Stephanie
Chapman
ab,
Catherine
Brookes
bc,
Michael
Bowker
bc,
Emma K.
Gibson
bd and
Peter P.
Wells
*bd
aUniversity of Southampton, Southampton, SO17 1BJ, UK
bUK Catalysis Hub, Research Complex at Harwell, RAL, Oxford, OX11 0FA, UK. E-mail: peter.wells@rc-harwell.ac.uk
cCardiff University, Cardiff, CF10 3XQ, UK
dUniversity College London, London, WC1H 0AJ, UK
First published on 8th December 2015
The performance of Mo-enriched, bulk ferric molybdate, employed commercially for the industrially important reaction of the selective oxidation of methanol to formaldehyde, is limited by a low surface area, typically 5–8 m2 g−1. Recent advances in the understanding of the iron molybdate catalyst have focused on the study of MoOx@Fe2O3 (MoOx shell, Fe2O3 core) systems, where only a few overlayers of Mo are present on the surface. This method of preparing MoOx@Fe2O3 catalysts was shown to support an iron molybdate surface of higher surface area than the industrially-favoured bulk phase. In this research, a MoOx@Fe2O3 catalyst of even higher surface area was stabilised by modifying a haematite support containing 5 wt% Al dopant. The addition of Al was an important factor for stabilising the haematite surface area and resulted in an iron molybdate surface area of ∼35 m2 g−1, around a 5 fold increase on the bulk catalyst. XPS confirmed Mo surface-enrichment, whilst Mo XANES resolved an amorphous MoOx surface monolayer supported on a sublayer of Fe2(MoO4)3 that became increasingly extensive with initial Mo surface loading. The high surface area MoOx@Fe2O3 catalyst proved amenable to bulk characterisation techniques; contributions from Fe2(MoO4)3 were detectable by Raman, XAFS, ATR-IR and XRD spectroscopies. The temperature-programmed pulsed flow reaction of methanol showed that this novel, high surface area catalyst (3ML-HSA) outperformed the undoped analogue (3ML-ISA), and a peak yield of 94% formaldehyde was obtained at ∼40 °C below that for the bulk Fe2(MoO4)3 phase. This work demonstrates how core–shell, multi-component oxides offer new routes for improving catalytic performance and understanding catalytic activity.
Despite their industrial significance, the application of iron and bismuth molybdates is hindered by low surface areas (typically < 6 m2 g−1),4,5 which limit the catalytic performance. Recent catalysis literature is largely devoted to the study of nanoparticle (NP) catalysts. Such systems can afford high mass activities and often exhibit exceptional properties compared to those materials prepared by conventional routes. Since nearly 50% of industrial formaldehyde production uses bulk-phase ferric molybdate,6 the development of nanoscale systems as a means of catalyst optimisation is hugely relevant.
Bowker,7 Brookes,8,9 and others10,11 have investigated haematite-supported molybdena, MoOx@Fe2O3, a shell@core system that comprises layers of molybdena deposited on haematite nanoparticles by thermal spreading,10 or incipient wetness impregnation.7–9,11 The shell@core catalyst was designed to isolate the catalytic contribution of surface molybdenum from that in the ferric molybdate bulk.8 By studying the model system, the catalyst activity has since been attributed to a well-dispersed layer of amorphous or low-dimensional crystalline molybdena (MoOx) at the surface.12 Notably, Brookes et al. reported improved catalytic activity using the MoOx@Fe2O3 catalyst, versus the ferric molybdate bulk phase, which was attributed to a 3–4-fold greater surface area in the former.9 It is thus proposed that by supporting the monolayer of catalytic molybdena at higher surface areas, an enhanced performance can be achieved.
Whilst dispersion of molybdena on inherently high surface area supports, such as alumina, titania and silica, has been documented,13 these materials favour inhomogeneous growth and distribution of molybdena. An exposed support reduces formaldehyde selectivity by promoting the formation of carbon monoxide and carbon dioxide by-products.14 In contrast, XAFS studies have shown that a solid state reaction between surface MoO3 crystallites and bulk haematite generates a subsurface layer of catalytic Fe2(MoO4)3.15 The interaction between the molybdena and ferric molybdate subsurface promotes high selectivity for formaldehyde; hence haematite is the favoured catalyst support.
In addition to applications as a support, nanoparticulate haematite finds widespread use in catalysis, including the oxidation of carbon monoxide,16 photocatalytic splitting of water,17 biomass conversion,18 and catalytic combustion.19 Although an inherently high energy of crystallisation of iron oxides affords surface areas in excess of 100 m2 g−1,20 high-temperature crystallisation of haematite leads to sintering and surface area losses.21
For the purpose of methanol oxidation, it is desirable to preserve the chemical properties of haematite and yet maintain higher surface areas to improve the catalyst performance. Herein, aluminium-doped haematite has been used as a support for a molybdena catalyst. The support has been found to retain the characteristics of hematite but a greater resistance to sintering has sustained a higher surface area of catalytic molybdena. Although the influence of the Al-dopant on the structure of iron oxide has been discussed extensively,22–24 to our knowledge, the application of such materials in the context of a catalyst support material is a novel application.
The shell@core catalysts were prepared by incipient wetness impregnation of haematite.7,9 An aqueous solution of ammonium heptamolybdate tetrahydrate (Sigma Aldrich) was prepared at a concentration to deliver the desired number of molybdena monolayer equivalents at the point of incipient wetness.
The nature of the aluminium incorporation in AlxFe(2−x)O3 was investigated by XPS, and this was consistent with isomorphous substitution of Al into the bulk (Fig. 1). The XPS spectrum did not indicate any preferential segregation of Al. This contrasts with ferric molybdate systems (Mo/Fe < 1.5), for which segregation is apparent by a surface Mo loading 2–3 times the theoretical bulk phase.25 In addition to the Al3+ peak at 119 eV, Fe 2p3/2 and 2p1/2 peaks at 725 eV and 711 eV, respectively, are consistent with high-spin Fe3+ of haematite. Lattice oxygen at a binding energy of 530 eV was identified.26
Fig. 1 Widescan XPS spectrum of AlxFe(2−x)O3 and table, inset, of the associated surface composition. |
XRD analysis of AlxFe(2−x)O3 was similar to that of haematite except that the peaks for AlxFe(2−x)O3 were displaced to a higher angle than in commercial α-Fe2O3 (Fig. 2). Under Bragg's Law, this implied a reduction in the d-spacing on introducing aluminium.26
Changes in lattice parameters were quantified by Rietveld refinement in the space group RcH (Table 1), which indicated reductions along both the a- and c-axes. Suppressed unit cell growth in aluminous haematite has been documented previously.23,27 Where α-Al2O3 and α-Fe2O3 share the corundum crystal structure, the formation of a solid solution of trivalent cations is relatively facile. In affecting isomorphous substitution of Fe3+ (r = 0.645 Å) by the smaller Al3+ ion (r = 0.535 Å),28 atomic displacement and altered electrostatic interactions introduce lattice strain. This is manifested as smaller crystallites and an enhanced surface area in Al-doped haematites.22,26 Raman spectroscopy confirmed the reduced particle dimensions of AlxFe(2−x)O3versus commercial haematite through a blue shift in the excitation bands originating from quantum confinement effects (Fig. 3).29 Additionally, where the substitution of Fe3+ with smaller Al3+ centres decreases atomic separation and concomitantly increases bond force constants, Raman bands have been shifted to a higher energy.23,30 Another consequence of the reduced grain size is peak broadening, since phonon confinement necessitates an increase in photon momentum distribution. Some loss of crystallinity on the introduction of Al (supported by a less intense XRD pattern) may have contributed to the broadened Raman bands.23,30–32
Haematite support | BET multipoint surface area/m2 g−1 | Optimised lattice parameters/Å | Optimised cell volume/Å3 | Scherrer crystallite size/nm | E a (CO2)/kJ mol−1 | |
---|---|---|---|---|---|---|
a | c | |||||
Theoretical34 | — | 5.0382 | 13.7721 | 302.7 | — | — |
Commercial | 21 | 5.0273 | 13.7245 | 300.4 | 37 | 182 |
AlxFe(2−x)O3 | 45 | 4.9933 | 13.6266 | 294.2 | 14 | 186 |
TPD analysis was used to compare the reaction profile of Al-doped haematite with that of phase-pure haematite. Mass spectrometric measurements revealed high-temperature CO2 and H2 production, which was attributed to methanol decomposition via the formate intermediate. Water was desorbed in two peaks, the first coincident with methanol (an aqueous solution), and the second high-temperature desorption was accompanied by CO2 as a product of combustion.
The combustion of methanol under oxidative conditions is typical of haematite catalysis.6 However, methanol TPD reveals a notable distinction between the behaviour of AlxFe(2−x)O3 and that of commercial haematite (Fig. 4). For the Al-doped sample, CO2 evolution is shifted to a higher temperature, which is consistent with the smaller particle size, noted previously. The decreased grain size leads to a concomitant increase in the surface free energy due to more acute curvature of the surface and a prevalence of low-coordinate sites. Where the adsorption of the bidentate formate intermediate satisfies surface valencies, mutual stabilisation yields stronger surface–adsorbate interactions that compel a larger energy input to initiate the reaction. Based on first order CO2 production,26 the Redhead equation has been used to approximate the activation energy of formate decomposition from the peak desorption temperature of CO2 (Table 1).33
Relative to the haematite supports, α-MoO3 has an exceptionally low surface area (<1 m2 g−1). Whilst our own work shows that the surface area of the shell@core catalyst is diminished with increasing monolayer coverage (Table 2),9 these systems still offer a higher surface area than bulk Fe2(MoO4)3; one of their principal advantages.
Theoretical number of MoOx-monolayers on the AlxFe(2−x)O3 support/ML | Surface area/m2 g−1 |
---|---|
0 | 46 |
1 | 45 |
3 | 35 |
6 | 30 |
The Mo content of 3 ML systems, determined by SEM-EDX analysis, correlated with their theoretical values (Table 3). XPS analyses provided evidence of surface segregation, with Mo loading in vast excess of a theoretical bulk phase (Table 3). If XPS was completely surface-specific, the 3 ML catalysts should show the same atom percent Mo, corresponding to the MoOx surface monolayer. However, XPS does, to some extent, penetrate the surface (∼1 nm depth).6 At a higher surface area, conversion of haematite to Fe2(MoO4)3 is more extensive and hence there is a greater concentration of Mo at the surface. Electron binding energies from XPS affirmed that Al, Fe and Mo cations were in their maximum oxidation states of +3, +3 and +6, respectively.35
Material | Surface area/m2 g−1 | SEM Mo content/wt% | XPS | ||||
---|---|---|---|---|---|---|---|
Theoretical | SEM-EDX | σ | Mo 3d peak position/eV | XPS Mo content/atom% | Mo content of theoretical bulk phase/atom% | ||
α-MoO3 | 1 | 66.7 | 62.2 | 0.4 | — | — | — |
Fe2(MoO4)3 | 3 | 48.7 | 46.1 | 0.4 | — | — | — |
3 ML MoOx@ISA | 15 | 3.0 | 2.8 | 0.1 | 232.065 | 5.12 | 1.1 |
3 ML MoOx@HSA | 35 | 7.3 | 7.6 | 0.2 | 232.743 | 5.29 | 2.0 |
The Mo surface-segregation, evidenced by XPS, is well-documented in the literature.8,9,11,36 Therefore, it is reasonable to treat Mo as being localised at the surface. On tuning the energy of the X-ray radiation to the molybdenum K-edge (20000 eV),37 XAFS can be used to analyse the surface structure of the shell@core catalysts.
The X-ray absorption near-edge structure (XANES) region was used to establish the coordination environment of Mo in the shell@core catalysts with 1, 3 and 6 ML surface coverages on the HSA support (Table 4). Two features are assigned in the XANES spectra (Fig. 5).7,38 The first is a pre-edge peak at 20010 eV, attributed to the dipole-forbidden, quadrupole-allowed 1s → 4d transition. The pre-edge absorption is weak for Mo in a distorted octahedral environment (as in molybdena) but increases in intensity with tetrahedral character (as in Fe2(MoO4)3). The second feature is an intense white line at ∼20020 eV from the dipole-allowed 1s → 5p transition, characteristic of Mo in (distorted) octahedral geometry.
Theoretical molybdena coverage/ML | Proportion of Mo in octahedral geometry | Proportion of Mo in tetrahedral geometry | R-Factor |
---|---|---|---|
1 | 1.000 | 0.000 | 0.0055310 |
3 | 0.605 | 0.395 | 0.0007580 |
6 | 0.275 | 0.725 | 0.0007589 |
Fig. 5 Normalised Mo-XANES for 1, 3, 6 ML MoOx@HSA catalysts, including an enlargement of the pre-edge region. |
The structure of the Mo-XANES is typical for a MoOx@Fe2O3 catalyst.8,9 From initial inspection, the increasing intensity of the pre-edge peak, with simultaneous diminution of the K-edge, indicates that the proportion of Mo in tetrahedral geometry increases with the level of surface doping. It is possible to quantify the proportion of octahedral and tetrahedral Mo by performing a linear combination fit (LCF), against suitable references (Table 4).
For the 1 ML catalyst, all Mo is in the (pseudo-)octahedral geometry and comprises the amorphous MoOx surface phase.9 At loadings in excess of 1 ML, an octahedral component from the MoOx monolayer remains but subsurface layers are increasingly converted to Fe2(MoO4)3. For the 3 ML and 6 ML catalysts, a tetrahedral component to Mo geometry indicates dopant levels in excess of monolayer coverage, where surplus Mo has reacted with the haematite core to form Fe2(MoO4)3.
The DRIFTS study of the 3ML@ISA and 3ML@HSA catalysts is consistent with a dual adsorption pathway,40 two methanolic surface species being detected (Fig. 6). The O–H group of methanol, adsorbed non-dissociatively, is identified by stretching vibrations at 3100–3500 cm−1 and a bending mode at ∼1370 cm−1. Also, a pair of intense bands at ∼2950 and 2850 cm−1 correspond to the symmetric stretch and the first overtone of the symmetric bend of the methanol C–H bonds, respectively.40 An analogous pair of bands in the C–H region (2930 and 2830 cm−1) identifies methoxy formed by dissociative chemisorption of methanol.
Fig. 6 3D maps of the C–H region of the IR spectra during TPD of methanol from the (a) 3ML@ISA and (b) 3ML@HSA catalysts. |
The C–H region is particularly informative and combination bands possess unusual intensities due to Fermi resonance between the symmetric C–H stretch and the first overtone of the corresponding symmetric deformation.40,41 Through the C–H band position, DRIFTS can reveal surface acid–base character. Over metal oxides of strong Lewis base character, surface oxygen or adsorbed hydroxyl may abstract the alcoholic proton from adsorbed CH3OH. Thus, IR can distinguish the low frequency C–H stretch of anionic methoxide on a basic support (∼2910/2800 cm−1) from the intermediate bands of methoxy on moderately Lewis-acidic oxides (2930/2830 cm−1) and the high frequency C–H vibrations of undissociated methanol on Lewis acidic surfaces (2960/2850 cm−1).42 In the case of the 3 ML catalysts, the strong Lewis acid character, particularly of the 3ML@ISA catalyst, is revealed by the greater intensity of the C–H stretches from adsorbed methanol, relative to methoxy. Overall, the C–H bands are less resolved in 3ML@HSA, which might indicate a greater range of adsorption sites, of varying affinity, for binding the methanolic adsorbate.
Furthermore, by tracking the C–H modes between 3000 and 2750 cm−1, the reaction of surface-bound methanol can be monitored. On applying a temperature ramp, the intensity of the bands associated with surface methoxylation decline, presumably accompanying the reaction of methanol/methoxy to formaldehyde and CO, as evidenced by TPD. Whilst molecular methanol may be the dominant surface species at low temperature, associated C–H bands are lost at lower temperatures compared to those of methoxy, reflecting the stronger surface interactions of the latter.
Some distinctions can be made between the DRIFTS responses of the 3ML@ISA and 3ML@HSA materials. Firstly, bands associated with methanol and methoxy intermediates are absent at 255 °C from the ISA catalyst, whereas these bands are still detected at 315 °C for 3ML@HSA. This result supports microreactor TPD, for which formaldehyde (and CO) were stabilised on the HSA catalyst.
For the 3ML@ISA catalyst, the two sharp peaks between 3600 and 3100 cm−1 are typical for a non-bonded –OH group, but may be attributed to methanol and hydroxy species, each making consistent interactions with the surface, such that their O–H stretching frequency is well-defined and more intense. In contrast, on exposing 3ML@HSA to methanol, a broad O–H band is observed, indicating extensive intermolecular hydrogen bonding to the adsorbate. The C–H stretching region analysed by DRIFTS previously revealed (from broad C–H stretching bands) a range of binding interactions between 3ML@HSA and the methanolic adsorbate. Variations in the Lewis acid–base interactions between the surface and adsorbate may modify the O–H bond polarity and hence the strength of the hydrogen bond interactions. Since the O–H bond force constant is sensitive to modifications in the intermolecular hydrogen bonding capabilities, a broad range of O–H stretching frequencies is detected.
In contrast, the Al-doped haematite operates a combustion pathway, converting methanol to CO2 (with a trivial quantity of CO at low conversion).43 The catalytic activity of the aluminous haematite is inferior to that of ferric molybdate, AlxFe(2−x)O3 achieves 50% methanol conversion (T50) at 20 °C above that for Fe2(MoO4)3, which reflects the stabilisation of surface intermediates on the HSA support.
However, when the haematite support is doped with molybdena, the reactivity of the resultant shell@core catalyst is dominated by the surface, where large energetic barriers to oxygen insertion favour formaldehyde desorption.25 Compared to the bulk phases, T50 is universally lower for the shell@core systems but decreases as the surface area, and hence access to active sites, is improved (Table 5).
System | Surface area/m2 g−1 | T 0/°C | T 50/°C | T 100/°C | T sel/°C | Maximum formaldehyde yield/% | T max/°C |
---|---|---|---|---|---|---|---|
AlxFe(2−x)O3 | 46 | 173 | 253 | 285 | — | 0 | — |
Fe2(MoO4)3 | 3 | 104 | 233 | 283 | 104–344 | 100 | 283–344 |
3ML@HSA | 35 | 143 | 206 | 277 | 143–245 | 94 | 244 |
3ML@ISA | 14 | 178 | 213 | — | 178–207 | 84 | 252 |
The TPPFR indicated that, of all the shell@core systems, a higher formaldehyde yield is attained with the 3ML@HSA catalyst and the maximum formaldehyde yield (Tmax) is also shifted to a lower temperature (Table 5). Given that, in TPD, the 3ML@HSA catalyst desorbs formaldehyde at the highest temperature of all the shell@core catalysts, the catalyst must therefore operate a high methanol conversion in the Tsel range.
Unlike the Fe2(MoO4)3 catalyst, the 3ML@HSA catalyst does not achieve 100% formaldehyde yield but between 190 and 255 °C the HSA shell@core catalyst outperforms the other systems, selectively oxidising methanol to formaldehyde with high yield. Significantly, the HSA supported molybdena catalyst achieves its maximum formaldehyde yield (94%) at Tmax ∼ 40 °C below that for the industrial bulk phase (Fig. 8).
Fig. 8 Formaldehyde yield of the bulk and shell@core iron molybdate catalysts under the TPPFR of methanol. |
In a novel preparation, AlxFe(2−x)O3 was coated with molybdena to yield a shell@core catalyst of surface area 35 m2 g−1. On confirming a Mo-rich exterior by XPS, surface analysis was performed by tuning XAFS to the Mo K-edge. All systems maintained a (pseudo)-octahedral Mo component of surface MoOx, though an increase in tetrahedral Mo at higher surface dosing was consistent with a more extensive subsurface Fe2(MoO4)3 formation. For the high surface area shell@core system, the sub-surface Fe2(MoO4)3 phase was identified by XRD, Raman and IR spectroscopies.
In the TPD of methanol, the reactivity of the shell@core catalysts was directed by the molybdena surface. Formaldehyde and CO products were consistent with the methoxy intermediate detected by DRIFTS. The TPPFR of methanol over the shell@core catalyst revealed an increase in peak formaldehyde yield with catalyst surface area, accompanied by a shift of the maximal yield to a lower temperature. Between 190 and 255 °C, the formaldehyde yield over the 3ML@HSA catalyst exceeded that of the other systems. Significantly, the 3ML@HSA catalyst reached the maximum formaldehyde yield (94%) ∼40 °C below that for the bulk phase. As such, further catalytic testing of the 3ML@HSA catalyst would focus on isothermal methanol oxidation between 190 and 255 °C to monitor formaldehyde yield over an extended reaction period. As is, the improved catalytic activity of the 3ML@HSA catalyst might be trialled for other reactions for which ferric molybdate catalysis is employed, such as propene,44,45 methane46 and toluene47 oxidation.
This journal is © The Royal Society of Chemistry 2016 |