Nickel Catalysts Supported Over TiO₂, SiO₂ and ZrO₂ for the Steam Reforming of Glycerol

Abstract

Ni-based catalysts supported on TiO₂, ZrO₂ and SiO₂ (in the form of mesoporous Santa Barbara Amorphous 15 (SBA-15) and amorphous dense nanoparticles), were employed in the steam reforming of glycerol. Each sample was prepared by liquid phase synthesis of the support followed by impregnation with the active phase and calcination at 800 °C or by direct synthesis through flame pyrolysis. Many techniques have been used to assess the physical chemical properties of both the fresh and spent catalysts, such as atomic absorption, N₂ adsorption/desorption, XRD, SEM, TEM, temperature-programmed reduction (TPR), X-ray photoelectron spectroscopy (XPS), Micro-Raman and FTIR spectroscopy. The samples showed different textural, structural and morphological properties, as well as different reducibility and thermal resistance depending on the preparation method and support. Some of these properties were tightly bound to catalyst performance, in terms of H₂ productivity and stability towards coking and sintering. A key parameter was the metal–support interaction, which strongly depended on the preparation procedure. In particular, the stronger the interaction, the more stable the metallic Ni clusters, which in turn lead to a higher catalytic activity and stability. Surface acidity was also taken into account, in which the nature of the acid sites was differentiated (silanols, titanols or Lewis acid sites). The characterisation of the spent catalysts also allowed us to interpret the deactivation process. The formation of multi-walled nanotubes was observed for every sample, though it was only in some cases that this led to severe deactivation.

Introduction

Different catalyst formulations have been proposed for the steam reforming (SR) of bioalcohols, in order to produce H₂ from renewable feedstocks.1–3 Ni showed to be one of the most promising active phases for such an application, especially if it was highly dispersed and thermally stabilised by the support. Coking and sintering are the main causes of catalyst failure. In spite of its lower activity and higher tendency to deactivation with respect to noble metal-based catalysts, Ni, if supported over Al₂O₃ in particular, is commonly used in industry for the steam reforming of hydrocarbons because of its low cost and wide availability. Ni is considered as a promising catalyst; this owes to its activity in CC, OH and CH bond cleavage and in the water–gas shift (WGS) reaction.

Glycerol is the main by-product in biodiesel production; therefore a glut in its market is expected in the next few years owing to a rise in biodiesel production. The pathway for the steam reforming of glycerol (GSR) is usually considered as similar to that of ethanol SR, even if this is not exhaustively understood. The reported mechanism for metal-catalysed GSR relates to the one previously proposed for hydrocarbon reforming.4, 5 GSR is an energy intensive process and CC cleavage is commonly considered to be the rate determining step. Glycerol can dehydrogenate and adsorb onto the metal, subsequent CC breakings lead to adsorbed CO that can be further oxidised through the WGS reaction, which is thermodynamically favoured at low temperature. The CO is then either converted to methane or desorbed. Methane can also be derived from CO₂ methanation.

The main reaction route is parallel to coking by CO decomposition and dehydration of the substrate to form surface olefin species, which may desorb, reformate or, regrettably, polymerise to form carbonaceous deposits.6 This is especially the case if a low water/glycerol feeding ratio is employed. Coke may also form as a result of the Bouduard reaction (CO disproportion), which may be thermodynamically favoured below 700 °C.7 Coke removal is possible by steaming and gasification, particularly at high temperature.2, 8

Detailed investigations into catalyst deactivation are available for methane or ethanol SR. The dehydration reaction pathway is mainly favoured by big Ni particles,9, 10 but also by strong surface acidity (e.g. in the case of Al₂O₃ supported samples). Considerable efforts have been devoted to develop non-acidic supports, such as MgAl₂O₄,2, 11 Ni_xMg_1−xO12 or MgO.13, 14 Up until now, these have mainly been adopted for the SR of ethanol. Unfortunately, such catalysts either showed unsatisfactory activity or, even when active, induced some scale-up problems owing to poor mechanical properties or formation problems. Alternatively, attempts have been made to limit surface acidity of common supports, for example, by impregnating alumina or zirconia with lanthanum oxide.6, 15

The best catalytic systems appear to be the ones in which the synergism between the metal and the support leads to metal stabilisation and decreases the rate of coke formation. MgO, CeO₂ and TiO₂ were used for their well-known ability to retard coke formation and to interact with metals that promote the catalytic activity as supports.16, 17 Furthermore, the modification of Ni/Al₂O₃ with Ce, Zr, Mg and La, brought about an improvement in hydrogen selectivity owing to surface Ni exposure in the case of Mg, steam activation for Zr, and stability of the metallic phase under the reaction conditions if Ce or La were added.18

Nevertheless, in spite of a growing number of papers on GSR, activity data are often contrasting and no clear relationships between the main physicochemical properties of the catalyst and its activity, selectivity and durability have been drawn.

Therefore, the aim of this work was to design, synthesise and characterise supported Ni catalysts for GSR.

TiO₂, SiO₂, mesoporous SBA-15 and ZrO₂ were chosen as supports owing to their different acidic and redox properties, and interaction with the metallic active phase. The catalysts were prepared by using different synthetic procedures, which were able to tune the thermal resistance and Ni dispersion. In particular, each sample was synthesised from the liquid phase, with the active phase deposited by impregnation, and calcined at 800 °C to impart proper thermal resistance for this high temperature application. In addition, a special preparation procedure, namely flame pyrolysis (FP) was employed to achieve high temperature stability and high metal dispersion. The latter technique proved effective for the preparation of thermally resistant samples for different high-temperature applications, such as the catalytic combustion of methane.19, 20 Furthermore, it led to unexpectedly high dispersion of the active phase in differently supported V-based catalysts even at relatively high loading (up to 10 wt %).21, 24 Therefore, it seems interesting to compare the effect of different preparation methods, which in principle lead to different dispersions of the active phase and interactions with the support.

Every sample was characterised by N₂ adsorption-desorption, temperature-programmed reduction/oxidation (TPR–TPO), atomic absorption (AA), X-ray photoelectron spectroscopy (XPS), SEM, TEM, XRD, FTIR and Micro-Raman spectroscopy. Attention was also paid to the characterisation of the spent catalysts, particularly with regards to the assessment of the nature and amount of coke deposits.

The main physicochemical properties of each sample have been discussed on the basis of the support and the preparation method. Activity and durability data for GSR were compared on the basis of catalytic properties.

Results and Discussion

Textural, structural and morphological characterisations

The textural properties of the samples prepared and the actual concentration of Ni are reported in Table 1.

Table 1. The main physicochemical properties of the freshly prepared samples. Some data have already been reported in Ref. 31

Sample	Preparation method	Ni loading^[a] [wt %]	SSA^[b] [m^2 g^−1]	Crystal size^[c] [nm]	Surface Ni^[d] [atom %]
[a] From atomic absorption analysis. [b] Based on the BET model. [c] NiO crystal size determined by the Scherrer equation and TEM between parentheses. [d] From XPS analysis.
NiTiL	TiO2 by precipitation with NaOH, calcined at 800 °C	6.7	4	31 (20–30)	0.9
NiSiL	SBA-15 calcined at 800 °C	8.9	309	21 (20–30)	0.8
NiZrL	ZrO2 by precipitation with NH3, calcined at 800 °C	8.8	43	18 (≈20)	1.2
NiTiF	FP	8.2	63	26 (20–30)	1.0
NiSiF	FP	9.6	211	18 (20–30)	1.4
NiZrF	FP	8.8	83	8 (10–15)	2.2

Different surface areas were obtained for the FP prepared samples, ranging from 60 to 200 m² g⁻¹, depending on the decomposition mechanism of the oxide precursor in the flame and on the type of solvent used.27, 28 Very low surface area was observed for NiTiL. In general, the surface area increased from titania to silica for both sets of samples and, accordingly, Ni concentration per m² decreased in the order TiO₂>ZrO₂>SiO₂.

The results of surface analysis (XPS) are summarised in Table 1; they are expressed in terms of relative atomic percentage.

The Ni fraction exposed on the support surface was always higher for samples obtained by FP. This was ascribed to Ni incorporation into the support during catalyst preparation.

TEM micrographs of the as-prepared catalyst NiTiL showed big particles that were relatively uniform in size (Figure 1 a). Some NiO nanoparticles were also evident, with a mean size of 20–30 nm (Table 1), in which most of the active phase was more dispersed on the support surface or was more likely to be incorporated into its structure. The latter hypothesis is consistent with the morphology of Ni clusters observed upon reduction at 700 °C under a 10 vol % H₂/He flow for 1 h, which mimics catalyst activation. The particle size of the NiTiL sample remained very uniform and was almost unaffected by thermal treatment; this reveals a suitable thermal stability of the support (Figure 1 b), also confirmed by SEM (Figure 2 a,b). By contrast, Ni particle size became less uniform. The enlargement of the previously existing metal oxide clusters upon reduction and the formation of much smaller Ni particles that were well-dispersed over the surface are shown in (Figure 1 b); this was likely to be formed from sintering of very dispersed NiO. The particle size of the NiSiL sample was even larger than that of NiTiL, whereas the ZrO₂ particle size was approximately 20–30 nm in the NiZrL sample.31

Figure 1. TEM analysis of the NiTiL sample a) as-prepared and b) reduced under 10 vol % H₂ flow at 700 °C for 1 h. Scale bars: 100 nm.

Figure 2. SEM analysis of samples a) NiTiL, b) NiTiL reduced under 10 vol % H₂ flow at 700 °C for 1 h and c) NiTiF. Scale bars: 2 μm.

Particles with a high homogeneity constituted the NiTiF sample, with sizes up to 30 nm as detailed by TEM (Figure 3 a). Similar dimensions were found for all the FP-prepared samples (Figure 3 a–c). A slightly lower homogeneity of the NiO particles was observed for NiSiL (Figure 3 d).

Figure 3. TEM micrographs of selected samples: a) NiTiF, b) NiSiF, c) NiZrF, d) NiSiL and e) NiZrL. Scale bars: 100 nm.

EDX analysis also confirmed the Ni loading with respect to atomic absorption, and repeated analyses in different zones demonstrated a rather uniform distribution of the active phase. The latter conclusion has also been supported by several maps that reveal an even incorporation of Ni into each oxide matrix in the case of FP-prepared samples.

Ni or NiO particle size that is similar to catalyst particle size has sometimes been found, which is likely to indicate a full coverage of the nanoparticle by the metal, as observed for instance in Figure 3 c, e relative to both zirconia-supported samples.

The TPR technique was employed to identify the different species present in the catalysts according to the different reduction temperatures. Reducibility data may help to evaluate the strength of interaction between the active phase and the support. In general, one may say that a low reducibility of the active phase indicates a strong interaction with the support. This may possibly lead to dispersed metal clusters that are stable even at high temperature. It is also well-known that metal/support interaction increases with calcination temperature. It was also reported that a lower reduction temperature may be ascribed to bigger particle size. Large Ni particles expose a small interface with the support surface, which leads to a weaker interaction and, thus, easier reduction.32 However, there is not full agreement on this point in the literature.

TPR–TPO–TPR cycles were performed. At first, the FP preparation procedure induced at least partial incorporation of Ni into the support matrix, which possibly led to a mixed oxide phase. It may be supposed that some reconstruction of the oxide may occur during metal reduction, so a second TPR enabled us to investigate the activated catalyst. Furthermore, activation may induce redistribution of the active phase or sintering in each sample.

The TPR pattern of NiTiL (Figure 4) revealed a sharp reduction peak, which suggests the presence of only one type of Ni-containing species. No “free” NiO, that is, none that interacts with the support was found (unsupported NiO has a reduction temperature of approximately 280 °C).33 By contrast, the presence of peaks at higher temperature reveals significant interaction of NiO with titania.34 Besides, such a high reduction temperature is compatible with the reduction of NiTiO₃,35 as revealed by XRD analysis (vide infra). This may be of utmost importance in stabilising Ni and achieving satisfactory catalytic activity. The second TPR pattern revealed the reversibility of the reduction treatment, but a slight shift of the reduction temperature towards lower values may indicate a lower interaction with the support or, possibly, some Ni sintering.

Figure 4. TPR–TPO–TPR cycle for the FP-prepared NiTiL.

The first TPR run of NiTiF (Figure 5) showed a series of broad and overlapping peaks that spanned a wide temperature range (300–700 °C), which could owe to the reduction of free surface NiO, bulk NiO and NiO, which have, respectively, stronger interactions with the TiO₂ surface, as discussed above. Metallic Ni can be oxidised at 300 °C, however, the second TPR showed a much simpler pattern, with only one peak centred at approximately 600 °C. This may indicate a stronger interaction with the support with respect to the as-prepared sample, and suggested the need of higher reduction temperature to reduce Ni after the first redox cycle. This feature also indicated the irreversibility of the process with respect to the transformation of “free” NiO particles into more uniform species that are strongly bound to the support, maybe as a mixed oxide. Catalyst NiTiF was found to be a bit more reducible than its above-reported homologue, even during the second reduction cycle.

Figure 5. TPR–TPO–TPR cycle for the FP-prepared NiTiF.

The TPR profile of the NiSiL catalyst31 showed different reduction zones, which indicates the presence of “free” NiO particles that don’t interact with the support (sharp peak) and the presence of NiO particles with stronger interactions with the silica support (very broad)36, 37 After oxidation, the second TPR run indicated an increased reducibility.31, 32 A broad and featureless peak ranging between 250 and 600 °C was observed with sample NiSiF, that shifted by approximately 50 °C at a higher temperature during the second TPR analysis, but remained unaltered in its shape. The metal was oxidised back at approximately 300 °C, as seen for the titania-supported catalyst. These data were interpreted as showing an almost perfectly reversible reduction process and very broad heterogeneity of Ni oxide sites. Nevertheless, in contrast with sample NiSiL, some increase of the metal support interaction after activation could indicate the retention of a high Ni dispersion.

The NiZrL catalyst had a higher temperature peak (with its maximum at 655 °C), which can be assigned to NiO particles that have a strong interaction with the ZrO₂ surface, and a peak at lower temperature (shoulder at approximately 450 °C), which is attributed to NiO species that have weak interactions with the support.38, 39 After oxidation, the sample became more reducible.

Also, for the FP NiZrF sample, two distinguishable NiO species may be found, which appear at lower temperature than those of NiZrL (approximately 350 °C and 380–510 °C). The subsequent TPO showed that Ni oxidation occurred at approximately 240 °C, whereas the last TPR run evidenced that the distinction between different Ni species was retained, though the former peak was less intense than the latter. Furthermore, the second peak became more intense and shifted towards higher temperature, which indicates the formation of stronger Ni–support interactions after the first treatment.

In general, NiO species impregnated over supports prepared by precipitation seem less reducible when fresh, compared with those synthesised by FP, and a tentative reducibility scale may be drawn (SiO₂>ZrO₂>TiO₂). The strength of the metal–support interaction is expected to increase in the opposite direction. However, during the second reduction cycle, an increase in the metal-support interaction strength is seen for the FP samples; this can probably be attributed to a reconstruction of the metallic crystallites after activation. By contrast, the reducibility increased after activation for the samples prepared in the liquid phase, which indicates a lower metal–support interaction.

XRD analysis of the as-prepared samples was performed to identify the different phases present in the samples (Figure 6). Approximate calculation of the NiO crystal size was obtained from the Scherrer equation (Table 1) and compared with TEM data.

Figure 6. XRD diffractograms of a) NiTiL, b) NiTiL after activation, c) NiTiF, d) NiSiL, e) NiSiF, f) NiZrL and g) NiZrF.

By comparing the diffractograms of the titania-supported samples with literature data, one may conclude that the NiTiL sample was mainly constituted by rutile, plus a fraction of ilmenite (NiTiO₃), as confirmed by the TPR profiles (vide supra). By contrast, the main component of NiTiF was rutile, with lower contributions of anatase and NiTiO₃. After activation, both the anatase and mixed oxide phases almost disappeared and the reflections of metallic Ni formed. The XRD pattern of the NiSiL sample confirmed the mesoporous structure of the SBA-15 support. By contrast, the NiSiF sample was amorphous, and only very broad NiO_x reflections with low intensity appeared. Therefore, it was not possible to unequivocally attribute the Ni-containing phase. The structure of the NiZrF sample was mainly tetragonal, which was expected as this phase is stable above 1100 °C. By contrast, NiZrL showed a more complex XRD pattern owing to the coexistence of both the tetragonal and monoclinic phases.

The average NiO_x crystal size, calculated from the Scherrer equation for samples with the same support was lower for the FP-prepared catalysts than for their homologues prepared by impregnation. This confirms the higher Ni-dispersion of the former, in spite of the very high calcination temperature attained in the flame. In general, metal dispersion increased in the order TiO₂<SiO₂<ZrO₂. The smallest Ni clusters were observed for NiZrF. This is related to the very good incorporation of Ni into the support, as testified by the NiO reducibility scale. The SiO₂ supported samples turned out to be the most reducible, that is, the ones with poorly stabilised Ni oxide particles that are possibly more prone to aggregation. Lower reducibility was progressively observed for ZrO₂ and TiO₂. However, the latter was characterised by the formation of a mixed oxide (thus justifying the much lower reducibility in the first reduction cycle). After reduction, rutile was the main TiO₂ phase, which usually has a low capability of hosting metals (Figure 6). Therefore, higher dispersion was achieved after activation for the ZrO₂ supported samples, especially if synthesised by FP.

Additional structural information may be drawn from skeletal FTIR spectra. The typical characterisation peaks for SiO₂-based materials at =1100 cm⁻¹ (shoulder at 1250 cm⁻¹), 800 and 450 cm⁻¹ were observed for the NiSiF sample and a broad peak was observed for TiO₂ at =680 cm⁻¹(NiTiF sample).^[40] The spectra of the zirconia-based samples were consistent with the formation of monoclinic ZrO₂ (band at =745 cm⁻¹), together with the most abundant tetragonal phase, for which the peaks are overlapped with the monoclinic phase in the low frequency region.

Typical features of the NiTiO₃ structure appeared in the skeletal IR spectra of NiTiL, characterised by peaks at =530, 410 and 320 cm⁻¹; these are in agreement with literature data.41 The shoulder at =690 cm⁻¹ can be attributed to the rutile phase, which is also detected by XRD. Another broad absorption at approximately =610 cm⁻¹ was detected for the NiTiL sample, but not assigned.

In summary, with the same support: the crystal structure was comparable for the samples prepared with different methods except for silica; Ni dispersion was always higher for the FP prepared samples, which also denoted increased reducibility after the second reduction cycle. If one compares different supports, the highest Ni dispersion was exhibited by zirconia.

FTIR analysis

Following a thermal treatment at 500 °C, pure powder spectra of activated and hydrogen-reduced NiTiF catalysts showed a clean surface that was almost completely dehydroxylated. Low temperature CO adsorption over the same sample gave rise to several IR bands in the CO carbonyl spectral region (Figure 7 a). As widely reported in the literature and summarised by Hadjiivanov et al., bands at =2184 cm⁻¹, shifting to =2191 cm⁻¹ at decreasing coverage, characterise carbonyl species over acidic coordinatively unsaturated Ti cations acting as Lewis sites of different strength (Ti⁴⁺).42 Another weak component at =2165 cm⁻¹, detected here as a broad shoulder tailing to =2139 cm⁻¹, was attributed to an interaction between CO and residual OH groups. These assignments were also confirmed by the behaviour of the bands after degassing: bands attributed to carbonyl interactions with Lewis sites were more resistant, whereas bands attributed to H-bonded CO, readily disappeared, as long as the weak bands attributed to the isolated OH stretching mode were restored in the spectra. It can be observed that only a few different bands related to Ni species can be identified in the region =1850–2130 cm⁻¹. The weak band at 2128 cm⁻¹ indicated the presence of geminal dicarbonyls over Ni⁺ ions, to which the associated band located at approximately =2090 cm⁻¹ could be masked by the strong band at =2064 cm⁻¹. This band disappeared only after evacuation at room temperature, which is in agreement with literature findings that reported Ni⁺–CO complexes are much more stable than Ni²⁺–CO complexes.

Figure 7. FTIR subtraction spectra of surface species arising from CO adsorption over: a) NiTiF reduced catalyst at 77 K and after warming to RT; the spectrum of CO adsorbed over pure titania is shown for comparison; b) NiTiL reduced catalyst at 77 K and after warming to RT. The activated surface has been subtracted.

Two intense bands at =2064 cm⁻¹ and 2036 cm⁻¹, with similar resistance to degassing, were observed in the region =2000–2100 cm⁻¹. These features were associated with CO linearly adsorbed onto Ni⁰ small clusters with different crystal face exposure, size and possibly a different metal–support interaction. An additional broad band at =2003 cm⁻¹, which shifted to =1992 cm⁻¹ at decreasing coverage, was detected in the region =2000–1980 cm⁻¹. This band corresponds to bridging CO over larger metal Ni⁰ centres. No bands associated with the presence of residual Ni²⁺–CO species of NiO or nickel titanate were detected following the reduction treatment. Thus, Ni was broadly reduced at the catalyst surface. On the other hand, the heterogeneity of the Ni species is in agreement with TPR results, which highlights the presence of at least three kinds of Ni metal particles, metal clusters strongly interacting with the support, and Ni ions with different reduction abilities. In fact upon degassing and warming, CO₂ was formed (band at approximately =2350 cm⁻¹) as a result of CO oxidation by Ni ions.

Following pivalonitrile (PN) adsorption (Figure 8 a), two complex bands were detected in the CN stretching region at =2278 and 2247 cm⁻¹ (with a shoulder at lower frequency). The former was attributed to an interaction between PN and the medium-strength Lewis sites (Ti ions), whereas the latter, which was strongly diminished after degassing at room temperature, was attributed to PN that was H-bound to residual exposed OH groups. The low relative intensity of this band was in agreement with the low degree of hydroxylation of this surface. Correspondingly, only a weak negative band could be detected in the OH stretching region of the subtraction spectrum (inset in Figure 8 a).

Figure 8. FTIR subtraction spectra of surface species arising from PN adsorption at RT and following degassing at increasing evacuation temperatures (ev) over: a) NiTiF reduced catalyst, inset: OH stretching region; b) NiTiL reduced catalyst. The activated surface has been subtracted.

Low temperature CO adsorption over NiTiL (Figure 7 b) gave rise to bands in the CO spectral region: at =2156 cm⁻¹ (an interaction between CO and OH groups that was possibly overlapped with CO adsorbed over Ni ions; this completely disappeared after degassing at low temperature), and at =2130 cm⁻¹ (weak), together with the band at =2095 cm⁻¹ (symmetric/asymmetric stretching modes of poly(di)carbonyl species stable at low temperature and characteristic of a dispersed Ni fraction). Bands at =2060 cm⁻¹, and possibly 2040 cm⁻¹, were assigned to carbonyl species on Ni⁰ crystals, as well as the band at =2020 cm⁻¹ (shoulder), for which the lower frequency suggested the assignation to larger Ni⁰ particles. No bands attributed to CO coordinated over exposed Ti centres can be detected (approximately = 2190 cm⁻¹) and, correspondingly, no bands attributed to PN interacting with Lewis centres were observed.

As a matter of fact, PN adsorption led to the detection of only one band (Figure 8 b), owing to the presence of H-bound species, which disappears after degassing at room temperature. This can be attributed to the collapse of surface area in these samples and also to the formation of a titanate phase, as evidenced by XRD and skeletal FTIR spectra (vide supra).

Thus, we detected carbonyl groups that were already coordinated over Ni metal particles after reduction at 500 °C, together with a fraction of Ni ionic species (Ni that strongly interacts with the surface and titanate species). This effect is in agreement with the TPR data, which indicates that the main reduction peaks appear at temperatures above 600 °C.

A detailed description of FTIR characterisation of the remaining samples has been provided elsewhere.31 As for the NiSiF FP-prepared sample, the typical features of silica were evident and remained almost unchanged after reduction. CO adsorption over this sample evidenced the presence of relatively extended metal particles that were characterised by bridging carbonyl species. Moreover, the formation of Ni⁺–polycarbonyl species was also suggested, although these were strongly reduced with respect to the NiTiF catalyst, and were present alongside some atomically dispersed Ni. This heterogeneity of Ni species facilely disappears after the two TPR reduction cycles.

PN adsorption mainly evidenced the presence of silanol groups (H-bound species that were weakly held) and of Lewis acid centres, which are likely attributed to the metal phase.

The FTIR spectra of the NiSiL catalyst did not evidence isolated free silanol bands in the activated sample spectra after the reduction treatment, but only highly disordered surface hydroxyl groups. CO adsorption confirmed the presence of silanols (weak physisorption) and extended Ni metal particles. PN adsorption confirmed the presence of weakly H-bound species only, without Lewis acidity ascribed to Ni, contrarily to sample FP. Catalyst NiZrF was characterised by very poor transmittance, which probably owes to the presence of reduced metal. Zr⁴⁺ Lewis acid sites were detected after both CO and PN adsorption.

Finally, NiZrL had exposed Zr ions acting as Lewis acid sites and quite large Ni metal particles.

Therefore, medium Lewis acidity attributed to exposed support ions, characterises both the titania- and zirconia-based catalysts. In the case of the former, this is accompanied by some weakly acidic OH groups, which are available for H-bound formation with probe molecules. By contrast, silica supported catalysts showed weak acidity owing to the presence of silanols, with Lewis acidity only induced by Ni particles for the Ni/SiO₂ FP catalyst.

A relatively heterogeneous population of Ni species was detected over the hydrogen-reduced samples (500 °C). This consisted of residual Ni⁺ ions, Ni clusters and larger metallic Ni particles, according to the metal reducibility scale extrapolated by TPR data.

Unfortunately, the FTIR characterisation of the spent catalysts was not possible owing to the presence of quartz powder, which diluted the sample, and coke deposits.

Catalytic activity for GSR

The catalytic activity was tested under isothermal conditions at 650 °C for 20 h on-stream after activation at 700 °C. Table 2 reports the main data on catalytic performance after 5, 10 and 20 h on-stream, that is, glycerol conversion, H₂ yield and selectivity to CO₂, CO and CH₄ (see the Experimental Section).

Table 2. Results of catalytic test runs for GSR.

	Fi [%]
	5 h on-stream					10 h on-stream					20 h on-stream
Sample	XC	Y	CO2	CH4	CO	XC	Y	CO2	CH4	CO	XC	Y	CO2	CH4	CO
NiTiL	97	88	85.9	2.7	11.4	99	89	80.5	2.7	16.8	98	82	67.2	2.6	30.2
NiTiF	100	97	91.5	0.7	7.7	100	95	89.6	1.2	9.3	100	85	86.7	0.8	12.5
NiSiL	87	82	87.8	1.3	10.9	88	81	86.0	1.1	12.9	100	87	70.2	1.5	28.3
NiSiF	97	82	75.9	1.5	22.6	90	71	62.7	1.7	35.6	71	55	53.4	2.4	44.2
NiZrL	95	89	88.1	0.0	12.0	97	92	92.2	0.0	7.8	82	80	91.1	0.0	8.9
NiZrF	87	80	89.6	0.0	10.4	90	87	92.8	0.0	7.2	87	84	93.5	0.0	6.5

A blank test at 650 °C on quartz (0.20 g, 0.355–0.500 mm), revealed a 3 % glycerol conversion to gaseous products after 3 h on-stream; the reactant was found in the condensed products and selectivities of 87 % to CO, 13 % to CH₄ and nil to CO₂ were also found. The test evidenced that without any catalyst, thermal decomposition was negligible at the selected temperature and almost no WGS reaction was promoted.

NiTiL and NiTiF showed comparable initial activity, with a glycerol conversion higher than 95 %, and a low selectivity to CO, even if the FP-prepared sample maintained a higher activity, a better CO and CH₄ selectivity and a higher H₂ productivity for the whole duration of the test. This is in agreement with a higher maintained dispersion of Ni after activation of NiTiF. Generally, the selectivity to CH₄ of the titania-supported samples remained stable for the whole duration of the test. In particular, if one considers the performances of NiTiL, despite the stable conversion of glycerol for 20 h on-stream, the WGS activity markedly decreased as illustrated in Figure 9, with a progressive increase in the selectivity to CO and a corresponding decrease of that to CO₂.

Figure 9. GSR at 650 °C on NiTiL, showing glycerol conversion (♦), hydrogen yield (▪) and CO₂ (▴), CO (×) and CH₄ (✶) fractions in the gas phase.

By comparing the performance of NiTiL with that of the same catalyst calcined at 500 °C instead of 800 °C and mainly constituted by anatase,43 it can be observed that NiO species supported on anatase were much more reducible than the catalysts supported on rutile and the mixed oxide phase. The samples calcined at 500 °C were also far less stable and had lower activities during the activity testing. Therefore, the formation of a strong metal–support interaction seems key to achieving a suitable Ni dispersion and activity. The metal–support interaction may be increased with an increase in the calcination temperature, but proper thermal stability of the formed Ni clusters is also needed to ensure stable performance with time-on-stream.

The NiSiL sample showed a stable conversion and hydrogen yield up to 20 h on-stream. Indeed, both parameters reached a stable value at 10 h on-stream (87 % and 82 %, respectively), to increase to 100 % and 87 % after 20 h on-stream. However, also in this case, CO selectivity increased during the test, revealing a progressive depletion of WGS activity, despite the excess of water. This feature could be attributed either to the loss of support specific area during the reaction or to Ni sintering, which owes to the huge amount of water vapour at high temperature combined with the lowest metal–support interaction within this series of catalysts.

Even though it is more stable than other silica mesoporous ordered supports,39, 44 SBA-15 may undergo a collapse of the structure under our reaction conditions, especially if it is not stabilised at high temperature. It is generally accepted that water is activated by the support during steam reforming of hydrocarbons.45–47 Therefore, a progressive decrease in its specific surface area (SSA) can result in a lower water activation. Another explanation could be that the active metallic sites inside the pores become progressively less accessible owing to a partial collapse of the mesoporous structure. The resulting decrease in metallic surface area affected the activity of the catalyst, but it cannot explain why the activity decrease only affected reactivity for the WGS reaction. Additionally, CH₄ could not be completely reformed under the present reaction conditions.

The conversion of glycerol achieved with NiSiF was higher than 95 % after 5 h on-stream, but it continuously ceased to reach approximately 70 % after 20 h on-stream. This sample was characterised by Lewis acidity, which is attributed to Ni. It is likely that this feature induced coke deposition over the active sites, progressively deactivating them.

ZrO₂ demonstrated an interesting support for this reaction, especially if one takes into account its WGS activity and selectivity to methane. Indeed, almost complete conversion was attained for NiZrL, though with some deactivation after 10 h on-stream. By contrast, even if the FP-prepared NiZrF showed a lower activity (≈87 % glycerol conversion), it demonstrated a surprising stability till the end of the test. This is in line with the highest C balance obtained with this catalyst during parallel tests for ethanol steam reforming at 500 °C, in which severe coking ruled out many other samples.31 Furthermore, complete methane reforming and very low selectivity to CO were achieved with both the zirconia-supported catalysts. This may be attributed to the ability of this support in the activation of steam if it interacts with metals, as already reported by Iriondo et al. for a ZrO₂ doped Ni/Al₂O₃ catalyst for GSR.18

All the catalysts showed similar initial activity, except for a lower conversion achieved with NiZrF, and this remained constant for the whole duration of the test.

Such results may be compared with similar samples calcined at 500 °C reported elsewhere.43 For instance, the same SBA-15-supported sample calcined at 500 °C, was much less active and deactivation led to a conversion of 48 % after only 20 h on-stream. Similarly, titania-supported catalysts calcined at 500 °C were found to be completely inactive for both the present reaction43 and for the steam reforming of ethanol. This evidenced how catalyst activity tightly depended on the calcination temperature of the support. Only the samples calcined at high temperature or synthesised by FP reached almost full conversion, coupled with a sufficient stability. A strong metal–support interaction, depending on support nature or whether attainable after calcination at high temperature, could account for this behaviour. The same effect of stabilisation was evidenced for the Pt/ZrO₂ system and it was associated with a strong Pt–Zr^δ+ interaction, which resulted in the formation of ZrO_x species on the Pt surface.48–50 In the case of a weak metal–support interaction, metal sintering may occur, thus leading to a strong decrease in the available metallic surface area and therefore of the catalytic activity.

A suitable stabilisation of the metallic phase by the support could also lead to a catalyst less prone to deactivation. The most effective parameter that determined the catalyst performance appears to be the reducibility of the Ni species, as a measure of their interaction with the support. The most active and stable catalysts were the ones that showed lower reducibility, which indicates a stronger interaction between the metal and the support. This parameter increased from silica to titania, and an element of the parameter that is important for attaining an even stronger interaction is the Ni incorporation in the support or the formation of a mixed oxide phase during the synthesis. For instance, the presence of a mixed oxide (NiTiO₃) was evidenced by XRD for both the titania-supported samples.

Other parameters, such as the Ni exposure or the textural properties of the support (SSA and pore morphology), seem to not deeply affect the catalyst performance.

As a matter of fact, however, the most critical aspect for GSR is catalyst durability. In the present study, the deposition of coke was relatively limited because of the huge excess of water in the feed (the catalytic tests were virtually performed in the no carbon deposition region with respect to temperature and water/glycerol ratio51), but GSR is a severe reaction that suffers from low H₂ productivity and strong deactivation compared with ethanol or methane reforming,52–54 owing to the high reactivity of glycerol.

During steam reforming, the catalyst may deactivate by the formation of olefins that can polymerise over the surface and possibly cover the active phase. Coke precursors can derive from dehydration, dehydrogenation and cyclisation of glycerol.55 One way to decrease coking is to operate with a high water to glycerol ratio, as in the case of the present study. This also improves hydrogen productivity, by increasing the WGS activity and methane reforming.

Ni particle size is often related to catalyst deactivation by coking and extensive studies have been conducted for the steam reforming of methane.10, 56–59 Usually, a low coking rate is exhibited by small Ni particles. In the present case we may observe that the most stable behaviour has been shown by NiZrF, which is indeed characterised by the smallest particle size. By keeping the support constant, catalyst stability versus time-on-stream was always higher with lower particle size, that is, for the FP-prepared materials, except for NiSiF. For the latter, the acidic character of the active phase may be invoked to explain the depletion of activity with time-on-stream.

The rate of deactivation may also be correlated with the acidity of the catalyst. As an example, the acidity of the support caused a rapid deactivation of the Ni/Al₂O₃ catalyst during glycerol reforming.60 The addition of La₂O₃ or CeO₂ to the Ni/Al₂O₃ catalyst was found to be effective for hindering deactivation. The basic character of CeO₂ was also related with the inhibition of undesired reactions61 and the same positive effect on H₂ and CO₂ productivity, owing to a reduction in acidity of Al₂O₃ by the addition of La₂O₃, which was evidenced by Iriondo et al.62 The employment of oxides, such as La₂O₃, CeO₂, TiO₂, ZrO₂, MgO, as supports or as dopants, allowed for superior and stable catalytic performances in GSR.18, 63, 64 Such oxides are also well-known for stabilising the Ni particles; they strongly interact with the metal. In Ni/Al₂O₃ catalysts doped with La, Ce, Mg, Zr, the “decoration” of metallic particles by dispersed species from the support has been also reported, which improve the performance of the catalyst.

The NiSiF sample was characterised by the presence of silanols and by Lewis acid sites attributed to Ni. By contrast, only silanols were observed for NiSiL. Both Lewis acid sites and silanols may in principle be subject to coke deposition. However, coking in the latter case only occurs on the support, and severe catalyst deactivation with time-on-stream may be ruled out. Indeed, sample NiSiL showed a stable behaviour during the whole test, even with increasing conversion. On the contrary, the FP sample showed full glycerol conversion at the beginning of the test, but it rapidly deactivated. This is attributed to progressive coking of the Ni active sites showing stronger Lewis acidity, provoking a decrease of catalytic activity until their complete deactivation. Based on the characterisation of the spent catalysts, we interpreted the deactivation of the FP sample by encapsulation of the active phase in a network of carbon nanotubes (vide infra). A similar conclusion is likely to be drawn for sample NiZrL deactivating after 10 h on-stream, and possibly to explain the low initial activity of NiZrF. However, it should be underlined that for the latter samples no evidence of Lewis acidity of Ni can be extrapolated from spectroscopic data because Lewis acidity of the metal would be covered by PN interacting with Zr Lewis acid sites.

With the TiO₂ and ZrO₂ supports, which induce medium Lewis acidity, coking occurs in case on the support (less critical) and there is no deactivation of the active phase.

To evaluate the amount of coke deposited on the samples during the catalytic test, thermogravimetric analysis (TGA) in air flow up to 1000 °C was performed on the spent catalysts after 20 h on-stream. The following scale of weight loss (always below 10 wt %) was obtained, almost regardless of the synthetic method: TiO₂<SiO₂<ZrO₂. It was only in the case of the ZrO₂-supported samples that different amounts of coke have been observed, which depend on the preparation procedure. Indeed, the weight loss was three times higher for NiZrF than for NiZrL.

Micro-Raman spectra have also been recorded over the spent catalysts (Figure 10). The typical D and G bands appeared and were attributed to the presence of multi-walled carbon nanotubes (MWCNTs).65, 66

Figure 10. Micro-Raman analysis of spent catalysts a) NiZrL, b) NiTiL, c) NiSiL, d) NiZrF, e) NiTiF and f) NiSiF after 20 h on stream. The characteristic D and G bands of carbon nanotubes are evidenced.

The attribution of most carbonaceous species to MWCNTs has been also confirmed by FE-SEM analysis, which clearly shows the morphology of carbon deposits over spent samples, as exemplified in Figure 11. Carbon nanotubes were evident on every used catalyst, with slightly different morphologies and distributions that depended on the support and, secondarily, on the preparation method. The nanotubes’ diameter ranged between 35 and 50 nm for most samples, that is, smaller with respect to those observed from NiTiF (Figure 11 a), for which the size was 55–75 nm. Furthermore, the surface of NiSiF was more fully and uniformly covered by the carbonaceous residue than other samples and the diameter was the lowest (35–40 nm).

Figure 11. FE-SEM analysis of spent catalysts; size of scale bars in parentheses: a) NiTiF with detail of nanotube size (100 nm), b) NiTiF (300 nm), c) NiSiL (200 nm), d) NiSiF (200 nm), e) NiZrL (100 nm) and f) NiZrF (200 nm).

TEM analysis for all the used samples after 20 h on-stream confirmed the presence of MWCNTs (Figure 12 a) and in addition allowed us to evidence possible sample sintering. The used NiTiL sample (Figure 12 b) somehow showed bigger support particle size (ca. 300 nm) with respect to the fresh one and increased Ni particle size (30–50 nm) with respect to the activated sample (Figure 1 b). Though the formation of MWCNTs can be noticed, the metal remains mainly exposed; this explains the sufficient durability of this sample. Similar conclusions also hold for NiTiF, for which the overall and Ni particle sizes increased to approximately 150–300 nm and 25–40 nm, respectively (Figure 12 c, d). The Ni size was therefore a bit lower in the case of the spent NiTiF catalyst with respect to NiTiL, though this possibly explains the better resistance for time-on-stream, especially for the WGS reaction.

Figure 12. TEM analysis of spent catalysts after 20 h on-stream; size of scale bars in parentheses: a) detail of multi-walled nanotube (100 nm), b) NiTiL (500 nm), c) NiTiF (500 nm), d) NiTiF (100 nm), e) NiSiL (200 nm), f) NiSiF (100 nm), g) NiSiF (500 nm), h) NiZrL (200 nm) and i) NiZrF (500 nm).

The mesoporous structure of NiSiL seems roughly maintained even after use; though in some zones the sample seemed amorphous. This would confirm the possible partial collapse of the above described structure as one of the causes of deactivation for the WGS reaction. However, some metal sintering also occurred, which led to non-uniform metal particle size (Figure 12 e), which is another possible reason for the decreasing performance for the WGS reaction. By contrast, poor accessibility to the metal phase was observed for NiSiF, in which the metal particles were mostly embedded into the MWCNTs (Figure 12 f, g). This extensive coking, mainly involving the active phase, could explain the rapid deactivation of this sample with time-on-stream.

Finally, MWCNTs also formed over both the zirconia supported samples, but the metal seems to remain dispersed and accessible, which justifies the good activity even after 20 h on-stream. In particular, NiZrL seems more thermally resistant as the support, which reached an average 40–50 nm particle size and seemed to remain uniformly covered by the metal (Figure 12 h). By contrast, the mean Ni particle size for NiZrF reached approximately 20 nm (Figure 12 i).

It can be concluded from these data that coking, primarily occurring on the most acidic Zr^IV Lewis sites, negatively affected the catalytic activity during the first hour on-stream for the sample prepared by FP. However, after initial deactivation, stable catalyst performance was achieved. On the contrary, a depletion of activity with time-on-stream was observed with sample NiZrL, Furthermore, it can be also concluded that such Lewis acid sites (owing to the metal or the support) are not involved at all in the WGS reaction, as their disappearance after more or less pronounced coking did not alter the selectivity to CO and CO₂.

In conclusion, severe catalyst deactivation may be related to metal sintering or to Lewis acidity owing to Ni active sites. Silanols and Lewis acid sites on the support are likely to cumulate coke without any depression of glycerol conversion. Residual methane seems less related to catalyst deactivation, as its concentration in the out-flowing gas only slightly increased for sample NiSiF (2.4 %), if deactivation occurred, otherwise, this remained constant.

Finally, some samples showed an increase in CO concentration in the outlet gas, which is attributed to the depletion of activity for the WGS reaction. In the case of NiSiF, this could be correlated to the general loss of activity of this sample, but such explanation cannot hold for other catalysts, in particular NiTiL, NiSiL and, to a much lesser extent, NiTiF. A first explanation may be related to some metal sintering, observed for the spent catalysts. The higher the Ni particle size, the lower the activity for the WGS reaction. Indeed, the zirconia supported sample, which showed high metal dispersion even after use, suffered less of WGS activity depletion.

Possibly, another tentative explanation for WGS activity loss versus time-on-stream recalls the presence of OH groups on the surface of activated catalysts. These were detected by FTIR analysis for both the silica supported samples and NiTiL, much less so for sample NiTiF, and not at all for ZrO₂. This parameter perfectly mimics the loss of WGS activity. It may be correlated to the support ability in water activation. Indeed, in a parallel investigation on the steam reforming of ethanol,^[31] we noticed a higher C loss owing to coke deposition on supports that were characterised by the presence of surface OH groups; this is much lower if Lewis acid sites were predominant. Therefore, we may conclude that coke deposition on the support does not directly affect glycerol conversion, as Ni exposed sites may still be active, provided that they do not deactivate for other reasons. However, the progressive coking of the support surface may limit water activation, with a predominantly negative effect on WGS activity.

Conclusions

Titania, silica and zirconia supported catalysts were prepared by different procedures, which induced variable SSAs, metal dispersions and metal–support interactions. All the samples were used for the steam reforming of glycerol (GSR).

NiZrF was found to be one of the most promising. Indeed, in spite of its low glycerol conversion (87 %), it was fully stable for the whole duration of the test. Furthermore, it had a high activity for methane reforming and for the water–gas shift (WGS) reaction, which led to selectivities of S=0 and S_CO<10 % for the whole test duration.

The metal–support interaction was found to be very important in determining the catalytic activity and stability of the catalyst. Indeed, the characterisation of the spent catalysts revealed that MWCNTs formed for every catalyst, but only in some cases (for instance for NiSiF) did they induce poor access to the active phase. Deactivation was also correlated to some metal sintering, sometimes evidenced by TEM images. Deactivation was also correlated to catalyst acidity. Possibly, a Lewis acid character, which owes to the Ni active sites, can account for a rapid loss of activity. By contrast, depletion of activity for the WGS reaction was mainly ascribed to metal sintering, though a possible effect of acidity also from this point of view cannot be excluded. Indeed, OH groups on the support surface may be covered by coke, which leads to poorer water activation on the coked support. Lewis acidity of the support mainly contributed to coke accumulation, but without any evident depression of activity for GSR or the WGS reaction.

Experimental Section

Catalyst preparation

Support synthesis in liquid phase

TiO₂ was prepared by a conventional precipitation method. TiOSO₄⋅x H₂SO₄⋅x H₂O (20 g, Sigma Aldrich, synthesis grade purity) was dissolved in distilled water (300 mL) at RT, then NaOH (Carlo Erba, 9 M) was added dropwise until the system reached pH 5.5. The precipitate was aged at 60 °C for 20 h, then repeatedly washed with distilled water and finally dried overnight at 110 °C.

SBA-15 was synthesised as previously reported,25 in the presence of Pluronic 123 (P123, Aldrich) as structure directing agent. Silicon hydroxide was calcined at 800 °C for 6 h.

ZrO₂ was prepared by a conventional precipitation method26 at pH 10.

Addition of the active phase

The active phase was added to each support by incipient wetness impregnation with an aqueous solution of the metallic precursor (Ni(NO₃)₂⋅6 H₂O, Sigma Aldrich, ≥98.5 %), in the proper concentration in order to obtain the desired Ni loading (10 wt %). The catalyst was dried overnight at 110 °C and then calcined at 800 °C for 4 h.

Catalysts synthesis by flame pyrolysis

A second set of samples was prepared in nanopowder form by means of a flame pyrolysis apparatus,19–24, 27, 28 with the aim to impart high thermal stability and tune the metal dispersion. The Ti-based sample was synthesised by using a solution of titanium(IV)isopropylate (Aldrich, 97 %, 0.67 M) in xylene, in which the concentration is referred to TiO₂. The SiO₂-supported sample was prepared by diluting TEOS (Fluka, 99 %, 0.67 m) in xylene, in which the concentration is referred to SiO₂, whereas the sample supported on ZrO₂ was produced from a Zr-acetylacetonate (Aldrich, 98 %) solution. Ni was added to such mother solutions by dissolving Ni^II acetate (Aldrich, 98 %) in propionic acid (Aldrich, 97 %) to achieve a nominal 10 wt % metal loading with respect to the support oxide and a 1:1 volume/volume solution of the two solvents. The solutions were fed to the nozzle by using a glass syringe (50 mL) with a flow rate of 2.2 mL min⁻¹ and a 1.5 bar pressure-drop across the nozzle, co-fed with 5 L min⁻¹ of O₂. (1 bar=100 000 Pa)

Catalysts were named as NiAB, in which A=Ti, Si or Zr with reference to TiO₂, SiO₂ and ZrO₂ carriers, respectively and B indicates the liquid-phase synthesis of the support (L) or flame pyrolysis (F).

Characterisation

To evaluate the actual metal concentration in the catalysts, atomic absorption spectroscopy measurements were performed on a Perkin–Elmer AAnalysis instrument after dissolution of the sample.

XRD patterns were collected on a Bruker D8 Advance diffractometer equipped with a Si(Li) solid-state detector (SOL-X) and a sealed tube providing CuK_α radiation. Phase recognition was possible by comparison with literature data.29 Crystal size was calculated with the Scherrer equation.

SSA and pore size distributions were evaluated through N₂ adsorption-desorption isotherms at −196 °C (Micromeritics, ASAP 2000 Analyser). Surface areas were calculated on the basis of the BET equation, whereas the pores size distributions were determined by the Barrett–Joyner–Halenda method, which was applied to the N₂ desorption branch of the isotherm. Prior to analysis, the sample was dried overnight at 110 °C and then degassed at the same temperature for 2 h.

XPS analysis was performed for each fresh sample by means of a monochromatised Surface Science Instrument (SSI).

TPR measurements were performed by placing the catalyst in a quartz reactor and heating at a rate of 10 °C min⁻¹ from RT to 800 °C in a H₂/Ar mixed gas stream (5 vol %) flowing at 40 mL min⁻¹. TPO was performed by heating at a rate of 10 °C min⁻¹ from RT to 800 °C in an O₂/He gas stream (5 vol %) flowing at 40 mL min⁻¹. TPR–TPO–TPR cycles were performed on all the samples.

SEM images were obtained by using a Philips XL-30CP electron microscope and the surface and elemental composition of catalysts was determined by using energy dispersive X-ray measurements (EDX). The scanning electron microscope was equipped with an LaB₆ source and an EDAX/DX4 detector. The acceleration potential voltage was maintained between 15 keV and 20 keV and samples were metallised with gold. Spent catalysts were analysed by field emission gun scanning electron microscopy (FEGSEM) on an LEO 1525 microscope, after metallisation with Cr.

TEM images of fresh and spent samples were obtained with a Philips 208 transmission electron microscope. The samples were prepared by putting one drop of an ethanol dispersion of the catalysts on a copper grid pre-coated with a Formvar film and dried in air. The particle size was averaged out from at least five pictures that were made on different zones of the catalyst.

FTIR spectra were recorded under static conditions by a Nicolet Nexus Fourier transform instrument, by using conventional IR cells connected to a gas manipulation apparatus. Pressed disks of pure catalyst and support powders (≈20 mg) were thermally pre-treated in the IR cell by heating under vacuum at 500 °C. The samples were heated in pure H₂ at 500 °C after this pre-treatment to reduce them (600 Torr, 2 cycles, 30 min each); this was followed by an evacuation step at the same temperature. CO adsorption experiments were performed at liquid nitrogen temperature, and were subsequently degassed upon warming.

PN adsorption experiments were performed over the reduced samples at RT and degassing was performed subsequently at increasing temperatures.

TGA was performed on spent samples by means of a Perkin–Elmer TGA7 instrument by heating the sample (20–30 mg) in air.

Micro-Raman sampling was made by an OLYMPUS microscope (model BX40) connected to an ISA Jobin–Yvon model TRIAX320 single monochromator, with a resolution of 1 cm⁻¹. The source of excitation was a Melles Griot 25LHP925 He-Ne laser that was used in single line excitation mode at λ=632.8 nm. The power focused on the samples was always less than 2 mW. The scattered Raman photons were detected by a liquid-nitrogen cooled charge coupled device (CCD, Jobin Yvon mod. Spectrum One).

The steam reforming of glycerol

GSR experiments were performed by using a micropilot continuous plant.30 The catalysts were loaded on a fixed bed quartz reactor, operating at atmospheric pressure. Each sample (200 mg, 0.355–0.500 mm), diluted with the same amount of quartz of equal grain size, was reduced in situ at 700 °C under an H₂ flow, with a flow rate of 25 cm³ min⁻¹, for 1 h. The time factor during activity testing was 1.8 g _GLY/(g _CAT*h), obtained by feeding 0.060 mL min⁻¹ of a 10 wt % glycerol aqueous solution by means of a diaphragm metering pump (Stepdos, KNF). He was used as a sweeping gas, with a flow rate of 30 cm³ min⁻¹. A pressure controller was placed before releasing the inlet gas to prevent overpressure phenomena. Vaporisation of the solution took place at 250 °C before entering the catalytic bed in a unit filled with quartz beads. The outlet gas was sent through a vent line or to the gas analysis apparatus, whereas condensables were removed through a refrigerated coil.

The analysis of the out-flowing gas was performed by using a gas chromatograph (Agilent, mod. 6890N) equipped with two columns connected in series (MS and Poraplot U) with thermal conductivity and flame ionisation detectors (TCD-FID), properly calibrated for the quantification of CH₄, CO, CO₂, H₂, whereas the gas flow was continuously monitored by an on-line volumetric flowmeter (BIOS Defender 530 L).

Typically, the reaction temperature was 650 °C for approximately 20 h on-stream, and repeated steady state analyses of the effluent gas were collected. The performance of the catalysts is presented here in terms of H₂ yield, C conversion and CO, CH₄ and CO₂ selectivity in the gaseous phase (F_i), based specifically on the following reaction [Eq. (1)] and Equations (2)–(4):(3), (4)

• (1)

• (2)

• (3)

• (4)

in which i=detected gaseous products (CO₂, CO and CH₄).

Acknowledgements

The authors are grateful to Regione Lombardia and the Consortium for Material Science and Technology (INSTM) for financial support. A.G. and V.D.S. would like to thank the Italian Ministry of Education, University and Research (Project “ItalNanoNet” protocol no. RBPR05JH2P). The valuable help of the PhD student Cesare Biffi and of the MoS graduating student Mauro Castellana is gratefully acknowledged.