Effect of cobalt supported on meso–macro porous hydrotalcite in Fischer–Tropsch synthesis

Abstract

Hydrotalcite based cobalt catalysts were prepared by a slurry precipitation method, followed by a slurry impregnation method. The prepared supports and catalysts were characterized by N₂ physisorption, mercury intrusion, chemisorption, TPR, TPD, SEM, TEM, TGA, DTA, and XRD techniques. Their catalytic performance for FTS was evaluated in a fixed-bed reactor with a H₂/CO molar ratio of 2, reaction temperature of 240 °C, and reaction pressure of 25 bar. The incorporation of alumina and kaolin enlarged the inter void between hydrotalcite clusters, which resulted in macro porosity. The cobalt catalyst supported on a bimodal pore structure induced by kaolin showed a more stabilized catalytic activity and better heavy hydrocarbon selectivity in the FTS reaction when compared to other catalysts. The catalytic performance of the prepared catalyst depended on the cobalt reducibility and diffusion efficiency, which were determined by the cobalt particle size and porosity.

---

1. Introduction

Today's high crude oil prices and the necessity for obtaining clean alternative fuels have increased the attractiveness of the Fischer–Tropsch Synthesis (FTS) as a process for producing liquid fuels. The FTS reaction has also been regarded as a key technology for the Gas-To-Liquids Floating Production Storage Offloading (GTL-FPSO) process that allows the use of stranded gas fields as attractive resources.^1,2 The essential target of FTS is to produce paraffin and olefins of different molecular weights while limiting the formation of methane and CO₂. Cobalt is favored for the synthesis of long chain hydrocarbons from natural gas-based synthesis gas because of its high activity, high selectivity for linear paraffin, high resistance towards deactivation, and low water-gas shift activity. Maximal exposure of cobalt to the gaseous reactants is usually achieved by dispersing the metal on a high-surface-area support.^3–6 Some examples include highly dispersed cobalt particles on porous supports, such as conventional ceramics (i.e., SiO₂,^7–10 TiO₂,¹¹ ZrO₂ (ref. 12) and Al₂O₃ (ref. 13–15)).

The physicochemical properties of the supports, including the surface area, pore diameter, pore volume, and metal-support interaction, have a pronounced effect on the cobalt particle size, dispersion, and reducibility, thereby affecting the catalyst performance for CO hydrogenation.^16,17 For instance, the catalytic activity is proportional to the number of active sites on the catalyst surface. The large surface area of the supports can provide a highly active surface; however, a support with a large surface area usually contains a great number of small pores, which results in poor diffusion of reactants, intermediates, and products. Under a diffusion-limited condition, the concentration of CO in the pores of the catalyst will therefore be decreased. As a result, the actual ratio of CO to H₂ will be decreased and the selectivity for lighter hydrocarbons will subsequently be enhanced. By contrast, the aggregation of cobalt particles occurs on supports with larger pore diameters, which leads to low cobalt dispersion and therefore low catalytic activity.¹⁸

Proposals to overcome these limitations have included the design of supports with two independently adjustable pore diameters i.e., bimodal porous materials. Many studies have described attempts to make supports with bimodal structures and then apply the FTS reaction, based on the advantages provided by the physical properties of the supports. For example, Tsubaki et al.¹⁹ prepared bimodal porous silica by directly introducing small particles of silica sol into the large pores of silica gel and then applying FTS using a slurry phase reactor. The resulting bimodal porous support with a mesopore range of 6 nm and 45 nm showed a higher CO conversion than a unimodal support with 3 nm or 50 nm pores. Park et al.²⁰ prepared a bimodal porous mixed oxide of ZrO₂–Al₂O₃ by a co-precipitation method and showed that this bimodal support with meso- and macro-pore ranges of 5 nm and 64 nm had higher activity.

The general form for compounds similar to hydrotalcite (HTlc) is [M(II)_1−xM(III)_x(OH)₂][Aⁿ⁻]_x/n·mH₂O, where the divalent ion can be Mg²⁺ or Zn²⁺ and the trivalent ion can be Al³⁺, Fe³⁺, or Cr³⁺. The compensatory anions may be NO³⁻, CO₃²⁻, SO₄²⁻, or Cl⁻. Generally, x can have values ranging from approximately 0.1–0.5.²¹ The use of HTlc-based catalysts has been recently reported for several processes in the energy field, such as hydrogen production from the reforming process.^22,23

Our previous work^24–26 suggested new methods for preparing spherical structured catalysts based on hydrotalcite-type materials for applications such as hydrogen production or synthesis of glycerol carbonate from biomass. Studies reporting the use of HTlc as a support or as a catalyst that includes the active metal as part of the structural core of the HTlc for FTS.^27,28

The objective of this study was to design a new HTlc-based support by slurry precipitation (SLP) and to incorporate the concept of bimodal structure with different two pore ranges, with the pore in the meso-pore range considered for cobalt dispersion and the pore in the macro-pore range considered for enhancing mass transfer. The outcomes of these investigations were confirmed through N₂ physisorption, mercury porosimetry, FT-IR, XRD, XRF, SEM, TEM, TGA, chemisorption, and temperature programmed techniques (TPR and NH₃-TPD).

2. Experimental

2.1. Catalyst preparation

2.1.1. Preparation of the hydrotalcite-based support. A commercial hydrotalcite catalyst, (Mg₂Al₂(OH)_4x+4CO₃·nH₂O) (MG70, procured from Sasol) was calcined at 500 °C for 5 h and denoted as s-(A). And the preparation ratios for other supports are listed in Table 1. The other supports were prepared by slurry precipitation method as the following below; calculated amount of MG70, kaolin [Al₂Si₂O₅(OH)₄, Sigma Aldrich] and aluminum nitrate [Al(NO₃)₃·9H₂O, Samchun] were dissolved in deionized water respectively. And precipitating agent, ammonium hydroxide solution [28.0% NH₃ basis, Sigma Aldrich] was separately prepared.

Table 1 Physical properties of hydrotalcite based supports by different preparation method

Catalyst	Preparation ratio (wt%)			N2 physisorption^a			Mercury intrusion^b
	Mg70	Aluminum nitrate	Kaolin	BET surface area (m^2 g^−1)	Pore size (nm)	Pore volume (cm^3 g^−1)	Mean pore size (nm)	Porosity (%)
a Measured by N2 physisorption [Moonsorp-II, KIST].b Measured by mercury intrusion [Carlo Erba Porosimeter 2000].
s-(A)	100	—	—	119	8.6	0.26	—	57.9
s-(B)	83.3	8.3	8.3	58	26.7	0.16	—	44.1
s-(C)	75.1	8.3	16.6	35	12.3	0.14	69.5	63.2
s-(D)	75.1	16.6	8.3	24	10.5	0.23	200	51.4

---

Both the solutions were simultaneously added to a 1 L round flask at rate of 10 ml min⁻¹ with constant stirring and maintaining the temperature during precipitation at 80 °C. And the precipitate was adjusted by ammonium hydroxide solution to a range pH of 9–10.

And then precipitated slurry was aged for 12 h, filtered, and then repeatedly washed with hot water. The precipitate thus obtained was dried at 60 °C for 12 h and subsequently calcined at 500 °C for 5 h. The prepared supports were designated as s-(B), s-(C), and s-(D).

2.1.2. Preparation of the supported cobalt catalysts. The cobalt supported catalysts were prepared by slurry impregnated methods with multiple steps.^29,30 All the catalysts containing 10 wt% Co on the prepared support were successively impregnated by using cobalt nitrate (Co(NO₃)₂·6H₂O, 98.0%; Junsei Chemicals, Japan) precursor solution.

They were subjected to further drying and calcined at 400 °C for 5 hours and denoted as c-(A), c-(B), c-(C), and c-(D). For comparison study, the commercial MG70 supported cobalt was regarded as reference.

2.2. Catalyst characterization

2.2.1. Nitrogen physisorption. A total of four supports and catalysts were analyzed for their capacity for nitrogen physisorption. The samples were placed in a quartz cell and combusted at 200 °C for 3 h to remove physisorbed materials. The physical properties (surface area, pore volume, and average pore diameter) were obtained using a Moonsorp II system (KIST, Korea) at −196 °C.

2.2.2. Mercury intrusion. Pore size measurements were performed by mercury intrusion using a Carlo Erba Porosimeter 2000. Prior to the analysis, the supports were evacuated and dried at 150 °C to remove physisorbed materials. A cylindrical pore model was assumed.

2.2.3. X-ray diffraction (XRD). The powder X-ray diffraction patterns were recorded for all the supports and catalysts at ambient temperature by wide angle X-ray spectroscopy on a D/MAX-2500 instrument (Rigaku Company) using CuKα radiation (λ = 1.54 × 10⁻¹⁰ m). The samples were crushed before measurement. The scans were recorded in the 2θ range between 10° and 90° using a step size of 0.04° and a step time of 15 s. Peaks were identified by comparison with standards in a database. The average cobalt oxide crystallite thickness was calculated from Scherrer's equation, using the (311) Co₃O₄ peak located at 2θ = 36.9°. A K factor of 0.89 was used in Scherrer's formula.^31,32 A high signal-to-noise ratio and, consequently, a minimal experimental error, were obtained by recording new scans only in the 2θ range between 32° and 43°. The average spherical Co₃O₄ particle size was calculated by multiplying the crystallite thickness by a factor of 4/3.³³ The Co₃O₄ particle size was converted to the corresponding cobalt metal particle size according to the relative molar volumes of metallic cobalt and Co₃O₄. The resulting conversion factor yielded the diameter d(Co₃O₄) of a given Co₃O₄ particle being reduced to metallic cobalt.

2.2.4. Hydrogen/carbon monoxide chemisorption. Hydrogen and carbon monoxide pulse chemisorptions were performed on an Auto Chem II 2920 (Micromeritics Company) 2010 instrument. The calcined catalyst was placed in a quartz reactor. Prior to pulse chemisorption, the sample was reduced in 10% H₂/Ar (or 10% Co/Ar) at 450 °C for 16 h.

The sample was then purged in helium at 400 °C for 1 h and cooled to 50 °C in flowing helium. Hydrogen pulse chemisorption (or carbon monoxide pulse chemisorption) was performed on the catalysts at 50 °C using 10% H₂/Ar (or 10% Co/Ar) and the gas dose was repeated at 5 min intervals until the hydrogen (or carbon monoxide) peaks on the chromatograph became identical. Furthermore, to calculate the dispersion, one cobalt surface atom is assumed to be covered by two hydrogen molecules (H/Cos = 2) and one cobalt surface atom is covered by one carbon monoxide molecule (H/Cos = 1).

The cobalt metal particle size was calculated from the cobalt metal dispersion by assuming spherical, uniform cobalt metal particles with a site density of 14.6 at per nm.^34,35 These assumptions give the following formula:

2.2.5. Temperature programmed reduction (TPR). Temperature-programmed reduction (TPR) experiments were performed on an Auto Chem II 2920 (Micromeritics Company) 2010 instrument. The calcined samples were loaded into U-shaped tubular quartz reactors heated by an electric furnace and then exposed to a reducing gas mixture consisting of 15% H₂ in argon while the temperature was increased from 100–900 °C. A cold trap containing a mixture of 2-propanol and dry ice was used to eliminate water and other condensable products from the product gas mixtures. The consumption of hydrogen was measured by comparing the thermal conductivity of the reference and product gas. Calibration was done by reduction of Ag₂O powder. The TPR was also used to estimate the degree of reduction. For the calcined catalysts, the hydrogen consumption of the reduced samples was also considered during calculation of the degree of reduction.

2.2.6. Scanning electron microscopy (SEM). SEM studies were performed using a Leica Stereoscan model 440 instrument manufactured by Scanning Electron Microscope, Inspect F (FEI Company). The powder samples were mounted on standard specimen stubs with adhesive tape or silver paste. The samples were coated with a thin layer of gold in a Polaran E-5000 coating unit to prevent charging of the sample. The electron beam parameters were kept constant during the analysis of the entire sample. The elemental mapping of metals was obtained by energy dispersive X-ray spectroscopy (EDS).

2.2.7. Transmission electron microscope (TEM). The TEM studies were carried out using a Super EDX-embedded TEM (FEI Company) equipped with a high brightness Schottky FEG and the SuperX EDS system, which includes four SDD EDS detectors symmetrically placed around the sample and a 16MP CMOS camera, the FEI Ceta 16M. The mapping experiments with energy dispersive X-ray spectroscopy (EDS) for each support metal were performed with the beam currents optimized to minimize beam damage and with a dwell time of 10 ms to further protect the specimen.

2.2.8. Thermal gravimetric analysis (TGA) and differential thermal analysis (DTA). The weight changes and the decomposition temperatures of the supports were investigated using a SDT Q800 (TA Company) thermal analyzer in a flow of air at the heating rate of 10 °C min⁻¹. The sample loading was typically 20–25 mg.

2.2.9. Temperature programmed desorption (TPD). The CO₂ temperature-programmed desorption (CO₂-TPD) analysis was performed in the same system used for TPR analysis. CO₂-TPD was used to obtain information about the basic sites of the catalytic samples. The temperature of desorption and the maximum CO₂ desorbed indicated the strength and number of basic sites. The sample was first reduced by heating for 3 h at 5 °C min⁻¹, from room temperature to 400 °C, under a flow of helium. After cooling to 100 °C under a flow of helium, CO₂ was fed at a rate of 50 ml min⁻¹ for 30 min. The temperature was ramped at 10 °C min⁻¹ from 50 to 900 °C and data were acquired. The basic site density was obtained by integration of the area under the curve. The average relative error in the basicity determination was lower than 3%.³⁶

2.3. Activity and selectivity measurements

Fischer–Tropsch synthesis was performed in a fixed-bed reactor (stainless steel tube, O.D = 12.7 mm and I.D = 10.2 mm). The apparatus has been described in detail elsewhere.^29,30 The reactor was placed in an electronic furnace to ensure a uniform temperature along the catalyst beds. The temperature of the electronic furnace was controlled by a proportional–integral–differential (PID) temperature controller. Separate mass flow controllers were used to add hydrogen, carbon monoxide, and nitrogen at the desired rates to the mixing vessel. The mole ratio of reactants and gas hourly space velocity (GHSV) was fixed at H₂/CO = 2 and 2000 h⁻¹, respectively, for the activity test. The samples were sieved (0.5 g, 300–400 μm) and diluted with 0.5 g α-Al₂O₃ (75–150 μm) as a diluting material. Since alumina has poor thermal conductivity, dilution of the catalyst allows good control of the temperature inside the catalytic bed. The maximum increase in temperature was experimentally verified with an axial thermocouple as 5 °C. The calcined catalysts were initially reduced in situ in hydrogen at 1 bar while the temperature was increased by 1 °C min⁻¹ to 450 °C. After 16 h of reduction, the catalysts were cooled to 180 °C and flushed with nitrogen gas for 3 h. The system was then pressurized to 25 bar, and a synthesis gas of molar ratio H₂/CO/N₂ = 2 : 1 : 0.5 (N₂ as the internal standard) was introduced into the reactor. Runaway and catalyst deactivation at startup were avoided by slowly increasing the temperature to the reaction temperature of 240 °C. The heavy hydrocarbons were collected in a heated trap (120 °C), while liquid products were collected in a cold trap (25 °C). The effluent gaseous product was analyzed on-line with a Agilent 7890A gas chromatograph (Agilent Technologies Company) equipped with a flame ionization detector (FID) and a GS Gaspro column (0.320 mm and 80 m) for C₁–C₉ hydrocarbons and with a thermal conductivity detector (TCD) and a Carbonsphere column (1.83 m × 1/8′′ × 2.0 mm) for hydrogen, nitrogen, carbon monoxide, carbon dioxide, water, and methane. The collected liquid/waxes products were analyzed off-line with an Agilent 7890 gas chromatograph equipped with a flame ionization detector (FID) and a HP-PONA column (50 m × 0.2 mm × 0.5 μm) for compounds over C₉.

2.4. Calculation method

The conversion of CO was calculated based on on-line GC analysis and defined as

The different product selectivity (i.e., the percentage of the converted CO that appeared as a given product) was calculated by on-line/off-line GC analysis and after subtraction of the amount of CO₂ in the product gas.

If the hydrocarbon is formed step-wise by insertion or addition of C₁ intermediates with constant growth probability (α), then the chain length distribution is given by the Anderson–Schulz–Flory (ASF) distribution. Assuming that α is independent of the hydrocarbon chain length, the following equation may be derived:

where W_n is the mass fraction of the species with carbon number n. The slope of the plot of log(W_n/n) against n gives the value of α. In this study, the product distribution of FTS was expressed by two independent ASF distributions with different chain growth probabilities (α₁, α₂) and the point of intersection of two ASF distributions.^37,38

The activity is expressed as Cobalt Time Yield (CTY, 10⁻⁴ mol_CO g_Co⁻¹ s⁻¹). The CTY is the number of moles of converted carbon monoxide per second divided by total amount of cobalt grams into the reactor, which was considered based on gas hourly space (GHSV) and cobalt dispersion determined by chemisorption.^5,6

The errors associated with GC analysis is 1%.

3. Results

3.1. The characterization of the hydrotalcite-based support

The isotherm of the supports and pore size distribution, as determined by the BJH model, are depicted in Fig. 1(a) and (b), respectively. As shown in Fig. 1(a), the isotherm of s-(A) was type IV with an H₂ hysteresis loop, indicating that the supports are composed of meso-pores with an ink-bottle structure. By contrast, the isotherms of the other supports [i.e., s-(B), s-(C) and s-(D)] exhibited composite isotherms consisting of types IV and II, indicating the presence of integrated meso-pore and macro-pore structures, respectively. The shape of the hysteresis loops of other supports fell within the H3 category,³⁹ which was attributed to slit-type pores associated with the inter-particle porosity generated in solids with plate-like morphology. The s-(A) showed a narrow Gaussian-like unimodal pore size distribution.

Fig. 1 Isotherm (a) and pore size distribution (b) of support obtained from nitrogen physisorption.

And its pore size distribution was matched with results of literature. The physical properties of hydrotalcite depends on the calcination temperature, which affected by the degree of losing hydroxyl groups (OH–) compensated by brucite plate and the existence of spinel compound. In this study, the calcination temperature of 500 °C gave their hexagonal shaped platy clusters with MgO crystal structure, which resulted in 8.6 nm of average pore size with narrow distribution as shown in Fig. 1 and 4.^24,44

However, other supports [i.e., s-(B), s-(C), and s-(D)] showed different average pore size with relatively broad distribution, depicted in Fig. 1(b). And it seemed that the incorporation of kaolin and alumina sources effected on the pore formation of support.

The macro-pores (above 50 nm), which cannot be investigated by nitrogen physisorption, were characterized by mercury intrusion to determine the features of the supports,⁴⁰ as shown in Fig. 2 and Table 1. The s-(A) support, as shown in Fig. 2, displayed a unimodal pore size with porosity of 57.9%, while the s-(B), which was fabricated with a higher preparation ratio of hydrotalcite, also had a unimodal pore size with porosity of 44.1%. Conversely, the s-(C) had a preparation ratio with a higher content of kaolin and showed a bimodal structure of 4 nm and 69.5 nm pore sizes, with a porosity of 63.2%. Similarly, s-(D), which had a preparation ratio with a higher aluminum nitrate content, also showed a bimodal structure of 4 nm and 200 nm pore sizes, with a porosity of 51.4%.

Fig. 2 The pore size distribution of support obtained from mercury intrusion.

The nitrogen physisorption and mercury intrusion results show an interesting inverse correlation between surface area and pore diameter. Increases in calcining temperature are well known to decrease the surface area while increasing the pore diameter due to acceleration of the sintering.^41,42 However, in the present study, the calcination temperature was fixed; only the preparation ratios were adjusted.

The SEM characterization of the support morphology can help to explain these results. The s-(A) support had hexagonal and platy clusters with a dimensional range between 2 and 4 μm, as shown in Fig. 3. The other supports s-(B)–(D), by contrast, had loosely packed aggregates of hydrotalcite clusters with enlarged inter-particle voids; i.e., few hexagonal shaped clusters were evident exist and the mostly plate-shaped forms were altered and smaller in support s-(D). These observations reflected the low surface area of 24.0 m² g⁻¹ compared to s-(A) at 120 m² g⁻¹.

Fig. 3 The morphology images of supports obtained from SEM.

The powder XRD patterns of the supports with the 2θ range of 10–80° are presented in Fig. 4. The s-(A) support shows diffraction patterns of an MgO-like phase (pericalse) or a magnesia–alumina solid indexed at 2θ = 43.0°, 62.0° and the hydrotalcite structure, designated as HY, indexed at 2θ = 11.2°, 23.1°, and 34.6°. The hydrotalcite phases were attributed to the calcination condition at 500 °C. The hydrotalcite structure was collapsed and converted to the mixed oxide (i.e., MgAl₂O₄) after the calcination process.^43,44 However, calcining at 500 °C was not sufficient to create the mixed oxide. The other support s-(B)–(D) commonly had the following diffraction patterns: a phase of MgAl₂O₄ indexed at 2θ = 36.7°, 44.7°, 59.4°, and 65.1°; kaolin (designated as Ka) indexed at 2θ = 20.8°, 26.6°, 28.3°, 39.3°, 50.6°, and 59.9°; and alumina (designated as Al) indexed at 2θ = 21.2°, 23.1° and 29.7°.^45,46 However, the intensity of each pattern arose from the different ratios in chemical composition; i.e., the support s-(C) had a higher intensity of kaolin while the support s-(D) had a higher intensity of alumina. Interestingly, the support s-(B) had a higher intensity of MgAl₂O₄. This may have reflected the relatively higher content of hydrotalcite, which may have formed spinel phases of MgAl₂O₄ due to the incorporation of alumina as an inorganic binder. This was also evidence that kaolin did not function as an inorganic binder to make the mixed oxide.

Fig. 4 XRD patterns of support before the impregnation.

Fig. 5 shows the HADDF images of the supports. The EDS mapping was conducted to examine the possible association between each metal. The Mg and Si elements represent the chemical sources of the hydrotalcite and kaolin, respectively. However, the Al dispersion over the entire analyzed area was difficult to discern, because elemental Al commonly exists in hydrotalcite and alumina; i.e., in the s-(A), the dispersion of Al entirely overlapped the dispersion of Mg. The formation of compounds like MgAl₂O₄ is also expected due to a potentially stronger interaction between Mg and Al; i.e., in the s-(B), the dispersion of Mg element entirely overlapped the dispersion of Al signal, as discussed in XRD studies. Nevertheless, the dispersion of Si maintained single-metal domains regardless of the dispersion of the Mg signals. Thus, kaolin may contribute the formation of the pore structure; i.e., in s-(B)–(D), the incorporation of kaolin to support enlarged meso-pore or formed secondary pores as the domains of the Si signals widened. However, the interpretation that the alumina is an inorganic binder that contributes to the formation of pore structure is more difficult to confirm due to the existence of mixed compounds.

Fig. 5 Mapping images of support obtained from TEM/EDS.

The results of the TGA analyses, performed to study the thermal behavior of the support, are shown in Fig. 6(a). Similar thermal behaviors with the two endothermic transitions were observed for s-(A) and s-(B). The thermal behavior of hydrotalcite was generally characterized by endothermic transitions. The first peak at a temperature below 200 °C might be attributed to the loss of interlayer water and weakly held surface water, without collapse of the hydrotalcite structure. The second and third peaks, at temperatures between 250–350 °C and at 400 °C might be attributed to decomposition of the inter layer anions in the brucite layer and the even deeper dihydroxylation of vicinal OH groups in the hydrotalcite, coupled with H₂O evolution. The result was the formation of Mg–Al mixed oxides.^39,47 However, degradation was more extensive for s-(A) than for s-(B) since the kaolin and alumina in the s-(B) slowed the degradation of hydrotalcite. By contrast, the degradation of s-(C) and s-(D) was stable with the endothermic transition and was assigned to dehydration or decomposition of alumina or kaolin resources at 200–300 °C.

Fig. 6 TGA and DTA curve of supports (a) and catalysts (b).

Aluminum nitrate undergoes a series of phase transformation after thermal treatment in air at atmospheric pressure. And endothermic dehydration begins and then transition of alumina phase (i.e., γ → δ → θ → α), which depends on the condition of thermal treatment.^29,30 i.e., γ-alumina phase formed with above 100 m² g⁻¹ of surface area at 500 °C.

In this study, the γ-alumina phase did not directly effect on the formation of pore structure but effect on the formation of spinel phase (i.e., MgAl₂O₄). However, contribution of spinel phase to the formation of pore structure is difficult to confirm.

Kaolin as clay material also undergoes a series of phase transformation after thermal treatment in air at atmospheric pressure. And endothermic dehydration begins at 550–600 °C to produce disordered meta-kaolin, Al₂Si₂O₇, but continuous hydroxyl loss (OH–) is observed up to 900 °C and has been attributed to gradual oxolation of the meta-kaolin as following eqn (6)^25,45

Si₂O₅(OH)₄Al₂ → Al₂Si₂O₇ + 2H₂O (6)

During the calcination, the structure of kaolin collapses releasing two molecule of water. In this study, the formation of meta-kaolin did not confirmed by XRD results. But calcination of kaolin at 500 °C disordered their structure, which may enlarge the inter voids between hydrotalcite clusters, resulted in formation of pore structure.

The CO₂-TPD data for the supports are shown in Fig. 7. The basicity of the supports obtained can be characterized qualitatively by the diverse CO₂ desorption profile in three ranges of temperature (100–200 °C, 200–450 °C and >450 °C). All supports had three CO₂ desorption profiles, which were related to several types of basic sites and were caused by different types of carbonate coordination in the interlayer space of the hydrotalcite. Several species, such as mono-dentate, bi-dentate, and bicarbonate anions, are known to form during the saturation of CO₂ with hydrotalcite.⁴⁸ The first peak, at below 200 °C, was mainly attributed to bicarbonate species formed by the interaction of CO₂ with hydroxyl groups (OH–) in the hydrotalcite. The second peak, at 200–450 °C, was tentatively attributed to bi-dentate carbonate species formed by the interaction of CO₂ with Mg–O (or Al–O) pairs in the hydrotalcite. The third peak, at over 450 °C, was mainly attributed to mono-dentate or unidentate carbonate species formed by the interaction of CO₂ with low-coordination anions (O₂⁻) at the corners or edges.⁴⁹ The medium and high strength basicity of s-(A) at 200–450 °C and at 450 °C was attributed to the MgO phases. In the s-(B), however, the medium basicity was increased and high strength basicity was decreased, which might be due to the increased presence of MgAl₂O₄ phases, as shown in the XRD studies. A lower basicity of Mg–Al binary oxides than MgO was reported in the literature.⁵⁰ By contrast, the s-(C) and s-(D) supports, which contained kaolin or alumina, showed decreased CO₂ desorption and basicity strength. The incorporation of kaolin or alumina could have led to increased acidity or reduced formation of MgAl₂O₄ (or MgO).

Fig. 7 CO₂ TPD profiles of supports.

3.2. The characterization of supported Co-based catalysts with different pore structures

Table 2 shows that the surface areas ranged from 24–119 m² g⁻¹, the average pore diameters were 8.6–26.7 nm, and the pore volumes were 0.14–0.26 cm³ g⁻¹. These physical quantities were significantly decreased after introduction of the cobalt precursor. The average losses in the initial surface area and pore volume were 17 and 11.5%, respectively. These reductions were attributed to a possible partial pore blockage by loading of cobalt oxide clusters during the preparation.²

Table 2 Physicochemical properties of different catalysts

Catalyst	N2 physisorption^a			XRD		Chemisorption
	BET surface area (m^2 g^−1)	Average pore diameter (nm)	Pore volume (cm^3 g^−1)	d(Co3O4)^b nm	d(Co^0)^c nm	H2^d (μmol g^−1)	CO^e (μmol g^−1)	CO/H^f	d(Co^0)^e nm	Dispersion^e (%)
a Measured by N2 physisorption [Moonsorp-II, KIST].b Co3O4 crystallite size determined by XRD (Scherrer equation).c Co^0 crystallite size and Co3O4 crystallite size by d(Co^0) = 0.75d(Co3O4).d Amount of H2 adsorbed determined by H2 chemisorption technique.e Amount of CO adsorbed determined by CO chemisorption technique. And based on the CO chemisorption H/Cos = 1, dispersion (%) = total carbon monoxide atoms chemisorbed/total number of Co atoms were calculated.f Calculated from H2 and CO uptakes as (no. of CO molecules)/(no. of H atoms).
c-(A)	99	8.0	0.23	8.5	6.4	46.94	116.4	1.23	14.5	6.86
c-(B)	49	22.6	0.14	12.0	9.0	37.55	104.1	1.38	16.3	6.13
c-(C)	29	11.0	0.12	19.4	14.5	63.41	65.96	0.52	25.6	3.88
c-(D)	20	10.3	0.20	17.7	13.3	31.71	75.22	1.18	22.5	4.43

---

The types of cobalt species were examined by powder XRD patterns obtained from fresh catalysts, as shown in Fig. 8. The phase of Co₃O₄ (denoted by ★) indexed at 2θ = 31.1°, 36.8° and 59.2°, and the phase of CoO (denoted by ☆) indexed at 2θ = 42.8° with very low crystallinity.^15–17 These phases appeared after the impregnation with cobalt. The crystallite sizes of Co₃O₄ [311] at 2θ = 36.8° were calculated using the Scherrer equation and are listed in Table 2. The crystallite size of Co₃O₄ depended mostly on the pore structure. The catalysts c-(A) and c-(B) with unimodal structure had a crystallite size of Co₃O₄ that was smaller than the mean pore diameter. By contrast, the crystallite size of Co₃O₄ was significantly larger than the meso-pore diameter in catalysts c-(C) and c-(D) that had bimodal structure. These findings imply that most of the cobalt particles will be located on the external surface of the supports. However, the pore structures of s-(C) and s-(D) were highly branched, as revealed by the SEM results. Therefore, cobalt oxide clusters could be formed and interconnecting from one pore to adjacent pores inside the support. Therefore, the cobalt particle size could possibly be larger than the pore size distribution results from the mercury porosimetry study would indicate. This could indicate another possibility whereby the lower specific surface area and the existence of macro-porosity for s-(C) and s-(D) led to aggregated cobalt oxide clusters.

Fig. 8 XRD patterns of supported cobalt catalyst before the FTS reactions.

Table 2 summarizes the sizes of the cobalt crystallites or particles as determined by XRD, H₂ chemisorption, and CO chemisorption. The H₂ and CO chemisorption results for all the catalysts, based on the same amount of cobalt loading (10 wt%) for each catalyst, are shown in Table 2. The H₂ chemisorption data for a support with a unimodal pore structure indicate that catalysts with the smallest meso-pore diameter and the highest surface area [e.g., c-(A)] had medium amounts of H₂ adsorbed, while the catalysts with the largest meso-pore diameter [e.g., c-(B)] had lower amounts of H₂ adsorption. However, the catalyst with a bimodal pore structure and a medium meso-pore diameter with the lowest surface area had the lowest amount of H₂ adsorption, while the catalyst with medium meso-pore diameter [e.g., c-(C)] had the highest H₂ adsorption. During the calcination step, the interconnected chemical compounds derived from different composition ratios led to the formation of different pore networks and made possible the formation of irreducible cobalt species by aggregation of the particles. Previous reports also supported the concept that the pores of a catalyst could be blocked by irreducible cobalt species.⁵¹ However, CO adsorption was increased with a decrease in the metallic cobalt crystallite size. In addition, the trends of the cobalt particle sizes determined by XRD paralleled the results for CO chemisorption. Since only certain crystal faces of cobalt are collected by XRD, the average sizes of cobalt crystallites estimated by Scherrer equation were much smaller than the sizes of cobalt particles determined by CO chemisorption.^5,33

The adsorbed CO/H ratio should be between 0.5 (for bridged adsorption of CO) and 1 (for linear adsorption of CO).⁵² The ratios found here were about 1.28, 1.38, and 1.18 for the c-(A), c-(B), and c-(D), respectively, or close to 1, which suggested that the linear form of CO was the main form adsorbed on the catalysts. The ratio for the c-(C) was about 0.52, suggested that the bridged form of CO was mainly adsorbed on that catalyst.

The TPR profiles are shown in Fig. 9. All the catalysts have the distinct reduction peaks. Since hydrotalcite, kaolin, and Al₂O₃ cannot be reduced in that temperature range, the observed reduction peaks should only be related to the reduction of Co species. The first reduction peak appearing at the low temperature (at 250–350 °C) was attributed to the reduction of Co₃O₄ to Co, the second one at the high temperature was ascribed to the reduction of Co interacting with supports (i.e., Co_xO_y–Al₂O₃). However, distinguishing between these compounds and support phases (i.e., MgAl₂O₄) was difficult in the XRD analysis because the phases overlapped or had similar indexes.

Fig. 9 TPR profiles of supported cobalt catalyst.

The intensity of the first peak was affected by the cluster size; i.e., the catalyst c-(C) had the largest intensity of the first peak. The intensity of the first peak varied in the following order: c-(C) > c-(D) > c-(B) > c-(A), which matched the pattern of the cobalt cluster size as estimated by XRD and CO chemisorption. The first peak was affected by the interaction, which was derived from the chemical properties of the support; i.e., for catalyst c-(D), the first peak was positioned at the highest temperature. The maximum reduction temperature of the first peak varied in the following order: c-(D) > c-(C) > c-(A) > c-(B), which matched with the contents of the alumina source in the preparation ratio. Based on these findings, the alumina content led to stronger interactions, whereas the MgO derived from hydrotalcite led to weaker interactions between cobalt and the support.

A TEM mapping technique was used to obtain a better understanding of the location of cobalt clusters and possibility of Co associations with other elements, as shown in Fig. 10. The majority of the cobalt matched well with the dispersion of Mg, although some part of the cobalt was dispersed over the Al element. However, the dark areas for cobalt elements varied with the preparation ratio of the support; i.e., the catalyst c-(C) showed greater darker areas for the part of cobalt cluster over the Mg while part of the cobalt cluster over the Al was lighter. On the contrary, the majority of cobalt over the Al was darker for catalyst c-(D). These findings confirmed the formation of irreducible compounds, as discussed for the TPR analysis (i.e., Co_xO_y–Al₂O₃). Cobalt was typically less dispersed over the Si.

Fig. 10 Mapping images of catalysts obtained from TEM/EDS.

3.3. Fischer–Tropsch activity and selectivity

The activity of CO with different selectivity was obtained under mild FTS conditions. Table 3 shows the average values for catalytic performance when the catalytic activity was stabilized after 20 h. In the present study, the sequence of activity (CTY, cobalt time yields) was c-(C) > c-(A) > c-(D) > c-(B). The data could be divided two domains, based on the characterization studies. The one domain related to the unimodal pore distribution was mostly dominated by the cobalt dispersion. For the c-(A) and c-(B) catalyst, the c-(A) had the higher CTY with a smaller cobalt cluster size than did c-(B), which depended on the number of active sites. Iglesia et al.⁹ reported that the cobalt time yield was increased with increasing dispersion, while the intrinsic activity for FTS reaction was more dependent on the number of active sites. However, the methane selectivity was higher for the c-(A) catalyst, with its small pore diameter, than for the c-(B), with wider pore diameters. The diffusional limitation of CO that occurred over small pore diameters increased the H₂/CO ratio, which resulted in higher methane selectivity. Recently, J. A. Lapszewicz et al.¹⁸ reported that methane formation rose sharply with increasing H₂/CO over a catalyst with an average pore size below 20 nm.

Table 3 The catalytic activity and hydrocarbon selectivity for prepared hydrotalcite based catalyst

Catalyst	Activity (CTY)^a	Selectivity (mol%)
		CH4	C2–C4	C5+	O/(O + P)^b
a CTY = 10^−4 molCO gCo^−1 s^−1.b The O/(O + P) value was calculated from the ratio of olefin divided by total hydrocarbons (olefin + paraffin) in the range of C2–C4.
c-(A)	5.79	11.2	2.4	86.4	0.54
c-(B)	2.00	8.4	5.4	86.2	0.80
c-(C)	7.11	14.3	2.6	83.1	0.22
c-(D)	4.27	15.4	4.1	80.5	0.50

---

The other domain related to the bimodal pore distribution was mostly dominated by the cobalt reducibility. The CTY were higher for c-(C) than c-(D) due to the higher reducibility. The highly reducible cobalt clusters (electron rich-state) of the c-(C) catalyst enhanced the facile dissociation of CO and increased adsorption of bridged-type CO, which resulted in a higher CO depletion rate. Several reports have supported these considerations with similar observations. Cluster size and reducibility of cobalt species are acknowledged to play an important role in catalytic activity, and the degree of reduction of the cobalt cluster can be greatly altered by the electron density of cobalt surfaces, which results in the CO adsorption.^32,33,54 And Goodman et al.⁵³ investigated the effects of cobalt cluster size and concluded that the TOF value remained constant but the CTY varied for cobalt cluster sizes above 10 nm.

Comparing the performance of c-(A) and c-(C), the reducibility seemed to have more impact on the catalytic activity than dispersion. The adjusted pore structure by incorporation of kaolin and alumina induced enlarged cobalt crystallite, which resulted in reduced interaction of cobalt with the support as the reducibility increased.^5,12 The previous work pointed out the activity and selectivity appeared to be more closely related to the properties of the support rather than to Co dispersion.⁵³

In this study, the pore structure had impact on the selectivity of heavy hydrocarbons. The c-(C) has a lower value of the olefin to paraffin ratio than c-(A). This effect on chain length is dependent on the re-adsorption of 1-olefin under the tested conditions;^11,36,55 i.e., the c-(C) had a lower value of O/(O + P), which indicated a high probability for secondary reactions to proceed to form heavy hydrocarbons as represented in Fig. 12.

And the more catalytic stability was observed in c-(C). Fig. 11 shows the cobalt time yields of c-(A) and c-(C) against the time on stream during the FT synthesis. The FTS sharply decreases in the first 20 hours and then levels off. Two different steps are distinguishable: the slope of the lines for the first and the second deactivation steps are different. The deactivation curve slopes steeply at first and then moderately when the time on stream exceeds 20 h over c-(C). The loss of activity for the first deactivation step can be calculated using the following linear correlation; i.e., the c-(A) had the following linear correlation initial reaction time:

X_CTY = −0.1559T_(h) + 9.0856

Fig. 11 Time on stream of cobalt time yields (CTY) for (a) c-(A) and (b) c-(C).

By contrast, the catalyst c-(C) had the following linear correlation initial reaction time:

X_CTY = −0.1962T_(h) + 11.418

The linear deactivation mode suggests that the deactivation rate is zero order to cobalt time yields. When the time on stream exceeded 20 h, the catalyst deactivation could be simulated with a power law expression; i.e., for c-(A), the deactivation rate was assumed by the following equation:

X_CTY = 17.00T_(h)^−0.29

Assuming the deactivation rate of c-(A) is:

After integration and data reduction by least squares fit, the power order (n) of c-(A) can be determined as 4.4.

In case of c-(C), the deactivation rate was assumed by the same procedure:

X_CTY = 8.88T_(h)^−0.032

The power order (n) of c-(C) can be determined as 32.3. That value for c-(C) is in the range that ordinary metal catalysts would experience during sintering. This result indicated that deactivation was caused by exterior factors, such as partial pressure of water, which caused the loss of the active site.^8,15,35 The FTS temperature used in the present study was low enough to enhance the cluster growth at the catalyst surface, but water vapor seemed to increase the oxidation reduction cycles on the catalyst surface, which in turn led to cluster growth or sintering.¹⁴ By contrast, the c-(A) was more affected by accumulation of the longer chained hydrocarbons over the pore. Botes et al.⁵⁶ hypothesized that longer chains desorb at slower rates due to stronger physical interactions, which may include effects of diffusion and physisorption. This assumption was adapted by Todic et al.,⁵⁷ who proposed a comprehensive model based on Botes's assumption. In their model, they assumed that desorption of 1-olefin was irreversible and chain-length dependent. This explanation could be supported by the by the Anderson–Schulz–Flory (ASF) product distribution, as shown in Fig. 12.

Fig. 12 ASF plot based on normal paraffin (hydrocarbons) for (a) c-(A) and (b) c-(C).

The alpha value was higher for c-(A) than for c-(C) in the range below the C₃–C₁₀. However, the alpha value higher for c-(C) was for c-(A) over the C₁₀. This observation matched the trends in the olefin to paraffin ratio, as shown in Table 3. These results suggested that the enhanced porosity of c-(C) imparted by the presence of macro-pores reduced the occurrence of mass diffusional restrictions during the FTS reaction when compared to c-(A).

Tsubaki et al.⁸ investigated the effect of a bimodal pore-structured FT catalyst and concluded that a ZrO₂–silica bimodal support is good for enhancing the diffusion efficiency of FT products, resulting in an increased TOF. The pores were filled with wax and water, which slowed pore diffusivity significantly. Diffusion limitations may result from limitations of the arrival of CO to the active sites or through the limited removal of reaction products.^9,13,17 In other words, the supported cobalt cluster with enhanced porosity was less damaged by the accumulated liquid or waxes under FTS reaction.

The results of XRD patterns and TGA profiles of a used catalyst could be used to illustrate this point. The XRD patterns of a used catalyst are shown in Fig. 8. The phase of CoO(☆) [200] indexed at 2θ = 42.4°, alpha metallic cobalt (fcc, the plane of [111]) and beta metallic cobalt (hcp, the plane of [002][101]) are marked. The size of CoO [200] calculation has not followed the Scherrer-equation due to the overlapped phases derived from the support, but the intensity of CoO was clearly higher for the used catalyst than for a fresh catalyst, in the case of c-(A) and c-(B). The thermal behavior of the catalyst, obtained from TGA profile, is shown in Fig. 6(b). The sequence of weight changes agreed with the trends in porosity. Interestingly, the catalyst c-(A) underwent significant weight changes and an intense endothermic transition, which corresponded with the dehydrated water and decomposed wax composition covering the catalyst.^3,5,34 By contrast, the c-(C) showed stable weight changes and a small endothermic transition due to the enhanced porosity imparted by the existence of macro-pores.

4. Conclusion

The characterization and activity tests of these catalysts lead to the following conclusions:

First, hydrotalcite-based supports were prepared by a slurry precipitation method with variations in the weight ratio, which included an inorganic binder to adjust the pore structure. The alumina source not only increased the porosity but also increased the possibility of MgAl₂O₄ formation, while kaolin enlarged the inter voids between hydrotalcite clusters and created a bimodal structure with enhanced the porosity.

Second, the majority of the cobalt was well dispersed over the MgO, which was converted from the hydrotalcite phase. The participation of alumina did not change the pore size, but it created irreducible cobalt species. The residual binders on the outer surface of the pore were assumed to readily form irreducible phases. On the contrary, kaolin caused pore size adjustments but did not create irreducible phases.

Lastly, the cobalt supported on a bimodal structure was less damaged by the accumulated liquid or waxes during the FTS reaction. Interestingly, the bimodal support with the relatively higher kaolin content was good at relieving the diffusion limitation of FT products, which resulted in stabilized catalytic activity. The catalytic performance of the prepared catalyst depended on cobalt reducibility and the diffusion efficiency, which were both determined by cobalt particle size and porosity.

Acknowledgements

This work was supported by Korea Institute of Science and Technology (Project No. 2E26570) and funded by Ministry of Trade, Industry and Energy, Korea. (Project No. 20142010102790).

References

1. S. Thomas and R. A. Dawe, Energy, 2003, 28, 1461–1477.
2. J. S. Jung, S. W. Kim and D. J. Moon, Catal. Today, 2012, 185, 168–174.
3. R. Oukaci, A. H. Singleton and J. G. Goodwin, Appl. Catal., A, 1999, 186, 129–144.
4. J. Hu, F. Yu and Y. Lu, Catalysts, 2012, 2, 303–326.
5. A. Y. Khodakov, W. Chu and P. Fongarland, Chem. Rev., 2007, 107, 1692–1744.
6. B. H. Davis, Top. Catal., 2005, 32, 143–168.
7. N. Tsubaki, S. Sun and K. Fujimoto, J. Catal., 2001, 199, 236–246.
8. Y. Zhang, M. Shinoda and N. Tsubaki, Catal. Today, 2004, 93–95, 55–63.
9. E. Iglesia, Appl. Catal., A, 1997, 161, 59–78.
10. A. M. Saib, M. Claeys and E. van Steen, Catal. Today, 2002, 71, 395–402.
11. G. Jacobs, W. Ma and B. Davis, Catalysts, 2014, 4, 49–76.
12. Y. Liu, J. Chen, K. Fang, Y. Wang and Y. Sun, Catal. Commun., 2007, 8, 945–949.
13. A. M. Hilmen, D. Schanke and A. Holmen, Catal. Lett., 1996, 38, 143–147.
14. J. Li, G. Jacobs, T. Das, Y. Zhang and B. Davis, Appl. Catal., A, 2002, 236, 67–76.
15. J. Zhang, J. Chen, J. Ren and Y. Sun, Appl. Catal., A, 2003, 243, 121–133.
16. Ø. Borg, S. Eri, E. a. Blekkan, S. Storsæter, H. Wigum, E. Rytter and A. Holmen, J. Catal., 2007, 248, 89–100.
17. H. Xiong, Y. Zhang, S. Wang and J. Li, Catal. Commun., 2005, 6, 512–516.
18. J. A. Lapszewicz and X. Z. Jiang, Catal. Lett., 1992, 13, 103–115.
19. Y. Zhang, Y. Yoneyama, K. Fujimoto and N. Tsubaki, Top. Catal., 2003, 26, 129–137.
20. S. J. Park, J. W. Bae, J. H. Oh, K. V. R. Chary, P. S. S. Prasad, K. W. Jun and Y. W. Rhee, J. Mol. Catal. A: Chem., 2009, 298, 81–87.
21. R. Xu and H. C. Zeng, Chem. Mater., 2001, 13, 297–303.
22. D. J. Moon, Catal. Surv. Asia, 2010, 15, 25–36.
23. M. E. Rivas, J. L. G. Fierro, R. Guil-López, M. A. Peña, V. La Parola and M. R. Goldwasser, Catal. Today, 2008, 133–135, 367–373.
24. D. J. Moon, J. S. Jung, E. H. Yang, N. Y. Kim and Y. S. Noh, et al., Pseudo Hydrotalcite Catalyst for preparing Glycerol Carbonate, Korea Patent 10-1516374, 2015.
25. D. J. Moon, S. W. Kim, J. S. Jung, E. H. Yang, N. Y. Kim and S. Y. Lee, et al., Process for preparing hydrotalcite catalyst for aqueous phase reforming of alcohols, Korea Patent 10-1516374, 2015.
26. D. J. Moon, J. S. Jung, E. H. Yang, J. T. Jung and S. W. Kim, et al., Process for preparing nickel based catalysts for scr of natural gas, US Pat., 9242228, 2016.
27. A. Di Fronzo, C. Pirola, A. Comazzi, F. Galli, C. L. Bianchi, A. Di Michele, R. Vivani, M. Nocchetti, M. Bastianini and D. C. Boffito, Fuel, 2014, 119, 62–69.
28. Y.-T. Tsai, X. Mo, A. Campos, J. G. Goodwin and J. J. Spivey, Appl. Catal., A, 2011, 396, 91–100.
29. J.-S. Jung, J.-S. Lee, G. Choi, S. Ramesh and D. J. Moon, Fuel, 2015, 149, 118–129.
30. J.-S. Jung, G. Choi, J.-S. Lee, S. Ramesh and D. J. Moon, Catal. Today, 2014, 250, 102–114.
31. R. Bechara, D. Balloy, J.-Y. Dauphin and J. Grimblot, Chem. Mater., 1999, 11, 1703–1711.
32. V. Patil, Soft Nanosci. Lett., 2012, 02, 1–7.
33. S. Rane, O. Borg, E. Rytter and A. Holmen, Appl. Catal., A, 2012, 437–438, 10–17.
34. M. Sadeqzadeh, H. Karaca, O. V. Safonova, P. Fongarland, S. Chambrey, P. Roussel, A. Griboval-Constant, M. Lacroix, D. Curulla-Ferr, F. Luck and A. Y. Khodakov, Catal. Today, 2011, 164, 62–67.
35. G. R. Moradi, Catal. Commun., 2003, 4, 27–32.
36. J. I. Di Cosimo, V. K. Díez, M. Xu, E. Iglesia and C. R. Apesteguía, J. Catal., 1998, 178, 499–510.
37. A. N. Pour, Y. Zamani, A. Tavasoli, S. M. Kamali Shahri and S. A. Taheri, Fuel, 2008, 87, 2004–2012.
38. A. Nakhaei Pour and M. R. Housaindokht, J. Ind. Eng. Chem., 2013, 20, 591–596.
39. S. Miyata, Clays Clay Miner., 1980, 28, 50–56.
40. S. P. Rigby, New methodologies in mercury porosimetry, Elsevier Masson SAS, 2002, vol. 144.
41. J. A. Toledo-Antonio, R. Gutiérrez-Baez, P. J. Sebastian and A. Vázquez, J. Solid State Chem., 2003, 174, 241–248.
42. A. Gaber, M. A. Abdel-Rahim, A. Y. Abdel-Latief and M. N. Abdel-Salam, Int. J. Electrochem. Sci., 2014, 9, 81–95.
43. E. Kanezaki, Solid State Ionics, 1998, 106, 279–284.
44. T. Hibino and A. Tsunashima, Chem. Mater., 1998, 4055–4061.
45. A. Elimbi, H. K. Tchakoute and D. Njopwouo, Construction and Building Materials, 2011, 25, 2805–2812.
46. M. L. Granizo, M. T. Blanco-Varela and A. Palomo, J. Mater. Sci., 2000, 35, 6309–6315.
47. J. Theo Kloprogge and R. L. Frost, Appl. Catal., A, 1999, 184, 61–71.
48. T. S. Saleh, K. Narasimharao, N. S. Ahmed, S. N. Basahel, S. a. Al-Thabaiti and M. Mokhtar, J. Mol. Catal. A: Chem., 2013, 367, 12–22.
49. A. Sternig, O. Diwald, S. Gross and P. V. Sushko, J. Phys. Chem. C, 2013, 117, 7727–7735.
50. H. V. Lee, J. C. Juan, N. F. Binti Abdullah, R. Nizah Mf and Y. H. Taufiq-Yap, Chem. Cent. J., 2014, 8, 30.
51. A. H. Kababji, B. Joseph and J. T. Wolan, Catal. Lett., 2009, 130, 72–78.
52. G. R. Johnson, S. Werner and A. T. Bell, ACS Catal., 2015, 5, 5888–5903.
53. Z. Wang, S. Skiles, F. Yang, Z. Yan and D. W. Goodman, Catal. Today, 2012, 181, 75–81.
54. A. Fielicke, P. Gruene, G. Meijer and D. M. Rayner, Surf. Sci., 2009, 603, 1427–1433.
55. T. Bhatelia, C. Li, Y. Sun, P. Hazewinkel, N. Burke and V. Sage, Fuel Process. Technol., 2014, 125, 277–289.
56. F. G. Botes, B. Van Dyk and C. McGregor, Ind. Eng. Chem. Res., 2009, 48, 10439–10447.
57. B. Todic, T. Bhatelia, G. F. Froment, W. Ma, G. Jacobs, B. H. Davis and D. B. Bukur, Ind. Eng. Chem. Res., 2013, 52, 669–679.

---