- Email: [email protected]
Nanocluster iron oxide-silica aerogel catalysts for methanol partial oxidation
Applied Catalysis A: General 285 (2005) 196–204 www.elsevier.com/locate/apcata
Nanocluster iron oxide-silica aerogel catalysts for methanol partial oxidation Chien-Tsung Wang *, Shih-Hung Ro Department of Chemical Engineering, National Yunlin University of Science and Technology, Yunlin 640, Taiwan, ROC Received 25 September 2004; received in revised form 21 February 2005; accepted 24 February 2005 Available online 18 March 2005
Abstract Nanostructured pure and silica-supported iron oxide materials have been prepared by the aerogel approach. Pure iron oxide powder derived from sol-gel ferric acetylacetonate formed agglomerates of 5–30 nm small crystallites of hematite and maghemite according to TEM identification of crystal faces. Depositing ferric species to mesoporous silica aerogels generated 1–5 nm particles in the amorphous matrix. They were evaluated for methanol oxidation in an ambient fixed-bed flow reactor from 225 to 300 8C. Product selectivity and oxidation activity were dependent upon iron dispersion and reactor operation. The formation of dimethyl ether was mainly related to the bulk phase and Lewis acidity of iron oxide. Active catalysts that were selective to formaldehyde and methyl formate required appropriate iron dispersion on the silica surface, including a strong electronic interaction. Methoxy transformation to formaldehyde and formate species was found to be a function of surface temperature based on a Fourier transform infrared (FT-IR) analysis. Low to moderate reactor temperature and short catalyst contact time favored methanol conversion to formaldehyde. The formation of methyl formate was found to compete with that of formaldehyde. The dependence of response time on oxygen feed attenuation suggests that mobile lattice oxide ions participate in the surface reaction and that oxygen molecules help to maintain surface iron sites highly oxidized for Lewis chemisorption and redox electron transfer. A good correlation between microstructures and reaction characteristics is proposed. # 2005 Elsevier B.V. All rights reserved. Keywords: Aerogel; Nanoparticles; Nanoclusters; Iron oxide; Silica; Methanol oxidation
1. Introduction Aerogels, nanomaterials normally prepared by sol-gel polymerization and supercritical drying, have had great potential for use in molecular recognition and chemical conversion, such as sensors, adsorbents and catalysts [1–3]. Silica aerogels are highly porous and consist of mesopores supported by a nanoparticle cross-linking framework. High porosity and surface area are attractive properties for catalytic chemisorption and favor accessibility for reaction molecules to active centers. In the NOx reduction with propane reaction, catalytic activity of aerogels is higher than that of their homologues xerogels because of the higher * Corresponding author. Tel.: +886 5 5342601x4623; fax: +886 5 5312071. E-mail address: [email protected] (C.-T. Wang). 0926-860X/$ – see front matter # 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.apcata.2005.02.029
effective diffusivity of reaction gas that is attributed to the high porosity and larger pore size [4,5]. Kim et al. [6] found that alumina-supported nickel aerogels showed a remarkably lower coking rate in methane reforming due to highly dispersed nickel particles than the rates for similar catalysts prepared by an impregnation method. High thermal stability makes aerogels, such as zirconia–silica [7], suitable for the destruction of volatile organic compounds. Recently, aerogels as catalysts have been successfully expanded into high-pressure flow reactors as well [8]. Partial oxidation of methanol has many commercial applications such as pathways to formaldehyde, methyl formate, and dimethyl ether. Production of formaldehyde over iron-molybdenum oxide catalysts is of particular industrial importance [9]. Dimethyl ether has been recognized as a promising clean alternative fuel in the next generation due to its perspective value for low
C.-T. Wang, S.-H. Ro / Applied Catalysis A: General 285 (2005) 196–204
pollutant emission and its non-toxic character [10]. Steam reforming with this ether provides a useful hydrogen source for fuel cells [11]. Further, methanol oxidation is also an excellent probe reaction to characterize oxidation catalysts using flow reactor data. The selectivity pattern and the formation rates of the reaction products provide information on the surface nature and dispersion of active sites at the molecular level [12]. Strong Lewis acidity helps to promote dehydration of methanol to dimethyl ether over metal oxide catalysts, such as niobia or alumina [13]. The high basic character of oxide catalysts tends to oxidize methanol to adsorbed formate species, followed by further decomposition to carbon oxides [12]. If redox sites are present on an oxide surface, such as molybdena, vanadia or zirconia, then formaldehyde or methyl formate is mainly produced [13–15]. However, when metal oxides generally considered as acidic types are deposited on silica, redox active sites then prevail on the catalyst surface during methanol oxidation [16]. Therefore, mixed oxides have been of great interest for use as catalysts because of the opportunity to adjust the variation in surface acidity. Iron oxide itself exhibits strong acidic and weak redox properties [17]. Most known iron oxide-silica catalysts were prepared with iron nitrate and with nitrogencontaining bases as gelation agents [18]. The present study aims to employ iron acetylacetonate simply with water in sol-gel synthesis at the molecular level, and to investigate the catalytic reactivity of silica coordinated and bulk iron oxide aerogels. Attempts have been made to demonstrate new data of aerogel structure comparable to those made from ferric salts. The methanol oxidation reaction is selected as the test probe to investigate the surface nature by aerogel evaluation in an ambient flow reactor. Detailed examination of reactor parameters, such as reaction temperature, catalyst contact time and oxygen feed, aids to determine a correlation between catalyst structure and product selectivity. Fourier transform infrared (FT-IR) and X-ray photoelectron (XPS) techniques were used to characterize surface intermediates and active iron oxide sites. The primary pathways under consideration and their gaseous standard heats of reaction (product basis) are as follows [19]. Dehydration of methanol to dimethyl ether: 2CH3 OH ! CH3 OCH3 þ H2 O; DH ¼ 19:7 kJ=mol:
(1)
197
2. Experimental 2.1. Aerogel preparation The iron and silicon oxides were synthesized with a typical two-stage sol-gel procedure. An example of preparation is described for the 20 wt.% (weight percent) iron oxide–silica aerogel. Ferric acetylacetonate (4.427 g, Fluka, 97%) and tetramethyl orthosilicate (10.138 g, Fluka, 98%) were mixed and dissolved in methanol (85 g, Tedia, 99.9%). Water triply deionized (6.564 g) was then added into the precursor solution for hydrolysis and condensation. An oxide wet gel solution formed, and complete mixing continued for another few hours. The final gel solution in a Pyrex glass liner was then placed inside a 316 S.S. autoclave reactor (American Engineering), and high temperature supercritical methanol drying followed. The overall aerogel yield per batch production is about 98 wt.%. 2.2. Catalytic evaluation Catalytic reaction tests by methanol oxidation were carried out in a fixed-bed flow apparatus at ambient pressure. A feed mixture was prepared by flowing nitrogen gas through liquid methanol placed inside a temperaturecontrolled Pyrex glass saturator. Finally, an oxygen flow was injected into the reactor. The feed stream normally at 0.7 g/min contains 2.2 mol% methanol, 5.2 mol% oxygen, and nitrogen for the balance. After being grounded into very fine powder, 0.5 g of aerogel was packed inside a tubular Pyrex glass tube (20 mm i.d.) placed in a temperature PID controlled oven. Contact time was estimated based on specific packing density and volumetric flow rate. The hydrocarbon species in the reactor inlet and effluent (methanol, dimethyl ether, formaldehyde, and methyl formate) were nicely separated with a Cowax 10 capillary column (Supelco, 60 m 0.53 mm 2.0 mm film thickness) and analyzed on a gas chromatograph (China Chromatography, Model 9800) with a flame ionization detector (Varian) in series with a thermal conductivity detector (Varian). Formic acid was never detected. In order to completely avoid total oxidation to carbon oxides, we maintained a low conversion of below 25%. The reaction temperature was incrementally raised from 225 to 300 8C, and for each step a 45-min interval was allowed for taking conversion and selectivity data at steady state.
Partial oxidation of methanol to formaldehyde: CH3 OH þ 12O2 ! HCHO þ H2 O; DH ¼ 38:0 kJ=mol:
2.3. Structural characterization (2)
Partial oxidation of methanol to methyl formate: 2CH3 OH þ O2 ! HCOOCH3 þ 2H2 O; DH ¼ 104:1 kJ=mol:
(3)
BET surface areas of aerogels were determined by 77 K nitrogen gas adsorption measurements (Micromeritics, Model ASAP 2010). Transmission electron micrographs (TEM) were taken on a JEOL 2000SX instrument. Surface morphology was examined by taking field emission scanning electron micrographs (FE-SEM, JEOL, and
198
C.-T. Wang, S.-H. Ro / Applied Catalysis A: General 285 (2005) 196–204
JSM-6340F). Fourier transform infrared spectra were obtained on a spectrometer (Perkin-Elmer, Model 1600) by introducing probe gas molecules (e.g., methanol) for adsorption onto an aerogel self-supporting wafer. Surface elements were examined with X-ray photoelectron spectroscopy on an ESCA instrument (Thermo VG Scientific, Model Sigma Probe) using Al Ka radiation (1486.6 eV) to excite photoelectrons. The binding energy scale was precisely calibrated by taking the adventitious C1s peak at 285 eV as a reference. Experimental data points were resolved by a curve-fitting procedure, and the fitting function was chosen to be the superposition of two doublets with Gaussian–Lorentzian line shapes. Crystalline structure was examined on a powder X-ray diffractometer (Mac Science, MXP3) equipped with a nickel monochromator using a Cu ˚ ). The patterns Ka radiation (20 mA, 30 kV, and l = 1.54 A were recorded in the 2u range of 5–908 with a scan rate of 48/ min. All fresh samples before analysis were subjected to pretreatment in air for 2 h at 500 8C.
Fig. 1. Field emission scanning electron micrograph of 20% iron oxide– silica aerogel.
3. Results and discussion 3.1. Aerogel structure Table 1 presents a summary of all aerogels evaluated as catalysts for this work. The weight percent of iron oxide is given as the sample composition. The pure Fe2O3 aerogel is in the form of reddish brown powder with BET surface area of 17 m2/g. Other silica-supported iron oxide samples appear monolithic and possess large surface areas of 764– 870 m2/g. Fig. 1 is a field emission scanning electron micrograph for the 20% iron oxide–silica aerogel after it was treated at 500 8C. This porous monolith exhibits an average BJH pore size of 12.1 nm (nitrogen desorption based), whose skeleton consists of 30–60 nm spherical particles connected in series (TEM not shown). Fig. 2 shows X-ray diffraction (XRD) patterns of two aerogels after they were treated at 500 8C for 2 h. Hematite (a-Fe2O3) was determined as the main crystalline phase in the pure Fe2O3 powder, according to the strong intensity of the Bragg scattering peaks in Curve a. The broadening near
Fig. 2. XRD patterns of aerogels after being treated at 500 8C for 2 h, (a) pure iron oxide and (b) 20% iron oxide–silica.
the peak baseline suggests the presence of very small crystalline particles (a few nanometers) [20]. For the 20% Fe2O3–SiO2 aerogel, the broad peak at 2u = 23–278 in Curve b is typically characteristic of amorphous silica. Further enlarging the spectrum helps one to see a low-intensity
Table 1 Structural properties and surface reduction of aerogels Aerogela
0.5% Fe2O3–SiO2 1% Fe2O3–SiO2 5% Fe2O3–SiO2 10% Fe2O3–SiO2 20% Fe2O3–SiO2 Fe2O3 a b c
Surface area by BET (m2/g)b
805 870 820 836 764 17
Packing density (g cm3)b
0.235 0.204 0.211 0.232 0.255 1.426
Weight percent-based sample composition. Measurements were taken for aerogels after being treated in air for 2 h at 500 8C. Post-reaction samples were collected directly from the reactor.
Fe2O3 particle size by TEM (nm)b
Not detected Not detected 1–2 2–3 4–5 5–30
Fe2+/Fe3+ atomic ratio by XPS Freshb
Spentc
– – – – 0.33 0.43
– – – – 0.37 0.48
C.-T. Wang, S.-H. Ro / Applied Catalysis A: General 285 (2005) 196–204
broad band at 2u = 31–378. We considered that this weak reflection was most likely associated with the presence of aFe2O3 in the silica matrix; on the contrary, Fabrizioli et al. [21] assigned it to the presence of maghemite (g-Fe2O3) or magnetite (Fe3O4) in silica at 600 8C. Other phase assignments to amorphous iron oxide nanoclusters or some crystallites were also not excluded. Possible factors for a discrepancy between all reported versions involve the selection of iron precursor and the progressive transformation of g- to a-forms with heat treatment [22–26]. Empirically, iron oxide made from an alkoxide precursor similar in nature to the one we used exhibits a hematite phase at 500 8C as determined by XRD analysis [27]. Further identification of the crystalline feature was done on a transmission electron microscope. For the 10% iron oxide–silica sample, TEM observations showed tiny dark spots spherical in shape randomly distributing in the silica matrix without massive aggregation (Fig. 3a). Upon increasing the Fe loading to 20 wt.%, many small particles like ‘‘islands’’ were observed to emerge on the silica surface (TEM not shown). The primary particle size ranges from 1 to 5 nm, increasing with iron oxide loading (Table 1). TEM examinations by bringing the sample into focus still produce no direct evidence that can indicate the crystalline characteristics of these small particles contained in the silica support. Fig. 3b is a TEM image taken for iron oxide nanocrystallites (100% Fe2O3) with grain size about 10–
199
30 nm. Some of these small basic units agglomerated to form 250 to 450 nm particles shaped like ellipses, others aggregated together to form clusters surrounding the large grains (Fig. 3c). The hematite detected by XRD patterns in Fig. 2a is directly related to the large agglomerates [29]. From Fig. 3d, we determined a straight a-Fe2O3 character by referring to a TEM measured d-spacing value (face index) of ˚ (1 0 4) [28] for 12–20 nm crystallites. On the other 2.69 A hand, Fig. 3e presents the presence of g-Fe2O3 crystallites ˚ (2 2 0) [28], with the d-spacing value (face index) 2.95 A ˚ ˚ and more other g data including 5.90 A (1 1 0), 4.82 A ˚ (1 1 3), and 2.09 A (4 0 0) [28] were measured for crystallites below 15 nm. The absence of well-defined crystalline peaks for g-Fe2O3 in Fig. 2 is probably because these particles are too small to be detected by XRD. Thus, these examinations demonstrate the coexistence of both crystalline phases of iron oxide nanoparticles. It has been reported that smaller g-Fe2O3 particles tend to aggregate with a consequent increase of the particle size, which makes the gto-a transition easier [29,30]. Cannas et al. [20] proposed that not only does the particle size affect the phase transition but also the progressive g-to-a transformation was accompanied by a contemporary growth of the g phase at the expense of the residual amorphous phase. Phase attribution to Fe3O4 would not be possible in the present work at 500 8C, since subsequent transformation to g-Fe2O3 occurs rapidly at temperature as low as 100 8C [20]. On the contrary, high
Fig. 3. Transmission electron micrographs of aerogels after being treated at 500 8C for 2 h, (a) 10% iron oxide–silica and (b–e) pure iron oxide.
200
C.-T. Wang, S.-H. Ro / Applied Catalysis A: General 285 (2005) 196–204
temperature reduction to produce metallic iron particles was never taken into account, due to the required calcination temperature being above 700 8C [31,32]. Binding energies (BE) of photoelectrons from surface iron and oxygen ions of aerogels were measured by X-ray photoelectron spectroscopy. The Fe2p spectra in Fig. 4a are characteristics of surface iron cations in the 20% iron oxide– silica aerogel calcined at 500 8C. The spots correspond to the experimental data. After deconvolution from the Fe2p3/2 broad band, two intense peaks with BE values of 711.4 and 709.5 eV were attributed to Fe3+ (Fe2O3) and Fe2+ ions (FeO), respectively [33]. Other data to confirm the high spin Fe3+ species came from the presence of a satellite peak at 715.7 eV (shake-up) [33]. Thus, iron ions in two different states coexist on the Fe–Si oxide surface, and the energy separation by 1.9 eV in between is due to a charge effect [34]. The BE value of Fe3+2p3/2 photoelectron in the bulk Fe2O3 was measured as 710.9 eV, and is lower than that for the Fe3+ bonded to silica. This indicates a strong electronic interaction between silica and iron on account of an unequal electron density and electronegativity, which is similar to the information reported in Refs. [21,35,36]. The absence of the Fe0 peak, typically located at 706.8 eV [33], excludes the possibility of iron reduction by high temperature. Fig. 4b shows two XPS bands for O 1s photoelectron of the pure Fe2O3 aerogel. One large peak at 529.8 eV was assigned to the oxygen ions bonded to iron (Fe–O), and the other at 531.5 eV is due to a hydroxide (Fe–OH) [37]. Brønsted acid
sites are very likely to represent the surface hydroxyl groups [31]. When a water molecule is bonded to a Lewis acid center, like Fe3+, a Brønsted acid site then forms [38]. The Lewis acidity of iron species is more thermally stable than Brønsted acidity [39]. Computation of the Fe2+/Fe3+ atomic ratio listed in Table 1 was done based on the peak area ratio of two corresponding Fe2p3/2 bands by way of graphical integration. When the fresh and the spent catalysts are compared, an increase in the ratio by 12%, suggests that a surface reduction has occurred after two 12-h reaction runs for methanol oxidation. Grabowska et al. [40] found a high surface reduction about 40% in the alkylation reaction due to deactivation and proposed a correlation between catalyst activity and relative contribution of the oxidation states of two Fe ions. Moreover, the atomic ratios between the Fe–Si and Fe oxides are also quite different (Table 1). There should be a higher oxidizing potential on the surface where iron is bonded to silica because of an electron density inequality between Fe and Si. Accordingly, the surface acidity or basicity has changed. Fig. 5 presents FT-IR spectra for methanol adsorbed on oxidized silica and 20% iron oxide–silica aerogels, followed by thermal evacuation at 150 and 200 8C, respectively. The sharp band at 1466 cm1 (das(CH3)) in Curve a was assigned to the methyl C–H asymmetric deformation mode [41] and was a characteristic of Si–OCH3 species. The enhanced intensity for this band in Curve b arose from overlapping with the absorbance band of surface Fe–OCH3, hence producing a peak shoulder near 1438 cm1. Iron adsorbed formaldehyde, indicated by the 1650 cm1 (n(CO)) band [41], was first observed at 150 8C. Another treatment at 200 8C (Curve c) produced iron bonded formate species, with two bands at 1567 and 1674 cm1 (nas(COO)) [41] and an intense band at 1377 cm1 (d(C–H)). Absorbance bands at 1610 cm1 (d(H2O)) [41] were attributed to molecular H2O produced from oxidation of adsorbed –OCH3 to –OCH2
Fig. 4. XPS spectra of fresh aerogels after being treated in air at 500 8C for 2 h, (a) Fe2p photoelectron of 20% iron oxide–silica and (b) O1s photoelectron of pure iron oxide.
Fig. 5. FT-IR spectra of methanol adsorbed on oxidized aerogels and after evacuation: (a) silica at 200 8C for 10 min, (b) 20% iron oxide–silica at 150 8C for 10 min, and (c) 20% iron oxide–silica at 200 8C for 10 min.
C.-T. Wang, S.-H. Ro / Applied Catalysis A: General 285 (2005) 196–204
at 150 8C and further oxidation to –OC(O)H at 200 8C. These IR bands are characteristics of surface intermediates, and will be used to compare with the flow reactor data. 3.2. Catalytic properties Methanol oxidation was used to examine the catalytic activity and surface nature of all iron-based aerogel materials prepared in this work. The main objective was to obtain a correlation between chemical dispersion and reaction characteristics. Fig. 6 shows space-time yields of products as a function of iron oxide mass loading in catalyst from 0.5 to 100%, measured at a reaction temperature of 300 8C. Conversion of methanol in the flow reactor was controlled below 23% to avoid total combustion. The yield rate of dimethyl ether (DME, CH3OCH3) increased monotonically with iron oxide content and reached a maximum on the Fe2O3 (100 wt.% in abscissa). More worthily noted is an abrupt elevation in the DME curve at Fe loadings between 1 and 5 wt.%. Meanwhile, the first presence of iron oxide particles in the silica matrix was measured, too (Table 1). If further taking a statistical analysis over the yield range for loadings between 5 and 100 wt.%, we were quite satisfied with an approximately linear correlation (r2 = 0.96). This ends up with an important conclusion that the formation activity of dimethyl ether is mainly related to the bulk phase and Lewis acidity of iron oxide. In contrast, at the catalyst composition of 0.5–1 wt.%, the DME yields were considered in appropriate relation to some acidic iron sites adjacent to the silica support, since bulk oxide particles were not observed at all in these samples by TEM. The surface reaction to this product begins with a dissociative chemisorption of methanol molecules on iron Lewis centers (Fe3+–O). Then two adjacent Lewis-bonded methoxy groups (CH3O–Fe–OH) condense to form dimethyl ether and water, and this process is followed by desorbing and leaving two oxidized Fe3+ sites behind. The oxide
Fig. 6. Product yield as a function of iron oxide loading in catalyst for methanol oxidation at 300 8C. CH3OH/O2/N2 = 2.2/5.2/92.6 mol%; contact time 0.21 s.
201
surface would never fully reduce, if the gas oxygen were supplied. The oxygen effect will be described below. The catalytic effect of Brønsted acidity associated with the iron oxide site is considered to be negligible [17]. Brønsted acid sites are present at low temperature and not regenerated as the temperature is raised, because a small amount of water formed during methanol dehydration is instantaneously removed under the reactor conditions [42]. Selectivity in the study is defined as follows. Selectivity (%) = (moles of product formed/moles of methanol converted) 100 SR, where SR is the stoichiometric ratio of methanol to product as given by reactions in Eqs. (1)–(3). Continued from Fig. 6, the yield of formaldehyde (HCHO) reached a maximum at the 5 wt.% composition, where small iron oxide particles first precipitates out from the silica surface (Table 1). It has been reported that formaldehyde was the primary product of methanol oxidation formed at monolayer coverage of the active catalyst when supported on various oxides, for example, 6– 7 wt.% MoO3/TiO2 and 15 wt.% ZrO2/SiO2 [13,14]. Obviously, this tells that iron oxide dispersed in silica at a near saturation solubility, such as 5 wt.%, possesses optimum coordination in structure and good reactivity towards partial oxidation to formaldehyde. McCormick and Alptekin [35] found that the iron highly coordinated to silica was most reactive, with a high selectivity to formaldehyde instead of carbon oxide, during partial methane oxidation over FePO4 supported on several oxides. Owning to a basicity difference between the O in methoxy and the O in silica ligand (less nucleophilic), adsorbed methoxy groups are easily oxidized into surface formaldehyde in accompany with a reduction of Fe3+ to Fe2+. This step occurred on the 150 8C surface already (Fig. 5b). The re-oxidation of reduced iron ions by gas oxygen is required for maintaining a high surface oxidizing potential, and will be proven below. For methyl formate (MF, HCOOCH3) in Fig. 6, the spacetime yield increases with iron oxide loading and reaches a maximum at the 20 wt.% composition, where the conversion of methanol (23%) is higher than any of the others measured at 300 8C. This formate is more likely to be produced by redox transformation over less acidic and more basic sites [12]. The strong interaction between iron and silica is required to produce this surface nature, as indicated by XPS data. Gao et al. [14] reported that increasing the nucleophilic character of oxygen ions of an oxide catalyst usually increased the production of methyl formate and carbon oxides during methanol oxidation. The basic character of this 20% Fe–Si aerogel was agreeable to a previous result of high conversion (96% at 300 8C) and high selectivity to carbon dioxide (90%) for methanol oxidation in a supercritical CO2 reactor [17]. Similarly, on V2O5/TiO2 the active sites responsible for the formation of methyl formate are V centers placed near to the nucleophilic oxygen ions, probably arising from the TiO2 phase or the V/Ti oxide interphase [41].
202
C.-T. Wang, S.-H. Ro / Applied Catalysis A: General 285 (2005) 196–204
Fig. 7. Effect of reaction temperature on methanol oxidation over 20% iron oxide–silica aerogel.
With regard to a quantum dispersion effect of iron when supported on silica, attempts have been done to evaluate a silica aerogel doped with 0.5 wt.% iron oxide (atomic ratio Fe/Si = 3.78 103). As observed, the highest conversion for methanol oxidation was below 5% at 300 8C, and only dimethyl ether in a trace amount was produced. As a result, extrapolation of conversion data gave rise to a negligible activity for the silica support, and the chemical inertness of silica at the present reactor conditions can be attributed to its structure being difficult to reduce [13]. Therefore, iron species were proven to be reactive for methanol oxidation due to acidity and reducibility, and the competitive reaction with the silica support determines the product selectivity pattern. The success of using silica as a support of reducible or acidic oxide species is well described in Refs. [43,44]. Fig. 7 shows how reaction temperature affects methanol conversion and product selectivity on the 20% iron oxide– silica aerogel. In the temperature range of 225–300 8C, the conversion increased monotonically up to 23%. Dimethyl ether was the most dominant product. The selectivity increases with temperature to reach a maximum around 90% at 285 8C. Formaldehyde was favorably produced at low to moderate temperature with a low conversion where the conversion is also low, whereas the formation of methyl formate was progressively activated at moderate to high temperature. The highest selectivity to methyl formate
appeared at a temperature higher than that where the maximum selectivity to formaldehyde was measured. This apparently indicates that the formation of methyl formate is competitive with that of formaldehyde. The FT-IR results (Fig. 5) regarding intermediate transformations at different temperatures are consistent with these reactor data. On heating under vacuum, oxidation of surface methoxy to adsorbed formaldehyde takes place on the catalyst surface of 150 8C (1650 cm1 in Fig. 5b). Further oxidative reaction to surface formates appears at 200 8C (1567 and 1674 cm1 in Fig. 5c). These are in good agreement with the results of Busca et al. [41]. Kinetic parameters of the activation energy (Ea;CH3 OH ) and the rate constant pre-exponential factor (ko) listed in Table 2 were estimated from Arrhenius plots based on five conversion points between 250 and 300 8C. All data computation and fitting were based on a fixed-bed reactor design and a statistical analysis at a 95% confidence limit. The apparent rate model for use was assumed to be firstorder with respect to methanol under the excess of oxygen employed (O2/CH3OH = 2.36), as suggested by many authors [13,45]. A plot of ln ko versus Ea;CH3 OH revealed a good linear correlation (r2 = 0.98), indicating a compensation effect with an isokinetic temperature around 270 5 8C [46]. These estimated energy values of 86.0–103.9 kJ/mol are not much different from 82 to 95 kJ/mol on V2O5/SiO2 [15] and 92 to 96 kJ/mol on Mo/SiO2 [47], and even higher than 85 kJ/mol required for breaking the C–H bond of methoxy [45]. Table 2 also shows first-order activation energy data for methanol dehydration, Ea;CH3 OCH3 , estimated using dimethyl ether production rates between 225 and 300 8C (six points) with a linear regression (r2 = 0.99). They are comparable to 67–71 kJ/mol on alumina for methanol dehydration to the ether [48]. The surface activity (Rs;CH3 OH ) in Table 2 has induced a correlation between iron oxide loading and surface nature. The influence of oxygen-to-methanol feed ratio on methanol oxidation was investigated by adjusting oxygen and methanol contents in feed (constant contact time 0.21 s). Fig. 8 illustrates two selectivity profiles as a function of oxygen-to-methanol molar ratio, measured at 275 8C over the 20% iron oxide–silica aerogel. The optimum O2/CH3OH ratios that maximize the formation of formaldehyde and methyl formate are around 1.0 and 2.5, respectively, and
Table 2 Arrhenius parameters and surface activity on aerogel catalysts Aerogel
ko (m3/(h kgcat.))a
Ea;CH3 OH (kJ/mol)a
Ea;CH3 OCH3 (kJ/mol)b
Rs;CH3 OH (mol/s m2)c
0.5% Fe2O3–SiO2 1% Fe2O3–SiO2 5% Fe2O3–SiO2 10% Fe2O3–SiO2 20% Fe2O3–SiO2 Fe2O3
3.34 104 2.38 108 1.09 1010 3.15 1010 9.44 1010 1.76 109
39.6 54.3 80.8 27.4 96.7 28.5 99.3 14.3 103.9 15.7 86.0 25.4
52.1 4.8 54.7 6.4 63.6 3.2 65.9 4.5 76.7 19.2 129.8 11.3
9.7 106 6.2 102 2.9 7.3 23.3 20.6
a b c
Ea
Computed by a rate model RCH3 OH ¼ ko eRT CCH3 OH from conversions at 250–300 8C. Ea Computed by a rate model RCH3 OH ! CH3 OCH3 ¼ ko eRT CCH3 OH from dimethyl ether production rates at 225–300 8C. Surface activity for first-order methanol oxidation at 300 8C.
C.-T. Wang, S.-H. Ro / Applied Catalysis A: General 285 (2005) 196–204
Fig. 8. Effect of oxygen-to-methanol molar ratio on methanol oxidation over 20% iron oxide–silica aerogel at 275 8C.
both values exceed their stoichiometric ratios, as given in Eqs. (2) and (3). A more reasonable explanation is that most of the oxygen fed to reactor has been consumed to produce a great amount of dimethyl ether in the reactor outlet, and only some has gone for practical use in the redox pathways to formaldehyde and formate. The methanol conversion (MeOH, X%) increased from 3 to 5.8% when O2/CH3OH reached 1.0, and increased additionally to 9.7% at O2/ CH3OH = 3.2. The selectivity (DME, S%) to dimethyl ether did not change and remained at 89.5 0.5%. Therefore, oxygen in the fluid phase plays a crucial role in promoting the methanol oxidation reaction. This can be clearly evidenced simply by an obvious decrease in the conversion and redox selectivity after removing the oxygen gas supply to the reactor. In order to further elucidate an influence of gas oxygen, we measured response time for the formation rate to an instant oxygen gas removal. Fig. 9 presents dimethyl ether
Fig. 9. Dimethyl ether activity loss curves in response to oxygen removal at different temperatures over 20% iron oxide–silica aerogel. Initial DME space-time yield rates (time = 0): 0.8 mmol/(s gcat.) at 250 8C, 1.5 mmol/ (s gcat.) at 275 8C, and 2.7 mmol/(s gcat.) at 300 8C.
203
Fig. 10. Effect of contact time on methanol oxidation over 20% iron oxide– silica aerogel at 275 8C. CH3OH/O2/N2 = 2.2/5.2/92.6 mol%; total gas flow 13.6–50.7 g/h.
activity loss curves as functions of on-stream time and reaction temperature over 20% iron oxide–silica. The DME activity % is defined as 100% the ratio of the space-time DME yield rate at any time without oxygen to the initial yield rate if 5.2 mol% oxygen is present in feed. The DME formation proceeded just for a few minutes and the activity rapidly fell down, once the oxygen gas supply was cut off. The higher the reactor temperature increases, the faster the response becames, and the more the activity drops down. For methyl formate, the activity rapidly dropped down by 80– 92% in only 10 min at three reactor temperatures, indicating a high need for gas oxygen during the formation. For formaldehyde, a sluggish response was obtained, such as a 75% activity off within 120 min and at 275 8C. These observations suggest that mobile lattice oxide ions participate in the surface reaction of methanol oxidation and gas oxygen acts to keep surface iron ions that have probably been reduced during the reaction at a high oxidation state. The effect of catalyst contact time on methanol oxidation is illustrated in Fig. 10. At a constant feed composition (O2/ CH3OH = 2.36), the conversion increased from 5.7 to 16.3% at 275 8C when the contact time was extended from 0.86 to 3.23 s over 20% iron oxide–silica aerogel. Short residence times facilitate the formation of formaldehyde over single iron sites. Longer residence times favor the formation of dimethyl ether and methyl formate [41]. Obviously, enough time through the reactor helps to promote the consecutive oxidation of formaldehyde to formate species. The competitive formation of the formate with formaldehyde is evidenced accordingly.
4. Conclusions Nanoclusters of pure and silica-supported iron oxide aerogel materials have been evaluated to exhibit a wide
204
C.-T. Wang, S.-H. Ro / Applied Catalysis A: General 285 (2005) 196–204
product selectivity pattern of methanol oxidation to dimethyl ether, formaldehyde and methyl formate. Pure iron oxide aerogel made from sol-gel ferric acetylacetonate is found by XRD analysis to form hematite agglomerates of small crystallites, about 5–30 nm. The basic units are identified by TEM to be in both hematite and maghemite domains. For silica supported aerogels, addition of iron oxide at 5 wt.% and higher produces 1–5 nm particles in the amorphous matrix. The formation of dimethyl ether is directly related to the bulk phase and Lewis acidity of iron oxide. Active catalysts that are selective to formaldehyde and methyl formate require optimum dispersion of iron oxide on the silica support, which is associated with a strong interaction between two cations. Formaldehyde is favorably produced at low to moderate temperature and at short catalyst contact time. The formation of methyl formate is competitive with that of formaldehyde. Measurement of response times to oxygen feed attenuation suggests a crucial role of gas oxygen in maintaining surface iron sites at a high oxidizing potential for better Lewis chemisorption and redox electron transfer. A good correlation between microstructures and reaction characteristics over the aerogel catalysts is proposed.
Acknowledgement This work is financially supported by The National Science Council in Taiwan under grant number NSC 912214-E-224-003.
[11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22]
[23]
[24] [25] [26] [27] [28] [29] [30] [31]
[32] [33] [34]
References [1] D.Y. Sasaki, C.J. Brinker, C.S. Ashley, C.E. Daitch, K.J. Shea, D.J. Rush, US Patent 6057377 (2000). [2] A. Khaleel, P.N. Kapoor, K.J. Klabunde, NanoStruct. Mater. 11 (4) (1999) 459. [3] G.M. Pajonk, Catal. Today 35 (1997) 319. [4] H. Hirashima, C. Kojima, K. Kohama, H. Imai, V. Balek, H. Hamada, M. Inaba, J. Non-Cryst. Solids 225 (1998) 153. [5] C. Hoang-Van, O. Zegaoui, P. Pichat, J. Non-Cryst. Solids 225 (1998) 157. [6] J.-H. Kim, D.J. Suh, T.-J. Park, K.-L. Kim, Appl. Catal. A: Gen. 197 (2000) 191. [7] L.M. Hair, P.R. Coronado, J.G. Reynolds, J. Non-Cryst. Solids 270 (2000) 115. [8] C.T. Wang, R.J. Willey, Catal. Today 52 (1999) 83. [9] A.P.V. Soares, M. Farinha Portela, A. Kiennemann, L. Hilaire, J.M.M. Millet, Appl. Catal. A: Gen. 206 (2001) 221. [10] A.M. Rouhi, Chem. Eng. News 29 (1995) 37.
[35] [36] [37] [38] [39] [40] [41] [42] [43] [44] [45] [46] [47] [48]
K. Takeishi, H. Suzuki, Appl. Catal. A: Gen. 260 (2004) 111. J.M. Tatiboue¨ t, Appl. Catal. A: Gen. 148 (1997) 213. H. Hu, I.E. Wachs, J. Phys. Chem. 99 (1995) 10911. X. Gao, J.L.G. Fierro, I.E. Wachs, Langmuir 15 (1999) 3169. G. Deo, I.E. Wachs, J. Catal. 146 (1994) 323. J.-M. Jehng, H. Hu, X. Gao, I.E. Wachs, Catal. Today 28 (1996) 335. C.T. Wang, R.J. Willey, J. Catal. 202 (2001) 211. P. Fabrizioli, T. Bu¨ rgi, A. Baiker, J. Catal. 206 (2002) 143. S.I. Sandler, Chemical and Engineering Thermodynamics, John Wiley & Sons, New York, 1999. C. Cannas, D. Gatteschi, A. Musinu, G. Piccaluga, C. Sangregorio, J. Phys. Chem. B 102 (1998) 7721. P. Fabrizioli, T. Bu¨ rgi, M. Burgener, S. van Doorslaer, A. Baiker, J. Mater. Chem. 12 (2002) 619. C. Cannas, M.F. Casula, G. Concas, A. Corrias, D. Gatteschi, A. Falqui, A. Musinu, C. Sangregorio, G. Spano, J. Mater. Chem. 11 (2001) 3180. F. del Monte, M.P. Morales, D. Levy, A. Fernandez, M. Ocan˜ a, A. Riog, E. Molins, K. O’Grady, C.J. Serna, Langmuir 13 (1997) 3627. M.F. Casula, A. Corrias, G. Paschina, J. Non-Cryst. Solids 293–295 (2001) 25. L. Casas, A. Roig, E. Rodrı´guez, E. Molins, J. Tejada, J. Sort, J. NonCryst. Solids 285 (2001) 37. B. Orel, M. Macˇ ek, F. Sˇ vegl, K. Kalcher, Thin Solid Films 246 (1994) 131. L. Armelao, G. Granozzi, E. Tondello, P. Colombo, G. Principi, P.P. Lottici, G. Antonioli, J. Non-Cryst. Solids 192–193 (1995) 435. R.M. Cornell, U. Schwertmann, The Iron Oxides, Wiley–VCH, 1996. M.S. Multani, P. Ayyub, Condens. Matter News 1 (1) (1991) 25. P. Ayyub, M.S. Multani, M. Barma, V.R. Polkar, R. Vijayaraghavan, J. Phys. C: Solid State Phys. 21 (1988) 2229. S. Bordiga, R. Buzzoni, F. Geobaldo, C. Lamberti, E. Giamello, A. Zecchina, G. Leofanti, G. Petrini, G. Tozzola, G. Vlaic, J. Catal. 158 (1996) 486. R. Joyner, M. Stockenhuber, J. Phys. Chem. B 103 (1999) 5963. B.M. Weckhuysen, D. Wang, M.P. Rosynek, J.H. Lunsford, J. Catal. 175 (1998) 347. A.P.V. Soares, M. Farinha Portela, A. Kiennemann, L. Hilaire, Chem. Eng. Sci. 58 (2003) 1315. R.L. McCormick, G.O. Alptekin, Catal. Today 55 (2000) 269. J.-N. Lin, J.-H. Chen, C.-Y. Hsiao, Y.-M. Kang, B.-Z. Wan, Appl. Catal. B: Environ. 36 (2002) 19. ¨ zer, F. Tepehan, Sol. Energy Mater. Sol. Cells 56 (1999) 141. N. O R.Q. Long, R.T. Yang, J. Catal. 194 (2000) 80. C.-T. Wang, R.J. Willey, J. Non-Cryst. Solids 225 (1998) 173. H. Grabowska, R. Klimkiewicz, W. Tylus, P.J. Godowski, Appl. Catal. A: Gen. 240 (2003) 111. G. Busca, A.S. Elmi, P. Forzatti, J. Phys. Chem. 91 (1987) 5263. M. Raimondo, A. De Stefanis, G. Perez, A.A.G. Tomlinson, Appl. Catal. A: Gen. 171 (1998) 85. J.-M. Jehng, J. Phys. Chem. B 102 (1998) 5816. J.-M. Jehng, H. Hu, X. Gao, I.E. Wachs, Catal. Today 28 (1996) 335. L.J. Burcham, I.E. Wachs, Catal. Today 49 (1999) 467. M. Boudart, G. Djega-Mariadassou, Kinetics of Heterogeneous Catalytic Reactions, 1984. C. Louis, J.-M. Tatiboue¨ t, M. Che, J. Catal. 109 (1988) 354. R. Tleimat-Manzalji, D. Bianchi, G.M. Pajonk, React. Kinet. Catal. Lett. 51 (1) (1993) 29.
Nanocluster iron oxide-silica aerogel catalysts for methanol partial oxidation Chien-Tsung Wang *, Shih-Hung Ro Department of Chemical Engineering, National Yunlin University of Science and Technology, Yunlin 640, Taiwan, ROC Received 25 September 2004; received in revised form 21 February 2005; accepted 24 February 2005 Available online 18 March 2005
Abstract Nanostructured pure and silica-supported iron oxide materials have been prepared by the aerogel approach. Pure iron oxide powder derived from sol-gel ferric acetylacetonate formed agglomerates of 5–30 nm small crystallites of hematite and maghemite according to TEM identification of crystal faces. Depositing ferric species to mesoporous silica aerogels generated 1–5 nm particles in the amorphous matrix. They were evaluated for methanol oxidation in an ambient fixed-bed flow reactor from 225 to 300 8C. Product selectivity and oxidation activity were dependent upon iron dispersion and reactor operation. The formation of dimethyl ether was mainly related to the bulk phase and Lewis acidity of iron oxide. Active catalysts that were selective to formaldehyde and methyl formate required appropriate iron dispersion on the silica surface, including a strong electronic interaction. Methoxy transformation to formaldehyde and formate species was found to be a function of surface temperature based on a Fourier transform infrared (FT-IR) analysis. Low to moderate reactor temperature and short catalyst contact time favored methanol conversion to formaldehyde. The formation of methyl formate was found to compete with that of formaldehyde. The dependence of response time on oxygen feed attenuation suggests that mobile lattice oxide ions participate in the surface reaction and that oxygen molecules help to maintain surface iron sites highly oxidized for Lewis chemisorption and redox electron transfer. A good correlation between microstructures and reaction characteristics is proposed. # 2005 Elsevier B.V. All rights reserved. Keywords: Aerogel; Nanoparticles; Nanoclusters; Iron oxide; Silica; Methanol oxidation
1. Introduction Aerogels, nanomaterials normally prepared by sol-gel polymerization and supercritical drying, have had great potential for use in molecular recognition and chemical conversion, such as sensors, adsorbents and catalysts [1–3]. Silica aerogels are highly porous and consist of mesopores supported by a nanoparticle cross-linking framework. High porosity and surface area are attractive properties for catalytic chemisorption and favor accessibility for reaction molecules to active centers. In the NOx reduction with propane reaction, catalytic activity of aerogels is higher than that of their homologues xerogels because of the higher * Corresponding author. Tel.: +886 5 5342601x4623; fax: +886 5 5312071. E-mail address: [email protected] (C.-T. Wang). 0926-860X/$ – see front matter # 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.apcata.2005.02.029
effective diffusivity of reaction gas that is attributed to the high porosity and larger pore size [4,5]. Kim et al. [6] found that alumina-supported nickel aerogels showed a remarkably lower coking rate in methane reforming due to highly dispersed nickel particles than the rates for similar catalysts prepared by an impregnation method. High thermal stability makes aerogels, such as zirconia–silica [7], suitable for the destruction of volatile organic compounds. Recently, aerogels as catalysts have been successfully expanded into high-pressure flow reactors as well [8]. Partial oxidation of methanol has many commercial applications such as pathways to formaldehyde, methyl formate, and dimethyl ether. Production of formaldehyde over iron-molybdenum oxide catalysts is of particular industrial importance [9]. Dimethyl ether has been recognized as a promising clean alternative fuel in the next generation due to its perspective value for low
C.-T. Wang, S.-H. Ro / Applied Catalysis A: General 285 (2005) 196–204
pollutant emission and its non-toxic character [10]. Steam reforming with this ether provides a useful hydrogen source for fuel cells [11]. Further, methanol oxidation is also an excellent probe reaction to characterize oxidation catalysts using flow reactor data. The selectivity pattern and the formation rates of the reaction products provide information on the surface nature and dispersion of active sites at the molecular level [12]. Strong Lewis acidity helps to promote dehydration of methanol to dimethyl ether over metal oxide catalysts, such as niobia or alumina [13]. The high basic character of oxide catalysts tends to oxidize methanol to adsorbed formate species, followed by further decomposition to carbon oxides [12]. If redox sites are present on an oxide surface, such as molybdena, vanadia or zirconia, then formaldehyde or methyl formate is mainly produced [13–15]. However, when metal oxides generally considered as acidic types are deposited on silica, redox active sites then prevail on the catalyst surface during methanol oxidation [16]. Therefore, mixed oxides have been of great interest for use as catalysts because of the opportunity to adjust the variation in surface acidity. Iron oxide itself exhibits strong acidic and weak redox properties [17]. Most known iron oxide-silica catalysts were prepared with iron nitrate and with nitrogencontaining bases as gelation agents [18]. The present study aims to employ iron acetylacetonate simply with water in sol-gel synthesis at the molecular level, and to investigate the catalytic reactivity of silica coordinated and bulk iron oxide aerogels. Attempts have been made to demonstrate new data of aerogel structure comparable to those made from ferric salts. The methanol oxidation reaction is selected as the test probe to investigate the surface nature by aerogel evaluation in an ambient flow reactor. Detailed examination of reactor parameters, such as reaction temperature, catalyst contact time and oxygen feed, aids to determine a correlation between catalyst structure and product selectivity. Fourier transform infrared (FT-IR) and X-ray photoelectron (XPS) techniques were used to characterize surface intermediates and active iron oxide sites. The primary pathways under consideration and their gaseous standard heats of reaction (product basis) are as follows [19]. Dehydration of methanol to dimethyl ether: 2CH3 OH ! CH3 OCH3 þ H2 O; DH ¼ 19:7 kJ=mol:
(1)
197
2. Experimental 2.1. Aerogel preparation The iron and silicon oxides were synthesized with a typical two-stage sol-gel procedure. An example of preparation is described for the 20 wt.% (weight percent) iron oxide–silica aerogel. Ferric acetylacetonate (4.427 g, Fluka, 97%) and tetramethyl orthosilicate (10.138 g, Fluka, 98%) were mixed and dissolved in methanol (85 g, Tedia, 99.9%). Water triply deionized (6.564 g) was then added into the precursor solution for hydrolysis and condensation. An oxide wet gel solution formed, and complete mixing continued for another few hours. The final gel solution in a Pyrex glass liner was then placed inside a 316 S.S. autoclave reactor (American Engineering), and high temperature supercritical methanol drying followed. The overall aerogel yield per batch production is about 98 wt.%. 2.2. Catalytic evaluation Catalytic reaction tests by methanol oxidation were carried out in a fixed-bed flow apparatus at ambient pressure. A feed mixture was prepared by flowing nitrogen gas through liquid methanol placed inside a temperaturecontrolled Pyrex glass saturator. Finally, an oxygen flow was injected into the reactor. The feed stream normally at 0.7 g/min contains 2.2 mol% methanol, 5.2 mol% oxygen, and nitrogen for the balance. After being grounded into very fine powder, 0.5 g of aerogel was packed inside a tubular Pyrex glass tube (20 mm i.d.) placed in a temperature PID controlled oven. Contact time was estimated based on specific packing density and volumetric flow rate. The hydrocarbon species in the reactor inlet and effluent (methanol, dimethyl ether, formaldehyde, and methyl formate) were nicely separated with a Cowax 10 capillary column (Supelco, 60 m 0.53 mm 2.0 mm film thickness) and analyzed on a gas chromatograph (China Chromatography, Model 9800) with a flame ionization detector (Varian) in series with a thermal conductivity detector (Varian). Formic acid was never detected. In order to completely avoid total oxidation to carbon oxides, we maintained a low conversion of below 25%. The reaction temperature was incrementally raised from 225 to 300 8C, and for each step a 45-min interval was allowed for taking conversion and selectivity data at steady state.
Partial oxidation of methanol to formaldehyde: CH3 OH þ 12O2 ! HCHO þ H2 O; DH ¼ 38:0 kJ=mol:
2.3. Structural characterization (2)
Partial oxidation of methanol to methyl formate: 2CH3 OH þ O2 ! HCOOCH3 þ 2H2 O; DH ¼ 104:1 kJ=mol:
(3)
BET surface areas of aerogels were determined by 77 K nitrogen gas adsorption measurements (Micromeritics, Model ASAP 2010). Transmission electron micrographs (TEM) were taken on a JEOL 2000SX instrument. Surface morphology was examined by taking field emission scanning electron micrographs (FE-SEM, JEOL, and
198
C.-T. Wang, S.-H. Ro / Applied Catalysis A: General 285 (2005) 196–204
JSM-6340F). Fourier transform infrared spectra were obtained on a spectrometer (Perkin-Elmer, Model 1600) by introducing probe gas molecules (e.g., methanol) for adsorption onto an aerogel self-supporting wafer. Surface elements were examined with X-ray photoelectron spectroscopy on an ESCA instrument (Thermo VG Scientific, Model Sigma Probe) using Al Ka radiation (1486.6 eV) to excite photoelectrons. The binding energy scale was precisely calibrated by taking the adventitious C1s peak at 285 eV as a reference. Experimental data points were resolved by a curve-fitting procedure, and the fitting function was chosen to be the superposition of two doublets with Gaussian–Lorentzian line shapes. Crystalline structure was examined on a powder X-ray diffractometer (Mac Science, MXP3) equipped with a nickel monochromator using a Cu ˚ ). The patterns Ka radiation (20 mA, 30 kV, and l = 1.54 A were recorded in the 2u range of 5–908 with a scan rate of 48/ min. All fresh samples before analysis were subjected to pretreatment in air for 2 h at 500 8C.
Fig. 1. Field emission scanning electron micrograph of 20% iron oxide– silica aerogel.
3. Results and discussion 3.1. Aerogel structure Table 1 presents a summary of all aerogels evaluated as catalysts for this work. The weight percent of iron oxide is given as the sample composition. The pure Fe2O3 aerogel is in the form of reddish brown powder with BET surface area of 17 m2/g. Other silica-supported iron oxide samples appear monolithic and possess large surface areas of 764– 870 m2/g. Fig. 1 is a field emission scanning electron micrograph for the 20% iron oxide–silica aerogel after it was treated at 500 8C. This porous monolith exhibits an average BJH pore size of 12.1 nm (nitrogen desorption based), whose skeleton consists of 30–60 nm spherical particles connected in series (TEM not shown). Fig. 2 shows X-ray diffraction (XRD) patterns of two aerogels after they were treated at 500 8C for 2 h. Hematite (a-Fe2O3) was determined as the main crystalline phase in the pure Fe2O3 powder, according to the strong intensity of the Bragg scattering peaks in Curve a. The broadening near
Fig. 2. XRD patterns of aerogels after being treated at 500 8C for 2 h, (a) pure iron oxide and (b) 20% iron oxide–silica.
the peak baseline suggests the presence of very small crystalline particles (a few nanometers) [20]. For the 20% Fe2O3–SiO2 aerogel, the broad peak at 2u = 23–278 in Curve b is typically characteristic of amorphous silica. Further enlarging the spectrum helps one to see a low-intensity
Table 1 Structural properties and surface reduction of aerogels Aerogela
0.5% Fe2O3–SiO2 1% Fe2O3–SiO2 5% Fe2O3–SiO2 10% Fe2O3–SiO2 20% Fe2O3–SiO2 Fe2O3 a b c
Surface area by BET (m2/g)b
805 870 820 836 764 17
Packing density (g cm3)b
0.235 0.204 0.211 0.232 0.255 1.426
Weight percent-based sample composition. Measurements were taken for aerogels after being treated in air for 2 h at 500 8C. Post-reaction samples were collected directly from the reactor.
Fe2O3 particle size by TEM (nm)b
Not detected Not detected 1–2 2–3 4–5 5–30
Fe2+/Fe3+ atomic ratio by XPS Freshb
Spentc
– – – – 0.33 0.43
– – – – 0.37 0.48
C.-T. Wang, S.-H. Ro / Applied Catalysis A: General 285 (2005) 196–204
broad band at 2u = 31–378. We considered that this weak reflection was most likely associated with the presence of aFe2O3 in the silica matrix; on the contrary, Fabrizioli et al. [21] assigned it to the presence of maghemite (g-Fe2O3) or magnetite (Fe3O4) in silica at 600 8C. Other phase assignments to amorphous iron oxide nanoclusters or some crystallites were also not excluded. Possible factors for a discrepancy between all reported versions involve the selection of iron precursor and the progressive transformation of g- to a-forms with heat treatment [22–26]. Empirically, iron oxide made from an alkoxide precursor similar in nature to the one we used exhibits a hematite phase at 500 8C as determined by XRD analysis [27]. Further identification of the crystalline feature was done on a transmission electron microscope. For the 10% iron oxide–silica sample, TEM observations showed tiny dark spots spherical in shape randomly distributing in the silica matrix without massive aggregation (Fig. 3a). Upon increasing the Fe loading to 20 wt.%, many small particles like ‘‘islands’’ were observed to emerge on the silica surface (TEM not shown). The primary particle size ranges from 1 to 5 nm, increasing with iron oxide loading (Table 1). TEM examinations by bringing the sample into focus still produce no direct evidence that can indicate the crystalline characteristics of these small particles contained in the silica support. Fig. 3b is a TEM image taken for iron oxide nanocrystallites (100% Fe2O3) with grain size about 10–
199
30 nm. Some of these small basic units agglomerated to form 250 to 450 nm particles shaped like ellipses, others aggregated together to form clusters surrounding the large grains (Fig. 3c). The hematite detected by XRD patterns in Fig. 2a is directly related to the large agglomerates [29]. From Fig. 3d, we determined a straight a-Fe2O3 character by referring to a TEM measured d-spacing value (face index) of ˚ (1 0 4) [28] for 12–20 nm crystallites. On the other 2.69 A hand, Fig. 3e presents the presence of g-Fe2O3 crystallites ˚ (2 2 0) [28], with the d-spacing value (face index) 2.95 A ˚ ˚ and more other g data including 5.90 A (1 1 0), 4.82 A ˚ (1 1 3), and 2.09 A (4 0 0) [28] were measured for crystallites below 15 nm. The absence of well-defined crystalline peaks for g-Fe2O3 in Fig. 2 is probably because these particles are too small to be detected by XRD. Thus, these examinations demonstrate the coexistence of both crystalline phases of iron oxide nanoparticles. It has been reported that smaller g-Fe2O3 particles tend to aggregate with a consequent increase of the particle size, which makes the gto-a transition easier [29,30]. Cannas et al. [20] proposed that not only does the particle size affect the phase transition but also the progressive g-to-a transformation was accompanied by a contemporary growth of the g phase at the expense of the residual amorphous phase. Phase attribution to Fe3O4 would not be possible in the present work at 500 8C, since subsequent transformation to g-Fe2O3 occurs rapidly at temperature as low as 100 8C [20]. On the contrary, high
Fig. 3. Transmission electron micrographs of aerogels after being treated at 500 8C for 2 h, (a) 10% iron oxide–silica and (b–e) pure iron oxide.
200
C.-T. Wang, S.-H. Ro / Applied Catalysis A: General 285 (2005) 196–204
temperature reduction to produce metallic iron particles was never taken into account, due to the required calcination temperature being above 700 8C [31,32]. Binding energies (BE) of photoelectrons from surface iron and oxygen ions of aerogels were measured by X-ray photoelectron spectroscopy. The Fe2p spectra in Fig. 4a are characteristics of surface iron cations in the 20% iron oxide– silica aerogel calcined at 500 8C. The spots correspond to the experimental data. After deconvolution from the Fe2p3/2 broad band, two intense peaks with BE values of 711.4 and 709.5 eV were attributed to Fe3+ (Fe2O3) and Fe2+ ions (FeO), respectively [33]. Other data to confirm the high spin Fe3+ species came from the presence of a satellite peak at 715.7 eV (shake-up) [33]. Thus, iron ions in two different states coexist on the Fe–Si oxide surface, and the energy separation by 1.9 eV in between is due to a charge effect [34]. The BE value of Fe3+2p3/2 photoelectron in the bulk Fe2O3 was measured as 710.9 eV, and is lower than that for the Fe3+ bonded to silica. This indicates a strong electronic interaction between silica and iron on account of an unequal electron density and electronegativity, which is similar to the information reported in Refs. [21,35,36]. The absence of the Fe0 peak, typically located at 706.8 eV [33], excludes the possibility of iron reduction by high temperature. Fig. 4b shows two XPS bands for O 1s photoelectron of the pure Fe2O3 aerogel. One large peak at 529.8 eV was assigned to the oxygen ions bonded to iron (Fe–O), and the other at 531.5 eV is due to a hydroxide (Fe–OH) [37]. Brønsted acid
sites are very likely to represent the surface hydroxyl groups [31]. When a water molecule is bonded to a Lewis acid center, like Fe3+, a Brønsted acid site then forms [38]. The Lewis acidity of iron species is more thermally stable than Brønsted acidity [39]. Computation of the Fe2+/Fe3+ atomic ratio listed in Table 1 was done based on the peak area ratio of two corresponding Fe2p3/2 bands by way of graphical integration. When the fresh and the spent catalysts are compared, an increase in the ratio by 12%, suggests that a surface reduction has occurred after two 12-h reaction runs for methanol oxidation. Grabowska et al. [40] found a high surface reduction about 40% in the alkylation reaction due to deactivation and proposed a correlation between catalyst activity and relative contribution of the oxidation states of two Fe ions. Moreover, the atomic ratios between the Fe–Si and Fe oxides are also quite different (Table 1). There should be a higher oxidizing potential on the surface where iron is bonded to silica because of an electron density inequality between Fe and Si. Accordingly, the surface acidity or basicity has changed. Fig. 5 presents FT-IR spectra for methanol adsorbed on oxidized silica and 20% iron oxide–silica aerogels, followed by thermal evacuation at 150 and 200 8C, respectively. The sharp band at 1466 cm1 (das(CH3)) in Curve a was assigned to the methyl C–H asymmetric deformation mode [41] and was a characteristic of Si–OCH3 species. The enhanced intensity for this band in Curve b arose from overlapping with the absorbance band of surface Fe–OCH3, hence producing a peak shoulder near 1438 cm1. Iron adsorbed formaldehyde, indicated by the 1650 cm1 (n(CO)) band [41], was first observed at 150 8C. Another treatment at 200 8C (Curve c) produced iron bonded formate species, with two bands at 1567 and 1674 cm1 (nas(COO)) [41] and an intense band at 1377 cm1 (d(C–H)). Absorbance bands at 1610 cm1 (d(H2O)) [41] were attributed to molecular H2O produced from oxidation of adsorbed –OCH3 to –OCH2
Fig. 4. XPS spectra of fresh aerogels after being treated in air at 500 8C for 2 h, (a) Fe2p photoelectron of 20% iron oxide–silica and (b) O1s photoelectron of pure iron oxide.
Fig. 5. FT-IR spectra of methanol adsorbed on oxidized aerogels and after evacuation: (a) silica at 200 8C for 10 min, (b) 20% iron oxide–silica at 150 8C for 10 min, and (c) 20% iron oxide–silica at 200 8C for 10 min.
C.-T. Wang, S.-H. Ro / Applied Catalysis A: General 285 (2005) 196–204
at 150 8C and further oxidation to –OC(O)H at 200 8C. These IR bands are characteristics of surface intermediates, and will be used to compare with the flow reactor data. 3.2. Catalytic properties Methanol oxidation was used to examine the catalytic activity and surface nature of all iron-based aerogel materials prepared in this work. The main objective was to obtain a correlation between chemical dispersion and reaction characteristics. Fig. 6 shows space-time yields of products as a function of iron oxide mass loading in catalyst from 0.5 to 100%, measured at a reaction temperature of 300 8C. Conversion of methanol in the flow reactor was controlled below 23% to avoid total combustion. The yield rate of dimethyl ether (DME, CH3OCH3) increased monotonically with iron oxide content and reached a maximum on the Fe2O3 (100 wt.% in abscissa). More worthily noted is an abrupt elevation in the DME curve at Fe loadings between 1 and 5 wt.%. Meanwhile, the first presence of iron oxide particles in the silica matrix was measured, too (Table 1). If further taking a statistical analysis over the yield range for loadings between 5 and 100 wt.%, we were quite satisfied with an approximately linear correlation (r2 = 0.96). This ends up with an important conclusion that the formation activity of dimethyl ether is mainly related to the bulk phase and Lewis acidity of iron oxide. In contrast, at the catalyst composition of 0.5–1 wt.%, the DME yields were considered in appropriate relation to some acidic iron sites adjacent to the silica support, since bulk oxide particles were not observed at all in these samples by TEM. The surface reaction to this product begins with a dissociative chemisorption of methanol molecules on iron Lewis centers (Fe3+–O). Then two adjacent Lewis-bonded methoxy groups (CH3O–Fe–OH) condense to form dimethyl ether and water, and this process is followed by desorbing and leaving two oxidized Fe3+ sites behind. The oxide
Fig. 6. Product yield as a function of iron oxide loading in catalyst for methanol oxidation at 300 8C. CH3OH/O2/N2 = 2.2/5.2/92.6 mol%; contact time 0.21 s.
201
surface would never fully reduce, if the gas oxygen were supplied. The oxygen effect will be described below. The catalytic effect of Brønsted acidity associated with the iron oxide site is considered to be negligible [17]. Brønsted acid sites are present at low temperature and not regenerated as the temperature is raised, because a small amount of water formed during methanol dehydration is instantaneously removed under the reactor conditions [42]. Selectivity in the study is defined as follows. Selectivity (%) = (moles of product formed/moles of methanol converted) 100 SR, where SR is the stoichiometric ratio of methanol to product as given by reactions in Eqs. (1)–(3). Continued from Fig. 6, the yield of formaldehyde (HCHO) reached a maximum at the 5 wt.% composition, where small iron oxide particles first precipitates out from the silica surface (Table 1). It has been reported that formaldehyde was the primary product of methanol oxidation formed at monolayer coverage of the active catalyst when supported on various oxides, for example, 6– 7 wt.% MoO3/TiO2 and 15 wt.% ZrO2/SiO2 [13,14]. Obviously, this tells that iron oxide dispersed in silica at a near saturation solubility, such as 5 wt.%, possesses optimum coordination in structure and good reactivity towards partial oxidation to formaldehyde. McCormick and Alptekin [35] found that the iron highly coordinated to silica was most reactive, with a high selectivity to formaldehyde instead of carbon oxide, during partial methane oxidation over FePO4 supported on several oxides. Owning to a basicity difference between the O in methoxy and the O in silica ligand (less nucleophilic), adsorbed methoxy groups are easily oxidized into surface formaldehyde in accompany with a reduction of Fe3+ to Fe2+. This step occurred on the 150 8C surface already (Fig. 5b). The re-oxidation of reduced iron ions by gas oxygen is required for maintaining a high surface oxidizing potential, and will be proven below. For methyl formate (MF, HCOOCH3) in Fig. 6, the spacetime yield increases with iron oxide loading and reaches a maximum at the 20 wt.% composition, where the conversion of methanol (23%) is higher than any of the others measured at 300 8C. This formate is more likely to be produced by redox transformation over less acidic and more basic sites [12]. The strong interaction between iron and silica is required to produce this surface nature, as indicated by XPS data. Gao et al. [14] reported that increasing the nucleophilic character of oxygen ions of an oxide catalyst usually increased the production of methyl formate and carbon oxides during methanol oxidation. The basic character of this 20% Fe–Si aerogel was agreeable to a previous result of high conversion (96% at 300 8C) and high selectivity to carbon dioxide (90%) for methanol oxidation in a supercritical CO2 reactor [17]. Similarly, on V2O5/TiO2 the active sites responsible for the formation of methyl formate are V centers placed near to the nucleophilic oxygen ions, probably arising from the TiO2 phase or the V/Ti oxide interphase [41].
202
C.-T. Wang, S.-H. Ro / Applied Catalysis A: General 285 (2005) 196–204
Fig. 7. Effect of reaction temperature on methanol oxidation over 20% iron oxide–silica aerogel.
With regard to a quantum dispersion effect of iron when supported on silica, attempts have been done to evaluate a silica aerogel doped with 0.5 wt.% iron oxide (atomic ratio Fe/Si = 3.78 103). As observed, the highest conversion for methanol oxidation was below 5% at 300 8C, and only dimethyl ether in a trace amount was produced. As a result, extrapolation of conversion data gave rise to a negligible activity for the silica support, and the chemical inertness of silica at the present reactor conditions can be attributed to its structure being difficult to reduce [13]. Therefore, iron species were proven to be reactive for methanol oxidation due to acidity and reducibility, and the competitive reaction with the silica support determines the product selectivity pattern. The success of using silica as a support of reducible or acidic oxide species is well described in Refs. [43,44]. Fig. 7 shows how reaction temperature affects methanol conversion and product selectivity on the 20% iron oxide– silica aerogel. In the temperature range of 225–300 8C, the conversion increased monotonically up to 23%. Dimethyl ether was the most dominant product. The selectivity increases with temperature to reach a maximum around 90% at 285 8C. Formaldehyde was favorably produced at low to moderate temperature with a low conversion where the conversion is also low, whereas the formation of methyl formate was progressively activated at moderate to high temperature. The highest selectivity to methyl formate
appeared at a temperature higher than that where the maximum selectivity to formaldehyde was measured. This apparently indicates that the formation of methyl formate is competitive with that of formaldehyde. The FT-IR results (Fig. 5) regarding intermediate transformations at different temperatures are consistent with these reactor data. On heating under vacuum, oxidation of surface methoxy to adsorbed formaldehyde takes place on the catalyst surface of 150 8C (1650 cm1 in Fig. 5b). Further oxidative reaction to surface formates appears at 200 8C (1567 and 1674 cm1 in Fig. 5c). These are in good agreement with the results of Busca et al. [41]. Kinetic parameters of the activation energy (Ea;CH3 OH ) and the rate constant pre-exponential factor (ko) listed in Table 2 were estimated from Arrhenius plots based on five conversion points between 250 and 300 8C. All data computation and fitting were based on a fixed-bed reactor design and a statistical analysis at a 95% confidence limit. The apparent rate model for use was assumed to be firstorder with respect to methanol under the excess of oxygen employed (O2/CH3OH = 2.36), as suggested by many authors [13,45]. A plot of ln ko versus Ea;CH3 OH revealed a good linear correlation (r2 = 0.98), indicating a compensation effect with an isokinetic temperature around 270 5 8C [46]. These estimated energy values of 86.0–103.9 kJ/mol are not much different from 82 to 95 kJ/mol on V2O5/SiO2 [15] and 92 to 96 kJ/mol on Mo/SiO2 [47], and even higher than 85 kJ/mol required for breaking the C–H bond of methoxy [45]. Table 2 also shows first-order activation energy data for methanol dehydration, Ea;CH3 OCH3 , estimated using dimethyl ether production rates between 225 and 300 8C (six points) with a linear regression (r2 = 0.99). They are comparable to 67–71 kJ/mol on alumina for methanol dehydration to the ether [48]. The surface activity (Rs;CH3 OH ) in Table 2 has induced a correlation between iron oxide loading and surface nature. The influence of oxygen-to-methanol feed ratio on methanol oxidation was investigated by adjusting oxygen and methanol contents in feed (constant contact time 0.21 s). Fig. 8 illustrates two selectivity profiles as a function of oxygen-to-methanol molar ratio, measured at 275 8C over the 20% iron oxide–silica aerogel. The optimum O2/CH3OH ratios that maximize the formation of formaldehyde and methyl formate are around 1.0 and 2.5, respectively, and
Table 2 Arrhenius parameters and surface activity on aerogel catalysts Aerogel
ko (m3/(h kgcat.))a
Ea;CH3 OH (kJ/mol)a
Ea;CH3 OCH3 (kJ/mol)b
Rs;CH3 OH (mol/s m2)c
0.5% Fe2O3–SiO2 1% Fe2O3–SiO2 5% Fe2O3–SiO2 10% Fe2O3–SiO2 20% Fe2O3–SiO2 Fe2O3
3.34 104 2.38 108 1.09 1010 3.15 1010 9.44 1010 1.76 109
39.6 54.3 80.8 27.4 96.7 28.5 99.3 14.3 103.9 15.7 86.0 25.4
52.1 4.8 54.7 6.4 63.6 3.2 65.9 4.5 76.7 19.2 129.8 11.3
9.7 106 6.2 102 2.9 7.3 23.3 20.6
a b c
Ea
Computed by a rate model RCH3 OH ¼ ko eRT CCH3 OH from conversions at 250–300 8C. Ea Computed by a rate model RCH3 OH ! CH3 OCH3 ¼ ko eRT CCH3 OH from dimethyl ether production rates at 225–300 8C. Surface activity for first-order methanol oxidation at 300 8C.
C.-T. Wang, S.-H. Ro / Applied Catalysis A: General 285 (2005) 196–204
Fig. 8. Effect of oxygen-to-methanol molar ratio on methanol oxidation over 20% iron oxide–silica aerogel at 275 8C.
both values exceed their stoichiometric ratios, as given in Eqs. (2) and (3). A more reasonable explanation is that most of the oxygen fed to reactor has been consumed to produce a great amount of dimethyl ether in the reactor outlet, and only some has gone for practical use in the redox pathways to formaldehyde and formate. The methanol conversion (MeOH, X%) increased from 3 to 5.8% when O2/CH3OH reached 1.0, and increased additionally to 9.7% at O2/ CH3OH = 3.2. The selectivity (DME, S%) to dimethyl ether did not change and remained at 89.5 0.5%. Therefore, oxygen in the fluid phase plays a crucial role in promoting the methanol oxidation reaction. This can be clearly evidenced simply by an obvious decrease in the conversion and redox selectivity after removing the oxygen gas supply to the reactor. In order to further elucidate an influence of gas oxygen, we measured response time for the formation rate to an instant oxygen gas removal. Fig. 9 presents dimethyl ether
Fig. 9. Dimethyl ether activity loss curves in response to oxygen removal at different temperatures over 20% iron oxide–silica aerogel. Initial DME space-time yield rates (time = 0): 0.8 mmol/(s gcat.) at 250 8C, 1.5 mmol/ (s gcat.) at 275 8C, and 2.7 mmol/(s gcat.) at 300 8C.
203
Fig. 10. Effect of contact time on methanol oxidation over 20% iron oxide– silica aerogel at 275 8C. CH3OH/O2/N2 = 2.2/5.2/92.6 mol%; total gas flow 13.6–50.7 g/h.
activity loss curves as functions of on-stream time and reaction temperature over 20% iron oxide–silica. The DME activity % is defined as 100% the ratio of the space-time DME yield rate at any time without oxygen to the initial yield rate if 5.2 mol% oxygen is present in feed. The DME formation proceeded just for a few minutes and the activity rapidly fell down, once the oxygen gas supply was cut off. The higher the reactor temperature increases, the faster the response becames, and the more the activity drops down. For methyl formate, the activity rapidly dropped down by 80– 92% in only 10 min at three reactor temperatures, indicating a high need for gas oxygen during the formation. For formaldehyde, a sluggish response was obtained, such as a 75% activity off within 120 min and at 275 8C. These observations suggest that mobile lattice oxide ions participate in the surface reaction of methanol oxidation and gas oxygen acts to keep surface iron ions that have probably been reduced during the reaction at a high oxidation state. The effect of catalyst contact time on methanol oxidation is illustrated in Fig. 10. At a constant feed composition (O2/ CH3OH = 2.36), the conversion increased from 5.7 to 16.3% at 275 8C when the contact time was extended from 0.86 to 3.23 s over 20% iron oxide–silica aerogel. Short residence times facilitate the formation of formaldehyde over single iron sites. Longer residence times favor the formation of dimethyl ether and methyl formate [41]. Obviously, enough time through the reactor helps to promote the consecutive oxidation of formaldehyde to formate species. The competitive formation of the formate with formaldehyde is evidenced accordingly.
4. Conclusions Nanoclusters of pure and silica-supported iron oxide aerogel materials have been evaluated to exhibit a wide
204
C.-T. Wang, S.-H. Ro / Applied Catalysis A: General 285 (2005) 196–204
product selectivity pattern of methanol oxidation to dimethyl ether, formaldehyde and methyl formate. Pure iron oxide aerogel made from sol-gel ferric acetylacetonate is found by XRD analysis to form hematite agglomerates of small crystallites, about 5–30 nm. The basic units are identified by TEM to be in both hematite and maghemite domains. For silica supported aerogels, addition of iron oxide at 5 wt.% and higher produces 1–5 nm particles in the amorphous matrix. The formation of dimethyl ether is directly related to the bulk phase and Lewis acidity of iron oxide. Active catalysts that are selective to formaldehyde and methyl formate require optimum dispersion of iron oxide on the silica support, which is associated with a strong interaction between two cations. Formaldehyde is favorably produced at low to moderate temperature and at short catalyst contact time. The formation of methyl formate is competitive with that of formaldehyde. Measurement of response times to oxygen feed attenuation suggests a crucial role of gas oxygen in maintaining surface iron sites at a high oxidizing potential for better Lewis chemisorption and redox electron transfer. A good correlation between microstructures and reaction characteristics over the aerogel catalysts is proposed.
Acknowledgement This work is financially supported by The National Science Council in Taiwan under grant number NSC 912214-E-224-003.
[11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22]
[23]
[24] [25] [26] [27] [28] [29] [30] [31]
[32] [33] [34]
References [1] D.Y. Sasaki, C.J. Brinker, C.S. Ashley, C.E. Daitch, K.J. Shea, D.J. Rush, US Patent 6057377 (2000). [2] A. Khaleel, P.N. Kapoor, K.J. Klabunde, NanoStruct. Mater. 11 (4) (1999) 459. [3] G.M. Pajonk, Catal. Today 35 (1997) 319. [4] H. Hirashima, C. Kojima, K. Kohama, H. Imai, V. Balek, H. Hamada, M. Inaba, J. Non-Cryst. Solids 225 (1998) 153. [5] C. Hoang-Van, O. Zegaoui, P. Pichat, J. Non-Cryst. Solids 225 (1998) 157. [6] J.-H. Kim, D.J. Suh, T.-J. Park, K.-L. Kim, Appl. Catal. A: Gen. 197 (2000) 191. [7] L.M. Hair, P.R. Coronado, J.G. Reynolds, J. Non-Cryst. Solids 270 (2000) 115. [8] C.T. Wang, R.J. Willey, Catal. Today 52 (1999) 83. [9] A.P.V. Soares, M. Farinha Portela, A. Kiennemann, L. Hilaire, J.M.M. Millet, Appl. Catal. A: Gen. 206 (2001) 221. [10] A.M. Rouhi, Chem. Eng. News 29 (1995) 37.
[35] [36] [37] [38] [39] [40] [41] [42] [43] [44] [45] [46] [47] [48]
K. Takeishi, H. Suzuki, Appl. Catal. A: Gen. 260 (2004) 111. J.M. Tatiboue¨ t, Appl. Catal. A: Gen. 148 (1997) 213. H. Hu, I.E. Wachs, J. Phys. Chem. 99 (1995) 10911. X. Gao, J.L.G. Fierro, I.E. Wachs, Langmuir 15 (1999) 3169. G. Deo, I.E. Wachs, J. Catal. 146 (1994) 323. J.-M. Jehng, H. Hu, X. Gao, I.E. Wachs, Catal. Today 28 (1996) 335. C.T. Wang, R.J. Willey, J. Catal. 202 (2001) 211. P. Fabrizioli, T. Bu¨ rgi, A. Baiker, J. Catal. 206 (2002) 143. S.I. Sandler, Chemical and Engineering Thermodynamics, John Wiley & Sons, New York, 1999. C. Cannas, D. Gatteschi, A. Musinu, G. Piccaluga, C. Sangregorio, J. Phys. Chem. B 102 (1998) 7721. P. Fabrizioli, T. Bu¨ rgi, M. Burgener, S. van Doorslaer, A. Baiker, J. Mater. Chem. 12 (2002) 619. C. Cannas, M.F. Casula, G. Concas, A. Corrias, D. Gatteschi, A. Falqui, A. Musinu, C. Sangregorio, G. Spano, J. Mater. Chem. 11 (2001) 3180. F. del Monte, M.P. Morales, D. Levy, A. Fernandez, M. Ocan˜ a, A. Riog, E. Molins, K. O’Grady, C.J. Serna, Langmuir 13 (1997) 3627. M.F. Casula, A. Corrias, G. Paschina, J. Non-Cryst. Solids 293–295 (2001) 25. L. Casas, A. Roig, E. Rodrı´guez, E. Molins, J. Tejada, J. Sort, J. NonCryst. Solids 285 (2001) 37. B. Orel, M. Macˇ ek, F. Sˇ vegl, K. Kalcher, Thin Solid Films 246 (1994) 131. L. Armelao, G. Granozzi, E. Tondello, P. Colombo, G. Principi, P.P. Lottici, G. Antonioli, J. Non-Cryst. Solids 192–193 (1995) 435. R.M. Cornell, U. Schwertmann, The Iron Oxides, Wiley–VCH, 1996. M.S. Multani, P. Ayyub, Condens. Matter News 1 (1) (1991) 25. P. Ayyub, M.S. Multani, M. Barma, V.R. Polkar, R. Vijayaraghavan, J. Phys. C: Solid State Phys. 21 (1988) 2229. S. Bordiga, R. Buzzoni, F. Geobaldo, C. Lamberti, E. Giamello, A. Zecchina, G. Leofanti, G. Petrini, G. Tozzola, G. Vlaic, J. Catal. 158 (1996) 486. R. Joyner, M. Stockenhuber, J. Phys. Chem. B 103 (1999) 5963. B.M. Weckhuysen, D. Wang, M.P. Rosynek, J.H. Lunsford, J. Catal. 175 (1998) 347. A.P.V. Soares, M. Farinha Portela, A. Kiennemann, L. Hilaire, Chem. Eng. Sci. 58 (2003) 1315. R.L. McCormick, G.O. Alptekin, Catal. Today 55 (2000) 269. J.-N. Lin, J.-H. Chen, C.-Y. Hsiao, Y.-M. Kang, B.-Z. Wan, Appl. Catal. B: Environ. 36 (2002) 19. ¨ zer, F. Tepehan, Sol. Energy Mater. Sol. Cells 56 (1999) 141. N. O R.Q. Long, R.T. Yang, J. Catal. 194 (2000) 80. C.-T. Wang, R.J. Willey, J. Non-Cryst. Solids 225 (1998) 173. H. Grabowska, R. Klimkiewicz, W. Tylus, P.J. Godowski, Appl. Catal. A: Gen. 240 (2003) 111. G. Busca, A.S. Elmi, P. Forzatti, J. Phys. Chem. 91 (1987) 5263. M. Raimondo, A. De Stefanis, G. Perez, A.A.G. Tomlinson, Appl. Catal. A: Gen. 171 (1998) 85. J.-M. Jehng, J. Phys. Chem. B 102 (1998) 5816. J.-M. Jehng, H. Hu, X. Gao, I.E. Wachs, Catal. Today 28 (1996) 335. L.J. Burcham, I.E. Wachs, Catal. Today 49 (1999) 467. M. Boudart, G. Djega-Mariadassou, Kinetics of Heterogeneous Catalytic Reactions, 1984. C. Louis, J.-M. Tatiboue¨ t, M. Che, J. Catal. 109 (1988) 354. R. Tleimat-Manzalji, D. Bianchi, G.M. Pajonk, React. Kinet. Catal. Lett. 51 (1) (1993) 29.