Synthesis of nanodispersed oxides of vanadium, titanium, molybdenum, and tungsten on mesoporous silica using atomic layer deposition

Transcription

1 Topics in Catalysis Vol. 39, Nos. 3 4, October 2006 (Ó 2006) 245 DOI: /s Synthesis of nanodispersed oxides of vanadium, titanium, molybdenum, and tungsten on mesoporous silica using atomic layer deposition Jose E. Herrera*, Ja Hun Kwak, Jian Zhi Hu, Yong Wang, and Charles H. F. Peden Institute for Interfacial Catalysis, Pacific Northwest National Laboratory, 999, MS K8-98Richland, WA 99352, USA The advantages of the atomic layer deposition (ALD) method for preparation of tungsten, vanadium, titanium, and molybdenum oxide catalyst supported on mesoporous silica are discussed, with emphasis on the importance of synthesis conditions on dispersion, structure and activity of the resulting materials. A suite of complementary techniques such as DRS-UV/Vis, BET, 1 H-NMR, XRD, and TEM were used to study the structural properties of the supported metal oxides, and probe reactions such as 2-butanol dehydration and ethanol partial oxidation were used to demonstrate the potential advantages of the ALD-prepared catalysts. Specifically, highly dispersed oxides of titanium, molybdenum, and tungsten oxide on mesoporous silica were synthesized using the ALD method. It is also demonstrated that attainment of high dispersions of vanadium oxide on mesoporous silica requires the presence of at least a single layer of titanium oxide due to the well-known poor interaction between vanadia and silica. The highly dispersed catalysts prepared here by ALD methods exhibited superior catalytic performance relative to those prepared using conventional incipient wetness impregnation. KEY WORDS: atomic layer deposition; tungsten oxide; vanadium oxide; titanium oxide; molybdenum oxide; ethanol oxidation; 2-butanol dehydration; UV/Vis-DRS; 1 H-NMR. 1. Introduction Nanotechnology has become a very active and vital area of research, which is rapidly developing in industrial sectors and spreading to almost every field of science and engineering. The literature continually reports new discoveries of unusual phenomena as material structures approach the nanometer scale due to substantial alteration of their fundamental physical and chemical properties. Among these materials, nano-sized metal and metal oxide particles have occupied a central place in heterogeneous catalysis for many years, long before recognition of nanotechnology. Nanotechnology has now opened the possibility to control the shape, size and chemical state of these particles. Thus, it is in the synthesis and design of a heterogeneous catalyst, that nanotechnology offers the most significant benefits such as exquisite control over the formation of the active site, the chemical environment around the active site and the binding sites to the support. In other words, control of the catalyst structure and therefore its chemical properties at the atomic scale. Among the different synthetic approaches for use of nanotechnology in heterogeneous catalysis synthesis, atomic layer deposition (ALD) processes offer exactly these potential advantages control of the catalyst synthesis at atomic level. ALD was originally introduced in the early 1970s for fabrication of polycrystalline luminescent ZnS:Mn and amorphous Al 2 O 3 insulator films [1, 2] by Suntola and coworkers, and now ALD is an established method for thin film *To whom correspondence should be addressed. jose.herrera.perea@pnl.gov deposition and conformal coating. An ideal ALD growth presupposes a very controlled deposition process in which the substrate is impregnated by the precursor(s) through chemisorption followed by surface reaction and desorption of the precursor byproducts. Several reports have been published on this subject in which the terms ALD, atomic layer epitaxy (ALE), atomic layer processing (ALP), and layer by layer growth are used almost synonymously to describe a sequential type coating process. Although ALE and ALD refer mainly to gas-phase deposition processes, the same principles of alternating adsorption and reaction of precursors can also be applied in solution. This method has been sometimes called liquid-phase atomic layer epitaxy (LPALE) or by a more common and descriptive name, successive ionic layer adsorption and reaction (SILAR) [3]. Due to its inherent surface control, ALD has become a very popular method for catalyst preparation [4] because more strict control afforded by ALD processing allows the study of interaction mechanisms and the build up of predictable structures. Indeed, the first studies involving atomic layer deposition onto porous substrates with the aim of preparing heterogeneous catalysts were carried out over a decade ago [5 10]. In all cases, materials with improved and controlled dispersion have been obtained. For example, zirconia, chromia, and vanadia have been deposited on silica and alumina using ZrCl 4, Cr(acac) 3 and VO(C 5 H 7 O 2 ) 2 as precursors, respectively [10 12]. Importantly, the rate of propylene formation in oxidative dehydrogenation of propane was shown to double on ALD-prepared catalysts relative to impregnated vanadium pentoxide catalysts [13] /06/ /0 Ó 2006 Springer Science+Business Media, Inc.

2 246 J.E. Herrera et al./synthesis of nanodispersed oxides using ALD There is an additional advantage for ALD processes since the growth on the porous surface can be expediently followed step-wise during preparation, i.e. chemisorption, bonding formation, precursor decomposition, and desorption from the surface. Because of the large surface area of the support, a relative large amount of surface species is available for the analysis so that surface reactions (e.g. ligand exchange or addition reaction) during synthesis can be studied using a variety of characterization techniques. The process can then be understood at a molecular level to obtain valuable information on the interaction with the support. For the past several years, we have been developing an ALD method in the liquid phase for grafting of metal oxide species on the surface of mesoporous scaffolds [14 16]. By keeping the system under rigorous dry conditions using organic solvents during grafting, we can avoid the presence of oligomeric species which are normally formed in aqueous solutions. Using this approach we are able to graft tungsten, molybdenum, titanium, zirconium, and vanadium oxides on the surface of silica-based mesoporous materials. Moreover, using a sequential ALD process, we are able to either generate a monolayer of TiO 2 which in turn works as an anchoring phase, important especially for dispersing vanadia species in these systems. By using this approach, we are able to achieve highly dispersed materials at very high loadings (up to 50% weight of metal oxide species). Additionally, the materials prepared show high stability toward high temperature treatments as the results of the characterization by UV Vis/DRS, Raman, X-ray diffraction, TEM, FTIR, MAS 51 V-NMR and MAS 1 H- NMR clearly indicate. In the present contribution, we report a comparative study of a series of such ALDprepared catalysts with conventional impregnated samples on mesoporous supports (MCM-41 and SBA-15). Their catalytic activities were compared using, selective oxidation of ethanol and dehydration of butanol as probe reactions. The results thus obtained clearly indicate the advantages of this novel synthesis method for the control of the morphology of metal oxide domains on mesoporous supports, as well as their improved catalytic activity and stability. 2. Experimental 2.1. Materials synthesis MCM-41 mesoporous silica was obtained from Mobil. The resultant BET surface area and the average pore size of the MCM-41 material after calcination at 500 C for 2 h are 690 m 2 /g and 3.4 nm respectively, as determined by N 2 adsorption. The SBA-15 material was synthesized using a protocol previously reported [17], and has an 860 m 2 /g surface area and an average pore size of 7 nm after calcination at 500 C for 4 h. For dehydration of the mesoporous silica surface, the supports were suspended in anhydrous toluene and refluxed for 3 h under a N 2 atmosphere. Tungsten precursor solutions were prepared by first dissolving WCl 6 in approximately 150 ml of pure toluene at RT, followed by addition of 20 ml of ethanol. The solution was then refluxed, with N 2 bubbled through the liquid phase to avoid the presence of water. The reflux continued until no HCl was detected in the nitrogen exhaust. The amount of WCl 6 added to this organic solution was varied in order to obtain different loadings of tungsten oxide on SBA-15. The maximum loading for a single deposition is based on the assumption that three silanol groups on the mesoporous silica can hydrolyze one WCl 6 molecule. With this assumption, one monolayer of WO x (100% coverage) corresponds to 30 wt.% of WO 3 in SBA-15. Samples at 25% (7.5 wt.% WO 3 ), 50% (15 wt.% WO 3 ), 75% (22.5 wt.% WO 3 ), 100% (30 wt.% WO 3 ) and 200% (60 wt.% WO 3 ) monolayer coverage were prepared. After cooling the tungsten precursor solution to room temperature, the dehydrated mesoporous silica in toluene solution was added and the mixture refluxed overnight in a N 2 atmosphere. The reaction mixture was filtered and washed with toluene several times until no WCl 6 could be detected in the washing solvent, and the solid was then collected and dried in an oven at 120 C for 30 min. This sample was designated as as-synthesized. The assynthesized sample was calcined in flowing dry air at 400 C and 500 C respectively, to generate two different calcined samples. Two additional samples of SBA-15 impregnated with an aqueous solution of ammonium metatungstate at the incipient wetness point were also prepared for comparison. The amount of ammonium metatungstate used corresponds to 22.5 and 30 wt% loadings of tungsten oxide as described above. These samples were also calcined at 400 C in flowing dry air. For the case of titania, titanium (IV) isopropoxide was added to the mixture after refluxing MCM-41 in toluene following the procedure described above. The amount of titanium precursor added corresponds to the one needed to get a 1.33 TiO 2 units/nm 2 on the MCM- 41 surface (10 wt.% TiO 2 ). In this case, the reaction mixture was refluxed for an additional 15 h under nitrogen. After cooling to room temperature, the reaction mixture was filtered and washed with toluene several times until no titanium isopropoxide could be detected in the washing solvent, and the solid was then collected and dried in an oven at 120 C for 2 h. Finally, the sample was calcined at 400 C in air. The resultant solid was refluxed again in toluene and then vanadium triethoxide was added to the toluene suspension. The amount of VO(OEt) 3 added corresponds to the one needed to get a loading of 12 wt.% V 2 O 5. The mixture was refluxed, filtered and calcined at 400 C. This sample is referred to as VO x /TiO 2 /MCM41. Two additional samples of vanadium oxide on MCM-41 were prepared without the titania interfacial layer. The first sample we

3 J.E. Herrera et al./synthesis of nanodispersed oxides using ALD 247 prepared using the ALD deposition as aforementioned, and is referred to as ALD-VO x /MCM41. The second sample was prepared by conventional impregnation using vanadyl acetylacetonate and ethanol as a solvent; this sample is referred to as RI-VO x /MCM41. In all cases the amount of vanadium loaded was the same as in the VO x /TiO 2 /MCM41 sample. For the case of molybdena, MoOCl 4 was added to the mixture after refluxing SBA-15 in toluene following the same procedure as described for MCM-41 dehydration. The amount of MoOCl 4 added corresponds to the one needed to get a loading of 20 wt.% MoO 3. The reaction mixture was refluxed for an additional 15 h under nitrogen. After cooling down to room temperature, the reaction mixture was filtered and washed with toluene several times until no MoOCl 4 can be detected in the washing solvent, and the solid was then collected and dried in an oven at 120 C for 2 h. Finally, the sample was calcined at 400 C in air. An additional sample of SBA-15 impregnated with an aqueous solution of ammonium heptamolybdate at incipient wetness point was also prepared for comparison purposes. This sample was also calcined at 400 C in flowing dry air. Both the ALD and regular impregnated samples have identical loading of molybdenum oxide XRD The X-ray diffraction experiments were conducted on a Philips PW3040/00 X Pert MPD system equipped with a Cu source (k = A ), a vertical h-h goniometer (220 mm radius), and focusing optics (Bragg-Brentano geometry). The specimens were mounted in an alumina cavity-type holder (18 mm ) for analysis. Data analysis was accomplished using JADE (Materials Data, Inc., Livermore, CA) as well as the Powder Diffraction File database (2003 Release, International Center for Diffraction Data, Newtown Square, PA) TEM Transmission electron microscopy was performed on a JEOL 2010 high-resolution analytical electron microscope operating at 200 kv. An attached energy dispersed X-ray (EDX) analyzer was used for elemental analysis. For the TEM analysis, a suspension of the powdered sample was deposited on a lacey carboncoated copper grid BET measurements A Micromeritic Tristar Gas Adsorption system was used to measure the BET surface area using N 2 adsorption isotherms. The pore size distribution was calculated from the N 2 desorption isotherms using the Barrett Joyner Halenda method. All samples were degassed in flowing air at 150 C for 4 h before the adsorption experiments H-NMR Solid-state 1 H-NMR spectra were measured using a Varian/Chemagnetics CMX Infinity 300 MHz NMR equipped with a Varian/Chemagnetics 7.5 mm HX MAS probe, operating at a spectral frequency of MHz. All spectra were externally referenced to tetramethylorthosilicate at 0 ppm and were obtained using 1 s recycle delay and 5 khz spinning rate UV/Vis Optical absorption DRS experiments were conducted on a Varian Cary 5G UV-Vis-NIR spectrophotometer with an internal integration sphere for diffuse reflectance measurements. The powder samples were loaded in a quartz cell and were measured in the region of nm. All fresh samples were dried at 120 C before measurement. A halon white (PTFE) reflectance standard was used as baseline unless otherwise indicated Catalytic testing The catalytic properties of the V and Mo containing samples were evaluated for partial oxidation of ethanol, while the tungsten containing samples were evaluated for 2-butanol dehydration. The experiments were carried out at atmospheric pressure in a micro tubular fixed-bed reactor connected to an on-line Agilent 3000 micro GC, with a temperature conductivity detector and a Plot Q column. Typically around 40 mg of catalyst were dispersed on a quartz frit and held in the middle of a quartz flow reactor (0.7 cm i.d.). All samples were treated in flowing dry air (0.8 cm 3 /s) at 400 C for 1 h prior to catalytic measurements. For the partial oxidation experiments, the catalyst bed temperature was first lowered to 130 C, and then a syringe pump was used to inject ethanol into a flowing mixture (80 sccm) of 5%O 2 /He (Airgas) at kpa and 100 C to give a constant ethanol pressure (0.85 kpa). The outlet of the reactor to the micro GC was heated to avoid condensation of the products. The catalytic activities were measured in a temperature range of C. Ethanol conversion and product selectivity were calculated on a carbon atom basis, expressed as molar ratio of ethanol transformed to ethanol fed, and ratio of ethanol transformed to each product to total ethanol transformed, respectively. In the case of 2-butanol dehydration, the catalyst bed temperature was first lowered to 100 C after the pretreatment in air. 2-butanol was then introduced into the reactor by controlled injection using a syringe pump into a flowing (80 sccm) He stream at kpa and 100 C to give a constant 2-butanol pressure (0.5 kpa). All activity measurements were made after stabilizing the system for 1 h under reaction conditions.

4 248 J.E. Herrera et al./synthesis of nanodispersed oxides using ALD 3. Results and discussion 3.1. BET measurements The porosity data measured on some of the samples under study are summarized in table 1. As mentioned above, the surface area for the bare MCM-41 support is close to 690 m 2 /g. Based on the metal oxide loading, the surface areas for the WO x /MCM41, VO x /MCM41 and TiO 2 /MCM41 samples are expected to be 586, 605, and 619 m 2 /g, respectively (assuming a negligible contribution of the tungsten oxide, titania, and vanadia phases). However, the measured surface area for these samples was 557, 622, and 591 m 2 /g, respectively. This result indicates that in the case of the VO x /MCM41 sample, the addition of vanadia results in a small increase to the overall surface area. The difference (17 m 2 /g) might indicate a contribution of a segregated, particulate vanadium oxide phase. A contrasting behavior is observed in the case of the WO x /MCM41 and TiO 2 /MCM41 samples where, as clearly observed from table 1, in this case the surface area obtained is lower than expected. In the same way, the VO x /TiO 2 / MCM41 sample shows also a decrease on the surface area (490 m 2 /g vs. the expected value, 546 m 2 /g). These results point to a loss of internal surface of the mesoporous silica. This might be related to either pore blocking, or to a homogeneous monolayer-like covering of the inner pores with either a tungsten oxide or a titania/vanadia phase. To identify the reason for this apparent loss in surface area, we performed a detailed investigation of the pore size distribution from the N 2 desorption isotherms using the classical Barrett Joyner Halenda method. The results obtained on each of the samples are shown in figure 1 and are also summarized in table 1. Several observations can be made from a detailed analysis of the results. First, notice that the VO x /MCM41 sample has almost the same pore size distribution as the bare MCM-41 material, indicating that in this case the vanadia phase is not uniformly incorporated inside the mesopores. This result is in agreement with the measurements of the BET surface area which also pointed to the presence of a V 2 O 5 phase segregated from the silica support. A contrasting behavior is observed when the Figure 1. Pore size distributions for MCM-41, WO x /MCM41 (15 wt.% WO 3 ), VO x /TiO 2 /MCM41 (10 wt.% TiO 2, 12 wt.% V 2 O 5 ), VO x /MCM41 (12 wt.% V 2 O 5 ) and TiO 2 /MCM41 (10 wt.% TiO 2 ). All samples were prepared by ALD. pore size distribution obtained for the WO x /MCM41 and TiO 2 /MCM41 samples are compared to the one obtained for bare MCM-41. In this case, a clear decrease in the average size of the mesopores is apparent. This together with the decrease in surface area observed in the BET measurements strongly indicates that the WO x and TiO 2 phases are more uniformly dispersed inside the mesopores of the support. Moreover, this result rules out blocking of the mesopores as a reason for the observed decrease of the surface area as well. The results obtained for the VO x /TiO 2 /MCM41 sample require further consideration. First, when the pore size distribution for this sample is compared to the one obtained for TiO 2 /MCM41 there is not a further shift to lower values but an increase on pore size although there is a clear decrease in the surface area when vanadium is incorporated. The reason for this apparent contradicting behavior may be linked to the overall surface coverage of titania and vanadia in this particular sample. The surface coverage of TiO 2 loaded into the MCM-41 support corresponds to 1.33 TiO 2 units/nm 2 of silica (8.0 TiO 2 units/nm 2 are required for a complete monolayer coverage). The loading of vanadia to this sample corresponds to 1.33 VO x units/nm 2 as Table 1 BET-surface area results obtained for different samples Sample Expected surface area (m 2 /g) Surface area (m 2 /g) Pore volume (ml) Pore size a (nm) MCM VO x /MCM41 (12 wt.% V 2 O 5 ) WO x /MCM41 (15 wt.% WO 3 ) TiO 2 /MCM41 (10 wt.% TiO 2 ) VO x /TiO 2 /MCM41 (10 wt.% TiO 2, 12 wt.% V 2 O 5 ) The values reported as expected surface area are calculated based on the wt.% of support in the sample and assuming negligible contribution of the tungsten oxide, titania, and/or vanadia phase to the overall surface area. a Calculated from the desorption isotherms.

5 J.E. Herrera et al./synthesis of nanodispersed oxides using ALD 249 well; it is clear that this additional amount of vanadia is not enough to create a complete monolayer coverage on the support. Van Hengstum et al. [18] have calculated the cross-sectional area of a VO 2.5 group to be nm 2. If we assume that the value for the crosssectional area of a TiO 2 unit is similar, then only approximately 15% of the mesoporous silica surface is covered with TiO 2 before vanadia deposition. Thus, the addition of VO x might not lead only to the formation of a second layer of vanadium oxide on top of the TiO 2 layer but also to dewetting of the silica surface by titania, possibly explaining why an increase of the average pore size is observed UV/Vis spectroscopy of the fresh catalysts We have used UV/Vis-DRS to get a deeper insight on the state of the vanadia, molybdena, and tungsten oxide species present on the samples. The information obtained on the band energy gap is particularly useful to evaluate the dispersion and local structure of d 0 transition metal compounds [19 22]. Several methods have been proposed to estimate the band energy gap of these materials by using optical absorption spectroscopy. A general power law form has been suggested by Davis and Mott [23], which relates the absorption coefficient with the photon energy. The order of this power function is determined by the type of transition involved. In the particular case of tungsten metal oxides, Barton et al. [19] have recommended using the formalism of an indirect allowed transition and therefore to use the square root of the Kubelka Munk function multiplied by the photon energy. By plotting this new function versus the photon energy, the position of the absorption edge can then be determined by extrapolating the linear part of the rising curve to zero. The values thus obtained carry information about the average domain size of the oxide nanoparticles since, as in the case of a particle in a box, the energy band gap decreases as the domain size increases [24]. Based on the position of the absorption bands, a relative comparison can be made between the energies of the samples under investigation and those of references of known domain size. This type of analysis is shown in figure 2, which compares the absorption edges values obtained for several WO x species of known domain size together with those of nine SBA-15-supported catalysts with different tungsten loadings and thermal treatments. For the case of the samples obtained after calcination at 400 C, the values obtained ( ev) lie closer to the values corresponding to (NH 4 ) 10 W 12 O 41, indicating an apparent octahedral coordination environment for these samples. Moreover, the variation on the edge energy values clearly indicates that the dispersion changes with loading. From this comparison it can be inferred that the W species in these catalysts have relatively small domain sizes even at a WO 3 loading as high as 60 wt.%, a result in agreement with the TEM and XRD observations (see below), which indicated a high degree of dispersion of tungsten oxide species over the SBA-15 support. In fact, previous studies by Barton et al. [19] have suggested that for edge energy values above 3.5 ev, the WO x species do not interact with each other to form bridging W O W bonds and that these species exist in a distorted octahedral symmetry. Figure 2b compares the edge energy values obtained for the sample with 30 wt.% WO 3 after three different thermal treatments. First, as mentioned above, the edge energy value obtained (3.2 ev) for the sample after calcination at 400 C corresponds to highly dispersed WO x species. In the case of the sample treated at 500 C a shift to lower energies is observed in the edge energy value. This is consistent with the formation of slightly larger WO x domains, probably through condensation (formation of bridging W O W bonds) at the expense of W-OH sites as observed by NMR spectroscopy [25]. Figure 2. Edge energy values obtained for different WO x /SBA15 catalysts as a function of (a) tungsten loading and (b) calcination temperature, for ALD ( ) and regular impregnated ()) samples. The values corresponding to analytical references of known domain size are also included for reference as dashed lines in the plot.

6 250 J.E. Herrera et al./synthesis of nanodispersed oxides using ALD Nevertheless, W certainly remains in a high state of dispersion as the values of the edge energies still remain close to 3.1 ev. A contrasting behavior is observed when this same sample is treated at 800 C. The optical absorption spectra show two different regions. A tail observed at values below 3.0 ev is a clear indication of the formation of stoichiometric WO 3 species, which agrees with the XRD observation (see below). However, part of the optical absorption spectra also showed an edge energy value close to 3.0 ev, which indicates that some of the tungsten species remained relatively well dispersed even at such a high temperature. While for the case of tungsten oxide species an indirect allowed transition formalism yields the best way to obtain edge energy values, in the case of vanadium oxide species it seems that this choice is based mostly on the best linear fit of the energy gap curve [26 28]. Wachs and coworkers have suggested using a direct allowed transition formalism in the Davis and Mott correlation [27] as the best way to obtain a linear fitting. The edge energy values obtained for the vanadia-containing samples by using this correlation are shown in figure 3. Here the results are compared to the values for some reference samples of known domain size obtained by Gao and Wachs [27]. This comparison supports the view of an extremely high dispersion for the sample obtained using TiO 2 /MCM41 as the support. Indeed, if the correlation developed by the same authors is used to obtain the average number of covalent V O V bonds (CVB), a value of 1.2 is obtained for the VO x /TiO 2 /MCM41 sample. In contrast, the energy band gap value obtained for two VO x /MCM41 samples prepared by ALD and RI corresponds to an average number of CVB close to 3.5, clearly indicating how the dispersion of the VO x phase is affected by the existence of the titanium oxide phase, incorporated through an ALD process. For the case of molybdena species, Weber [24] has proposed to use a direct allowed transition formalism as well. The edge energies values obtained by using this correlation on molybdena containing samples are shown in table 2. Here the results are compared to the values obtained by Weber on samples of known domain size. Figure 3. Average number of covalent V O V bonds present on different samples based on the correlation proposed in ref. [27]. The values corresponding to analytical references of vanadia are also included (adapted from ref. [27]). Table 2 Average number of nearest Mo neighbors present on different samples based on the correlation proposed in ref. [24] Sample Gap energy Number of nearest Mo neighbors MoO (Tetrabutyl ammonium) Mo 6 O 19 (NH 4 ) 6 Mo 7 O (Tetrabutyl ammonium) Mo 2 O 7 Na 2 MoO RI-MoO x /SBA ALD-MoO x /SBA As expected, the values for the band gap energies in the reference series decrease as the domain size increases. Those of the ALD samples lie between those of Mo 7 O 24 )6 and MoO 3, while the value for the sample obtained by conventional impregnation reveals the presence of largely agglomerated molybdenum oxide species. From this comparison, it can be inferred that the Mo species in the sample obtained by atomic layer deposition have relatively small domain sizes, although the presence of a small contribution of crystalline MoO 3 species could not be ruled out since a small tail was observed in the absorption spectrum (not shown) at values below 3.0 ev Solid-state 1 H-NMR We have used high-resolution NMR-MAS (magic angle spinning) to monitor catalyst preparation in order to get a clear insight on the atomic layer deposition processes taking place on the surface during catalyst synthesis. We have performed a thorough characterization of these materials during and after synthesis using 1 H and 51 V-NMR. Here we illustrate the results obtained while monitoring catalyst synthesis of the WO x /SBA15 catalysts; the reader is referred to the original references for a more comprehensive discussion [25, 29]. Figure 4 shows the 1 H-NMR spectra obtained for the bare SBA-15 mesoporous silica, the as synthesized 30%WO x /SBA15 catalyst, and after a calcination step at 400 C. Several features can be observed. First, the NMR spectrum obtained on the bare SBA-15 shows a very sharp peak at about 1.6 ppm which clearly indicates the presence of isolated silanol (Si- OH) groups [30]. The small tail at higher fields can be assigned to the combined contribution from both the silanol groups that are hydrogen bonded to physisorbed water and the adsorbed water molecules that are hydrogen bonded to silanol groups [25]. The spectrum obtained on the as synthesized sample, right before calcination, shows four peaks. The chemical shift values observed are located at 0.75, 2.0, 3.5,

7 J.E. Herrera et al./synthesis of nanodispersed oxides using ALD 251 Figure 4. 1 H-NMR-MAS spectra of (a) SBA-15, and of WO x /SBA15 (30 wt%. WO 3 ) after two subsequent treatments: (b) as-synthesized, and (c) calcined at 400 C. The left panel shows a schematic description of the state of the catalyst surface at each stage. and 6.9 ppm, respectively. The peaks centered at 2.0 and 6.9 ppm are assigned to methyl and aromatic protons respectively, which originate from residual toluene in the sample. The origin of the band at 3.5 ppm is related to hydrogen-bonded water on silanol groups as explained above. Notice that in this case a band at 1.6 ppm corresponding to free silanol groups is not observed and instead a very sharp peak at 0.75 ppm was detected. This peak at 0.75 ppm is assigned to tungsten hydroxyl groups (W-OH). This assignment is based on previous reports that the chemical shifts of Al-OH and Mg-OH hydroxyl protons are )0.3 ppm [31] and 0.5 ppm [32], respectively. Moreover, this assignment was further confirmed by changing the sample composition by increasing the tungsten loading as shown below. The disappearance of the peak related to Si-OH groups and the appearance of a peak assigned to W-OH species seem to confirm a full coverage of the silanol surface by tungsten oxide moieties inside the pore channel in the as synthesized sample. Finally, the NMR data obtained for the sample after calcination at 400 C in dry air shows three sharp peaks at 0.75, 1.61, and 3.5 ppm. As discussed above, the band at 1.61 ppm originates from the hydrogen of the isolated silanol groups. The reemergence of these Si-OH protons indicates that during calcination some of initial tungsten species coalesced forming clusters of WO x although they remain highly dispersed as the optical absorption data showed. This change in dispersion in turn produced domains of bare silica surface, leading to a regeneration of isolated silanol groups. Nevertheless, the presence of a strong W-OH peak is additional evidence that the tungsten oxide species on mesoporous silica are highly dispersed. To get a semiquantitative assessment of the dispersion and amount of tungsten oxide species covering the surface of the support, we performed a detailed 1 H-NMR analysis on four samples synthesized by ALD with different loadings of tungsten oxide right after a calcination step at 400 C in air. Figure 5 shows the ratio of the integrated area of the peak centered at 0.75 ppm (assigned to W-OH protons) relative to the integrated area of the peak centered at 1.61 ppm (isolated silanol protons) obtained from the 1 H-NMR spectra as a function of the tungsten oxide loading. It can be observed that this ratio increases with the tungsten loading indicating an increase in the amount of tungsten hydroxyl groups present on the surface relative to isolated Si-OH groups. This is an unambiguous indication of the increased consumption of the silanol groups by formation of tungsten oxide domains as the loading of WO x species increases in the sample. Moreover, the presence of a relatively high concentration of tungsten hydroxyl groups is a clear sign of the high dispersion of tungsten oxide species. Indeed, if a significant amount of crystalline tungsten oxide clusters were formed on the surface through condensation processes involving formation of W O W moieties, the value of the ratio of the intensity of the peak at 0.75 ppm to that at 1.6 ppm would dramatically decrease, which is clearly not the case even at loadings up to 30 wt.% of tungsten oxide XRD The X-ray diffraction patterns for as-synthesized 30 wt.% WO x /SBA15, 400 C calcined 30 wt.% WO x / SBA15, 20 wt% MoO x /SBA15, and 10 wt.% TiO 2, 12 wt.% V 2 O 5 VO x /TiO 2 /SBA15 samples are shown in figure 6. All samples were prepared by ALD. Several features can be observed. For instance, in figure 6a it is easy to identify the (100), (110), and (200) low angle diffraction peaks, that are assigned to the hexagonal Figure 5. Values obtained for the ratio of the integrated area of the peak at 0.75 ppm (W-OH protons) relative to the integrated area of the peak at 1.61 ppm (isolated Si-OH) as obtained from the 1 H-NMR spectra for WO x /SBA15 samples as a function of the tungsten oxide loading.

8 252 J.E. Herrera et al./synthesis of nanodispersed oxides using ALD Figure 6. XRD patterns obtained for WO x /SBA15 (30 wt.% WO 3 ) as-synthesized and after calcination at 400 C, MoO x /SBA15 (20 wt.% MoO 3 ) calcined at 400 C, and VO x /TiO 2 /SBA15 (10 wt.% TiO 2, 12 wt.% V 2 O 5 ) calcined at 400 C. structure of the mesoporous silica materials, in all of the patterns, clearly indicating that the nearly amorphous materials are characterized by a short-range order. All of the samples show this same characteristic, demonstrating that the mesoporous structure of the SBA-15 support is preserved after the synthesis process. More important, as figure 6b clearly shows, diffraction peaks are not observed at higher angles for any of the samples. These results clearly indicate that the tungsten oxide, molybdena, titania and vanadia phases are highly dispersed on the surface of the SBA-15 support, even at high metal oxide loadings. On the other hand, in the diffractograms obtained on the samples prepared by conventional impregnation (not shown), diffraction peaks at higher angles were observed. Comparison of these peaks with previously published works indicate that the peaks corresponded to crystalline phases of WO 3 [33], V 2 O 5 (JCPDS ) and molybdenum oxide (JCPDS ) TEM Transmission electron microscopy (TEM) images (including different sample orientations) for some of the materials obtained by ALD are presented in figure 7. Several observations can be made from a detailed analysis of the TEM micrographs. First, the TEM images clearly reveal well-ordered hexagonal arrays of mesopores (1D channels) for all the samples, which are characteristic of the mesoporous MCM-41 and SBA-15 materials [34]. Furthermore, no vanadia, titania, molybdena, or tungsten oxide particles can be observed on either the external or internal surfaces of the support in the samples prepared by ALD. This certainly reveals that these metal oxide phases could be readily incorporated into the channels of the support with little or no coalescence even at high metal oxide loadings. A contrasting behavior was observed, however, for the samples obtained using conventional aqueous impregnation, as exemplified for the case of a RI-WO x / SBA15 sample shown in figure 7. In this case, segregated tungsten oxide clusters larger than 10 nm can be clearly observed through the mesoporous silica structure. Although the XRD analysis did not show diffraction peaks corresponding to crystalline tungsten oxide on this sample, it is possible that these particulate tungsten oxide species lack long-range order. A more severe thermal treatment, notably calcination at 500 C, further segregated this tungsten oxide phase, as evidenced by clusters of submicron size in TEM micrographs (not shown) and the XRD peaks corresponding to crystalline tungsten oxide [35] Catalytic dehydration of 2-butanol and partial oxidation of ethanol The dehydration of 2-butanol was used to probe the acid base catalytic properties of the tungsten oxide samples, while the samples containing vanadia and molybdena were evaluated for their redox properties using partial oxidation of ethanol. For comparison purposes, catalytic activity measurements were also performed on samples prepared by conventional impregnation. In the case of the tungsten oxide containing samples, only dehydration products were detected (1-butene, cis- and trans-2-butenes for 2-butanol dehydration). For the case of 2-butanol it has been previously reported that bimolecular dehydration reactions leading to ethers are very slow compared to monomolecular pathways leading to alkenes, particularly at low alcohol concentrations [36 38]. Steady-state 2-butanol dehydration rates (per W-atom) as a function of reciprocal reaction temperature on both the ALD and conventional impregnated WO x /SBA15 catalysts are shown in figure 8a. For both samples, 2-butanol conversion increases with temperature. However, the activity of the sample obtained by ALD was markedly superior to that of the sample obtained by conventional

9 J.E. Herrera et al./synthesis of nanodispersed oxides using ALD 253 Figure 7. TEM micrographs obtained for VO x /TiO 2 /MCM41 (10 wt.% TiO 2, 12 wt.% V 2 O 5 ), MoO x /SBA15 (20 wt.% MoO 3 ), and two samples calcined at 400 C containing 30 wt.% WO 3, one prepared by ALD and another one by conventional aqueous impregnation (RI). impregnation. This behavior cannot be simply rationalized in terms of surface area, since the specific surface area and tungsten oxide loading were very similar for both samples. The same difference in activity behavior between the ALD and conventionally impregnated samples was also previously observed in the case of methanol dehydration [35], where a higher reaction temperature (300 C) is required for measurable activity than that for 2-butanol dehydration. Based on the TPO and UV/Vis characterization of the spent methanol dehydration catalysts, the superior catalytic performances of the ALD catalyst was attributed to less refractory coke formation due to more uniform acid strength and/or more improved thermal stability against sintering in the ALD sample. At the higher reaction temperatures required for methanol dehydration, the presence of water vapor (a byproduct of dehydration) is likely to enhance sintering by hydrolyzing Si O W bonds between WO x and the support, and mobilizing tungsten oxide species. The higher resistance of the catalyst prepared by ALD to this sintering process may reflect either a higher density of Si O W moieties or a more effective use of the available surfaces area of silica by ALD relative to conventional impregnation methods, resulting in an enhanced stability of these materials and, correspondingly, superior catalytic performance. The reader is referred to the original reference for a more comprehensive discussion [35]. For the case of the vanadia and molybdena catalysts, earlier studies have shown the suitability of the partial oxidation of ethanol as a model reaction for these systems. In fact, Oyama and Somorjai [39], on the basis of studies on silica-supported vanadia catalysts, have demonstrated that ethanol oxidation is a structureinsensitive reaction. Steady-state ethanol partial oxidation rates (per Mo or V atom) as a function of reciprocal reaction temperature for the molybdena and vanadia catalysts used in the present study are shown in figure 8b and c. In most of the cases shown here, total selectivity to acetaldehyde with little or negligible formation of other products such as ethylene, CO x, and acetic acid was observed at low reaction temperatures. The activity of the samples obtained by atomic layer deposition are remarkably superior to that observed for the conventionally impregnated catalysts, although in the case of the molybdena samples the difference in activity is not as dramatic. On the other hand, figure 8c also shows the activity profile for a sample obtained by a sequential ALD process in which a TiO 2 phase was grafted on the support before vanadia was dispersed. This particular sample shows the highest partial oxidation activity likely due, in part, to higher vanadia

10 254 J.E. Herrera et al./synthesis of nanodispersed oxides using ALD support by ALD, facilitating the formation of support metal oxide covalent bonds. Moreover, we have demonstrated that a sequential ALD process can either increase metal oxide loading with high dispersion, and/ or tailor the surface properties of mesoporous silica scaffolds to better graft specific metal oxides with desired catalytic activity. Using such a sequential ALD process, we synthesized a novel VO x /TiO 2 /MCM41 catalyst by dispersing VO x on a TiO 2 -grafted MCM-41 support, which exhibited superior performance for ethanol partial oxidation. Acknowledgments This work is supported by U. S. Department of Energy (DOE), Office of Basic Energy Sciences, Division of Chemical Sciences. The research was performed in the Environmental Molecular Sciences Laboratory, a national scientific user facility sponsored by the DOE Office of Biological and Environmental Research, and located at the Pacific Northwest National Laboratory. PNNL is operated for DOE by Battelle Memorial Institute under Contract# DE-AC06-76RLO Figure 8. (a) Steady-state dehydration rates (mol of 2-butanol converted per second per W atom) as a function of reciprocal reaction temperature for (n) ALD-WO x /SBA15 (30 wt.% WO 3 ) and (m) RI-WO x /SBA15 (30 wt.% WO 3 ). (b) Steady-state partial oxidation rates (mol of ethanol converted per second per Mo atom) as a function of reciprocal reaction temperature for (h) ALD-MoO x /SBA15 (20 wt.% MoO 3 ) and (n) RI-MoO x /SBA15 (20 wt.% MoO 3 ). (c) Steady-state partial oxidation rates (mol of Ethanol converted per second per V atom) as a function of reciprocal reaction temperature for( ) ALD-VO x /SBA15 (12 wt.% V 2 O 5 ), (d) RI-VO x /SBA15 (12 wt.% V 2 O 5 ) and ( ) ALD-VO x /TiO 2 /SBA15 (12 wt.% V 2 O 5 ). dispersion. In fact, this high activity may also be linked to the presence of vanadia species interacting with titania domains. Indeed, during ethanol partial oxidation studies Quaranta et al. [40] have reported a dramatic improvement in the catalytic activity of vanadia supported on titania-modified silica materials which seems to be a function of the amount of titania modifier in the catalyst. 4. Conclusions We have demonstrated that ALD is a novel and flexible synthetic approach to highly disperse oxides of tungsten, vanadium, molybdenum, and titanium on scaffolds such as mesoporous silica materials. Such ALD-prepared novel materials show superior catalytic activity and stability to those prepared by conventional impregnation. Their superior performance can be partially explained by the high degree of dispersion of the metal oxide species, as well as a stronger interaction with the support. This stronger interaction is the result of a careful grafting of the active species on the surface References [1] T. Suntola and J. Antson, US Patent no. 4,058,430 (1977). [2] T Suntola, Appl. Surf. Sci. 100/101 (1996) 391. [3] L. Niinisto, Curr. Opin. Solid State Mater. (1998) 147. [4] S. Auca, A. KytSkivi, E.-L. Lakomaa, U. Lehtovirta, M. Lindblad, V. Lujala and T. Suntola, Stud. Surf. Sci. Catal. 91 (1995) 966. [5] S. Haukka, E.-L. Lakomaa and T. Suntola, Stud. Surf. Sci. Catal. 120 (1998) 715. [6] J. Keränen, E. Iiskola, C. Guimon, A. Auroux and L. Niinisto, Stud. Surf. Sci. Catal. 143 (2002) 777. [7] E.-L. Lakomaa, A. Root and T. Suntola, Appl. Surf. Sci. 107 (1996) 107. [8] R.L. Puurunen, A. Root, P. Sarv, M.M. Viitanen, H.H. Brongersma, M. Lindblad and A.O.I. Krause, Chem. Mater. 14 (2002) 720. [9] A. Rautiainen, M. Lindblad, L.B. Backman and R.L. Puurunen, Phys. Chem. Chem. Phys. 4 (2002) [10] M. Lindblad, S. Haukka, A. Kytökivi, E.L. Lakomaa, A. Rautiainen and T. Suntola, Appl. Surf. Sci. 121/122 (1997) 286. [11] A. KytSkivi, E.-L. Lakomaa and A. Root, Langmuir 12 (1996) [12] A. Gervasinia, P. Carnitia, J. Keränen, L. Niinisto and A. Auroux, Catal. Today 96 (2004) 187. [13] J. Keränen, A. Auroux, S. Ek-Härkönen and L. Niinisto, Thermochim. Acta 379 (2001) 233. [14] Y. Wang, K.Y. Lee, S. Choi and C.H.F. Peden, Proceedings 17th NACS, Toronto, Ontario, Canada. June 3 8 (2001), p [15] S. Choi, Y. Wang, Z. Nie, J. Liu and C.H.F. Peden, Catal. Today 55 (2000) 117. [16] Y. Wang, C.H.F. Peden and S. Choi, Catal. Lett. 75 (2001) 169. [17] D. Zhao, J. Feng, Q. Huo, N. Melosh, G.H. Fredrickson, B.F. Chmelka and G.D. Stucky, Science 279 (1998) 548. [18] A.J. Van Hengstum, J.G. Van Ommen, H. Bosch and P.J. Gellings, Appl. Catal. 5 (1983) 207. [19] D.G. Barton, M. Shtein, R.D. Wilson, S.L. Soled and E. Iglesia, J. Phys. Chem. B 103 (1999) 630.

11 J.E. Herrera et al./synthesis of nanodispersed oxides using ALD 255 [20] A. Gutiérrez-Alejandre, P. Castillo, J. Ramirez, G. Ramis and G. Busca, App. Catal. A. Gen. 216 (2001) 181. [21] A. Khodakov, J. Yang, S. Su, E. Iglesia and A.T. Bell, J. Catal. 177 (1998) 343. [22] D. Wei and G.L. Haller, Proceedings 2nd Memorial G. K. Boreskov Intern. Conference, Novosibirsk, July (1997), p [23] E.A. Davis and N.F. Mott, Philos. Mag. 22 (1970) 903. [24] R.S. Weber, J. Catal. 151 (1995) 470. [25] J.Z. Hu, J.H. Kwak, J.E. Herrera, Y. Wang and C.H.F. Peden, Solid State Nucl. Magn. Reson. 27 (2005) 200. [26] A.A. Hossein, C.A. Hogarth and J. Beynon, J. Mater. Sci. Lett. 13 (1994) [27] X. Gao and I.E. Wachs, J. Phys. Chem. B 104 (2000) [28] G.A. Khan and C.A. Hogarth, J. Mater. Sci. Lett. 26 (1991) 412. [29] J.H. Kwak, J.E. Herrera, J.Z. Hu, Y. Wang and C.H.F. Peden, Appl. Catal. A (in press). [30] G.E. Maciel and P. D. Ellis, in: NMR Techniques in Catalysis, eds. A.T. Bell and A. Pines (Marcel Dekker Inc, New York, 1994), p [31] J.M. Miller and L.J. Lakshmi, Appl. Catal. A. Gen. 190 (2000) 197. [32] K.J.D. Mackenzie and M.E. Smith, Multinuclear Solid State NMR of Inorganic Materials,(Pergamon, New York, 2002). [33] Y. Wang, Q. Cheng, W. Yang, Z. Xie, W. Xu and D. Huang, Appl. Catal. A. 250 (2003) 5. [34] M.H. Lim and A. Stein, Chem. Mater. 11 (1999) [35] J.E. Herrera, J.H. Kwak, J.Z. Hu, Y. Wang, C.H.F. Peden, J. Macht and E. Iglesia, J. Catal. submitted. [36] C.D. Baertsch, K.T. Komala, Y.-H. Chua and E. Iglesia, J. Catal. 205 (2002) 44. [37] H. Knozinger and R. Kohne, J. Catal. 5 (1966) 264. [38] H. Knozinger, H. Buhl and K. Kochloefl, J. Catal. 24 (1972) 57. [39] S.T. Oyama and G.A. Somorjai, Catal. Sci. Technol. 1 (1991) 219. [40] N.E. Quranta, J. Soria, V. Cortes Corberan and J.L.G. Fierro, J. Catal. 171 (1997) 1.