Characterization of slurry phase iron catalysts for Fischer–Tropsch synthesis

Applied Catalysis A: General 186 (1999) 277–296

Characterization of slurry phase iron catalysts for Fischer–Tropsch synthesis Linda D. Mansker a , Yaming Jin a , Dragomir B. Bukur b , Abhaya K. Datye a,∗ a

Center for Micro-Engineered Materials, Chemical and Nuclear Engineering Department, University of New Mexico, Albuquerque, NM 87131-1341, USA b Chemical Engineering Department, MS 3122; Texas A&M University, College Station, TX 77843-3122, USA

Abstract The study of Iron Fischer–Tropsch catalysts by conventional powder X-ray diffraction (XRD) is complicated by the number and type of phases present (␣-Fe, various iron carbides, Fex C: 2 < x ≤ 3, magnetite, Fe3 O4 ). Peak overlap in the diffraction patterns and differences in the X-ray scattering ability of each phase can make quantitation difficult. This has led to the consensus that activity for Fischer–Tropsch (F–T) synthesis does not correlate with the bulk composition of the iron catalyst, as seen by X-ray diffraction. As we demonstrate in this paper, some of the problems with sample analysis arise from the difficulty in preserving microstructures and composition intact, as the sample is removed from the reactor and prepared for analysis. Our results indicate that Soxhlet extraction, a commonly used procedure to remove the wax from a catalyst, can cause changes in catalyst phase composition. We present a study of a slurry-phase iron F–T synthesis catalyst, where samples have been removed under inert atmosphere and care was taken to preserve the catalyst constituents intact. Quantitative Rietveld structural refinement, combined with step-scanned X-ray diffraction data, allows us to determine changes in composition and morphology in the working catalyst. We conclude that, in its most active form, this Fe catalyst contains the Fe7 C3 carbide with small amounts of alpha-iron (␣-Fe), while the ␹-carbide (hereafter designated as Fe5 C2 ), also present in differing amounts during each run, appears to be less active. Major changes in carbide particle size also occur during the course of the F–T synthesis run. The active catalyst contains a significant amount of highly dispersed carbide particles. ©1999 Elsevier Science B.V. All rights reserved. Keywords: Fischer–Tropsch synthesis; Promoted iron catalyst; Iron carbide; Catalyst characterization; Quantitative X-ray diffraction; Rietveld refinement

1. Introduction The Fischer–Tropsch (F–T) synthesis, which produces high-molecular weight hydrocarbon wax from syngas, is considered an effective solution to the problem of finding suitable substitutes for liquid fossil fu∗ Corresponding author. Tel.: +1-505-277-0477; fax: +1-505-277-1024 E-mail address: [email protected] (A.K. Datye)

els [1]. The F–T synthesis reaction converts syngas (H2 /CO, derived from natural gas or coal) to liquid hydrocarbon feed stocks, which can be used for further processing to chemicals, catalytically cracked to various API weight fuels, etc. The choice of catalyst depends upon the syngas ratio (H2 /CO). The H2 /CO ratio is dependent, in turn, upon the syngas precursor and the processing method. A reasonable H2 /CO feed ratio ranges from 1.8 to 2.5, which is typical of natural gas-derived syngas. Coal-derived syngas typ-

0926-860X/99/$ – see front matter ©1999 Elsevier Science B.V. All rights reserved. PII: S 0 9 2 6 - 8 6 0 X ( 9 9 ) 0 0 1 4 9 - 0

278

L.D. Mansker et al. / Applied Catalysis A: General 186 (1999) 277–296

ically produces an H2 /CO ratio of 0.5–1.7; therefore, the catalyst used must possess adequate water gas shift (WGS) activity (in which F–T synthesis product water and CO from the feed are converted to H2 and CO2 ) to make up the deficit in H2 for the F–T synthesis. Iron F–T synthesis catalysts effectively promote the water gas shift (WGS) reaction, making them most attractive for use with coal-derived syngas. Iron F–T synthesis catalysts are prepared by precipitation of a water-soluble iron species (such as Fe(NO3 )3 ), with or without promoters (such as KNO3 and Cu(NO3 )2 ) and binders (such as silica). The precipitated powder is then calcined to form hematite (Fe2 O3 ) and activated by pretreatment with H2 , CO or syngas, outside the reactor or in-situ. Activation alters the catalyst composition to what is thought to be a mixture of iron oxides (Fe2 O3 , Fe3 O4 ), various iron carbides (Fex C, 2 ≤ x ≤ 3), and iron metal (␣-Fe). There is no agreement over the nature of the working catalyst and the active phase(s) responsible for F–T synthesis. In the literature, we find studies suggesting that magnetite (Fe3 O4 ) may be the active phase [2–3], while other workers have linked the formation of magnetite to catalyst deactivation [4]. Huang et al. [5] reported that their “XRD patterns indicate that Fe3 O4 is the only crystalline phase present for the catalyst at the point of maximum activity, and this does not depend upon the activation gases”. Based on this observation, they imply that either “Fe3 O4 is the active phase, or that the active phase cannot be detected by XRD”. These studies illustrate some of the difficulties in the study of Fe F–T catalysts by X-ray diffraction. The starting iron oxide or the reduced iron (␣-Fe) is known to transform into iron carbides during reaction. Hence, there are numerous studies that propose iron carbide to be the active phase for F–T synthesis [6–12]. However, as shown in this paper, as well as in a companion paper in this special issue [13], a catalyst activated in CO exhibited very low activity initially, despite having been almost completely transformed to the monoclinic Fe5 C2 phase. Yet a third possibility, that ␣-Fe is the active phase, has also been proposed in the literature [14]. It is clear that previous attempts to relate activity to catalyst phase composition have not been successful, and are best summarized in the review by M.E. Dry who states that there is “no reason to relate the amount and nature of iron carbides to F–T synthesis activity” [15].

In previous work, X-ray diffraction has been considered of little use in providing information on the nature of the iron F–T synthesis catalyst, especially when compared with Mössbauer spectroscopy. The major reason is that Mössbauer spectroscopy provides quantitative information on the distribution of iron phases in these catalysts. However, as we show in this work, when X-ray diffraction is coupled with Rietveld structure refinement [16,17], a wealth of detailed information on bulk composition, particle morphology, and phase identity can be obtained. To gain a better understanding of iron F–T synthesis catalyst behavior, we have systematically examined working catalysts from a slurry phase reactor. The working catalyst exists as a mixture of highly reactive material suspended in product hydrocarbons. The iron catalyst oxidizes readily if not protected adequately from exposure to air or moisture upon sampling, separation from the waxy hydrocarbon product, or during characterization. We have analyzed the artifacts that may arise from sample handling and concluded that the commonly used technique of Soxhlet extraction may adversely affect catalyst composition and morphology. Our work has been performed on catalyst samples protected by the product hydrocarbons and demonstrates that XRD in conjunction with TEM may provide unique insight into the nature of working Fe F–T catalysts.

2. Experimental The study was performed as a cooperative effort of two laboratories: the catalyst reactivity study was performed at Texas A&M University (TAMU), whilst the characterization study was performed at the University of New Mexico (UNM). Specifically, we present an analysis of two F–T synthesis runs performed in a stirred tank reactor, which are part of a larger study of catalyst activation treatments conducted at TAMU. The catalysts used in both runs were prepared from the same precursor: the first catalyst was activated in H2 at 250◦ C (run SB-3425, hereafter called the H2 -activated catalyst)and the second was activated in CO at 280◦ C (run SA-0946, hereafter called the CO-activated catalyst). The details of catalyst preparation, pretreatment, reaction run conditions, and initial experimental behavior for F–T synthesis in a stirred tank reactor are presented elsewhere in this issue [13].

L.D. Mansker et al. / Applied Catalysis A: General 186 (1999) 277–296

Catalyst suspended in the waxy product was removed by dip tube under inert atmosphere from the working reactor, at different times during each run [18]. The importance of removing F–T synthesis catalyst slurry under inert atmosphere in order to preserve the composition and morphology of the catalyst has been discussed elsewhere [19]. Each slurry sample was split into two portions. One portion was used as-received for detailed X-ray and TEM analysis. A second portion was Soxhlet-extracted at TAMU, to separate the catalyst from the wax prior to analysis. All XRD analyses reported here were performed at UNM, with the exception of the Soxhlet-extracted powders from the CO-activated run (SA-0946), which were analyzed at TAMU. One slurry sample from the H2 -activated run (SB-3425) was concentrated in our lab by warming the slurry for several days at 150◦ C under inert atmosphere and letting the catalyst powder concentrate in the bottom portion of the vial. Material was cut from the bottom of the vial and scanned by XRD. The concentrated sample and the remainder of the slurry was then completely stripped of its wax under flowing inert at reaction temperature, for subsequent analysis by X-ray diffraction. Neutron diffraction studies of this sample are reported elsewhere [20]. X-ray diffraction data were collected at UNM using a Scintag PAD-V powder diffractometer with diffracted beam monochromator, operated in step-scan mode, using Bragg-Brentano (θ –2θ) geometry. Data were collected from 15◦ to 105◦ , at 0.002◦ per second for the H2 -activated catalyst (run SB-3425) and from 10◦ to 120◦ 2θ, 0.005◦ per second for the CO-activated catalyst in wax (run SA-0946). All scans have an average diffraction scan error in the order of ± 0.005◦ (± 0.0004 Å). X-ray diffraction data were collected at TAMU for the CO-activated catalyst after Soxhlet extraction (run SA-0946), using a Scintag XDS-2000 series powder diffractometer in fast scan (continuous) mode, Bragg-Brentano (θ –2θ ) geometry, .0167◦ per second scan rate, with an average diffraction scan error in the order of ±0.05◦ (±0.006 Å). In this paper, we will report angles to 2 decimal places, and d-spacings to 3 decimal places. None of the diffraction data were subjected to background or polarization correction, or K␣2 stripping, prior to analysis and interpretation. Slurries, concentrated slurries, and powders were all pack-mounted

279

in a zero-background substrate sample cell, without any preprocessing, addition of solvent, or grinding. Powders in oil were allowed to settle by gravity, then pack-mounted in the same manner as the slurries. Sample volume required was approximately 0.18 cc. Selected samples were periodically re-analyzed at UNM. Multiple data collections on selected samples, carefully preserved, from the H2 -activated catalyst synthesis run (SB-3425), were performed 24 h, 1 month, 4 months, 7 months, and 1 year apart, in order to check sample long-term stability, and to validate our analytical procedures. All phase identifications were made from powder diffraction patterns calculated from the single crystal data for each phase, by data comparison with reference diffraction patterns of pure, single-phase material (Fe5 C2 , Fe3 O4 , Fe2 O3 , or a-Fe) [20,21], or using the diffraction reference data contained in the International Center for Diffraction Data (ICDD), Joint Committee on Powder Diffraction Standards (JCPDS) powder diffraction data base [22], sets 1–42 (see Sections 3.1 and 3.3). High resolution transmission electron microscopy (HRTEM) of two samples from run SA-0946 has been described elsewhere [23]. The micrographs reported here illustrate the effects of Soxhlet extraction on the composition of the working catalyst in run SB-3425. Samples were prepared by embedding catalyst slurry in epoxy, and preparing thin slices using ultramicrotomy. The microtomed sections (∼ 40–50 nm thick) were carbon-coated and mounted on an holey carbon molybdenum grid, and examined with a JEOL 2010 TEM operated at 200 KeV.

3. Results 3.1. Calculated absolute powder X-ray diffraction intensities of major iron phases As an important step in the characterization of the iron Fischer–Tropsch catalyst, we will consider in detail how some of the commonly observed phases in the working catalyst appear, by X-ray diffraction, and how they may interfere with one another, when mixed. It is well-known that these materials [e.g., alpha-iron (␣-Fe), hexagonal Fe7 C3 , monoclinic Fe5 C2 or Fe2.5 C (also known as ␹ or Hägg carbide), and magnetite (Fe3 O4 )] show overlapping diffraction peaks in the

280

L.D. Mansker et al. / Applied Catalysis A: General 186 (1999) 277–296

region of interest, {25◦ ≤ 2θ ≤ 70◦ }. We do not include the pseudo-hexagonal (properly called monoclinic) ⑀0 -carbide (Fe2.2 C) that has been reportedly seen in other F–T reactor studies, because its crystal structure is ill-defined. Neither was it observed in any of the catalyst samples analyzed for this study. However, most previous analyses, including our own [8], implicitly assume that the absolute or relative intensities from each phase reflect the relative phase abundances. This is, in fact, not the case. The powder diffraction database [22] presents the data for each component in terms of relative diffraction intensities as a function of either degrees (◦ ) 2θ or the d-spacing, in Å. The most intense diffraction peak is set to a value of 100 counts per second or 100%, and all other peak intensities are adjusted accordingly. In principle, the powder diffraction database also lists a relative intensity ratio (RIR) factor for each phase, listed as I/Icor on the card, which is the ratio of a given compounds 100% relative intensity peak relative to the 100% relative intensity peak for corundum ( ␣-Al2 O3 ), in a 50 : 50 wt.% mixture. The I/Icor value is then used to ‘correct’ the raw 100% peak intensities of each phase for quantitation purposes, by accounting for differences in elemental and compound X-ray scattering cross-section [24]. The I/Icor (or RIR) is used to quantify phases in a mixture, as follows: Ii wi =k wj Ij k=

(1)

(I /Icor )j (I /Icor )i

where wi , wj , Ii and Ij refer to the weight fractions and diffraction peak intensities for phases i and j, respectively, and k is defined as the I/Icor for phase j divided by the I/Icor for phase i. The use of I/Icor as an empirical correction factor is based on an assumption that weight percents of the phases in a mixture are similar, and that effects such as preferred orientation, extinction, mixing inhomogeneities, and the breadth of the crystallite size distribution are small enough that the changes in the value of I/Icor are minimal. In the case of the Fischer–Tropsch catalysts, the deviation from equal weight percent mixtures can be substantial; intermediate, disordered, defect or distorted structures are common, particles of material for a given phase can deviate significantly from spherical (preferred ori-

entation or growth habit), and significant anisotropy can occur frequently. For the iron phases of interest for F–T synthesis, the only I/Icor listed in the powder diffraction database is that for hematite (␣-Fe2 O3, card # 33–664) [22] , so we calculated powder diffraction patterns for various iron phase mixtures, in order to estimate I/Icor for the phases not listed, then experimentally verified those of the oxide phases (Fe3 O4 , and Fe2 O3 ). We first calculated a diffraction pattern for a 50–50 wt.% mixture of ␣-Al2 O3 and the isostructural ␣-Fe2 O3 , using single crystal structures from the literature [25,26], then attempted to reproduce the data experimentally by preparing a 50 wt.% calcined hematite and calcined corundum mixture (both from Aldrich Chemical Company, 99.995% purity). The constants used to describe the instrumental parameters, background function, zero-point error, sample transparency, sample displacement, and crystallite sizes were taken from values determined in our laboratory. Results were encouraging: we obtained a calculated value of 2.39, and an experimental value of 2.49. These both compare favorably with the value of 2.4 in the literature, showing that either procedure yields a value that differs from the value of 2.4 listed in the powder diffraction database [22] by no more than 4%. Hence we are assuming that, for the next set of simulations, the error in experimental and calculated I/Icor will be comparable in magnitude. The details of the calculation and experimental verification method will be published elsewhere [20]. An I/Icor is estimated for each of the five phases, using the I /Icor |hem :   Ii Ihem Ii (calc) = (2) Icor Ihem Icor ref The calculated values are listed in Table 1 along with the experimental values of I /Icor |hem and I /Icor |mag determined in our laboratory [20]. Diffraction patterns were calculated using the DBWS-9411 Rietveld structure analysis program [27,28], following the method of phase quantification of Bish et al. and Hill [29–31]. The Rietveld method was developed by Hugo Rietveld in 1967, for crystal structure determination using neutron powder diffraction data [16,17], but its application to X-ray powder diffraction data has become widespread in the last decade. The method provides detailed information on

L.D. Mansker et al. / Applied Catalysis A: General 186 (1999) 277–296 Table 1 Absolute integrated peak intensities and relative intensity ratios for Fe phases (based on dp =15 nm for all phases) Phase

␣-Fe Fe7 C3 Fe5 C2 Fe3 O4 ␣-Fe2 O3 a b

wt.% Imax

I/Icor

0.20 0.20 0.20 0.20 0.20

1526 ± 32 4299 3.237 ± 066 (c) 9 0.355 ± 008 (c) 1 141.3 ± 3 (c)/147 (e)b 398 2.39 (c) ± 049/2.49 (e) 6.7

5750 12.19 1.34 532.2 9.191

I/I(Fe5C2) (c)a

(c): calculated. (e): experimental, UNM.

the structure, morphology and phase abundances in a polycrystalline sample from first principles. Using this method, quantifications are possible down to a few hundreds of ppm. It can also be used to determine particle size (average and distributions), crystal habit, particle shape (indirectly), particle surface roughness, stress and strain in the powder, the nature of lattice vacancies and defects. Fig. 1(a) shows the plot of the simulated diffraction data for the 5-phase mixture (20 wt.% each of ␣-Fe, Fe7 C3 , Fe5 C2 , Fe3 O4 , and ␣-Fe2 O3 ) based on single-crystal data from the literature [25,32–42]. The location and integrated intensities of the 100% relative intensity peaks for each phase have been indicated. The figure demonstrates the degree of diffraction peak overlap that is characteristic of these materials in the region, 25◦ –70◦ 2θ. It also illustrates that, when absolute intensities are compared, a mixture containing equal wt.% of the five Fe phases of interest can produce a diffraction pattern which appears visually to be a mixture of mostly ␣-Fe and Fe3 O4 , with little or no Fe7 C3 , Fe5 C2 or Fe2 O3 . For completeness, the powder patterns for each of the phases contributing to the mixture were calculated, and are plotted in Fig. 1(b). The diffraction patterns for ␣-Fe, Fe5 C2 , Fe3 O4 , and ␣-Fe2 O3 were checked against X-ray and neutron diffraction patterns obtained from pure, single-phase materials prepared in our laboratory [20]. The diffraction pattern for each phase has been normalized to a maximum intensity of 100, so that the details in each diffraction pattern can be seen. The experimental diffraction patterns, along with tables of d-spacings, 2θ angles, and relative intensities from both the calculated (Fig. 1(b)) and experimental reference patterns will be published elsewhere [20]. D-spacings, 2θ angles, and relative intensities from

281

experimental diffraction patterns similar to the calculated patterns of Fe7 C3 and Fe5 C2 are listed in the JCPDS database, cards 36–1248 and 17–0333 [22]. It is important to remember that the powder patterns shown here for the carbides and oxides have been constructed from the single crystal structures (referenced previously) without accounting for the phase disorder, significant structural distortion, lattice defects, non-spherical particle effects, or bimodal particle size distributions often seen in these catalysts [20]. Hence, the diffraction patterns of material formed in the working catalyst can vary somewhat from the ones shown here. The kind and number of defects and distortions which can be observed in the Fe phases are quite systematic in nature, and will be discussed in detail elsewhere [20]. Comparison of the estimated I/Icor values and the composite diffraction pattern (Fig. 1(a)) demonstrates that large differences in peak intensities can exist, even though the iron phases are combined at equal wt.%. It becomes clear that differences in X-ray scattering cross-sections must be considered when examining the Fe catalysts used for F–T synthesis. A simple examination of relative peak heights cannot be used to infer the relative amounts of a material present in the working catalyst, without use of the RIR (I/Icor ) correction factors.

3.2. Influence of soxhlet extraction on catalyst morphology Soxhlet extraction is commonly used to concentrate and separate the catalyst from the product wax. Unfortunately, because it employs solvent at its boiling point (in the order of 80◦ C) [13], Soxhlet extraction could induce undesirable changes in the powder, leading to deceptive results. We examined the H2 -activated catalyst as removed from the reactor after pretreatment suspended in oil (run SB-3425, TOS = 000), and compared it to the Soxhlet-extracted powder from the same sample. This sample is composed of ␣-Fe, with unreduced oxide and silica binder. The ␣-Fe is known to be very reactive and pyrophoric if not properly passivated. Thus, if the extraction process affects catalyst composition and morphology, it would be easily observed in this sample. Fig. 2 shows a comparison of the diffraction patterns of the activated catalyst

282

L.D. Mansker et al. / Applied Catalysis A: General 186 (1999) 277–296

Fig. 1. (a) Calculated X-ray diffraction pattern of a 20 wt.% mixture of each: Fe2 O3 , Fe3 O4 , ␣-Fe, Fe7 C3 , and Fe5 C2 ; (b): calculated pattern for each phase: Fe2 O3 , Fe3 O4 , ␣-Fe, Fe7 C3 , and Fe5 C2 .

suspended in oil, and the activated sample after Soxhlet extraction. The differences between these two diffraction patterns are quite subtle, and could be easily overlooked; however, closer examination of the

data reveals definite evidence of changes caused by Soxhlet extraction. The catalyst in oil contains a prominent peak of ␣-Fe, with fairly uniform, large crystallite size (sym-

L.D. Mansker et al. / Applied Catalysis A: General 186 (1999) 277–296

283

Fig. 2. X-ray diffraction pattern, H2 -activated run SB-3425, TOS = 000 h. Lower curve, sample in oil. Upper curve: powder after Soxhlet extraction.

metric, nearly Gaussian peak shape, and fairly narrow full-width at half maximum-FWHM). The ␣-Fe peak in the extracted catalyst has a broader FWHM and is asymmetrically shaped, which we attribute to the possible presence of a second phase, for example of Fe5 C2 phase overlapping with the ␣-Fe peak. The sample also contains Fe3 O4 . The diffraction peaks are very broad at the base, indicating that the Fe3 O4 may show a bimodal particle size distribution. The height of the ␣-Fe peak relative to the background is lower than seen in the sample in oil, indicating a smaller wt.% of ␣-Fe in this sample. Further details of the Rietveld refinement of these samples are included elsewhere [20]. TEM images of the H2 -activated catalyst (SB-3425, TOS=000 h.) are shown in Fig. 3. A low-magnification micrograph of the H2 -activated catalyst after Soxhletextraction (Fig. 3 a), shows irregularly shaped particles of ␣-Fe (dark) in a matrix that contains finer catalyst particles, silica, and some unreduced iron phase. Note the presence of a pronounced surface halo on the ␣-Fe particle surfaces, such as that seen on the particles highlighted by arrows in Fig. 3(a). The high-magnification image of one of these haloed particles (Fig. 3(c)) clearly shows the thickness of the surface layer is in the order of 5 nm. The inner core shows lattice fringes of about 2.03 Å, which are characteristic of the ␣-Fe (110) crystal planes. A similar

surface oxidation layer has been shown to result from the passivating oxide film formed on the surface of ␣-Fe after air exposure [43]. The micrographs of the activated, unreacted catalyst powder in oil (run SB-3425, TOS = 000) are shown in Fig. 3(b) and (d). The overall catalyst microstructure is similar to the extracted catalyst shown in Fig. 3(a) and (c), but the particles do not show same the thick, pronounced surface halos on their surfaces. The high magnification view of one of the primary particles (Fig. 3(d), examples indicated by arrow in Fig. 3(b)) shows a particle of irregular shape, with a much thinner surface halo. The H2 -activated catalyst sample (TOS = 000 h) shows evidence of subtle changes after Soxhlet extraction. While these subtle changes may be dismissed and considered trivial, it should be pointed out that ␣-Fe makes up a large fraction of this sample. In a sample with smaller amounts of ␣-Fe (as is seen under reaction conditions), all of the reactive ␣-Fe would be lost. We conclude that the observed composition and morphology in the extracted sample is probably not representative of the material in the reactor. These changes are detectable by X-ray diffraction and HRTEM. However, it was not known whether all of the samples subjected to Soxhlet extraction would show similar changes. Accordingly, the rest

284

L.D. Mansker et al. / Applied Catalysis A: General 186 (1999) 277–296

Fig. 3. TEM image of H2 -activated catalyst, run SB-3425, (a): TOS = 000 h, extracted powder, low magnification view (b): TOS = 000 h, slurry, low magnification view (c): TOS = 000 h, extracted powder, high magnification view (d): TOS = 000 h, slurry, high magnification view.

of the samples, slurry and Soxhlet-extracted, were examined using XRD. The diffraction patterns of the H2 -activated catalyst, run SB-3425, in the hydrocarbon wax and after Soxhlet-extraction, are shown in Fig. 4(a) and (b) (five samples each, as a function of TOS). The X-ray diffraction patterns for the CO-activated catalyst, SA-0946, are shown in Fig. 4(c) and (d) (six samples each, as a function of TOS).

Visual inspection of the slurry diffraction patterns in Fig. 4(a) and (c) (H2 -activated and CO-activated catalyst, respectively), reveals that several peaks remain unchanged with time. As will be discussed in the next section, these diffraction peaks come from the hydrocarbon wax. While the presence of wax peaks obscures the location of some of the Fe3 O4 peaks, the most prominent Fe3 O4 peaks (indicated in

L.D. Mansker et al. / Applied Catalysis A: General 186 (1999) 277–296

285

Fig. 4. X-ray diffraction patterns of H2 -activated catalyst, run SB-3425, slurry samples. The location of the prominent ␣-Fe and Fe3 O4 peaks are indicated. (b): X-ray diffraction patterns of H2 -activated catalyst, run SB-3425, extracted samples. The locations of ␣-Fe, Fe3 O4 , and Fe5 C2 peaks are also indicated. (c): X-ray diffraction patterns of CO-activated catalyst, run SA-0946, slurry samples. The positions of Fe3 O4 , Fe5 C2 , and Fe7 C3 peaks are indicated. (d): X-ray diffraction patterns of catalyst from CO-activated catalyst, run SA-0946, extracted samples. Note the loss of plane-dependent line broadening in the Fe5 C2 phase, with increasing TOS (except in TOS = 563), indicating that the particles of a given phase are either breaking up or renucleating. The samples show significant zero-point error, as an S-shaped displacement in the peaks, with TOS. Peak positions for Fe3 O4 (mag), Fe5 C2 (c), and Fe7 C3 (a) are indicated, based on the reference angles and/or easily recognizable peak-group shapes. The Fe7 C3 (a) does not occur in significant amounts in these samples, the locations are labeled for reference.

286

L.D. Mansker et al. / Applied Catalysis A: General 186 (1999) 277–296

Fig. 4. (Continued).

L.D. Mansker et al. / Applied Catalysis A: General 186 (1999) 277–296

Fig. 4(b)) should have been observed if Fe3 O4 were indeed present in the catalyst slurry. Since they are not observed, we conclude that Fe3 O4 is not present in the slurry samples in statistically significant amounts in the H2 -activated sample (Fig. 4(a)). In contrast, the same samples after Soxhlet extraction (Fig. 4(b)) show the presence of significant amounts of Fe3 O4 , which is clearly not present in the slurry samples. Note that the ␣-Fe peak, which appears consistently in the slurry samples, is completely absent in the extracted samples, except for TOS = 000 h. In the Soxhlet-extracted samples of the CO-activated catalyst (SA-0946, Fig. 4(d)), TOS = 000 h to TOS = 427, we observe a loss of the Fe7 C3 carbide. However, the sample taken at end-of run (TOS = 563 h, CO-activated catalyst) appears to contain much more magnetite and a carbide with a different structure than any other samples. We conclude that Soxhlet-extraction effects can vary in severity. The variation could arise from extent of solvent outgassing, poor sealing of the extraction apparatus, etc. For completeness, TEM analysis was performed on the Soxhlet-extracted sample from end-of-run, H2 -activated catalyst, TOS = 384 (Fig. 5(a) and (c)) and on a slurry sample near end-of-run, H2 -activated catalyst, TOS = 330 (Fig. 5(b) and (d)). The activities of the catalyst at TOS = 330 and 384 did not differ very much, thus permitting the comparison. The Soxhlet-extracted sample (H2 -activated run, TOS = 384 h, Fig. 3(a)) contains primary particles of irregular shape covered with a surface halo. Higher magnification views (Fig. 3(c)) illustrate that these surface layers exhibit lattice fringes consistent with Fe3 O4 . The rough surface layer seems to indicate that oxidation of the particle surface has occurred. The low-magnification micrograph (Fig. 5(b)) of the slurry catalyst sample (H2 -activated run, TOS = 330 h) shows particles with morphology similar to that seen after Soxhlet extraction, and also show layers on the surface. While a passivating surface oxide formed at room temperature is often amorphous [43], beam-induced heating during TEM analysis transforms it into crystalline magnetite (Fe3 O4 ). The high-magnification image of this catalyst (Fig. 5(d)), shows that the surface layers observed in the slurry sample remain amorphous, suggesting that the surface layer does not represent a passivating film. Energy loss spectroscopy of similar surface layers

287

obtained after CO activation reveals that the surface layers contain amorphous carbon [21]. The differences between the sample in slurry and after Soxhlet extraction once again indicate that oxidation of the sample occurred either during or after Soxhlet extraction. Based on all these results, we conclude that Soxhlet extraction of the slurry samples is undesirable. We see evidence of oxidation of the reduced iron phases which could occur either during extraction or upon subsequent exposure to air. Hence subsequent work is based on the study of catalysts in the wax. Fig. 6 shows six diffraction patterns from the H2 -activated catalyst (run SB-3425, TOS = 233 h). The diffraction patterns are divided into three groups, from top to bottom: the Soxhlet-extracted powder (EP), catalyst concentrated by stripping the slurry of wax under flowing inert gas (SP), catalyst concentrated by sedimentation at 150◦ under inert atmosphere for several days (CC), as-received slurry (SL), wax obtained by stripping under flowing inert (SW), and filtered product wax from the reactor (FW). Note the remarkable difference between the two ‘clean’ catalyst powders, EP and SP. The ␣-Fe peak is completely gone from the Soxhlet-extracted powder (EP), which also contains a broad Fe3 O4 peak not present in the original sample. A large Fe5 C2 peak is seen also in the EP (refer to Fig. 2, extracted), while the SP appears to contain ␣-Fe, with trace carbide (based on a visual inspection of peak heights).

3.3. Interference of the crystalline wax with catalyst in the slurryof the crystalline wax with catalyst in the slurry Comparison of the slurry diffraction pattern (SL) to that of the waxes (EW and SW) shows that the wax accounts for many of the diffraction peaks present in the slurry (Fig. 6). In order to determine if entrainment were a problem, the ‘catalyst-free’ waxes (FW and SW) were examined by atomic absorption spectroscopy (AAS) to determine how much catalyst solid, if any, remained suspended in them. The product wax (FW) showed a total iron content of 80 ppm, while the stripped wax (SW) showed a total iron content of 207 ppm. Both are well below our detection limit of 350 ppm for XRD, so we concluded that iron entrained

288

L.D. Mansker et al. / Applied Catalysis A: General 186 (1999) 277–296

Fig. 5. TEM image of the H2 -activated catalyst, run SB-3425 (a): TOS = 384 h, extracted powder, low magnification view (b): TOS = 330 h, slurry, low magnification view (c): TOS = 384 h, extracted powder, high magnification view (the powder was mounted directly upon the TEM grid, without embedding in epoxy and ultramicrotomy). (d): TOS = 330 h, slurry, high magnification view.

in the separated wax could not contribute significantly to the wax diffraction pattern. The high molecular weight hydrocarbon wax obtained with an high ␣ catalyst can be quite crystalline, and its diffraction pattern can overlap and obscure the diffraction patterns of the various Fe phases of interest.

The wax peaks could therefore, be easily misidentified as several iron phases, as is evident from the diffraction peak positions listed in Table 2. We attempted to ameliorate the wax interferences by physically separating the catalyst from the wax. The as-received slurry was concentrated, first, by warming it at 150◦ C for

L.D. Mansker et al. / Applied Catalysis A: General 186 (1999) 277–296

289

Fig. 6. X-ray diffraction pattern of catalyst components, H2 -activated catalyst, run SB-3425, TOS = 233 h. From top to bottom: the Soxhlet-extracted powder (EP), catalyst concentrated by stripping the slurry of wax under flowing inert gas (SP), catalyst concentrated by sedimentation at 150◦ under inert atmosphere for several days (CC), as-received slurry (SL), wax obtained by stripping under flowing inert (SW), and filtered product wax from the reactor (FW). The ␣-Fe peak seen in the slurry sample grows in size as the catalyst is concentrated, while the wax peaks diminish in intensity. Note the remarkable difference between the Soxhlet extracted powder (EP) and the inert-stripped powder (SP). Table 2 d-spacings, 2θ, and Irel in the wax product diffraction pattern d-spacing, Å ◦

2θ angle, % Intensity, filtered wax % Intensity, stripped wax

2.986

2.569

2.545

2.485

2.108

1.613

1.444

29.9 79.96 62.58

34.9 45.76 42.31

35.24 48.79 42.99

36.12 100 66.87

42.86 98.39 100

57.04 30.18 25.84

64.46 24.74 19.11

several days and allowing the powder to settle by gravity (CC, Fig. 6). After analyzing by XRD, we recombined CC with the rest of the slurry, and stripped the sample of the wax in flowing inert at reaction temperature (SP, Fig. 6). The concentrated slurry (CC), when compared with the as-received slurry (SL) and the inert-stripped powder (SP), indicates that the concentration and stripping methods do a far better job of maintaining catalyst composition and morphology than does Soxhlet extraction. The second approach was to analyze the wax structure using Rietveld refinement using a low-density polyethylene (LDPE) single

crystal structure [44–49]. The wax diffraction pattern could be successfully deconvoluted from that of the catalyst, as discussed in detail elsewhere [20]. Re-examination of the slurry sample (H2 -activated catalyst, TOS = 384) 7 months after first exposure to air revealed that the ␣-Fe peaks, present in the first scan at 44.74◦ and 64.94◦ 2θ , were not present in the same sample 7 months later. The repeatability runs are discussed in detail elsewhere [19,20,50]. This seemed to indicate that Fe catalyst suspended in the wax is so reactive that it can still undergo changes over time, if not well-protected from air. However, all the

290

L.D. Mansker et al. / Applied Catalysis A: General 186 (1999) 277–296

samples analyzed in this study were removed from the reactor under inert atmosphere and stored under inert atmosphere until they were used for analysis. Hence, none of the samples were subject to the air exposure problems discussed above. The data appear to indicate that stripping away the wax and carefully passivating the powder is more effective at preserving the catalyst morphology and composition than Soxhlet extraction. However, the catalyst remains highly reactive, and changes over time. We conclude that F–T synthesis catalyst samples must be protected under an inert atmosphere, even in the waxy product. This shows that it is important that samples should be analyzed in a timely manner.

3.4. Fischer–Tropsch reaction results Catalyst activity results, expressed in terms of an apparent reaction rate constant, for the H2 -activated (SB-3425) and CO-activated (SA-0946) catalyst, are included here as Fig. 7. The apparent rate constant was calculated from experimental data assuming that the reaction rate has a first-order dependence on hydrogen partial pressure, and that reactor behaves as a perfectly mixed flow reactor [51]. The value of the apparent rate constant is representative of the catalyst activity during testing at different reaction conditions (different flow rate, reaction pressure, reaction temperature and/or synthesis gas feed composition). Syngas (H2 + CO) conversions in these two runs were generally high (65–80%) and their numerical values up to 140 h on stream were reported by Bukur et. al. [13]. Table 3 shows reaction conditions used during these two runs at different times on stream, as well as the syngas conversion at selected times on stream. The catalyst activated in CO at 280◦ C (SA-0946) shows a lengthy induction period characterized by relatively low initial activity, followed by a steady rise up to about 100 h on stream. The run conditions were changed to maintain a high conversion throughout the run. The catalyst reaches its maximum activity near TOS = 113 h, where a slurry sample was obtained. A moderate decline in catalyst activity occurs towards the end of the run (450–550 h). The catalyst activated in H2 at 250◦ C (SB-3425) reaches its maximum activity in 4–10 h, followed by a period of continuously decreasing activity. The CO-activated catalyst shows

better activity maintenance than the H2 -activated catalyst. In the next section, we will relate the observed variations in catalyst activity to the phase composition and morphology.

3.5. Catalyst evolution with time on stream We will now describe the evolution of the working catalyst microstructure, based on the slurry sample diffraction data (Fig. 4(a) and (c)), then discuss what this tells us about this particular catalyst’s active components, over time. We will contrast the behavior of the catalyst after activation in CO to that of the same catalyst activated in H2 . The sample compositions were determined using quantitative Rietveld structure refinement [16,17,29–31], or the estimated I/Icor for a given phase with its 100% peak intensity, as observed in the diffraction pattern. Quantitation data for the H2 -activated catalyst (SB-3425) and the CO-activated catalyst (SA-0946) are reported in Tables 4 and 5, respectively. The data are shown as wt.% crystalline components, ignoring the product wax, silica binder, and unreduced oxide, if any. Two consistency tests were applied to the H2 -activated catalyst results (Table 4). First, wt.% at TOS = 111 h were calculated using the estimated I/Icor and the 100% peak heights, and then using the more rigorous Rietveld structure analysis method. The results are in good agreement with one another. The second consistency test involved applying the I/Icor and Rietveld quantitation methods to the data from TOS = 233, slurry and inert-stripped powder (SP). The high intensity of scattering from the ␣-Fe phase, as manifested in the I/Icor values reported in Table 1, means that the carbide peak will be difficult to see by X-ray, when the ␣-Fe is present. In spite of this, the quantitation of the slurry data appears consistent with the quantitation obtained from the inert-stripped powder [20]. The sample at TOS = 000 h primarily contains ␣-Fe. The diffraction pattern also shows the presence of the silica binder and unreduced oxide phase as diffuse humps in the diffraction pattern, which were not quantified for this study. The crystallites are spherical in shape, and show a narrow particle size distribution, which is in reasonable agreement with the TEM image reported in Fig. 3. The next sample, taken

L.D. Mansker et al. / Applied Catalysis A: General 186 (1999) 277–296

291

Fig. 7. Fischer–Tropsch reactivity curves for H2 -activated run SB-3425, and CO-activated run SA-0946. The results are plotted in the form of a pseudo-first-order rate constant, referenced to 250◦ C. The rate constant units are in mmol of CO converted per g Fe per MPa pressure per hour. The reactor was operated at temperatures ranging from 260◦ C to 266◦ C and varying space velocities. The process conditions and actual conversion at several times on stream are shown in Table 3.

at TOS = 111 h, shows that practically all the sample appears to have transformed into the hexagonal carbide, Fe7 C3 , with trace ␣-Fe. The diffraction pattern shows diffraction-plane-dependent line broadening in the carbide (Fig. 4(a)), which indicates that the catalyst particles have become rough and non-spherical [20]. The catalyst is active until approximately TOS = 160 h, when a fairly sharp drop-off occurs. The feed gas space velocity was decreased 23% (Table 3), and the system took time to reach a new steady state. The catalyst activity begins to fall quickly after TOS = 233 h. A process upset occurred at TOS = ∼278 h, then the H2 /CO ratio was reduced slightly at TOS = 311 h (Table 3). We now see the appearance of monoclinic carbide with Fe5 C2 stoichiometry at TOS = 330 h. The catalyst contains five times as much Fe5 C2 as Fe7 C3 . At TOS = 330 h.

the space velocity was decreased 45% (Table 3). By TOS = 384 h, the catalyst appears to have essentially transformed into the Fe5 C2 carbide. Its average dp has doubled from the value at 330 h, indicating that sintering has occurred. The average particle sizes listed in Table 4, TOS = 330 h, compare favorably with the average particle sizes of seen in the TEM images in Fig. 3(a). The slight differences can be explained by two phenomena. First, XRD determines a volume-average particle size, while TEM determines an number-average particle size. Second, XRD is sensitive to the size and shape of the crystalline domains while TEM sees the entire particle. Hence, if the particles are polycrystalline aggregates instead of single crystals, the XRD estimate of particle size will be smaller than the TEM size. The TEM image does show that many

292

L.D. Mansker et al. / Applied Catalysis A: General 186 (1999) 277–296

Table 3 Process changes with corresponding % conversion TOS, h

% Conversion

SA-0946: Change 0 113 133 229 231 354 427 443 563

Sample; T = 260◦ C; P = 200 PSIG; H2 /CO = 0.67; SV = 2.34 Nl/g-cat/h Sample SV = 1.8 Nl/g-cat/h Sample P = 300 PSIG; SV = 2.64 Nl/g-cat/h Sample Sample H2 /CO = 0.6; SV = 2.0 Nl/g-cat/h Sample

36.1 76.6 76.0 80.2 76.4 81.1 81.0 81.5 81.9

(2 h) (112 h) (131 h) (228 h) (235 h) (351 h) (426 h) (442 h) (561)

SB-3425: Change 0 111 159 233 278.5 279 282 311 330 330 355 384

Sample; H2 /CO = 0.67; T = 260◦ C; SV = 2.34 Nl/g-cat/h; P = 200 PSIG Sample SV = 1.80 Nl/g-cat/h Sample Power out T = 200◦ C T = 260◦ C H2 /CO = 0.6 Sample SV = 1.0 Nl/g-cat/h T = 266◦ C Sample

75.8 70.8 70.2 73.5 71.8 --70.0 69.8 65.1 81.7 80.0 80.1

(1.5 h) (110 h) (157 h) (230 h) (276 h) (287 h) (310 h) (329 h) (340 h) (353 h) (383 h)

Table 4 Wt.% and particle sizes of iron phases in SB-3425, slurry samples TOS, h

␣-Fe

Wt.% 0 100 111 0.02 (QRSR)0.03 (I/Icor ) 233 3.09 330 0.42 384 0.004 Particle sizes, nm 0 14 111 38 233 31 330 38 384 37

Fe7 C3

Fe5 C2 (so-called ␹)

Fe3 O4 (magnetite)

0 99.98(QRSR)98.25 (I/Icor ) 96.91 14.32 2.47

0 0 0 85.26 97.52

0 0 0 0 0

5 47 5 5

– – – 8 15

– – – – –

of the particles in Fig. 3(a) are not single crystals, and high resolution images, such as the one in Fig. 3(c), show the lattice planes in the crystal permitting identification of the individual crystal phase. Average phase composition is more difficult to estimate from TEM, since it requires high resolution imaging of numerous crystallites and matching the lattice planes with the expected spacings for each phase. Our limited survey of these particles indicated that

the majority of them showed lattice fringes of 2.14 Å, which is consistent with the presence of Fe5 C2 . Hence, the TEM of the slurry at TOS = 330 h. is in good agreement with the phase composition results of the Rietveld structure analysis. The same is true of the sample from TOS = 384 h. The loss of activity in this catalyst appears to be related to the loss of the ␣-Fe and Fe7 C3 phases and concurrent growth of the Fe5 C2 phase, by TOS = 330.

L.D. Mansker et al. / Applied Catalysis A: General 186 (1999) 277–296

293

Table 5 Wt.% and particle sizes of iron phases in SB-3425, slurry samples TOS, h Wt.% 0 113 229 354 427 563 Particle sizes, nm 0 113 229 354 427 563

␣-Fe – – 0.01 0.01 0.02 – – – 52 19 57 –

Fe7 C3

Fe5 C2 (so-called ␹)

13.41 46.17 49.46 53.05 64.61 12.19

86.38 52.85 49.48 46.66 35.37 87.40

13.7 12 11 13 15 30

26.9 12 8 8 19 77

This behavior suggests that ␣-Fe and Fe7 C3 may be necessary components of an active catalyst, and Fe5 C2 is less active for F–T synthesis. Although not listed in the table, the relative wt.% and volume-average molecular weights of wax were monitored in the sample series, as a function of TOS. Rietveld refinement of the slurries gave an average total organic weight percent of 99.98% (34.99 wt.% MWavg = 2500 g/mole, 64.99 wt.% MWavg = 1500 g/mole), for all four samples, TOS = {111, 233, 330, and 384} h [20]. The composition and crystallite size data for the CO-activated catalyst run (SA-0946) are listed in Table 5. Once again, the data shown are wt.% crystalline components, ignoring the amounts of product wax, silica binder, and unreduced oxide, if any. The initial slurry sample (TOS = 000 h) consists mainly of monoclinic Fe5 C2 with no defects or distortions, some hexagonal Fe7 C3 , and trace Fe3 O4 . The ratio of monoclinic carbide to hexagonal carbide is 5. The particle size distributions are fairly uniform, showing little bimodality. The sample taken at TOS = 113 h shows a very different sample composition, independent of wax and other amorphous material. The catalyst contains distorted monoclinic carbide with Fe5 C2 stoichiometry, hexagonal Fe7 C3 , and trace Fe3 O4 . The particle size distribution is now bimodal. The particles have become rough and non-spherical, which manifests as diffraction-plane-dependent line broadening in the diffraction pattern (Fig. 4(c)) [20]. The TEM

Fe3 O4 (magnetite) 0.22 0.99 1.05 0.28 0.01 0.41 19 48 44 142 146 158

images of this sample, [23], show the presence of nanometer-sized carbide particles co-existing with larger carbide particles. This coincides with a maximum in the activity versus time curve for this run. The space velocity was decreased 23% at TOS = 133 h (Table 3). The activity drops slightly, then is relatively constant for a time. More samples were taken at TOS = 229, 354 and 427 h. These contain ␣-Fe, monoclinic Fe5 C2 , hexagonal Fe7 C3 , and Fe3 O4 . The trace amounts of ␣-Fe and Fe3 O4 remain essentially constant during this period. Referring to Table 3, note that the pressure and space velocity were increased 50% at TOS = 231 h. The activity begins to drop off again after TOS = 427 h. At TOS = 443 h, the syngas ratio was decreased slightly, then the space velocity was reduced 25% (Table 3). The sample at TOS = 427 h contains all four Fe phases mentioned previously, but the wt.% of hexagonal Fe7 C3 carbide has increased again, with a corresponding decrease in wt.% of the monoclinic Fe5 C2 carbide. Note that the phase compositions of the CO-activated catalyst at TOS = 563 h are similar to those seen in the unreacted sample after pretreatment (TOS = 000 h) and in the samples from TOS = 330 and 384 h, H2 -activated catalyst (SB-3425). Though not shown here, TEM micrographs of the CO-activated catalyst taken at TOS = 000 h [23] and those of the H2 -activated catalyst at TOS = 330 and 384 h. look very similar. These catalysts also show similar activity at their respective TOS. We can see smooth, unfaulted

294

L.D. Mansker et al. / Applied Catalysis A: General 186 (1999) 277–296

Fe5 C2 particles of uniform size in the TOS = 000 sample, in Fig. 5(a) and (b), H2 -activated catalyst. We interpret the data to show that the initial low kinetic activity of the CO-activated catalyst can be attributed to the initial phase composition and smooth, uniformly sized particles. The ‘induction period’ observed at the beginning of the run corresponds to the time required for the transformation of some of the catalyst material into Fe7 C3 carbide, with concurrent crystallite break-up in both phases [20]. The mild deactivation towards the end of the run seems to be caused by an increase in the Fe5 C2 carbide relative to the Fe7 C3 carbide, and an increase in crystallite sizes for the both phases [19–21,23]. The Fe7 C3 carbide is not observed in significant amounts in either catalyst during the activation treatment, but seems to form under reaction conditions at elevated pressure.

3.6. Relationship to previous work Note that previous studies of the carbide phase transitions [52–54] have not reported the hexagonal Fe7 C3 carbide. Instead, most previous work has analyzed catalyst deactivation in terms of the interconversion of the so-called ⑀0 -carbide of Barton and Gale [55] (Fe2.2 C) with the ␹-carbide (Fe2.5 C-also Fe5 C2 or Hägg carbide) structure of Sènateur [39]. The previous work shows that the type of the carbide formed initially depends upon the pretreatment used: (a) CO reduction, (b) H2 reduction followed by CO, or (c) syngas (H2 + CO) reduction. Jung and Thompson [54] performed an in-situ XRD study and reported finding the ⑀0 -carbide after reducing the catalyst precursor in H2 , followed by reaction in syngas (H2 /CO = 3.2) at 503◦ K. After some time on-stream, the catalyst transformed from the ⑀0 phase to the ␹ phase, with a corresponding drop in reactivity. They attributed the observed catalyst deactivation to the ⑀0 →␹ phase transformation. This work was performed at low pressures (atmospheric) and for short times on-stream (∼20 min to ∼20 h). The main reaction product was CH4 — hence their reaction conditions were not typical of those used for F–T synthesis, which is generally conducted at pressures in the order of 10 atm or more, and which can sometimes take over 100 h to reach steady-state activity.

There are several studies where catalysts used for F–T synthesis at elevated pressures have been analyzed after reaction. In the study by Dictor and Bell [9], used samples were Soxhlet extracted before analysis. It is not possible to make a direct comparison of the phase compositions with those in our study, because of the uncertainties in the extent of phase transformations induced by Soxhlet extraction. However, we note some common features. Dictor and Bell [9] found that both their active catalysts showed peaks at 2.01 and 2.02 Å. In one case they identified this as coming from ␣-Fe. In the second catalyst, this peak was identified as ␹-carbide. In fact, the most intense peak corresponding to the Fe7 C3 carbide occurs at 2.018 Å and the most intense peak from ␣-Fe occurs at 2.028 Å. In view of the peak overlap between the carbides and ␣-Fe, it is entirely possible that the phase compositions in their catalysts were very similar to those observed in our work. XRD was used mainly as a qualitative tool in the work of Dictor and Bell [9], and this is generally true of most previous work where XRD has been used. Previous work at TAMU at high pressure [52,53] used samples which were protected under nitrogen during sample removal and X-ray diffraction analysis. All these runs involved a fixed bed reactor in which the sample composition varied along the reactor bed. The present study used a slurry reactor in which catalyst composition is uniform throughout the reactor. Hence, we make comparisons with the catalyst removed from the top of the fixed bed, since it is least susceptible to oxidation by the product H2 O. It is interesting to note that the catalyst showing stable and high activity contained ␣-Fe along with the carbide phases.

4. Summary and conclusions We report a careful, systematic analysis of catalysts used for reaction runs in a stirred tank reactor under conditions similar to those used for commercial F–T synthesis. The catalysts were obtained from two runs. One run used catalyst activated in CO while the other used catalyst activated in H2 . The two catalysts showed dramatically different compositions after pretreatment, and exhibited very different activity profiles upon reaction. The H2 -activated catalyst reached its peak activity at around TOS = 25 h, while

L.D. Mansker et al. / Applied Catalysis A: General 186 (1999) 277–296

the CO-activated catalyst reached its peak activity at about TOS = 100 h (Fig. 7). The CO-activated catalyst showed better activity maintenance than the H2 -activated catalyst as a function of time. We used X-ray diffraction as the primary analytical tool to characterize the catalyst as a function of time onstream in the reactor, then attempted to correlate phase compositions and morphologies with the observed activity. We can draw several conclusions from this study: • Fe catalysts separated from the product wax by Soxhlet extraction can exhibit significant changes in composition and/or morphology, hence it is preferable to rely on analyses of the catalyst in the hydrocarbon wax or develop less severe methods of catalyst–product separation. • Diffraction methods can be used effectively to analyze the reacted catalyst samples suspended in the waxy product, without further processing, when combined with rigorous crystal structure analysis. The accuracy of this method depends on the concentration of the catalyst in the wax, the absorption characteristics of the various iron phases, the length of time over which the data are collected, and the soundness of the iron phase crystal structure models employed. • The absolute intensities of the 100% diffraction lines from the various iron phases can vary by over three-orders of magnitude in a sample in which the product wax can constitute 99+ wt.% of the analytical sample. This implies that a purely visual examination of X-ray diffraction patterns resulting from the catalyst cannot always provide clues to the relative abundance of the various phases present, unless one accounts for differences in the X-ray scattering cross-sections. • The catalyst activated in CO contains monoclinic Fe5 C2 with some unreduced oxide, while the catalyst activated in H2 is composed of ␣-Fe. Once F–T synthesis begins, each catalyst readily transforms into a mixture in which significant amounts of hexagonal Fe7 C3 and trace amounts of ␣-Fe are present. In both cases, the catalyst appears to lose activity due to conversion of significant amounts of hexagonal Fe7 C3 into monoclinic Fe5 C2 and accompanying growth in crystallite size. The catalyst shows low activity when monoclinic Fe5 C2 carbide is the predominant phase.

295

• We suggest that hexagonal Fe7 C3 carbide is an important constituent of Fe F–T catalysts, and is seen in significant amounts only in catalysts used at high pressures (10 atm). We find that the hexagonal Fe7 C3 carbide is always accompanied by a trace amount of ␣-Fe in the working catalyst samples. The Fe5 C2 phase is also present in the mixture in significant amounts where high activity is sustained over time, as seen in the CO-activated catalyst (SA-0946). However, a key feature is the presence of highly-dispersed carbide particles in the active catalyst. Acknowledgements This work was supported under the U.S. Department of Energy University Coal Research Grant DE-FG22-95PC95210, with partial support from NSF HRD 93-53208 at the University of New Mexico, and by U.S. Department of Energy Contract DE-AC22-94PC93069 at Texas A&M University. We further acknowledge the X-ray Powder Diffraction Laboratory, Department of Earth and Planetary Sciences, University of New Mexico. References [1] L.K. Rath, J.R. Longanbach, Energy Sources 13 (1991) 443. [2] F.B. Blanchard, J.P. Raymond, B. Pommier, S.J. Teichner, J. Mol. Catal. 17 (1982) 171. [3] J.T. McCartney, L.J.E. Hofer, B. Seligman, J.A. Lecky, W.C. Peebles, R.B. Anderson, Ind. Eng. Chem. 57 (1953) 730. [4] D.J. Duvenhage, R.L. Espinoza, N.J. Coville, Stud. Surf. Sci. Catal. 88 (1994) 351. [5] C.S. Huang, L. Xu, B.H. Davis, Fuel Sci. Technol., International. 11 (1993) 639. [6] G.B. Raupp, W.N. Delgass, J. Catal. 58 (1979) 348. [7] J.A. Amelse, L.H. Schwartz, J.B. Butt, J. Catal. 72 (1981) 95. [8] M.D. Shroff, D.S. Kalakkad, K.E. Coulter, S.D. Kohler, M.S. Harrington, N.B. Jackson, A.G. Sault, A.K. Datye, J. Catal. 156(2) (1995) 185. [9] R.A. Dictor, A.T. Bell, J. Catal. 97 (1986) 121. [10] J.W. Niemantsverdriet, A.M. van der Kraan, W.L. van Dijk, H.S. van der Baan, J. Phys. Chem. 84(25) (1980) 3363. [11] J.W. Niemantsverdriet, A.M. van der Kraan, J. Catal. 72 (1981) 375. [12] D.S. Kalakkad, M.D. Shroff, S.D. Köhler, N.B. Jackson, A.K. Datye, Appl.Catal. A 133 (1995) 335. [13] D.B Bukur, X. Lang, Y. Ding, Appl. Catal., this issue. [14] D.J. Dwyer, G.A. Somorjai, J. Catal. 52 (1978) 291. [15] M.E. Dry, in: J.R. Anderson, M. Boudart (Eds.), Catalysis-Science and Technology, 1980, Chap. 15, p. 160.

296

L.D. Mansker et al. / Applied Catalysis A: General 186 (1999) 277–296

[16] H.M. Rietveld, Acta Crystallographica 22 (1967) 151. [17] H.M. Rietveld, J. Appl. Crystallogr. 2 (1969) 65. [18] D.B. Bukur, X. Lang, Y. Ding, Preprints of the American Chemical Society Division of Fuels Chemistry 42(2) (1997) 623. [19] L.D. Mansker, Y. Jin, A.K. Datye, Proc. Coal Liquefaction and Solid Fuels 1997 Conf. proc., electronic publication, Federal Energies and Technologies Center, Pittsburgh, Pennsylvania, 1997. [20] L.D. Mansker, Dissertation, Ph.D. University of New Mexico, 1999. [21] Y. Jin, Ph.D. Dissertation, University of New Mexico, 1999. [22] International Center for Diffraction Data (ICDD), Joint Committee on Powder Diffraction Standards (JCPDS) powder diffraction data base, sets 1–42, 1994. [23] Y. Jin, A.K. Datye, Proc. Int. Congress on Electron Microscopy; Cancun 1998, vol. II, Inst. Of Physics Publishing, 1998, p. 379. [24] W.L. Bragg, W.H. Bragg, in: W.L. Bragg, (Ed.), The Crystalline State: A General Survey, The Crystalline State, vol. 1, G. Bell, London, 1949. [25] M.I. Antipin, V.G. Tairelson, M.P. Flugge, R.G. Gerr, Iu.T. Struchkov, R.P. Oserov, Doklady Akadameii Nauk SSSR 281(4) (1985) 854. [26] P. Thompson, D.E. Cox, J.B. Hastings, J. Appl. Crystallogr. 20 (1987) 79. [27] A. Sakthivel, R.A. Young Rietveld Analysis Program DBWS-9411, Release 30.03.95, software, 1995. [28] R.A. Young, J. Appl. Crystallogr. 28 (1995) 366. [29] D.L. Bish, S.A. Howard, J. Appl. Crystallogr. 21 (1988) 86. [30] R.J. Hill, C.J. Howard, J. Appl. Crystallogr. 20 (1987) 467. [31] R.J. Hill, The Rietveld method, in: R.A. Young, (Ed.), International Union of Crystallography, Oxford University Press, Oxford, 1993, p. 61.

[32] [33] [34] [35] [36] [37] [38] [39] [40] [41] [42] [43] [44] [45] [46] [47] [48] [49] [50] [51] [52] [53] [54] [55]

M. Dirand, L. Afqir, Acta Metallurgica 31(7) (1983) 1089. M.E. Fleet, Acta Crystallographica B37 (1981) 917. M.E. Fleet, Acta Crystallographica B38 (1982) 1718. M.E. Fleet, Acta Crystallographica C40 (1984) 1491. M.E. Fleet, J. Solid State Chem. 62 (1986) 75. F.H. Herbstein, J.A. Snyman, Inorg. Chem. 3(6) (1964) 894. H. Okudera, K. Kihara, T. Matsumoto, Acta Crystallographica B52 (1996) 450. J.P. Sénateur, Annales de Chimie 2 (1967) 103. D. Senczyk, Phase Transitions 43 (1993) 153. R.L. Withers, L.A. Bursill, J. Appl. Crystallogr. 13 (1980) 346. R.W.G. Wyckoff, Crystal Structures 3rd ed., Interscience, New York, 1960. M.D. Shroff, A. K Datye, Catal. Lett. 37(1-2) (1996) 101. D.L. Dorsett, B. Moss, Polymer 24(3) (1983) 291. B. Moss, D.L. Dorsett, Acta Crystallographica A40 (1984) C318. D.L. Dorsett, Polymer 27(9) (1986) 1349. H.L. Hu, D.L. Dorset, Acta Crystallographica B45 (1989) 283. D.L. Dorsett, Acta Crystallographica B51 (1995) 1021. D.L. Dorsett, Polymer 38(2) (1997) 247. N.B. Jackson, L.D. Mansker, R.J. O’Brien, B.H. Davis, A.K. Datye, Stud. Surf. Sci. Catal. 111 (1997) 501. W.H. Zimmermann, D.B. Bukur, Can. J. Chem. Eng. 68 (1990) 292. D.B. Bukur, M. Koranne, X. Lang, K.R.P.M. Rao, G.P. Huffman, Appl. Catal. 126 (1995) 85. D.B. Bukur, K. Okabe, M.P. Roysynek, C. Li, D. Wang, K.R.P.M. Rao, G.P. Huffman, J. Catal. 155 (1995) 353. H. Jung, W.J. Thomson, J. Catal. 134 (1992) 654. G.H. Barton, B. Gale, Acta Crystallographica 17 (1964) 1460.