2.1 Preparation of the catalysts

The SBA-15 silica support was synthesized according to the method reported by Zhao et al. [12]. After filtration, the powders were calcined at 500 °C (heating rate 2 °C·min⁻¹) for about 9 h in a muffle furnace (Nabertherm, LE6/11) in order to remove the organic template and the Cl ions. The nickel nitrate [Ni(NO₃)₂·6H₂O] or cerium nitrate [Ce(NO₃)₃·6H₂O] salts (from Sigma–Aldrich) were then impregnated using the “two-solvent” technique described elsewhere [13–15] that favors the deposition of the metal precursor species inside the pores of the SBA-15 supports. The cerium content was fixed to 6 wt% and that of nickel to 5 wt%. After each impregnation, the sample was calcined in the air at 450 °C for 5 h prior to further treatment. Even if calcination in the air has been reported to be detrimental to the dispersion of the active phase compared to a treatment in a neutral or reducing atmosphere [4,16,17], this calcination atmosphere was chosen in order to obtain catalysts that are similar to those usually reported in the literature for this type of materials. The three prepared samples were referred to as Ni–Ce/SBA-15 when the Ce addition preceded the Ni addition, Ce–Ni/SBA-15 when the nickel precursor was added first and Co-NiCe/SBA-15 when Ni and Ce were co-impregnated together. Two additional reference catalysts were prepared following the same procedure, but with only Ni or Ce addition (samples Ni/SBA-15 and Ce/SBA-15, respectively).

2.2 Catalytic test

The catalytic tests were conducted at atmospheric pressure in a continuous U-shaped fixed-bed flow reactor using 20 mg of catalysts loaded on a quartz wool plug. Prior to the catalytic reaction, each sample was reduced in situ at 650 °C for about 2 h in 5 volume% H₂/Ar (30 mL·min⁻¹). After that, the temperature was cooled down to 200 °C, then increased from about 200 °C up to 800 °C at a rate of 5 °C·min⁻¹. The molar ratio of CH₄ to CO₂ was 1:1 and the gas hourly space velocity (GHSV) was 264 L·g⁻¹·h⁻¹. For stability measurements, the temperature was cooled down to 500 °C and maintained at this temperature for 12 h. The effluent gases were analyzed using a Micro-GC Inficon equipped with a thermal conductivity detector (TCD). The conversions during activity and stability measurements were calculated as follows (in and out relate to the measurements at the reactor entrance and exit, respectively):

2.3 Characterization techniques

The pore volumes and surface areas were determined from N₂ adsorption–desorption isotherms recorded on an ASAP 2020 (Micromeritics) apparatus. Prior to experiments, the samples were degassed under vacuum for 2 h at 250 °C. XRD patterns were registered on a Bruker D8 diffractometer (Bragg-Brentano, copper λ = 1.5418 Å). The diffractograms were recorded both at low (2θ between 0.5° and 5°) and wide angles (2θ between 5° and 95°) to respectively (i) ascertain the pore structure arrangement of the supports and (ii) evaluate the nature and mean size (calculated using the Scherrer equation) of the supported metal nanoparticles. The morphology and homogeneity of the silica grains were identified by SEM (scanning electron microscopy) on a Hitachi SU-70 SEM-FEG. TEM (transmission electron microscopy) images were registered on a 2010 JEOL JEM-200 electron microscope operating at 200 keV (LaB6 gun) and equipped with an energy dispersive X-ray spectroscopy detector (EDS PGT, spatial resolution below 10 nm) for chemical analyses. TPR experiments were carried out on a Micromeritics Autochem 2910 Instrument equipped with a trap for water removal before TCD detection. H₂ consumption was followed during heating in 5 volume% H₂/Ar (20 mL·min⁻¹) at a rate of 10 °C·min⁻¹ from 200 °C to 650 °C.

3.1 Catalytic activity

The comparison of activity between the different samples is presented first in order to identify if the order of addition of the metals played a role towards catalytic performances of the samples. After reduction of the catalysts at 650 °C under 5% H₂/Ar for 2 h, the reactor was cooled down to about 200 °C and catalytic measurements were carried out while raising again the temperature up to around 750 °C. The variations of the conversions of reactants (CH₄ and CO₂) as a function of temperature are shown in Fig. 1. Both conversions increased with temperature, reflecting the endothermic character of the dry reforming reaction. However, CH₄ conversion was always slightly higher than that of CO₂, over the entire temperature range, suggesting some (although limited) decomposition of methane into hydrogen and carbon (CH₄ → 2H₂ + C) as a side reaction [6,18].

In the absence of nickel (Ce/SBA-15 reference catalyst, data not shown), no catalytic activity was observed, thus validating that the active sites for the dry reforming reaction are the metallic nickel nanoparticles. The most active catalyst was Ni–Ce/SBA-15 (Ce impregnated first) on which 100% of CH₄ conversion and 90% of CO₂ conversion were reached at around 600 °C (Fig. 1a, a′). On the contrary, the lowest activity was observed for Ce–Ni/SBA-15 (Ni impregnated first) on which the same conversions as above were attained at 700 °C (Fig. 1c, c′). The co-impregnation led to better results than Ce–Ni/SBA-15, yet lower than Ni–Ce/SBA-15, but similar to those of the reference catalyst Ni/SBA-15. Therefore, it appears that Ce has to be added prior to Ni in order to ensure a positive effect towards activity, whereas its addition after Ni leads on the contrary to an activity loss. This could be due to a blocking by Ce of some channels where Ni species were previously deposited, thus, leading to a hindering of the access to the metal active phase.

All catalysts were relatively stable during the 12 h of catalytic running (data not shown), not only in terms of activity but also with respect to their selectivity. However, the selectivity (reflected in the H₂:CO molar ratio near unity) varied between the different catalysts, as for their activity, in the order Ni–Ce/SBA-15 > Co-NiCe/SBA-15 > Ni/SBA-15 > Ce–Ni/SBA-15. Thus, the H₂:CO molar ratio was 0.96 on Ni–Ce/SBA-15, whereas it was only 0.77 on Ce–Ni/SBA-15. Such a ratio below 1 suggests the occurrence of side reactions, such as RWGS reverse water–gas shift reaction [4,6] that therefore occur at higher extent on the Ce–Ni/SBA-15 catalyst than on the Ni–Ce/SBA-15 one. The more limited access to active sites assumed above from conversion data could play a role, enhancing the possibility of side reactions. Therefore, the order of addition of nickel and cerium is important and the best catalyst in terms of both activity and selectivity is obtained when Ni is deposited on the mesoporous support that already contains CeO₂ (Table 2).

3.2 Structural properties and morphology

The low-angle X-ray diffractograms of the SBA-15 support (uncalcined and calcined) and of the calcined Ni–Ce/SBA-15 and Ce–Ni/SBA-15 samples are shown in Fig. 2. After calcination of the fresh SBA-15 silica at 500 °C for 9 h, the structuring organic agent in the pores was fully eliminated and the silica walls were condensed. This resulted in a decrease in the unit cell parameter from 94.5 Å for the parent uncalcined SBA-15 (Fig. 2a) to 81.6 Å for the calcined SBA-15 support (Fig. 2b) as calculated from the (100) main diffraction peak. For all samples, the well-resolved peak indexed as (100) and the tiny (110) and (200) peaks are characteristic of the highly ordered p6mm hexagonal mesoporous structure. This structure was preserved after impregnation and calcination treatments (Fig. 2c, d), with however some collapse, as indicated by the lower intensity of the peaks.

In spite of this collapse attributable to successive impregnation in solution and calcination, high pore volumes and surface areas were still detected after Ni and Ce deposition (Table 2). All the Ce–Ni/SBA-15, Ni–Ce/SBA-15 and co-NiCe/SBA-15 samples showed a type-IV N₂ adsorption–desorption isotherm (Fig. 3) typical of mesoporous materials. The H2-type hysteresis loop observed with a smooth adsorption step and a sharper desorption step revealed non-uniform pore shapes and/or sizes. For all three samples, the BET surface area was in the range 550–650 m²·g⁻¹, and the total pore volume was between 0.3 and 0.4 cm³·g⁻¹ (Table 2). As for the conversion data, the values were the highest for Ni–Ce/SBA-15, in which Ce was impregnated first. Nevertheless, the pore diameter was around 3.4 nm whatever the order of Ni and Ce addition, suggesting that, if present, the effect of accessibility inhibition discussed above would rather consist in partial channel blocking than in channel diameter restrictions.

The structure and morphology of the samples are further depicted in Fig. 4. As can be seen from the MEB pictures (Fig. 4a, b), the SBA-15 grains of the support were small, almost spherical and homogeneous in size, with a length of about 0.3–0.4 μm and diameters in the range 0.4–0.5 μm. This morphology was maintained after catalytic study (spent catalysts). Moreover, TEM images taken both in parallel (Fig. 5a) and perpendicular (Fig. 5b) grain orientations clearly showed the highly ordered linear mesoporous system and hexagonal pore structure characteristic of the SBA-15 structure for all samples.

3.3 Nanoparticles dispersion and consequences on activity

The darker regions seen on the grey silica grains in the TEM images correspond to metal nanoparticles with higher electronic density than the silica support (Fig. 5). These nanoparticles often appeared regularly aligned along the channels (Fig. 5c) and they were well dispersed inside the pores, even after catalytic tests (Fig. 5d), whatever the sample.

Such preservation of the dispersion upon reaction was also illustrated by EDS measurements. Indeed, in spite of some heterogeneity in the Ni and Ce enrichment between grains, the local compositions were on average not different before and after catalytic tests, with for instance atomic Ni/Si ratios between 0.06 and 0.18 and Ce/Si ratios in the range 0.02 to 0.07 for the co-NiCe/SBA-15 catalyst. This shows the efficiency of the two-solvent impregnation method for dispersing rather homogeneously the nanoparticles inside the grains of the SBA-15 support.

Nevertheless, some differences in mean particle sizes were found depending on the order of metal impregnation. This is even better illustrated by the XRD data of calcined samples (Fig. 6 and Table 3). The broad signal observed on all the diffractograms at around 22° is due to the presence of silica with amorphous silica walls. The narrower peaks at 2Θ = 37°, 44°, and 63° visible above this signal correspond to nickel oxide nanoparticles (JCPDS 89-5881) with a face-centered cubic lattice (Fm3¯m) and a unit cell parameter equal to 8.35 Å. The peaks at 2Θ = 29°, 34°, 48°, 57°, and 70° are assignable to cerium oxide (JCPDS 81-0792) with a face-centered cubic lattice and a cell parameter of 5.41 Å. All these peaks are slightly asymmetric, indicating that the nanoparticles are not spherical, but rather slightly elongated, as also seen on TEM images.

The average nanoparticle sizes were calculated by applying the Scherrer equation to the (400) peak for NiO and to the (220) peak for CeO₂ (Table 3), assuming spherical shapes. It thus appeared that NiO and CeO₂ nanoparticles as small as 5–6 nm and 3–4 nm, respectively, were formed over the Ni–Ce/SBA-15 and co-NiCe/SBA-15 catalysts. With respect to NiO, this size is the same as for the Ce-free Ni/SBA-15 sample (NiO nanoparticles with diameter of 6 nm). It is worth noting that the crystalline domains thus calculated for NiO are greater than the diameter of the pores (around 3.5 nm, Table 1). However, this can be explained by the elongated shape of the particles (zoom in Fig. 5b) with diameter around 3–4 nm, smaller than their length (between 4 and 8 nm), forming worm-like aggregates along the channels (Fig. 5c). This also agrees with the slightly asymmetric XRD diffraction peaks for NiO (Fig. 6).

Therefore, contrarily to previous reports (references in Table 1), it is seen that the addition of Ce does not necessarily reduce the size of the nickel particles. In the present samples, this size was indeed the same in the three following cases: sample with only Ni, sample co-impregnated with Ni and Ce at the same time, and sample submitted to Ce impregnation first followed by Ni addition. In contrast, the addition of Ni first followed by Ce (Ce–Ni/SBA-15 sample) induced an enlargement of the NiO particle size up to 14 nm (Table 3). At this point, it has to be recalled that the deposited Ni-based nanospecies in this last sample were submitted, during preparation, to a second impregnation step in solution (for Ce deposition) followed by a new calcination step, which may have provoked some instability of the nanoparticles and additional sintering. In fact, this could be the main explanation for the lowest catalytic activity of the Ce–Ni/SBA-15 catalyst, in addition to the accessibility effects discussed above. Indeed, the Ni nanoparticles were on the contrary kept very small on the most active, stable and selective Ni–Ce/SBA-15 catalyst. Therefore, the impregnation of cerium first followed by addition of nickel is more favorable and this order of impregnation was interestingly also validated in a study on Ni/La₂O₃/γ-Al₂O₃ catalyst [18], in which lower Ni dispersions and catalytic performances were obtained when impregnating nickel first and then lanthanum.

Interestingly, further information on the spent catalysts was provided by electron microscopy study. Firstly, nickel dispersion after stability tests, carried out in reaction conditions for 12 h at 500 °C (after heating at 800 °C), was found as good as before reaction, as for instance visualized in Fig. 5d. Secondly, at this temperature, only very rare carbon nanotubes were observed, attesting to the absence of significant coke formation, in line with the above-mentioned close CH₄ and CO₂ conversions found on all catalysts.

Finally, additional insights into the influence of the order of Ni and Ce addition on the Ni active phase were obtained by TPR (Fig. 7 and Table 4). It is known that Ni²⁺ is directly reduced to Ni⁰ without passing by intermediate oxides [19]. Two main reduction processes can be distinguished in the ranges 200–400 and 400–650 °C. The first peak (200–400 °C) is relatively the less intense one for the Ce–Ni/SBA-15 sample, while it is the highest one for Ni–Ce/SBA-15 catalyst (Table 4). This suggests that easily accessible and more reducible NiO nanoparticles were present on the Ni–Ce/SBA-15 catalyst, in line with porosity data, thus, contributing to its highest performances compared to the other samples. On the contrary, the addition of Ce after Ni reduces the accessibility to the NiO nanoparticles as well as their reducibility (NiCe–Ni/SBA-15 catalyst).

3.4 Comparative catalytic performances

As indicated in the introduction, the preservation of high nickel dispersion combined with an easy access to the active sites appears as a prerequisite to ensure good catalytic performances in dry reforming of methane. In this respect, the smaller Ni-based particles obtained with or without pre-addition of Ce are more active than the larger ones formed when the Ni-containing sample is submitted to new impregnation/calcination for Ce addition.

However, regardless of the order of impregnation, the catalysts tested at medium temperatures in this study showed a better catalytic behavior than the mesoporous catalysts from the literature listed in Table 1. The activity performances at 600 °C of our catalysts and of Ni/SBA-15 catalysts from the bibliography containing or not Ce are compared in Fig. 8. At this temperature, more than 80% CH₄ and CO₂ conversions are achieved on our samples, while they are around 50% [5] and 60% [6] in bibliographic reports.

A first reason for the good performances of our catalysts lies in the formation of very small NiO nanoparticles; this is not only favored by the mesoporosity of the SBA-15 support, but also strongly ameliorated by the two-solvent method used for metal deposition that helped dispersing and stabilizing the active phase inside the channels, with almost no metal deposited outside. Thus, excellent nickel dispersions were obtained, with very small particles sizes below 6 nm, compared to higher sizes reported for comparable materials in the literature (Table 1). The high dispersion in the Ni–Ce/SBA-15 sample is also the result of the enhanced reducibility of the metal species in this sample (Table 4).

Besides, thermodynamic features have to be also considered, since the dry reforming reaction is known to be thermodynamically limited, thus usually requiring high temperatures to operate in pure reactant mixtures. Nevertheless, thermodynamic data calculated with the HSC chemistry software prove that 90% of CH₄ conversion can be achieved at a temperature of 575 °C, calculated with respect to the dilution conditions used in our study. Similarly, another study [20] reported an equilibrium temperature of 710 °C for 90% CH₄ conversion at the dilution conditions used in their study. Therefore, dilution effects could play a role in this reaction, in which the number of formed molecules is doubled in comparison to the reactants. Also, we may assume that the SBA-15 silica support can act as a membrane where the diffusion of H₂ outside the pores is higher than that of the remaining products and reactants, shifting the equilibrium to the side of products and increasing the conversions slightly above thermodynamics. Work is in progress to clarify these points.