Catalytic properties of cobalt and nickel ferrites dispersed in mesoporous silicon oxide for ethylbenzene dehydrogenation with CO₂

Abstract

A polymeric precursor method was applied to synthesize catalysts of the general formula MFe₂O₄ (M = Co and Ni) in order to contribute to studies on ethylbenzene dehydrogenation in the presence of CO₂. The catalysts were characterized by TG, H₂-TPR, XRD, Mossbauer spectroscopy (MS), N₂ adsorption/desorption isotherms and TPD-CO₂. Investigations using XRD and MS revealed that the spinel structures of CoFe₂O₄ and NiFe₂O₄ phases were formed. The catalytic results suggested that materials with a spinel structure are particularly interesting for ethylbenzene dehydrogenation. Compared to the other catalysts synthesized the sample containing cobalt ferrite showed higher conversion and good styrene selectivity. The analysis of the spent catalyst (DRX) showed that the CoFe₂O₄ phase was stable under the reaction conditions and that a coke (TG) deposit was more pronounced for the NiFeSi.

---

1. Introduction

Catalytic ethylbenzene dehydrogenation in the presence of steam (reaction (1)) is a representative process for the production of styrene, an important monomer for synthetic polymers.^1–4

C₆H₅–CH₂CH₃(g) ⇌ C₆H₅–CH=CH₂(g) + H₂(g) (1)

The catalysts used in the above process are usually based on iron oxide, which can be further modified by the addition of promoters, such as potassium, chromium, calcium and cerium.⁵ Due to relatively fast catalytic deactivation, many research groups have dedicated their efforts to synthesize catalysts with better stability.^6–8

Currently, researchers have studied the use of carbon dioxide (CO₂) as a co-feeder gas in ethylbenzene dehydrogenization (reaction (2)) with the aim of introducing an alternative route to the steam process.^7–10

C₆H₅–CH₂CH₃(g) + CO₂(g) ⇌ C₆H₅–CH=CH₂(g) + CO(g) + H₂O(g) (2)

With this goal in mind, a large amount of catalysts with diverse composition have been tested in ethylbenzene dehydrogenization in the presence of CO₂.^7–11 Nevertheless, the results observed point to the need for more research to improve catalytic stability and thus to decrease the process cost.

Various publications have shown that, for catalysts containing hematite as the active phase, magnetite forms during the dehydrogenation reaction, which subsequently affects the conversion of ethylbenzene to styrene.^4,5,9,12 On the other hand, there are some publications pointing to materials with spinel structures, such as ferrites, as presenting remarkable catalytic activity and stability in this reaction.^13,14 In addition to their attractive redox and acid/base properties, spinel structures are versatile compounds offering structural stability.¹⁵ Therefore, materials of the general type MFe₂O₄ can be used in heterogeneous catalysis.^15–18 Specifically in the ethylbenzene dehydrogenization reaction, the ferrites of Co and/or Ni may be interesting due to their dehydrogenation ability. Additionally, these ions (Co²⁺ or Ni²⁺) may occupy different positions in the spinel structure, changing the interactions of CO₂ with the catalyst surface and the redox properties of the Fe³⁺, which is the main site for ethylbenzene adsorption.¹⁹

The synthesis of ferrites of Co and Ni with nanometric size has already been described in the literature.^20–22 Nevertheless, in order to apply these materials (ferrites of Co and Ni with nanometric size) in heterogeneous catalysis, it is necessary to avoid the sintering process. In this context, we report the synthesis and characterization of MFe₂O₄ (M = Ni or Co) compounds embedded in SiO₂ with the purpose of studying their catalytic performance for ethylbenzene dehydrogenation in the presence of carbon dioxide.

2. Experimental

2.1. Catalyst preparation

The precursor polymeric method used in this report was based on the chelation of cations (metals) by citric acid in aqueous solution.^23–25Nickel nitrate hexahydrate {Ni(NO₃)₂·6H₂O}, cobalt nitrate hexahydrate {Co(NO₃)₂·6H₂O}, iron nitrate nonahydrate {Fe(NO₃)₃·9H₂O}, tetraethylorthosilicate (TEOS), citric acid monohydrate (CA){C₆H₈O₇·H₂O} and ethylene glycol (EG) {C₂H₆O₂} were used as starting materials. A CA/metal ratio of 2 : 1 (mol) was used for all samples; the metal amount was the sum of Fe³⁺, Co²⁺ or Ni²⁺ and Si⁴⁺, and the mass ratio of CA/EG was kept at 2/3. The different samples were denoted by the general formula MFeSi, where M was the transition metal (Co or Ni) used. In addition, a sample containing only iron oxide and silicon oxide was prepared and labeled FeSi.

For the synthesis of the CoFeSi sample, 0.0085 mol and 0.017 mol of Co and Fe precursors, respectively, were dissolved in distilled water at room temperature. CA (0.151 mol) was dissolved in ethanol at 50 °C followed by the addition of 0.05 mol of a Si precursor (TEOS) with stirring. Next, the aqueous solution was added to the CA-TEOS-ethanol solution and stirred for 15 min at 50 °C. Subsequently, the EG was added and the mixture was stirred at 100 °C until it became a viscous resin. The resin was treated at 300 °C for 1 h under air, and the resulting precursor composite was milled and treated at 700 °C under a flow of air for 1 h.

2.2. Catalyst characterization

The calcination step was followed by thermogravimetric analysis (TG) using a 10 °C min⁻¹ heating rate under a flow of air at 50 ml min⁻¹ for 10 mg of a sample. The crystal structure of the metal oxides was characterized by X-ray diffraction analysis (XRD) with a CuKα irradiation source (λ = 1.5418 Å, 30 kV and 20 mA). Transmission Mössbauer spectra were recorded at room temperature and at 4.2 K in constant acceleration mode with a ⁵⁷Co (Rh) source. The Mössbauer data were fitted to discrete Lorentzian functions using the least-square fitting routine of the NORMOS^® software package. All isomer shift values (δ) were quoted relative to αFe.

Specific surface area and pore volume of the catalysts were determined by N₂ adsorption/desorption isotherms at liquid nitrogen temperature. Before N₂ adsorption, the catalyst sample (40 to 50 mg) was evacuated at 200 °C for 2 h. The TPR analyses (temperature-programmed reduction) of the catalysts were carried out in the range of 100–950 °C in a quartz reactor by passing a 5% H₂/N₂ flow (30 ml min⁻¹) over the sample at a heating rate of 10 °C min⁻¹. The H₂ consumption was monitored by a thermal conductivity detector (TCD) connected to the reactor outlet. The CO₂ temperature-programmed desorption was carried out in the range of 40–500 °C under a He flow (10 °C min⁻¹, 30 ml min⁻¹). The catalysts (200 mg) were preheated under He flow (30 ml min⁻¹) at 550 °C for 1 h, after which the temperature was decreased to 60 °C and the He flow was changed to pure CO₂ (30 ml min⁻¹ for 0.5 h). The desorbed CO₂ was detected by an online thermal conductivity detector (TCD) after passing through a trap (−20 °C) to remove any trace of water.

To assess the amount and the properties of coke deposited, the spent catalysts were analyzed by thermogravimetric analysis (TG) using a 10 °C min⁻¹ heating rate under a flow of air at 50 ml min⁻¹ and Fourier Transform Infrared (FTIR) in the range of 400 to 4000 cm⁻¹ using KBr pellets containing 1.0 wt% of the sample.

2.3. Catalytic activity

Catalytic tests were carried out in a fixed bed stainless steel microreactor using 150 mg of a sample at 550 °C and atmospheric pressure with a CO₂ to ethylbenzene molar ratio of 30. The molar flow rate of ethylbenzene was controlled at 1.446 mmol h⁻¹. Prior to reaction, the catalyst was heated under a N₂ flow (30 ml min⁻¹) up to the reaction temperature, and the reaction mixture (CO₂, ethylbenzene and N₂) was then introduced at a flow rate of 30 ml min⁻¹. The resulting catalytic ethylbenzene conversion was analyzed by gas chromatography using an instrument equipped with a FID and a nonpolar capillary column (30 m × 0.25 mm × 0.25 μm, similar to AT-1). The ethylbenzene conversion (C_Et) was calculated according to eqn (1), and the styrene selectivity (S_St) was calculated according to eqn (2).

C_Et = (ethylbenzene_in − ethylbenzene_out)/ethylbenzene_in (1)

S_St = (styrene_out)/n(hydrocarbon products) (2)

3. Results and discussion

3.1. Thermal analysis

As in previous publications,^23–25 the TG/DTG curves (not shown) could be divided into two temperature ranges for weight loss. The first range (40–300 °C) (near 10 wt%) was related to the evaporation of adsorbed water and any remaining ethylene glycol. The second stage (300–450 °C) was related to the decomposition of the residual organic material of the polymeric chains formed by polyesterification^25,26 and subsequent production of CO₂ by the decomposition of residual organic precursors (near 50 wt%). After that, no weight loss was detected up to 700 °C, which suggested that the decomposition of organic material had finished.

Employing organic precursors (CA and EG), which were removed as volatile compounds during the calcination step in the preparation of the catalysts, resulted in cavities and the simultaneous rearrangement of the solid, which might have been responsible for the morphological properties displayed by the material (pore diameter and volume).

3.2. X-Ray diffraction

After the calcination process (700 °C under airflow), the catalyst powders were analyzed by X-ray diffraction (XRD) to gather information on the solid phase formed (Fig. 1). In spite of the low crystalline organization, the XRD patterns of the catalyst CoFeSi suggested the formation of a spinel structured phase (CoFe₂O₄, JCPDS 02-1045). Likewise, the sample NiFeSi also pointed to the spinel structure of nickel ferrite (NiFe₂O₄, JCPDS 44-1485) in addition to the presence of nickel oxide (NiO, JCPDS 78-0423). The FeSi sample showed a hematite pattern (α-Fe₂O₃, JCPDS 86-2368) and a maghemite phase (γ-Fe₂O₃, JCPDS 39-1346).

Fig. 1 X-Ray diffraction profile of the catalysts after the calcination process.

Due to the similarity in the diffraction profile of the nickel ferrite and cobalt ferrite with the maghemite phase (γ-Fe₂O₃), the presence of a γ-Fe₂O₃ phase could not be initially ruled out for CoFeSi and NiFeSi samples.

From the results shown above, it seemed that nickel ferrite (NiFe₂O₄) was easier to form than cobalt ferrite (CoFe₂O₄) despite the fact that the ionic radii of Fe³⁺ (0.67 Å) were smaller than the radii of Ni²⁺ (0.83 Å) and closer to Co²⁺ (0.68 Å). It therefore seemed reasonable to assume that a replacement of the Fe³⁺ ions by the Co²⁺ ions in both the tetrahedral A and octahedral B sites of the cation sublattices with the corresponding decrease in the lattice parameters²⁷ would be more favourable than by the larger Ni²⁺ ions.

To clarify the XRD results, the diagrams were analyzed by means of the Rietveld refinement procedure²⁸ from which the crystallographic lattice parameters of the samples were determined. To achieve this, the DBWS9807 program described by Young et al.²⁹ was applied. By this method it was possible to quantify the different phases formed and the resulting data are listed in Table 1. The crystallite sizes of the samples were calculated from the XRD patterns using the Debye–Scherrer formula.³⁰ The calculated results illustrated in Table 1 indicated that the catalysts were nanometre-sized crystalline powders and probably with a high number of crystalline defects.

Table 1 Relative mass percentage and grain size of the phases observed

Crystalline phase	Sample
	CoFeSi		NiFeSi		FeSi
	Mass^a (%)	Gs/nm	Mass^a (%)	Gs/nm	Mass^a (%)	Gs/nm
a Mass of crystalline phase (wt%). Gs = grain size (nm).
NiFe2O4			86.2	11.4
CoFe2O4	100.0	4.5
α-Fe2O3					88.5	13.2
γ-Fe3O4					11.5	4.1
NiO			13.8	15.7

---

3.3. Mössbauer spectroscopy

As observed in the XRD results, the small size of the particles synthesized and the slight difference in the lattice parameters of the maghemita (γ-Fe₂O₃) and ferrite phases (CoFe₂O₄ and NiFe₂O₄) made the XRD insufficient for characterization of the powders. As shown in Fig. 2a, there were two sextets very close to one another and a superparamagnetic central doublet in the Mössbauer spectra of FeSi at room temperature. In addition, one six-line subspectra with an isomer shift of δ = 0.30 mm s⁻¹ relative to αFe, a quadrupole shift of ε = −0.20 mm s⁻¹, and a magnetic hyperfine field Bhf = 51.5 tesla were also observed. This type of pattern is typical of hematite (α-Fe₂O₃). The other one six-line subspectra had δ = 0.50 mm s⁻¹, a quadrupole shift of ε = ∼0.02 mm s⁻¹, and Bhf = 50 tesla, which are all characteristics of maghemite.³¹

Fig. 2 ⁵⁷Fe Mössbauer spectra of the ferrites, at room temperature (300 K) and at 4.2 K. FeSi (a), CoFeSi (b) and NiFeSi (c).

Alternatively, the Mössbauer spectrum of the FeSi sample was also acquired at a low temperature because the influence of superparamagnetism could be counteracted by reducing the sample temperature. The superparamagnetic relaxation was suppressed at 4.2 K (Fig. 2a), but the presence of different iron environments in the sample was evidenced by two six-line subspectra. This result corroborated with XRD observations on the presence of hematite and maghemite phases in the samples.

Similarly, as shown in Fig. 2b and c, the room temperature Mössbauer spectra of the CoFeSi and NiFeSi samples exhibited the superposition of a magnetic split sextet pattern and a superparamagnetic doublet. The Mössbauer spectrum at 4.2 K for the sample CoFeSi was fitted with four subspectra. Two six-line subspectra with isomer shifts of δ = 0.25 and 0.38 mm s⁻¹ (relative to αFe), quadrupole shifts of ε = 0.02 and −0.08 mm s⁻¹, and magnetic hyperfine fields of Bhf = 50.3 and 52.3 tesla were attributed to cobalt ferrite. These hyperfine parameters corresponded to a distribution of Fe in site A (tetrahedral) and site B (octahedral). Nevertheless, a broad sextet with B_hf = 30–49 tesla was also observed, which was probably due to cobalt ferrite with an extremely small particle size. The broad sextet is typical for Fe particles in the magnetic relaxation regime. Moreover, at 4.2 K, great deals of magnetic moments were blocked, as evidenced by the appearance of a broad and asymmetric magnetic sextet.³² In Fig. 2b (at 4.2 K), we see that 90% of particles were blocked and the remaining central doublet (10%) suggested that there were very small superparamagnetic particles (less than 4 nm). These results were in agreement with the mean diameter for a CoFeSi sample (4 nm), as estimated by XRD analysis (Table 1).

The Mössbauer spectrum at 4.2 K for the NiFeSi sample was fitted with three magnetic subspectra. The hyperfine field values for sites A (tetrahedral) and B (octahedral) were 49.6 and 52.8 tesla, respectively. The corresponding isomer shift values were 0.31 and 0.40 mm s⁻¹ relative to αFe. In addition, one broad sextet with B_hf = 30–51 tesla was observed, which was probably due to nickel ferrite with a very small particle size. In addition, the decrease of the magnetic hyperfine field from the bulk value could be attributed to the very small particle size.³³ In the spectra of NiFeSi (Fig. 2c, 4.2 K), unlike CoFeSi (Fig. 2b, 4.2 K), we saw that 100% of the particles were blocked, suggesting that the size of the nanoparticles was larger than 4 nm of diameter. Thus, the Mössbauer results were in agreement with the X-ray diffraction patterns, which suggested a particle diameter of 11.4 nm (Table 1).

3.4. Temperature-programmed reduction

To acquire more information on the properties of the catalysts and to clarify the chemical environment of the iron oxide, which influences the reducibility of the catalyst, temperature-programmed reduction (TPR) analyses were carried out (Fig. 3). At first the TPR pattern of the FeSi catalyst suggested four peaks of reduction. However, this pattern is somewhat complex and was decomposed into five well-defined Gaussian-shaped peaks and an incomplete one at high temperature (agreement factor of 0.997).

Fig. 3 Temperature-programmed reduction (TPR) profiles of the samples.

The H₂ consumption attained a maximum at 430 °C and was related to the reduction of Fe³⁺ to Fe²⁺, probably due to the formation of the magnetite phase (Fe₃O₄).^9,12 In addition, the peak at 575 °C, which overlapped with the first (430 °C) and the other peaks at higher temperatures, might have been due to the reduction of Fe²⁺ to Fe⁰. This low reduction temperature (575 °C) might have been the result of the small particle diameter, as indicated by XRD analysis (Table 1). Furthermore, the H₂ consumption peak at 575 °C was probably due to Fe²⁺ reduction because of the amount of H₂ consumption observed for the peaks at higher temperatures, which was lower than the peak at 430 °C. This phenomenon suggested either incomplete Fe²⁺ reduction or the formation of FeO instead of the desired Fe₃O₄ phase at 430 °C. Furthermore, certainly the three H₂ consumption peaks in the temperature range of 600–820 °C (with a maximum at 645, 705 and 780 °C), as well as a peak at temperatures higher than 900 °C, were associated with Fe²⁺ reduction to produce metallic iron.^12,34

A possible explanation for the presence of four H₂ consumption peaks at temperature ranges higher than 600 °C for the FeSi sample might be drawn from the textural properties. Metal oxide particles placed inside of porous materials are more difficult to reduce compared to metal oxides positioned on external surfaces.^12,35 Therefore, the peak at 645 °C might have been due to Fe²⁺ reduction outside of the porous channel, while the peaks at 705 and 780 °C and higher than 900 °C might have been from the reduction of Fe²⁺ inside the porous channel. Furthermore, the TPR profile of the FeSi sample suggested (as did the XRD) a high proportion of iron in the Fe³⁺ oxidation state, which was interesting for dehydrogenation.^4,9,12

The TPR profile of the NiFeSi catalyst had a starting H₂ consumption process near 200 °C, and a shoulder at 520 °C was clearly observed. The pattern was decomposed into five well-defined Gaussian-shaped peaks and an incomplete one at high temperature (agreement factor of 0.998). The peak at 414 °C could be attributed to the Ni²⁺ and Fe³⁺ (to Fe²⁺) reduction processes. Similar to the FeSi sample, the peak at 426 °C could be attributed to the reduction of Fe³⁺ to Fe²⁺. The Ni²⁺ reduction in a temperature range was typical for a nickel oxide phase dispersed in SiO₂,³⁶ which was determined to be present by XRD analysis (Fig. 1). Nevertheless, the peak at 500 °C could be attributed to the reduction of Ni²⁺ in the spinel phase, which had an intense interaction with the iron oxide and might have been localized in octahedral or tetrahedral positions. The peaks at temperatures higher than 600 °C were assumed to be the result of Fe²⁺ reduction.

The TPR profile of the CoFeSi catalyst presented a poorly defined pattern, but it was successfully decomposed into seven well-defined Gaussian-shaped peaks and an incomplete one at high temperature (agreement factor of 0.998). The weak peak at 345 °C was associated to the reduction of the Co₃O₄ phase (Co³⁺ to Co²⁺). The low H₂ consumption observed corroborated with the XRD analysis, which did not indicate the presence of this phase. The presence of this phase might have been missed, however, due to a smaller proportion. The peak at 436 °C was attributed mainly to the reduction of Fe³⁺ (to Fe²⁺). Similar to the NiFeSi sample, the peaks at 518 °C and at 524 °C could be also attributed to the reduction of Fe³⁺ to Fe²⁺ and the reduction of Co²⁺ to Co⁰, both in the spinel phase and possibly localized in octahedral or tetrahedral positions. The peaks for temperatures higher than 600 °C might have been due to Fe²⁺ reduction influenced by the cobalt presence and/or textural properties.

Different from the NiFeSi sample the CoFeSi catalyst showed low H₂ consumption peaks in the temperature range of 600 °C to 700 °C. In addition, it is observed a well defined H₂ consumption peak at 840 °C, which may be due to the CoFe₂O₄ phase reduction.

The H₂ consumption peaks that appeared at temperatures higher than 600 °C were particularly interesting for ethylbenzene dehydrogenation because they provided evidence that the reaction temperature (550 °C) did not completely reduce the transitions metals.

Assuming that the iron oxide species (Fe³⁺) was the main active site for ethylbenzene dehydrogenation,⁶ the TPR profiles suggested that the addition of nickel additives could not delay the deactivation process during the catalytic test. However, and at the same level of importance, the TPR patterns pointed to a mixture of different oxide phases, as did XRD and Mössbauer analysis. In addition to the low particle diameter (Table 1), this heterogeneity of phases in the catalyst composition suggested that these samples might present a significant number of defects in the surface, an interesting phenomenon for the catalytic process.

On the other hand, the H₂ consumption profile in a temperature range higher than 600 °C for the CoFeSi sample suggests that cobalt addition could delay the catalyst deactivation due to a reduction process during the catalytic test carried out at 550 °C.

3.5. Specific surface area measurements

From the perspective of the catalysis and the catalytic characterization, the surface area determination was an important parameter for application. Thus, the N₂ adsorption/desorption isotherms were measured and the isotherm profile and pore diameter distribution are presented in Fig. 4a and b. The samples CoFeSi and NiFeSi showed similar isotherm profiles that were characteristic of mesoporous materials (type IV according to IUPAC classification and the H2 hysteresis loop). This behavior indicated that the pore structure was complex and consisted of interconnected networks of pores of different sizes and shapes. On the other hand, the sample FeSi showed isotherms of type II characteristics of mesoporous powders (H3 hysteresis loop). A type H3 loop is usually assigned to slit-shaped pores due to non-rigid aggregates of particles.³⁷

Fig. 4 N₂ adsorption/desorption isotherms of the catalysts (a). Pore diameter distribution of the samples (b) (calculated using the BJH model).

The profiles of pore diameter distribution are presented in Fig. 4b (according to the BJH method) and exhibited a narrow pore size distribution, with the majority of the diameters located in the range of 16 to 238 Å. Additionally, Table 2 lists the surface area, total pore volume and average pore diameter of the studied powders. In spite of the similarity in the average pore diameter and pore diameter distribution (Fig. 4b) for the samples, the addition of transition metals (Co and Ni) to the catalyst promoted the synthesis of samples with better textural properties. For instance, the catalyst NiFeSi possessed higher specific surface area (49% higher) and total pore volume (40% higher) in comparison with the FeSi sample, while the sample CoFeSi showed a specific surface area 18% higher than the FeSi sample.

Table 2 Textural properties of the samples determined by N₂ adsorption isotherms

Sample	S/m^2 g^−1	V p/cm^3 g^−1	D p/Å
CoFeSi	346	0.55	63
NiFeSi	439	0.76	70
FeSi	294	0.54	74

---

Such textural effects in the porosity of the synthesized solids might be related to the ratio of residual organic precursors, as verified in the TG analysis. As such, samples with a spinel structure (CoFeSi and NiFeSi) were seen to display a higher ratio of residual organic substances. After calcination, this can result in differences in the characteristics of the pores, which are perceptible in the values for the surface area, pore volume, pore diameter and changes in the hysteresis behaviour.

3.6. Temperature-programmed desorption

It is accepted that the basic sites play an important role in ethylbenzene dehydrogenation.¹⁹ Therefore, it was interesting to determine the effect of nickel and cobalt addition on the basicity of the catalyst samples. To verify the relative basicity, temperature-programmed desorption of CO₂ (TPD-CO₂) experiments was carried out (Fig. 5). The data in Fig. 5 are in arbitrary units, and the intensity of the deflection was not quantified with regard to the number of moles of CO₂. Nevertheless, the experimental conditions were rigorously the same for all samples, and thus it was possible to compare the areas of the TPD profiles.

Fig. 5 Temperature-programmed desorption of CO₂ (TPD-CO₂) for the catalysts after calcination.

The TPD patterns of NiFeSi and FeSi samples showed one peak for CO₂ desorption, while the catalyst CoFeSi presented two peaks for desorption. The maximum desorption peak for NiFeSi was observed around 90 °C, whereas the maximum for FeSi was around 118 °C. On the other hand, the desorption curves corresponding to the CoFeSi sample were around 97 and 368 °C, respectively. The patterns of TPD-CO₂ were decomposed for all samples into two or three well-defined Gaussian-shaped peaks. Therefore, different types of basic sites (or CO₂ to metal oxides interactions) might be distinguished.

The first CO₂ desorption temperature range (53–200 °C) was attributed to weak basicity. However, the second CO₂ desorption temperature range (337–420 °C) was typical of a medium basicity. Thus, the peak of CO₂ desorption at 370 °C suggested that the CoFeSi sample had a higher basic site strength. Consequently, it was expected that the CoFeSi would have the highest ethylbenzene conversion, followed by FeSi and then NiFeSi. Considering the amount of CO₂ desorbed from the sample (the total areas of the TPD profiles), the number of basic sites followed the order: FeSi > CoFeSi > NiFeSi. NiFeSi stood out as a solid with relatively low basicity, while FeSi had a large amount of basic sites.

In general, basic catalysts can improve the adsorption of acidic CO₂ to supply the surface with oxygen, suppressing the deposition of coke.³⁸ Additionally, the number and strength of the basic sites of the catalyst had an effect on the catalytic performance.³⁹ Several research papers^12,40 have shown that the catalytic activity for ethylbenzene dehydrogenization is increased with improved CO₂ adsorption ability.

3.7. Catalytic performance

The ethylbenzene conversion and styrene selectivity for the catalytic samples are depicted in Fig. 6a and b. All the samples presented a similar activity and showed a reasonable selectivity in ethylbenzene dehydrogenization. However, some important differences were observed.

Fig. 6 Catalytic performance as a function of time for the catalysts: ethylbenzene conversion (a), styrene selectivity (b).

The CoFeSi sample demonstrated a higher ethylbenzene conversion and styrene selectivity than either FeSi or NiFeSi catalyst. Initially, the NiFeSi and FeSi catalysts presented a similar ethylbenzene conversion (near 25%); however, the catalytic deactivation was higher for the FeSi sample.

Analysis of the byproducts observed for the samples can help in the interpretation of catalyst properties. Considering benzene selectivity (Table 3), which is the main byproduct for all the catalysts tested, it is important to point out that the NiFeSi sample showed the highest benzene selectivity (near 23%) up to 120 min. At first, a superior benzene production meant higher catalyst acidity and consequently, more significant carbon deposition. Taking into account the TPD-CO₂ measurements, the NiFeSi sample presented the lowest ability for CO₂ adsorption by mass of any catalyst tested, which meant higher acidity relative to the FeSi and CoFeSi samples.

Table 3 Catalytic selectivity observed for by-products

Samples	Selectivity^a (%)
	Benzene	Toluene	Ethylene + ethane + methane
a Average values for reaction period up to 120 min.
CoFeSi	2.5	0.8	1.1
NiFeSi	23.5	1.3	0.8
FeSi	8.6	—	—

---

The deactivation of catalysts may occur due to the reduction of the active phase and/or by blocking of the surface by carbonaceous deposits.¹² To investigate the cause of catalyst deactivation during ethylbenzene dehydrogenization, studies were conducted to verify possible changes in the chemical properties of the catalysts, as well as determining the amount and nature of coke deposits on the catalysts. For this, XRD analysis after the catalytic tests were performed, and the results are depicted in Fig. 7. These data made it possible to verify some differences from the fresh catalysts (Fig. 2).

Fig. 7 X-Ray diffraction profiles of the samples after the catalytic tests.

The pattern presented in Fig. 7 for the FeSi sample pointed to Fe³⁺ reduction during the reaction process. In addition, the diffraction pattern of the FeSi sample pointed to magnetite phase formation (Fe₃O₄ JCPDS 26-1136) and the maghemite phase was still detected (γ-Fe₂O₃ JCPDS 39-1346).

Through careful XRD pattern (Fig. 7) analysis, refinement based on the Rietveld method²⁸ was carried out, and the weight percentages for the different phases detected are listed in Table 4. Interestingly, for the FeSi sample, maghemite was the major phase (89 wt%) in the used sample but not the fresh sample (Table 1), suggesting that the maghemite phase was formed from hematite during the reaction process simultaneously to magnetite phase formation. From the literature,⁴¹ however, the hematite phase has a superior stability to maghemite. In spite of the fact that experimental data presented here are not enough to support this model, a possible explanation for the observed result was the formation of maghemite particles within a shell of hematite. These core–shell particles were then partially reduced during ethylbenzene dehydrogenation, maintaining the maghemite in the interior. If Fe²⁺ was not as active in ethylbenzene dehydrogenation as Fe³⁺,¹³ the core–shell particle formation theory might explain the observed decreases in the conversion of ethylbenzene presented in Fig. 6a.

Table 4 Relative mass percentage of the phases observed after the catalytic process

Crystalline phase (%)	Samples
	CoFeSi	NiFeSi	FeSi
NiFe2O4	—	27	—
CoFe2O4	100	—	—
γ-Fe2O3	—	—	89
Fe3O4		—	11
NiO	—	64	—
Ni	—	9	—

---

On the other hand, it is necessary to carry out analysis in order to clarify if the reaction conditions (CO₂ presence) are favourable to maghemite formation.

The reaction process also promoted the partial reduction of the NiFeSi sample, forming metallic nickel (Ni, JCPDS 01-1258). However, the NiFeSi pattern suggested that a ratio of the structures of nickel ferrite (NiFe₂O₄, JCPDS 44-1485) and nickel oxide (NiO, JCPDS 78-0423) was maintained after 24 h of reaction time. On the other hand, the relative mass percentage (Table 4) calculated indicated a decrease in the wt% of NiFe₂O₄ and an increase for NiO. These results suggested that the Fe³⁺ from the spinel phase was reduced, which then likely promoted magnetite phase formation without Ni²⁺ reduction. This hypothesis was suggested because the NiO wt% for the used sample (Table 4) was higher than for the fresh sample (Table 1). This observation might be related to a superior stability for the Ni²⁺ species relative to the Fe³⁺ or that a higher proportion of tetrahedral sites were occupied by Ni²⁺. The disproportionate occupation would consequently lead to less exposure to the reduction process. The metallic nickel formed might have been from the reduction of the NiO detected in the fresh sample. Furthermore, it is well known that nickel in the metallic state (Ni⁰) is able to promote a C–H bond disproportionation reaction,⁴² which for some reactions is responsible for a high ratio of coke deposition.^13,42 Thus, it was expected that the NiFeSi sample would present a high amount of coke deposition.

In contrast to the other samples, the CoFeSi catalyst seemed to not undergo phase modification due to the reaction process. The XRD pattern of CoFeSi after the reaction was very similar to that for the fresh catalyst (Fig. 1), which indicated a superior stability for the CoFe₂O₄ phase under the reaction conditions. If the change did occur, it was in a very low proportion. In addition, taking into account the by-product selectivity (Table 3), the CoFeSi sample showed a lower benzene selectivity, which suggested a low coke deposition.

Therefore, the XRD results strengthen the suggestion that the H₂ consumption peak at 840 °C showed by the CoFeSi sample (Fig. 3) may be due to CoFe₂O₄ phase reduction.

In spite of the valuable information acquired from the catalytic test expressed in Fig. 6, the reaction time used for the FeSi sample was only of 6 h due to the high deactivation. Therefore, catalytic tests with a period of 24 h were carried out for the samples CoFeSi and NiFeSi. The respective ethylbenzene conversions were 13% and 10%, which suggested the same rate of deactivation for both samples over 24 h. Additionally, the ethylbenzene conversion observed for the NiFeSi suggested that high surface area alone was not a determining factor for superior catalytic conversion. In spite of the higher surface area presented by the NiFeSi catalyst (Table 2), the ethylbenzene conversion values were lower than the CoFeSi sample. Therefore, the CoFeSi catalyst was the most active solid (conversion near 17% after 4 h) due to the combined effects of basicity, surface area and reducibility. Despite the H₂-TPR profile of CoFeSi resembling that of NiFeSi, the XRD analysis of the spent catalysts revealed that the spinel structure of the CoFeSi sample was maintained, presenting a hard reducibility under the ethylbenzene dehydrogenation conditions and the temperature-programmed desorption of CO₂ revealed a medium basicity for the CoFeSi sample, justifications for its better catalytic performance.

To find another reason for the decrease in ethylbenzene conversion, the spent catalysts were analyzed by thermogravimetric analysis (TG) (Fig. 8).

Fig. 8 Thermogravimetric profile of the samples carried out under synthetic air flow. NiFeSi and CoFeSi after the catalytic test of 24 h, FeSi after 6 h of time on stream.

This thermogravimetric analysis of the spent catalysts was carried out to estimate the amount of coke deposited, as revealed by the reduction in weight due to coke oxidation in air. Notice that the thermal analysis results (Fig. 8) pointed to a lower coke deposition for FeSi, while the NiFeSi catalyst presented a higher coke deposition. The thermograms of three coked catalyst samples revealed coke contents of 45.0, 71.2, and 3.2 wt% for the CoFeSi, NiFeSi and FeSi samples, respectively. However, it is necessary to notice the time on stream of each catalyst before the TG analysis.

The catalytic test data, TPD-CO₂ measurements, X-ray diffraction and thermal analysis of the spent catalysts were in good agreement and showed that the NiFeSi sample presented higher coke deposition, lower ethylbenzene conversion and higher benzene selectivity. These properties were probably due to the lower CO₂ adsorption ability (Fig. 5) and the formation of metallic nickel (Fig. 7) in this sample. Furthermore, these properties could explain the lower selectivity for styrene of this solid.

The thermograms of coked samples exhibited a coke oxidation temperature range of 300 to 450 °C for all the catalysts, although the sample FeSi also showed coke elimination (burning) with an inflection point at 540 °C (Fig. 8), which is related to amorphous carbon.³⁵ Such behaviour (a low temperature burning) was characteristic of either a carbon deposit with a low molecular weight or a low C/H ratio^43–45 but also of carbonaceous material containing oxygen in its structure.^43,46

To study the molecular structure of the coke deposited on the spent catalysts, the samples were dissolved in concentrated HF and the material (coke) extracted was analyzed by Fourier transformed infrared spectroscopy (FT-IR) (Fig. 9).

Fig. 9 Fourier-transform infrared (FTIR) spectroscopy of material (coke) extracted from the samples, after the catalytic tests.

The spectroscopy results revealed the presence of aliphatic C–H (2918 cm⁻¹ with low intensity) and a band at 1600 cm⁻¹ for frame vibrations of benzene.⁴⁷ Additionally, a peak at 740 cm⁻¹ was due to the C–H in-plane bending.⁴⁸ All these peaks illustrated that the precursors of the coke were mainly saturated hydrocarbons.

Further inspecting the FTIR spectrum, a peak at 1400 cm⁻¹ was due to the C–H deformation vibration of a vinyl group bonded to oxygen.⁴⁹ However, this peak could also be assigned to the C=O group of a carboxylate or carbonate-type species.⁵⁰ The presence of a C=O group in the carbonaceous material was reinforced by a peak with low intensity at 1705 cm⁻¹ for FeSi. In addition, the bands at 1200, 1160 and 1110 cm⁻¹ were assigned to OH, COC and CO groups.⁵¹ The presence of an OH group in the carbonaceous samples was also suggested by a wide peak at 3418 cm⁻¹; however, the presence of residual water in the carbonaceous sample could not be ruled out.

The coke precursors were mainly composed of aliphatic hydrocarbons as well as oxygen-containing and polyaromatic structures. The adsorption peaks of the FT-IR for aliphatic hydrocarbons and oxygen-containing moieties were much weaker than the peaks for the aromatic hydrocarbons. Therefore, there was a larger amount of aromatic hydrocarbons in the coke precursor. Accordingly, the results from TG and FT-IR analysis suggested that the coke was generated from ethylbenzene but also from CO₂.

The observation from the data of TG and FTIR that the coke can be deposited as much of the ethylbenzene as of the carbon dioxide corroborates with the value of ΔG° at 550 °C (−7,9 kJ mol⁻¹) for the reaction (3).

CO₂ + 2H₂ ⇌ 2H₂O + C (3)

In Table 5 are given the relative values of the amount of deposited coke and the ethylbenzene converted for each catalyst, according to the period of reaction time. The data in Table 5 show that the catalyst NiFeSi presents a high ratio of coke formation. Similarly, the data obtained by TG indicate a significant amount of coke in the sample CoFeSi, if compared to the catalyst FeSi.

Table 5 Amount of coke deposited and ethylbenzene converted during the catalytic test

Property	Samples
	CoFeSi	NiFeSi	FeSi
Etbz conv. = total of ethylbenzene converted during the reaction test. X = mass of coke (mg) divided by reaction time and by Etbz conv. (mg).
Coke/wt%	45.0	71.2	3.2
Mass of coke/mg	122.72	370.83	4.95
Reaction time/h	24	24	6
Etbz conv./mg	569.86	417.53	77.27
X	0.0089	0.0370	0.0106

---

Conversely, if it is taken into account the reaction time that each catalyst was submitted and the total amount of ethylbenzene converted during this period, it is verified that the ratio of coke deposition for the catalyst CoFeSi is similar (or somewhat lower) to the FeSi sample.

Therefore, although the coke deposition was not suppressed, the addition of Co can be interesting because it can be observed a superior ratio of ethylbenzene conversion without the depreciation of the selectivity for styrene.

Considering the X value from Table 5, the catalytic deactivation (Fig. 6a) could not be attributed mainly to the coke deposition, in spite of the fact that its presence can make difficult the active site accessibility for ethylbenzene.

The catalytic deactivation observed, however, can be attributed mainly to the partial reduction of iron oxide, as suggested by the TPR profile (Fig. 3). As pointed out by Fig. 6a, a higher ratio of catalytic deactivation takes place in the first two hours of time on stream. This behaviour corroborates well with the XRD suggestions, which pointed out that the catalyst samples have a high number of crystalline defects. For the catalytic process a surface with crystalline defects may be more active, because generally a highest surface free energy is observed in surfaces with lower coordination surface atoms. On the other hand, it may present a superior tendency of deactivation.

Accordingly, the XRD profile of the CoFeSi sample after the catalytic test did not pointed out a new phase formation due to the low crystalline diameters or the amorphous characteristic.

4. Conclusions

In this study, cobalt and nickel ferrites dispersed in mesoporous silicon oxide were synthesized by a polymeric precursor method and tested as catalysts for ethylbenzene dehydrogenation in the presence of CO₂. The CoFeSi catalyst presented higher ethylbenzene conversion and good styrene selectivity, which were ascribed to the poor reducibility under the reaction conditions and the existence of medium strength basic sites. Conversely, the NiFeSi sample showed higher coke deposition, lower ethylbenzene conversion and high benzene selectivity, probably as a result of the lower CO₂ adsorption ability and the formation of metallic nickel during the reaction process for this material. Furthermore, the lower stability observed for the solid FeSi may be correlated with the formation of Fe₃O₄ during the ethylbenzene reaction. Therefore, in spite of the high coke deposition, the results suggested that materials with a spinel structure, such as ferrites of cobalt, might be particularly useful for ethylbenzene dehydrogenation. On the other hand, as the results suggested that ethylbenzene and CO₂ contribute to coke deposition, it is necessary to carry out studies in order to better understand the CO₂ role in the ethylbenzene dehydrogenation with CO₂.

Acknowledgements

The authors acknowledge the “Universidade Federal do Ceará” (UFC), CNPq/CT-PETRO, and Dr J.M. Sasaki (Laboratório de Raios X) for XRD measurements. T.P. Braga, B.M.C. Sales and A.N. Pinheiro express gratitude for scholarships from CNPq and CAPES.

References

1. E. H. Lee, Catal. Rev.–Sci. Eng., 1973, 8, 285–305.
2. V. Zhyznevskiy, R. Tsybukh and V. Gumenetskiy, React. Kinet. Catal. Lett., 2000, 71, 209–215.
3. K. Othmer, Encyclopedia of Chemical Technology, Willey, New York, 1984, p. 770.
4. M. S. Ramos, M. S. Santos, L. P. Gomes, A. Albornoz and M. C. Rangel, Appl. Catal., A, 2008, 341, 12–17.
5. A. B. Styles, in Applied Industrial Catalysis, ed. B. E. Leach, Academic Press, New York, 1983, p.137.
6. B. D. Herzog and H. F. Rase, Ind. Eng. Chem. Prod. Res. Dev., 1984, 23, 187–196.
7. N. Mimura and M. Saito, Catal. Today, 2000, 55, 173–178.
8. G. Carja, Y. Kameshima and K. Okada, Microporous Mesoporous Mater., 2008, 115, 541–547.
9. T. P. Braga, E. Longhinotti, A. N. Pinheiro and A. Valentini, Appl. Catal., A, 2009, 362, 139–146.
10. M. Surgino, H. Shimada, T. Turuda, H. Miura, N. Ikenaga and T. Suzuki, Appl. Catal., A, 1995, 121, 125–137.
11. Y. Ohishi, T. Kawabata, T. Shishido, K. Tanaki, Q. Zhang, Y. Wang and K. Takehira, J. Mol. Catal. A: Chem., 2005, 230, 49–58.
12. T. P. Braga, A. N. Pinheiro, C. V. Teixeira and A. Valentini, Appl. Catal., A, 2009, 366, 193–200.
13. R. M. Freire, A. L. Pinheiro, A. C. Oliveira, F. F. de Sousa, A. P. Ayala, J. M. Filho, P. T. C. Freire and A. C. Oliveira, Appl. Catal., A, 2009, 359, 165–179.
14. A. Miyakoshi, A. Ueno and M. Ichikawa, Appl. Catal., A, 2001, 216, 137–146.
15. N. Ballarini, F. Cavani, S. Passeri and K. Wilson, Appl. Catal., A, 2009, 366, 184–192.
16. S. Kameoka, T. Tanabe and A. P. Tsai, Appl. Catal., A, 2010, 375, 163–171.
17. J. Ak, T. Ghaddar, A. Ghanem and H. El-Rassy, J. Mol. Catal. A: Chem., 2009, 312, 18–22.
18. E. Manova, T. Tsoncheva, D. Paneva, J. L. Rehspringer, K. Tenchev, I. Mitov and L. Petrov, Appl. Catal., A, 2007, 317, 34–42.
19. Y. Joseph, M. Wuhn, A. Niklewski, W. Ranke, W. Weiss, C. Woll and R. Schlogla, Phys. Chem. Chem. Phys., 2000, 2, 5314–5319.
20. S. M. Montemayor, L. A. García-Cerda, J. R. Torres-Lubián and O. S. Rodríguez-Fernández, Mater. Res. Bull., 2008, 43, 1112–1118.
21. E. Manova, T. Tsoncheva, Cl. Estournès, D. Paneva, K. Tenchev, I. Mitov and L. Petrov, Appl. Catal., A, 2006, 300, 170–180.
22. A. A. Mirzaei, A. Beig babaei, M. Galavy and A. Youssef, Fuel Process. Technol., 2010, 91, 335–347.
23. A. Valentini, N. L. V. Carreño, L. F. D. Probst, E. R. Leite and E. Longo, Microporous Mesoporous Mater., 2004, 68, 151–157.
24. A. Valentini, N. L. V. Carreño, L. F. D. Probst, A. Barison, A. G. Ferreira, E. R. Leite and E. Longo, Appl. Catal., A, 2006, 310, 174–182.
25. A. Ries, A. Z. Simões, M. Cilense, M. A. Zaghete and J. A. Varela, Mater. Charact., 2003, 50, 217–221.
26. J. C. Forti, P. Olivi and A. R. de Andrade, Electrochim. Acta, 2001, 47, 913–920.
27. C. F. Windisch Jr, K. F. Ferris, G. J. Exarhos and S. K. Sharma, Thin Solid Films, 2002, 420–421, 89–99.
28. H. M. Rietveld, J. Appl. Crystallogr., 1967, 2, 65–71.
29. R. A. Young, A. Sakthivel, T. S. Moss and C. O. Paiva-Santos, J. Appl. Crystallogr., 1995, 28, 366–367.
30. M. I. Mendelson, J. Am. Ceram. Soc., 1969, 52, 443–446.
31. M. Gotić, G. Koščec and S. Music, J. Mol. Struct., 2009, 924–926, 347–354.
32. M. Rajendran, R. C. Pullar, A. K. Bhattacharya, D. Das, S. N. Chintalapudi and C. K. Ajumdar, J. Magn. Magn. Mater., 2001, 232, 71–83.
33. J. Jacob and M. A. Khadar, J. Appl. Phys., 2010, 107(11), 114310–114310.
34. J. C. Ryu, D. H. Lee, K. S. Kang, C. S. Park, J. W. Kim and Y. H. Kim, J. Ind. Eng. Chem. (Washington, D. C.), 2008, 14, 252–260.
35. B. Pawelec, L. Daza, J. L. G. Fierro and J. A. Anderson, Appl. Catal., A, 1996, 145, 307–322.
36. J. Guo, Z. Hou, J. Gao and X. Zheng, Fuel, 2008, 87, 1348–1354.
37. G. A. El-Shobaky, A. M. Turky, N. Y. Mostafa and S. K. Mohamed, J. Alloys Compd., 2010, 493, 415–422.
38. L. Xiancai, W. Min, L. Zhihua and H. Fei, Appl. Catal., A, 2005, 290, 81–86.
39. V. R. Choudhary, S. A. R. Mulla and B. S. Uphade, Fuel, 1999, 78, 427–437.
40. R. Dziembaj, P. K. Strowski and L. Chmielarz, Appl. Catal., A, 2003, 255, 35–43.
41. C. Cannas, D. Gatteschi, A. Musinu, G. Piccaluga and C. Sangregorio, J. Phys. Chem. B, 1998, 102, 7721–7726.
42. A. W. Chester, Ind. Eng. Chem. Res., 1987, 26, 863–869.
43. H. Hattori, Chem. Rev. (Washington, DC, U. S.), 1995, 95, 537–558.
44. A. Effendi, K. Hellgardt, Z.-G. Zhang and T. Yoshida, Catal. Commun., 2003, 4, 203–207.
45. M. Guisnet and P. Magnoux, Appl. Catal., A, 2001, 212, 83–96.
46. F. Cavani and F. Trifirò, Appl. Catal., A, 1995, 133, 219–239.
47. Y. He and Y. He, Catal. Today, 2002, 74, 45–51.
48. Y. W. Chen, Q. H. Li and W. L. Huang, Organic instrumental analysis, Hunan University Publishing Company, Hunan, 1st edn, 1987.
49. S.-H. Wang and P. R. Griffiths, Fuel, 1985, 64, 229236.
50. W. Wei, J. A. Moulijn and G. Mul, J. Catal., 2009, 262, 1–8.
51. T. Jiang and K. Xu, Carbon, 1995, 33, 1663–1671.

---