Fischer-Tropsch Performance
The CO conversion and product selectivity for olefins, alcohols and C5+ was measured throughout the reaction for 0–270 h and is presented in Fig. 1 as averages for all the samples. The average CO conversion of the 2% Mn sample was the highest (49%) whilst the 1% Mn sample was marginally lower at 46%. This indicates that the Mn promoter improved conversion, even at low weight percentages, possibly due to smaller crystallite size of the FCC/HCP Co phases. However, at increasing Mn loading (5–10 wt. %), conversion was also lower at 36%. The 10% Mn sample which was reduced at 450°C had an average conversion rate at 38% throughout the experiment which was similar to the 0 wt. % sample.
At higher weight percentages of Mn (3–10 wt. %) there was an increase in alcohol and olefin selectivity. The selectivity to olefins and alcohols was highest for the 3% Mn sample. C5+ selectivity was the highest for the 1 and 2% Mn catalysts (77–80%) suggesting that the Mn promoted chain growth at low weight percentages. The 1 and 2% Mn catalysts exhibited high C5+ selectivity and lower olefin and alcohol selectivity. The 10% Mn catalyst which was reduced at 450°C showed similar selectivity as the 0% Mn catalyst. Catalyst stability was investigated throughout the reaction and the results for the 2% Mn catalyst are presented in Figure S2 where it was found that activity and selectivity was relatively constant from 0–270 h (up to 12 days under FT conditions). Furthermore, Fig. 2 illustrates that the catalytic activity of the pellets remained consistent in a long-term study lasting 1500 h, showcasing the recoverability of the catalysts with resumption of catalytic activity even after shutdown periods.
Mean XRD-CT pattern Refinement
The mean XRD-CT patterns (illustrated in Fig. 3 in the 3.4–4.0 2θ° range) were produced by summing the patterns for all pixels which provided a starting point for analysis. Evidence of the crystalline TiO2 support is shown through the (111) rutile peak at 3.56°. The anatase TiO2 polymorph is not present in this range but is evident in the full pattern (Figure S6). A summary of the Rietveld refinement results is presented in Table 1 (complete results in Table S2). It was found that the Co2C phase increased (peaks at 3.68° and 3.93°) with Mn loading whilst the cobalt metal phases (peak at 3.81°) decreased. The Co was mostly reduced in the Co-containing species as the samples were recovered from the reactor under predominantly reducing conditions and the wax products remaining on the catalysts prevented oxidation. There was, however, a small percentage (3 wt. %) of CoO detected in the 10% Mn sample with an expanded lattice parameter (4.29 Å). A fraction of this could be MnO given that MnO was not separately refined. This is due to Co and Mn having similar scattering factors. Previous research has also pointed towards the presence of mixed-oxide spinels with an expanded lattice parameter (Co1 − xMnxO) 28,55. Consistent with previous work, no other crystalline Mn-containing species were detected 28. The mean pattern in Figure S7 highlights the presence of the different Co phases in the 3% Mn sample which contains the Co metal phases (FCC and HCP) and the Co2C phase as well as the difference between the refined and experimental data. The FCC (200) peak was excluded from the refinements as it was found to have reduced intensity due to the possible presence of stacking faults 56. Whilst this aided the fit of the FCC (100) peak, the presence of stacking faults results in more complex peak shapes and asymmetry that would require more elaborate modelling to improve the fit further 56. The full pattern is presented in Figure S6, where the support phases (TiO2), anatase and rutile, were identified whilst MnTiO3 was formed in the 10% Mn sample reduced at 450°C.
For the 1 and 2% Mn loading samples, less than 1 wt. % of Co2C phase was present. However, above 3% Mn loading, the wt. % of Co2C noticeably increased from 3.6 wt. % (3% Mn) to 7.1 wt. % (10% Mn). This increase in Co2C formation corresponds with increasing alcohol and olefin selectivity of the catalyst. The crystallite size of the Co FCC phase was found to decrease with increasing Mn loading which correlates with previous research 57. This smaller crystallite size in the 1 and 2% Mn samples was a likely cause for the increased activity and C5+ selectivity in these samples. The total Co metal (FCC + HCP + Co2C) wt. % was found to be approximately 8 wt. % (lower than the known 10 wt. %) indicating that some of the Co was not found in the XRD data due to small crystallite sizes or disordered Co being present. The low wt. % (0.5-2%) of CoO present and the retention of Co2C phase affirm the efficacy of the catalyst wax passivation method.
A 5% Mn sample was extracted after 150 h of reaction, and it was found that the Co2C phase had already formed at this intermediate stage. The refined weight percentage of the Co2C phase increased from 7.5 wt. % at 150 h to 8.9 wt. % at 300 h and the crystallite size increased from 7.5 nm to 9.3 nm. The data indicates that carbide had already become the dominant Co phase by 150 h with only marginal further increase observed until 300 h for the samples. This data only shows the bulk average analysis across the samples, akin to the analysis from a laboratory X-ray diffractometer study, where an increase in overall Co2C was found with increasing Mn loading. Using this data in combination with XRD-CT analysis enables not just a bulk analysis to be performed but also spatially resolved phase and particle size correlation.
Table 1
Rietveld Refinement results of the XRD-CT mean patterns illustrating the Co phase wt. % and CS (crystallite size in nm) in the samples extracted after 150 and 300 h. The Co2C wt. % increases with Mn loading whilst the Co metal phases (FCC and HCP) decrease with increasing Mn wt. % beyond 5%. The 10 wt. % samples were reduced at 450°C.
|
|
Mn wt. %
|
Time (h)
|
|
150
|
300
|
Phase
|
|
5%
|
0%
|
1%
|
2%
|
3%
|
5%
|
10%
|
10%a
|
Co2C
|
Wt. %
|
7.5
|
-
|
0.6
|
0.9
|
3.6
|
8.9
|
7.1
|
-
|
|
CS
|
7.5
|
-
|
-
|
6.5
|
10.6
|
9.3
|
11.7
|
-
|
FCC
|
Wt. %
|
0.4
|
3.5
|
3.3
|
3.6
|
1.6
|
0.1
|
0.0
|
2.9
|
|
CS
|
3.54
|
10.5
|
8.1
|
7.7
|
5.2
|
-
|
-
|
11.2
|
HCP
|
Wt. %
|
0.2
|
4.9
|
4.8
|
3.8
|
3.1
|
0.2
|
0.7
|
5.3
|
|
CS
|
-
|
2.6
|
2.2
|
2.2
|
7.1
|
-
|
9.9
|
2.9
|
a Reduction at 450 °C
XRD-CT Refinement
The XRD-CT spatial mappings presented in Fig. 4 illustrate the weight percentages of the different phases present in the catalyst and their crystallite size with spatial resolution. At lower Mn loadings, Co metal (in both FCC and HCP phases) is uniformly distributed, while at higher loadings, it tends to concentrate towards the centre of the extrudates. This spatial distribution is further elucidated in Fig. 5, which displays XRD patterns of consecutive layers within the 3% Mn catalyst. Here, the Co2C phase predominates at the periphery of the extrudates, while FCC and HCP Co metal phases are concentrated at the core. The 3% Mn sample exhibits the highest concentration of the Co HCP phase (7–8 wt. %) compared to the 0–2% Mn samples (3–4 wt. %) at the central region of the extrudates, evident in the XRD patterns of the successive layers in Figure S8. This observation was not captured in the analysis of the mean XRD patterns, where lower HCP content was found for the 3% Mn sample. This highlights the efficacy of the XRD-CT technique in detecting phase distribution heterogeneities, revealing distinctions between bulk and specific regions. Notably, the 3% Mn sample represents the point just before the significant formation of Co2C and an increase in the HCP phase. This suggests an equilibrium between cobalt carburization and reduction back to HCP cobalt, favouring the latter during the reduction process due to their structurally similar hexagonal configuration. Above this equilibrium point in Mn loading, cobalt carbide phase predominates, while below it, the HCP phase prevails. Strikingly, the 3% Mn sample also exhibited the highest alcohol and olefin selectivity, signifying that an optimal equilibrium between the HCP and Co2C phases was reached in this sample, contributing to an enhanced selectivity towards oxygenates.
FCC and HCP Co phases were both present in the 10% Mn (reduced at 450°C) sample and there was no presence of carbides, exhibiting a similar distribution of phases as the 0% Mn sample. Both catalysts also exhibited low selectivity for alcohols and olefins. This is due to the Mn being locked away in the Mn titanate phase which prevents its promotional effects that require the Mn to be close to the cobalt 58. Only a small percentage of CoO was found in the samples indicating that most of the cobalt was reduced however more was found to be present at higher Mn loadings, likely due to the presence of mixed oxide spinels (Co1 − xMnxO). The Rwp (residual weight percentage) was relatively constant at around 7–8% indicating a good fit for all the patterns.
Like in the mean refinements, it was found that the FCC crystallite size decreases with increasing Mn loading which corresponds to previous research 28. The crystallite sizes of the HCP phase (2.5 nm) are much smaller than the FCC phase (6–10 nm) except for the 3% Mn catalyst where larger HCP particles were found to be formed at the centre of the extrudates. Larger Co2C particles (14 nm) are found to be formed on the periphery of the 3% Mn catalyst compared to the 5 and 10% Mn catalysts (10 nm). This further supports the hypothesis of a Co2C to HCP phase transformation equilibrium. This also agrees with previous research that found that Mn in close proximity to larger cobalt particles would carburize 59. Small CoO crystallites were found to be present at higher Mn loadings (5 and 10 wt. %), which were expected to be mixed-oxide spinels (Co1 − xMnxO), indicating that the Mn inhibited reduction of Co. The phase distribution and crystallite sizes did not vary significantly along the length of the pellet and XRD-CT cross-section maps of the samples at alternative positions within the pellets are presented in Figure S10.
Spatial maps of the wax were produced (Fig. 6) by refining polyethylene (C3H6)n and it was found that that more wax content remained on the catalyst with increasing Mn loading. This correlates with research that found that olefin products are more likely to re-adsorb on the catalyst surface, due to greater solubility in synthesis liquids, and undergo secondary reactions such as hydrogenation 60,61.
NMF Analysis of XRD-CT data
NMF analysis was performed to ensure a thorough identification of all phases present in the catalyst. The analysis revealed the presence of five components as further increasing the component count led to the inclusion of components that resembled noise. Due to the homogeneous presence of the support and crystalline wax (C3H6)n, all components included these phases. In Figure S11, masked images display the component locations, while Figure S12 illustrate the computed XRD patterns of the components and Table S4 contains the Rietveld refinement results. Component 1 (FCC/HCP phases) was concentrated at the catalyst centre with higher Mn loading, while Component 3 (Co2C) was primarily located on the catalyst periphery, consistent with the XRD-CT findings. Furthermore, component 2 was found to contain the MnTiO3 phase which was found only in the 10% Mn catalyst reduced at 450°C. Components 4 and 5, demonstrated a co-location of HCP and Co2C phases but only accounted 2.4 and 0.3 wt. % of the data intensity respectively. The NMF analysis confirmed that no additional phases were present that were not detected by the XRD-CT analysis.
PDF-CT data analysis
The summed PDF patterns were produced by summing all the Bragg data for each pixel and then Fourier transforming the result. Real-space refinement was performed on the patterns and the results are presented in Table S6. The refined mean patterns from 0–10 Å are presented in Fig. 7 where there are changes to the Co-Co peak at 2.5 Å with increasing Mn due to the formation of cobalt carbide. Similar to the XRD refinement, the PDF results indicated increased Co2C formation with increasing Mn loading, first appearing at 3% Mn, whilst Co (FCC and HCP) phases were present at lower Mn loadings (0 and 1 wt. %). There was a higher percentage of FCC/HCP cobalt detected in the PDF than in the XRD indicating the presence of small cobalt metal particles. A larger percentage of CoO is detected for the 10% Mn sample indicating that the Mn was present in small mixed-oxide spinels (Co1Mn1 − xO) or small MnO particles (1 nm) that could not be detected by XRD. Previous research has also found that small Co particles oxidise to CoO 38.
The PDF-CT wt% and crystallite size maps are presented in Fig. 8. The results from the PDF-CT refinement were consistent with the findings from the XRD-CT refinement where an increasing Mn loading led to an increase in Co2C formation on the extrudate periphery and decreasing Co (FCC and HCP) content in the centre. Similar to the XRD results, the crystallite size of the FCC particles was found to decrease with increasing Mn loading, correlating with previous research 28. A higher HCP wt. % was found in the centre of the extrudates at higher Mn loadings with the highest concentration present in the 3% Mn sample. This aligns with the XRD results, revealing a phase equilibrium between the Co2C and HCP phases, with the Co2C phase prevailing at elevated Mn loading levels.
NMF analysis was also performed on the PDF-CT data to ensure that all phases including non-crystalline phases were detected. The spectra were decomposed into 5 components although only 3 were significant. Two of the components (2 and 4) were of much weaker intensity comprising only 0.1–0.2 wt. % of the total intensity. The component locations are shown in Figure S14 and their respectives PDFs are illustrated in Figure S15 while Table S8 contains the Rietveld refinement results. Component 1 contained FCC/HCP Co located at the centre of the pellets at higher Mn loading with the presence of small CoO nanoparticles. Component 5 contained the Co2C phase that was present on the periphery in the 3–10% Mn samples. Component 3 contained the MnTiO3 phase co-located with FCC/HCP Co which was exclusively present in the 10% Mn sample reduced at 450°C. The NMF analysis confirmed the previous PDF-CT results serving as an efficient tool for spatial decomposition of PDF-CT data into its constituent chemical components.
XRF mapping experiments
µ-XRF images are presented in Figure S16 which provide a visual insight into the cross-sections and lateral faces of the 3% Mn samples both before and after undergoing a 300 h reaction. The images reveal that Mn and Co are colocated in the samples and there are no significant alterations in their distribution observed before and after the extended reaction period. This opposes recent research that found that Mn Oxide and Co were mobile under reaction conditions 59.
XAS Results
To understand how the environment of the Mn species in the catalysts varied with different weight loadings, X-ray Absorption Near Edge Structure (XANES) were collected of the Mn K-edge and compared to a variety of oxidic standards (Fig. 9).
All systems shared a common pre-edge feature at 6540 eV, which is attributed to electronic transitions from the 1s to 3d levels. This feature however is known to be largely insensitive to the precise environment compared to the main-edge feature 62. Similarly, there is a significant feature at 6560 eV, attributed to dipole-allowed 1s to 4p transitions, which is particularly pronounced in the 1–2% Mn samples 63. The XANES data shows excellent agreement with the 1–2% Mn samples (Fig. 9, red and blue respectively), suggesting they are in very similar environments. However, as the loading increases, in the main-edge (between 6545 and 6550 eV), there is a notable shift to lower energies, as the Mn weight loading increases, suggesting a reduction in oxidation state. Comparing these spectra with known reference compounds (MnO, MnO2, Mn3O4, Mn2O3 and MnTiO3), Figure S17 shows that the 1–2% Mn systems have a strong resemblance to Mn3O4 and Mn2O3 (Figure S17A). This suggests that at least part of the Mn in these catalytic systems are Mn3+ in an octahedral environment. The pre-edge peak in the 1 and 2 wt. % however is around 1 eV slightly lower than the Mn3O4 and Mn2O3 species, this is likely due to a slightly lower average oxidation states in the samples, compared to the oxidic references, or due to their existence as smaller nanoparticles, and not bulk oxides. Above these loadings (> 2 wt. % Mn, Figure S17B) a signal at 6553 eV begins to evolve, which coincides with the main feature in the reference MnO spectra. The 10% Mn sample reduced at 450°C presented similar features to the MnTiO3 reference, suggesting that post-reaction the Mn active sites alloy with the TiO2 support, forming a ternary phase. Overall, the XANES study suggests that at lower weight loadings (1–2% Mn) the Mn environment resembles Mn3O4 and Mn2O3. At higher weight loadings (3–10% Mn) there are significant features from MnO appearing, showing a notable change from the Mn3O4 and Mn2O3 environment.