Elsevier

Journal of Catalysis

Volume 299, March 2013, Pages 67-80
Journal of Catalysis

Structure sensitivity of the Fischer–Tropsch activity and selectivity on alumina supported cobalt catalysts

https://doi.org/10.1016/j.jcat.2012.11.013Get rights and content

Abstract

Identifying the active site on the surface of a heterogeneous catalyst is one of the biggest challenges in the field of catalysis research. Especially, in the case of structure sensitive and heterogeneously catalyzed reactions, the knowledge of the active site/ensemble would result in a great advantage in the quest to design tailored catalyst displaying the desired activity and selectivity. In the Fischer–Tropsch synthesis, the multitude of reaction products as well as the large number different reaction pathways does result in additional difficulties in the search for the active site/ensemble. In the here presented work, we were able to conduct a thorough study of the CO hydrogenation reactions on nano-sized alumina supported cobalt crystallite model catalysts. By evaluating the full product spectrum, it was possible to deconvolute the structure sensitivity of the various reactions and to gain further insight into the nature of the present reaction mechanisms. It was therefore possible to measure decreasing carbon monoxide turn over frequency with decreasing cobalt crystallite site, paralleled with an increased selectivity toward methane and branched hydrocarbons at a constant olefin selectivity. Although these trends were observed to be independent of time on stream, the activity did change drastically upon the initial exposure to reaction conditions. CO-TPD studies provided direct evidence for the observed size dependencies.

Graphical abstract

A thorough study of the CO hydrogenation reactions on nano-sized alumina supported cobalt crystallite model catalysts was conducted. By evaluating the full product spectrum, it was possible to deconvolute the structure sensitivity of the various reactions and to gain further insight into the nature of the present reaction mechanisms.

  1. Download : Download high-res image (75KB)
  2. Download : Download full-size image

Highlights

► Full catalyst characterization before and after FT experimentation. ► Time resolved alumina supported cobalt structure sensitivity. ► Full selectivity and activity analysis. ► Supporting TPD experimentation.

Introduction

In the field of heterogeneous catalysis, especially in the case of catalysts which contain expensive metals, the optimization of the catalytic system in order to obtain a maximum mass specific activity in conjunction with the desired product selectivity is of utmost importance to reduce the running costs and ultimately the cost of the final product and profitability of a process. For catalytic reactions, one approach to optimize the reactivity per mass unit of catalyst is to increase the dispersion of the active material, that is, decrease the crystallite size, in a bid to maximize the active surface area. However, not in all catalytic reactions, a decrease in catalyst crystallite size is reported to result in an increase in activity. The most prominent reactions displaying this so-called structure sensitivity are the carbon monoxide oxidation over nano-sized gold crystallites [1] and the Fischer–Tropsch synthesis (FTS).

In 1984, Boudart and McDonald proposed that a structure sensitive reaction seems to require a so-called multiple atom site or ensemble to show activity [2]. When comparing iron crystallites of average sizes between 1 and 17 nm in their FTS experiments, they showed that the turnover rate for methane formation increased slightly with increasing crystallite size, while the selectivity for C2+ hydrocarbons decreased slightly. The differences in crystallite size were achieved through different metal loadings of the catalyst, varying from 0.5 to 2.8 wt.%. The active phase was deposited onto the support by means of pH-controlled precipitation and impregnation. The authors noted that the observed changes were much smaller than typical effects of promoters, and they commented that changes in the active phase including carbide formation and sintering may have played a role.

In the same year, Reuel and Bartholomew published a study on different supported cobalt catalysts and the effect of the crystallite size and the support on the FT activity and selectivity [3]. Similar to Boudart and McDonald [2], they used pH-controlled precipitation, impregnation, and different metal loadings in their catalyst preparation. The results of their experiments showed that the activity of a catalyst and the turn over frequency (TOF), that is, metal surface specific activity, increases with increasing loading or, respectively, with increasing average crystallite size. Furthermore, the average hydrocarbon chain length increased with increasing average crystallite size. In a later publication, Bartholomew et al. proposed that their measured crystallite size effects on the CO conversion may have been clouded by effects of varied degrees of reduction in the cobalt phase in the different catalyst samples due to metal-support interaction [4].

In 1990, Ho and co-workers tried to isolate the effect of the cobalt crystallite size from other variables [5]. For this purpose, they used silica as support, to minimize metal-support interactions, and completely reduced catalyst samples. The catalysts were prepared by incipient wetness impregnation with metal loadings from 1 to 10 wt.%, and average cobalt crystallite sizes of 4–10 nm were obtained. In this range, the experimental results showed no change of the TOF.

Iglesia and co-workers stated in 1992 that in a range of average cobalt crystallite sizes between 10 and 210 nm, no change in the TOF could be observed [6]. Earlier reported changes in TOF or in selectivity were attributed either to incomplete reduction in cobalt or to the reaction conditions at which those experiments were carried out which would favor the production of lighter products. According to Iglesia et al. at industrial relevant conditions, effects of cobalt crystallite size would not play a role in catalyst performance while also stating that cobalt crystallites smaller than 5–6 nm might oxidize at the water rich conditions at high conversion [7]. The possibility of oxidation was later confirmed to be feasible for crystallites smaller than 4–5 nm in a thermodynamic analysis [8], [9], [10].

Recent publications from the group of de Jong [11], [12], [13] described a different observation than the one reported by Iglesia [7]. In order to avoid metal-support interactions, the cobalt particles were loaded onto carbon nano-fibers. The loading process was carried out by means of incipient wetness impregnation, ion adsorption, and homogeneous deposition precipitation. The cobalt loading varied between 0.8 and 22 wt.%, while the average cobalt crystallite sizes varied between 3 and 30 nm. Comparing the cobalt phase before and after the reaction, it could be stated that no oxidation (neither bulk nor surface) of small crystallites had taken place as assumed by Iglesia [7]. Also, no crystallite agglomeration or sintering could be measured in post-run catalyst analyses. The experiments were carried out at two different reaction conditions with pressures/temperatures of 1 bar/220 °C and 35 bar/250 °C. In both cases, the results were very similar. While above a certain cobalt crystallite size (6 nm for 1 bar and 8 nm for 35 bar), the TOF remained unchanged, a sharp decrease in activity below these crystallite sizes was found. Furthermore, it was reported that the methane selectivity increases with smaller crystallite size as well as the paraffin to olefin ratio, which indicates increased hydrogenation activity on these crystallites. A study by Prieto et al. [14] also reported a decreasing trend of activity with decreasing cobalt crystallite sizes. However, the results obtained for cobalt crystallite sizes below 6.8 nm might have been clouded by the formation of strong interactions between the cobalt species and the silylated ITQ-2 delaminated zeolite chosen as support. The selectivity trends observed in this study are however inverse to previous results, that is, decreasing methane selectivity and increasing C5+ selectivity with decreasing crystallite size. No explanation for this inverse trend was given by the authors, and no further selectivities (for olefin, branched hydrocarbons and/or oxygenates) were reported.

Similar trends have recently also been described for other FT active metals. Mabaso [15], Mabaso et al. [16] and Barkhuizen [17] reported a decrease in the turn over frequency for supported iron crystallites smaller than 6–8 nm. The model catalysts in this work were prepared by means of a novel reverse micelle precipitation method, and crystallites with narrowly distributed sizes (between 2 and 16 nm on average) were prepared and deposited onto carbon and alumina supports. Similar to the study by Bezemer et al. [11], increased methane selectivity was reported for smaller crystallites, while olefin to paraffin ratios remained apparently unaffected. It was noted that, although similar activity and selectivity patterns were observed over time on stream (i.e., at initial stages of testing and after several days), the results may have been affected by changes of iron phases and iron crystallite size (sintering), and no full post-run catalyst characterization was reported. It was further noted that effects of unoptimized potassium promotion may have clouded their results. A follow-up study by Cheang [18] has however conclusively shown that the effects reported by Mabaso [15] are indeed due to iron crystallite size, but they can further be influenced by water levels during the reaction.

Detailed studies on alumina supported ruthenium crystallites with average sizes between 2 and 11 nm by Welker [19], Welker et al. [20] and Barkhuizen et al. [17] also showed an increase in specific catalyst activity (TOF) with a simultaneous decrease of methane selectivity for increasing crystallite sizes. In these studies, the catalysts were prepared via an impregnation method using reverse micelle systems and detailed pre- and post-run catalyst characterization showed that effects of sintering did not impact on the results obtained. Effects of oxidation by water, which generally are considered unlikely on this noble metal, could also be excluded as some of the tests were done at very low synthesis gas conversion levels (<0.1%). Like in the studies by Bezemer et al. [11], Welker [19] and Welker et al. [20] also observed decreasing olefin contents with smaller crystallite size. They also studied the formation of oxygenates and branched compounds, but no clear trends with respect to crystallite size could be found. It was noted that activity and selectivity can change dramatically over time on stream, in particular in the very first minutes of catalyst testing. It was concluded that, since the crystallite sizes had not changed significantly during testing, surface reconstruction of the crystallites may have occurred upon exposure to synthesis gas, which are believed to cause dramatic performance changes of the model catalysts [11], [21]. In an attempt to understand site requirements for the formation of FT products, Welker [19] and Welker et al. [20] also used organo-metallic ruthenium complexes of varied nuclearity (2–6 metal atoms). It was shown that these compounds do indeed show C–C bond formation and C–O bond cleavage. All clusters showed high methane selectivity, while the overall activity was dependent on the nuclearity, and the different ligands used. In line with the trends observed for crystallites with a decreasing number of Ru atoms in the cluster, the TOF also decreased. It could not be proven whether gas phase CO was incorporated in the produced hydrocarbons.

In a study by Ojeda et al. [22], rhodium nano-crystallites of different size (<5–30 nm on average) supported on alumina have been tested in CO hydrogenation. Again, a reverse micelle method was applied in which the metal crystallites were prepared via reduction of the metal precursor followed by impregnation onto the support. Rhodium shows some FT activity [23], although generally shorter chains and higher oxygenate selectivities are obtained due to the lower tendency of this metal to promote C–O bond cleavage. Again, smaller crystallites displayed lower specific activity and higher methane selectivity (60 C% for crystallites with a diameter of <5 nm compared to 42 C% for crystallites with a diameter of 30 nm). Interestingly, greatly enhanced selectivity of valuable oxygenates was observed on the smaller crystallites (26 C% for crystallites with a diameter of <5 nm compared to 17 C% for crystallites with a diameter of 30 nm). It was shown that enhanced metal-support interaction of small crystallites lead to partially oxidized rhodium atoms which may be responsible for increased oxygenates formation. Alternatively, it has been stated by Schulz [24] that sites of low coordination, which should preferably be present on corners and edges of small crystallites, might be responsible for CO insertion-type reactions [25], [26] leading to oxygen containing products. In particular, metals like rhodium and cobalt are known to promote such reactions, and homogeneous complexes of these metals are effective catalysts, for example, olefin hydroformylation [27].

Most published studies which report a change of TOF and selectivity with changing metal crystallite size are consistent regarding the resulting effects. With decreasing metal crystallite size, the selectivity of methane increases with a concurrent decrease in the selectivity for higher hydrocarbons. In all these cases, the TOF is lower for the smallest crystallites tested compared to the larger ones. The only exception is the publication by Boudart and McDonald [2] who report a decrease in the selectivity of higher hydrocarbons with increasing crystallite size. This might be due to the different definition of higher hydrocarbons as Boudart and McDonald [2] define C2+ selectivities while others referred to C5+ selectivities. Further differences might be the result of crystallite size and phase changes which were not detected due to incomplete catalyst characterization.

The exact origin of these observed size effects is still under debate. Three main effects have been proposed:

  • 1.

    Structural effect: At different metal crystallite sizes different surface sites are exposed leading to different activities and selectivities [2], [28], [29], [30], [31].

  • 2.

    Electronic effect: At different crystallite sizes the activation energy for the dissociation of hydrogen is favored over that of carbon monoxide resulting in changes in selectivity [11], [13], [19].

  • 3.

    Oxidation effect: Smaller crystallites are easily re-oxidized under Fischer–Tropsch conditions leading to changes in activity due to a loss in active material [7], [8], [9], [10].

In the present study, we attempt to obtain novel insights into the activity and selectivity dependence on the crystallite size in the case of the cobalt catalyzed Fischer–Tropsch synthesis. A set of alumina supported cobalt model catalysts (cobalt crystallite size 2–10 nm) is prepared using a newly developed, reverse micelle based preparation technique [32] and fully characterized in the oxidic, reduced and spent, that is, after exposure to FT synthesis conditions, state. By targeting low carbon monoxide conversions (<10%), the possibility of re-oxidation of nano-sized cobalt crystallites upon exposure to product water as suggested in literature [8], [10] can be excluded. Compared to previous studies, the full characterization of the catalyst at various stages of its life as well as the high temporal resolution of the activity and selectivity measurements during the FTS allows us a more in-depth view into the crystallite size effect as well as its changes under exposure to Fischer–Tropsch conditions (here reaction temperature of 190 °C, synthesis gas pressure of 9.9 bar, H2/CO ratio of 2 and a space velocity of 7.2 mlsynthesis gas/gcatalyst min (STP)).

Section snippets

Catalyst preparation

A series of alumina supported cobalt model catalyst were prepared for this study. Exact preparation details were published previously [32]. In short, nano-sized Co3O4 are prepared via calcination of the precipitate obtained from the reaction of ammonia solution with cobalt nitrate containing reverse micelles (for compositions see Table 1). The size of the obtained cobalt oxide crystallites can be targeted by tuning the size of the reverse micelles through changes in the water to surfactant (in

Characterization of the spent model catalysts

The model catalysts are tested for 25 h under mild Fischer–Tropsch conditions. The size and nature of the crystallites in the model catalysts at the initial stages of the reaction, that is, the freshly reduced cobalt crystallites, have previously been characterized (see Table 2) [32] and can be correlated to the initial activity and selectivity values obtained. To study the changes of crystallite size and cobalt phase, the catalysts are cooled to room temperature under argon flow and passivated

Conclusions

Nine alumina supported nano-sized cobalt catalysts were fully characterized and studied for their activity and selectivity for the Fischer–Tropsch synthesis. Several previously on Co, Fe, Ru, and Rh reported trends with regard to changing crystallite size could be confirmed and novel insights into the FT reaction mechanism on nano-sized metal crystallites could be obtained.

  • 1.

    With increasing crystallite size, the surface specific activity (TOF) increases without reaching a maximum/leveling off in

References (79)

  • M. Haruta

    Catal. Today

    (1997)
  • R.C. Reuel et al.

    J. Catal.

    (1984)
  • E. Iglesia et al.

    J. Catal.

    (1992)
  • E. Iglesia

    App. Catal. A: General

    (1997)
  • J. van de Loosdrecht et al.

    Catal. Today

    (2007)
  • G.L. Bezemer et al.

    J. Catal.

    (2006)
  • G. Prieto et al.

    J. Catal.

    (2009)
  • H. Schulz et al.

    Catal. Today

    (2002)
  • M. Ojeda et al.

    Appl. Catal. A: General

    (2004)
  • M.A. Vannice

    J. Catal.

    (1977)
  • H. Schulz

    Top. Catal.

    (2003)
  • M. Claeys et al.

    Surf. Sci. Catal.

    (2004)
  • M. Claeys et al.

    Surf. Sci. Catal.

    (2000)
  • R. van Hardeveld et al.

    Surf. Sci.

    (1969)
  • N. Fischer et al.

    Catal. Today

    (2011)
  • I. Lisiecki

    Coll. Surf. A

    (2004)
  • M. Boutonnet et al.

    Coll. Surf.

    (1982)
  • M. Boutonnet et al.

    Curr. Op. Collid. Interface Sci.

    (2008)
  • S. Eriksson et al.

    Appl. Catal. A: General

    (2004)
  • J.J. Mortensen et al.

    J. Catal.

    (1999)
  • C. Klink et al.

    Surf. Sci.

    (1996)
  • J.E. Kirsch et al.

    Surf. Sci.

    (2003)
  • I.M. Ciobîca et al.

    Surf. Sci.

    (2008)
  • A.M. Saib et al.

    Catal. Today

    (2010)
  • J.A. Moulijn et al.

    Appl. Catal. A: General

    (2001)
  • E. van Steen et al.

    Appl. Catal. A: General

    (1999)
  • P. Biloen et al.

    Adv. Catal.

    (1981)
  • H. Schulz et al.

    Stud. Surf. Sci. Catal.

    (1997)
  • H. Schulz et al.

    Stud. Surf. Sci. Catal.

    (1994)
  • M.E. Dry et al.

    Stud. Surf. Sci. Catal.

    (2004)
  • J.J.C. Geerlings et al.

    Appl. Catal. A: General

    (1999)
  • E. Iglesia et al.

    J. Catal.

    (1991)
  • J. Gaube et al.

    J. Mol. Catal. A: Chem.

    (2008)
  • H. Schulz et al.

    Appl. Catal. A

    (1999)
  • P. Biloen et al.

    J. Catal.

    (1977)
  • T.A. Dorling et al.

    J. Catal.

    (1969)
  • H. Schulz et al.

    Fuels Proc. Technol.

    (1988)
  • M. Boudart et al.

    J. Phys. Chem.

    (1984)
  • C.H. Bartholomew et al.

    In Eleventh North American Meeting of the Catalysis Society

    (1989)
  • Cited by (0)

    View full text