Stability of Colloidal Iron Oxide Nanoparticles on Titania and Silica Support.

Using model catalysts with well-defined particle sizes and morphologies to elucidate questions regarding catalytic activity and stability has gained more interest, particularly utilizing colloidally prepared metal(oxide) particles. Here, colloidally synthesized iron oxide nanoparticles (Fe x O y -NPs, size ∼7 nm) on either a titania (Fe x O y /TiO2) or a silica (Fe x O y /SiO2) support were studied. These model catalyst systems showed excellent activity in the Fischer-Tropsch to olefin (FTO) reaction at high pressure. However, the Fe x O y /TiO2 catalyst deactivated more than the Fe x O y /SiO2 catalyst. After analyzing the used catalysts, it was evident that the Fe x O y -NP on titania had grown to 48 nm, while the Fe x O y -NP on silica was still 7 nm in size. STEM-EDX revealed that the growth of Fe x O y /TiO2 originated mainly from the hydrogen reduction step and only to a limited extent from catalysis. Quantitative STEM-EDX measurements indicated that at a reduction temperature of 350 °C, 80% of the initial iron had dispersed over and into the titania as iron species below imaging resolution. The Fe/Ti surface atomic ratios from XPS measurements indicated that the iron particles first spread over the support after a reduction temperature of 300 °C followed by iron oxide particle growth at 350 °C. Mössbauer spectroscopy showed that 70% of iron was present as Fe2+, specifically as amorphous iron titanates (FeTiO3), after reduction at 350 °C. The growth of iron nanoparticles on titania is hypothesized as an Ostwald ripening process where Fe2+ species diffuse over and through the titania support. Presynthesized nanoparticles on SiO2 displayed structural stability, as only ∼10% iron silicates were formed and particles kept the same size during in situ reduction, carburization, and FTO catalysis.

Introduction

Supported metal catalysts are of great importance in a wide range of catalytic processes.[1,2] The activity and selectivity of supported catalysts can be determined by the size, shape, and distribution of the active metal nanoparticles on the support, as well as metal–support interactions. As the catalytic reaction takes place on the surface of the metal nanoparticles, it is often the aim to obtain small particles, uniformly distributed on the support surface.[3] However, metal particles are typically prone to deactivation under industrially relevant conditions, prompting great efforts to design more stable catalysts.[4−7]

Many factors can contribute to catalyst deactivation, but a major cause is the loss of active metal surface area by nanoparticle growth.[8] In literature, two mechanisms are commonly used to describe the particle growth, namely, (i) Ostwald ripening and (ii) particle diffusion and coalescence. Mechanism i ensues due to the transport of atomic species diffusing from small nanoparticles to larger particles, via the support surface or gas/liquid medium. Depending on the metal, the diffusion of atomic species can be accelerated by reactive gasses such as H2, CO, or H2O.[9−13] Mechanism ii involves particle migration over the surface and coalescence to form larger particles. Different factors can affect the particle growth such as the initial particle size, particle-size distribution, and composition of the metal catalysts.[14] Furthermore, metal–support interactions have known to slow down or accelerate particle growth and can play a role in the particle growth mechanism.[15,16]

Strong interactions between the support and metal can influence the particle growth mechanisms.[17,18] Partial or complete encapsulation of metal nanoparticles by the support can occur when a reduced metal oxide is exposed to high-temperature treatments. In other cases, particles can be stabilized by the support due to the strong interaction, which may hinder the particle growth.[19,20] However, this effect can accelerate particle growth also, as atoms/ions from small metal nanoparticles can diffuse more easily over the support to form larger particles.[20] One of the supports prone to a strong interaction with metal nanoparticles is titania.

Titania is a reducible support and can go through several crystalline phases, namely, anatase, rutile, and brookite under different reduction temperatures. Due to its reducibility, an extra challenge arises in the preparation of supported metal particles on titania. Until today, many studies have focused on elucidating the interactions between titania and metal particles such as Fe, Co, Rh, Ir, Ni, Pd, and Pt.[17,18,21−25] When the heat treatments necessary to create metal particles are performed, unreducible metal titanates can form.[24,26] This was found in the preparation for cobalt and iron on titania catalysts where often, during the reduction, cobalt and iron titanates form that cannot be reduced to the metallic state at typical reduction temperatures. When the metal oxide solid solutions cannot be converted into the active metallic iron phase, this causes deactivation of the catalyst.[27−29]

Iron and cobalt are used in Fischer–Tropsch catalysis, which involves the conversion of synthesis gas (CO/H2) into hydrocarbon products. Depending on the catalyst used, the product selectivity can be tuned to either the formation of linear paraffins (cobalt-based catalysts) with high molecular weight or the production of olefins (iron-based catalysts), which is sometimes called Fischer–Tropsch-to-olefins (FTO).[26,30] Iron-based FTO catalysts are metal particle size, shape, and support sensitive.[31−33] Iron particles were previously supported on several materials such as carbon, silica, alumina, and titania where impregnation or coprecipitation is used as the synthesis method to prepare these catalysts. However, using an oxidic support such as silica, alumina, or titania can be challenging as nonreducible aluminates, silicates, and titanates can be formed during the heat treatment steps, restricting activation of the catalysts.[34−38]

An alternative synthesis method that separates the particle synthesis step from the attachment to the support is colloidal synthesis. Metal oxide particles can readily be made, and precursor solutions are not in contact with the metal oxide support. This means that the interaction of the metal particles with the support is substantially different compared to, e.g., impregnated or coprecipitated catalysts, because metal precursor contact with the support is avoided.[39−41] Using colloidal particles with the well-defined particle size, particle size distribution, shape, and composition can contribute to differentiate the catalyst properties that influence activity, selectivity, and stability.[42−44] In recent literature, colloidal synthesis methods were developed to synthesize iron oxide nanoparticles (FeO-NPs) with great control over the size and shape. These FeO-NPs can subsequently be attached to different support materials, with the objective to arrive at more stable catalysts.[33,45,46]

In this research, colloidal iron oxide nanoparticles were used to understand the growth of iron oxide supported on a silica or titania support. The colloidal iron oxide nanoparticles (FeO-NPs) were synthesized with a narrow particle size distribution and average particle sizes of 7 nm. These FeO-NPs were deposited on both a titania and silica support after which the well-distributed model catalysts were obtained (FeO/SiO2 and FeO/TiO2). The catalysts were subjected to FTO conditions where it was found that the particle size of the used FeO/SiO2 catalyst remained the same, whereas the iron oxide nanoparticles on the FeO/TiO2 catalyst had on average grown larger. This growth was further investigated for both catalysts using transmission electron microscopy (TEM) and scanning transmission electron microscopy (STEM) equipped with an energy-dispersive X-ray (STEM-EDX) detector for quantitative elemental mapping. Mössbauer, X-ray photoelectron spectroscopy (XPS), and temperature-programmed reduction (TPR) were used to obtain information about the iron species during or after hydrogen reduction. It was found that the reduction step caused the growth for FeO/TiO2 at temperatures above 300 °C. Instead, an Ostwald ripening mechanism was inferred where Fe2+ species diffuse over and through the titania support.

Experimental Section

Iron nanoparticles of approximately 7 nm were synthesized according to the procedure described by Casavola et al.[33] A 100 mL three-necked round bottom flask was filled with 1.2 mmol (0.35 g) hexadecanediol (≥ 98% purity), 0.75 mmol (0.21 g) oleylamine (70% purity), 0.43 g of oleic acid (90% purity), and 10 mL of 1-octadecene (90% purity). The flask was connected to a Schlenk line via a reflux condenser. The mixture was heated to 120 °C while stirring at 650 rpm with a glass-coated magnetic stirring bar. The flask was slowly brought to vacuum, and the mixture was degassed for 30 min. One millimole (0.21 g) iron pentacarbonyl (Fe(CO)5) (99.99% purity) was weighed in a glovebox, and 1 mL of 1-octadecene was added. The three-necked round bottom flask was flushed three times with nitrogen gas and allowed to cool to 90 °C before injecting the Fe(CO)5 solution into the heated mixture. The mixture was then heated to 290 °C with 10 °C/min. The temperature was maintained for 1 h to allow iron nanoparticle formation, and afterward, the obtained colloidal suspension was washed three times by adding five drops of toluene and a large excess of isopropanol followed by centrifugation (2700 rpm, 15 min). Lastly, the iron nanoparticles (FeO-NPs) were dispersed in approximately 2 mL of toluene. (All chemicals were obtained from Sigma-Aldrich unless stated otherwise).

The FeO-NPs were attached to a silica or titania support using a heating-up method.[33] A weight of 740 mg of the silica (Ox50 support Aerosil Evonik) or 800 mg of titania (P25 support Aeroxide Degussa Evonik) was placed in a three-necked round bottom flask, which was attached to a Schlenk line set-up via a reflux condenser. Ten milliliters of octadecene was mixed with the as-synthesized FeO-NP dispersion, and the mixture was added to the support material while stirring at 400 rpm with a magnetic glass-coated stirring bar. To remove the toluene present in the FeO-NP dispersion, the system was evacuated slowly while heating to 120 °C for half an hour. Next, the system was flushed three times with nitrogen gas and subsequently heated to 200 °C for 30 min. The mixture was allowed to cool to room temperature after which it was further processed in air. To remove all non-attached particles, the catalysts were washed six times with a hexane/acetone mixture (3: 1) followed by centrifugation (2700 rpm, 5 min). Subsequently, the catalysts were dried by a three-step drying method: at 60 °C in stagnant air for 1 h, at 120 °C in stagnant air for 3 h, and finally at 80 °C under vacuum for another 3 h. This three-step drying method was used to dry the catalysts and remove the organic ligands from the iron oxide particles as has been previously shown.[33] The FeO-NP on titania will be referred to as FeO/TiO2, and the FeO-NP on silica will be referred to as FeO/SiO2.

Characterization

Nitrogen-physisorption measurements were conducted at −196 °C using a Micromeritics Tristar 3000 to study the pore size and surface area of the support materials. The pore size distributions were determined from the adsorption branch of isotherms by the BJH method (between 1.7 and 300 nm). The total pore volume was taken as the single point pore volume at P/P0 = 0.995.

X-ray diffraction (XRD) measurements were performed on a Bruker D2 phaser diffractometer with a fixed slit using Co Kα1 radiation with λ = 1.78897 Å. The measurements were taken at angles between 20 and 80 °2θ with an increment of 0.15 °2θ and a scan speed of 0.8–1 °2θ/s.

The elemental composition of the catalysts was determined by inductively coupled plasma atomic emission spectroscopy (ICP-AES) with a Thermo Jarrell Ash model ICAP 61E trace analyzer. The samples were prepared by an aqua regia extraction where the powdered sample (approximately 125 mg) was added to 1.5 mL of HNO3 (65%) and 4.5 mL of HCl (30%) (1:3 ratio) in a destruction vessel and left overnight at 90 °C on a hot plate. The vessels were heated to 160 °C without caps until a gel was formed in less than 60 min. About 20 mL of (5%) HNO3 was added to the vessels, closed, and subsequently heated to 90 °C overnight once more. Thereafter, the vessels were weighed to determine the dilution and measured to determine the iron content.

The catalysts were imaged using (scanning) transmission electron microscopy ((S)TEM). Sample (either FeO-NP or FeO/support) was dispersed in ethanol and drop-casted on 300 mesh copper grids with formvar (FeO-NP) or the lacey carbon (FeO/support) film. The samples were investigated using a Talos F200X (ThermoFisher Scientific) equipped with an X-FEG electron source operated at 200 kV. Energy-dispersive X-ray (EDX) maps were acquired using the same apparatus with a Super-XTM EDX detector in a STEM mode. Average particle sizes were determined by measuring around 300 metal nanoparticles per sample using ImageJ program.

To prepare the samples for (S)TEM, a droplet of dispersed FeO/TiO2 in ethanol was added onto a copper lacey grid. To determine the local and average iron loading of the supported catalysts, STEM-EDX spectra were collected from ∼0.5 μm2 areas of samples. The spectra were acquired using Velox software (FEI) with a 10-min acquisition time. Ten different spectra were used to calculate the average loading of the sample, while areas from these spectra revealed the local iron loading in the regions where no iron nanoparticles were observed. Spectrum background correction and the peak analysis were performed using the software-integrated methods.

In vacuo transfer XPS using a Kratos AXIS Ultra 600 spectrometer equipped with a monochromatic Al Kα X-ray source (hν(Al Kα) = 1486.6 eV) was employed to study the iron particle growth on the titania and silica supports. Survey scans were recorded at a pass energy of 160 eV (a step size of 0.5 eV) with the background pressure kept below 5 × 10–9 mbar.

A high-temperature reaction cell (Kratos, WX-530) was used to pretreat the sample, which was supported on an aluminum stub, allowing in vacuo sample transfer into the XPS measurement chamber. The reduction was performed in a 66% H2 in Ar flow at atmospheric pressure at three reduction temperatures, viz., 240, 300, and 350 °C for 2 h with 5 °C/min. After reduction at each temperature, the sample was cooled down and subsequently transferred to the XPS analysis chamber in vacuo. Quantitative analysis was done by measuring the peak areas of specific elemental core lines (Fe2p, Si2p, and Ti2p) in the region scans and by applying appropriate atomic sensitivity factors.

Transmission 57Fe Mössbauer spectra were collected at 120 K with a sinusoidal velocity spectrometer using a 57Co(Rh) source. Velocity calibration was carried out using an α-Fe foil at room temperature. The source and the absorbing samples were kept at the same temperature during the measurements. The Mössbauer spectra were fitted using Mosswinn 4.0 program.[47] The experiments were performed in a state-of-the-art high-pressure Mössbauer in situ cell developed at Reactor Institute Delft.[48] The high-pressure beryllium windows used in this cell contain 0.08% iron impurity whose spectral contribution was fitted and removed from the final spectra. Reduction was carried out for 2 h at 240, 300, and 350 °C (H2/Ar = 2, 1 bar), and after each temperature, a spectrum was obtained.

To investigate the reduction of the catalysts, temperature-programmed reduction (TPR) was performed. A weight of 25–50 mg of catalyst (75–150 μm sieve fraction) was reduced in a 15 mL/min H2 gas flow while increasing the temperature with 5 °C/min ramp up to 900 °C. The off-gas was analyzed using a thermal conductivity detector (TCD), and certain components were analyzed using a mass spectrometer.

To investigate the effect of the reduction step, 10–20 mg of catalyst sieved to a grain size of 75–150 μm was mixed with 150–200 mg of SiC (212–450 μm sieve fraction) and transferred to a tubular glass reactor. The catalyst was then reduced in a flow of a 15 mL/min H2/Ar gas mixture (H2/Ar = 2: 1 v/v) for 2 h at temperatures ranging from 240 to 350 °C. The temperatures were reached with a heating ramp of 5 °C/min. The GHSV during these experiments was ∼10,000 h–1. A CO reduction was performed with the same samples as described above. FeO/TiO2 was dried for 2 h at 350 °C before performing the reduction step to eliminate all residual water. The catalyst was then reduced in a flow of 15 mL/min CO/Ar gas mixture (CO/Ar = 2:1 v/v) for 2 h at 350 °C.

Catalyst Performance

Fischer–Tropsch reactions were carried out using an Avantium Flowrence 16-port parallel fixed-bed reactor setup. Around 10–20 mg of catalyst sieved to a grain size of 75–150 μm was mixed with 200 mg of SiC (212–450 μm sieve fraction) and transferred to a stainless-steel reactor. The catalyst was then reduced in situ at 3 bar with a 103.3 mL/min H2/He gas mixture (H2/He = 1:2) for 2 h at 350 °C and carburized at 3 bar and 290 °C for 1 h with a 132 mL/min CO/H2 gas mixture (a ratio of 2:1). Catalysis was performed for 100 h at 300 °C with GHSV = 72,600–103,700 h–1 and 2:1 CO/H2 at 10 bar. All temperatures were preceded by a heating ramp of 5 °C/min.

Results

Iron oxide nanoparticles (FeO-NPs) were synthesized via a colloidal route and thereafter analyzed using transmission electron microscopy (TEM), see Figure A. The average particle size was 7 nm with a narrow size distribution (±1 nm), as can be seen from the histogram in Figure B. Particles dried on the TEM grid were separated by ∼2 nm corresponding to the length of the organic ligands used in this synthesis method.[49,50]

Figure 1

Colloidal iron oxide nanoparticles. (A) Transmission electron microscopy image of the colloidal particles dried on a TEM grid. (B) The histogram of the particle size distribution.

The TiO2 and SiO2 supports were analyzed using nitrogen physisorption to obtain the BET surface area, see Table and Figure S1 in the Supporting Information. Both materials had similar specific surface areas and a broad pore size distribution, with titania showing larger total pore volume and more (∼25–75 nm) mesopores than silica.

Table 1

Surface Area of both Supports, Iron Oxide Particle Size, and Iron Weight Loading

sample name	support BET surface area (m2/g)a	iron oxide particle size(nm)b	iron weight loading (wt %)c
FexOy/SiO2	45	7.3 (±0.8)	3.3
FexOy/TiO2	42	7.1 (±1.0)	2.9

The BET surface area was determined by nitrogen physisorption at −196 °C.

Particle sizes were obtained by measuring ∼300 particles from TEM images.

Iron weight loading was determined by ICP analysis.

Particles were attached to both support materials and analyzed using TEM. Samples were named FeO/SiO2 when attached to the silica support (Figure A) and FeO/TiO2 when attached to the titania support (Figure B). The FeO-NPs were homogeneously attached to the silica support, while on the titania support, empty regions were observed also, revealing heterogeneities in particle distribution.

Figure 2

TEM micrographs of the attached iron oxide nanoparticles. (A) Particles attached to SiO2 and (B) particles attached to TiO2. The insets show the histograms of the particle sizes.

The inset histograms show that the FeO-NP size distributions were similar for both catalysts and the as-synthesized FeO-NPs in Figure . The average particle sizes of the FeO-NPs and the iron weight loadings were determined by TEM and ICP, respectively, see Table . The weight loading of both catalysts was around 3 wt %, while the particle sizes remained 7 nm on average.

X-ray diffractograms were obtained for both catalysts and a blank for the TiO2 support, see Figure . The iron on silica showed a broad peak around 24 °2θ originating from the amorphous silica[49] and also peaks, which correspond to the magnetite or maghemite phase of iron oxide. The iron on titania only showed peaks from two different titania phases (rutile and anatase, in line with the expectation of this titania material[50]). For this catalyst no iron oxide peaks could be identified, which is probably caused by the strong signal from the titania overlapping with the iron oxide peaks between 42 and 75 °2θ.

Figure 3

X-ray diffractograms of FeO/SiO2, blank TiO2, and FeO/TiO2. The different phases identified in FeO/TiO2 are anatase (black square) and rutile (black diamond) (offsets of 30 counts for TiO2 and 50 counts for FeO/TiO2), while for FexOy/SiO2 amorphous silica (black circle) and Fe3O4 magnetite or γ-Fe2O3 maghemite (black arrow) were found.

The performance of the two catalysts was investigated in the Fischer–Tropsch-to-olefin reaction in which both were subjected to syngas (a CO/H2 ratio of 2:1) at 10 bar for 100 h time on stream (TOS) after in situ reduction in H2 and carburization in a mixture of H2 and CO gas. The reduction was used to both reduce the iron oxide to metallic iron and remove any residual organic ligands on the surface of these particles.[6] The activity in iron time yield (FTY) was plotted against the time on stream (Figure A). FeO/SiO2 had a higher activity compared to FeO/TiO2. For FexOy/SiO2, an activity decrease of 30% was found while FeO/TiO2 showed a 60% decrease after a 100 h time on stream. The selectivities (at time on stream (TOS) = 100 h) of both catalysts can be found in the Supporting Information, Table S1, and were comparable for both catalysts. Moreover, when an impregnated iron on the silica catalyst from previous literature is compared to the colloidally prepared catalysts in this research,[51] the colloidal model catalyst in present research showed higher activity (FTY) most probably due to the well-developed particles in the fresh catalysts.

Figure 4

FTO catalysis results and TEM images of FeO/SiO2 and FeO/TiO2 after 100 h TOS under FTO conditions at 10 bar. (A) Catalytic activities shown as FTY (iron time yield) in mol CO converted per gram iron per second as a function of time-on-stream (TOS). (B) TEM image of the used FeO/SiO2 catalyst. (C) Particle size histogram of the fresh and used FeO/SiO2. (D) TEM image of the used FeO/TiO2 catalyst. (E) Scanning transmission electron microscope (STEM) image with an energy-disperse X-ray (EDX) map overlay of the used FeO/TiO2 catalyst showing the iron in red. (F) Particle size distribution of fresh and used FeO/TiO2.

The used catalysts were investigated using TEM. The FeO/SiO2 catalyst showed only little change after the catalytic reaction, with average FeO-NP sizes of 7 nm and a standard deviation of ±0.9 nm, as can be seen in Figure B. The particle size distribution of the used silica-supported catalyst overlapped the distribution of the fresh catalyst almost perfectly (Figure C), indicating that the structural stability of this the iron nanoparticles on silica is impeccable. The 30% decrease in catalytic activity in Figure A was therefore not attributed to particle growth but most probably due to other contributing factors such as coke or iron silicate formation.[26] Surprisingly, the morphology of the FeO/TiO2 catalyst had dramatically changed after catalysis, as shown in Figure D,E. FeO-NP could no longer be distinguished from TiO2 particles in the TEM images (Figure D). An energy-dispersive X-ray (EDX) map shown as an overlay of a STEM image can be found in Figure E. Iron nanoparticles are indicated in red color and showed larger particles with an average size of 48 nm (±20 nm), see Figure F. As the decrease in catalytic activity could not be solely explained by the growth in particles, especially for the silica, the reduction pretreatment was investigated in more depth.

The reduction step of the catalysts was investigated at 1 bar. A temperature-programmed reduction measurement was done for both FeO/TiO2 and FeO/SiO2 (see Figure ) under hydrogen from 200 to 450 °C with 5 °C/min. The reduction of both catalysts initiated around 275 °C where FeO/SiO2 showed a two-step reduction. The reduction behavior of the two catalysts is relatively similar to work done in previous literature of iron on carbon supports.[41] The peak of the reduction in the TPR measurement displayed a shift toward lower temperatures for FeO/TiO2, which was consistent with previous literature as surface titanium ions can induce the reduction of surface Fe3+ species.[52]

Figure 5

Temperature-programmed reduction experiments of iron oxide nanoparticles with a ramp of 5 °C/min up to 425 °C. In green, the TPR of FeO/SiO2, while in purple, the TPR of FeO/TiO2 is shown.

To investigate the reduction step in more detail, STEM images with an EDX map (Figure ) as an overlay of FeO/TiO2 after hydrogen reduction for 2 h at either 240, 275, 300, 325, or 350 °C are shown, with temperatures increasing from left to right, respectively. At reduction temperatures of 240, 275, and 300 °C, the average particle size remained 7 nm (Figure A–C), as also shown by the particle size distribution histograms in Figure F. Increasing the reduction temperature up to 325 °C resulted in a broader particle size distribution (Figure G) and a less homogeneous particle distribution on the titania support (Figure D). Moreover, TEM images (Figure S2 in the Supporting Information) revealed that the spherical iron oxide nanoparticles from the fresh catalysts had now changed to hemispherical particles, indicating a decreased wetting angle with the titania surface, which indicates a strong interaction with the titania support.[20] Analyzing particles after a reduction of 2 h with a temperature of 350 °C, showed particles with an average of 40 nm in size and a broad particle size distribution (Figure E,G). The additional particle growth observed with the used catalysts (Figure F) could explain the decrease in catalytic activity shown in Figure A and originate from particle growth due to iron carbide formation as observed in previous research.[46] The iron particle sizes were close to the TiO2 support particle size (∼42 nm), making it difficult to distinguish iron particles from support particles using TEM as was also observed for FeO/TiO2 after FTO (Figure D).

Figure 6

STEM images and particle size distributions of FeO/TiO2 after reduction using H2/Ar for 2 h at different temperatures. (A–E) STEM images with EDX map overlay (iron in red) of reduction temperatures ranging from 240 °C (left) to 350 °C (right). (F) FeO-NP size distributions after reduction temperatures 240, 275, and 300 °C. (G) Particle size distributions after reduction at 325 and 350 °C.

EDX spectra were used to quantify the iron content on the support material in regions where no iron oxide nanoparticles were observed (for preparation details see the Experimental Section) for all the reduction temperatures mentioned above. In the EDX spectrum originating from Area 1 (Figure A), copper, iron, titanium, carbon, oxygen, and silicon could be found where copper, carbon, and silicon were signals coming from the grid used for the preparation. In areas such as Area 1, an iron signal could clearly be distinguished (Figure B), indicating that the iron was present as dispersed species below imaging resolution. The iron content (w/w) of this given area (see Figure A) was measured and divided by the Fe content detected in the whole STEM-EDX image. This gave the iron content found in the support for the bare titania support, the fresh catalyst, and the reduced catalyst at different temperatures (Figure C). The bare titania support contained no iron, while the fresh catalyst already had 20% of the initial 2.9 wt % iron present in a highly dispersed form. Moreover, an increase of this iron content was found with increasing reduction temperature up to 80% at 350 °C, meaning that roughly 2.3 wt % iron was situated on the support in a highly dispersed form. Furthermore, at this temperature, no hemispherical particles were found anymore. This showed that the increase in reduction temperature increased the iron oxide wetting onto and reaction with the titania support. From Figures and 7, it is concluded that the reduction step induced iron redistribution onto the support (∼80% of iron present) with concomitant particle growth (∼20% of iron present).

Figure 7

EDX measurements of FeO/TiO2. (A) An EDX overlay on a STEM-HAADF image indicating an area (Area 1) showing no iron oxide nanoparticles. (B) Carbon, oxygen, silicon, titanium, iron, and copper can be distinguished in the EDX spectrum from Area 1. Copper, carbon, and silicon originated from the copper lacey grid used to disperse the sample on. (C) The iron content in the particle-free support regions (w/w) is plotted for the bare TiO2, fresh FeO/TiO2, and the reduced FeO/TiO2 samples at temperatures ranging from 275 to 350 °C.

To investigate the influence of water formed during reduction with hydrogen,[5,13,53,54] the reduction of the FeO/TiO2 catalyst was performed with CO instead of H2. The histogram in Figure S3A in the Supporting Information showed once more an increase in the average particle size (∼24 nm) as well as a broad particle size distribution. The EDX map in Figure S3B revealed the larger particles, but also a low number of 7 nm particles was observed (indicated by the arrow in Figure S3A). Moreover, to assess the influence of heat treatment as such on the particle growth, the fresh catalyst was heated to 350 °C for 2 h in argon, see Figure S3C. The size histogram of the heat-treated catalyst overlapped the histogram of the fresh catalyst. As the particles still grew in a water-free environment and as the particle size stayed the same size during a heattreatment of 350 °C, this led to the conclusion that neither water nor heat was main parameter for the growth of the particles on titania but rather the reduction treatment.

XPS wide-scan spectra were measured of both catalysts after reduction with hydrogen at 240, 300, and 350 °C, see Figure S4 in the Supporting Information. Due to inelastic scattering within the samples, photoelectrons generated more than 2–5 nm below the surface do not contribute to the signal. Therefore, the surface atomic ratio of Fe/Si and Fe/Ti provide information about size and structure of supported iron phases.[55,56] For FeO/SiO2, it was found that the Fe/Si ratios stayed similar after the increase in reduction temperatures (Table ) confirming that the iron particles were not altered as also observed from the TEM images (Figure B).

Table 2

X-ray Photoelectron Spectroscopy Results with Surface Atomic Ratios of Both Catalysts after Different Reduction Temperatures

Quantitative analysis was done by measuring the peak areas of the Fe2p, Si2p, and Ti2p core lines and by applying appropriate atomic sensitivity factors in the survey scans. The estimated error of the surface atomic ratios was ±0.01 at/at for Fe/Si and ± 0.03 at/at for Fe/Ti.

Schematic representations were based on XPS, STEM-EDX, TEM, and Mössbauer data.

However, for FeO/TiO2, the Fe/Ti ratio did change with increasing reduction temperatures at an H2 atmosphere, showing an increased ratio at 300 °C where after the ratio decreased again at 350 °C (Table ). This increase in the Fe/Ti ratio indicated a higher dispersion of iron over the titania support during the reduction. We explain this by a strong interaction of iron and titania in comparison to iron and silica. After reduction at the highest temperature, the Fe/Ti ratio decreased; however, this is in qualitative agreement with the growth of particles shown by TEM (Figure E).

Mössbauer results were obtained for FeO/TiO2 and FeO/SiO2 to investigate the phase transitions of the iron species at the different reduction temperatures, which are shown in Figure S5 and Table S2 in the Supporting Information. The quadrupole splitting (QS) value in Table S2 (a measure of the charge asymmetry around the Fe atoms) can be used to identify the different iron phases and their dispersion on the surface of the catalysts at different reduction temperatures.[57] FeO/SiO2 showed a phase transition of Fe2+ and Fe3+ species toward Fe2+ and Fe0 when reduced at higher temperatures (Table S2). FeO species were already found at 240 °C, indicating that at this temperature the iron oxide nanoparticles were partially reduced.[58] Furthermore, at a reduction temperature of 350 °C, 12% Fe2+ species were observed to interact strongly with the silica support, indicating that the loss in catalytic activity found in Figure A can be explained by a limited amount of formation of iron silicates. Upon reduction of the FeO/TiO2 catalyst, Fe2+ and Fe3+ species were found also, but the QS value of Fe2+ was higher, indicating that Fe2+ was strongly interacting with the titania support already at 240 °C.[59−62] This was observed in previous literature also, as iron and titania can relatively easily form iron titanate species.[54] The Fe2+ species for FeO/TiO2 had a QS distinctive for amorphous FeTiO3 species,[63] which decreased when the reduction temperature was increased, indicating sintering of the iron into large iron (oxide) particles.[64]

TEM results indicated a spreading of 80% of the iron species onto the support as Fe2+, and Mössbauer simultaneously showed a similar amount (70%) of amorphous FeTiO3 with increasing reduction temperature. In addition, XPS showed that as the reduction temperature increases, the iron distributed onto the surface of the titania. For common growth mechanisms, particle growth is initiated by species moving over the support; however in this case, the iron is not situated on the support but has formed a new phase due to a strong interaction between the iron oxide and titania. Therefore, it is postulated that the growth of the iron particles occurred via an Ostwald ripening process, which involves Fe2+ species from smaller particles moving through the support to form larger particles (Table , last entry). The final particle size of the FeO-NP (∼40 nm) after reduction at 350 °C is close to the pore size of the titania support (38 nm) and suggests growth of the iron oxide due to Ostwald Ripening in confinement, as earlier observed for example for nickel catalysts.[65]

In literature, it is often mentioned that iron silicates can form also, just as iron titanates, under a reducing atmosphere at elevated temperatures.[26,38] Therefore, it is remarkable that in this study the iron titanates induced growth, while the small amount of iron silicates that was seen in Mössbauer did not induce any growth. When using an impregnation method, a precursor solution of iron ions is in direct contact with the support, which may give rise to iron silicate formation[51] opposed to the use of colloids where the silica is in direct contact with preformed and distinct iron oxide nanoparticles. It is hypothesized that due to this difference in synthesis method, the colloidal attached particles are more stable, considering that it might be a slow process for solid iron oxide to react with in an unreducible support.

Conclusions

In this research, colloidally obtained iron oxide nanoparticle (FeO-NP) model catalysts were synthesized and attached to either a silica support (FeO/SiO2) or a titania support (FeO/TiO2). During the Fischer–Tropsch-to-olefin reaction, both catalysts deactivated over time with FeO/SiO2 deactivating less than FeO/TiO2. When investigating the used catalysts, it was found that an initial particle size of 7 nm (±1 nm) did not change after catalysis for FeO/SiO2, indicating that the deactivation was most likely due to coke or a limited extent of iron silicate formation. However, the titania-supported iron oxide particles had grown from the initial 7 nm to much larger particles 48 nm with a broad size distribution of ±20 nm.

This significant difference in structural stability was most prominent upon reduction and further investigated using TPR, STEM-EDX imaging, and quantitative EDX measurements combined with information from Mössbauer and XPS. Interestingly, it appeared that during reduction temperatures >300 °C the iron oxide particles on average grew with the largest growth observed at 350 °C. STEM-EDX and TEM images showed only particles larger than 20 nm and dispersed iron species below imaging resolution. Fe/Ti surface atomic ratios from XPS indicated that dispersion/wetting of iron particles increased on the titania surface at 300 °C, and substantial iron particle growth took place at a temperature of 350 °C. Mössbauer spectroscopy showed that 70% of iron was in the Fe2+ form at 350 °C, with peaks of Fe2+ from amorphous FeTiO3 species. These three observations suggest that an amorphous FeTiO3 layer on the TiO2 support was formed due to the strong interactions of TiO2 with the iron oxide. The Fe2+ species on the titania support also gave rise to an Ostwald ripening mechanism where species move over and through the support during the reduction. For SiO2, this was not the case, as this support is unreducible and did not easily form a mixed oxide support to a significant extent using presynthesized iron oxide particles. Finally, this study shows that, by separating the synthesis of the iron oxide nanoparticles from the attachment step, catalysts can be obtained, which are extremely helpful in enhancing structural stability and/or in revealing mechanisms of nanoparticle growth processes.