3.1. Characterisation
Figure 1.a-c illustrates the PXRD diffractograms of the catalysts, before and after catalytic tests. Calcined Fe/CBEA catalyst showed characteristic peaks of α-Fe2O3, most of which were not observed in the reduced catalyst. Instead, the reduced catalyst showed Fe0 peaks at 2θ = 44.7° and 65° and residual α-Fe2O3 peaks at 35.98° and 62.83°. However, there were no Fe0 or α-Fe2O3 peaks detected in the used catalyst which indicated leaching of Fe from the catalyst support. The residual reaction solution slowly turned to red colour over a period of few days, indicating presence of iron oxides in the solution. Therefore, Fe/CBEA catalyst was not considered further.
Both the fresh and the used T-Fe/MIL-101 catalyst showed peaks corresponding to Fe3O4, suggesting that the catalyst was stable after the reaction. However, the α-Fe2O3 peaks observed in Fe/MIL-101 (Figure S1, ESI) which did not reduce to Fe0 in T-Fe/MIL-101.
T-MIL-88B catalyst showed peaks corresponding to both Fe3O4 and Fe0, which remained steady after a single run of 48 h reaction time and 5 cycles of 21 h each. Only Fe3O4 peaks have been reported after the thermal treatment of MIL-88B at 500°C under nitrogen atmosphere23. However, due to the reducing atmosphere used in this study, some of iron oxide nanoparticles reduced to Fe0. No evidence of iron carbide was found in the PXRD results.
During the thermal transformation of MOFs, first, the linkers break from the metal oxide clusters. After that, the metal oxide clusters agglomerate and reduce depending upon the chemical environment. MIL-88B consists Fe3O clusters coordinated by six carboxylate ligands and three adsorbed water molecules, depending on the synthesis method (Figure 2a). Based on our earlier computational study of thermal transformations in Zr-based MOFs24, we expect the following physiochemical transformations in MIL-88B upon thermal treatment. First, the adsorbed water molecules desorb, and at c.a. 100°C the MOF is expected to change the morphology25. Near the decomposition temperature, some of the linkers start detaching from the cluster. Unlike Fe oxide nanoparticles encapsulated in MIL-101(Cr), where the movement of nanoparticles is less hindered and can easily agglomerate, in MIL-88B, the Fe3O metal clusters are part of the framework and hence remain less mobile. After detachment of the organic linkers, the linkers go through thermolysis, resulting in formation of small gaseous molecules such as CO and CO2. At high temperature, hydrogen is expected to dissociate on iron and likely to catalyse the decarboxylation of linkers, reducing the Fe-O coordination. Without the oxygen from carboxylate groups, formation of single Fe3O4 phase is stoichiometrically not possible in Fe3O. Hence, promoted decarboxylation in H2 environment is likely to increase the abundance of a mixed Fe/Fe3O4 metal nanoparticles. Figure 2 shows the proposed mechanism of thermal evolution of MIL-88B(Fe).
Figure 3.a-f shows the TEM images of MIL-101, Fe/MIL-101, T-Fe/MIL-101, MIL-88B, T-MIL-88B, and used T-MIL-88B, respectively. MIL-101 shows the characteristic octahedral shape of ca. 200-300 nm size (Figure 3a and Figure S2a of ESI)26. After impregnation of Fe over MIL-101, agglomerates of Fe nanoparticles were observed on MIL-101 (Fe/MIL-101) with approximately 50-100 nm in size (Figure 3b), whereas after thermal transformation, T-Fe/MIL-101 exhibited approximately 5-30 nm particles (Figure 3c). The emergence of these smaller nanoparticles is likely due to the thermal transformation of Fe/MIL-101 in reductive atmosphere, where the deconstruction of linkers leads to breakage of the Fe agglomerates. Figure 3d and Figure S2b (ESI) show the characteristic fusiform rod shaped morphology of MIL-88B with ~360 nm length and 90 nm width23. After thermal transformation, T-MIL-88B shows a narrow range of Fe0/Fe3O4 nanoparticle which are well-dispersed over the carbonaceous support (Figure 2e). The amount of Fe on T-MIL-88B is 49.3%, with 13.7% C and negligible amount of H, N and S (Table S1, ESI), which indicates that original MOF structure is completely transformed into porous carbon. Figure 3f shows that the T-MIL-88B catalyst retains its structure after 48h of reaction. Figure 3.g-h illustrates the particle size distribution (PSD) for T-MIL-88B and used T-MIL-88B, respectively. 525 and 476 particles were measured from multiple images which showed most of the particles in 4-16 nm for both fresh and used T-MIL-88B, respectively. The peaks were observed at 8 nm with average particle sizes of 9.7 and 9.1 nm for fresh and used T-MIL-88B, respectively which suggested that the studied catalyst is stable and potentially reusable for this reaction.
The surface oxidation state of Fe in the different catalysts was evaluated by X-Ray photoelectron spectroscopy (XPS) study, as shown in Figure 4. For T-MIL-88B (Figure 4.a), Fe 2p3/2 XPS spectrum exhibited three peaks, including a peak at 706.9 eV corresponding to metallic iron 27. Moreover, the other two peaks at 710.1 and 712.3 eV which are correlated to Fe+2 and Fe+3 oxidation state of iron and the satellite peaks for these aforementioned oxidation state appeared at 716.6 and 719.8 eV 28. In the Fe 2p region of T-MIL-88B, Fe2p1/2 and Fe2p3/2 peaks are situated 710.1 and 723.8 eV, where, the spin orbital splitting is 13.7 eV that indicated the presence of Fe3O4 in T-MIL-88B 29. Fe3O4 may exist as mixed FeO and Fe2O3 states, which appears from Fe+2 and Fe+3 oxidation states 30. The present XPS study shows that Fe3O4 is the dominant species on the surface, where the amount of Fe+2 was 60.4% and Fe+3 was 21.0%, whereas Fe0 was 18.6%. Therefore, the ratio of Fe0 to Fe3O4 was accounted as 1/4.38 in T-MIL-88B.
The XPS spectra of Fe 2p3/2 in T-Fe/MIL-101 exhibited two peaks at 711.7 and 712.4 eV which is related to Fe+2 and Fe+3 along with two satellite peaks at 718.1 and 722.4 eV. Furthermore, Fe2p1/2 and Fe2p3/2 of Fe+2 appeared at 711.7 and 725.4 eV and the spin orbital splitting is 13.7 eV which interpreted the existence of Fe3O4 in T-Fe/MIL-101. Metallic Fe peak is absent in this catalyst which is in good agreement with PXRD results. For Fe/MIL-101 catalyst, Fe 2p3/2 XPS spectra also contained both Fe+2 and Fe+3 at 711.7 and 713.4 eV, respectively. However, the spin orbit splitting for Fe2p1/2 and Fe2p3/2 is 14.1 eV (711.7 and 725.8 eV) which suggested the absence of Fe3O4 phase.
Figure 4b represented the Cr XPS spectra of MIL-101, Fe/MIL-101 and T-Fe/MIL-101 catalysts. In MIL-101, Cr 2p XPS spectra contained only one peak at 577.6 eV which is corresponds to Cr+3 oxidation state31. For Fe/MIL-101, Cr XPS spectra attributed to two peaks at 577.2 and 578.8 eV which are mainly resembles with Cr+3 and CrO3 32. The negative binding energy shift (0.4 eV) of Cr+3 as compared to Cr+3 present in MIL-101 is most likely due to the interfacial electronic interaction (charge transfer) between Cr and Fe after the inclusion of Fe in MIL-10128. The Cr spectra for T-Fe/MIL-101, Cr XPS spectra mainly consisted with Cr+3 peak at 577.1 eV and the amount of CrO3 is very less as compared to Fe/MIL-101 which may be due to the thermal transformation of Fe/MIL-101 under hydrogen atmosphere that reduces the oxidised Cr species on catalyst surface.
The C 1s XPS spectra for Fe/MIL-101 (Figure 3c) shows three different types of C peak at 285, 286.4 and 288.5 which belongs to C-C, C-O-C and O-C=O33. The C 1s XPS spectra of both T-Fe/MIL-101 and T-MIL-88B contains only two peaks corresponding to C-C and C-O-C, whereas, the O-C=O peak is absent, which may be due to the thermal transformation of both Fe/MIL-101 and MIL-88B under hydrogen atmosphere reducing the oxygen content in the catalyst.
A thermogravimetric analysis of Fe/MIL-101 and MIL-88B has been represented in Figure 5. For Fe/MIL-101, the weight loss in the range of 50-250°C is because of the evaporation of water and removal of free terephthalates inside the pore of MOF34. Thereafter, the main weight loss in the temperature range of 270 to 670°C is due to the degradation of organic ligand in the framework of MOF which is attributed to the collapse of the framework34. The weight loss of MIL-88B before 250°C corresponds to the removal of water and excess DMF from the framework35. For MIL-88B, the weight loss occurs in the temperature ranges of 300 to 500°C due to the degradation of H2BDC and the breakdown of the framework. The step in the TGA profile of between 550-650°C is most likely due to the carbonization of the framework and the formation of Fe3O4–carbon composites35.
3.2. Catalyst Activities
3.2.1. Role of Fe based zeolite and MOF catalysts
Figure 6.a-c illustrates the yield and selectivity of AA via aqueous phase CO2 reduction with iodomethane at various pressures. All the catalysts showed some activity for AA production; however, T-MIL-88B was clearly the most active and selective catalyst with best yield of 504 mmol/gcat.L and AA selectivity of 92.4%. Based on stoichiometric calculation, it is equivalent to 80.6% conversion of CH3I into AA. Both Fe/CBEA and T-Fe/MIL-101 provide lower activity for CO2 hydrogenation and >90% selectivity for FA production. With increasing pressure, the yield increased initially but the AA selectivity peaked at 60 bar for both Fe/CBEA and T-Fe/MIL-101. However, the AA yield and selectivity increases with increasing pressure for T-MIL-88B. Since Fe was present in the structural framework of T-MIL-88B, the thermally transformed catalyst consists of - embedded active metal sites dispersed evenly in a carbon matrix23. The high AA activity and the selectivity over T-MIL-88B catalyst is most likely due to the presence of both Fe0 and Fe3O4 which assist the hydrogenation and C-C coupling reactions, respectively36, 37.
3.2.2. Extent of reaction with time
Figure 7.a illustrates the extent of reaction over T-MIL-88B to produce AA and FA via CO2 hydrogenation with CH3I as the starting material in the aqueous media. The reaction proceeds via formation of FA as the initial product, whereas AA was not detected until after 8h of reaction. The AA yield and selectively sharply increased between 12 to 24 h, thereafter gradually increasing to 657.6mmol/gcat.L and 98.8%, respectively, at 48h as the reaction approached equilibrium conversion. Based on the initial CH3I concentration (10 mmol), 100% conversion at 100% selectivity for AA was achieved, within the range of measurement errors. However, as discussed later, CO2 first converts into FA and after reaching the maximum yield (377.4 mmol/gcat.L) at 8h, the FA yield decreases sharply until the end of reaction at 48 h when the FA yield was measured at 8.1 mmol/gcat.L. However, since CH3I is consumed by this time, the residual FA cannot convert into AA. Therefore, for the CO2 hydrogenated into carboxylic acids, the selectivity of AA is 98.8%.
When CH3OH (10 mmol) was used as a reactant with LiI as a co-catalyst (Figure 7b), in otherwise identical reaction conditions, the reaction generates in situ CH3I and hence the peak of FA is broader than Figure 7a. The AA yield and selectivity increased more gradually and achieved a similar yield of 590.1 mmol/gcat.L at 81.7% selectivity after 48 h, which is equivalent to 94% conversion of CH3OH into AA. The in-situ production of CH3I slowed down the conversion of FA into AA, which may be due to mass transfer limitation.
3.2.3. Catalyst reusability
Figure 8 shows that the catalytic activity dropped initially but after three cycles, there was no significant decline in AA yield and selectivity. The PXRD of the used catalyst after five cycles (Figure 1.c), and the TEM image (Figure 3.f) and PSD (Figure 3.h) of used catalyst after 48 h confirmed that the structure is stable and there was no sintering or agglomeration of Fe and Fe3O4 nanoparticles in T-MIL-88B. The initial loss in activity is likely due to the loss of small particles of the catalyst which could not be recollected in centrifuge.
3.2.4. Proposed Reaction Pathway
Reaction mechanism of hydrocarboxylation of methanol in an organic solvent proceeds via reaction of CH3OH with LiI to produce CH3I and LiOH which is similar to the carbonylation of methanol (Monsanto processes) followed by formation of CH3Rh*I due to the insertion of CH3I into a Rh* complexing catalyst 13. Further, CO2 is inserted into CH3-Rh bond to produce CH3COORh*I. Finally, CH3COOH is formed via reduction of CH3COORh*I with H2 molecule in the presence of Ru* to produce HI as an intermediate. Whereas, LiI is regenerated in situ via HI formation which reacts with LiOH to produce H2O and LiI. However, here we show aqueous phase methanol hydrocarboxylation in which the reaction pathway deviates from the published works and FA is formed as an intermediate.
First, we show that FA can react with CH3I in water over T-MIL-88B in H2 atmosphere (Figure 9). The conversion of FA closely follows AA yield and after 24 h of the reaction FA conversion of 91.5% is achieved with 100% AA selectivity.
Next, we show aqueous phase hydrocarboxylation of CH3OH using T-MIL-88B as catalyst and LiI as co-catalyst. Here both liquid and gas samples were collected after 48 of reaction. The liquid sample showed only the presence of HCOOH and CH3COOH with 81.7% acetic acid selectivity (Figure 7.b). Whereas gas analysis did not detect any carbonaceous molecules apart from CO2 (ESI, Figure S4), which eliminates the methanol carbonylation route for AA production.
Figure 10 shows the proposed reaction pathway for acetic acid production via hydrocarboxylation of CH3OH over T-MIL-88B. CO2 and H2 adsorbed over the catalyst and converted into FA, which may desorb. Subsequently, the adsorbed formate species reacts with iodomethane (CH3I) to allow C-C coupling reaction to take place which generates an acetate species and HI as the by-product. Finally, acetate species is converted into acetic acid, whilst LiI might be regenerated from LiOH and HI (step 8).