University of Birmingham Nickel–silica core@shell catalyst for methane reforming

Steam reforming of methane from biogas at a small scale could potentially provide a source of hydrogen for applications such as electricity generation via fuel cells. The efﬁ-ciency of the reforming process is dependent upon an effective catalyst and thus this work aimed to produce a highly active catalyst for methane reforming which is resistant to deactivation. A nickel e silica core@shell catalyst was synthesized by a deposition e precipitation method. The catalyst was characterized by XRD, SEM, TPR and infrared spectroscopy techniques. TEM analyses of sections of the catalyst embedded in resin conﬁrmed that the catalyst had a core@shell structure. Both forms of nickel phyllosilicate 1:1 and 2:1 were identiﬁed in the catalyst structure. The performance of the catalyst in methane steam reforming was investigated. The catalyst showed relatively high methane conversion 85% at 750 (cid:1) C. an


Introduction
Hydrogen could potentially become the major energy carrier in the future, since it is not polluting and is a very energy efficient fuel. There has been a huge amount of research in last few years to develop hydrogen generation systems [1]. In rural areas, hydrogen can be produced by reforming of gas from biomass gasification [2,3] or by reforming of gas from anaerobic digestion [4,5]. Biohydrogen can be produced directly from biomass using a dark fermentation process [6,7]. Thus the motivation for this work was to investigate reforming processes with regard to their possibility of producing hydrogen from gasification and anaerobic digestion in rural areas.
The majority of hydrogen today (98%) is generated from fossil materials, mostly via steam reforming of natural gas [6]. Methane steam reforming (1) is a highly endothermic process, which requires an efficient heat supply: The most common type of commercial steam reformers are fixed bed reactors in which a catalyst is usually packed in a set of tubes placed into a furnace. The temperature in the reactors is around 675e1000 C at a pressure of around 30 bar [8,9].
The methane steam reforming reaction (1) is associated with the wateregas shift reaction (2) which results in increased hydrogen production: Possible catalyst deactivation may occur if a lower than 1:1 stoichiometric ratio of water to methane is used during methane steam reforming with a Ni catalyst, since it promotes carbon deposition over the catalyst. Coke formation under such conditions is due to methane decomposition (3) and the Boudouard reaction (CO disproportionation) (4) which are favourable when there is not enough water for the stoichiometry of methane steam reforming [10,11]: 2CO4C þ CO 2 DH 0 298 À 173 kJ$mol À1 : On the other hand, the high molar ratio water/methane can also lead to deactivation of the catalyst and has high energy consumption [11]. Carbon can be also formed during reduction of CO to carbon (5): In addition to steam reforming, hydrogen can be produced from methane by dry reforming, catalytic partial oxidation, bireforming, autothermal reforming and tri-reforming of methane. Typical catalysts for steam reforming are bi-functional with active metal particles and support material with high surface area. Due to their high activity, long stability and low cost, nickel based catalysts are normally used for methane reforming.
Silica is attractive as a catalyst support since it has strong structural robustness, is stable even at elevated temperature and is chemically inert. It presents the advantage of mechanical strength and usually has a strong interaction with metal. Loaiza-Gil et al. [12] used Ni/SiO 2 catalyst synthesized by the ammonia method to obtain w60% conversion of methane (at 700 C). Similarly methane conversion of w50% was obtained by Guo et al. [13] with catalysts prepared by impregnation and by plasma decomposition method (CH 4 :H 2 O 1:1, 700 C). Matsumura and Nakamori [14] tested a methane steam reforming catalyst prepared by impregnation 20 wt%Ni/SiO 2 and obtained considerably lower methane conversion of only w20% (at 500 C), with their catalyst deactivating completely after 4 h. Zhang et al. [15] obtained w2% conversion of methane for a catalyst prepared by impregnation during direct internal reforming of methane in fuel cell with Ni/SiO 2 catalyst and up to 11% conversion after covering the same catalyst by a zeolite protective shell (650 C, Ni-12.5% wt). Using a Ni/silica catalyst prepared by impregnation for methane reforming with CO 2 can result in rapid deactivation [16]. However, catalyst prepared by co-assembly method can result in up to 90% methane conversion (Ni 6% wt, 800 C) [17].
Silica nanostructures can be prepared using the Stö ber method [18e24], which results in particles which are spherical and have size distribution in the range 0.05e2 mm. Close control of the particle size is important for catalyst preparations and can be controlled by modification of reaction conditions, for example by changing compound concentrations [18,21]. Silica spheres can be used as a support for a catalyst with core@shell structure, which could provide a high metal dispersion for steam reforming of methane. Coreeshell structures also offer lower internal diffusion resistance, since metal nanoparticles formed the shell are not buried in the support framework, such as the silica core. Metal particles in a core@shell structure usually have strong chemical and physical interaction with the support matrix and thus are highly resistant to sintering and deposition. The core@shell structure of catalysts may allow reduction the amount of metal used, since the core does not contain metals, which could be significant for the economics of expensive noble metal catalyst industrially. Additionally, sometimes the metal nanolayer shows higher catalytic activity than pure metal [25].
Core@shell catalysts can be synthesized by a number of different techniques, which include; depositioneprecipitation method [18,20,26,27], ultrasonic treatment of the support and nickel salt solution [11,28], solegel/reduction method [29], reduction with a previously prepared core [30] or by sequential reduction process [25]. An alternative method of core@shell structure preparation was proposed by Libor and Zhang [31], who used the opposite zeta potential of silica particles and nickel nanocrystals to create a nickel shell over a silica core.
In the present study, the synthesis and characterisation of a nickel supported by silica core@shell structured catalyst are reported. The prepared catalyst has good catalytic activity for methane reforming. Silica spherical particles were prepared according to the Stö ber method. A nickel shell was synthesized using the depositioneprecipitation method. The catalysts were prepared via a literature method, and extensive characterisation is reported here in order to understand and explain their behaviour in the methane reforming reactions. The goal of this paper is to demonstrate the efficacy of nickel silica core@shell particles as a catalyst for methane reforming processes.

Catalyst preparation
The following method was used to prepare the catalyst with a core@shell structure, which briefly was made from spherical silica particles covered by a nickel shell.

Preparation of spherical silica particles
Silica particles were prepared according to the Stö ber process [18e24]. This physicochemical process includes the hydrolysis of alkyl silicates followed by polycondensation under alkaline condition in alcohol solution. Tetraethyl orthosilicate (TEOS) was chosen as a silica source. A mixture of ethanol with ammonium hydroxide was blended for 5 min followed by quick stepwise addition of TEOS. The mixture was stirred for 1 h in a sealed flask at room temperature. The starting reagents contained: 13.24 mol l À1 of ethanol, 1.08 mol l À1 of ammonia as a catalyst/to stabilise the suspension and 0.33 mol l À1 of TEOS. The precipitate was separated from suspension by centrifugation and washed with deionised water.

Covering SIO 2 particles by nickel
Silica particles were covered with nickel using the depositioneprecipitation method [18,20,26,27]. The advantage of this method is that resulting metal particles are smaller and more resistant to sintering than those produced by impregnation. This method also allows to deposit high nickel loading compared to cationic exchange method [32]. The depositioneprecipitation method is based on a slow and homogeneous basification of the solution containing nickel precursor and suspended support material. Introduction of hydroxyl ions was performed by slow decomposition of added urea at a temperature above 90 C. SiO 2 particles prepared in the first step were dispersed in water by sonication at a concentration i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 8 ( 2 0 1 3 ) 1 4 5 3 1 e1 4 5 4 1 of 15 g l À1 . Nickel(II) nitrate hexahydrate with concentration 5.2 mmol l À1 was added to the suspension, after which the mixture was stirred slowly and heated using a stirring hot plate. After reaching 90 C, 50 mmol l À1 of urea was quickly added stepwise to the mixture. Following the introduction of hydroxyl ions, the nickel(II) phase slowly precipitated over the support surface. Nickel precipitate interacted with the silica surface and slowly covered the silica particles and after aging for 3 h the core@shell particles were separated using centrifugation and washed with water. The single coated catalyst had nickel concentration w3.6 wt% (3%Ni@SiO 2 ). The depositioneprecipitation process was repeated three times to achieve the desired nickel concentration w11 wt% (11%Ni@SiO 2 ), with this triple coated catalyst being used for the remainder of the catalytic tests. After drying in air at 105 C overnight, the prepared catalyst was calcined at 700 C for 4 h in air with a moderate heating rate of 5 C min À1 to prevent nickel particles from sintering. Calcination also resulted in removal of residual organic compounds and nitrate ions remaining from the catalyst preparation procedure. After calcination and reduction, the nickel catalyst was considered as activated in readiness for the reforming process.

Characterisation
The BET surface area, average pore diameter and total pore volume were determined from N 2 adsorptionedesorption cycles performed on a Micromeritics ASAP2010. Samples were degassed at 120 C prior to the experiment. Temperature programmed reduction (TPR) experiments were carried out using a Micrometrics Autochem II Chemisorption Analyser. Samples of ground catalyst 0.1 g were preheated at 500 C for 1 h under an argon atmosphere and then cooled to room temperature to remove moisture and gases adsorbed from the air. Finally, the temperature was increased to 900 C, at a rate of 10 C min À1 in 10% H 2 /Ar. Hydrogen consumption was recorded using a thermal conductivity detector (TCD). The structure of the catalyst was analysed by X-ray powder diffraction (XRD) using a Bruker D8 Advance. Analyses were carried out at room temperature in the two-theta range from 20 to 80 . Scanning speed was 2.2 min À1 for all samples. The sample reduced at 650 C ( Fig. 6D) was additionally analysed using scanning speed 0.7 min À1 in two-theta range 35e55 min À1 . The chemical composition of the catalyst was investigated using X-ray fluorescence spectrometry (XRF) on a Bruker S8 Tiger. For analysis 0.5 g of each sample was ground, mixed with wax and pressed to form a pellet. The catalysts were analysed using Fourier transform infrared spectroscopy. Absorption spectra were collected in range of wavenumbers 400e4000 cm À1 using a Bruker Tensor 37. Samples of the catalyst were mixed with KBr in a proportion of 1:100.
The morphological properties of the tested catalyst and the support were measured using a Scanning Electron Microscope (SEM) Philips Xl-30 and a Jeol 7000F. For the Philips XI-30 instrument the accelerating voltage was 15e20 kV. Samples were deposited onto a copper tape and coated by gold to improve surface conductivity. For the Jeol 7000F instrument samples were coated by carbon and the accelerating voltage was 15 kV. The particle size distribution was measured from electron micrographs. The particle morphology was examined by transmission electron microscopy (TEM). Samples of the catalyst were suspended in ethanol by sonication. A carbon film on a golden micro-grid was used as a sample holder. Samples were examined by the JEOL 1200EX microscope operated at 80 kV. To expose the core@shell structure of prepared catalyst it was necessary to crush the deposited shells. Samples of the catalyst were dehydrated in ethanol and embedded in low viscosity acrylic resin (LR-White) overnight. The resin blocks were trimmed on a microtome and then examined with the JEOL TEM.
The temperature-dependent mass change profiles (TG) were analysed using a thermogravimetric analyser (Netzsch 209F1). The mass change of fresh catalyst during catalyst reduction (gas mixture 5%H 2 /N 2 at flow 50 ml min À1 ) was recorded over the temperature range 25e1000 C, with a heating rate of 5 C min À1 and sample mass of 10e20 mg. Samples of spent catalyst were examined by a temperature programmed oxidation (TPO) method, by heating samples of used catalyst (10e20 mg) in air at a flow rate of 50 ml min À1 from room temperature (w20 C) to 900 C at a rate of 10 C/ min. The recorded mass reduction in the temperature range 575e625 C was assumed to be due to carbon burning, since adsorbed gases would desorb at lower temperatures and the Ni@SiO 2 was assumed to be resistant to thermal decomposition over that temperature range.

Catalytic reaction
The catalytic activity of the prepared catalyst 11%Ni@SiO 2 was evaluated by testing for methane steam reforming in a continuous flow fixed bed reactor. The stainless steel reactor with inner diameter 5 mm operated at atmospheric pressure and was equipped with a gas flow control system and an online gas chromatograph (GC). For each test, 1.0 g of the catalyst (as a powder <250 mm) was loaded into the reactor supported by quartz wool. In order to achieve close to isothermal behaviour and to avoid a gas pressure drop and particles clogging, the catalyst was diluted with 2 g of molecular sieve 13Â (1.4e2 mm). The molecular sieve was selected as an inert particle, which does not catalyse the reaction. In order to check this assumption, a reaction was carried out with only the molecular sieve without catalyst in the reaction tube. The methane concentration changed by only w0.1%. The reactor with a catalyst bed length of 150 mm was installed into a tubular furnace equipped with a temperatureprogrammed system. A set of mass flow controllers (MFC) regulated flow of reactants to the reactor. Prior to the reaction, the catalyst was reduced "in situ" at 650 C using hydrogen flow 10 ml min À1 for 1 h and a heating rate of 5 C min À1 . After the reduction process, the reactor was purged by nitrogen (100 ml min À1 , 5 min), a gas mixture of CH 4 and He (carrier gas) was then switched into the reactor. Deionised water was fed to the reactor by a pump 0.02e0.12 ml min À1 five minutes after commencing the flow of methane, the delayed start being used since Song et al. [33] reported that introduction of H 2 O before methane can lead to catalyst deactivation. During the first few minutes of the reaction, methane was diluted by nitrogen present in the reactor. Steam was generated and mixed with methane and helium in the top part of the reactor (filled with the molecular sieve to aid dispersion of the flow) before reaching the catalyst bed. The steam reforming reaction was carried out at a constant temperature of 550 C, 650 C or 750 C. Gas flow rates during all experiments were constant: CH 4 25 ml min À1 , He 10 ml min À1 . The molar ratio of feed gases were respectively CH 4 :H 2 O:He 1:3:0.4; 1:1:0.4 or 1:6:0.4.
Gaseous products at the reactor outlet were analysed by on-line gas chromatography (Agilent Technologies 7890A) after cooling and condensing water by passing through an ice trap. The gas chromatograph was equipped with seven columns, five valves and three detectors: a flame ionization detector (FID) and two thermal conductivity detectors (TCD). The combination of three packed columns (2Â HayeSep Q and 1x Molsieve 5A) and two valves were used for separation of CH 4 , CO 2 and CO and a molecular sieve column was used for H 2 separation. The steam reforming reaction was carried out for 1 h before the first sample was taken, with products being subsequently analysed every 30 min.
The methane conversion and the H 2 yield were calculated according to the following equations: where CH 4in and H 2 O in are the reactants inlet molar flow rates (mol s À1 ) of CH 4 and H 2 O respectively, CH 4out , H 2out , CO 2out and CO out are the outlet molar flow rates (mol s À1 ) of produced species respectively CH 4 , H 2 , CO 2 and CO.

Results and discussion
3.1. Catalyst characterisations

X-ray florescence spectrometry
The chemical composition of catalyst used for methane steam reforming was determined by X-ray fluorescence spectrometry ( Table 1). The nickel loading measured by XRF for silica particles after only a single nickel coating was w3.6%, whilst the catalyst prepared by triple coating had w11 wt% of nickel loading and Ni/Si molar ratio around 0.5. Except nickel and silicon, no other elements were detected. Other elements constituting the catalyst such as oxygen from nickel or silica oxide, or oxygen with hydrogen from the hydroxide group cannot be detected by XRF.

Morphological properties
It was observed that the colour of the prepared catalyst changed during preparation. White silica particles became celadon coloured after covering by nickel. Following calcination the nickel covered catalyst surface changed colour to light grey and after the reduction process the catalyst became dark grey. The colour change is a good indicator of the catalyst condition, showing that under some conditions of methane reforming part of catalyst became oxidized, as indicated by a light grey colour. The fact that the colour change was due to nickel phase oxidation was confirmed by TPO analyses of the reduced catalyst. Silica particles generated in the Stö ber process were found to be spherical and monodispersed. The surface of the uncoated silica spheres ( Fig. 1A and B) was relatively smooth. The dimensions of the prepared silica catalysts ranged from 0.45 to 0.85 mm with average particle size around 0.7 mm. Some particles formed clusters made from 2 to 3 particles due to aggregation during the preparation process. Wang et al. [21] reported that at low (w20 C) reaction temperature coagulation of silica spheres is possible during the Stö ber process. They also reported that when ethanol was used as a solvent, coagulation of silica particles could occur. Both conditions were accomplished in the experiment reported here and therefore aggregation of some particles was expected. Singh et al. [24] obtained smaller particles of around 0.15 mm diameter using a similar method of preparation, although a lower ammonium concentration was used.
The particle morphology changed significantly after coating by nickel, as evident from Fig. 1C and D which present the morphology of the prepared catalyst after the reduction process. Since no uncoated silica surface was detected, cracks or breaks in nickel coating were not apparent on catalyst grain structure, even after calcination or reduction processes. This suggests a strong interaction between the silica core and nickel shell. The prepared nickel catalyst had ferromagnetic properties after reduction and therefore any uncoated silica particles would be easily separated from those covered by nickel by using a magnet. This method can also be used for separation of grains of oxidised catalyst. The average particle size increased after coating silica spheres with a nickel shell. The particle-size range was from 0.68 to 1.0 mm with average diameter around 0.9 mm. The number of clusters of 2e3 particles was higher than that for bare silica. This is due to particle agglomeration during coating or coalescence of coating layers during calcination which could be avoided by the use of stabilising surfactant. Fig. 2A and B presents TEM micrographs of the calcined catalyst 11%Ni@SiO 2 . TEM analyses confirmed results achieved by SEM, again showing that the particles had a spherical shape and were covered by a nickel shell made from needlelike particles. This explains the high surface area of prepared catalyst compared to bare silica particles ( Table 2). Bare silica particles, or particles with broken shell were not detected and the shell thickness was too large to detect the core@shell structure. The prepared shells appear to be homogenous because large metallic particles, normally characteristic for catalyst prepared by impregnation method, were  not present. It may be noted that particles were rather uniformly covered by a nickel shell.
To expose and image the core@shell structure of prepared catalyst it was necessary to crush the shells of the calcined 11%Ni@SiO 2 using the sectioning method described in Section 2.2. Cutting ultrathin sections exposed the silica cores as shown in the TEM micrographs ( Fig. 3A and B). Particles with cracked shells, shells partially removed to expose the silica core (white parts) and some bare silica particles (white) are visible in the images. Table 2 lists the textural information of the calcined and reduced catalyst and of pure silica spheres based on characterisation of N 2 sorption. The surface area of bare silica was around 3.0 m 2 g À1 , being slightly lower than the obtained silica with specific surface area of silica 5.5 m 2 g À1 obtained by Singh and co-workers [24] using a similar method. The surface area significantly increased to above 68 m 2 g À1 after covering silica spheres by a nickel shell due to the fact that the nickel layer had a porous structure. Increasing the amount of nickel on the silica particles from 3.6 to 11 wt% led to a 3-fold increase in catalyst surface area, and also increased the total pore volume and decreased the average pore diameter. The reduction process did not change the average pore diameter however the surface area and the total pore volume were slightly reduced. This is consistent with literature data [34] that the surface area of nickel based catalysts can be decreased after catalyst reduction. Pu et al. [18] reported that the BET surface area of a nickel shell prepared by depositioneprecipitation can be around 290 m 2 g À1 . Burattin et al. [32] observed that using non-porous silica as a support material can result in the   formation of a Ni(II) phase with smaller metal particles compared to those formed on porous silica. They concluded that this was a result of better interaction of support with Ni and formation of better crystallized, less reducible metal particles.

X-ray diffraction
XRD patterns of the catalyst sample 11%Ni@SiO 2 after calcination and after reduction at 650 C and 900 C are shown in Fig. 4. The absence of diffraction peaks for silica suggests that the silica was amorphous and absence of diffraction peaks related to nickel oxide suggests that there are no free NiO crystals. This suggests that nickel was in an amorphous phase or there was a high dispersion of small free crystals that were below the XRD detection limit. Heating did not result in significant changes in the XRD patterns. For catalyst reduced at 650 C (Fig. 4, curves B and D), at 2Q values between 44 and 47 a weak and broad peak were observed. For catalyst reduced at 900 C the broad peak slightly increased. The shoulder was in the range specific for Ni 0 and for nickel silicide Ni 3 Si, which can suggest the formation of a Ni 0 phase. The formation of nickel silicide during reduction of nickel(II) phase was reported in the literature [35]. The pattern for the sample of 11%Ni@SiO 2 catalyst after calcination did not show any diffraction peaks (Fig. 4, curve A). It is possible that after catalyst reduction a small amount of nickel in crystal form appeared on the catalyst surface.

Temperature programmed reduction
It is known that Ni/SiO 2 catalyst prepared by impregnation has a major hydrogen consumption peak at ca. 430 C and small hydrogen consumption peak around 500 C [12,14,36,37].
The first reduction peak (430 C) is usually attributed to the reduction of NiO, which has a weak interaction with the silica support and the second peak (500 C) is attributed to NiO that interacts strongly with the support. The same shape of TPR but slightly shifted to higher temperature was achieved by Pompeo et al. [38].
There are several methods for increasing the reduction temperature of a catalyst prepared by impregnation. Pan et al. [36] noticed that treatment of the catalyst after impregnation by plasma discharge can shift the catalyst reduction peaks to higher temperatures. After the plasma treatment, the second TPR reduction peak was more intense and wider whilst the first was smaller, with interaction between nickel particles and support matrix being enhanced. Higher catalyst calcination temperature also leads to the reduction of nickel species being shifted to higher temperature [39,40], indicating greater integration between NiO and the oxide support. For a catalyst prepared by co-precipitation the nickel reduction can be shifted to higher temperature by a reflux procedure [40]. Also preparation of catalyst by an ammonia method [12] favours the formation of nickel having a strong interaction with the support.
Reduction of the Ni@SiO 2 catalyst prepared by depositioneprecipitation took place at higher temperatures than the Ni/SiO 2 catalyst prepared by impregnation [12,14,36e38], showing stronger interaction between nickel and silica for the former catalyst sample. This interaction can enhance resistance for coke deposition and prevent sintering of nickel particles. According to the literature [12,18,27,41,42] during depositioneprecipitation of nickel with silica spheres, usually only nickel hydroxide and nickel phyllosilicate is formed, although formation of nickel(II) basic carbonate is also possible [20]. Decomposition of urea during nickel depositioneprecipitation results in basification of aqueous solution, whereupon nickel hydroxide formed from nickel salt reacts with silicic acid to form NieOeSi bonds [42]. Two forms of nickel silicates can be derived from nickel hydroxide  [27,41]. Burattin et al. [27] suggested that during depositioneprecipitation only 1:1 nickel phyllosilicate is formed and 2:1 phyllosilicate is developed during depolymerisation of 1:1 nickel phyllosilicate.
The tested catalyst was calcined in preparation at 700 C, a temperature at which a significant part of the nickel  hydroxide and nickel carbonate should decompose. As noted by Lehmann et al. [41] after catalyst calcination at temperatures above 500 C the presence of Ni(OH) 2 is rather unlikely. Burattin et al. [26,27,32,35] observed that 1:1 nickel phyllosilicate is reduced at temperatures in the range 450e650 C on Ni/SiO 2 catalysts. On the other hand, 2:1 nickel phyllosilicate requires higher temperatures in the range 690e760 C for complete reduction. They also noticed that crystallinity of bulk nickel phyllosilicate strongly depends on the molar ratio OH À /Ni 2þ , Ni/Si and on the depositioneprecipitation time [26]. Reduction of the tested catalyst 11%Ni@SiO 2 started beyond 350 C and the maximum hydrogen consumption was at 580 C. The secondary consumption peak was at 690 C, with also a minor reduction peak obtained at 410 C. The appearance of three separate peaks during catalyst reduction suggests the occurrence of different Ni phases on the catalyst surface. From the literature [27], nickel precipitate obtained in the absence of silica can be reduced at w320 C. Therefore, interaction of nickel with silica support shifts nickel reduction to a higher temperature.
The TPR profile of tested catalyst (11%Ni@SiO 2 catalyst, Fig. 5, curves B and C) shows two distinct peaks in the range typical for phyllosilicate and thus it can be concluded that both forms of nickel phyllosilicate are present. The hydrogen consumption above 450 C (Fig. 5, curve C) with the maximum reduction peak at 580 C can be attributed to the reduction of 1:1 nickel phyllosilicate. Part of the same hydrogen adsorption peak can be also attributed to unsupported 2:1 nickel phyllosilicate. The last peak at 690 C can be associated with reduction of 2:1 nickel phyllosilicate.
The small hydrogen consumption at 410 C observed by TPR (Fig. 5, curve C) can be attributed to unsupported nickel hydroxide [26] or possibly NiO crystallites without support interaction. XRD scans (Fig. 5) did not show any peaks corresponding to NiO, therefore a small amount of NiO below XRD detection limit cannot be ruled out. Also unsupported 1:1 nickel phyllosilicate could be reduced at that temperature. However, a definitive assignment of the form of obtained phyllosilicate cannot strictly be made based on only the reduction temperature, at it also depends upon the depositioneprecipitation time and on the silica structure [25]. In addition, the reducibility is also correlated with the catalyst crystallinity [27,41]. Therefore, the reduction temperature can change with different catalyst preparation procedures.
The temperature programmed reduction using TG was carried out to explain the reduction process of the Ni@SiO 2 catalyst. The thermogravimetric profile of the calcined 11% Ni@SiO 2 catalyst as a function of temperature is presented in Fig. 6. Two mass-loss steps were observed upon heating, although the total mass lost was only around 3.3%, suggesting that the catalyst is thermally stable. The initial change in the catalyst mass is attributed to the release of unbound water, with derivative thermogravimetry (DTG) peak at temperature 51 C. The release of physically adsorbed water on the catalyst    surface and hydroxyl anions reduced the mass of the catalyst by 0.5%. Since the catalyst was calcined at 700 C during preparation, most water had already been removed. Beyond 110 C a gradual 0.6% lineal mass reduction was observed. As there was no steep drop in the catalyst mass during that period the amount of nickel hydroxide or nickel carbonate in deposited layer was minor, since decomposition of these compounds would be expected to result in significant mass loss.
A more significant change in the catalyst mass due to the sample reduction occurred at temperatures in the range 650e700 C. The maximum hydrogen consumption (Fig. 5) in that temperature range was attributed to the 2:1 nickel phyllosilicate. According to the literature the nickel phyllosilicate is the most thermo stable form of the Ni(II) phase. As reported by Burattin et al. [35], the nickel phyllosilicate decomposes to Ni 0 metal particles via NiO as an intermediate product during catalyst reduction. After heating to 1000 C the residual mass was w96.7%. At that temperature the nickel phyllosilicate should totally decompose to NiO and SiO 2 and eventual nickel hydroxide to NiO [32]. Pompeo et al. [38] suggested that there is also possible formation of Ni 2 SiO 4 metallic species on silica surface after reduction of Ni/SiO 2 catalyst.
The possibility that part of the catalyst mass reduction at the temperature above 700 C was due to the removal of impurities cannot be ruled out, since the catalyst was calcined at 700 C and impurities stable to higher temperatures may not have been destroyed during calcination. It must be also noted that the presence of support sometimes modifies the decomposition temperature.

Infrared spectroscopy
The FT-IR spectra for bare silica and Ni@SiO 2 catalyst were similar (Fig. 7), suggesting that no significant changes in the silica structure occurred after coating silica particles with nickel shells. Spectra of all samples presented characteristic absorbance bands for silica without significant change in their position. Bands attributed to physically adsorbed water occurred at 3100-3670 cm À1 from the stretching of OeH bonds and 1628 cm À1 from the bending of HeOeH bond. Silica bands occurred at 813 cm À1 due to the symmetric vibration of silicon atoms (SieOeSi) and 473 cm À1 representing the bending mode of SieOeSi bond. The asymmetric vibration of SieOeSi bond represents an intensive group of bands at 1030-1220 cm À1 with a peak at 1108 cm À1 [21,23,24,41]. A shoulder at the left side of that band (around 1180 cm À1 ) was due to the external linkage of SiO 4 . A shoulder at the right side (at about 950 cm À1 ) can be assigned to the stretching vibrations of SieO À from surface silanol groups [41]. The latter was slightly reduced when nickel nanoparticles were loaded upon silica spheres. Similar decrease of the surface silanol vibration after phyllosilicate formation was observed by Lehmann et al. [41]. They explained the decrease in intensity of the silanol vibration as due to the formation of nickel phyllosilicate initiated by the hydrolytic adsorption of uncharged nickel hydroxoaqua complexes at silanol surface sites. Some authors [21] suggest that the very intensive band at wave number 1030e1220 cm À1 can confirm the formation of high density silica network.
An absorption band at 3746 cm À1 corresponds to the isolated silanol groups SieOH [41,43]. That band became slightly more intense after coating silica with a nickel shell (Fig. 7). The NieO stretching vibration (w480 cm À1 ) was impossible to detect because of the intensive SieOeSi band at 473 cm À1 .
After coating silica particles by nickel absorption bands at 670 cm À1 and 711 cm À1 appeared. According to the literature [26,27,41] both bands can be assigned to the formation of nickel phyllosilicate. The band at 670 cm À1 represents the d OH vibration and can be assigned to the formation of 1:1 nickel phyllosilicate or 2:1 nickel phyllosilicate [27,41]. The band at 711 cm À1 can be related only to 2:1 nickel phyllosilicate and was due to the tetrahedral SieO mode [27,41]. These two weak peaks confirm the XRD results, showing low crystallinity of formed nickel shell. Two forms of phyllosilicate were indicated from the weak absorption peaks attributed to the stretching of Y OH band, namely 2:1 nickel phyllosilicate structure at 3629 cm À1 and 1:1 nickel phyllosilicate at 3649 cm À1 [27,41]. After calcination (Fig. 7, curve B) the vibration bands specific for nickel phyllosilicate (3649, 3629, 711 and 670 cm À1 ) became slightly more intense than those of the reduced catalyst (Fig. 7, curve C). That may suggest that the reduction process slightly increased the crystallinity, in agreement with the XRD results presented in Fig. 4. However, because the IR absorbance obtained for nickel phyllosilicate had low intensity, it is difficult to estimate only from the IR results whether the tested catalyst contained a mixture of 1:1 and 2:1 phyllosilicate or only one of those compounds. In spite of this, the IR results support the aforementioned observation of the catalyst temperature programmed reduction profile and it can be concluded that both forms of nickel phyllosilicate were present in tested 11%Ni@SiO 2 catalyst.

Catalytic performance
The Ni@SiO2 catalyst prepared by depositioneprecipitation gave a respectable methane conversion compared to values of silica-supported catalysts reported in the literature. Figs. 8  and 9 show results from the catalytic tests of the nickel silica core@shell catalyst synthesized by the depositioneprecipitation method (11%Ni@SiO 2 ) for methane steam reforming. All the reactions were carried out in the fixed bed reactor at atmospheric pressure with constant methane partial pressure and ratio of methane feed flow rate to catalyst mass. Fig. 8 shows the methane conversion as a function of the reaction time under five different reaction conditions. The reactions were carried out isothermally at three different temperatures: 550 (Fig. 8, curve A), 650 (B) and 750 C (C) at CH 4 :H 2 O molar ratio 1:3. To analyse the effect of water concentration, the methane steam reforming reaction was carried out at different CH 4 :H 2 O feed molar ratio: 1:1 (D), 1:3 (B) and 1:6 (E) at the reaction temperature 650 C. No other compounds were detected in the produced gas except the main reactants and products: CO 2 , CO, H 2 and CH 4 . Water vapour in the exhaust stream was condensed before analysis, so that only the gases were detected. The catalyst was found to be stable during a test of 4 h, as the results indicate near constant characteristics of a methane conversion, hydrogen yield and CO selectivity in the tested range. The catalyst showed relatively high methane conversion, which rose with increasing reaction temperature, from around 35% at 550 C to w80% at 650 C. Further temperature increase to 750 C changed the methane conversion only slightly to 85%. Also with the temperature increase, the hydrogen yield increased from 20% at 550 C to 45% at 650 and 750 C. In line with this, the CO yield increased from 3% at 550 C to 45% at 650 C. At 750 C selectivity shifted towards CO 2 and the CO concentration in the produced gas slightly decreased. This suggests that at higher temperature the water gaseshift reaction (2) became more effective. The molar ratio of H 2 /CO (Fig. 9) changed from almost 60 at 550 C to 5e6 at 650e750 C. After the reaction the spent catalyst was inspected. It was found that following the low reaction temperature of 550 C, the part of catalyst at the bottom (inlet zone) of the reactor changed colour to light grey indicating catalyst deactivation. This effect was most likely due to oxidation of nickel particles on the silica surface by steam. In reactions of longer duration methane conversion could be decreased by this effect.
At the reaction temperature 550 C an interesting observation was that the molar ratio of H 2 /CO x was around 3.9, suggesting that almost all produced CO was converted via the water gaseshift reaction (equation (2)) into CO 2 , the major produced carbon species. During reactions at 650 and 750 C the H 2 /CO x ratio was around 3, independently of water addition (in the tested range). The fact that the molar ratio of H 2 / CO x was below 4, suggests that the amount of hydrogen produced by methane decomposition (3) was insignificant. Those results were confirmed by the absence of coke on the catalyst surface after the 4 h reaction (Table 3). When only the methane steam reforming reaction takes place by equations (1) and (2), a ratio of H 2 /CO x in the range 3e4 is expected. Production of gases with the ratio of H 2 /CO x higher than stoichiometric from the steam reforming reaction would suggest that some hydrogen was produced by methane decomposition as given in equation (3). Modern plants for commercial methane steam reforming use the ratio of water/ methane around 2.5. Higher ratios of 4e5 can result in a better hydrocarbon conversion but a bigger mass flow requires larger equipment that increases the investment and operating cost [44].
To test the effect of water concentration molar ratios CH 4 :H 2 O in the range 1:1 to 1:6 were investigated. At the stoichiometric water to methane ratio of 1:1 the methane conversion was below 50%. Increasing the partial steam pressure to give CH 4 :H 2 O molar ratio 1:3 significantly improved the methane conversion to above 80%. However, the hydrogen yield increased only slightly from 45 to 50%. The higher amount of water (1:6 CH 4 :H 2 O molar ratio) slightly reduced the methane conversion but significantly reduced the hydrogen yield to below 20%. At higher CH 4 :H 2 O molar ratio, water competed with methane for active sites upon the catalyst resulting in reduction of methane conversion. At molar ratio CH 4 :H 2 O 1:3 and 1:6 it was concluded that the   excess of water beyond the stoichiometric value of 1:1 did not take part in the reaction, since the excess amount of water was condensed from produced gas by the ice trap. This explains the hydrogen yield reduction during the reaction at 1:6 CH 4 :H 2 O molar ratio. Interestingly, with increasing the water partial pressure, the CO selectivity decreased and the CO 2 formation increased. The water gas shift reaction (2) became more intensive. This suggests that CO rather than CO 2 is produced by a mechanism competitive with adsorption and dissociation of H 2 O.
The nickel supported by silica core@shell catalyst was found to have high coke resistance. TPO characterisation of spent catalyst showed that carbon deposited on the catalyst surface only during the reaction at low temperature (550 C) and the reaction at low water ratio CH 4 :H 2 O 1:1 ( Table 3). The insignificant reduction in methane conversion during the reaction at 550 C was probably due to carbon deposition. TG analyses of used catalysts showed that the combustion of deposited coke occurred at around 600 C. This suggests that the deposited carbon was mostly in filamentous and graphitic form.

Conclusions
Silica particles prepared according to the Stö ber process were covered with nickel using the depositioneprecipitation method. TEM analyses of particles previously embedded in LR White resin and sectioned were successfully applied to confirm that the prepared catalyst had a core@shell structure. The shell that covered silica spheres was mostly comprised of 1:1 and 2:1 nickel phyllosilicate. The catalyst exhibited good activity for methane steam reforming. Methane conversion increased with increasing reaction temperature. At 750 C methane conversion was around 85%. The tested catalyst was effective even at stoichiometric water/methane ratio, although the optimum CH 4 :H 2 O molar ratio for methane steam reforming was 1:3. The strong interaction of metal with support resulted in high coke resistance and catalyst showed good stability for the first 4 h. The nickel@silica catalyst was demonstrated to have good potential for methane reforming. i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 8 ( 2 0 1 3 ) 1 4 5 3 1 e1 4 5 4 1