Characterization of CoCu- and CoMn-Based Catalysts for the Fischer–Tropsch Reaction Toward Chain-Lengthened Oxygenates
The need for a sustainable energy supply in the face of depleting oil reserves has reignited the importance of Fischer–Tropsch (FT) synthesis technology. Presently, the FT process is practiced at the industrial scale to predominately produce synthetic diesel-type fuels and lubricants. More recently, the possibility of hydrogenating CO toward oxygenates, and not just hydrocarbons, has been explored. We have developed a series of CoCuMn and CoMnK catalysts prepared via the oxalate co-precipitation route that are capable of forming oxygenates with desirable selectivity. Upon H2-assisted thermal decomposition of the resultant mixed metal Co1Cu1Mn1 oxalates, catalysts naturally exhibited a cobalt core–copper shell configuration with Mn5O8 dispersed throughout the catalyst nanoparticle as determined via Atom Probe Tomography (APT). We suggest structural changes are induced by the CO and H2 reactants to form the catalytically active phase under real-time reaction conditions as demonstrated by corroborative Density Functional Theory calculations and experimental evidence. APT studies also show that a Co4Mn1K0.1 catalyst post reaction contained a cobalt carbide phase as determined from a Co/C ratio of 2/1. Manganese and potassium were found only in the outermost part of the particle. Both catalysts were found to contain the presence of a Mn5O8 oxidic phase before and post reaction which we attribute to the high activity toward oxygenates of these two catalysts.
KeywordsFischer–Tropsch Oxygenates Co-based catalysts Atom Probe Tomography Density Functional Theory
In the breadth of the early 1920s, two scientists by the name of Franz Fischer and Hans Tropsch at the Kaiser Wilhelm Institut fϋr Kohlenforschung in Mϋlheim an der Ruhr, Germany, observed the catalytic reaction of H2 and CO to produce hydrocarbons and oxygenates . Different from the original patent by BASF (Badische Anilin und Soda-Fabrik) in 1913 , they realized that chain-lengthened products contain multiple –CH2– units and therefore need H2/CO ratios larger than one. With the depletion of natural oil reserves and the need for a sustainable energy supply, the Fischer–Tropsch (FT) synthesis has regained importance over the recent past. It is being practiced on the industrial scale to mainly produce diesel-type fuels and lubricants with straight-chained hydrocarbons .
Various types of catalysts have been developed to transform H2 and CO into oxygenated hydrocarbons. The most notable catalysts include Rh, MoS2 and modified FT CoCu ones . Ruthenium-based catalysts are favored for synthesis of ethanol and other C2+ oxygenates such as acetic acid and acetaldehyde [7, 8]. The rarity and expense of the element renders it incompatible with large scale FT applications. Molybdenum sulfide (MoS2) catalysts have found success in the formation of long chain alcohols [9, 10]. However, MoS2 catalysts typically need to be regenerated with sulfur which is not ideal considering sulfur typically poisons other catalysts used in downstream separation processes at refineries and other chemical plants. Here on, we focus on the development of modified FT CoCu and CoMn catalysts for the synthesis of long chain alcohols and aldehydes, respectively. Other reviews are available for alternate catalysts considered for higher alcohol formation [6, 11, 12, 13, 14, 15, 16, 17]. After a short introduction into “early times” research of higher alcohol formation through CO hydrogenation, we will mainly develop aspects of relevant structural and compositional properties of CoCu- and CoMn-based catalysts on the basis of experimental investigations using 3D Atom Probe Tomography (APT), Transmission Electron Microscopy (TEM) and theoretical calculations by Density Functional Theory (DFT).
2 Development of Modified CoCu and CoMn Catalysts for Oxygenates Formation
The concept of oxygenated hydrocarbons (“higher alcohols”) using FT Synthesis technology was first pioneered by the “Institut Français du Petrole” (IFP) in the late 1970s and early 1980s. IFP brought to light the possibility of manufacturing C1–C6 alcohols for use in motor fuels using CoCu catalysts . Designs of CoCu catalysts were adjusted so that higher alcohols were obtained under comparable operational conditions of pre-existing methanol synthesis processes while favoring desired alcohol length and selectivity . Typical IFP catalysts contained cobalt, copper, at least one metal oxide and an alkali metal promoter . Yet, catalyst preparation remained ambiguous and sometimes lacked reproducibility leaving much skepticism in the community up until now.
Within recent years, a series of CoCu- and CoMn-based catalysts were prepared via oxalate precursors (see Sect. 3.2) by Xiang and Kruse [20, 21]. Both long- and short-chained alcohols and aldehydes were found to form at favorable selectivities. The addition of metal oxides tuned the behavior of the CoCu catalysts in terms of WGS activity, CO conversion and selectivity toward alcohols and olefins . Most notably, long-chain alcohols were synthesized by a “CoCuMn” catalyst (atomic amounts of elements in this case: 1:1:1; however, other relative compositions were also investigated) containing a Mn promoter/dispersant in the form of Mn5O8 . In the absence of Cu, CoMn was found to form alcohols as well as aldehydes . Further addition with potassium, usually considered a “structural” promoter, resulted in a CoMnK catalyst that strongly favored selectivity toward aldehydes using feeds with low H2/CO ratios . Exchanging K for a Li promoter, i.e. employing CoMnLi catalysts, turned out to form appreciable amounts of olefins . Quite clearly, addition of metal oxide and alkali promoters are valuable tools for tuning selectivity to desirable oxygenates and olefins (the latter being interesting high-value products, too).
3 The Nature of CoCu- and CoMn-Based Catalysts
CoCu-based catalysts for higher alcohol synthesis via CO hydrogenation are frequently considered to combine both FT chain lengthening (Co) and WGS (Cu) properties. This (oversimplifying) picture is occasionally extended to associating Co metal to be responsible for CO dissociation and the Cu metal to keep the CO molecule intact. We note that while there is no doubt that a stepped Co surface can dissociate the C–O triple bond there is presently neither experimental nor theoretical proof that the chain lengthening mechanism to produce oxygenates involves C–C coupling of (partially hydrogenated) surface carbon atoms. Instead, a CO insertion mechanism may be a serious option to explain both chain growth and oxygen functionalization. One has to keep in mind that both metals show only limited miscibility, see Fig. 1. There is, however, no doubt about a synergy effect between these two elements resulting in a catalyst which produces chain-lengthened hydrocarbons with terminal alcohol functionality.
Alkali metal promoted CoMn- differ from CoCu-based catalysts by forming both alcohols and aldehydes. It has been suggested that the presence of an alkali metal, especially K, promotes the reconstruction of the active catalytic phase during the induction period of the reaction. Such reconstruction results in the formation of bulk Co2C dispersed by a Mn5O8 oxidic phase with Mn2+ and Mn4+ mixed valence states . The Mn5O8–Co2C interface is presently advocated as being essential for the formation of the active sites of oxygenates production. Similar conclusions were also drawn in other literatures when focused on the synthesis of short-chain olefins from FT synthesis . It should be emphasized, however, that the atomic configurations of “active sites” for oxygenate and/or olefin production are not known at present. Yet it seems they differ from those that are relevant for saturated hydrocarbon production.
3.1 Structural and Compositional Analysis of CoCu- and CoMn-Based Catalysts
A number of fundamental questions regarding the very nature of CoCu-based catalysts have been raised throughout the literature. Co is accepted to be responsible for hydrocarbon chain growth whereas the role of Cu is not yet clear . As seen in Fig. 1, the metallic phase of an activated CoCu-based catalyst (actually a “CoCuMn” hybrid), prepared according to the oxalate route (see Sect. 3.2) in our laboratory, is core–shell structured. APT clearly reveals the low miscibility of cobalt and copper giving rise to a cobalt core–copper shell (Co@Cu) structure (before reaction). Indeed, one would expect copper to be surface-enriched since it has a lower Gibbs free surface energy as compared to cobalt . The oxalate route of catalyst preparation obviously favors self-assembling of Co and Cu atoms in a Co@Cu configuration.
As to the details of Fig. 1, the occurrence of Cu in the core and, vice versa, of Co in the shell, are rather limited. Counting Cu atoms in the Co core leads to concentrations reflecting the thermodynamic maximum of 9% miscibility of Cu into Co . Conversely, Co atoms are also present in the Cu-dominated shell. Atoms of Mn are dispersed throughout in both the shell and core of the nanoparticle, occasionally appearing as precipitates. Atoms of O are found in the shell of the nanoparticle. This is expected since the CoCuMn catalyst must be passivated by oxygen chemisorption after thermal oxalate decomposition due to the pyrophoric nature of the material. Since Me–O (metal–oxygen) surface species may undergo preferential field evaporation during the initial stages of the APT analysis, we suspect that the actual amount of oxygen atoms appearing in Fig. 1 is underrepresented.
It is unlikely that the Co@Cu,Mn core–shell nature of the CoCuMn particle remains intact during the reaction. Even though an after-reaction APT analysis is not yet available, DFT and X-ray Photoelectron Spectroscopy (XPS) show a CO-induced restructuring in a CoCu model catalyst . Before demonstrating the result of swapping Cu for Co atoms in the presence of adsorbed CO, it is instructive to consider the experimental results obtained for inverse Cu@Co core–shell structures. The groups of Somorjai and Spivey both investigated synthetically manufactured Cu@Co core–shell nanoparticles [27, 28, 29]. Exposing such particles to both oxidizing and reducing atmospheres revealed the tendency for Cu to segregate to the surface in oxidizing atmospheres . Once the Co@Cu configuration was reached, it remains stable when exposing it to H2. This suggests that Co@Cu core–shell nanoparticles are stable under environmental and net vacuum conditions which is in agreement with the APT results presented above and the DFT calculations in the absence of adsorbates (see Fig. 2).
Based on the experimental literature, it can be posited quite reasonably that the presence of adsorbed CO will cause surface reconstruction of as-prepared CoCu catalysts. Specifically testing the surface prevalence of Co on the Cu/Co(0001) surface, Fig. 3 shows the extent to which this can be expected. The lowest DFT surface energies are found when 0.75 ML of CO (green triangles in Fig. 3) are adsorbed in mixed on-top and triple-bridged positions on a surface terminated with a 100% Co surface layer. In fact, at 0.5 and 0.75 ML of CO, the surface energy is rather flat from 25 to 100% surface Co enrichment. Remarkably, fairly low CO surface coverages cause Cu to swap against Co and that the amount of surface Co varies non-linearly with the CO coverage. Furthermore, it is quite interesting to state that the p(2 × 2) reconstructed surface is more stable at 0.5 ML of CO than the same surface at 0.25 ML of CO. In both cases, p(2 × 2) overlayer structures are formed with Co(CO)2 configurations in the former case and on-top CO positions in the latter case, respectively. Our theoretical results of a CO-induced reconstruction are also in agreement with earlier reports by Falo et al.  and Cerenco et al.  according to which CoCu nanoparticles suffer surface enrichment in Co and depletion in Cu under syngas exposure.
When turning to CoMn-based catalysts in the absence of Cu, less information is presently available about the virgin metallic phase. Therefore, the question for the possible existence of a core–shell structure as encountered for CoCu and CoCuMn catalysts has to be postponed. On the other hand, a common feature of Mn-containing catalysts, i.e. CoCuMn and CoMnK, is the simultaneous occurrence of a Mn5O8 phase in which Mn is present in both 2+ and 4+ valence states. Recently, we demonstrated by post-reaction TEM analysis that CoMnK catalysts, favoring long chain aldehydes at low CO/H2 ratios, contain a stable Mn5O8 phase (see Fig. 4) [21, 30]. Furthermore, cobalt went from a metallic phase before the reaction to a cobalt carbide (Co2C) phase after the reaction. This was somewhat unexpected considering earlier characterization studies clearly showed Co remained in a metallic state during the catalytic reaction . In the meantime, a reaction-induced Co2C phase has been reported and considered by several laboratories to contribute to alcohol formation in promoted Co and CoCu catalysts [21, 33, 34, 35, 36]. Other competing reports for CoCu catalysts argued that the presence of Co2C may inhibit oxygenate functionality [26, 36, 37]. Furthermore, Co2C nanoprisms have been supposed to be responsible for lower olefins production . The presence of Co2C has also been attributed to elevated methane formation . Regardless of competing literature, we demonstrate by XRD (Fig. 4) that CoMnK catalysts undergo a reaction-induced transformation into a catalyst containing a Co2C phase along with a Mn5O8 phase which are most probably in intimate contact so as to develop the active sites responsible for long-chain aldehyde/alcohol formation.
We now turn to a preliminary APT composition profile of a Co4Mn1K0.1 catalyst previously subjected to the FT reaction (see Fig. 5). The catalyst had to be passivated by oxygen chemisorption before subjecting it to APT. Because of the delicate sample preparation and the easy fracture of the samples under conditions of high electric fields in APT, individual catalyst nanoparticles were first deposited on a Si needle specimen and then covered by Cr (and Ni) serving as internal reference and field desorbing at low field strength (see Fig. 5a, b). The embedded nanoparticles were then shaped into a sharp APT tip (less than 100 nm) using a technique known as SEM–FIB (Scanning Electron Microscopy–Focus Ion Beam) developed by Thompson et al. . This technique was originally designed to prepare bulk phase specimens for APT analysis, rather than nanoparticles. Different from Fig. 1, which shows a cross-sectional atom map of a nanosized Co1Cu1Mn1 particle, we show here the concentration profiles for selected elements following layer-by-layer field evaporation of a Co1Mn1K0.1 particle (Fig. 5c). Obviously, when uncovering this particle through continuous field evaporation of the Cr reference (moving from right to left along the x-axis in Fig. 5c), cobalt (Co) and carbon (C) are mainly field evaporated until a Co/C = 2/1 ratio is reached for the bulk. This is in accordance with the occurrence of a Co2C phase. Furthermore, low concentrations of manganese and potassium are detected in the outer layers of the particle. Note that these elements are essentially absent in the Co2C bulk. From this, we would suggest that an originally core–shell structured particle transforms into Co2C during the reaction and maintains low concentrations of Mn and K in the outermost layers of the particle. Such a Co@Co,Mn,K core–shell structure would ensure that all elements are in intimate contact so as to provide the unique catalytic properties of CoMnK catalysts in the FT synthesis to chain-lengthened aldehydes/alcohols. Our present efforts concentrate on capturing and analyzing full catalyst particles (rather than a fraction of it, as in Fig. 5). These efforts include the need to elucidate the role of Mn5O8 structures as revealed by TEM.
3.2 CoCu- and CoMn-Based Catalysts via Oxalate Co-precipitation: Key to Self-Assembled Core–Shell Structures
Mixed-metal oxalate precursors as prepared in our laboratory contain no generic support material. Binary CoCu catalysts, despite proving constant steady-state activity and selectivity for more than 24 h time-on-stream under FT reaction conditions , may suffer long-term performance stability. For CoCuMn and CoMnK catalysts, no such loss in performance was ever found during more than 78 h time-on-stream. The remarkable performance stability of these catalysts is likely due to the generation of an oxide phase, here Mn5O8, during the thermal decomposition of the ternary oxalate precursors.
Non-supported CoCu catalysts derived from the decomposition of a mixed-metal CoCu oxalate, a type of metal–organic-framework (MOF) structure, naturally exhibit a nanosized Co@Cu core–shell configuration . Thus, a phase separation occurs during oxalate decomposition (see Fig. 6). It is important to note that this phase separation does not involve Co and Cu to nucleate and grow into separate metal particles. H2-assisted Temperature Programmed Decomposition (H2-TPDec) clearly shows that the decomposition temperatures for co-precipitated CoCu oxalates are very different from those of the individual metal oxalates, CoC2O4 and CuC2O4 . This is particularly true for mixed oxalates with Co/Cu atomic ratios larger than one (see Fig. 6). Conversely, for Co/Cu ratios < 1, the low temperature feature (265 °C) was observed to intensify. This can be straightforwardly correlated with the occurrence of CuC2O4 decomposition. Note that the Co analogue, if present, would have been expected to appear at temperatures of around 375 °C. Its absence in Fig. 6 provides further yet indirect proof of a CoCu oxalate co-precipitate containing both Co and Cu along with chelating oxalate groups, in a polymeric MOF structure. The fact that mainly CO2 is formed during TPDec demonstrates that the activated catalyst is metallic.
Consequently, considerable amounts of CO appear in the TPDec spectrum, the ratio of CO/CO2 thereby roughly being in agreement with the stoichiometry of the decomposition, if the atomic composition of the actual CoCuMn catalyst is taken into account.
In summary, the oxalate route of catalyst preparation has proven to be simple, fast and reliable. Characterization of as-prepared oxalates by TPDec or XPS allowed distinguishing between metallic and oxidic phase formation . Cutting-edge methods like APT and Scanning Transmission Electron Microscopy [(S)TEM] revealed the structural and compositional details of the catalysts before and after reaction. This is an important step toward an elucidation of the chemical nature of the catalytically active phase, a prerequisite for developing realistic mechanistic concepts of the FT reaction.
4 Concluding Remarks
Under optimal conditions, CoCuMn catalysts favor long chain alcohols whereas CoMnK catalysts favor long chain aldehydes. The role of Mn is at least twofold: it may act as a dispersant and as a promoter. It may coexist with other metals in core–shell metallic structures and as a Mn5O8 oxidic phase which we consider key to increased oxygenate functionality of the catalysts. CoMnK catalysts favoring higher aldehydes developed a Co2C phase during the reaction where the Mn5O8 promoter remained stable.
Cutting edge techniques, such as APT, demonstrated CoCuMn catalysts naturally exhibit a Co@Cu,Mn core–shell structure. STEM demonstrated concomitant Mn5O8 clusters remained highly dispersed throughout the catalyst so as to guarantee high specific surface areas during reaction. Both DFT and experimental XPS findings agree that CoCu catalysts undergo significant reconstruction under reactive conditions. Determining the catalytically active phase responsible for alcohol formation is vital for the targeted design of such catalysts and still remains a challenge today.
Comparatively, CoMnK catalysts favoring formation of oxygenates—mainly aldehydes—also underwent significant reconstruction. Cobalt in the CoMnK catalyst originally existed in a metallic state which was then transformed into a Co2C phase as determined by in-situ and post reaction XRD analysis. The APT post reaction analysis of this catalyst demonstrates a cobalt carbide phase to be formed along with a surface layer containing all three elements, Co, Mn and K, The APT method has therefore demonstrated its unique capability of providing chemical maps of atoms, with unprecedented lateral resolution, across the bulk of individual nanoparticles.
The oxalate co-precipitation route used in our laboratory allows the mixing of Co, Cu and Mn metals into a common polymeric MOF structure. This technique bypasses bulk miscibility limits by locking metal ions in neighboring positions via chelating oxalate ligands. No generic support material is needed when preparing catalysts through oxalate precursors. The necessary metal–metal oxide systems form through temperature-programmed decomposition of oxalate precursors. As we demonstrated, this unique synthesis route allows mixing of metal ions at the instant of precipitation thereby creating synergies between elements which may be difficult to achieve via more classical methods of catalyst preparation. We anticipate that the oxalate route of catalyst preparation will enjoy further applications in the future. The CoCuMn and CoMnK examples investigated here in relation to the synthesis of higher oxygenates are highly encouraging.
J.-S.M. and N.K. were supported by the National foundation under contract no. CBET-1438227. G.C. was supported by the National Science Foundation Graduate Student Fellowship Program under contract number 1347973. J.V. was also supported by the National Science Foundation Graduate Student Fellowship Program. G.C. and J.V, Seattle Chapter ARCS Fellows, gratefully acknowledge financial support from the Achievement Rewards for College Scientists foundation. J.V. acknowledges the support of the U.S. Department of Energy, Office of Science Graduate Student Research program administered by the Oak Ridge Institute for Science and Education (ORISE) which is managed by ORAU under contract number DE-SC0014664. A portion of the research was performed using EMSL, a national scientific user facility sponsored by the Department of Energy’s Office of Biological and Environmental Research and located at PNNL. PNNL is a multiprogram national laboratory operated for the U.S. DOE by Battelle. We finally thank Simon Ringer (University of Sydney) and his group for providing one of us (N.K.) with the possibility of performing APT measurements on CoCuMn catalysts.
- 1.Fischer F, Tropsch H (1926) Über Die Direkte Synthese von Erdöl- Kohlenwasserstoffen Bei Gewöhnlichem Druck. (Erste Mitteilung). Eur J Inorg Chem 59:830–831Google Scholar
- 2.BASF (1913) Und Soda-Fabrik, Ludwigshafen a. German patent DE 2,93,787Google Scholar
- 5.Dean JA (1972) Lange’s handbook of chemistry. McGraw Hill, New YorkGoogle Scholar
- 18.Sugier A, Freund E (1978) Process for manufacturing alcohols, particularly linear saturated primary alcohols, from synthesis gas. US patent 4122110Google Scholar
- 27.Beaumont SK, Alayoglu S, Pushkarev VV, Liu Z, Somorjai GA (2013) Exploring surface science and restructuring in reactive atmospheres of colloidally prepared bimetallic CuNi and CuCo nanoparticles on SiO2 in situ using ambient pressure X-ray photoelectron spectroscopy. RSC 162:31–44. https://doi.org/10.1039/c2fd20145c CrossRefGoogle Scholar
- 29.Musselwhite N, Brooks CJ, Marcus MA, Guo J, Liu Z, Kruse N, Somorjai GA (2013) Surface composition changes of redox stabilized bimetallic CoCu nanoparticles supported on silica under H2 and O2 atmospheres and during reaction between CO2 and H2: in situ X-ray spectroscopic characterization. J Phys Chem C 117:21803–21809. https://doi.org/10.1021/jp405745n CrossRefGoogle Scholar
- 32.Carenco S, Tuxen A, Chintapalli M, Pach E, Escudero C, Ewers TD, Jiang P, Borondics F, Thornton G, Alivisatos AP, Bluhm H, Guo J, Salmeron M (2013) Dealloying of cobalt from CuCo nanoparticles under syngas exposure. J Phys Chem C 117:6259–6266. https://doi.org/10.1021/jp4000297 CrossRefGoogle Scholar
- 33.Lebarbier VM, Mei D, Kim DH, Andersen A, Male JL, Holladay JE, Rousseau R, Wang Y (2011) Effects of La2O3 on the mixed higher alcohols synthesis from syngas over Co catalysts: a combined theoretical and experimental study. J Phys Chem C 115:17440–17451. https://doi.org/10.1021/jp204003q CrossRefGoogle Scholar