Fischer-Tropsch synthesis on cobalt catalysts: the effect of water

Edd Anders Blekkan, Øyvind Borg, Vidar Frøseth and Anders Holmen

DOI: 10.1039/b601307b

1. Introduction

Modern GTL (Gas-to-Liquids) technology involves Fischer-Tropsch synthesis for converting natural gas derived synthesis gas to transportation fuels. The Fischer -Tropsch synthesis (FTS) produces a complex mixture of hydrocarbons, consisting of methane, C2+ olefins and paraffins (linear and branched) and oxygenates (mainly alcohols). The product distribution is very dependent on the type of catalyst used and on the reaction conditions. The active catalysts for Fischer-Tropsch synthesis are Fe, Co and Ru. Fe has a high water–gas shift activity and is used when the synthesis gas is produced from coal, i.e. when the water–gas shift reaction is desirable due to low H2/CO ratios in this syngas. Supported cobalt is the preferred catalyst for the Fischer-Tropsch synthesis of long chain paraffins from natural gas due to their high activity and selectivity, low water–gas shift activity and comparatively low price. A key element in improved Fischer-Tropsch processes is the development of active catalysts with high wax selectivity.

The Fischer-Tropsch synthesis follows a polymerization mechanism where a C1 unit is added to the growing chain. A simplified representation of the reaction network is shown in Fig. 1, where the key points are termination by either H -abstraction to give α-olefins or by hydrogenation to give n-paraffins.

The main secondary reactions are hydrogenation and readsorption of primary olefins. However, the reaction network is very complex and involves a large number of reactions. Although the Fischer-Tropsch synthesis has been known since the 1920's the exact mechanism is still a matter of debate. Three mechanisms have been proposed based on different species as the monomer: In the original carbide mechanism proposed by Fischer and Tropsch, CHx is the proposed monomer, in the enol mechanism proposed by Storch et al. oxymethylene (HCOH) is the species responsible for chain growth and in the CO insertion mechanism proposed by Pichler and Schulz the chain growth occurs through the insertion of CO into the metal–methyl bond. The carbide mechanism has been favored for a long time, but a recent study points to the CO insertion route as a more likely mechanism.

Mass transfer effects are very important for the selectivity in the Fischer-Tropsch synthesis. Even though the reactants are in the gas phase, the catalyst pores will be filled with liquid products. Diffusion in the liquid phase is about 3 orders of magnitude slower than in the gas phase and even slow reactions may become diffusion limited. Diffusion limitations may occur through limitation on the arrival of CO to the active points or through the limited removal of reactive products.

In the cobalt-catalyzed Fischer-Tropsch reaction, oxygen is mainly rejected as water and this will generate high partial pressures of water at the reactor exit for fixed-bed reactors. As a consequence of extensive back mixing in slurry reactors, high water concentrations and low reactant concentrations will exist throughout the entire reactor. Water will therefore always be present in varying quantities during the reaction. It has been shown that water affects the activity as well as the selectivity for the Fischer-Tropsch synthesis, but different results have been reported for the effect of water. Even though water apparently influences the activity of various Co catalysts in different ways, water increases the C5+ selectivity and decreases the CH4 selectivity for all Co catalysts.

Different explanations have been proposed to explain this effect, including the influence of water on the adsorbed carbon species on the surface and the reduction of secondary hydrogenation of primary olefins by water, thereby facilitating olefin readsorption and chain initiation.

There are contradictory observations on the effect of water on the performance of Co catalyst using different supports, and an obvious common explanation of the observed effects is not available. The purpose of the present work is to compile and review the studies dealing with the effect of water on cobalt based FTS catalysts. As mentioned, water influences the FT-synthesis in many ways, and we summarize the literature in terms of 3 main areas; deactivation, activity (kinetics) and selectivity. Modern cobalt catalysts are supported, in order to obtain high cobalt surface areas and high catalyst activities. The supports commonly used for cobalt-based Fischer-Tropsch catalysts are metal oxides such as alumina, silica and titania, and the major technology providers offering catalysts or processes use these supports. There are also studies using other supports as well as unsupported cobalt. The different supports used have differing properties, in terms of chemical composition and purity, degree of interaction with the cobalt, and also physical properties (porosity, pore sizes). Here we summarize the results obtained on the various systems, showing that the effect of water is indeed dependent on the support used.

2. Deactivation

Deactivation is a common and important phenomenon in FTS. Deactivation effects of water are recorded on all commonly used supports. The suggested mechanisms include oxidation, sintering and solid state reactions rendering cobalt inactive.

2.1 Unsupported cobalt

Das et al. studied FTS over bulk cobalt in a CSTR, and observed some deactivation at high levels (>20%) of water addition. Bertole et al. studied bulk cobalt in a SSITKA experiment, and found no deactivation using a dry feed, but significant loss of activity when water was added to the feed. Most of this activity could be regained by a treatment in hydrogen (re-reduction), indicating that oxidation of cobalt is responsible for some of the deactivation, but a fraction of the activity was not recoverable this way, and this was attributed to sintering due to the high steam partial pressure.

2.2 Alumina

The influence of water on γ-Al2O3 supported cobalt catalysts has been the subject of several studies. Fig. 2a illustrates how a higher conversion leads to more rapid deactivation, whereas Fig. 2b shows a typical experiment illustrating the deactivation of a Co-Re/(γ-Al2O3 catalyst at simulated high conversions, obtained by adding steam to the feed. Schanke et al. found that the presence of large amounts of water suppressed the activity of un-promoted as well as Re-promoted cobalt supported on γ-Al2O3. From X-ray photoelectron spectra it was concluded that re-oxidation of surface cobalt atoms or highly dispersed cobalt phases, and not bulk cobalt oxidation, was responsible for the loss in activity.

Holmen and coworkers also observed a loss in activity when water was introduced to un-promoted and Re-promoted cobalt deposited on γ-Al2O3. In a recent paper similar results were reported for CoRe supported on both narrow-pore and wide-pore γ-Al2O3, and permanent deactivation was observed when the inlet ratio H2O:H2 was 0.7. The same group reported that the rhenium-promoted catalysts lost activity more rapidly than their un-promoted counterparts. According to a study from the same group using SSITKA (steady-state isotopic transient kinetic analysis) the deactivation is caused by the loss of active sites, whereas the site activity of the remaining sites is unchanged by the water treatment. The authors suggested that oxidation of exposed cobalt atoms is the main cause for the deactivation process. Hilmen et al. applied a range of characterization techniques including chemisorption, XPS, TPR, and gravimetry to study the possibility of oxidation of cobalt by water. The oxidation of the catalyst is suggested to be limited to the surface or to oxidation of highly dispersed parts of the cobalt. The promoted samples reoxidise easier, but the direct influence of the promoter is not clear: Either the promoter also promotes the direct reoxidation of cobalt metal, or the effect is linked with the larger fraction of the highly dispersed cobalt that is reduced, but also more easily oxidized since the particle size is small.

The deactivation of Pt-, Ru-promoted as well as un-promoted Co/γ-Al2O3 was investigated by Jacobs et al. The more active catalysts also deactivated faster. The catalysts were recovered from the reactor and examined using X-ray Absorption Spectroscopic (XAS) techniques. XANES (X-ray Absorption Near Edge Structure) analysis gave evidence of oxidation of a fraction of the cobalt clusters to cobalt oxide (Co3O4) or cobalt aluminate-like species by water produced during the reaction. Only the small clusters interacting with the support and clusters deviating from bulk -like behaviour were oxidized, and although the oxidized species are difficult to differentiate the authors favored the formation of a cobalt-aluminate species based on curve-fitting and comparison with model substances. However, EXAFS (Extended X-ray Absorption Fine Structure) results also strongly suggested that a large part of the deactivation was caused by sintering.

Sintering was also proposed as an explanation for the deactivation of a CoRe/γ -Al2O3 catalyst. EXAFS analysis of catalysts withdrawn from a FTS reactor after increasing time-on-stream (TOS) showed steadily increasing average Co–Co co-ordination numbers, indicating that the cobalt cluster size increased with TOS. The authors discussed the effect of the Re promoter in terms of facilitating the reduction of smaller cobalt crystallites with strong interactions with the support surface. These smaller crystallites are suggested to be prone to sintering due to a higher surface energy. Based on earlier XPS-results from Schanke et al., showing that such small clusters are oxidized by water during FTS, the authors speculate that a dynamic oxidation-reduction cyclic mechanism can cause this sintering.

Li et al. investigated the effect of water for a platinum-promoted Co/γ-Al2O3 catalyst during Fischer-Tropsch synthesis in a CSTR-type reactor. The catalyst lost activity in the presence of water, and it was found that small quantities of water (3–25 vol%) led to mild and reversible deactivation, whereas large amounts of water (> 28 vol%) deactivated the catalyst more severely and permanently. The deactivation was attributed to the formation of cobalt oxide or cobalt aluminate.

Jacobs et al. studied the effect of water on Co/γ-Al2O3 catalysts containing 15 and 25 wt% Co. Focus was on high water partial pressures leading to permanent loss of activity (above 25% water). The catalysts were characterized using EXAFS, and exhibited differences in the degree of cobalt interaction with the support surface. The 15 wt% Co/γ-Al2O3 catalyst consisted of cobalt clusters with an average size close to 6 nm, while the 25 wt% Co/γ-Al2O3 catalyst had clusters about twice the size (12 nm), as measured both by hydrogen chemisorption and confirmed by XRD line broadening. When exposed to FTS with added water, the 15 wt% Co/γ-Al2O3 catalyst deactivated strongly when the 25% water was added to the feed. At lower levels of water addition (3–21%) the activity was inhibited (see below), but the activity was regained when the water was removed from the feed. The 25 wt% Co/γ-Al2O3<.sub> catalysts showed little effect of adding up to 20% water in the feed. Adding 25% water led to some irreversible activity loss, but much less than that observed with the 15% Co catalyst. When studied by EXAFS, these catalysts showed different changes in the structure. The derivative XANES spectra of the 15 wt% Co/g-Al2O3 catalyst was consistent with cobalt aluminate (CoAl2O4) formation (peak at 7717 eV), however, the interpretation of these spectra is difficult. A similar result was reported by the same group for a Pt-promoted 15 wt% Co/γ-Al2O3. The authors argue that the Pt promotes reduction of small clusters with strong support interaction, and that cobalt aluminate formation is favored by small cobalt clusters. The 25% Co/γ-Al2O3 catalyst behaved differently, and the XANES-spectra showed evidence of CoO formation due to oxidation by water. This was reversible, and reverted to Co metal after removal of the added water. Hence, the results indicate that the behaviour of the Co is linked to particle size, small particles form cobalt aluminate, whereas larger cobalt particles do not, they oxidize to form reducible CoO.

Goodwin and coworkers have studied the formation of cobalt aluminates. Jongsomjit et al. studied cobalt aluminate formation by TPR and RAMAN spectroscopy. The catalysts investigated were 25% Co/ γ-Al2O3 and Ru-promoted (0.5%)– 25% Co/γ-Al2O3. The Raman spectra of the samples after various pretreatments are shown in Fig. 3.

The Raman spectra of water-treated catalysts (Co-RWP and Co-Ru-RWP), which are reduced in hydrogen with 3% water present and subsequently passivated show broad bands between 400 and 750 cm-1, clearly different from the cobalt aluminate spinel, and the bands cannot be attributed to the support or cobalt metal or oxides (CoO or Co3O4). The authors suggest that these bands are due to non-stoichiometric surface Co alumina compounds, and support this speculation with TPR results, where the water-treated catalysts show larger peaks at higher reduction temperatures. In a separate study using TPR it was suggested that the hydrated alumina surface stabilizes cobalt oxidic species in strong interaction with the support. Dried samples form less nonreducible cobalt oxides, and it is therefore important to control the water vapor pressure.

Additional evidence of the importance of the amount of water (partial pressure) during the Fischer-Tropsch synthesis on the deactivation was provided by van Berge et al. Using Mössbauer emission spectroscopy (MES) they showed that the degree of oxidation of the γ-Al2O3 supported catalyst depended on the [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] ratio. Formation of both reducible and less reducible cobalt oxide was observed, and the relative ratio between these species was affected by the severity of the oxidation conditions. The catalysts investigated were promoted by Pt. These experiments were done at atmospheric pressure. The same group later reported further studies applying MES to alumina-supported catalysts, and also reported the design of a dedicated high-pressure cell. Applying the MES technique is challenging, since the catalyst has to contain a fraction of the radioactive isotope 57Co. Through model studies using 57Fe they showed that the isotope can be added to an existing catalyst (in this case a cobalt catalyst) by a simple incipient wetness impregnation followed by calcination and reduction. Investigations of the state of Co in a 15% Co/Al2O3 catalyst after treatment in a gas containing H2O:H2 in the ratio 1:1 showed that at a constant temperature (423 K) the degree of Co reduction increased with increasing total pressure from 1–10 bar, indicating that the bulk oxidation of cobalt is not possible at these conditions.

2.3 Silica

Several groups have reported deactivation of silica-supported cobalt catalysts. Holmen and coworkers have reported increased deactivation due to added (external) water in the feed to silica-supported Co catalysts. Kogelbauer et al. reported the formation of silicates. Catalysts recovered from FTS as well as catalysts deactivated by steam-treatment both showed fractions of non-reducible cobalt in TPR. The presence of metallic cobalt was a prerequisite for the silicate formation.

Bartholomew and coworkers described deactivation of cobalt catalysts supported on fumed silica and on silica gel. Rapid deactivation was linked with high conversions, and the activity was not recovered by oxidation and rereduction of the catalysts, indicating that carbon deposition was not responsible for the loss of activity. Based on characterization of catalysts used in the FTS and steam-treated catalysts and supports the authors propose that the deactivation is due to support sintering in steam (loss of surface area and increased pore diameter) as well as loss of cobalt metal surface area. The mechanism of the latter is suggested to be due to the formation of cobalt silicates or encapsulation of the cobalt metal by the collapsing support.

Chen et al. studied the deactivation of ZrO2-promoted cobalt on silica in the FTS using a fixed bed reactor at realistic reaction conditions and high conversions (65–90%). A fraction of the lost activity was regained by a rereduction, and the selectivity was reported to be unaffected by deactivation and regeneration. The authors identified hydrated cobalt silicates using IR and XRD analyses, and suggested that these species caused the permanent deactivation.