1 Introduction

Supported cobalt remains the favoured catalyst material for the synthesis of long-chain hydrocarbons from synthesis gas prepared from natural gas because of attractive features such as high activity, high selectivity for linear paraffins, low water–gas shift activity, and relatively low price compared to noble metals.

Since oxygen mainly is rejected as water in cobalt-catalysed Fischer–Tropsch synthesis (FTS), water will be present in varying quantities during the reaction. The influence of water on cobalt supported on common metal oxides such as α-Al2O3, γ-Al2O3, SiO2, and TiO2 has been extensively studied in literature. Water, either produced by the reaction or externally added, is known to decrease or to increase the activity. A unified explanation of these seemingly contradictory results is still not available. However, in all cases water (either co-fed or produced by the reaction) increases the C5+ selectivity. From this it follows that for the Fischer-Tropsch reaction the C5+ selectivity increases with increasing conversion. The magnitude of the effect of water on the C5+ selectivity depends on the catalyst [1]. Minderhoud et al. [2] and Kim [3] were among the first to describe the effect of water on the catalyst performance. The effect of water is well documented and described in several recent reviews [4–8].

The carbon nanofiber (CNF) supported cobalt catalysts have shown promising performance in FTS [9–11]. The CNF supports are interesting for several reasons such as chemical inertness, high purities, high mechanical strengths, high surface areas, high thermal stability, and tunable bulk density [12]. The present contribution deals with three different CNF supported cobalt catalysts which have been exposed to increased water concentration during the FTS reaction. The catalysts were prepared either by incipient wetness or wet impregnation and the CNF structure was either of the platelet or fishbone type.

2 Experimental

2.1 Catalyst Preparation

Platelet and fishbone CNF were prepared from CO decomposition on a Fe3O4 catalyst and CH4 decomposition on a Ni/Al2O3 catalyst, respectively. Details about the platelet fiber preparation have been published previously [13]. After growth, the CNFs were boiled in an HNO3 solution for 1 or 3 h in order to introduce surface oxygen groups and remove remaining catalyst particles.

Two catalysts containing 20 wt% cobalt were prepared by one-step incipient wetness impregnation of purified platelet and fishbone CNF with aqueous solutions of cobalt nitrate hexahydrate. One catalyst containing 12 wt% cobalt was prepared by wet impregnation of purified platelet CNF with a toluene/ethanol solution of cobalt nitrate hexahydrate. The solvent was evaporated under vacuum at room temperature.

All catalysts were dried at 393 K overnight after impregnation. Calcination at 573 K for 6 h in flowing nitrogen completed the preparation process. The temperature was increased by 2 K/min from ambient temperature to 573 K. Further pre-treatment was done in situ.

In the following, the general catalyst nomenclature WWCo-XYZ will be used. WW is the cobalt loading (WW = 12 for 12 wt% and WW = 20 for 20 wt %), X is the CNF structure (X = P for platelet and X = F for fishbone), Y is the impregnation method (Y = I for incipient wetness impregnation and Y = W for wet impregnation), and Z is the boiling time in the HNO3 solution during purification (Z = 1 for 1 h and Z = 3 for 3 h).

2.2 Support and Catalyst Characterisation

2.2.1 X-ray Diffraction

X-ray diffraction patterns were recorded at room temperature on a Siemens D5005 X-ray diffractometer using CuKα radiation. The scans were recorded in the 2θ range between 20° and 80° using a step size of 0.03°. The samples were ground to fine powders prior to measurement.

The average Co3O4 crystallite thickness was calculated from the Scherrer equation using the (311) peak located at 2θ = 36.9°. A K factor of 0.89 was used in the Scherrer formula. Lanthanum hexaboride was used as reference material to determine the instrumental line broadening. The average spherical Co3O4 particle size was calculated by multiplying the crystallite thickness by a factor of 4/3 [14].

2.2.2 Nitrogen Adsorption/Desorption

Nitrogen adsorption–desorption isotherms were measured on a Micromeritics TriStar 3000 instrument, and the data were collected at liquid nitrogen temperature, 77 K. The samples were outgassed at 523 K overnight prior to measurement.

The surface area was calculated from the Brunauer–Emmett–Teller (BET) equation while the total pore volume was calculated from the nitrogen desorption branch applying the Barrett–Joyner–Halenda method [15].

2.2.3 Hydrogen Chemisorption

Hydrogen adsorption isotherms were recorded on a Micromeritics ASAP 2010 unit at 313 K. The samples were evacuated at 313 K for 1 h, and then reduced in flowing hydrogen at 573 K for 16 h. The temperature was increased at 2 K/min from 313 to 573 K. After reduction, the sample was evacuated for 1 h at the reduction temperature and for 30 min at 373 K, before subsequently cooling it to 313 K. The adsorption isotherm was recorded at this temperature, in the pressure interval ranging from 15 to 500 mmHg. The amount of chemisorbed hydrogen was determined by extrapolating the straight-line portion of the isotherm to zero pressure. Furthermore, in order to calculate the dispersion, it was assumed that two cobalt surface atoms were covered by one hydrogen molecule [16].

The cobalt metal particle size was calculated from the cobalt metal dispersion by assuming spherical, uniform cobalt metal particles with site density of 14.6 at./nm2. These assumptions give the following formula [16]:

$$ d\left( {{\text{Co}}^{0} } \right) ( {\text{nm)}} = \frac{96}{D(\% )} $$
(1)

2.3 Activity and Selectivity Measurements

Fischer–Tropsch synthesis was performed in a fixed-bed reactor (stainless steel, 10 mm inner diameter) at 483 K, 20 bar, and H2/CO = 2.1. The samples were diluted with inert silicon carbide particles in order to improve the temperature distribution along the catalytic zone.

The samples were reduced in situ in hydrogen at 1 bar while the temperature was increased at 1 K/min to 573 K. After 16 h of reduction, the catalysts were cooled to 443 K. The system was then pressurised to 20 bar and synthesis gas of molar ratio H2/CO = 2.1 (and 3 % N2 as internal standard) was introduced to the reactor. To avoid run-away and catalyst deactivation at start-up, the temperature was increased slowly to the reaction temperature of 483 K. Water vapour was introduced to the reactor by passing deionised liquid water through an electrical vaporiser kept at 573 K. The total pressure and the flow rate of synthesis gas were kept constant during external water addition. Thus, the reactant partial pressures were reduced as water was introduced to the reactor.

The runs were divided into five periods, each of 26 h. The following conditions were used:

  • Period 1: Synthesis gas at flow rate 250 ml/min (F 1).

  • Period 2: Synthesis gas at reduced synthesis gas space velocity (F 2) to give initially 50 % CO conversion.

  • Period 3: Synthesis gas at flow rate F 2 and 20 mol% water vapour at reactor inlet.

  • Period 4: Synthesis gas at flow rate F 2 and 33 mol% water vapour at reactor inlet.

  • Period 5: Synthesis gas at flow rate F 2.

We have previously studied the role of water on the FTS performance of cobalt supported on metal oxides such as γ-Al2O3, SiO2, and TiO2 [1, 17, 18]. In order to have a reliable comparison, the same water concentrations in the reactor inlet were used for the catalysts in the present work (Period 3 and 4).

20 and 33 mol% added water correspond to a larger amount than would be produced for a total conversion of CO. Thus, the catalyst is being subjected to a much higher water partial pressure than would be encountered during normal synthesis. The water partial pressure given in this work was calculated as the average value between the reactor inlet and outlet.

Heavy hydrocarbons were collected in a heated trap (363 K) and liquid products were removed in a cold trap (298 K). The effluent gaseous product was analysed for hydrogen, nitrogen, carbon monoxide, carbon dioxide, water, and C1–C9 hydrocarbons using an HP5890 gas chromatograph equipped with a thermal conductivity detector (TCD) and a flame ionisation detector (FID).

The activity is reported as the hydrocarbon formation rate [g/(gcat·h)]. The C5+ selectivity was calculated by subtracting the amount of C1–C4 hydrocarbons and CO2 in the product gas mixture from the total mass balance.

3 Results and Discussion

3.1 Support and Catalyst Characterisation

X-ray diffraction patterns of the catalysts confirmed the presence of graphite and Co3O4. Thus, Co3O4 was the only crystalline cobalt species after calcination in nitrogen. The only difference between the diffractograms of the various catalysts was the line width of the Co3O4 peaks and, therefore, the crystallite size. The results are given in Table 1. Wet impregnation gave smaller cobalt particle sizes than the incipient wetness impregnation method. We attribute this difference to the more favourable wetting properties of the mixed toluene/alcohol (wet impregnation) compared to the aqueous solution of cobalt nitrate hexahydrate (incipient wetness impregnation) instead of the low cobalt loading for 12Co-PW3. In fact, previous results [10] have shown that two catalysts containing 12 wt% cobalt prepared by incipient wetness impregnation had larger particle sizes than 12Co-PW3. For the catalysts prepared by incipient wetness impregnation, slightly larger particles were present in catalyst 20Co-FI1 than in 20Co-PI1, most likely because the platelet CNF provided more edge sites for anchoring of cobalt particles compared to the fishbone type. It should also be mentioned that the particle sizes correlated with the inverse surface area of the CNF support (Tables 1, 2).

Table 1 Cobalt particle size data
Full size table
Table 2 Nitrogen sorption properties of the catalysts used
Full size table

3.2 Fischer–Tropsch Synthesis

Fischer–Tropsch synthesis was run at 483 K, 20 bar, and H2/CO = 2.1. The effect of water was evaluated at dry inlet, at two different water addition levels (20 and 33 mol%), and finally after removal of the co-fed water. Since the catalysts deactivated with time on stream irrespective of whether water was introduced to the synthesis gas or not, experiments were also done without addition of water to the feed.

As shown in Fig. 1a–c the reaction rates show similar patterns for all three catalysts without water addition. 20Co-PI1 has the highest activity. 12Co-PW3 has lower loading but much higher dispersion which more than outweights the loading difference. The reaction rate for the 12Co-PW3 catalyst expressed as g HC/gCo h is the highest among all catalysts before the addition of external water (Table 3). For 20Co-FI1 the activity level is significantly lower compared to 20Co-WI1. The dispersion is slightly lower on 20Co-FI1, but the difference in dispersion is not very large: 5.4 % for 20Co-PI1 and 4.7 % 20Co-FI1 and can not explain the difference in activity. Important factors for a possible explanation could be different support structures including metal-support interactions and different initial deactivation behaviour as will be discussed later.

Fig. 1
figure 1

FTS activity of catalysts 20Co-PI1 (a), 20Co-FI1 (b), and 12Co-PW3 (c) at 483 K, 20 bar, and H2/CO = 2.1

Full size image
Table 3 Extent of deactivation
Full size table

3.2.1 Effect of Water on the Reaction Rates

The impact of water on the reaction rates of catalysts 20Co-PI1, 20Co-FI1, and 12Co-PW3 is also shown in Fig. 1a–c. The space velocity of synthesis gas was decreased after 26 h in order to reach 50 % CO conversion. For all the catalysts, the increase in conversion and, accordingly, reactor water concentration led to an increase in the catalytic productivity as shown in Fig. 1a–c. Thus, at relatively low reactor water concentrations without any water added to the feed stream (conversion of CO <50 %), there is a positive correlation between the amount of water and the hydrocarbon formation rate. As shown in Fig. 2a–c, the partial pressure of water was less than 3 bar for all catalysts in absence of external water feeding. Similar results have been reported for cobalt supported on narrow and wide-pore γ-Al2O3, SiO2, and TiO2 [1, 17, 18]. To summarise, at moderate water concentrations, the activity always seems to increase with increasing water concentration for supported cobalt, irrespective of the support nature.

Fig. 2
figure 2

The effect of water partial pressure on the FTS activity of catalysts 20Co-PI1 (a), 20Co-FI1 (b), and 12Co-PW3 (c) at 483 K, 20 bar, and H2/CO = 2.1

Full size image

Presence of larger amounts of water is known to impact the reaction rates in different ways for cobalt based FTS catalysts. However, in this work, continuous addition of 20 mol% water after 54 h increased the catalytic productivity for all the catalysts, though most strongly for the platelet CNF supported samples: 20Co-PI1 and 12Co-PW3 (Fig. 1a–c). As shown in Fig. 2a–c, the partial pressure of water in this case was 6 bar. At the same time, the deactivation rate increased for all the catalysts. As shown in Fig. 1a–c, further increases in steam concentration, i.e. from 20 to 33 mol% after 80 h on stream had a less positive effect on the activity of all the samples. In this case, the partial pressure of water was approximately 8 bar. Finally, removal of water gave a clear negative change in the reaction rates for all the catalysts. As shown in Fig. 1a–c, the catalytic productivity after removal of water was low compared to the dry runs. Thus, large amounts of water resulted in significant deactivation. The extent of deactivation is commented in Sect. 3.2.2.

Since the total pressure was kept constant when water was co-fed along with synthesis gas, the partial pressures of hydrogen and carbon monoxide were reduced. A decrease in reactant partial pressures reduces the reaction rate. Experiments have been performed with He dilution instead of water to quantify the effect of reducing partial pressure of reactants [19]. Accordingly, the improved reaction rates upon water addition observed in some cases for 20Co-PI1, 20Co-FI1, and 12Co-PW3 (Fig. 1a–c) were actually higher than expected from FT kinetics.

At this time, we are not able to explain the positive effects on the reaction rates due to water addition. As mentioned, there are contradictory observations related to the effect of water on the reaction rates. Since the addition of water is positive for the reaction rates of unsupported cobalt and negative for cobalt catalysts supported on various supports such as narrow-pore γ-Al2O3 and SiO2, the support must play a role. The impact of water seems to be at least indirectly related to the pore size and strength of metal-support interaction in the catalytic system. At the lower level of added water, the effect is positive for cobalt dispersed on α-Al2O3 [20, 21], wide-pore γ-Al2O3 [18], wide-pore SiO2 [22], and wide-pore TiO2 [1] where the interaction between the metal and the support is comparably weak. For more strongly interacting systems such as cobalt on narrow-pore γ-Al2O3 [23], the effect of water on the reaction rates is negative. The removal of transport limitations via the water-rich intra-pellet liquids is also an important factor [7].

Temperature programmed reduction profiles can offer information about the severity of metal-support interactions. Reduction curves for the catalysts of the present investigations and cobalt supported on γ- and α-Al2O3 are given in Fig. 3. The two reduction peaks for the CNF supported catalysts can be assigned to a two-step reduction of Co3O4 to Co0 via CoO as intermediate species [24]. Figure 3 shows that the interactions between the support and the cobalt phases were similar for CoRe/γ-Al2O3 and 12Co-PW3. These two catalysts actually give opposite responses to addition of 20 wt%, namely negative and positive, respectively [18]. Thus, the severity of the interaction may not necessarily be the decisive factor. For this reason, we believe that the pore diameter is important. The effect of water may of course also be related to the particle size of cobalt. Large particles are known to be more resistant to oxidation by water than small particles [25]. However, in this work, the average Co0 particle size ranged from 8 to 20 nm (as determined from hydrogen chemisorption), and the effect was positive in all cases. Moreover, the water effect (20 mol%) was least pronounced for the fishbone CNF which exhibited the largest particles. Thus, it is not reasonable to ascribe the positive effect of the CNF based catalysts to large particles of cobalt. The water effect is more pronounced for the platelet CNF supported catalysts than for the fishbone CNF supported catalyst. This indicates that the surface properties of CNF supports might play a role and it is known that the platelet and fishbone CNFs have different edge sides on the graphite sheets.

Fig. 3
figure 3

Temperature programmed reduction profiles of catalyst 20Co-PI1, 20Co-FI1, 12Co-PW3, 20CoRe/γ-Al2O3, and 20CoRe/α-Al2O3

Full size image

A few papers have attempted quantitative descriptions and mechanistic explanations for the impact of water. Bertole et al. [26] studied the effect of water on an unsupported Re-promoted cobalt catalyst in a SSITKA setup, at 483 K, and pressures in the range 3–16.5 bar. The partial pressures of H2, CO, and H2O in the feed were varied, and the effect on activity, selectivity, and intrinsic activity was studied. It was evident that addition of water resulted in an increase in overall activity. Using SSITKA, Bertole et al. [26] were able to deconvolute the site activity from the site coverage. It was concluded that the intrinsic activity was unchanged by addition of water, hence the increased activity is linked with a larger inventory of intermediates leading to products. Krishnamoorthy et al. [27] also found a significant increase in CO conversion for a silica supported cobalt catalyst when water was added to the feed. The authors suggested that the water effects do not rise from new pathways introduced by water, by scavenging effects of H2O on the concentration of site-blocking unreactive intermediates, or by removing significant CO transport restrictions. Krishnamoorthy et al. [27] were left with only the possibility that water influences the relative concentrations of the active and inactive forms of carbon, present at low concentrations on Co surfaces. The authors were unable to give a mechanism by which such effects occurred.

Recently, DFT calculations have pointed to water-mediated proton hopping on iron oxide surfaces [28] and for Ru supported on SiO2 it has been indicated that H2O derived species may act as a promoter or co-catalyst [29].

3.2.2 Effect of Water on the Deactivation

Comparison of the dry and wet runs after water removal showed that the increased water concentration in the reactor gave irreversible deactivation. The extent of deactivation for the different catalysts is given in Table 3. It shows that the deactivation was most severe for the fishbone based catalyst, but also significant for the two other catalysts. Irreversible deactivation at sufficiently high water concentrations have previously been reported for a number of different catalytic systems, e.g. unsupported cobalt, for cobalt supported on γ-Al2O3, on SiO2 and on TiO2. Several reviews summarizing the effect of water on FTS have been given [5, 7, 8].

Several mechanisms have been proposed for the deactivation of Co Fischer–Tropsch catalysts including polymeric carbon formation (polycarbon), sintering, phase transformations such as reoxidation of cobalt sites, metal-support solid reactions and carbidization, Co reconstruction, poisoning, attrition [8]. It is known that bulk oxidation of cobalt is not feasible under realistic Fischer–Tropsch conditions. Thermodynamic calculations have shown that Co particles less than 4.4 nm may oxidize in steam-hydrogen environments corresponding to 75 % CO conversion and 220 °C as commonly encountered in FTS [25]. In situ studies have also confirmed this behaviour [30]. Summarizing, the experimental evidence for catalyst deactivation as a result of water addition is overwhelming as indicated in the reviews given above. However, the effect of water can be indirect for example through water-induced sintering.

Since CNF supported catalysts, and especially fishbone, have relatively weak interactions between the metal and the support, oxidation of cobalt metal or sintering seem most likely. In fact, the fishbone supported catalyst deactivated more strongly than the platetelet supported catalysts (Fig. 1a–c). It is known from several studies that CNF supported catalysts are more susceptible to catalyst deactivation than alumina supported catalysts [12]. However, by covering the CNF surface with a thin layer of SiO2 a stable catalyst was obtained [31]. We have previously attributed deactivation with time-on-stream to oxidation of cobalt rather than sintering without external addition of water [10], but water-induced sintering may very well be the main reason for catalyst deactivation also in this work. Other studies with CNF supported catalysts have indicated that sintering is important [11].

3.3 Effect of Water on the Selectivity

The effect of water on the C5+ selectivity is shown in Fig. 4a–c. The difference in selectivity on different catalysts before water addition is most relevant to be compared during period 2, which starts with the same conversion of 50 %. The 20Co-FI1 catalyst has the highest C5+ selectivity close to 85 %, slightly higher than the 20Co-PI1 catalyst. The 12Co-PW3 catalyst has only a C5+ selectivity of 80 % in agreement with the particle size effect [32]. The increase in C5+ selectivity with CO conversion from Period 1 to 2 can be related to the increased amount of water as the conversion increases. Figure 5a–c confirms this hypothesis. As indicated by the dotted lines, at all conditions, a clear positive correlation between the water concentration in the reactor and the C5+ selectivity existed. The same trend has been found for cobalt support on narrow and wide-pore γ-Al2O3, SiO2, and TiO2 [1]. However, the extent of C5+ enhancement with steam was found to depend significantly on the catalyst, as shown in Fig. 5a–c.

Fig. 4
figure 4

The effect of water on the C5+ selectivity of catalysts 20Co-PI1 (a), 20Co-FI1 (b), and 12Co-PW3 (c) at 483 K, 20 bar, and H2/CO = 2.1

Full size image
Fig. 5
figure 5

The effect of water partial pressure on the C5+ selectivity of catalysts 20Co-PI1 (a), 20Co-FI1 (b), and 12Co-PW3 (c) at 483 K, 20 bar, and H2/CO = 2.1

Full size image

Improved C5+ selectivity due to water addition is in accordance with previous results on all cobalt catalysts. Bertole et al. [33] concluded from isotopic transient kinetic studies that the rate of propagation to termination during chain-growth is strongly correlated with the steady-state amount of active carbon for all carbon number products.

3.4 Effect of Deactivation on the Selectivity

As shown in Fig. 1a–c, the catalyst deactivated with time on stream for all catalysts. As a result of the deactivation, the partial pressure of water decreased as shown in Fig. 2a–c. The deactivation rates were especially significant in the two water periods (Period 3 and 4). However, as shown in Fig. 5a–c, even though the water partial pressure (and CO conversion) decreased with time on stream, the selectivity did not initially decrease. Instead, the selectivity actually increased right after water was introduced. We suggest that this effect is caused by sintering of the cobalt particles at high water partial pressures. Sintering produces larger particles, and the C5+ selectivity for catalysts containing large particles is high [34].

4 Conclusions

The effect of water on the activity and selectivity for a series of CNF supported cobalt catalysts was investigated in a fixed-bed reactor at chain-growth conditions (483 K, 20 bar, H2/CO = 2.1). Addition of 20 and 33 mol% water to the reactor inlet increased the reaction rates and deactivation rates. Because of the high water concentrations, the catalysts were irreversibly deactivated, most likely because of cobalt oxidation and sintering. However, the effect of water on the C5+ selectivity was positive. The structure and thus the surface properties of the CNF supports might play a role for the water effect as well as for FTS without adding water in the feed stream.