrocarbon pyrolysis or hydrocarbon reforming processes including steam reforming, partial oxidation or autothermal reforming [1][2][3][4][5]. Biomass processes consisting of biological (bio-photolysis, dark fermentation, photo fermentation) and thermochemical (pyrolysis, gasification, combustion, liquefaction) methods as well as water splitting processes such as electrolysis, thermolysis or photolysis can be alternatively applied for the production of H 2 from renewable energy sources [2,6]. However, the latter approaches are facing some major obstacles mostly related to high cost and low H 2 yields. Currently, steam reforming of light hydrocarbons, including natural gas, ethane, propane, butane and liquified petroleum gas (LPG), are considered among the most promising and economical routes for hydrogen production [7]. Propane, which is the main component of LPG, has many advantages such as high energy density, compressibility to a transportable liquid at normal temperature and well-developed infrastructure which enable its use worldwide [8][9][10]. Moreover, propane can be stored and transferred as LPG through a wide distribution network or in high pressure cylinders in order to be supplied in remote places (e.g., agricultural, inaccessible or camping areas) or for domestic uses (e.g., households) [3,7].
Under propane steam reforming conditions, the water-gas shift reaction occurs simultaneously at low temperatures contributing to H 2 and CO 2 production, whereas CO/CO 2 methanation may also run in parallel yielding CH 4 and H 2 O. Methane can be also formed via hydrogenation of CH x species derived by the dissociative adsorption of propane on the catalyst surface or through propane decomposition accompanied by ethylene production. In certain cases, the C 2 H 4 , CH 4 and CO thus produced are further decomposed supplied from the JCPDS data base, whereas the primary crystallite size of M x O y (d MxOy ) was estimated according to Scherrer's equation: where θ is the angle of diffraction corresponding to the peak broadening, B is the fullwidth at half maximum intensity (in radians) and λ = 0.15406 nm is the X-ray wavelength corresponding to CuK a radiation. The specific surface area (SSA) of the supported Ni catalysts were measured by N 2 adsorption at 77 K (B.E.T. technique) using a Gemini III 2375 instrument (Micromeritics, Norcross, GA, USA). Carbon monoxide chemisorption measurements at 25 • C were applied for the determination of Ni dispersion and mean particle size using a modified Sorptomatic 1900 apparatus (Fisons Instruments, Glaskow, UK) and assuming a CO:Me stoichiometry of 1:1, an atomic surface area of 6.5 Å 2 and spherical particles. CO chemisorption measurements were used instead of H 2 chemisorption in order to avoid overestimation of Ni dispersion due to hydrogen spillover effects, which have been previously found to occur over supported Ni catalysts [26,27]. Nickel particle size was calculated according to the following equation: where d Ni is the mean crystallite diameter, ρ Ni (= 8.9 g·cm −3 ) is the density of Ni and S Ni [m 2 /g Ni ] is the surface area per gram of Ni. Transmission electron microscopy (TEM) images were obtained with a JEM-2100 system (JEOL, Akishima, Tokyo, Japan) operated at 200 kV (point resolution 0.23 nm) using an Erlangshen CCD Camera (Model 782 ES500W, Gatan Inc., Pleasanton, CA, USA). Samples were dispersed in water and spread onto a carbon-coated copper grid (200 mesh). Details related to the equipment and procedures used for catalyst characterization have been described in detail elsewhere [28].

Catalytic Performance Tests and Kinetic Measurements
The catalytic performance of the synthesized materials was studied in a tubular fixed-bed quartz reactor under atmospheric pressure using an apparatus which has been described in detail elsewhere [11]. The reaction conditions were as follows: temperature range 400-750 • C, H 2 O/C = 3.25, and gas hourly space velocity (GHSV) = 55,900 h −1 . The reactor was loaded with 150 mg of catalyst (particle diameter: 0.15 < d p < 0.25 mm) and placed in an electric furnace, where it was reduced in situ at 300 • C for 1 h under 50%H 2 /He flow (60 cm 3 min −1 ) to ensure that the Ni exists in its metallic phase prior to catalytic performance tests. Catalyst reduction was followed by heating at 750 • C under He and subsequent switch of the flow to the feed stream consisted of 4.5%C 3 H 8 + 0.15%Ar + 44%H 2 O (He balance). Argon was used as internal standard in order to account for the volume change. Water was fed through an HPLC pump (LD Class Pump, TELEDYNE SSI, PA, USA) into a vaporizer maintained at 180 • C and mixed with the gas stream coming from mass-flow controllers. A condenser immersed in an ice bath was placed at the exit of the reactor to condensate water prior to introduction of the gas stream to the analysis system. Reaction gases (He, 30%C 3 H 8 −1%Ar/He, H 2 ) are supplied from high-pressure gas cylinders (Buse Gas, Bad Hönningen, Germany) and are of ultrahigh purity. Measurements of reactants' and products' concentrations were obtained by stepwise decreasing temperature up to 400 • C. The effluent from the reactor was analyzed using two gas chromatographs (Shimadzu, Kyoto, Japan) which were connected in parallel. The procedure used for gas phase analysis was described in our previous study [11]. The conversion of propane (X C 3 H 8 ) was calculated using the following expression: Selectivity toward reaction products containing carbon was defined using Equation (5). The factor n corresponds to the number of carbon atoms in the corresponding molecule (e.g., for CO is 1, for C 2 H 4 is 2 etc.): Selectivity toward hydrogen production was defined as the concentration of hydrogen produced divided with the concentration of all products containing hydrogen according to Equation (6). The factor m represents the number of hydrogen atoms in the corresponding molecule (e.g., for CH 4 and C 2 H 4 is 4).
The intrinsic reaction rates for propane steam reforming reaction were measured for low propane conversions (X C 3 H 8 < 10%) by varying W/F using the following expression: where R C3H8 is the molar rate of C 3 H 8 consumption (mol s −1 g cat −1 ), [C 3 H 8 ] in , [C 3 H 8 ] out , are the inlet and outlet concentrations (v/v) of C 3 H 8 , respectively, F in and F out are the total flow rates in the inlet and outlet of the reactor (mols −1 ), respectively, and W is the mass of catalyst (g cat ).
Turnover frequencies (TOFs) of propane conversion were estimated following Equation (8) taking into account the measurements of both the reaction rates and nickel dispersions: where AW Ni is the atomic weight of nickel (g Ni /mol Ni ), X Ni is the nickel loading (g Ni /g cat ) and D Ni is the dispersion of nickel.

In Situ FTIR Spectroscopy
In situ Fourier transform infrared (FTIR) experiments were carried out using an iS20 FTIR spectrometer (Nicolet, Thermo Fischer Scientific, Waltham, MA, USA) equipped with an MCT detector, a KBr beam splitter and a diffuse reflectance (DRIFT) sampling system (Specac, Orpington, UK) accompanied by an environmental chamber suitable for the study of diffusely reflecting solid samples in a controlled atmosphere. A flow system equipped with mass flow controllers, a steam saturator and a set of valves used for controlling the gas stream interacted with the catalyst surface, was directly connected to the gas inlet of the environmental chamber.
In a typical experiment, the catalyst powder was placed in the sampling system and heated at 500 • C in flowing helium for 10 min and then reduced under hydrogen flow at 300 • C for 30 min. The flow was then switched to He and the temperature was increased at 500 • C. After remaining 10 min at this temperature the sample was cooled at 100 • C. While cooling, the background spectra were recorded at the desired temperatures. Finally, the flow was switched to the reaction mixture, which consisted of 0.5%C 3 H 8 +5%H 2 O (in He). Steam was introduced to the system via an independent He line passing through a saturator containing water maintained at 60 • C. The resulting gas mixture was fed to the DRIFT cell through stainless steel tubing maintained at 60 • C by means of heating tapes. A Nanomaterials 2021, 11, 1948 5 of 17 spectrum was collected at 100 • C after 15 min-on-stream followed by a stepwise increase of temperature up to 500 • C. During heating, spectra were recorded at selected temperatures after an equilibration for 15 min. In all experiments, the total flow through the DRIFT cell was 30 cm 3 min −1 . Reaction gases (He, 2%C 3 H 8 /He, H 2 ) are supplied from high-pressure gas cylinders (Buse Gas, Bad Hönningen, Germany) and are of ultrahigh purity.

Catalyst Characterization
The XRD patterns of Ni/M x O y catalysts are shown in Figure 1 100 °C. While cooling, the background spectra were recorded at the desired temperatures. Finally, the flow was switched to the reaction mixture, which consisted of 0.5%C3H8 +5%H2O (in He). Steam was introduced to the system via an independent He line passing through a saturator containing water maintained at 60 °C. The resulting gas mixture was fed to the DRIFT cell through stainless steel tubing maintained at 60 °C by means of heating tapes. A spectrum was collected at 100 °C after 15 min-on-stream followed by a stepwise increase of temperature up to 500 °C. During heating, spectra were recorded at selected temperatures after an equilibration for 15 min. In all experiments, the total flow through the DRIFT cell was 30 cm 3 min −1 . Reaction gases (He, 2%C3H8/He, H2) are supplied from high-pressure gas cylinders (Buse Gas, Bad Hönningen, Germany) and are of ultrahigh purity.

Catalyst Characterization
The XRD patterns of Ni/MxOy catalysts are shown in Figure 1     In the case of Ni/TiO 2 and Ni/YSZ catalysts an additional weak peak located 44.5 • was appeared corresponding to Ni (111) plane (JCPDS Card No. 04-0850). The absence of peaks corresponding to metallic Ni for the rest catalysts investigated is due to the low Ni loading and/or particle size. The primary crystallite size of the supports was estimated according to Scherrer's formula at the diffraction angles corresponding to (440) plane for Al 2 O 3 , (−111) plane for ZrO 2 , (111) plane for CeO 2 , (111) plane for YSZ and (101) plane for TiO 2 , and it was found to be 6.0 nm for Ni/Al 2 O 3 , 10.5 nm for Ni/CeO 2 , 15.0 nm for Ni/ZrO 2 , 20.9 nm for Ni/YSZ and 21.8 nm for Ni/TiO 2 ( Table 1).
The SSAs of Ni catalysts supported on metal oxide (M x O y ) carriers were estimated equal to 39 m 2 /g for Ni/ZrO 2 , 11 m 2 /g for Ni/YSZ, 66 m 2 /g for Ni/Al 2 O 3 , 39 m 2 /g for Ni/CeO 2 and 41 m 2 /g for Ni/TiO 2 (Table 1). Results of Ni dispersion (D Ni ) and mean particle size (d Ni ) estimated from CO chemisorption meaurements are summarized in Table 1. Generally, low Ni dispersions were estimated for all the investigated catalysts, most possibly due to the high Ni content (5 wt.%) in agreement with previous studies [4,6]. Higher Ni dispersion of 11.9% and smaller particle size of 8.5 nm was found for Ni/CeO 2 catalyst, whereas Ni/TiO 2 exhibited the lowest value of Ni dispersion of 2.8% and the largest particle size of 36.1 nm. Figure 2 shows representative TEM images and selected area electron diffraction (SAED) patterns obtained from Ni/YSZ, Ni/CeO 2 and Ni/TiO 2 catalysts. In all cases Ni particles appear as fairly homogeneously distributed spherical particles with average sizes of 20 nm for Ni/YSZ, 10 nm for Ni/CeO 2 and 30 nm for Ni/TiO 2 , in agreement with those estimated according to CO chemisorption measurements (Table 1). It should be noted that, based on the results of Table 1, the mean particle size of Ni is similar to the average size of the corresponding metal oxide used as support for all the investigated catalysts. This may hinder distinguishing between Ni particles and the MxOy carrier in TEM images. Thus, SAED analysis was performed to calculate the d-spacing in an attempt to further discerned Ni particles from those of metal oxide support ( Figure 2, It should be noted that, based on the results of Table 1, the mean particle size of Ni is similar to the average size of the corresponding metal oxide used as support for all the investigated catalysts. This may hinder distinguishing between Ni particles and the M x O y carrier in TEM images. Thus, SAED analysis was performed to calculate the d-spacing in an attempt to further discerned Ni particles from those of metal oxide support ( Figure 2, Table 2). Results indicated that, in all cases, Ni particles are present in TEM images as evidenced by the appearance of the (111)

Influence of the Nature of the Support on Catalytic Activity
The influence of the nature of the support on catalytic performance for the propane steam reforming reaction has been investigated using Ni catalysts (5 wt.%) supported on five different commercial metal oxide powders (ZrO 2 , YSZ, TiO 2 , Al 2 O 3 , CeO 2 ). The results obtained are shown in Figure 3A, where propane conversion is plotted as a function of reaction temperature. It is observed that, among the investigated catalysts, Ni/ZrO 2 is the most active one, exhibiting measurable C 3 H 8 conversions at temperatures higher than 400 • C and achieving complete conversion at 750 • C. Although Ni/YSZ is activated at similar temperatures as Ni/ZrO 2 , the conversion curve of propane is shifted toward higher temperatures. This is also the case for Ni/Al 2 O 3 and Ni/CeO 2 catalysts, which present similar performance. The latter catalysts are less active than Ni/YSZ below 550 • C, but are able to reach higher X C3H8 at higher temperatures. The titania-supported catalyst becomes active above 500 • C, with the propane conversion curve being shifted at remarkably higher temperatures. In all examined cases, the carbon balance was satisfactory, with a deviation of <1%.
Results of specific reaction rate measurements are presented in the Arrhenius diagram of Figure 3B, where it is observed that the TOF of propane conversion increases in the order Ni/TiO 2 < Ni/CeO 2 < Ni/Al 2 O 3 < Ni/YSZ < Ni/ZrO 2 , with its value at 450 • C being more than one order of magnitude higher when Ni is dispersed on ZrO 2 compared to TiO 2 , and approximately 2.5 times higher than that of Ni/Al 2 O 3 . It should be mentioned that, as discussed above, the mean particle size of Ni varies significantly for the investigated catalysts from 8.5 nm for Ni/CeO 2 to 36.1 nm for Ni/TiO 2 . If the Ni particle size were similar for this set of catalysts then the order of catalytic activity could be somewhat different. Interestingly, no trend was observed between the specific reaction rate and Ni particle size or M x O y crystallite size or M x O y surface area. This indicates that either Nanomaterials 2021, 11,1948 8 of 17 these parameters do not affect catalytic activity or most possibly each of them contributes in a different manner to the reaction rate, resulting in the observed catalyst ranking. It should be noted that all catalysts have been reduced at 400 • C prior to physicochemical characterization measurements. Although the values of SSA or d MxOy or d Ni may were different if catalyst pre-reduction was carried out at 750 • C, which is the onset reaction temperature for catalytic performance experiments, the trend of catalytic properties with respect to the nature of the support is not expected to vary due to the catalyst pretreatment at different temperatures, at least to such an extent that would affect the catalyst ranking for the propane steam reforming reaction. the order Ni/TiO2 < Ni/CeO2 < Ni/Al2O3 < Ni/YSZ < Ni/ZrO2, with its value at 450 °C being more than one order of magnitude higher when Ni is dispersed on ZrO2 compared to TiO2, and approximately 2.5 times higher than that of Ni/Al2O3. It should be mentioned that, as discussed above, the mean particle size of Ni varies significantly for the investigated catalysts from 8.5 nm for Ni/CeO2 to 36.1 nm for Ni/TiO2. If the Ni particle size were similar for this set of catalysts then the order of catalytic activity could be somewhat different. Interestingly, no trend was observed between the specific reaction rate and Ni particle size or MxOy crystallite size or MxOy surface area. This indicates that either these parameters do not affect catalytic activity or most possibly each of them contributes in a different manner to the reaction rate, resulting in the observed catalyst ranking. It should be noted that all catalysts have been reduced at 400 °C prior to physicochemical characterization measurements. Although the values of SSA or dMxOy or dNi may were different if catalyst pre-reduction was carried out at 750 °C, which is the onset reaction temperature for catalytic performance experiments, the trend of catalytic properties with respect to the nature of the support is not expected to vary due to the catalyst pretreatment at different temperatures, at least to such an extent that would affect the catalyst ranking for the propane steam reforming reaction. The apparent activation energies (Ea) of the propane steam reforming reaction were calculated from the slopes of the fitted lines of Figure 3B. The results showed that the nature of the metal oxide carrier significantly affects Ea, which takes values between 102 kJ/mol for Ni/CeO2 and 154 kJ/mol for Ni/ZrO2 without presenting any trend with respect to catalytic activity (Table 1). This can be explained taking into account that, as it will be discussed below, several reactions run in parallel under the present experimental conditions each one of which is influenced by the nature of the support in a different manner resulting in the observed random variation of Ea with catalytic activity. The results are in agreement with our previous study where it was found that the apparent activation en- The apparent activation energies (E a ) of the propane steam reforming reaction were calculated from the slopes of the fitted lines of Figure 3B. The results showed that the nature of the metal oxide carrier significantly affects E a , which takes values between 102 kJ/mol for Ni/CeO 2 and 154 kJ/mol for Ni/ZrO 2 without presenting any trend with respect to catalytic activity (Table 1). This can be explained taking into account that, as it will be discussed below, several reactions run in parallel under the present experimental conditions each one of which is influenced by the nature of the support in a different manner resulting in the observed random variation of E a with catalytic activity. The results are in agreement with our previous study where it was found that the apparent activation energy for the reaction of steam reforming of propane over Rh catalysts supported on a variety of metal oxides does not present any trend with the activity order [11]. Figure 4 shows the selectivities toward reaction products as a function of temperature over the supported Ni catalysts investigated. In all cases the main products detected were H 2 , CO 2 , CO and CH 4 with their selectivities being significantly varied with temperature. In particular, for Ni/ZrO 2 catalyst ( Figure 4A), both hydrogen (S H2 ) and CO 2 (S CO2 ) selectivities decrease from 99 to 78% and from 98 to 58%, respectively, with increasing temperature in the range of 390-505 • C followed by an increase of methane selectivity (S CH4 ) up to 32.5%, indicating the occurrence of CO 2 methanation reactions. Carbon dioxide consumption continues with further increase of temperature above 505 • C contrary to S H2 which progressively increases reaching 99% at 720 • C. Consumption of CO 2 is followed by production of CO providing evidence that the reverse WGS (RWGS) reaction is enhanced at high temperatures. Moreover, S CH4 decreases above 505 • C and becomes practically zero at 720 • C, implying that the reaction of methane steam reforming occurs contributing to the observed increase of both S H2 and S CO . It should be noted that selectivity toward reaction products containing carbon was defined as the concentration of each product containing carbon at reactor effluent over the concentration of all products containing Nanomaterials 2021, 11,1948 9 of 17 carbon (5), whereas S H2 was defined as the concentration of hydrogen produced divided by the concentration of all products containing hydrogen (6). Therefore, the values of Sc n and S H2 cannot be correlated based on the stoichiometry of the reactions taking place under propane steam reforming conditions. Nanomaterials 2021, 11,1948 10 of 17 The influence of the nature of the support on the activity of Ni catalysts for propane steam reforming reaction was also investigated by Harshini et al. [16], who found tha Ni/LaAlO3 was more active than Ni/Al2O3, while Ni/CeO2 exhibited intermediate per formance. The optimum activity of the former catalyst was attributed to the small N nanoparticles dispersed on LaAlO3 surface. Although the effect of the support nature on propane steam reforming activity has not been widely studied over Ni catalysts, certain properties of metal oxide carriers may help explain the results of Figure 3. For example the use of YSZ as support, which exhibited high activity in the results of the presen study, has been found to suppress carbon deposition over Rh-Ni catalysts by providing lattice oxygen, which facilitates carbon removal and enhances the dissociation of C-C bond under reaction conditions [3]. The prevention of coke formation, occurring either via hydrocarbons decomposition or CO dissociation, by the lattice oxygen of the suppor  Qualitatively similar results were obtained for the rest of the investigated Ni catalysts, with the main differences being related to the values of the selectivities toward the reaction products, which reflect the extent of each reaction taking place with respect to the nature of the support. In particular, the observed decrease of S H2 and S CO2 at low temperatures and the simultaneous increase of S CH4 are higher for the most active Ni/ZrO 2 ( Figure 4A) and Ni/YSZ ( Figure 4B) catalysts, followed by Ni/Al 2 O 3 ( Figure 4C) and Ni/CeO 2 ( Figure 4D), whereas it is eliminated for Ni/TiO 2 ( Figure 4E). As a result methane production increases in the order of Ni/TiO 2 < Ni/CeO 2 < Ni/Al 2 O 3 < Ni/YSZ < Ni/ZrO 2 which is consistent with the order of catalytic activity. This can be clearly seen in Figure 5 where the TOF at 450 • C is plotted as a function of methane selectivity obtained at the same temperature for all the investigated catalysts. It is observed that the specific reaction rate increases from 0.018 s −1 to 0.33 s −1 following the above catalyst ranking and accompanied by a parallel increase of S CH4 from 0 to 29%. The results indicate that there is a clear relationship between catalytic activity and methane production.
It should be noted that besides CO 2 hydrogenation, CH 4 can be also produced via CO hydrogenation. However, the contribution of the latter reaction does not seem to be significant for the results of the present study taking into account the low S CO below 500 • C and its progressive increase with temperature. Moreover, methane formation may also take place via hydrogenation of CH x species formed following the dissociative adsorption of propane on Ni surface and the subsequent hydrogenation of the so-formed C 3 H x species [29,30]. As it will be discussed below, the formation of CH x species intermediates may be the key reaction since it has been proposed that they interact with the hydroxyl groups or lattice oxygen of the support producing CO or CO 2 and H 2 [3,29,30].
The influence of the nature of the support on the activity of Ni catalysts for propane steam reforming reaction was also investigated by Harshini et al. [16], who found that Ni/LaAlO 3 was more active than Ni/Al 2 O 3 , while Ni/CeO 2 exhibited intermediate performance. The optimum activity of the former catalyst was attributed to the small Ni nanoparticles dispersed on LaAlO 3 surface. Although the effect of the support nature on propane steam reforming activity has not been widely studied over Ni catalysts, certain properties of metal oxide carriers may help explain the results of Figure 3. For example, the use of YSZ as support, which exhibited high activity in the results of the present study, has been found to suppress carbon deposition over Rh-Ni catalysts by providing lattice oxygen, which facilitates carbon removal and enhances the dissociation of C-C bond under reaction conditions [3]. The prevention of coke formation, occurring either via hydrocarbons decomposition or CO dissociation, by the lattice oxygen of the support has been also demonstrated over Ni/CeO 2 -Al 2 O 3 [19]. Moreover, the addition of manganese oxide on Ni/Al 2 O 3 was found to act as an oxygen donor that is transferred to Ni particles leading to rapid decomposition and oxidation of C 3 H 8 and CH 4 or C 2 H 4 that may be produced under reaction conditions, resulting in further H 2 production and improvement of the catalyst lifetime [9]. It has been also found that activation of steam followed by H 2 formation may be favored over metal catalysts supported on "reducible" metal oxides through generation of oxygen defects, resulting in improved propane steam reforming activity and resistance to coke formation [3,13,31]. Based on previous studies, the reducibility of the supports used in the present study is expected to vary significantly. It is well known that Al 2 O 3 is a hardly reducible metal oxide characterized by low oxygen storage capacity contrary to TiO 2 and CeO 2, which are strongly reducible metal oxides, or ZrO 2 and YSZ, which are characterized by intermediate oxygen mobility [32]. Based on the above, Ni/TiO 2 should be also active for the title reaction, taking into account that titania support is characterized by high oxygen storage capacity [32,33]. However, the results of Figure 3 clearly show that: (a) this catalyst was the least active one and (b) the catalytic activity is increased in the order Ni/TiO 2 < Ni/CeO 2 < Ni/Al 2 O 3 < Ni/YSZ < Ni/ZrO 2 , which cannot be correlated with the reducibility of the support. Therefore, it is evident that support reducibility is not among the key parameters affecting the catalytic activity of Ni according to the results of the present study. The low activity of Ni/TiO 2 catalyst may be related to the fact that Ni/TiO 2 has lower Ni dispersion and larger Ni particles, which was previously suggested to suppress both propane steam reforming [15,16], and the intermediate (CO 2 or CH x ) hydrogenation reactions, in excellent agreement with the results of our previous study [28]. However, large Ni particles may not be solely responsible for the low activity of Ni/TiO 2 taking into account that no trend was observed between TOF and Ni particle size for the investigated catalysts.
catalytic activity is increased in the order Ni/TiO2 < Ni/CeO2 < Ni/Al2O3 < Ni/YSZ < Ni/ZrO2, which cannot be correlated with the reducibility of the support. Therefore, it is evident that support reducibility is not among the key parameters affecting the catalytic activity of Ni according to the results of the present study. The low activity of Ni/TiO2 catalyst may be related to the fact that Ni/TiO2 has lower Ni dispersion and larger Ni particles, which was previously suggested to suppress both propane steam reforming [15,16], and the intermediate (CO2 or CHx) hydrogenation reactions, in excellent agreement with the results of our previous study [28]. However, large Ni particles may not be solely responsible for the low activity of Ni/TiO2 taking into account that no trend was observed between TOF and Ni particle size for the investigated catalysts. It should be mentioned that a different activity order was reported in our previous study over supported Rh catalysts, where it was found that Rh/TiO2 was the most active catalyst with TOF being one order of magnitude higher compared to that measured for Rh/CeO2 [11]. This implies that the nature of the metallic phase may affect metal/support interactions leading to variations on propane steam reforming activity and/or possibly changes on the type of active sites on the catalyst surface. It should be mentioned that a different activity order was reported in our previous study over supported Rh catalysts, where it was found that Rh/TiO 2 was the most active catalyst with TOF being one order of magnitude higher compared to that measured for Rh/CeO 2 [11]. This implies that the nature of the metallic phase may affect metal/support interactions leading to variations on propane steam reforming activity and/or possibly changes on the type of active sites on the catalyst surface.

Long Term Stability Test
The long-term stability of Ni/ZrO 2 catalyst, which exhibited the highest activity, was investigated at 650 • C using the same experimental conditions as those used in catalytic performance tests. In this experiment, the catalyst was reduced in situ at 300 • C under 50%H 2 /He flow followed by heating at 650 • C. The flow was then switched to the reaction mixture and determination of the conversion of propane and product selectivity started. The system was shut down overnight, while the catalyst was kept at room temperature under a He flow. The next day the catalyst is heated to 650 • C in the He flow, followed by switching of the flow to the reaction mixture and determination of X C3H8 and the product selectivity as a function of time. Results obtained are shown in Figure 6, where X C3H8 and S H2 , S CO2 , S CO and S CH4 are plotted as functions of time-on-stream. As it can be seen the catalyst presents excellent stability for more than 30 h-on-stream. Propane conversion and hydrogen selectivity were varied in the range of 95-99% and 97-98%, respectively. The selectivity toward methane was low (3-4%) whereas S CO and S CO2 exhibited similar values ranging between 46 and 50%. The carbon balance was found to be satisfactory during the stability test, with a deviation lower than 1-2%. selectivity as a function of time. Results obtained are shown in Figure 6, where XC3H8 and SH2, SCO2, SCO and SCH4 are plotted as functions of time-on-stream. As it can be seen the catalyst presents excellent stability for more than 30 h-on-stream. Propane conversion and hydrogen selectivity were varied in the range of 95-99% and 97-98%, respectively. The selectivity toward methane was low (3-4%) whereas SCO and SCO2 exhibited similar values ranging between 46 and 50%. The carbon balance was found to be satisfactory during the stability test, with a deviation lower than 1-2%.

DRIFT Studies
The interaction of selected catalysts with the reaction mixture was also investigated employing in situ FTIR spectroscopy. Experiments were conducted in the temperature range of 100-500 °C using a feed composition of 0.5%C3H8 + 5%H2O (in He) and the results obtained are shown in Figure 7. It is observed that the spectrum recorded at 100 °C ( Figure 7A, trace a) for the pre-reduced Ni/TiO2 catalyst is characterized by two negative bands at 3787 and 3676 cm −1 which can be attributed to losses of ν (OH) intensity of at least two different types of free hydroxyl groups, which are either originally present on TiO2 surface or created via H2O adsorption. Two weak peaks were also detected in the ν (C-H) region, located at 2987 and 2966 cm −1 (trace a) due to C-H stretching vibrations in methyl groups (CH3,ad) and to symmetric C-H vibrations in methylene groups (CH2,ad), respectively [20,31,[34][35][36]. These peaks are more obvious in Figure 8A (trace a) where selected spectra in the narrow range of 3200-2400 cm −1 are presented. Moreover, a band at 1642 cm −1 followed by a shoulder at 1560 cm −1 can be discerned, which have been previously assigned to carbonate species associated with TiO2 support [37][38][39][40][41][42]. An increase of temperature results in progressive separation of the latter two bands which are both

DRIFT Studies
The interaction of selected catalysts with the reaction mixture was also investigated employing in situ FTIR spectroscopy. Experiments were conducted in the temperature range of 100-500 • C using a feed composition of 0.5%C 3 H 8 + 5%H 2 O (in He) and the results obtained are shown in Figure 7. It is observed that the spectrum recorded at 100 • C ( Figure 7A, trace a) for the pre-reduced Ni/TiO 2 catalyst is characterized by two negative bands at 3787 and 3676 cm −1 which can be attributed to losses of ν (OH) intensity of at least two different types of free hydroxyl groups, which are either originally present on TiO 2 surface or created via H 2 O adsorption. Two weak peaks were also detected in the ν (C-H) region, located at 2987 and 2966 cm −1 (trace a) due to C-H stretching vibrations in methyl groups (CH 3,ad ) and to symmetric C-H vibrations in methylene groups (CH 2,ad ), respectively [20,31,[34][35][36]. These peaks are more obvious in Figure 8A (trace a) where selected spectra in the narrow range of 3200-2400 cm −1 are presented. Moreover, a band at 1642 cm −1 followed by a shoulder at 1560 cm −1 can be discerned, which have been previously assigned to carbonate species associated with TiO 2 support [37][38][39][40][41][42]. An increase of temperature results in progressive separation of the latter two bands which are both shifted toward lower wavenumbers. A new band at 1430 cm −1 can be also discerned in the spectra obtained at 350 • C (trace f) which is also due to carbonate species. This peak may be also present in the spectra obtained at lower temperatures but couldn't be clearly observed due to the low signal-to-noise ratio in the region below 1700 cm −1 . The intensities of bands assigned to carbonate species are progressively decreased above 200 • C. This decrease is accompanied by the detection of a weak peak at 2021 cm −1 [38,[43][44][45][46], which is characteristic of linear-bonded CO on reduced nickel sites (Ni • ), indicating that carabonate species are further decomposed yielding CO and most possibly also CO 2 in the gas phase. The weak bands in the ν (C-H) region are present on the spectra obtained up to 500 • C ( Figure 7A, trace i) implying that CH x species are thermally stable and remained adsorbed on the catalyst surface. [47][48][49]. The intensity of the latter bands increases with increasing temperature up to 300 °C and diminishes upon further heating at 350-400 °C. This is also the case for the band at 1640 cm −1 , indicating that bicarbonate species are decomposed above 450 °C. At temperatures higher than 400 °C (traces h-i) two new broad bands seem to be developed at ca 1540 and 1350 cm −1 . The broadness of these bands implies that they may contain contributions from more than one species with their corresponding bands being overlapped.  This can be clearly seen in the spectrum obtained at 500 °C (trace i) where four bands can be clearly discerned located at 1558, 1520, 1370 and 1331 cm −1 . Those detected at 1520 and 1331 cm −1 have been previously attributed to bidentate carbonates [49,50], whereas those located at 1558 and 1370 cm −1 can be assigned to bidentate formate species [51] adsorbed on ZrO2 surface. The appearance of the latter bands is accompanied by evolution of three peaks in the ν(CO) region due to CO linearly adsorbed on reduced Ni sites (2021 cm −1 ) and bridged bonded CO (1909 and 1858 cm −1 ) [43,46,[52][53][54].
Interestingly, CHx species are eliminated from the spectra obtained above 400 °C followed by evolution of CH4 in the gas phase, as evidenced by the detection of the 3016 cm −1 band (traces h-i). Production of CH4 at the expense of CHx species can be clearly seen in Figure 8B where the spectra obtained at 100 and 450 °C in the wavenumber range of 3200-2400 cm −1 are presented.
Based on the above it can be suggested that the reaction of steam reforming of propane over Ni/ZrO2 catalyst proceeds via a dissociative adsorption of propane on metallic Ni leading to the formation of C3Hx species, which are further decomposed toward CHx species and probably carbon oxides due to the presence of H2O adsorbed on the support surface. This may result in the formation of the bicarbonate species (1640, 1540 and 1425 A similar experiment was conducted over the most active Ni/ZrO 2 catalyst and the results obtained are presented in Figure 7B. It is observed that the interaction of catalyst with the reformate mixture at 100 • C (trace a) results in the appearance of bands corresponding to bicarbonate species (1640 cm −1 ) [47][48][49], CH x species (2983 and 2966 cm −1 ) [31,36,47] as well as by negative bands (3749 and 3675 cm −1 ) related to the consumption of surface OH groups [31,36,48]. Increase of temperature at 200 • C (trace c) leads to the progressive development of two bands at 1540 and 1425 cm −1 due to bicarbonate species [47][48][49]. The intensity of the latter bands increases with increasing temperature up to 300 • C and diminishes upon further heating at 350-400 • C. This is also the case for the band at 1640 cm −1 , indicating that bicarbonate species are decomposed above 450 • C. At temperatures higher than 400 • C (traces h-i) two new broad bands seem to be developed at ca 1540 and 1350 cm −1 . The broadness of these bands implies that they may contain contributions from more than one species with their corresponding bands being overlapped.
This can be clearly seen in the spectrum obtained at 500 • C (trace i) where four bands can be clearly discerned located at 1558, 1520, 1370 and 1331 cm −1 . Those detected at 1520 and 1331 cm −1 have been previously attributed to bidentate carbonates [49,50], whereas those located at 1558 and 1370 cm −1 can be assigned to bidentate formate species [51] adsorbed on ZrO 2 surface. The appearance of the latter bands is accompanied by evolution of three peaks in the ν(CO) region due to CO linearly adsorbed on reduced Ni sites (2021 cm −1 ) and bridged bonded CO (1909 and 1858 cm −1 ) [43,46,[52][53][54].
Interestingly, CH x species are eliminated from the spectra obtained above 400 • C followed by evolution of CH 4 in the gas phase, as evidenced by the detection of the 3016 cm −1 band (traces h-i). Production of CH 4 at the expense of CH x species can be clearly seen in Figure 8B where the spectra obtained at 100 and 450 • C in the wavenumber range of 3200-2400 cm −1 are presented.
Based on the above it can be suggested that the reaction of steam reforming of propane over Ni/ZrO 2 catalyst proceeds via a dissociative adsorption of propane on metallic Ni leading to the formation of C 3 H x species, which are further decomposed toward CH x species and probably carbon oxides due to the presence of H 2 O adsorbed on the support surface. This may result in the formation of the bicarbonate species (1640, 1540 and 1425 cm −1 ) detected at low temperatures on the surface of the support [20]. Part of CH x species are hydrogenated above 400 • C, yielding methane in the gas phase (band at 3016 cm −1 ) whereas the rest interact with adsorbed water producing formates (bands at 1558 and 1370 cm −1 ) and, eventually, CO species adsorbed on the Ni surface (bands at 2021, 1909 and 1858 cm −1 ). Formates may also interact with hydroxyl groups producing H 2 and carbonate species (1520 and 1331 cm −1 ), which are further decomposed to CO 2 [20] . It should be noted, however, that, under certain conditions, CH x species may be also dehydrogenated producing C and H 2 . Surface carbon is either accumulated on the catalyst surface resulting in catalyst deactivation (which is not the case here) or it interacts with the hydroxyl groups or the lattice oxygen of the support yielding CO or CO 2 [3,16,19,31,55].
The ability of CH x species to be converted to methane and/or formates on the surface of Ni/ZrO 2 may imply that CH x species are more weakly adsorbed and/or more reactive on the surface of this catalyst, thus resulting in higher overall activity for the propane steam reforming reaction. This agrees well with results of Figure 5, where it was shown that catalytic activity increases progressively with increasing methane selectivity. If as discussed above the origin of adsorbed CO is formate species, the high reactivity of CH x species over Ni/ZrO 2 is also indirectly confirmed by the significantly higher population of carbonyl species observed over this catalyst. On the other hand, although CH x species were also detected on the surface of Ni/TiO 2 catalyst, no band due to gas phase methane or formate species was observed, indicating that CH x species cannot be further converted in the temperature range investigated. The increase of S CH4 in parallel with the increase of catalytic activity ( Figure 5) indicates that the reactivity of CH x species is strongly related to the conversion of propane to the desired products. Although detailed mechanistic studies should be conducted to further explore the reaction pathway, it can be suggested that CH x species are key reaction intermediates for the reaction of propane steam reforming.

Conclusions
The results of the present study show that the catalytic activity of Ni and the product distribution for the propane steam reforming reaction depend strongly on the nature of the support. The specific reaction rate measured for Ni/ZrO 2 catalyst was found to be more than one order of magnitude higher compared to that measured for Ni/TiO 2 . The intermediate production of CH 4 is strongly influenced by the type of metal oxide used as support following the same trend with that of catalytic activity. DRIFTS studies provided evidences that the most active Ni/ZrO 2 catalyst is able to convert the intermediate produced CH x species to CH 4 and/or formate species and, subsequently, to carbon oxides and H 2 . In contrast, CH x species seem to be less reactive when Ni is dispersed on TiO 2 , thus resulting in a lower reaction rate. In addition to the high activity of Ni/ZrO 2 catalyst, it was also found to be stable for more than 30 h on stream and therefore, is a promising candidate for the production of H 2 for fuel cell applications.