9836
J. Phys. Chem. B 2000, 104, 9836-9841
Evidence for Structure Sensitivity in the Thermally Activated and Photocatalytic Dehydrogenation of 2-Propanol on TiO2 David Brinkley and Thomas Engel* Department of Chemistry, UniVersity of Washington, Box 351700, Seattle, Washington 98195-1700 ReceiVed: NoVember 16, 1999; In Final Form: August 10, 2000
The reaction CH3CHOHCH3 + 1/2O2 f CH3COCH3 + H2O has been studied on the (110) and (100) planes of TiO2 using molecular beam techniques. We find that the reaction can proceed through thermally activated and photocatalytic pathways to produce acetone and propene. On the (110) surface, in the absence of light the total reaction probability for an incident 2-propanol molecule, Preaction ) Pacetone + Ppropene, is 0.03. Pacetone is less than 0.01. In the presence of light with hυ > 3.2 eV, Ppropene ∼ 0, and Pacetone ) 0.15 and the reaction proceeds readily for T < 180 K, indicating that it is not thermally activated. On the (100) surface, we also observe thermally activated and photocatalytic pathways. However, the branching ratio is very different than that on the (110) surface. We observe that, for the thermally activated channel, Preaction ) 0.13, and for the photocatalytic pathway, Preaction ) Pacetone ) 0.03. Whereas the (110) surface shows a high selectivity for the photocatalytic pathway, the (100) surface shows a high selectivity for the thermal pathway. We discuss this difference in terms of the different adsorption site geometry on these surfaces. For the (100) surface sites, bridging oxygens are closer to Ti4+ binding sites than on the (110) plane. This facilitates proton transfer, which is necessary for the thermal reaction pathway. The photocatalytic pathway is dominant at the (110) surface sites because hydrogen abstraction proceeds more rapidly from the cation resulting from hole trapping than through proton transfer from the neutral molecule.
1. Introduction The potential use of TiO2 as a photocatalyst for the oxidation of organic contaminants in water and air streams has made understanding the fundamental interactions of adsorbates with this substrate essential for further progress in this field. TiO2 is an excellent candidate for photocatalysis because its valence and conduction band positions are compatible with many redox reactions suitable for the oxidation of organic contaminants.1,2 TiO2 is also chemically stable under redox reaction conditions and has a band gap that corresponds to near UV radiation energy, which is abundant in the solar spectrum. The underlying goal of this study is to obtain a better understanding of what types of surface sites are necessary for photocatalytic reactions and the role that these sites play in the reaction mechanism, using as a model reaction the photooxidation of 2-propanol. Few photooxidation studies have been carried out on singlecrystalline TiO2 surfaces.3-6 To date, most experiments involving photocatalytic oxidation of alcohols on TiO2 have been performed on powders or colloidal particles.1,7-12 Though useful for identifying and quantifying product yields, it is difficult to characterize surface structures and reactive sites in these materials. This makes understanding the role that specific sites play in the reactivity of powders and colloids difficult. However, single-crystalline studies of TiO2 in UHV offer a way to control the surface structure and therefore the reaction sites at which photooxidation occurs. One method of gaining understanding in this area is to compare the reactivity of single crystals with different surface structures such as TiO2 (110) and (100) rutile planes. It would be of greater relevance to applied photocatalytic studies if it were possible to carry out single-crystal studies on the anatase phase of TiO2. However, because the anatase to rutile transition occurs near 775 K, macroscopic single crystals in the
Figure 1. Schematic representation of the (a) (100) surface side and face view and (b) (110) surface side and face view. The solid spheres are Ti atoms while the hollow spheres are lattice oxygen atoms. The letter “b” denotes bridging oxygen atoms.
anatase phase are not available. The major difference between the (110) and (100) planes is the proximity of bridging oxygen atoms (two-coordinate O2-) to adsorbates, which can be seen in Figure 1. In both planes, bridging oxygen atoms sit above the plane of Ti4+ atoms; however, on the (100) plane, the fivecoordinate Ti4+ binding sites bind two bridging oxygens, whereas on the stoichiometric (110) plane the Ti4+ binding sites do not directly interact with bridging oxygen atoms. The effect that this difference in site geometry has on the relative rates of thermal and photochemical reactions will be addressed in this paper. The TiO2 (100) surface was chosen for this study because its structure provides a good complement to work that has been completed on the (110) plane. Like the (110) plane, the (100) plane is one of the lower energy rutile planes. This suggests that the (100) plane is likely to be one of the dominant facets of the nanocrystalline particles that make up the powders currently used in photocatalytic oxidation processes. This
10.1021/jp9940710 CCC: $19.00 © 2000 American Chemical Society Published on Web 09/28/2000
Structure Sensitivity in 2-Propanol on TiO2
J. Phys. Chem. B, Vol. 104, No. 42, 2000 9837
assumes that these powders have an equilibrium structure. In addition, the (100) surface may demonstrate a different reactivity for the photocatalytic oxidation of 2-propanol than those on the (110) surface, which we have studied previously.6,13 This direct comparison between different single-crystal structures, in conjunction with the appreciable literature for the oxidation of small organics on TiO2 powders, will enable us to better understand which surface sites participate in photocatalysis and which sites are inactive. Both surfaces also exhibit some reactivity for the thermally activated conversion of 2-propanol to acetone and propene. A further goal of this study is to determine if the branching ratio between photocatalytic and thermal reactivity is specific to the site geometry. 2. Experimental Section Experiments were carried out in a UHV molecular beam surface scattering apparatus.18 The beam portion consists of two stages. The nozzle-skimmer chamber is pumped with a 5000 L s-1 oil diffusion pump. This stage is followed by a collimationchamber pumped by a 700 L s-1 oil diffusion pump. The UHV chamber is pumped by a 510 L s-1 turbomolecular pump and a titanium sublimation pump. This chamber has capabilities for LEED, XPS, and ISS and allows dosing of a second gas through a pinhole aperture located close to the crystal. Two quadrupole mass spectrometers are mounted in the chamber. The first allows direct line of sight analysis of the beam. The mass spectrometer used to detect the scattered water is mounted in a separate chamber pumped by a 170 L s-1 turbomolecular and a titanium sublimation pump. The beam-detector angle is 90° and the skimmer-surface and surface-ionizer distances are 21 and 5 cm, respectively. Typically, both the angle of incidence of the beam and the scattering angle were 45°. None of the results reported here were sensitive to variations of these angles. The flux of 2-propanol for the molecular beam used in all experiments results reported in this work are between 7.8 × 1013 and 1.0 × 1014 cm-2 s-1, and the ratio of O2/2-propanol was (8:1). The results reported below were not sensitive to these parameters within a wide range. The photon flux of the 1000-W Xe light source in the energy range between 3.2 and 4.0 eV (greater than the band gap energy of 3.2 eV for TiO2) was 8 × 1016 cm-2. The spectral radiance in this energy region fell linearly from a maximum at 3.1 eV to very small values at 3.8 eV. Polished single-crystalline TiO2 samples were obtained in 10 × 5 × 1 mm rectangles from Princeton Scientific Corp. The (1 × 1) and (1 × 3) TiO2 (100) surfaces were prepared as follows. After mounting, the crystal was placed in the UHV chamber and heated to 900 K for several hours until it became dark blue. The best (1 × 1) surface, evidenced by sharp LEED patterns, was attained by first sputtering the sample with an Ar+ beam (1 × 10-6 Torr) for 20 min at a beam voltage of 500 eV and an emission current of 15 mA.14 After sputtering, the sample was held at 750 K and annealed in 1 × 10-6 Torr of O2 for 20 min. The oxygen was then pumped out of the chamber and the sample was allowed to cool. The (1 × 3) phase may be achieved from the (1 × 1) phase by annealing in a vacuum at 750 K for 15-20 min. This change in phase was identified both by LEED and by the emergence of emission in Ti(2p3/2) XPS between 454 and 458 eV of binding energy as seen in Figure 2. It has been established that the low binding energy tail between 454 and 458 eV of the Ti(2p3/2) peak indicates the presence of Ti3+ sites on the TiO2 rutile surfaces.15,16 The (1 × 1) phase may be recovered from the (1 × 3) phase by sputtering for 10 min followed by annealing in O2 at 750 K for as little as 5 min.
Figure 2. XP spectra in the Ti(2p1/2 and 2p3/2) emission region for the three surfaces investigated.
3. Results 3.1. Temperature-Programmed Desorption (TPD) of Reactants and Products. The absolute coverages for adsorbates on TiO2 surfaces are generally not known. Therefore, all coverages cited throughout this paper are based upon assigning one monolayer (ML) to the density of Ti4+ sites on a particular surface, which is 5.2 × 1014 molecules/cm2 for the (110) plane17-19 and 1 ML ) 7.4 × 1014 molecules/cm2 for the (100) plane. These values correspond to monolayer coverages as established for water adsorption by Henderson et al.14 The coverage for a monolayer is easily identified by the appearance of a sharp low-temperature peak in the TPD trace. Upon appearance of this feature the area beneath the TPD curve is integrated (mass spectrometer signal versus time). We assign this integrated area to one monolayer coverage for adsorbates such as water, 2-propanol, and acetone. All fraction coverages reported in this work are based upon the comparison of the integrated area under the fractional coverage TPD to the integrated area under the monolayer TPD trace. For gases such as propene and O2, the coverages were determined by calibrating the mass spectrometer at a known pressure using tabulated values for ionization gauge sensitivities and measuring the area under TPD traces relative to 2-propanol areas for which the monolayer coverage was determined as discussed above. Temperature-programmed desorption experiments were carried out with 2-propanol seeded in a helium molecular beam, which was dosed onto the TiO2 (100) single crystal at 120 K. Three significant features are evident in the TPD experiment shown in Figure 3. Initially, two high-temperature features grow in concurrently at 350 and 475 K. The higher temperature feature is quite broad and saturates rapidly with no discernible shift upon increasing coverage. The high desorption temperature of this strongly bound 2-propanol in combination with the appearance of propene produced in the same temperature region from the thermal dehydration reaction channel (see below) strongly suggest that this feature can be attributed to dissociatively adsorbed 2-propanol in which the propoxy occupies a bridging oxygen site.20 The small coverage of this high-temperature feature (∼0.10 ML) indicates that it is a minority species. The 350 K feature shifts to 280 K with increasing coverage. The saturation limit of this feature is difficult to gauge due to its overlap with other features in the TPD curves; however, dosing
9838 J. Phys. Chem. B, Vol. 104, No. 42, 2000
Brinkley and Engel
Figure 4. Acetone desorption rate (in arbitrary units) as a function of the temperature, which was increased at a rate of 5 K s-1. The curves have been offset vertically to better show the individual traces resulting from the different initial coverages shown in the figure. Figure 3. 2-Propanol desorption rate (in arbitrary units) as a function of the temperature, which was increased at a rate of 5 K s-1. The curves have been offset vertically to better show the individual traces resulting from the different initial coverages shown in the figure.
at 250 K gives a peak with a maximum of 280 K that has an integrated area equal to ≈0.40 ML. Because of the lack of spectroscopic data and the temperature range of this peak, it is unclear whether the 2-propanol desorbing in this feature is dissociatively or molecularly adsorbed. A lower temperature feature appears at 225 K and shifts slightly with increasing coverage, saturating at 210 K. The 2-propanol desorbing from this feature is associated with molecular adsorption at fivecoordinate titanium sites.17 Finally, a sharp peak develops at 160 K, which is associated with the desorption of multilayer 2-propanol. Temperature-programmed desorption experiments were also carried out with 2-propanol seeded in a molecular oxygen beam (8:1 O2:2-propanol). No new features were observed in the desorption traces, but the relative yield of different reaction products changed. By comparison, the TPD curves for the (110) surface showed similar low-temperature peaks, but exhibited no desorption above 400 K.6 TPD traces of molecular oxygen were taken with molecular beams containing only O2 and those with the O2/2-propanol ratio given above. No significant differences were observed. These TPD traces reveal a broad feature stretching from 150 to 400 K, which is peaked at 220 K. The total coverage under this peak was measured at 0.03 ML with a dosing temperature of 130 K. As for 2-propanol, oxygen TPD traces show significant differences in the TPD features for the (110) and (100) planes. In particular, there is no analogue on the (100) plane of the 450 K peak on the (110) surface that has been attributed to molecular adsorption at steps and defect sites.6,21 TPD experiments carried out for acetone/He molecular beams are shown in Figure 4. The acetone TPD curves have an initial peak that grows in at 300 K and becomes a shoulder of the second feature, which develops with increasing coverage at 250 K. As the coverage continues to increase, a third peak becomes evident at 180 K, followed by multilayer desorption of acetone at 150 K. The major difference between acetone and 2-propanol adsorption on TiO2 (100) is the lack of any state above 400 K in the acetone TPD. The absence of this feature suggests that unlike 2-propanol, acetone is not dissociatively adsorbed on either the (100) or the (110) plane.
Figure 5. These temperature-programmed oxidation (TPO) traces show the reactant, 2-propanol, and the products, acetone and propene, desorption rates (in arbitrary units) as a function of temperature, which was increased at a rate of 5 K s-1.
3.2. Thermal Oxidation. We discuss the thermal and photochemical oxidation of 2-propanol on the (100) surface separately, beginning with the thermally activated process. The total thermal reaction probability has been measured using TPD on the TiO2 (100) surface with and without molecular oxygen being present in the molecular beam. Figure 5 shows acetone, propene, and 2-propanol TPD traces obtained simultaneously from the TiO2 (100) surface, which was dosed with an oxygen/ 2-propanol molecular beam to a coverage just above one monolayer for 2-propanol. Acetone, propene, and water were the only products detected during the TPD experiment. Subsequent XPS measurements showed that the sample surface contained no carbon residues after the reaction. This was the case for surfaces dosed with oxygen beams as well as for those dosed with He/2-propanol beams. Acetone was detected in a broad desorption peak between 200 and 350 K, which was centered at 275 K. The total integrated area under this peak gave 0.03 ML of acetone produced during TPD for the 2-propanol beam seeded in He. However, the acetone yield increased to 0.09 ML, upon the introduction of O2 rather than He as the carrier gas in the
Structure Sensitivity in 2-Propanol on TiO2 2-propanol molecular beam. The thermal dehydration of 2-propanol to propene was detected in a diffuse peak between 400 and 600 K. The amount of propene produced in this reaction channel was 0.04 ML. This quantity remained constant for 2-propanol coverages as low as 0.3 ML and was not affected by the presence of molecular oxygen. The sum of the dehydration and dehydrogenation channels provides a total thermal reaction probability of 0.07 for a single monolayer of 2-propanol dosed on the TiO2 (100) surface in the absence of O2, with an increase to 0.13 in the presence of molecular oxygen. The temperatures at which acetone appears make it clear that the formation of gas-phase acetone is desorption-limited; it is quite likely that acetone is produced on the surface at lower temperatures. The temperature at which propene appears indicates that its evolution is determined by the decomposition temperature of the surface complex that liberates propene. It is likely that the surface reaction complex was formed at lower temperatures, which implies that it is not meaningful to associate the temperature with the reaction probability. We note that if the reaction probability were measured at steady state by detecting the products desorbed into the gas phase, the rate would be determined by the desorption of acetone rather than the reaction of 2-propanol to form acetone. 3.3. Photocatalytic Oxidation. The photoreactivity has been probed at steady state by measuring the 2-propanol scattered signal with and without light. These experiments were performed with both constant He/2-propanol and O2/2-propanol molecular beams incident on the surface.6 A UV light source6 was impinged on the sample, and the decrease in reflected 2-propanol signal was monitored as a function of surface temperature. The absolute conversion efficiency of 2-propanol to reaction products can be determined from these data.6 No photoreactivity was detected when the sample was dosed with a He/2-propanol beam during irradiation with UV light. This demonstrates that lattice oxygen cannot be activated by UV light under our reaction conditions. However, the inclusion of O2 in the molecular beam in concert with UV light produced a decrease in the reflected 2-propanol signal to 97% of its initial value. This corresponds to an average probability for the photoreaction on the TiO2 (100) plane of 0.03. This reaction probability was constant over a temperature range of 300-500 K. This is the range for which the appearance of gas-phase acetone is not desorption-limited and for which the surface coverage of the reactive species is not too low to limit the reaction probability. Simultaneously, acetone, CO, CO2, propene, and water were monitored as possible photoreaction products via mass spectrometry. The only photoproducts detected were acetone and water, which desorbed in a fast transient signal that decayed to a steady-state rate of production. This is the same behavior that was observed on the (110) surface.6 We have also measured the reactivity under transient conditions. The production of acetone in the integrated transient signal was found to have the temperature dependence that would be expected. At temperatures below 300 K the integrated transient signal corresponded to 0.01 ML of acetone produced. It is low because the production of gas-phase acetone is desorptionlimited. Between 300 and 475 K, the production increased to 0.03 ML, while at higher temperatures, the amount of acetone produced decreased to zero as the surface coverage of the reactants decreases.6 The quantity of acetone produced in this transient measurement is in good agreement with the steadystate photoreactivity detected in the steady-state 2-propanol uptake experiments.
J. Phys. Chem. B, Vol. 104, No. 42, 2000 9839 We note that because of signal-to-noise considerations, the probabilities for thermal and photocatalytic reactions have been measured in different ways. However, in both methods, acetone formation can only be detected at temperatures that are higher than those at which acetone is formed because the reaction is desorption-limited. The difference between the two methods lies in the surface residence time characteristic for each. The residence time is several seconds for the TPD experiment and less than a second for the beam scattering method.6 As our previous measurements of the photocatalytic reaction on TiO2 (110) showed, the average reaction probability is limited by the fraction of sites that are active.6 We found that the product of the average reaction probability and the fraction of surface sites that are active is the same for the steady state molecular beam method and for UV irradiation at 180 K followed by thermal desorption. These results show that, over the time range measured, the reaction probability is independent of the residence time. This suggests that diffusion from inactive sites to active sites is slow for 2-propanol on the strongly corrugated TiO2 surfaces. These observations lead us to conclude that the measured reaction probability for this reaction will not depend on the residence time over the range that we are investigating in this study. Our observation that the photocatalytic reaction probability measured on the (100) surface using the steadystate and the transient techniques is the same supports this assumption. 4. Discussion 4.1. Why Does the Dehydrogenation Reaction Proceed Thermally on the (100) Surface and Photocatalytically on the (110) Surface? The results obtained in this study show a total reactivity and product distribution very similar to that of the (110) rutile plane of TiO2. However, the reaction channels utilized by the two planes are completely different. As we have shown previously,6 on the (110) plane, the dehydrogenation reaction is dominated by the photoreaction channel. This results in a photocatalytic reaction probability of 15% and a thermal reaction probability of only 3%. By contrast, the reaction on the (100) plane is dominated by the thermal reaction channel, resulting in a thermal reaction probability of 13% and a photoreaction probability of only 3%. The disparity in the reaction channel utilized by these two planes can be best understood by examining the types and proximity of their adsorption sites. One of the major differences between the (110) and the (100) rutile planes is their ability to dissociate alcohols and water through proton-transfer reactions. It has been determined that dissociation of water on the (100) plane is quite efficient, up to 50% of a monolayer.14,19 In contrast, the (110) plane exhibits
