1 Introduction

In Fischer–Tropsch synthesis (reactions i and ii), iron catalysts are often selected for converting syngas having low hydrogen to carbon monoxide ratio (e.g., 0.7–1.0, derived from coal or biomass) [1]. This is because iron catalysts possess intrinsic water–gas shift activity (reaction iii) that is used to increase the H2/CO ratio in situ in order to provide the H2 needed to further convert CO.

  1. (i)

    (2n + 1)H2 + nCO = CnH2n+2 + nH2O

  2. (ii)

    2nH2 + nCO = CnH2n + nH2O

  3. (iii)

    CO + H2O = H2 + CO2

With pressure mounting to decrease net emissions of greenhouse gas emissions, the utilization of renewable resources, such as biomass, is receiving greater attention, making iron catalysts an attractive option. Moreover, iron catalysts have high activity for FT and may serve as a lower cost alternative to cobalt, which due to its low WGS activity, would require the process to have an additional upstream water–gas shift unit.

Promoters for iron catalysts are different from those used for cobalt. Cobalt catalysts are often promoted with a second metal (e.g., Pt, Ru, Re) [2] to facilitate the reduction of cobalt oxides—perhaps by a H2 dissociation and spillover mechanism or a chemical effect. In contrast, in addition to structural promoters (e.g., Si), iron catalysts often contain an alkali or alkaline earth metal promoter, and often contain Cu as well. Potassium, as a result of its basic character, has been proposed to facilitate carburization during catalyst activation [3, 4], as well as influence the adsorption of the H2 and CO reactants on the catalyst surface (e.g., promoting the dissociative adsorption of CO and preventing excessive adsorption of hydrogen) resulting in lower methane and other light gas selectivities, higher C5+ selectivity, higher chain growth probability, and higher olefin selectivity.

Potassium has been investigated by numerous researchers over the years. Kölbel and Schneidt [5] reported that promoting precipitated iron catalysts with 0.05–0.2%K2CO3 increased surface area by up to 60%, increased the heat of reversible adsorption of CO by up to 50%, and the capacity for CO adsorption by up to 80%. Dosing with 0.6%K2CO3 was found to increase the yield of olefins from 22 to 60%. Kölbel, investigating Fe–Cu–Si catalysts, found that not only did potassium addition enhance CO chemisorption, but it also suppressed the chemisorption of hydrogen. Potassium was suggested to donate electrons to iron, promoting CO bond scission. Hydrogen, on the other hand, was proposed to donate electrons to iron such that electron donation from potassium to iron weakened the chemical bond between iron and hydrogen. Other studies have also supported the view that potassium acts to strengthen the Fe–C bond [6,7,8]. For example, Benziger and Madix reported that while S, O, and C tended to reduce the binding energies for CO and H, and inhibited the dissociation of CO, potassium was found to enhance the binding strengths of CO and H, and promoted the dissociation of CO. The enhanced binding energy of CO was proposed to be due to interactions between the K 4s orbital and the CO 2π* antibonding orbital.

Some studies have shown that potassium increases the catalytic activity of Fischer–Tropsch synthesis and water–gas shift and/or provides benefits in terms of selectivity. For example, Bukur et al. [9] found that 0.2–1 wt% K resulted in not only significant increases in FT and WGS activity, but also increases in selectivities to olefins and heavier products, with decreases in methane and C2-C4 light gas selectivities. In a systematic investigation of K-promoted Fe catalysts (100Fe:0K, 100Fe:0.5K, 100Fe:1.0K, 100Fe:2.0K), a selectivity comparison at constant conversion resulted in a systematic decrease in C1 and C2–C4 light gas selectivities, with systematic increases in C5+ selectivity and chain growth probability with increasing K content [10].

In that same investigation, a detailed study using Mössbauer spectroscopy and TEM was conducted to determine the effect of potassium content on deactivation rate. The unpromoted and heavily promoted catalysts (100Fe:2K) deactivated rapidly within 200 h of time on-stream at 270 °C, 13 atm, H2/CO = 0.7, and SV = 3 SL/gcat h. On the other hand, moderate levels (e.g., 100Fe:0.5 K) resulted in much higher stability. Interestingly, the characterization data indicated that the unpromoted Fe catalyst deactivated by oxidization to Fe3O4, while the 2K promoted catalyst deactivated by the formation of more carbon-rich iron carbides or intense carbon buildup as a function of time onstream. Thus, the carburization rate during FTS should be rapid enough to prevent oxidation of active iron carbides (e.g., χ-Fe5C2) but slow enough to avoid excessive carbon buildup or the formation of less active iron carbides having higher C/Fe ratios.

To date, all of these studies focus solely on how K affects the state of Fe as well as catalyst performance. There has been a scarcity in literature that report on the chemical state of potassium or how the structure of potassium changes during the course of the FTS reaction. In the current contribution, soft X-rays were used to probe the K absorption edge of potassium. Samples were retrieved from the reactor as a function of time on-stream, and solidified in the solid FT wax for analysis. A comparison was made of XANES spectra of the catalysts with relevant reference compounds, and changes in XANES spectra with catalyst time on-stream were examined.

2 Experimental

2.1 Catalyst Preparation

The precipitated iron catalyst parent batch was prepared by using a ferric nitrate solution obtained by dissolving iron (III) nitrate nonahydrate (1.17 M) in deionized water. A controlled flow of the iron nitrate mixture was added to the precipitation vessel together with a stream of ammonium hydroxide (14.8 M) that was added at a rate to maintain a pH of 9.0. The slurry was recovered using a vacuum filter and the solids were washed twice with deionized water. The final filter cake was dried for 24 h in an oven at 110 °C with flowing air. For this study, the iron catalyst base powder was then impregnated with varying amounts of aqueous potassium nitrate solution to produce a desired composition of Fe:K = 100:2.0 and 100:5.0 (atomic ratios). The K ratio between the two catalysts was confirmed by ICP to be 2.4, only slightly less than the nominal value of 2.5. Catalysts were then dried at 110 °C overnight, followed by calcination at 350 °C in flowing air for 4 h.

2.2 Catalyst Characterization

2.2.1 Surface Area and Pore Size Distribution

The measurement of BET surface area and porosity of the calcined catalysts was conducted using a Micromeritics Tri-Star system. Before performing the test, the temperature was gradually ramped to 160 °C and the sample was evacuated for 12 h to approximately 50 mTorr. The BET surface area, pore volume (single point), and average pore radius (single point) were obtained for each sample.

2.2.2 Temperature Programmed Reduction of CO

Temperature programmed reduction (TPR) of CO was conducted using a Zeton-Altamira AMI-200 unit with a thermal conductivity detector (TCD). TPR was performed using 10%CO/He at a flow rate of 30 cm3/min and catalyst samples were heated from 50 to 525 °C and held at 525 °C for 1 h. An isopropyl alcohol/dry ice trap was used to remove CO2 from the effluent during TPR.

2.2.3 X-ray Absorption Spectroscopy

Potassium K-edge XANES spectra of catalysts and reference compounds were recorded in fluorescence mode at the Soft X-ray Micro-characterization Beamline (SXRMB) at the Canadian Light Source, Inc. Catalyst samples were removed from the continuously stirred tank reactor (CSTR) under realistic FTS conditions and sealed in the wax product for analysis to prevent exposure to air. The spectra were background subtracted and normalized with a Victoreen function and further normalized using a two-polynomial method with degree 1 for both the pre- and post-edge regions using WinXAS [11].

Iron K-edge XANES spectra (recorded at SSRL) were processed in a similar manner. For iron EXAFS data, after background removal and normalization, spectra were converted to k-space and background subtracted in k-space using a cubic spline. The spectra were truncated over the range of 2.5–10 Å−1. Fe K-edge data were fitted with theoretically generated spectra from structural models assuming the presence of χ-Fe2.5C (Hägg carbide) and Fe3O4 (magnetite). Hägg carbide has a monoclinic Bravais lattice with a C 2/c space group. Lattice parameters are a = 11.504 Å, b = 4.524 Å, C = 5.012 Å, and β = 97.60°. Magnetite, on the other hand, is isometric hexoctahedral in structure with a = 8.391 Å and a space group of Fd3m. Using Atoms software [12], structural information of each contribution was converted to spatial coordinates, which were employed by FEFF [13] to determine scattering paths (Tables 1, 2). These scattering paths were inputted into FEFFIT [14] to generate theoretical χ(k), the fitting parameters of which were varied to fit the experimental data. Parameters included a global lattice expansion coefficient, α; the overall energy shift applied to each path, Δe0; Debye–Waller factor, σ, and amplitude coefficient to account for the contribution of each phase, and particle size.

Table 1 Path parameters generated by FEFF (single scattering) for Hägg carbide
Full size table
Table 2 Path parameters generated by FEFF (single scattering) for magnetite
Full size table

2.3 Catalyst Testing

The FTS experiments were conducted using a 1 L CSTR equipped with a magnetically driven stirrer with turbine impeller, a gas-inlet and a vapor outlet line, and a stainless steel (SS) fritted filter (2 micron) placed external to the reactor. A tube fitted with a SS-fritted filter (0.5 micron opening), extending below the liquid level of the reactor, was used to withdraw reactor wax in order to maintain a nearly constant liquid level in the reactor. Another SS dip-tube (1/8ʺ OD) extended to the bottom of the reactor and was used to withdraw aliquots of the catalyst/wax slurry from the reactor at different synthesis times. Separate mass flow controllers were used to regulate the flow of hydrogen and carbon monoxide at the desired rate. The gases were premixed in an equalization vessel before entering the reactor. Carbon monoxide was passed through a vessel containing lead oxide-alumina to remove traces of iron carbonyls. The mixed gases entered the CSTR below the stirrer, which was operated at 750 rpm. The reactor slurry temperature was maintained constant (±1 °C) by a temperature controller.

Catalyst (15 g) was added to C30 oil (310 g) in the CSTR to produce a slurry that contained about 5% iron oxide. The reactor temperature was then raised to 270 °C at a rate of 1 °C/min. The catalyst was activated using CO at a space velocity of 3.0 sl/h/gcat at 270 °C and 175 psig for 24 h. At the end of the activation period, a sample of activated catalyst was withdrawn through the dip-tube of the CSTR. The FTS reaction was then started by adding the synthesis gas mixture to the reactor at a space velocity of 3.0 sl/h/gcat and a H2/CO ratio of 0.7. The conversions of CO and H2 were obtained by gas-chromatography analysis (micro-GC equipped with thermal conductivity detectors) of the reactor effluent. The reaction products were collected in three traps maintained at different temperatures: a hot trap (200 °C), a warm trap (100 °C) and a cold trap (0 °C). The products were separated into different fractions (rewax, wax, oil and aqueous phase) for quantification. However, the oil and the wax fractions were mixed prior to GC analysis.

3 Results and Discussion

The BET surface area and pore size distribution results of un-promoted and various potassium promoted iron catalysts are shown in Table 3. If Fe2O3 is the primary contributor to surface area, addition of 2 and 5 K should lead to decreases in surface area to about 42.7 and 41.9 m2/g, respectively. However, more significant decreases were observed, to 29.0 and 31.2 m2/g, respectively. At the same time, increases in pore diameter were observed by 42 and 31% (single point data), suggesting that potassium addition caused some pore blocking, especially of narrower pores.

Table 3 BET surface area and porosity measurements of un-promoted and potassium promoted iron catalysts
Full size table

The CO TPR profiles of un-promoted and various potassium promoted iron catalysts are shown in Fig. 1. Based on TPR-XANES/EXAFS results [3], iron activation in CO proceeds in a series of steps, including: (1) reduction of Fe2O3 to Fe3O4; (2) further reduction of Fe3O4 to a defect-laden form resembling FeO; and (3) carburization of FeO to FexCy. The first peak in the profile, centered at close to 300 °C, is likely that of step (1), while the remaining oxide reduction steps occur in the higher temperature broad asymmetric peak. Carburization also likely begins in the asymmetric peak and continues at higher temperature and during the hold step. Based on the peak intensities during the hold procedure, CO uptake during carburization/carbon formation is more significant with addition of potassium, and as a function of potassium loading.

Fig. 1
figure 1

CO TPR profiles of 100Fe:0K, 100Fe:2K, and 100Fe:5K catalysts

Full size image

The effect of potassium promoter loading on CO conversion versus time-on-stream for the iron catalysts is shown in Fig. 2. At similar activation and reaction conditions (temperature, pressure and GHSV) but varying the potassium loading (2 and 5 K), high initial CO conversions (~80%) were observed for both catalysts. The activity of the catalysts gradually decreased with time-on-stream. Both catalysts exhibited similar deactivation trends. The effect of potassium promoter loading was investigated in our previous studies [10, 15], and a similar deactivation trend was observed for the 100Fe:2.0K catalyst. The deactivation of these catalysts was explained by the results of Mössbauer spectroscopy and TEM techniques. The effect of deactivation on product selectivities for the 100Fe:2.0K catalyst is presented in Fig. 3. With decreasing CO conversion or with time, olefin selectivity decreased and the corresponding paraffin and oxygenate selectivities increased. The 100Fe:5.0K catalyst also exhibited similar product selectivities as compared to the 100Fe:2.0K catalyst (not shown for the sake of brevity). The oxygenate selectivity, and especially the aldehydes selectivity, was found to increase with time (i.e., decreasing CO conversion). Changes in product selectivities with time may be due to the formation of a carbon layer on the surface of the catalyst. The effect of potassium promoter versus time on hydrocarbon selectivity of 100Fe:2.0K catalyst is shown in Fig. 4. There were no significant changes observed regarding hydrocarbon selectivity (methane, C2-C4 and C5+) of the catalyst either with time or CO conversion. However, the carbon dioxide selectivity decreased slightly with time. Generally, for cobalt and ruthenium based catalysts, the selectivity to lower hydrocarbons (C1-C4) decreases and that of the higher hydrocarbon selectivity (C5+) increases with increasing CO conversion, whereas for an iron catalyst, the product selectivity remains more or less constant at conversions less than 70% for potassium promoted iron catalysts [16].

Fig. 2
figure 2

CO conversion against time on stream for various potassium loaded iron catalysts

Full size image
Fig. 3
figure 3

CO conversion and product selectivities against time on stream of 100Fe:2.0K catalyst

Full size image
Fig. 4
figure 4

Hydrocarbon and CO2 selectivity against time on stream for a 100Fe:2.0K catalyst

Full size image

XANES spectra at the potassium K-edge for catalysts retrieved from the CSTR as a function of time on-stream is reported in Fig. 5. Within 20 eV of the edge jump, the near edge region consists primarily of a lower energy asymmetric peak with two low energy shoulders and a higher energy peak. Interestingly, with increases in time on-stream, the higher energy peak systematically increases in intensity and shifts to higher energy.

Fig. 5
figure 5

Normalized K K-edge XANES spectra of the 100Fe:2K catalyst (black) following activation in CO at 425 °C for 6 h and, moving along the trajectory of the arrow, after running Fischer–Tropsch synthesis for (red) 24 h, (green) 48 h, (blue) 95 h, and (pink) 165 h

Full size image

Potassium K-edge absorption spectra of numerous potassium compounds were extensively investigated by Sugiura and Muramatsu [17]. However, somewhat surprisingly, very little has been published on the potassium K-edge spectra of alkali doped catalysts. The authors examined families of potassium compounds and observed similarities (as well as some differences) in spectral features within each family. In reference to the current work, a similar splitting of peaks was found to occur in the family of compounds K2SO4, K2CrO4, and K2MnO4 investigated by Sugiura and Muramatsu [17]. Within that family, the authors assigned the low energy peak to electronic transitions (i.e., their Peak A pre-edge shoulder was assigned to 1s to 3d, while the main Peak B was assigned to the 1s to 4p transition) and the higher energy peak to a continuum state (i.e., multiple scattering).

Related to the potassium K-edge spectra of the K/Fe catalysts obtained in this work, the assignments by Sugiura and Muramatsu [17] make logical sense, as the peak for electronic transitions should remain stationary while the multiple scattering peak should be sensitive to changes in the local atomic environment. It has been speculated that the K compound in Fe-based FTS catalysts may be K2CO3 or KHCOO, or a related species. To test the idea that the second peak is due to multiple scattering, reference compounds were prepared of K2CO3 and KHCOO supported on silica and for each compound two different concentrations were used; in addition, spectra of the bulk reference compounds were also analyzed. These reference spectra are provided as (c) and (d) in Fig. 6. The results indicate that increasing the concentration of the reference compound supported on silica or using the bulk reference compound did not significantly alter the energy position of the first peak; this indicates the first peak is likely due to electronic tranisitions. However, increasing the concentration of the dispersed reference compound on silica, or moving from supported reference compound to the bulk reference compound, the second peak moves systematically to higher energy; this is consistent with the assignment by Sugiura and Muramatsu [17] that the peak is a continuum multiple scattering feature. In the case of potassium formate, the peak assigned to multiple scattering splits into two peaks in moving from supported to bulk reference compound, while in the case of potassium carbonate, a similar change occurs; however, the bulk reference peak for multiple scattering exhibits asymmetric character suggesting the presence of two contributions. Note that in Fig. 6d, due to the effect of self-absorption on XANES intensity for the potassium formate reference compound in fluorescence mode, a second spectrum was taken using a TEY detector (dashed line). As shown in Table 4, Peak C (bold) of KHCO2 moves from 3.6170 to 3.6175 to 3.6212 keV in moving from 0.97%KHCO2/SiO2 to 2.38%KHCO2/SiO2 and then to the bulk HCO2 compound. Likewise, Peak C (bold) of K2CO3 moves from 3.6170 to 3.6174 to 3.6208 keV in moving from 0.97%K2CO3 to 2.38%K2CO3 and further to the bulk K2CO3 compound.

Fig. 6
figure 6

Normalized K K-edge XANES spectra of a 100Fe:2K and b 100Fe:5K catalysts (black) after activation and (red) following the Fischer–Tropsch synthesis reaction test; c (black) 0.97% K2CO3/SiO2, (pink) 2.38% K2CO3/SiO2, and (blue) bulk K2CO3; and d (black) 0.97% KHCO2/SiO2, (pink) 2.38% KHCO2/SiO2, and (green) bulk KHCO2

Full size image
Table 4 K-K edge XANES peak positions for potassium formate on SiO2 and in bulk form
Full size table

Spectra for the CO-activated catalyst and the catalyst following the FTS reaction are also shown. Peaks due to electronic transitions are listed as A, B, and B′, consistent with the notation of Sugiura and Muramatsu [16]. Peak C was found to move to higher energy for both the 2 and 5 K promoted catalysts following FTS in comparison with the freshly carburized catalyst. Figure 5 emphasizes that the change in the position of Peak C (bold) occurs systematically as a function of time on-stream during the FTS reaction test. As shown in Table 5, Peak C moves from 3.6175 to 3.6188 keV for the 100Fe:2K catalyst and from 3.6180 to 3.6194 keV for the 100Fe:5K catalyst. These changes are consistent with the view of Sugiura and Muramatsu that Peak C is a continuum peak and that the local environment is changing to one of increased size/ordering with time on-stream. Also, the fact that Peak C in the spectrum of the CO-activated 5K catalyst is located at a higher energy (3.6180 keV) than that of the 2K catalyst (3.6175 keV) lends further support to the view that Peak C is a multiple scattering peak whose energy moves to higher values with increased size/ordering.

Table 5 K-K edge XANES peak positions for (black) CO-activated and (red) used catalysts after Fischer–Tropsch synthesis, including (a) 100Fe:2K and (b) 100Fe:5K
Full size table

These results reveal three important points: (1) the catalyst contains potassium in a well-dispersed state that does not match well that of bulk reference compounds; (2) the potassium compound located on the iron catalyst surface has similar characteristics to either potassium carbonate or potassium formate, and is likely a related compound; and (3) with time on-stream, the results strongly suggest that sintering of this surface potassium compound occurs with time, as the shift in energy position of Peak C, assigned to multiple scattering, is consistent with increasing size and crystalline order.

Sintering of potassium may be problematic for two reasons. First of all, sintering of potassium may exacerbate pore blocking, especially if there is excessive carbon being formed (e.g., the case of high potassium promoter loading [10]). Secondly, sintering removes surface potassium that otherwise promotes CO scission occurring on the Fe carbide surface. In the case of the un-promoted catalyst, deactivation by oxidation was found to occur [10]. Thus, there is the possibility, and especially in the case of low alkali loading, that sintering may remove a critical amount of promoter from the surface that is required to stabilize iron in a carburized state that is needed for conducting both the FTS and WGS reactions. However, although we attempted to investigate catalysts having low potassium content, the spectra were found to be too noisy to draw any conclusion.

In the current context, where high promoter loadings are utilized, loss of carbide by oxidation was not a key issue associated with deactivation. Figure 7 shows Fe K-edge XANES spectra of catalysts and reference compounds; the end-of-run (listed as EOR) catalysts display the characteristic line shape of Fe carbide consistent with a form close to that of χ-Fe5C2. Figure 8 and Table 6 indicate that excellent fittings of the spectra with a Hägg carbide model could be obtained without a significant contribution from the oxide. In Mössbauer spectroscopy, the carbide identified in the EOR 2K catalyst was Fe2.2C and residual Fe3O4 was just 7%; thus, it is evident that while EXAFS provides useful supporting information, it should probably not be used to distinguish between closely related carbides.

Fig. 7
figure 7

Normalized XANES spectra at the K-edge of Fe for reference compounds and end-of-run (EOR) 100Fe:5K and 100Fe:2K catalysts. Note that the Fe carbide reference was following a CO-TPR/XANES experiment for a Fe:1.5K catalyst (details provided elsewhere [3])

Full size image
Fig. 8
figure 8

EXAFS spectra at the K-edge of Fe for end-of-run 100Fe:5K and 100Fe:2K catalysts, including (left) raw χ(k) versus k spectra; (middle) (red) filtered χ(k) versus k and (filled circles) result of the EXAFS fitting; and (right) (blue) raw k1-weighted Fourier transform magnitude and (filled circles) result of the EXAFS fitting

Full size image
Table 6 Results of EXAFS fittings for end-of-run catalyst considers the presence of Hägg carbide
Full size table

4 Conclusions

For the first time, the K-edge of potassium was examined by XANES for investigating the changes occurring in the spectra of potassium promoter for an iron catalyst during the FTS reaction. CO-activated and end-of-run spectra for 100Fe:2K and 100Fe:5K catalysts presented peaks corresponding to electronic transitions and multiple scattering continuum states. The shift in the peak tentatively assigned to multiple scattering to higher energy was consistent with sintering of potassium promoter during the course of the reaction test. The spectra resembled that of dispersed potassium carbonate or potassium formate, although a related compound is also possible. The end-of-run catalyst retained its iron carbidic state, as demonstrated by XANES and EXAFS spectra at the iron K-edge; thus, removal of potassium due to sintering did not lead to deactivation by oxidation. However, sintering of potassium may exacerbate a deactivation mechanism involving pore blocking by excessive carbon formation at high K loading. The 100Fe:2K catalyst was previously found to produce excessive carbon in Mössbauer spectroscopy and TEM studies. Based on the CO TPR results and similar deactivation profile, the 100Fe:5K catalyst appears to have the same stability issue, which may be compounded by potassium sintering.