Role of Palladium in Iron Based Fischer−Tropsch Catalysts Prepared

Dec 14, 2010 - Melanie Minnermann, Suman Pokhrel, Karsten Thiel, Robert Henkel, .... Miriam Schubert , Suman Pokhrel , Andreas Thomé , Volkmar Zielas...
0 downloads 0 Views 2MB Size
1302

J. Phys. Chem. C 2011, 115, 1302–1310

Role of Palladium in Iron Based Fischer-Tropsch Catalysts Prepared by Flame Spray Pyrolysis† Melanie Minnermann,O,‡ Suman Pokhrel,O,§ Karsten Thiel,| Robert Henkel,⊥ Johannes Birkenstock,# Torsten Laurus,∇ Ardalan Zargham,∇ Jan-Ingo Flege,∇ Volkmar Zielasek,‡ Edyta Piskorska-Hommel,∇ Jens Falta,∇ Lutz Ma¨dler,*,§ and Marcus Ba¨umer*,‡ Institute of Applied and Physical Chemistry, UniVersity of Bremen, Leobener Strasse NW2, 28359 Bremen, Germany, Foundation Institute of Materials Science (IWT), Department of Production Engineering, UniVersity of Bremen, Badgasteiner Strasse 3, 28359 Bremen, Germany, Fraunhofer Institute for Manufacturing Technology and Applied Materials Research, Wiener Strasse 12, 28359 Bremen, Germany, Institute for Pure and Applied Chemistry, Carl-Von-Ossietzky UniVersity Oldenburg, Ammerla¨nder Heerstrasse 114-118, 26111 Oldenburg, Germany, Central Laboratory for Crystallography and Applied Materials, UniVersity of Bremen, 28359, Bremen, Germany, and Institute of Solid State Physics, UniVersity of Bremen, Otto-Hahn-Allee NW1, 28359 Bremen, Germany ReceiVed: July 23, 2010; ReVised Manuscript ReceiVed: NoVember 26, 2010

Flame spray pyrolysis (FSP) is a novel technique for the fabrication of nanostructured catalysts with farreaching options to control structure and composition even in cases where complex composites need to be prepared. In this study, we took advantage of this technique to synthesize highly dispersed pure and Pddoped iron oxide nanoparticles and investigated them as Fischer-Tropsch (FT) catalysts. By systematically varying the Pd content over a large range from 0.1 to 10 wt %, we were able to directly analyze the influence of the Pd content on activity and selectivity. In addition to catalytic measurements, the structure and composition of the particles were characterized before and after these measurements, using transmission electron microscopy, adsorption measurements, X-ray diffraction, and EXAFS. The comparison revealed on the one hand that small Pd clusters (diameter: 1-2 nm) evolve from initially homogeneously distributed Pd and on the other hand that the iron oxide transforms into iron carbides depending on the Pd content. The presence of Pd influences the particle size in the pristine samples (8-11 nm) resulting in specific surface areas that increase as the Pd content increases. However, after activation and reaction the specific surface areas become similar due to partial agglomeration and sintering. In a fixed bed FT reaction test, enhanced FT activity was observed with increasing Pd content while the selectivity shifts to longer chain hydrocarbons, mainly paraffins. Mechanistic implications regarding the role of Pd for the performance of the catalysts are discussed. 1. Introduction In past few years, the Fischer-Tropsch (FT) reaction, allowing to convert syngas (CO + H2) into hydrocarbons for the production of fuels and chemicals, has received renewed attention. As syngas can be derived from coal, natural gas, and also biomass, it constitutes a flexible alternative to crude oil and thus could play an important role for the satisfaction of future energy demands. In addition, a FT-based fuel production provides advantages in terms of more stringent environmental regulations since cleaner (sulfur-free) fuels become available in this way. These factors in combination with rising crude oil prices and a shortage of natural oil resources thus reinforce †

Part of the “Alfons Baiker Festschrift”. * Corresponding authors. E-mail: [email protected]; lmaedler@ iwt.uni-bremen.de. ‡ Institute of Applied and Physical Chemistry, University of Bremen. § Department of Production Engineering, University of Bremen. | Fraunhofer Institute for Manufacturing Technology and Applied Materials Research. ⊥ Carl-von-Ossietzky University Oldenburg. # Central Laboratory for Crystallography and Applied Materials, University of Bremen. ∇ Institute of Solid State Physics, University of Bremen. O These two authors contributed equally to the work.

efforts of industry and academic research to improve the performance of FT catalysts.1-6 Nowadays, mainly Co and Fe-based catalysts are used for industrial FT synthesis, both of them having distinct advantages and disadvantages regarding their activity and selectivity, price, long-term stability, etc.7 To enhance their performance, promoters are added during the commercial catalyst production. To iron-based catalysts, for instance, structural and chemical promoters, such as nonreducible oxides (e.g., SiO2 or Al2O3) or K are added. Furthermore, Cu is often used to enhance the reducibility of the iron oxide catalysts. While the impact of these promoters on the catalytic behavior has been extensively studied,8-12 less attention has been paid to a replacement of Cu as a reduction promoter by noble metals (Pt, Pd, and Ru), although these metals have a higher hydrogenation activity than Cu13 and have proven to be effective in Co-based FT catalysts.14 In particular for Pd, only a few reports can be found in the literature regarding the effect of Pd promotion on the performance of Fe-based FT catalysts. For example, Pd was added to a Fe catalysts (Fe, SiO2, and K)15 and to an Fe-Mn catalyst,16 respectively. Both studies found changes in the catalysts’ activity and selectivity in the FT reaction upon Pd addition in form of an increased CO conversion rate and a decreased olefin to paraffin ratio, but did not elucidate how Pd interferes in the

10.1021/jp106860d  2011 American Chemical Society Published on Web 12/14/2010

Role of Palladium in Fischer-Tropsch Catalysts mechanism or exerts structural effects. On the contrary, more studies have been published using Pt to promote Fe-based FT systems.13,17-20 Here, catalytic investigations as well as more intensive structural characterizations of the catalysts were performed. In this context, Yu et al.13 observed that Pt facilitates reduction of Fe2O3 and carburization during the reaction, whereas Xu and Bartholomew17 correlated the enhanced catalytic activity with the formation of iron carbides. The commonly used synthesis technique to prepare iron-based FT catalysts consists of a precipitation (or coprecipitation) step to obtain the base material and a subsequent impregnation step to promote the catalyst with K.21 Additional doping with noble metals can also be achieved by impregnation.13,15 The need for a reproducible synthesis of the catalyst renders a precise control of all parameters during these synthesis steps, such as a constant pH value during precipitation or the rate with which the impregnation solution is added,7 necessary. Moreover, additional drying and calcination steps may influence particle size and particle size distribution.22 In view of these problems, the exploration of novel routes for the synthesis of FT catalysts appears rewarding. In this context, flame spray pyrolysis (FSP) as a gas phase synthesis technique seems particularly promising as it is ideally suited for the synthesis of catalysts with high surface areas and reproducible structural morphologies in a fast and automatic one-step approach.23,24 Moreover, the possibility to easily vary the composition of the product allows preparing tailored multicomponent materials. Against this background, the main objective of the present study was to elucidate the potential of flame spray pyrolysis (FSP) for the preparation of doped and undoped Fe-based FT catalysts with defined structure and composition. So far, such an approach was only tested for Co-based catalysts doped with small amounts of Ru as a reduction promoter.25 Pd was chosen as a dopant in this study as it is known for its high H2 solubility which then could lead to spillover processes supporting the iron oxide reduction.15,16,26 As the capabilities of the FSP technique enabled us to systematically vary the Pd doping, we furthermore used the systems to study the trends in the catalytic properties as a function of the Pd content. Therefore, we did not include other typical promoters, such as K, for instance. The effects of promoters and supports, which can be added by means of sophisticated FSP techniques as well,27 will be addressed in a future report. We will show that the nanoparticles prepared by FSP exhibit a high surface area and a homogeneous distribution of all components within the pristine material. Under reaction conditions, very small Pd nanoparticles form and the iron oxide transforms - dependent on the Pd content - into carbides. To the best of our knowledge, this is the first report about FSP used for the production of iron oxide nanoparticles serving as FT catalysts. 2. Experimental Section 2.1. FSP Synthesis of Fe3O4 or Pd/Fe3O4 Nanoparticles. A total of 50 mL of ferrous napthenate (Strem chemicals, 99.9% purity, 0.5 M by metal in xylene) was used for the preparation of pure iron oxide and a mixture (0.5 M) of 50 mL of ferrous napthenate (0.5 M by metal in xylene) with 89 mL of palladium acetylacetone (Strem chemicals, 99.9% purity, 0.01473 M in xylene) was used for the synthesis of material doped with 10% Pd. The other materials with lower contents of Pd in the series (0.1 and 1% Pd) were prepared similarly using less Pd

J. Phys. Chem. C, Vol. 115, No. 4, 2011 1303 acetylacetonate solution (0.89 and 8.9 mL of Pd(acac)2 solution). These precursor solutions are converted to highly dispersed nanoparticles in the flame spray pyrolysis (FSP) process. In the present case, the liquid precursor was delivered at rate of 5 mL/ min using a syringe pump. The precursor is then atomized by a two phase nozzle using 5 L/min O2 at a constant pressure drop of 1.5 bar at the nozzle tip. The spray is ignited by a supporting CH4 and O2 premixed flame (1.5 L/min, 3.2 L/min) forming a self-sustaining spray flame. The particles are formed by reaction, nucleation, surface growth, coagulation and coalescence in the spray flame environment and collected on glass fiber filters after the particles stream has been diluted with cold gas.28,29 2.2. XRD Measurements. For the X-ray diffraction measurements, the as-prepared undoped and Pd-doped nanoparticles were placed in circular sample holders with a diameter of 16 mm and then loaded into a Bruker D8 diffracting system. The diffractometer was configured in Bragg-Brentano geometry, equipped with a primary Johansson monochromator producing monochromatic Mo KR1 (λ ) 0.07093 nm) radiation. A ∼0.1° fixed divergence, 4° primary and 2.5° secondary Soller slits, and LynxEye detector (position sensitive in a range of 3° 2θ with 192 channels, yielding a channel width of 0.01563° 2θ) were used. Continuous scans in the range of 5-55° 2θ were applied with an integration step width of ∼0.03° 2θ and 30 s per step. The XRD patterns of the pure and Pd-doped catalysts after FT synthesis were measured using Cu-K-radiation. These samples were loaded in a PANalytical X’Pert MPD PRO diffracting system, with Cu-KR (λ)0.154 nm) radiation for powder X-ray diffraction. Primary and secondary Soller slits with 0.04 rad aperture, 0.25° fixed divergence, circular sample holders with 16 mm diameter, and X’Celerator detector (position sensitive in a range of 2.122° 2θ with 127 channels) were used. A continuous scan and an integration step width of 0.0334° 2θ were applied. 2.3. Surface Adsorption Measurements. The adsorptiondesorption isotherms of nitrogen (Brunauer-Emmett-Teller (BET) measurements) were measured at 77 K using a Quantachrome NOVA 4000e and Autosorb-1 gas sorption system as a function of relative pressure P/P0 over the range of 0.01-0.99. Prior to the measurement, the sample was outgassed at 200 °C under vacuum to determine the specific surface areas. The pore volume was estimated from the nitrogen uptake at P/P0 ) 0.99. Data were obtained by exposing or removing a known quantity of adsorbing gas in or out of a sample cell containing the solid adsorbent maintained at constant liquid nitrogen temperature. 2.4. EXAFS Measurements. The EXAFS (extended X-ray absorption fine structure) measurements were performed at beamline CEMO at Hasylab, DESY in Hamburg. Iron oxide samples doped with 1% and 10% Pd were investigated at the Pd K edge as well as Fe K edge before and after FT. The monochromator consisted of a pair of Si(311) crystals mounted on a rotation axis driven by an ex-vacuo goniometer covering the energy range between 4.4 and 43.4 keV. The samples were pressed into pellets and fixed onto the sample holder with an incident angle of about 45°. The measurements were performed in fluorescence mode using the 7 pixel Si(Li) detector. The evaluation of the EXAFS data (subtraction of the pre-edge background, normalization to the experimental edge step, subtraction a smooth atomic background from normalized absorption data, and Fourier transformation) was performed using the ARTEMIS program based on the FEFF 6 code.

1304

J. Phys. Chem. C, Vol. 115, No. 4, 2011

2.5. Electron Microscopy Measurements. To characterize the samples’ morphology (shape and nanoparticle size) after synthesis, activation in H2, and FT reaction, transmission electron microscopy (TEM) measurements were carried out using a FEI Tecnai F20 S-TWIN microscope equipped with a GATAN imaging filter (GIF2001). The microscope was operated at an accelerating voltage of 200 kV with a field-emission gun (FEG), resulting in a point resolution of 2.4 Å. The TEM and energy-filtered images were recorded with the slow-scan CCD camera integrated in the GIF (1024 × 1024 pixel array). No binning was used for the TEM images, while for the energy filtered images a binning of 2 × 2 was applied so that images with a size of 512 × 512 pixels were generated. Element maps were recorded using the three-window method; drift between successive images was corrected by a crosscorrelation algorithm (see ref 30 for details). In order to determine the composition of the nanoparticles, energy dispersive X-ray (EDX) microanalysis was performed in the scanningTEM (STEM) mode, achieving a spatial resolution of about 1 nm and an energy resolution of 136 eV. Samples were prepared by suspending the solid in ethanol and subjecting the suspension to ultrasonication for 5 min. Afterward, a drop of the solution was placed on a carbon coated copper grid (200 mesh). Particle size distributions were determined based on the measurement of 200 particles from different areas of the TEM grid. 2.6. Catalytic Measurements. The Fischer-Tropsch synthesis (FTS) measurements of the pure and Pd-doped iron oxide catalysts were carried out in a fixed-bed stainless-steel reactor (inner diameter 9 mm). About 0.1 g of the catalyst materialphysically mixed with MgO (acros organics, p.a.) so that a weight percentage of 10% Fe was obtained-was placed in the reactor using quarz wool to fix the sample. The temperature inside the reactor was monitored and controlled by an axially placed thermocouple being directly in contact with the catalyst. Three-way valves installed in the setup allowed the gas stream to either pass through the reactor or bypass it through a second tube. The latter was installed to measure the initial concentration of the educt stream and to pressurize the system. Prior to the FT reaction, the catalysts were activated in situ at 1 bar in a pure H2 flow of 60 mL/min at 623 K for 16 h (1 K/min heating rate; held for 1 h at 373 K). After reduction, the temperature was lowered to the reaction temperature of 573 K under H2 flow. Before pressurizing the system, the catalyst bed was flushed with Ar (30 mL/min) for 10 min. Then, the reaction pressure was increased up to 5 bar with a mixture of the reaction gases (flow rate: 30 mL/min; H2:CO:Ar volume ratio of 6:3:1, Ar as internal standard). The gas flow rates were controlled by mass flow controllers (Bronkhorst Ma¨tting). After reaction conditions were reached, the feed was switched from the bypass to the reactor defining the starting point of the experiment. The progress of the FT reaction was monitored over a time period of 10 h using an online gas chromatograph (HewlettPackard 5890 II) equipped with a thermal conductivity detector (TCD) and a flame ionization detector (FID). The use of a heatable ten-port valve allowed injecting the product stream periodically and the analysis by both detectors to be done simultaneously. To avoid possible condensations of the reaction products, the gas transfer lines between reactor and GC were constantly heated at 453 K. Analysis of H2, CO and Ar was performed with a molecular sieve column in combination with the TCD, while light hydrocarbons (C1-C9) were separated by using a DB-5 capillary column in combination with the FID. Since Ar was added as

Minnermann et al.

Figure 1. BET isotherm plots of pure and Pd-doped FT catalysts. Inset shows plots of [(P/P0)/V(1 - P/P0)] versus [P/P0] in the range between 0.05 and 0.3 that were used for the determination of the specific surface areas (R2 value: 0.99996).

TABLE 1: Specific Surface Area (SSA) and Particle Sizes Derived from BET and TEM of the Pristine Material sample

specific surface area (m2/g)

particle size (dBET) (nm)

particle size (dTEM) (nm)

undoped 0.1% Pd 1% Pd 10% Pd

95 ((3) 101 ((5) 136 ((3) 189 ((4)

12.4 ((0.2) 11.6 ((0.3) 8.7 ((0.2) 6.2 ((0.2)

12 ((3) 11 ((3) 9 ((3) 8 ((2)

an internal standard, CO conversion was calculated directly by comparing the CO/Ar peak ratio in the stream before reaction with the ratios observed for the gas stream leaving the reactor. Hydrocarbon selectivities of the product stream, as evaluated in the range of C1-C9 and reported below, represent product contents (in mol %). For this purpose, theoretical mass specific response factors have been used, following an incremental approach suggested by Kaiser.31 3. Results and Discussion 3.1. Characterization of the As-Prepared Catalysts. 3.1.1. BET. Figure 1 shows the adsorption/desorption isotherms of the undoped and Pd-doped Fe3O4 nanomaterial. Plots of [(P/ P0)/V(1 - P/P0)] versus [P/P0] in the range between 0.05 and 0.3 result in straight lines (correlation coefficient being higher than 0.999, see inset of Figure 1) from which the specific surface areas can be determined. These values can be used to calculate an average particle size by the equation32 dBET ) 6000/(FpSA) (in nm), where dBET is the average diameter assuming a spherical particle shape, SA represents the measured specific surface area of the powder in m2/g, and Fp is the density in g/cm.3 Specific surface areas of 95, 101, 136, and 189 m2/g for pure iron oxide and material doped with 0.1% Pd, 1% Pd, and 10% Pd, respectively, were derived from the BET data in this way. The corresponding particle sizes were found to decrease from ∼12 nm for the undoped sample to 6 nm for the sample doped with 10% Pd (see Table 1). The values thus reveal that the addition of Pd significantly reduces the particle size of the pristine material. In addition to the surface areas, also the pore size distributions were determined on the basis of the BJH33 (Barrett, Joyner, and Halenda) analysis, yielding pore diameters between 40 and 60 nm for all samples (see Figure 2). These results show that the individual particles are loosely packed, i.e., the extent of agglomeration is low. Accordingly, high porosities of 98% can be achieved.34 3.1.2. XRD. Representative XRD patterns of as-prepared iron oxides samples, undoped and doped with 0.1-10% Pd, are

Role of Palladium in Fischer-Tropsch Catalysts

Figure 2. Cummulative pore volume and pore size distribution for (a) undoped, (b) 1% Pd doped, and (c) 10% Pd doped catalysts before FT as determined on the basis of the BJH (Barrett, Joyner, and Halenda) analysis. The similarity of the plots for all catalysts suggests the same particle morphology and porous nature of the powder.

Figure 3. XRD patterns of the as-prepared iron oxide nanoparticles: (a) undoped, (b) 0.1% Pd, (c) 1% Pd, and (d) 10% Pd. The broad and enhanced background of the XRD patterns is due to diffuse scattering probably arising from significant non-Bragg scattering.

presented in Figure 3. The data reveal that the samples consist of a mixture of magnetite (Fe3O4) and hematite (Fe2O3). (The peak assignments were performed using PDF card no. 00-1190629 for magnetite and 00-033-0664 for hematite.) Magnetite crystallizes as a cubic phase (a ) b ) c ) 8.3960 Å, R ) β ) γ ) 90°, space group Fd3m (227), V ) 591.86 Å3, Z ) 8), hematite in a trigonal phase (a ) b ) 5.0356, c ) 13.7489 Å, R ) β ) 90°, γ ) 120°, space group R3jc (167), V ) 301.93 Å3, Z ) 6). The absence of peaks at 15.399 (100%) and 24.430 (20%) 2θ [Mo KR1] in all the powder patterns suggests the absence of PdO particles (P42/mmc, tetragonal, star entry 41-1107). (Metallic Pd is not expected under the chosen FSP conditions.) This observation excludes the formation of large PdO crystallites and suggests either a homogeneous distribution of Pd in the Fe3O4 matrix or, less likely, the presence of

J. Phys. Chem. C, Vol. 115, No. 4, 2011 1305 amorphous PdO clusters. (We will come back to this point in the next section.) Note that the broad and enhanced backgrounds of the XRD patterns are due to diffuse scattering, probably arising from significant non-Bragg scattering.35 Part of this diffuse scattering may be due to an amorphous fraction, part of it may also result from cystalline but very small (nm-sized) particles in the sample. With lattice parameters of 8.39 Å, crystallite sizes of 10 nm and less, corresponding to just 10-12 unit cells in a row, can give rise to rather broad X-ray diffraction peaks and diffuse scattering. In order to clarify this question, TEM was carried out. 3.1.3. TEM/STEM. The TEM images in Figure 4 show that the pristine FSP prepared material consists of slightly agglomerated iron oxide nanoparticles having roughly spherical shape. While the particle morphology essentially does not change due to Pd doping (compare 4 a and b), the particle sizes decrease from 12 ((3) nm for the undoped sample to 8 ((2) nm for the sample doped with 10% Pd, in very good agreement with the BET data (see Table 1). Concomitantly, the particle size distribution becomes narrower with increasing Pd amount (data not shown). In agreement with the XRD data, parts of the particles were found to be crystalline as revealed by the presence of lattice fringes, but an amorphous fraction could not be excluded. Based on the STEM-EDX elemental analysis carried out for the sample doped with 10% Pd, almost identical Fe:Pd ratios, corresponding to weight ratios of about 10:1, were obtained for different areas studied by EDX. As shown in Figure 4, even at brighter spots no Pd enrichment was detected by EDX. In agreement with that, the TEM images provide no indications of PdO particles (compare with TEM images in Figure 9 acquired with the same magnification after activation and reaction where these particles are visible). Accordingly, it can be assumed that Pd is homogeneously distributed in the iron oxide matrix rather than forming individual particles. It is important to note that the nominal compositions as chosen for the synthesis were reproduced in the EDX measurements, demonstrating that it is indeed possible to precisely adjust the composition of mixed systems using the FSP technique. 3.2. Catalytic Measurements. The performance of the four catalysts in FT synthesis was studied at 573 K, 5 bar and a total gas flow of 30 mL/min (H2/CO ) 2). For each measurement, about 0.1 g of the iron oxide catalyst (77 mg pure Fe each) was used. The CO conversions and selectivity patterns after 10 h time on stream are presented in Table 2 of all samples. (Note that steady-state conditions were not yet reached after 10 h time on stream. Thus, our investigation focuses on the initial behavior of the catalysts.) All FSP-synthesized catalysts were active for the FT reaction after activation in H2. Not unexpectedly, the activities are inferior to technical catalysts (CO conversions in the range of 5-10%), as no promoters or supports (except for the physical mixture with MgO), were employed. (While promotion with K is reported to increase the activity up to 50%,22 the use of supports is known to stabilize the catalyst surface preventing the particles from sintering.8) Our approach, however, allows us to directly investigate the influence of Pd. All doped samples show higher activities than the pure iron oxide sample. While the activities for the 0.1% Pd and 1% Pd sample are essentially the same, increasing the Pd content to 10% leads to a further activity increase. For this catalyst the FT activity is finally 1.7 times higher than for the undoped sample (after 10 h time on stream).

1306

J. Phys. Chem. C, Vol. 115, No. 4, 2011

Minnermann et al.

Figure 4. TEM (a-d) and STEM (e) images of the as-prepared undoped (a) and doped (10% Pd) iron oxide nanoparticles (b-e). EDX spectra (1, 2) taken at selected spots for the 10% Pd sample indicate a homogeneous distribution of Pd in the sample since no Pd enrichment has been observed even at brighter spots (wt % Fe:Pd ≈ 10:1). In agreement, no Pd or PdO particles are visible in (c, d) (in contrast to Figure 9).

TABLE 2: Catalytic Performance of Pure and Differently Doped Catalysts Prepared by FSP undoped

0.1% Pd

4.8 ((1)

CO conversion (%) 6.8 ((2) 6.5 ((2)

C1 C2 C4-C9

76 16 8

C2 C4 C5 C6 % olefins in linear HC (C6)

0.29 1.27 0.50 0.37 61

1% Pd

hydrocarbon (HC) selectivity 67 67 19 19 14 14 olefin/paraffin ratio 0.14 0.16 0.80 0.83 0.35 0.37 0.29 0.30 58 58

10% Pd 8.1 ((1) (%) 66 17 17 0.02 0.18 0.04 0.07 33

Interestingly, not only the final activity is influenced by noble metal doping, but also the evolution of the catalytic activity as a function of time. After 20 min, the CO conversion is higher for the undoped sample than for the catalyst containing 10% Pd (Figure 5). However, while being on stream, the activity decreases for the undoped sampe slightly and levels at a conversion of about 5%, while CO conversion of the Pd doped catalyst increases continuously as a function of time (Figure 5). In addition to the activity, also the selectivity is changed upon Pd doping. With increasing Pd content, the selectivity is shifted to longer chain hydrocarbons (C4-C9). Specifically, the C1 fraction is distinctly higher for the pure iron sample. Nevertheless, the main component for all four catalysts formed during the FT reaction is methane. Another factor affected by Pd is the olefin to paraffin ratio (cf. Table 2). While the pure iron catalyst shows the highest olefin/paraffin ratio, the value slightly decreases for the 0.1 and 1% Pd catalysts. Doping with a very high amount of Pd (10%) finally leads to a strong decrease in olefin formation due to strong hydrogenation activity of Pd, as exemplarily shown for the C6 products.

Figure 5. Evolution of the CO conversion in % (error bar: ∼ ( 1%) with time on stream shown for the undoped sample and the sample doped with 10% Pd. Reaction conditions: 573 K, 5 bar, H2:CO:Ar ) 6:3:1, 30 mL/min.

3.3. Characterization of the Catalyst Samples after Reaction. After 10 h time on stream, the structural and compositional changes of the samples that occurred under FT reaction conditions were investigated by BET and ex-situ XRD, TEM/ STEM and EXAFS measurements. In order to study the changes already induced by the activation of the catalysts, some electron microscopy measurements were carried out also after H2 reduction. (Note that, as the samples were physically mixed with MgO for the FT synthesis, the XRD patterns of the samples after the catalytic experiments exhibit strong MgO peaks.) 3.3.1. XRD Results. From the changes in the XRD pattern presented in Figure 6, it can be concluded that the undoped sample contains mainly Fe5C2, a iron carbide phase called Ha¨gg carbide. Apparently, the oxidic base material was reduced first to metallic iron and further on transformed to the carbide phase under reaction conditions. For the Pd-doped samples the situation is slightly different. Here a carbon-richer hexagonal Fe2C phase is observed in addition to Fe5C2.

Role of Palladium in Fischer-Tropsch Catalysts

J. Phys. Chem. C, Vol. 115, No. 4, 2011 1307

Figure 6. Rietveld refinements of the undoped and Pd-doped iron catalysts after FT synthesis. For the catalytic experiments the catalysts were physically mixed with MgO leading to the presence of MgO reflexes. (a) Undoped, (b) 0.1% Pd, (c) 1% Pd, and (d) 10% Pd. The XRD patterns were refined using structural models of MgO (ICSD77821), Fe (ICSD-64998), Fe3O4 (ICSD-82237), Fe2C (ICSD-76826), and Fe5C2 (ICSD-245334). Since the different iron carbide peaks are broad and appear very close to one another, they are presented in a single label (iron carbide). Apart from the fits, the original data and the residual values are also shown.

TABLE 3: Phase Composition of Undoped and Pd-Doped Fe Catalysts after FT Synthesis As Obtained by Analysis of the X-ray Diffraction Patterns sample

Fe5C2 (%)

Fe2C (%)

Fe (%)

Fe3O4 (%)

undoped 0.1% Pd 1.0% Pd 10% Pd

65 ((2) 64 ((2) 57 ((5) 79 ((8)

0 36 ((2) 43 ((4) 21 ((3)

30 ((2) 0.05 ((0.0) 0.04 ((0.0) 0.04 ((0.0)

5 ((3) 0 0 0

For a quantitative analysis of the phase composition after FT synthesis, Rietveld analysis of the diffraction patterns (see Figure 6) was performed. Using the BRASS program,36 full profile fitting (Rietveld refinement) was employed, yielding lattice constants, composition and microstructural parameters. (For the refinements, the crystal structure models of MgO, Fe, Fe3O4, Fe2C, and Fe5C2 corresponding to the entries No. 77821, 64998, 82237, 76826, and 245334, respectively, from the inorganic crystal structure database ICSD were employed.) The Rietveld refinements based on structural models of metallic Fe and the iron carbides yielded reasonable agreement with the experimental data. (The goodness-of-fit and Rwp values for the undoped catalyst and the catalysts doped with 0.1% Pd, 1% Pd, and 10% Pd after FT was found to be 3.08, 7.38, 4.06, and 3.63 and 11.33, 3.42, 8.46, and 8.15, respectively). The data, presented in Table 3, reveal that (a) for the undoped sample some metallic Fe is present in addtion to the Ha¨gg carbide and (b) for the Pd containing catalysts, only carbides were formed with the Ha¨gg carbide tending to increase with the Pd content at the expense of the additional Fe2C phase. So, Pd apparently has a strong influence on the carburization of the catalysts and increases the carbon uptake. A possible explanation for this effect is a more efficient reduction of the iron oxide to metallic Fe in the presence of Pd. Furthermore, the smaller particle sizes

Figure 7. TEM images of the four different catalysts after FT reaction showing the formation of core-shell particles with different sizes depending on the Pd content.

in these samples identified by the TEM provide a larger active surface area of metallic Fe possibly resulting in a higher and more rapid carbon uptake by CO dissociation.19,37 The removal of oxygen formed during the dissociation as well may also be accelerated by atomic hydrogen supplied by Pd. 3.3.2. Microscopy InWestigations after Reaction. In addition to XRD, the samples were investigated by TEM and STEM. For all samples, two types of particles are visible in the images compiled in Figure 7. The first type resembles the original morphology of the pristine samples. In addition to these small, partially agglomerated particles, however, larger core-shell like particles with different diameters in the range of 10 to 100 nm are present. As revealed by TEM after the activation of the catalysts, these structures are formed already during H2 activation. Notably, the core-shell particles are on average larger for the undoped catalyst (56 ( 20 nm) than for the Pd containing samples for which the average particle size is nearly independent of the Pd content. Yet, the standard deviation of the particle size distribution decreases slightly as the Pd doping increases: 0.1% Pd (25 ( 9 nm), 1% Pd (26 ( 8 nm), and 10% Pd (29 ( 5 nm). A quantitative evaluation of the TEM images furthermore revealed that the number density of core-shell particles is higher for the doped samples as compared to the pure sample. Since XRD measurements in combination with Rietveld analysis can only provide information about the average compositions of the crystalline areas of the samples, EF-TEM mappings were employed to identify the chemical composition of the structural features observed by TEM. Specifically, elemental mappings of iron, oxygen and carbon were recorded for the pure and 10% Pd containing sample, the latter being the most active in FT catalysis (Figure 8). In the case of the large core-shell particles, the mappings show that the core always contains iron. For the Pd-doped sample also larger amounts of carbon are detected as concluded from the bright spots in the C mapping for the 10% Pd sample after FT reaction. This is in a good agreement with the results obtained from Rietveld analysis, indicating the preferential formation of iron carbides for the Pd containing samples. In

1308

J. Phys. Chem. C, Vol. 115, No. 4, 2011

Minnermann et al. TABLE 4: Specific Surface Area of the Different Catalysts As Derived from BET Measurements after Activation and after FT Reactiona

sample undoped

BET surface area after activation (m2/g)

BET surface area after FT reaction (m2/g)

[change in % compared to the as-prepared material]

[change in % compared to the as-prepared material]

76 ((6) [20((9)]

92 ((7) [3((9)] 64 ((7) [37((7)] 83 ((7) [39((5)] 78 ((7) [59((4)]

0.1% Pd 1% Pd Figure 8. EF-TEM mappings for the pure and 10% Pd-doped iron catalyst after FT. Despite the different magnifications (undoped: 41 000×, 10%Pd: 78 000×) it is obvious that the 10% Pd-doped sample shows a higher amount of carbon in the core of the particles.

Figure 9. TEM (a and b) and STEM (c) images of the catalyst doped with 10% Pd sample after FT reaction. Pd clusters (1-2 nm, white arrows) are located all over the surface of the catalyst particles as indicated by the bright spot (1) in the STEM image (c) and identified by EDX.

case of the pure Fe catalyst, only small amounts of carbon could be identified in the core of the particles rendering metallic iron probable as a main constituent (in agreement with XRD). The shell of these particles consists of iron and oxygen, the latter being especially visible in form of the bright contrast in the oxygen mapping. It is most likely that this shell has been formed during exposure to air (after removal from the reactor and the transport to the TEM) leading to the oxidation of the metallic or carbidic iron, respectively.3,8,38 The morphology of the smaller particles visible in Figures 7 and 9 (the black arrows in Figure 9, panels a and b, point to some of these particles) changes to a somewhat more dense structure as compared to the pristine nanoparticles, an effect that could result from agglomeration and sintering during activation and reaction. The EF-TEM mappings shown in Figure 8 illustrate that many of the small particles contain iron and oxygen. This result is surprising at first sight, as XRD revealed no oxide component after FT synthesis but only carbide as a main constituent. Apparently, these small particles must be amorphous so that they cannot be detected by XRD. In

10% Pd

87 ((6) [54((9)]

a The error bars essentially arise from the difficulties to separate the MgO (mixed to the catalyst for the FT reaction) from the catalyst after use. The residual MgO content for each sample was determined by AAS.

agreement with that, higher magnification TEM images (Figure 9b) prove that these small particles are indeed not crystalline. The oxidation of iron carbides (as well as metallic iron) during the FT reaction forming Fe3O4 as a deactivating species is wellknown in the literature and reviewed in ref 3. According to these studies, however, this process leads to crystalline magnetite particles.39 Since smaller particles have been observed to carburize faster in syngas,40,41 it is thus more likely that the observed small particles consisted originally of iron carbide and were subsequently oxidized after exposing them to air. In summary, the following picture evolves: although in the case of the Pd-doped samples the whole catalyst transforms into carbides forming large and small particles, in the case of the pure Fe catalyst carbidic and metallic Fe exists, the former phase mainly being present in small particles and the latter in large particles. In both cases, metallic Fe is expected to form during the reductive activation which apparently sinters partly explaining the large particles. Under FT reaction conditions carburization then occurs. While all small particles can easily form carbides, only for the doped catalysts Pd can induce carbide formation rapidly also for the large particles while Fe still remains metallic in case of the undoped system. However, the question is still open in which form the Pd is present and comes into action. From XRD results, the fate of the Pd is still unclear. Yet, by TEM, very small (1-2 nm) crystallites could be found all over the 10% Pd sample (Figure 9, white arrows point to some of these Pd particles), which are not visible in the images of the as-prepared samples (cf. Figure 4). STEM together with EDX elemental analysis revealed that these crystallites have a high content of Pd. Thus, it can be concluded that small Pd particles form during the activation, which in turn can act as hydrogenation catalyst and supplier of atomic hydrogen. Due to their very small size, their low amount and their potentially low degree of crystalline order, it is expected that these particles remain invisible to XRD. 3.3.3. BET Measurements after Reaction and ActiWation. BET measurements after activation and reaction were performed to investigate the changes in surface area. Compared to the surface areas of the as-prepared materials (listed in Table 1), a decrease after activation is observed (see Table 4). This decrease is particularly pronounced for the sample doped with10% Pd showing a 54% change (189 ( 4 to 87 ( 6 m2/g), whereas the surface area of the pure iron sample only decreases by 20% (95 ( 5 to 76 ( 6 m2/g).

Role of Palladium in Fischer-Tropsch Catalysts

Figure 10. The moduli of the Fourier transforms of the EXAFS data recorded at the Fe and Pd K edges for the samples doped with 1 and 10% of Pd.

These findings can be correlated with the TEM results after activation. In the case of the 10% Pd sample the partial transformation of the originally small particles to larger ones (30 ( 5 nm) provides a reasonable explanation for the decrease of the specific surface area. In addition a higher degree of agglomeration of the smaller particles may also contribute to this effect. While the formation of larger particles was also observed for the undoped sample, the distinctly smaller number of such particles is in agreement with the moderate decrease of the BET surface area. Once reduced in H2, the BET surface area of the 10% Pd doped sample is nearly unaffected by the FT reaction, indicating only a negligible further decrease in specific surface area. In contrast, the BET surface area of the undoped sample increases slightly after FT reaction as compared to the value after activation. This finding might suggest the formation of small Fe3O4 particles as a product of the initial deactivation process (Figure 5).42 3.3.4. EXAFS Results. In order to verify the picture regarding the impact of Pd doping on the particle structure and chemical state, EXAFS measurements of 1% and 10% Pd-doped iron oxide catalysts before and after FT were performed at the Fe K edge and Pd K edge, respectively. Figures 9a-d presents the moduli of the Fourier transform of the EXAFS oscillation for studied samples, clearly confirming the strong structural change during the activation and FT reaction. Specifically, a second main peak at larger interatomic distances occurs in the Fourier transform of the Fe EXAFS signal for the samples after FT reaction pointing to a more metallic surrounding of the Fe atoms as expected for a partial reduction and carbide formation of the iron oxide particles during activation and the FT reaction.43 Since the peak at the original position is still present, the EXAFS data also support the assumption that the samples after FT and exposure to air still contain (noncrystalline) iron oxide which was not detectable by XRD but visible in the EF-TEM mappings. For the Pd EXAFS data a similar trend is found (Figure 10). The peaks shift toward significantly larger interatomic distances which is consistent with a change of the Pd oxidation state from palladium oxide (peak position at ∼1.5 Å) to metallic palladium (peak at ∼2.5 Å). The main difference found when comparing

J. Phys. Chem. C, Vol. 115, No. 4, 2011 1309 the 1% and the 10% Pd containing nanoparticles is the more pronounced evolution of a peak at about 2.5 Å in the 10% Pd spectra after the FT reaction (Figure. 9d), confirming the formation of metallic clusters.44 The EXAFS findings are thus in a good agreement with the EF-TEM studies. 3.4. Mechanistic Implications. Apart from the insight regarding the structural transformation and activation of the Febased FSP nanoparticles for FT synthesis, our results demonstrate the beneficial role of Pd for such FT catalysts. Pd leads to a higher degree of carburization. In literature the role of carbides in general and especially their different phases as active FT catalysts is still under debate.41,45-47 However, the formation of iron carbides is nowadays considered to be essential for obtaining high FT synthesis activities and high CO conversion rates.2,17,42,45 Therefore, an enhanced carbide formation as observed for the Pd doped samples should be beneficial for the activity of iron-based FT catalysts, as indeed shown by our catalytic results. The higher carbide content can in turn be explained by a better reducibility of catalyst. Small Pd particles (1-2 nm) formed during activation process facilitate the reduction of iron oxide. The resulting metallic Fe particles can then easily form carbides by dissociation of CO. The beneficial role of Pd in terms of a higher reducibility of the material becomes obvious also in the development of the initial activities. Whereas the undoped system shows deactivation in the first 600 min, a continuous increase of the activity is found in the case of the Pd containing systems probably due to an efficient removal of atomic oxygen by atomic hydrogen formed on the Pd particles.48 So while Pd enhances the activity due to a number of factors directly or indirectly related to its role as a reduction catalyst, it also influences the selectivity pattern. The analysis of the product spectrum shows that the addition of Pd results in a decrease of olefin formation once again in agreement with the hydrogenation ability of Pd. 4. Conclusion The promoting effect of Pd on the structural properties as well as the catalytic behavior of iron-based catalysts in the FT reaction was studied in this work. For this purpose, flame spray pyrolysis (FSP) was used to produce nanosized crystalline Fe3O4 particles in the range of 8-12 nm in a fast one-step approach. The BET surface area of the obtained particles increases with increasing Pd amount due to a decrease in particle size. Structural analysis of the as-prepared samples indicates that Pd is well dispersed in the iron oxide matrix as Pd aggregates could not be observed. Therefore, FSP-synthesized particles can serve as a well-defined system to investigate the effect of Pd promotion on iron-based FT catalysts. In a fixed-bed FT reaction test, the pristine iron oxide nanoparticles showed FT activity. Doping this material with Pd enhances the activity of the iron catalyst while the selectivity is shifted to longer chain hydrocarbons and the formation of methane is decreased. Moreover, Pd decreases the selectivity of olefins in the product due to its catalytic hydrogenation ability. By TEM and XRD after FT reaction, we could prove that Pd affects the reducibility as well as the carburization of the nanoparticles. While the enhanced reducibility after H2 activation was only indirectly revealed by BET measurements after activation and also by a higher number of larger, sintered particles, TEM, Rietveld and EF-TEM analysis clearly indicate the improved carburization of metallic iron with increasing Pd amount. Pd supports the complete transformation of metallic iron to iron carbides (Fe2C5 and Fe2C), which are known as

1310

J. Phys. Chem. C, Vol. 115, No. 4, 2011

active species in FT. In contrast, the pure iron catalyst still contains metallic iron after FT reaction. In essence, FSP was proven to be a very suitable technique to produce catalysts with variable compositions, which can be very precisely controlled. The addition of other essential promoters such as alkali or structural promoters is necessary to reach sufficient activity levels for industrial FT synthesis. For this purpose, FSP is again a convenient technique, allowing a one-step synthesis of all the components which will be part of future investigations. Acknowledgment. We gratefully acknowledge financial support of the German Science Foundation (DFG) and the EWE AG. We thank Willian G. Menezes for some of the TEM measurements. Furthermore, we are grateful to Prof. Dr. F. Ro¨ssner (University of Oldenburg) for fruitful discussion and advice in setting up the catalytic experiments. References and Notes (1) Schulz, H. Appl. Catal. A-Gen. 1999, 186 (1-2), 3–12. (2) Herranz, T.; Rojas, S.; Pe´rez-Alonso, F. J.; Ojeda, M.; Terreros, P.; Fierro, J. L. G. J. Catal. 2006, 243 (1), 199–211. (3) de Smit, E.; Weckhuysen, B. M. Chem. Soc. ReV. 2008, 37 (12), 2758–2781. (4) Maitlis, P. M.; Zanottib, V. Chem. Commun. 2009, 13, 1619–1634. (5) Dry, M. E. Appl. Catal. A-Gen. 2004, 276 (1-2), 1–3. (6) Cheng, J.; Hu, P.; Ellis, P.; French, S.; Kelly, G.; Lok, C. M. J. Phys. Chem. C 2010, 114 (2), 1085–1093. (7) Khodakov, A. Y.; Chu, W.; Fongarland, P. Chem. ReV. 2007, 107 (5), 1692–1744. (8) Jin, Y.; Datye, A. K. J. Catal. 2000, 196 (1), 8–17. (9) Dry, M. E. , FT catalysts. In Fischer-Tropsch Technology; Elsevier Science: Amsterdam, 2004; Vol. 152, pp 533-600. (10) Gaube, J.; Klein, H. F. Appl. Catal., A 2008, 350 (1), 126–132. (11) Anderson, R. B.; Seligman, B.; Shultz, J. F.; Kelly, R.; Elliott, M. A. Ind. Eng. Chem. 1952, 44 (2), 391–397. (12) de Smit, E.; de Groot, F. M. F.; Blume, R.; Havecker, M.; KnopGericke, A.; Weckhuysen, B. M. Phys. Chem. Chem. Phys. 2010, 12 (3), 667–680. (13) Yu, W. Q.; Wu, B. S.; Xu, J.; Tao, Z. C.; Xiang, H. W.; Li, Y. W. Catal. Lett. 2008, 125 (1-2), 116–122. (14) Diehl, F.; Khodakov, A. Y. Oil Gas Sci. Technol. 2009, 64 (1), 11–24. (15) Luo, M. S.; O’Brien, R.; Davis, B. H. Catal. Lett. 2004, 98 (1), 17–22. (16) Richter, B.; Baerns, M. Chem.-Z. 1985, 109 (12), 395–399. (17) Xu, J.; Bartholomew, C. H.; Sudweeks, J.; Eggett, D. L. Top. Catal. 2003, 26 (1-4), 55–71. (18) Vannice, M. A.; Garten, R. L. J. Mol. Catal. 1976, 1 (3), 201– 222. (19) Xu, J.; Bartholomew, C. R. J. Phys. Chem. B 2005, 109 (6), 2392– 2403.

Minnermann et al. (20) Guczi, L. Catal. ReV. Sci. Eng. 1981, 23 (3), 329–376. (21) Bukur, D. B.; Lang, X. S.; Rossin, J. A.; Zimmerman, W. H.; Rosynek, M. P.; Yeh, E. B.; Li, C. P. Ind. Eng. Chem. Res. 1989, 28 (8), 1130–1140. (22) Mabaso, E. I. Nanosized Iron Crystallites for Fischer-Tropsch Synthesis; University of Cape Town: Cape Town, South Africa, 2005. (23) Strobel, R.; Baiker, A.; Pratsinis, S. E. AdV. Powder Technol. 2006, 17 (5), 457–480. (24) Strobel, R.; Pratsinis, S. E. J. Mater. Chem. 2007, 17 (45), 4743– 4756. (25) Teoh, W. Y.; Setiawan, R.; Ma¨dler, L.; Grunwaldt, J. D.; Amal, R.; Pratsinis, S. E. Chem. Mater. 2008, 20 (12), 4069–4079. (26) Nolan, P. E.; Lynch, D. C.; Cutler, A. H. J. Phys. Chem. B 1998, 102 (21), 4165–4175. (27) Teoh, W. Y.; Amal, R.; Ma¨dler, L. Nanoscale (in press). (28) Teoh, W. Y.; Amal, R.; Ma¨dler, L.; Pratsinis, S. E. Catal. Today 2007, 120 (2), 203–213. (29) Ma¨dler, L.; Kammler, H. K.; Mueller, R.; Pratsinis, S. E. J. Aerosol. Sci. 2002, 33 (2), 369–389. (30) Hofer, F.; Grogger, W.; Kothleitner, G.; Warbichler, P. Ultramicroscopy 1997, 67, 83–103. (31) Kaiser, R. Chromatographie in der Gasphase; Bibliographisches Institut: Mannheim, 1969. (32) Teoh, W. Y.; Ma¨dler, L.; Beydoun, D.; Pratsinis, S. E.; Amal, R. Chem. Eng. Sci. 2005, 60 (21), 5852–5861. (33) Barrett, E. P.; Joyner, L. G.; Halenda, P. P. J. Am. Chem. Soc. 1951, 73 (1), 373–380. (34) Ma¨dler, L.; Roessler, A.; Pratsinis, S. E.; Sahm, T.; Gurlo, A.; Barsan, N.; Weimar, U. Sens. Actuator B-Chem. 2006, 114 (1), 283–295. (35) Pokhrel, S.; Birkenstock, J.; Schowalter, M.; Rosenauer, A.; Ma¨dler, L. Cryst. Growth Des. 2010, 10 (2), 632–639. (36) Birkenstock, J.; Fischer, R. X.; Messner, T. BRASS-The Bremen RietVeld Analysis and Structure Suit; 2009. (37) Pour, A. N.; Housaindokht, M. R.; Tayyari, S. F.; Zarkesh, J. J. Nat. Gas Chem. 2010, 19 (3), 284–292. (38) Janbroers, S.; Louwen, J. N.; Zandbergen, H. W.; Kooyman, P. J. J. Catal. 2009, 268 (2), 235–242. (39) Ning, W.; Koizumi, N.; Chang, H.; Mochizuki, T.; Itoh, T.; Yamada, M. Appl. Catal., A 2006, 312, 35–44. (40) Heon, J.; Thomson, W. J. J. Catal. 1992, 134 (2), 654–667. (41) Raupp, G. B.; Delgass, W. N. J. Catal. 1979, 58 (3), 348–360. (42) Shroff, M. D.; Kalakkad, D. S.; Coulter, K. E.; Kohler, S. D.; Harrington, M. S.; Jackson, N. B.; Sault, A. G.; Datye, A. K. J. Catal. 1995, 156 (2), 185–207. (43) Ribeiro, M. C.; Jacobs, G.; Davis, B. H.; Cronauer, D. C.; Kropf, A. J.; MarshaW, C. L. J. Phys. Chem. C 2010, 114 (17), 7895–7903. (44) Okumura, K.; Kato, K.; Sanada, T.; Niwa, M. J. Phys. Chem. C 2007, 111 (39), 14426–14432. (45) Riedel, T.; Schulz, H.; Schaub, G.; Jun, K. W.; Hwang, J. S.; Lee, K. W. Top. Catal. 2003, 26 (1-4), 41–54. (46) Niemantsverdriet, J. W.; Vanderkraan, A. M.; Vandijk, W. L.; Vanderbaan, H. S. J. Phys. Chem. 1980, 84 (25), 3363–3370. (47) Eliason, S. A.; Bartholomew, C. H. Appl. Catal. A-Gen. 1999, 186 (1-2), 229–243. (48) Li, S.; O’Brien, R. J.; Meitzner, G. D.; Hamdeh, H.; Davis, B. H.; Iglesia, E. Appl. Catal. A-Gen. 2001, 219 (1-2), 215–222.

JP106860D