Photooxidation of Acetone on TiO2(110): Conversion to Acetate via

Zhi-Tao Wang , Michael A. Henderson , and Igor Lyubinetsky. ACS Catalysis 2015 5 ... Till Cremer , Stephen C. Jensen , and Cynthia M. Friend. The Jour...
3 downloads 0 Views 183KB Size
Download PDF
12062

J. Phys. Chem. B 2005, 109, 12062-12070

Photooxidation of Acetone on TiO2(110): Conversion to Acetate via Methyl Radical Ejection Michael A. Henderson Interfacial Chemistry and Engineering Group, Pacific Northwest National Laboratory, P.O. Box 999, MS K8-93, Richland, Washington 99352 ReceiVed: February 11, 2005; In Final Form: April 20, 2005

It is generally held that radicals form and participate in heterogeneous photocatalytic processes on oxide surfaces, although understanding the mechanistic origins and fates of such species is difficult. In this study, photodesorption and thermal desorption techniques show that acetone is converted into acetate on the surface of TiO2(110) in a two-step process that involves, first, a thermal reaction between acetone and coadsorbed oxygen to make a surface acetone-oxygen complex, followed second by a photocatalytic reaction that ejects a methyl radical from the surface and converts the acetone-oxygen complex into acetate. Designation of the photodesorption species to methyl radicals was confirmed using isotopically labeled acetone. The yield of photodesorbed methyl radicals correlates well with the amount of acetone depleted and with the yield of acetate left on the surface, both gauged using postirradiation temperature programmed desorption (TPD). The thermal reaction between adsorbed acetone and oxygen to form the acetone-oxygen complex exhibits an approximate activation barrier of about 10 kJ/mol. A prerequisite to this reaction is the presence of surface Ti3+ sites that enable O2 adsorption. Creation of these sites by vacuum reduction of the surface prior to acetone and oxygen coadsorption results in an initial spike in the acetone photooxidation rate, but replenishment of these sites by photolytic means (i.e., by trapping excited electrons at the surface) appears to be a slow step in a sustained reaction. Evidence in this study for the ejection of organic radicals from the surface during photooxidation catalysis on TiO2 provides support for mechanistic pathways that involve both adsorbed and nonadsorbed species.

1. Introduction The study of radical species generated during heterogeneous photocatalysis has been of keen interest.1-6 While radicals formed from water, OH groups, or O2 have received considerable attention because of their expected participation in photolytic surface reactions, there is also evidence for the formation and reactions of organic radicals,3,5,6 usually from reactions involving OH, OOH, or O/O2 radicals and radical anions. These latter species are formed from charge-transfer events associated with charge carriers (holes and electrons) generated by photon absorption in the photocatalyst. While the surface most often serves as host to reactions of photogenerated radicals, in some cases radical species are believed to be ejected from the surface by a photon-initiated event (aka, “photodesorption”) and participate in reactions in the gas phase, the solvent or the physisorbed layer surrounding the photocatalyst. In this study, the rutile TiO2(110) surface is employed under ultrahigh vacuum (UHV) conditions to show that a primary photocatalytic pathway in acetone photooxidation on TiO2 involves ejection of methyl radicals from the surface. The TiO2(110) surface has become the prototypical oxide singlecrystal surface on which to study surface chemistry.7 Previous photocatalysis studies of organics adsorbed on this surface include alkyl halides,8-14 acetate,15 2-propanol,16-18 and trimethyl acetic acid.19-23 This work focuses on understanding the mechanism of the heterogeneous photooxidation of acetone on TiO2. Acetone is of importance in photocatalytic * To whom correspondence [email protected].

should

be

addressed.

E-mail:

remediation because it is a common organic solvent and airborne contaminant. Acetone photooxidation has been extensively studied on powder TiO2 (anatase and rutile) materials.24-44 Many of these studies employ acetone photooxidation as a means of evaluating the performance of various TiO2 photocatalyst preparations or methods. Other studies focus on the kinetics of the overall oxidation process, terminating in CO2 gas, and how trace amounts of water vapor affect these kinetics. Several groups have paid considerable attention to the reaction mechanism(s) and surface intermediates. For example, Attwood et al.34 used EPR to characterize radical species generated during photooxidation of acetone on Degussa P-25. These authors detected evidence for an organic radical of the form CH2OO• which they assigned to the peroxy species CH3C(dO)CH2OO•. This species presumably formed from reaction of O2 with the radical CH3C(dO)CH2•, which in turn was formed from a reaction of acetone with an O radical anion species. Using FTIR spectroscopy during UV illumination, Coronodo et al.27 detected buildup of adsorbed acetate as the major surface intermediate during acetone photooxidation on TiO2, with formate as a secondary species. These authors also detected acetaldehyde and formaldehyde as adsorbed intermediates. Two groups have observed photooxidation of acetone occurring through dimerized acetone species formed thermally on TiO2. Using NMR, Xu and Raftery25 observed that small amounts of mesityl oxide (CH3C(dO)CHdC(CH3)2) and diacetone alcohol ((CH3C(dO)C(OH)(CH3)2), formed from the thermal coupling of acetones, were photooxidized on TiO2 to formate. Additionally, these authors observed a second photooxidation pathway involving acetone molecules thermally converted to adsorbed propylene oxide

10.1021/jp0507546 CCC: $30.25 © 2005 American Chemical Society Published on Web 05/27/2005

Photooxidation of Acetone on TiO2(110) species that in turn were photooxidized to acetate groups. This group reached similar conclusions using FTIR.26 El-Maazawi and co-workers24 observed two channels of acetone photooxidation on TiO2 powder using FTIR, one involving mesityl oxide and the other involving molecularly adsorbed acetone. In both cases, only C1 surface intermediates were detected. In contrast to studies on TiO2 powders, results in this study on the single-crystal TiO2(110) surface provide fundamental insights into the photooxidation of acetone on a surface with a uniform surface structure and under conditions in which surface coverages (of acetone, oxygen and water) are more easily controlled. One may argue that a UHV study of a single-crystal face of the rutile form of TiO2 has nominal relevance to the conditions and materials employed under applied conditions. However, similarities between results reported here and those observed in several powder studies suggest that the insights obtained from this study have direct relevance toward understanding the photooxidation of acetone on TiO2. In particular, this study shows that coupled thermal and photochemical reactions are involved in acetone photooxidation on TiO2, and that the conversion of C3 acetone to C2 acetate involves ejection of a methyl radical from the TiO2 surface. Methyl radical photodesorption from surfaces is not uncommon in the literature, having been observed for adsorbed methyl halides8-10,45-47 and in one case for adsorbed trimethyl gallium.48 This study shows that methyl radical photodesorption is a key step in the photooxidation of acetone on TiO2.

J. Phys. Chem. B, Vol. 109, No. 24, 2005 12063 onto the tip of a single strand, 0.6 mm diameter fused silica fiber optic cable, that directed the light through a UHVcompatible feedthrough onto the crystal face. The use of a fiber optic system enabled exclusive light exposure to the crystal without exposure to extraneous surfaces (e.g., the sample holder). The photon flux for energies above the TiO2 band gap was calibrated using a cutoff filter and a photodiode detector. Photodesorption measurements were performed with UV light incident to the crystal face and with the crystal oriented at a 45° angle from the entrance to an apertured mass spectrometer. Photon exposures were started and ended using a mechanical shutter situated on the lamp’s housing. A feedback system, also mounted on the lamp’s housing, regulated the lamp in order to provide a constant photon flux during periods of prolonged exposure. Tests of the transparency of TiO2(110) crystals of similar thickness and bulk color showed almost complete attenuation of the output of the fiber optic by the crystal. However, about 30-40% of the incident light was reflected from the surface. No accounting for this amount of light reflection was made in the photochemical cross section measurements because such a correction would not change the order of magnitude of the resulting cross sections. A typical photon flux (of photons with energies in excess of the TiO2 band gap) at the crystal was on order of 1.4 × 1017 photons/cm2 s. Exposure of the TiO2(110) crystal at 95 K to this photon flux resulted in a rise in the crystal temperature of no more than 10 K. 3. Results and Discussion

2. Experimental Section The UHV chamber and techniques used in this study have been described in more detail elsewhere.49 The base pressure of the chamber was 2 × 10-10 Torr. The TiO2(110) crystal used in this study was obtained from First Reaction with dimensions of 10 × 10 × 1.5 mm. The surface was cleaned first by Ar+ ion sputtering with 3 keV ions until the major impurities were eliminated and then with 500 eV ions in order to smooth out the extensive damage caused by the higher energy ions. Both sputter treatments were conducted with the crystal temperature at 850 K in order to facilitate faster segregation of impurities and repair of the sputter damage. After sputtering, annealing at 850 K was continued for 30 min. Contaminant levels were monitored with secondary ion mass spectrometry (SIMS) during sputtering and afterward by Auger electron spectroscopy (AES). The most persistent impurity was K, which required about 15 h of sputtering to diminish its influence on surface chemistry to an undetectable level. Water TPD50 was used to detect and quantify surface oxygen vacancies resulting from reduction of the crystal via vacuum annealing and ion sputtering. After the above treatment, the crystal surface possessed an oxygen vacancy population of about 0.07 ML. This level did not significantly change with additional annealing at 850 K unless more sputtering was done with the surface hot. Research-grade acetone and d6-acetone were obtained from Aldrich and were further purified using LN2 freeze-pumpthaw cycles. Acetone was exposed to the TiO2(110) surface using a calibrated directional doser, while oxygen was passed through a LN2 trap and dosed by backfilling the chamber. Acetone TPD experiments were performed with a heating rate of 2 K/s and involved adsorption at 95 K. In this study, ‘1 ML’ is defined based on the surface site density of five-coordinated Ti4+ cation sites on the ideal TiO2(110) surface (5.2 × 1014 molecules/cm2). UV irradiation was accomplished using a 100 W Hg arc lamp. A water filter was used to remove IR. The light was focused

3.1. TPD Results. Two acetone coverages, 1 and 0.25 ML, were selected for studies on the UV photooxidation of acetone adsorbed on TiO2(110). Figure 1 shows the mass 42 TPD trace for 1 ML of acetone adsorbed on reduced TiO2(110) at 95 K without irradiation (bottom trace) and the same coverage exposed to UV light at 95 K in a background of 5 × 10-7 Torr O2 for various time periods. Mass 42 was selected because it conveys information about both the amount of unreacted acetone and amount of the immediate photooxidation product of acetone, namely acetate. This is in agreement with FTIR results by Coronodo et al.27 for the photooxidation of acetone on TiO2 powder (70% anatase). These authors also observed C1 surface intermediates (formate and formaldehyde) that likely originated from the C1 fragment leftover after conversion of C3 acetone to C2 acetate. As will be shown below, this C1 fragment is likely a methyl radical, which is lost to vacuum in these studies but likely reacts on the surface of another particle or in the gas phase with other species under high-pressure conditions employed in typical photocatalytic studies that utilize high surface area TiO2 powders. Acetate thermally decomposes on TiO2 single-crystal surfaces to gas-phase ketene at about 600 K,7 and mass 42 is the major QMS signal for gaseous ketene. No carbon-containing photooxidation product of acetone other than acetate was detected by TPD. Although the mass 42 signal from QMS cracking of acetone is only about 10% of the mass 43 fragment intensity,51 the signal at mass 42 was sufficient to track changes in the acetone coverage due to photolysis. Additionally, the desorption features from acetone and ketene did not overlap in TPD. As shown in Figure 1, UV irradiation of 1 ML acetone on TiO2(110) in 5 × 10-7 Torr O2 resulted in depletion of acetone from the surface and an increase in the surface coverage of acetate groups, which thermally decomposed to ketene in TPD at ∼620 K. The desorption profile of the unreacted acetone was slowly altered as the photolysis period extended, eventually evolving into two peaks at about 200 and 300 K after irradiation

12064 J. Phys. Chem. B, Vol. 109, No. 24, 2005

Figure 1. Mass 42 TPD spectra from 1 ML acetone adsorbed on TiO2(110) at 95 K and irradiated with UV light for various time periods in 5 × 10-7 Torr O2. The inset shows TPD peak area data from these and other data for acetone (filled circles) and ketene (empty triangles) as a function of UV exposure.

for 180 min. The splitting of the acetone desorption profile into two peaks resulted from background adsorption of water during the photolysis. The ‘valley’ position in the acetone TPD spectrum marks the location of water desorption (not shown). Previous work has shown that coadsorbed water compresses acetone into high coverage islands, some of the acetone in these islands desorbs prior to water desorption in TPD.52 (The influence of water on acetone photooxidation will be discussed in a future work.) The inset to Figure 1 shows acetone and ketene coverage changes, as gauged by the mass 42 TPD peak area, versus the UV irradiation time. The mass 42 TPD peak data for acetone were normalized to one and can be interpreted as absolute coverages since the starting acetone coverage was 1 ML. The total acetone coverage (filled circles) decreased exponentially, indicative of a first-order process, to about half the initial coverage after a 180 min UV exposure. The mass 42 TPD peak area data for ketene (open triangles) were also normalized by the same factor as that for acetone, but cannot be interpreted in absolute terms since the relationship between the acetate coverage (yielding ketene in TPD) and mass 42 TPD peak area is not known. TPD of adsorbed acetic acid on TiO2(110) was attempted as a calibration, but coincident contributions at mass 42 in the 500 to 750 K range from ketene and molecular acetic acid desorption events complicated the analysis. Nevertheless, the rate of acetate accumulation (ketene evolution) mirrored the decay curve for acetone. This result is consistent with ketene being the only detectable C-containing TPD product after UV irradiation of adsorbed acetone. The fact that the acetate yield

Henderson

Figure 2. Mass 42 TPD spectra from 0.25 ML acetone adsorbed on TiO2(110) at 95 K and irradiated with UV light for various time periods in 5 × 10-7 Torr O2.

does not attenuate with prolonged UV irradiation indicates that the rate of acetone photooxidation was at least 2 orders of magnitude faster than that of acetate photooxidation under these conditions. Figure 2 shows data similar to that in Figure 1 except for an initial acetone coverage of 0.25 ML. The mass 42 trace at the bottom of the figure is in the absence of UV irradiation. In a similar manner as in Figure 1, the mass 42 signal due to acetone attenuated and the signal due to acetate (ketene) increased as the UV irradiation period increased. (A weak, unidentified mass 42 feature was also observed at ∼140 K, presumably from adsorption of some background species on the partially empty TiO2(110) surface during UV irradiation.) Because the initial acetone coverage was low, the acetone TPD spectrum was not complicated by acetone-acetone repulsions that prevail at higher coverages53 or when water is present.52 The shift in the acetone TPD peak from ∼300 to 375 K was due to formation of an acetone-oxygen surface complex,53 which will be discussed in more detail below. It is unclear from these data whether the acetone-oxygen complex was formed during UV irradiation at 95 K or subsequently during TPD between coadsorbed acetone and oxygen. Figure 3 shows the acetone TPD peak area data from Figures 1 and 2 as a function of UV photon exposure. (See the Experimental Section for the procedure used to calibrate the photon flux.) The figure shows curves for attenuation of the two acetone initial coverages (1 and 0.25 ML) using TPD data from three QMS cracking fragments of acetone (mass 42, 43, and 58). Effective cross sections for acetone photooxidation on TiO2(110) were determined by plotting the natural log of the acetone TPD peak area remaining after each irradiation time, normalized to that from the initial acetone coverage, versus the photon exposure. For both initial coverages, all three masses

Photooxidation of Acetone on TiO2(110)

Figure 3. Acetone photooxidation cross section plots obtained from TPD data using three acetone mass spectrometer cracking signals (42, 43, and 58) at two different surface coverages (0.25 and 1.0 ML).

yielded complimentary curves. The effective cross section for photooxidation of 1 ML acetone on TiO2(110) in 5 × 10-7 Torr O2 was 3 × 10-22 cm2, whereas this value was an order of magnitude higher for an initial acetone coverage of 0.25 ML. While these values are 5 to 6 orders of magnitude lower than typical collisional cross sections (∼10-16 cm2), it is important to note that they reflect the overall cross section for the initial step in acetone photooxidation on TiO2(110), which includes photon absorption, carrier diffusion to the surface, and charge transfer processes. Since the bulk of the crystal, where presumably the majority of the charge carriers are generated, is consistent throughout these studies, changes in the effective cross section in Figure 3 can be linked to surface effects. These data suggest that the acetone photooxidation rate on TiO2(110) is sensitive to the acetone coverage. The rate, as reflected by the cross section measurements of Figure 3, decreased at high acetone coverage. As will be shown below, the most likely cause of the coverage sensitivity of the cross section comes from the inability of gas-phase O2 to interact with an organic covered surface. Previous work on trimethyl acetate (TMA) covered TiO2(110) has shown that the rate of TMA photooxidation accelerates as patches of organic-free surface are made available to gas-phase O2.20 In the case of 1 ML acetone, its photooxidation product (acetate) remained strongly bound to the surface, inhibiting O2 adsorption. Figure 3 also reveals that the cross section plots for both coverages do not intersect the y-axis at the origin, but instead at values of about -0.18 for the 1 ML case and -0.25((0.1) for the 0.25 ML. In both cases, the absence of an intersection at the origin indicates the presence of an initial cross section that was at least an order of magnitude greater than that seen at longer UV irradiation periods. The y-axis intercept values correspond to ‘instantaneous’ amounts of acetone photooxidation equivalent to ∼0.16

J. Phys. Chem. B, Vol. 109, No. 24, 2005 12065

Figure 4. Photodesorption spectra for various masses obtained during UV irradiation at 200 K of 0.75 ML acetone (lower traces) and d6acetone (upper traces) on TiO2(110) in 5 × 10-8 Torr O2.

ML for the 1 ML case and ∼0.06 ( 0.02 ML acetone in the 0.25 ML case. Since the surface pretreatment was the same for the two coverages examined, the difference in the amount of acetone photooxidized in the initial period likely stemmed from differences in availability of surface sites and/or in the degree of intermolecular interactions. One might have expected that if a set coverage of ‘special’ surfaces sites were responsible for the fast process, then the y-axis intercepts for both acetone coverages would reflect equivalent amounts of photooxidation in the fast stage. This, however, was not observed. 3.2. Photodesorption Results. Photodesorption studies were conducted to probe the fast acetone photooxidation process revealed in Figure 3 from TPD data. Although one cannot assume that photooxidation of an adsorbed species should result in photodesorption, photodesorption studies offer advantages to post-irradiation TPD measurements as a method for characterizing surface photocatalysis. First, observations of photodesorption provides insights into what bonds are being broken during photolysis. Second, it is possible to extract dynamical information with appropriately constructed pulsed light sources. Third, using postirradiation TPD to track the amount of depleted acetone or accumulated acetate can be misleading if heating the surface initiates thermal chemistry that inadvertently might be ascribed to photocatalysis. This latter point is a consideration for the acetone/TiO2(110) system because previous measurements indicate that a small fraction (∼6%) of adsorbed acetone thermally decomposes when coadsorbed with oxygen.53 Figure 4 shows a series of photodesorption traces from UV irradiation at 200 K of 0.75 ML d0-acetone (lower traces) and 0.75 ML d6-acetone (upper traces) adsorbed on TiO2(110), both in 5 × 10-8 Torr O2. (The choice of a 200 K substrate temperature is significant and will be discussed below.) The time designated as ‘0 s’ corresponds to the start of UV exposure, with the light blocked from the sample at 300 s. In both

12066 J. Phys. Chem. B, Vol. 109, No. 24, 2005 experiments, the surface was dosed with a multilayer acetone coverage at 95 K, and the temperature was then ramped to and held at 200 K in 5 × 10-8 Torr O2 for 4 min prior to UV irradiation. This treatment resulted in a reproducible acetone coverage of about 0.75 ML. The time resolution of the data acquisition was low (cycle time on the order of 250 ms), and the time period for opening the shutter to expose UV to the crystal was about the same. Data in Figure 4 shows several photodesorption ‘spikes’ on exposure of the surface to UV light. For the d0-acetone (hereafter referred to as ‘acetone’) case, intense spikes were detected at masses 15, 29, and 30, with weaker spikes at masses 13 and 14. No photodesorption signal was detected due to acetone (see mass 43 signal). Although signals at masses 16, 28, 32, and 44 could not be resolved from the high backgrounds of O2, CO, and CO2 resulting from backfilling the chamber with 5 × 10-8 Torr O2, no other masses yielded detectable photodesorption signals. The ratio of masses 29 to 30 is consistent with that of gas phase formaldehyde.51 In the d6-acetone case, there was a small mass 30 signal attributable to DCO from QMS cracking of D2CO, but the parent signal was obscured by O2. In the acetone case, the ratios of masses 13 to 14 to 15 were not consistent with those of methane. Mass 16 was not of assistance in confirming this conclusion because QMS cracking of O2 dominated at this mass. Instead, a similar photodesorption experiment involving 0.75 ML d6-acetone (upper traces) showed a photodesorption spike at mass 18 (CD3), but not at mass 20 (CD4). This indicates that the spikes in the mass 15 signal for acetone and in the mass 18 signal for d6acetone were from methyl radical photodesorption. The mass 15 to mass 14 to mass 13 ratios are also consistent with gasphase mass spectrum of methyl radical.54,55 These data indicate that UV irradiation of adsorbed acetone on TiO2(110) results in cleavage of a C-C bond and ejection of a methyl group from the surface. Repetition of these photodesorption experiments under various conditions of acetone coverage, O2 pressure and substrate temperature resulted in only the mass signals from methyl radicals and formaldehyde. In these experiments, the signal ratios of methyl radical to formaldehyde were generally about 4 to 1 despite the changes in the surface conditions. On the basis of these observations, the formaldehyde signals are attributed to methyl radical reactions on the walls of the chamber and/or QMS and not from formaldehyde photodesorption from the surface. (Given that these measurements were conducted under UHV conditions, the probability of gas phase reactions involving the ejected methyl radicals is essentially nonexistent.) This conclusion provides a plausible explanation for the generation of formaldehyde and formate species on powder TiO2. That is, that ejected methyl radicals form new C1 species by reactions occurring away from the surface they were ejected from. Multiplication of the mass 15 signal in Figure 4 (by 20) reveals two rates of methyl radical photodesorption. A fast component, which constitutes the initial spike, attenuated rapidly after about 5 s, settling into a nonzero value that was maintained until the light was blocked at 300 s. (The mass 18 signal from photolysis of d6-acetone was too noisy to resolve the slow rate due to much larger contributions at this mass from background water in the chamber.) These data were used to determine the acetone photooxidation cross section. Point-by-point integration of the mass 15 photodesorption signal in Figure 4 as a function of time gave the relative amount of methyl radical ejected up to a given irradiation time. Determination of the total amount of acetone photooxidation, which enabled normalization of the yield of methyl radical

Henderson

Figure 5. Acetone photooxidation cross section plot obtained from the methyl radical (mass 15) photodesorption trace of Figure 4.

ejected, was obtained by conducting TPD after the photodesorption experiment. TPD before and after the photolysis treatments shown in Figure 4 reveal that about 50% of the adsorbed acetone photooxidized during the 300 s irradiation period. From these data, a natural log plot of the acetone coverage retained at UV irradiation time ‘t’ versus the corresponding UV photon exposure from the photodesorption data in Figure 4 is plotted in Figure 5. The total photon exposure in this experiment was ∼4 × 1019 cm-2, whereas the photon exposure regime in Figure 3 was on the order of 1021 cm-2. Nevertheless, the data in Figure 5 (from the methyl radical photodesorption signal of Figure 4) shows two regimes just as seen in the data of Figure 3. The slow rate yielded a cross section of ∼6 × 10-21 cm2, which is close to the 3 × 10-21 cm2 value obtained for large photon exposures in the TPD measurements. The similarity of these two cross section values indicates that the conversion of acetone to acetate, as gauged by TPD, and the ejection of methyl radicals, as gauged by photodesorption, both probe the same process: photocatalytic cleavage of a C-C bond in an adsorbed acetone species. As in the case of the TPD data, the linear fit of the slow process in Figure 5 does not intercept at the origin. The initial, fast component had a cross section on the order of 10-18 cm2, which is only a factor of 100 from that of a typical collisional cross section (∼10-16 cm2). This suggests that the efficiency of the fast process is on the order of 1%, but likely somewhat higher since a large percent of the incident flux is absorbed too deep in the TiO2(110) crystal to be of use in initiating surface processes. The position of the y-axis intercept for the slow rate can be used as a maximum for the amount of acetone photooxidation in the fast step. The -0.4 value corresponds to a depletion of about 35% of the initial acetone coverage or ∼0.26 ML acetone. In contrast, only about 15% of the initial acetone coverage (∼0.11 ML) was decomposed during the slow process.

Photooxidation of Acetone on TiO2(110)

Figure 6. Methyl radical (mass 15) photodesorption spectra from UV irradiation in UHV at 105 K of 0.25 ML acetone adsorbed at 95 K on (a) clean and (b and c) preoxidized TiO2(110). (See text for details on the preoxidation procedure.) The only difference between the spectra in parts b and c is that after acetone adsorption the surface in (c) was preheated to 243 K and recooled to 105 K before UV irradiation.

3.3. Preheating Studies. As mentioned above, there is particular significance to conducting the photodesorption measurements at 200 K. Figure 6 illustrates the impact that substrate temperature has on the photooxidation rate of acetone on TiO2(110). In Figure 6a, 0.25 ML acetone was exposed to UV light at 105 K without having preheated the surface or exposing it to oxygen. In this case, the surface was populated with ∼7% oxygen vacancy sites in addition to the dosed acetone. There was no evidence for methyl radical photodesorption or for any surface photocatalysis based on subsequent TPD analysis (not shown). These data illustrate the need for O2 in order to see acetone photooxidation on TiO2(110). For the spectra in Figure 6b and 6c, the surface with 7% oxygen vacancies was treated with 20 L O2 at 95 K, heated to RT, and recooled to 105 K for acetone adsorption. The data in Figure 6b were collected during UV irradiation at 105 K immediately after acetone adsorption, whereas the data in Figure 6c were collected after first preheating the acetone + oxygen adlayer to 243 K and then recooling to 105 K prior to UV irradiation. Comparison of parts b and c indicates that the initial yield of methyl radicals was over 4 times greater if the acetone + oxygen adlayer was heated prior to UV irradiation. Figure 7 shows a continuation of the type of experiments shown in Figure 6 to a variety of preheating conditions between 105 and 300 K. The data show a rise in the methyl radical yield after preheating the acetone + oxygen adlayer to temperatures in excess of about 175 K. These data suggest that acetone and oxygen mixtures do not rapidly photoreact on TiO2(110) unless some thermal activation barrier has been crossed prior to UV

J. Phys. Chem. B, Vol. 109, No. 24, 2005 12067

Figure 7. Methyl radical (mass 15) photodesorption yield as a function of surface preheating temperature using the same experiment parameters discussed for Figure 6.

exposure. Since the data in Figures 6 and 7 were collected during UV irradiation at 105 K, one can assume that a thermodynamically favorable reaction takes place after heating above ∼175 K. On the basis of previous HREELS work that shows vibrational features associated with an acetone-oxygen complex grow in after heating acetone + oxygen adlayers from 90 K to above about 235 K,53 the data in Figures 6 and 7 suggest that methyl radical ejection from adsorbed acetone requires first that acetone be activiated by coadsorbed oxygen according to this reaction:

η1-acetone(a) + O(a)/O2(a) f {acetone-oxygen}(a) (1) where η1-acetone refers to acetone molecules coordinated to Ti4+ sites through lone pair electrons on the oxygen end of the molecule (this is the main form of adsorbed acetone on TiO2 surfaces 53), {acetone-oxygen}(a) refers to the complex, and the designation ‘a’ indicates an adsorbed species. The complex has not been adequately characterized to an extent needed for confidence in any particular structural model. However, it is clear from HREELS that the C-O bonds in the complex are midway between single and double bonds.53 Coupling of thermal processes with photooxidation of acetone on a TiO2 photocatalyst have also been demonstrated by Vorontsov and coworkers.43 These authors observed an increase in the rate of acetone photooxidation when the catalyst was heated from 40 to 100 °C, and this was attributed to thermal oxidation of an intermediate species such as formaldehyde. The nature of the oxygen species participating in the reaction is also unclear since both atomic and molecular oxygen exist on TiO2(110) after exposure of O2 to vacancy defects (followed by heating to >200 K).56-58 The photodesorption yield data as a function of preheating temperature shown in Figure 7 were

12068 J. Phys. Chem. B, Vol. 109, No. 24, 2005 used to construct a crude Arrhenius plot that yielded an activation energy of about 10 kJ/mol for reaction 1. This low energy barrier explains why some acetone-oxygen complex is formed on the TiO2(110) surface at 95 K on exposure to gaseous acetone possessing a RT internal energy distribution. 3.4. Acetone Photooxidation Mechanism. Results in Figures 6 and 7 indicate that photocatalytic C-C bond cleavage in adsorbed acetone on TiO2(110) depends on a thermal reaction between acetone and a surface oxygen species (resulting from O2 adsorption at surface electronic defects). Two acetone photooxidation rates (fast and slow) are seen in both the TPD and photodesorption data. The fast reaction can be expressed in the following manner:

fast: {acetone-oxygen}(a) + UV f acetate(a) + CH3(g) (2) Thermal decomposition of the acetone-oxygen complex occurs above 350 K and preferentially yields gas-phase acetone and not adsorbed acetate.53 Therefore, conversion of the complex into acetate using UV light at substrate temperatures below RT must be nonthermal in nature. Data discussed below suggests that the substrate mediates this process (via electron-hole (e-/ h+) pairs) instead of it occurring through direct excitation of the complex. At this point, it is unclear whether holes or electrons drive reaction 2, but the former is the more likely candidate. As mentioned above, formation of the acetoneoxygen complex requires O2 adsorption on the TiO2(110) surface, which in turn is only observed if Ti3+ sites are present.7,57 A substantial surface coverage of adsorbed oxygen species was enabled by starting with a vacuum annealed surface that possessed 7% oxygen vacancy sites. Given that the rate of acetone photooxidation was at least 2 orders of magnitude faster in the initial few seconds of UV irradiation than at longer times, one can assume that the rate of the slow period was not limited by the flux of e-/h+ pairs reaching the surface because the photon flux was the same throughout the experiment. As mentioned above, one exception to this would be if the incident UV light directly excited the acetone-oxygen complex (i.e., was not mediated by e-/h+ pair generation in TiO2) while the slow rate required some TiO2mediated excitation. This latter condition is difficult to test for without knowledge of the structural and optical properties of the acetone-oxygen complex. The optical properties of gas phase acetone are well-known (see below), but one cannot assume that this knowledge is transferable to the acetoneoxygen complex. In fact, the absence of photooxidation for purely η1-adsorbed acetone (see Figure 6), which is essentially structurally identical to gas phase acetone, indicates that the acetone-oxygen complex is significantly different in terms of its optical and/or redox properties. (The redox properties describe the thermodynamics associated with electron-mediated reduction and hole-mediated oxidation of the complex using charges in their respective positions in the TiO2 potential energy diagram.) In either case (both fast and slow processes are e-/h+ pair mediated or only the slow process is), the photocatalytic steps associated with the slow process are different in some way than those in the fast process. Two possibilities for the slow portion of the acetone photooxidation reaction are discussed below. The key similarity to keep in mind for the fast and slow reactions is that both involve methyl radical photodesorption and acetate formation. The first model assumes that fast process involves activation of the complex (reaction 2) and the slow process, occurring after all surface acetone-oxygen complex is consumed, involves direct activation of adsorbed acetone via C-C bond cleavage forming

Henderson an acetyl intermediate (reaction 3). The acetyl intermediate, with a structure that maintains a CH3-CO configuration, then must react with an oxygen species to form the final surface product acetate (reaction 4).

slow: η1-acetone + UV f CH3 + acetyl(a)

(3)

acetyl(a) + O(a) f acetate(a)

(4)

While reaction 3 may or may not involve charge carriers generated in the TiO2 surface, this proposed process resembles what is known to occur in gas-phase photochemistry of acetone (refs 59-64 and references therein). A summary of the gasphase photolysis of acetone from these references will help in excluding reactions 3 and 4 as a viable mechanism for the slow photolysis process. The C-C bond dissociation energy in acetone is about 81 kJ/mol or 3.5 eV.65 While this energy is clearly within the spectral range of the Hg arc lamp used in this study, direct photolysis of acetone with 3.5 eV photons (∼355 nm) would involve an unallowed S f S transition, the probability of which is very small in the gas phase. In fact, the photon absorption cross section of gas-phase acetone at wavelengths >330 nm is