Jacob A.
Moulijn
*ab,
J. Ruud
van Ommen
b,
Aristeidis
Goulas
b,
David
Valdesueiro
b,
Jana
Juan-Alcañiz
c,
Kar-Ming
Au-Yeung
c,
Leo
Woning
c and
Jaap A.
Bergwerff
c
aDelft IMP B.V, Molengraaffsingel 8, 2629 JD Delft, The Netherlands
bDelft University of Technology, van der Maasweg 9, 2629 HZ Delft, The Netherlands
cKetjen, Nieuwendammerkade 1-3, 1022 AB Amsterdam, The Netherlands
First published on 23rd May 2023
The feasibility of gas phase deposition using a Ti alkoxide precursor for precise surface modification of catalysts was demonstrated by modifying a mesoporous alumina support with a Ti oxide overcoat. Titanium tetra-isopropoxide yields a Ti oxide layer that covers homogeneously the alumina surface. Uniformity of the deposited TiO2 was verified by SEM-EDX, on both intra-particle and inter-particle levels. Only a few atomic layer deposition (ALD) cycles were required in order to obtain Ti contents with a relevance for industrial application. The pore size distribution of the overcoated catalyst support was barely affected by the coating process. Synthesized CoMo catalysts based on the Ti-alumina carrier showed up to 40% higher activity compared to a catalyst supported on pristine alumina, in hydroprocessing under industrial testing conditions. The TiO2 coating appeared to be stable, showing no agglomeration characteristics after reaction as corroborated by TEM-EDX. ALD provides a scalable route with low waste generation for the production of precisely structured TiO2–Al2O3 hydroprocessing catalyst supports.
Unfortunately, compared to aluminas, titanium oxides generally have small surface areas and a poor thermal stability,12 which negatively affects catalyst performance. The negative effect of the low surface area becomes especially pronounced at the high active metal loadings, which are required for the production of ultra-low-sulphur diesel (ULSD). Thus, it is difficult to assess the true potential of TiO2 as a support for industrial hydroprocessing catalysts. An elegant alternative design proposal involves the modification of a commercially available alumina with favourable porosity characteristics (pore volume, pore size distribution, specific surface area) with a thin overcoat of titanium oxide. In this way, the effect of TiO2 addition to hydroprocessing performance can be assessed without the interfering effect of the pore structure. One of our goals was to provide a fair comparison of titania and alumina for application as support for hydroprocessing catalysts.
Solvent-based methods are conventionally applied in the synthesis of hydroprocessing catalysts. We opted for the use of atomic layer deposition (ALD), a highly precise gas phase deposition technique, in synthesizing a TiO2–Al2O3 support of a high-quality CoMo hydroprocessing catalyst.13–17 The applicability of ALD for the preparation of heterogeneous catalysis and their respective catalytic supports has been discussed extensively in literature.13,18 For hydroprocessing, ALD was successfully implemented for the synthesis of a Pd-based catalyst19 and the modification of a Pd catalyst with a TiO2 overcoat.20 Furthermore, MoO3 active phase growth on mesoporous alumina by ALD and its application in oxidative desulphurization has been reported.21 ALD synthesis was anticipated to lead to a homogeneously distributed Ti layer, fully covering the accessible surface of the support. Moreover, contrary to conventional used synthesis methods it provides the possibility to avoid the use of solvents, thus drastically minimizing the generation of synthesis process-related waste.22
In early ALD studies, Lakomaa et al.23 first reported the growth of a TiO2 oxide coating onto a mesoporous silica catalyst support by TiCl4 and H2O. Haukka et al.24,25 implemented a single TiCl4 exposure on mesoporous SiO2 catalyst support. After thermal decomposition of the deposited species, scanning electron microscopy coupled with energy dispersive X-ray spectroscopy (SEM-EDX) of silica particle cross-sections revealed an even distribution of the TiO2 particles, with no enrichment of the outermost surface of the support.25 In the follow-up work of Lakomaa et al.,26 the homogeneity of the Cr and Ti species spatial distribution was again verified. A high ratio of Cl/Ti was observed in all cases, and residual Cl was only reported to be removable by thermal treatment with H2O vapor at 450 °C.
In addition, Haukka et al.27 described the modification of mesoporous γ-alumina by titanium tetra-isopropoxide (TTIP). Lindblad et al.28 deposited TiO2 into porous Al2O3 using TTIP and air at 450 °C. Keränen et al.29 used TTIP to modify mesoporous SiO2 with TiO2, in order to prepare a catalyst support for subsequent vanadia deposition. Treatment at elevated temperature for the ligand removal was required, similar to the work based on the TiCl4 precursor. Lu et al.30 and Yang et al.31 described the growth of TiO2 into silica gel powder, using TTIP and H2O at lower temperature (150 °C and 200 °C respectively). SEM-EDX revealed that the Ti species were deposited more effectively on the outer surface of the SiO2 support.31
Catalyst performance can be heavily influenced by the existence of impurities that can strongly absorb into the support surface. Such is the case for the Cl ligands of the Ti halide precursor. Although a ligand removal step at high temperature can eliminate the Cl impurities, ALD schemes operated at -relatively- low temperatures are deemed more feasible for large-scale catalyst synthesis. For that reason, we selected the alkoxide precursor, titanium tetra-isopropoxide for our ALD scheme. A comprehensive list of TTIP-based ALD processes for TiO2 growth is reported in the review article of Niemelä et al.32
In summary, the objective of this study was (i) to demonstrate the feasibility of TTIP as ALD precursor for the synthesis of highly-uniform overcoats on mesoporous catalyst supports at deposition temperatures that enable feasible large-scale catalyst synthesis, and (ii) to establish the superior properties of titania compared to alumina as support for hydroprocessing catalysts.
Mesoporous γ-alumina, produced by Ketjen as support for hydroprocessing catalysts, was used as the catalyst support. This material has a monomodal pore size distribution containing mesopores, with 90% of the pore volume corresponding to pore size below 12 nm, and no significant amount of macropores. The specific surface area is 271 m2 g−1, estimated by N2 physisorption. In all experiments the γ-alumina is used as particles at sieve fractions of 125–300 μm, obtained by crushing extrudates, followed by sieving. The precursors used for the deposition of TiO2 were titanium(IV) isopropoxide 98% (Strem Chemicals) and de-mineralized water (Veolia). Both precursors were contained in 600 mL stainless-steel bubblers. The bubbler of TTIP was heated at 90 °C while the water bubbler was kept at room temperature. Pressurized N2 (grade 5.0, Linde Gas) was used as carrier and fluidization gas. In all the experiments the initial mass of Al2O3 powder was 3 g, the flow of N2 was 0.8 l min−1, and the reaction temperature was 180 °C. Before starting the deposition process, the powder was pretreated by heating to the reaction temperature (180 °C) and flushing with N2 for up to 9 h to remove physisorbed water molecules from the surface.
After the ALD process, the samples were further submitted to a mild calcination at 350 °C to ensure removal of any remaining organics and transformation of titanium hydroxide into the respective oxide. Calcination was carried out in a static oven with a ramp rate of 5 °C min−1 and a dwell time of 2 h.
The Ti loading of the three samples (Fig. 1) increases linearly with the number of ALD cycles. It should be noted that the bulk Ti concentrations estimated from ICP-OES do not give information on the Ti distribution over the particle surface. From the data in Fig. 1 the growth-per-cycle (GPC) can be estimated. The GPC is the ratio of the volume of TiO2 deposited per cycle and the corresponding surface area. Fig. 1 shows that per cycle ca. 4 wt% Ti is deposited, corresponding with ca. 7 wt% TiO2. The surface area of the alumina used is 271 m2 g−1 and the density of the deposited TiO2 is assumed to be 3.8 g cm−3. We consider 1 g of product, containing 0.93 g of alumina with a surface area of 271 × 0.93 m2 g−1 and 0.07 g of TiO2 with a volume of 0.07/3.8 cm3. The corresponding GPC equals to 0.07 nm. The value of the GPC is approximately double than the GPC of a comparable ALD process that was estimated for Si wafers by ellipsometry.31 For an interpretation of the GPC value it is meaningful to compare it with the GPC value corresponding to a theoretical full coverage of the deposited material, calculated from the stoichiometry of the deposition reaction, taking place as depicted in Scheme 1.35 This scheme shows a stoichiometry of Ti/OH of one to two. For the calculation of a theoretical full coverage with TiO2 the concentration of the OH surface groups should be known. The OH concentration of alumina surfaces has been reported in literature.36,37 It is a strong function of the pretreatment temperature of the alumina. At the temperature of this study, 180 °C, the OH concentration is 10–12 OH/nm2.36 Based on this value, the concentration of TiO2 for the conditions of the ALD synthesis in this study is estimated to be 5–6 TiO2 species per nm2 alumina. The surface area of the alumina is 271 m2 g−1. From these data the TiO2 concentration per g of alumina at full coverage is calculated to be 1.36 × 1021 to 1.63 × 1021 TiO2 species per g alumina, corresponding with a TiO2 content of 18–21.5 wt% on the alumina basis. The value of the mass fraction of TiO2 per unit mass of product can be calculated according to the relationship depicted in Scheme 2. The resulting value for the theoretical full coverage with TiO2 is 15–18 wt%.
![]() | ||
Fig. 1 Titanium content (bulk, measured by ICP-OES) of the samples overcoated with 1, 2 and 3 TTIP ALD cycles. The line indicates a linear fit. |
![]() | ||
Scheme 1 Simplified interaction of a hydroxylated alumina surface with TTIP and H2O during the ALD overcoating reactions. |
![]() | ||
Scheme 2 Conversion of TiO2 concentration per g of alumina at full coverage, to TiO2 concentration per g of overcoated catalyst support product. |
As shown in Fig. 1 the Ti content after 3 deposition cycles is ca. 12 wt%, and thus ca. 20 wt% TiO2. This number shows that after 3 cycles, the amount of Ti-oxide deposited corresponds to a theoretical full coverage provided that large amounts of 3-D structures are not present. Per cycle ca. 2 TiO2/nm2 are deposited. It should be noted that the GPC value of 0.07 nm is an average over the surface.
SEM-EDX analysis of particle cross-sections was carried out to assess the extent of TiO2 deposition in the inner pore structure of the particles. A semi-quantitative Ti distribution was obtained. An impression of the Ti distribution inside the TiO2–Al2O3 particles after 1, 2 and 3 cycles is given in Fig. 2. The Ti concentration increased at increasing number of cycles. In the individual particles the distribution is rather homogeneous with a sharp, relatively concentrated layer of 10–20 μm at the edges of the particles.
For all three samples the concentration differences between particles are rather small. By combining bulk and cross-sectional Ti content measurements, the TiO2 concentration in the centre of the particles is compared to the bulk composition. Note that in this comparison the amount of Ti is expressed as the wt% of oxides in the total sample, the common way to express composition of hydroprocessing catalysts. The TiO2 concentration in the centre was measured by taking the average of several spot measurements near the centre of 10 different particles (for more details see ESI†).
The obtained line scans (Fig. 3) show the distribution of Ti species along the cross section of the three samples. Except for the aforementioned enrichment in the particle rim, the observed concentration is, in each case, essentially constant over the radial position. This enrichment rim of the particles is in all three cases about the same (ca. 20% of the total amount of the TiO2). Thus, compared to the ideal case of fully homogeneous deposition, roughly 80% of the Ti-oxide is homogeneously distributed over the inner particle surface. As shown in Fig. 4 the content values in the particle centre show close to linear correlation with the bulk concentration. They are roughly 20–30% lower than the bulk concentration.
Despite the homogeneous distribution of the deposited species, coating the inner pores could induce a significant decrease in the specific surface area of the alumina support due to pore blockage. To assess this, the pore size distribution was estimated for one of the samples coated with a single ALD cycle (7.3 wt% TiO2), using N2 physisorption (Fig. 5). Similar pore size distributions between the uncoated and coated samples are observed, with a maximum in the pore size distribution curve at the same pore diameter. The delta pore volume (normalized as ml of N2 adsorbed per g of alumina support) is only slightly larger than expected, based on the density of bulk TiO2. It is concluded that the differences between the alumina and the Ti-coated alumina are minor: no extensive pore blocking has taken place.
The effect of TiO2 deposition on the surface area of the Al2O3–TiO2 materials measured by hexane adsorption (Table 1) confirms that when normalized on γ-alumina (Al2O3) basis, no noticeable decrease is observed.
All the catalysts based on the Ti-coated supports showed significantly higher activity than the reference catalyst. Up to 40% higher relative activity in HDS per Mo atom was obtained for the sample that was modified with 2 ALD overcoat cycles. It is clear that the introduction of the highly dispersed TiO2 in the catalysts has a beneficial effect on activity, in line with previous reports. We speculate that the observed optimum in activity is related to the dispersion of the Ti species. As the Ti content increases with the number of ALD cycles, approaching the order of a monolayer coating (18–21.5 wt%), the dispersion of the Ti species might be slightly decreasing for the highest Ti loading. Further evaluation of the porosity characteristics for varied Ti contents, and a careful study on the catalyst acidity characteristics and the dispersion of the active phases are expected to enable a better understanding of the reasons behind the increase of the catalytic activity.
High resolution TEM images of spent catalysts were used to determine the morphology and dispersion of the CoMoS active phase, based on the sample after a single ALD cycle (7 wt% TiO2) as shown in Fig. 8. The CoMoS slabs can be observed as black lines, which are in fact projections of the 2-dimensional MoS2 slabs as observed side-on. Mostly single slabs of 3–7 nm can be observed, but in areas with higher local metal loading, some stacking of the slabs was evident. The slabs positioning seems to follow the contour of the support particles, in agreement with previous visualization via 3-dimensional TEM imaging.38,39 No noticeable formation of separate cobalt-sulfide particles was observed. Altogether it can be concluded that the introduction of TiO2 in the support via ALD has not resulted in any drastic modification of the active phase. Quantification of the dispersion of the active phase is notoriously difficult due to (i) sample inhomogeneity which makes it difficult to obtain a representative set of images for quantification, (ii) the potential presence of small CoMoS slabs that cannot be traced in the current TEM resolutions (and could contribute to a large amount of active sites), (iii) the large error associated to counting and size measurement of active sites, (iv) the large error associated to counting and size measurement of the observed slabs. As the main objective of this study was to determine the applicability of ALD for the deposition of TiO2 overcoats in relevant mesoporous catalyst support materials and its positive effect on catalyst activity, we did not attempt any further quantification.
![]() | ||
Fig. 8 High resolution TEM images on several sample locations (a–d) of a spent overcoated CoMoS catalyst (7 wt% TiO2 after 1 TTIP ALD cycle) obtained after the catalytic performance testing. |
The results strongly suggest that the deposition of TiO2 did not have a significant effect on the dispersion of the CoMoS phase. We conclude that the positive effect of TiO2 overcoating on the catalytic performance is not the result of an increased dispersion, in agreement with the suggestion of a modification of the support (a lower support active phase interaction) or of a direct function for titanium sites similar to Ni and Co.10,11 A modification of the electronic structure of the active phase could be the origin,5,40 while the TiO2-surface could also play a role in reaction pathways that facilitate the HDS reaction, such as isomerization of substituted dibenzothiophenes.41–43 Determining the exact effect of the TiO2 overcoat on the activity is beyond the scope of this study.
ALD of TiO2 onto porous Al2O3 supports—such as mesoporous powders typically used in catalysis applications—has already been showcased in the 1990s pioneering ALD research collaboration of Microchemistry and Neste.23–28 Further research30,31,44 has proven the benefits of ultrathin coatings in catalytic applications. This is the first report that presents extensive evidence on the intra-particle and inter-particle distribution of the Ti-oxide overcoat. A clear performance benefit, under industrially-relevant conditions, is observed and attributed to the overcoat.
Footnote |
† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d3cy00335c |
This journal is © The Royal Society of Chemistry 2023 |