Photochemistry of 1, 1, 1-Trifluoroacetone on Rutile TiO2 (110)

Feb 12, 2010 - The ultraviolet (UV) photon-induced photodecomposition of 1,1,1-trifluoroacetone (TFA) adsorbed on the rutile TiO2(110) surface has bee...
0 downloads 0 Views 2MB Size
16900

J. Phys. Chem. C 2010, 114, 16900–16908

Photochemistry of 1,1,1-Trifluoroacetone on Rutile TiO2(110)† R. T. Zehr, N. A. Deskins,‡ and M. A. Henderson* Institute for Interfacial Catalysis, Pacific Northwest National Laboratory P.O. Box 999, MS K8-87 Richland, Washington 99352 ReceiVed: NoVember 03, 2009; ReVised Manuscript ReceiVed: January 12, 2010

The ultraviolet (UV) photon-induced photodecomposition of 1,1,1-trifluoroacetone (TFA) adsorbed on the rutile TiO2(110) surface has been investigated with photon stimulated desorption (PSD), temperature programmed desorption (TPD), and density functional theory (DFT). TFA adsorbed molecularly on the reduced surface (8% oxygen vacancies) in states desorbing below 300 K with trace thermal decomposition observed in TPD. Adsorption of TFA on a preoxidized TiO2(110) surface (accomplished by pre-exposure with 20 L O2) led to formation of a new TFA desorption state at 350 K, assigned to decomposition of a TFA-diolate species ((CF3)(CH3)COO). No TFA photochemistry was detected on the reduced surface. UV irradiation of TFA on the oxidized surface depleted TFA in the 350 K state with TFA molecules in other TPD states unaffected. PSD measurements revealed that both carbonyl substituents (CH3 and CF3), as well as CO, were liberated during UV exposure at 95 K. Postirradiation TPD showed evidence for both acetate (evolving as ketene at 650 K) and trifluoroacetate (evolving as CO2 at 600 K) as surface-bound photodecomposition products. The CO PSD product was not due to adsorbed CO, to mass spectrometer cracking of a CO-containing PSD product, or from background effects, but originated from complete fragmentation of an unidentified adsorbed TFA species. Thermodynamic analysis using DFT indicated that the photodecomposition of the TFA-diolate was likely not driven by thermodynamics alone as both pathways (CH3 + trifluoroacetate and CF3 + acetate) were detected when thermodynamics shows a clear preference for only one (CF3 + acetate). These observations are in contrast to the photochemical behavior of acetone, butanone, and acetaldehyde on TiO2(110), where only one of the two carbonyl substituent groups was observed, with a stoichiometric amount of carboxylate containing the other substituent left on the surface. We conclude that fluorination significantly alters the electronic structure of adsorbed carbonyls on TiO2(110) in such a way as to promote multiple channels of photofragmentation. Factors that dictate the partitioning between the three TFA channels are not related to photon energy (above that of the TiO2 band gap) but likely to the electronic structure of the charge transfer excited state. 1. Introduction There has been growing research interest in using the photocatalytic properties of TiO2 to perform useful chemical transformations.1-5 Much research has been focused on using TiO2 as a photocatalyst to break down volatile organic pollutants.1,3 Adsorption of a photon with energy greater than the TiO2 bandgap excites an electron to the conduction band (CB) while leaving behind a hole in the valence band (VB). Once separated, the electron and hole may migrate to the surface and be available to perform charge transfer chemistry on adsorbed molecules. Many studies focus exclusively on the end points of photomineralization (i.e., conversion to CO2 and H2O) in an attempt to gauge the relative efficiency of their particular photocatalyst formulation. However, understanding the detailed mechanism leading to photomineralization is valuable for the rational design of photocatalysts. The use of single crystal model catalysts under the well-controlled conditions of ultrahigh vacuum (UHV) affords exquisite control of reaction conditions, allowing the study of elementary reaction steps in much greater detail than is possible under operando conditions. †

Part of the “D. Wayne Goodman Festschrift”. * To whom correspondence should be addressed. E-mail: [email protected]. ‡ Current address: Department of Chemical Engineering, Worcester Polytechnic Institute, 100 Institute Road, Worcester, MA 01609-2280.

Work described here is an extension of our previous studies aimed at understanding the mechanisms for carbonyl photooxidation on TiO2. These previous studies (acetone,6-9 butanone,9 and acetaldehyde10) have established several detailed steps in the photooxidation of such organic carbonyl molecules on TiO2(110). In the absence of adsorbed oxygen, the adsorbed carbonyls are bound to the surface in an η1 fashion through an electrostatic interaction between the lone pair of the carbonyl oxygen and a Ti4+ site. The η1-bound carbonyl molecules do not undergo detectable photochemistry. The main steps involve first a thermal reaction between the carbonyl (RC(O)R′) and adsorbed oxygen to form a diolate species (η2-R(R’)COO(a),), as shown in reaction 1

η1 - RC(O)R′(a) + O(a) f η2 - R′(R′)COO(a)

(1)

where R and R′ are the carbonyl substituent groups (e.g., H and CH3 in acetaldehyde), and “a” represents adsorbed species. The degree of charge on the oxygen adatom reactant and the diolate product involved in this first step are presented as neutral simply because they are not known. Reaction 1 is reversible in TPD. The diolate structure was assigned based on the appearance of a symmetric O-C-O stretching mode at 1425 cm-1 observed with HREELS for the photoactive species of acetone on oxidized TiO2(110) and on oxygen isotopic scrambling in

10.1021/jp910507k  2010 American Chemical Society Published on Web 02/12/2010

Photochemistry of TFA on Rutile TiO2(110)

J. Phys. Chem. C, Vol. 114, No. 40, 2010 16901

TPD.11 The activation energy to form the diolate is low, typically in the 8-10 kJ/mol range.6,10 The identity of the oxygen species that reacts with the carbonyl is likely an oxygen adatom8,12,13 because diolate can be formed under conditions in which no molecular oxygen is present (e.g., after RT adsorption of O2). The second step involves substrate-mediated photochemistry leading to the decomposition of the diolate species, formation of an adsorbed carboxylate and photodesorption of one of the substituent groups of the carbonyl (reaction 2 or 3)

η2 - R(R′)COO(a) + hν(h+) f R(g) + R′COO(a)

(2) or

η2 - R(R′)COO(a) + hν(h+) f R′(g) + RCOO(a)

(3) The charge carrier responsible for this photochemistry is not definitively known, but a VB hole is the likely candidate because sp3 hybridization in the diolate makes it unlikely that the adsorbate possesses electron acceptor states near the CB minimum. The amount of charge on the diolate likely plays a role in determining how much energy the neutralized diolate achieves after charge transfer with a VB hole. When the two carbonyl substituents are different, the final products might be predicted by considering the thermodynamics of the two possible outcomes (R-carboxylate + R′ radical versus to R′-carboxylate + R radical). Since both possible channels start from a common point, only a comparison of the final products is necessary. Bearing in mind that the relative stabilities of the resulting carboxylates cannot be ignored, one can start by comparing the enthalpies of formation for the respective gas phase radicals of these substituents. For example, the photochemistry of butanone (CH3CH2(CdO)CH3) on oxidized TiO2(110) has the potential of photodesorbing CH3 (∆Hf°gas ) 146.7 kJ/mol) or CH3CH2 (∆Hf°gas ) 118.8 kJ/mol).14 Only the more stable ethyl radical is observed during PSD.9 Similarly, the photochemistry of acetaldehyde (CH3(H)CO) has the potential of photodesorbing either CH3 or H (∆Hf°gas ) 218 kJ/mol),14 but only the more stable methyl radical is observed during PSD.10 In this paper, we show that this simplistic model breaks down for the photochemistry of trifluoroacetone on oxidized TiO2(110). One might expect that the extremely exothermic enthalpy of formation for trifluoromethyl radical (∆Hf°gas ) -465.7 kJ/mol)14 would lead to preferential ejection of CF3 as the primary PSD product from TFA-diolate photodecomposition on TiO2(110) on the surface. Instead, a substantial fraction of adsorbed TFA-diolate species photofragment to CH3 and adsorbed CF3COO. Results presented here point to the need for a deeper understanding of charge transfer excited states in order to fully understand partitioning of this species during photochemistry on TiO2(110). 2. Methods The UHV chamber used in these experiments has been described in detail elsewhere.15 The base operating pressure for the system during these experiments was 2 × 10-10 torr. The TiO2(110) crystal (Princeton Scientific) was cleaned and bulk reduced by cycles of Ar+ sputtering (2 kV) at high temperature (850 K), followed by UHV annealing at 850 K. The crystal was judged clean and well-ordered by means of AES, SIMS,

and H2O TPD. The reduced crystal was dark blue and contained ∼8% surface O-vacancy defects as determined by H2O TPD.15 1,1,1-Trifluoroacetone (97%, Aldrich) and trifluoroacetic acid (99.8%, Ficher) were further purified by a series of freezepump-thaw cycles before use. Gas lines were passivated by exposure to the specific gas for several hours before first use. UV irradiation was accomplished using a 100 W Hg arc lamp and fused silica fiber optic light delivery system as described in previously.6 The typical flux of photons with energies greater than the 3.0 eV TiO2 band gap was ∼2.0 × 1016 photons/cm2s. Irradiation of the TiO2(110) crystal under these conditions resulted in less than 5 K temperature increase at 95 K. Dosing of TFA on the TiO2 surface was achieved using a triply differentially pumped molecular beam doser that delivered the reagent in a ∼5 mm spot centered on the TiO2 crystal, thus minimizing the exposure of the rest of the chamber to TFA. The molecular beam doser was calibrated for TFA dosing using water as a standard reference, assuming the sticking coefficient for TFA at 95 K was unity. At the TFA backing pressures used for the first stage (∼1-2 Torr), roughly 85-90% of the flux emitted from the doser was directed (in the beam) and ∼10-15% was diffuse (depending on the selected TFA backing pressure). Because of the low sticking of O2 on TiO2(110) at 95 K and above, oxygen was introduced by backfilling the UHV chamber using a variable leak valve. Computer-controlled pneumatic valves located between the leak valve and the gas handling manifold, as well as between the leak valve and a rough vacuum line, allowed precise control of chamber backfilling. Typically, the surface was prepared for photochemistry by exposure to 20 L O2 at 95 K followed by dosing of excess TFA at 95 K. The surface was then heated to 200 K to complete the reaction of TFA and oxygen to form the photoactive TFA-diolate species. Briefly heating to 200 K also desorbed the TFA multilayers and weakly bound TFA molecules leaving a consistent final coverage TFA diolate (see below). Irradiation was initiated once the crystal cooled to 95 K. Coverages are expressed as monolayers (ML), where 1 ML is defined as 5.2 × 1014 sites/cm2, the number density of Ti4+ sites on the ideal TiO2(110) surface. Molecular modeling calculations of the rutile TiO2(110) surface were performed using the CP2K program.16-18 The simulations were carried out at the DFT level. CP2K uses the Gaussian plane wave (GPW) method whereby the electronic wave functions are described by Gaussian functions, while the electron density is described by planewaves. Core electrons (1s for C, O, and F; 1s2s2p for Ti) were represented by static pseudopotentials19,20 since the electrons in these orbitals change little in different environments. Valence electrons were represented by a double-zeta Gaussian basis set. Reciprocal space was sampled by the Γ-point. The PBE exchange correlation functional21 was used in this work. The slab method was used to portray the (110) surface. A four layer slab, where each layer consisted of an O-Ti-O structure, which represented a (4 × 2) surface cell was used. In total the clean surface consisted of 192 atoms. A model of the preoxidized surface consisted of a single O adatom adsorbed at a 5-coordinate Ti4+ site on the surface; this model is consistent with what is obtained experimentally from exposure of O vacancy sites on UHV annealed surface to gas phase O2. Reaction of various ketones with the O-Ti4+ moiety leads to surface-adsorbed diolate species. Thermodynamics for decomposition of the TFA-diolate was modeled by removing R-groups to gas phase. Cited heats of reaction refer to 0 K and include energy minimizations of the gas and adsorbed phases.

16902

J. Phys. Chem. C, Vol. 114, No. 40, 2010

Zehr et al.

Figure 1. TPD spectra from various coverages of trifluoroacetone adsorbed on the UHV annealed TiO2(110) surface. The inset depicts the integrated TPD peak areas of the labeled peaks.

Figure 2. TPD traces from various coverages of trifluoroacetone adsorbed on the preoxidized TiO2(110) surface. The inset depicts the integrated TPD peak areas of the labeled peaks.

3. Results and Discussion 3.1. Thermal Chemistry. Figure 1 presents a series of TPD traces from various coverages of TFA adsorbed on the vacuum reduced TiO2(110) surface. The inset graph shows the relative areas of the labeled peaks as a function of coverage. TPD of the lowest TFA coverages showed a broad and weak feature at 315 K labeled (a). This feature did not saturate with increasing coverage even after saturation of the first layer, and is assigned to desorption of the small fraction of nondirectional TFA emitting from the molecular beam doser that impinged on the sample holder. A more substantial feature, labeled (b), grew in at 265 K with additional TFA exposure, shifting to lower temperature with increasing TFA exposure. Feature (b) saturated at 220 K with a TFA coverage of ∼0.4 ML. A new peak, labeled (c), centered at 155 K, appeared for coverages greater than 0.4 ML. The 155 K peak grew in and shifted to 130 K as the TFA coverage increased. This peak saturated at a cumulative TFA coverage of 0.95 ML corresponding to a saturation coverage of ∼1 ML. Peaks (b) and (c) are assigned to TFA desorption from the first layer. Peak (c) is likely due to desorption of TFA from a highly compressed TFA overlayer that relieves stress resulting from nearly every available Ti4+ site being occupied with TFA. After ∼50% of the first layer coverage desorbed, the adlayer relaxed and the interadsorbate repulsion became secondary to the TFA-surface binding interaction in terms of affecting the desorption behavior. This model is consistent with the behavior of acetone on TiO2(110).11 The similarities in the acetone and TFA coverage dependence in TPD is also consistent with the similar gas phase dipole moments of the two molecules (both being ∼2.9 D22,23). Performing a Redhead analysis24 on the 265 K feature in Figure 1 (0.04 ML in coverage), assuming a pre-exponential factor of ν ) 1013 s-1, gave a desorption energy of 69 kJ/mol. The value of 69 kJ/mol should be regarded as an estimate of the desorption energy given that Redhead’s analysis in only reliable for truly first order desorption where the desorption kinetics are not a function of coverage.25 Finally, a multilayer

peak, labeled (d), formed at 115 K for coverages greater than 1 ML. The multilayer peak could not be populated further due to the ∼95 K minimum temperature achieved in these studies. A small fraction of the first layer TFA coverage decomposed on the reduced TiO2(110) surface. Weak signals consistent with C2-3 hydrocarbons and CF2-3 fragments (not shown) appeared at 520 K where no parent (mass 43) desorption was observed. Because of the very low intensity of these signals, no conclusive identification were made. The integrated intensity of the strongest decomposition peak at mass 50 was less than 1% of that of the monolayer at mass 43, indicating that TFA thermal decomposition on the reduced TiO2 surface was a minor channel, occurring at coverages well below that of the surface oxygen vacancy population. Figure 2 shows a series of TFA TPD experiments carried out on a TiO2(110) surface previously exposed to 20 L of O2 at 95 K. The inset shows the relative peak areas of the labeled features as a function of TFA coverage. Preoxidation of the surface has a significant influence on the TPD of TFA on TiO2(110). The most obvious effect was the presence of a significant amount of TFA desorption above 300 K that was not observed on the reduced surface. This new TFA TPD intensity, labeled (b′), grew in with near first order behavior in that the peak temperature (360 K) did not shift with coverage. TFA groups that constituted this new TPD state might be thought of as having contributed to the (b) state on the reduced surface (see Figure 1) since the amount of TFA desorbing from the (b) state in Figure 2 (at 250 K) on the oxidized surface decreased from the ∼0.4 to ∼0.2 ML coverage with preoxidation. Peak (b′) saturated at a coverage of ∼0.29 ML. Redhead analysis of the initial low coverage (b′) peak assuming a preexponential factor of ν ) 1013 s-1 gave a desorption energy of 96 kJ/mol. The (b′) state is assigned to the reaction of adsorbed TFA with preadsorbed oxygen to form a TFA-diolate species akin to what has been seen on TiO2(110) with acetone, butanone, and acetaldehyde. The reaction was reversible as the diolate thermally decomposed during TPD liberating TFA and presumably leaving oxygen atoms on the surface. A fourth peak, labeled

Photochemistry of TFA on Rutile TiO2(110)

Figure 3. TPD traces from a submonolayer coverage of trifluoroacetone adsorbed on the 18O2 preoxidized TiO2(110) surface.

(c), forming in the temperature range between 130 and 170 K, was seen for TFA coverages greater than 0.49 ML. This feature is essentially identical to the Figure 1(c) peak seen on the reduced surface. Finally, a multilayer peak, labeled (d), formed at 113 K for coverages greater than 1 ML. The experiments in Figure 2 were repeated, except with 18O2 instead of 16O2, to verify whether oxygen-isotope scrambling occurred. Figure 3 shows mass 43 and 45 traces from a submonolayer coverage of TFA on the 18O2 preoxidized TiO2(110) surface. Mass 43 reflects desorption of the 16Ocontaining parent molecule. Occurrence of signal at mass 45 can only arise from scrambling of 18O into the desorbing TFA. As with the acetone case,11 the source of this scrambling was decomposition of the TFA-diolate formed from reaction of TFA with adsorbed 18O. As was seen with acetone, 18O ended up preferentially in the diolate state but also in molecularly adsorbed TFA (the 250 K TPD feature), indicating some degree of interchange during TPD between the diolate state and the molecularly adsorbed state. No 18O exchange was seen in the lower temperature desorption states of TFA (not shown). These data are consistent with the diolate model for reactions of organic carbonyls with adsorbed oxygen on TiO2(110). As on the reduced surface, a small fraction of TFA adsorbed onto the oxidized surface irreversibly reacted, in this case with coadsorbed oxygen to form acetate during the TPD experiment (not shown). The surface bound acetate thermally decomposed to ketene at 600 K.26,27 The coverage of acetate formed by TFA decomposition was quantified using acetic acid TPD.28,29 Comparison of the two signals gave an estimated acetate coverage from TFA decomposition of ∼0.01 ML. Formation of acetate from TFA would necessitate the loss of the trifluoromethyl substituent, however, no fluorine containing fragments were observed in TPD, most likely due to the low coverages of fluorocarbon fragments involved. 3.2. Photon-Induced Chemistry. The TFA-diolate from reaction of TFA with the oxidized TiO2(110) surface is photoactive, undergoing photofragmentation when irradiated with UV light. (As shown in the Supporting Information, use of cutoff filters established that the photochemistry occurred

J. Phys. Chem. C, Vol. 114, No. 40, 2010 16903

Figure 4. PSD spectra for various masses obtained during UV irradiation at 95 K of 1 ML of trifluoroacetone dosed on oxidized TiO2(110) and preheated to 200 K.

via a substrate-mediated excitation process.) No TFA photochemistry was observed on the reduced surface without the presence of adsorbed oxygen. Figure 4 depicts PSD data for TFA on oxidized TiO2(110) exposed to UV light at 95 K. For these experiments, excess TFA was dosed on the preoxidized TiO2(110) surface and briefly heated to 200 K resulting in a reproducible starting coverage of 0.5 ML, of which more than half was TFA-diolate and the remained being other TFA species (e.g., molecularly adsorbed TFA). The UV lamp shutter was opened at time t ) 0 and the subsequent evolution of photofragments was monitored with the mass spectrometer as a function of time. Several PSD signals exhibited a sharp spike in intensity on UV exposure, followed by near-exponential signal decay. Unlike the photochemistry exhibited by acetone,6-9 butanone,9 and acetaldehyde10 on oxidized TiO2(110), the photofragmentation of TFA was more diverse and complex, resulting in several fragments being ejected into the gas phase. Signals were detected at masses 13-15, 19, 28, 31, 50, and 69 during UV irradiation. The absence of signals at other masses precluded PSD of a variety of species. For example, absence of signal at mass 43 precluded the photodesorption of the parent or any other CH3CO-containing species. Measurements with the QMS ionizer off and the QMS tuned for ion detection verified that none of the detected signals were due to desorption of ions, thus negating the possibility of CF3- PSD. Identification of the PSD products was achieved using published mass spectral data. For example, the integrated intensities of the mass 13-15 fragments were consistent with the expected ratios for methyl radical cracking in the ionizer of the mass spectrometer.6,30,31 Unlike the cases for acetone, butanone, and acetaldehyde, where only alkyl radicals were photodesorbed, the relatively intense mass 28 signal indicated different photochemistry for TFA. PSD signal in the mass range of 26-30 was only observed for mass 28, consistent with an assignment of this fragmentation channel to CO and not a C2 species. Similarly, absence of other higher mass signals (e.g., mass 42 from ketene or mass 44 from CO2) that might have contributed signal at mass 28 indicated that this signal was due to CO. For example, the possibility

16904

J. Phys. Chem. C, Vol. 114, No. 40, 2010

Zehr et al.

Figure 5. Comparison of the CH3 (mass 15) and CO (mass 28) PSD signals as a function of substrate temperature during UV irradiation. Adlayers in both cases were prepared under identical conditions.

that the mass 28 signal arose from mass spectrometer cracking of a transiently stable PSD species, such as CH3CO or CF3CO (that might also contribute signal for CH3 or CF3), was excluded based on the absence of any PSD signals other than those indicated above. The temporal profile for CO from photodecomposition of TFA diolate was slower than that of the methyl radical, an observation one might attribute to the ability of the TiO2(110) surface to partially thermalized the photodesorbing CO at 95 K. CO TPD results by Linsebigler and co-workers32 reveal that while nearly a complete monolayer of CO can be stabilized on TiO2(110) at 95 K, CO is not stable on oxidized TiO2(110) above 175 K. On the basis of these results, the surface should not exert a significant retention effect on CO in these PSD experiments for substrate temperatures above ∼175 K. Figure 5 clearly shows that the temporal profiles of both CH3 and CO were unaffected by substrate temperature in the range between 95 and 250 K. (In this figure, a TFA diolate adlayer was prepared as was done for Figure 4, except that photolysis was performed at 250 K instead of at 95 K.) This comparison suggests that the tail in the CO PSD is not due to transient retention of CO by the surface. It is also interesting that the temporal evolution of methyl radical showed no effect of substrate temperature. This suggests either that methyl radicals do not interact strongly with the TiO2(110) surface or (more likely) that their trajectories after photofragmentation were highly oriented away from the surface. In addition to CO and CH3, Figure 4 shows that PSD signals were detected at masses 19, 31, 50, and 69. The intensities of this family of signals were considerably less than those of CH3 or CO, making analysis of their temporal profiles difficult. We assign these signals to mass spectrometer cracking of CF3 radicals. The relative intensities of the 19, 31, 50, and 69 peaks were approximately 1:9:10:1. Tarnovsky and co-workers33 have measured the absolute dissociative ionization cross sections of CF3 by electron impact. They found that the 70 eV electron impact ionization cross sections of CF3 to F+, CF+, CF2+, and CF3+ were 0.35 ((30%), 0.68 ((20%), 0.76 ((20%), and 0.38 ((18%) Å2, respectively, corresponding to ratios of 4.6:8.9:10: 4.7. The relative PSD intensities in our data at masses 31 and

Figure 6. PSD yields of CH3 (mass 15), CO (mass 28) and CF3 (mass 50) as a function of preannealing temperature. PSD conducted at 95 K.

50 matched the relative CF+ and CF2+ cross sections well; however, the PSD intensities for CF3+ and F+ were lower by comparison. This discrepancy may result from differences in ion transmission efficiencies (the mass spectrometer used in this study is equipped with additional ion optics for ion energy analysis). With only a few exceptions, the mass spectrometer cracking of small, CF3-containing organics typically have mass 69 as the most intense fragment in the 19-31-50-69 series.14 The intensities of these ions from cracking of TFA in this study followed this trend: ∼0.01:8.5:4.2:10 (data not shown). The observation that our mass 69 signal was far from the most intense PSD signal in this series is consistent with these signals not originating from cracking of a CF3-containing molecule, but from CF3 itself. Observations of three PSD products (CH3, CF3, and CO) is unique among the carbonyls studied thus far (acetone,6-9 butanone,9 and acetaldehyde10), which showed only one channel of photodesorption. Figure 6 tracks the relative (not absolute) PSD yields of CH3, CF3, and CO (obtained by integration of each PSD trace over the first 2 s of UV irradiation; see Figure 4) as a function of the preannealing temperature to ascertain if each signal originated from the same source. The conditions for each experiment in terms of TFA and O2 exposure were identical (see Figure 4 description) except that the preannealing temperature for each TFA + O2 adlayer was varied, with all PSD data collection at 95 K. Without preheating (the 95 K points), the PSD signals of all products were low, but not zero because arriving TFA molecules had sufficient thermal energy to form some TFA-diolate even at 95 K. The yields increased as the surface was preannealed, a condition which allowed for more efficient diolate formation.6,10 The yields of all three PSD products abruptly dropped to zero as the preannealing temperature was increased above 350 K due to thermal decomposition of TFA-diolate and desorption of TFA (see TPD in Figure 2). This suggests that all three products are linked to TFA on the oxygen-modified surface. However, the CH3 yield (mass 15) followed a different temperature dependence than the CF3 and

Photochemistry of TFA on Rutile TiO2(110)

J. Phys. Chem. C, Vol. 114, No. 40, 2010 16905

Figure 7. C16O (mass 28) and C18O (mass 30) PSD traces from coadsorption of 16O-labeled TFA with 18O2 on TiO2(110).

Figure 8. TPD spectra for various masses obtained before (black lines) and after (red lines) 5 min UV irradiation of 1 ML trifluoroacetone dosed on oxidized TiO2(110) and preheated to 200 K.

CO yields. While the CF3 and CO yields increased with preannealing temperature up to 300 K (and then diminished), the CH3 yield maximized at (or below) 200 K and gradually diminished with increasing preannealing temperature. These observations do not preclude some convolution of more than one PSD process in the evolution of the products (especially since formation of CO implies concomitant formation of some corresponding amounts of CH3 and CF3). They also do not suggest that CH3 PSD came from molecularly bound TFA, which desorbed below 300 K, because this form of adsorbed TFA is not photoactive on TiO2(110) (see supplemental section). Instead, these data suggest that at least two forms of photoactive TFA must exist on the surface, both of which required the presence of oxygen on the surface, and that a temperature change affected the relative populations or photoreaction probabilities of these. The supposition that more than one photoactive TFA species was present on the O-modified surface is further supported by PSD experiments involving preadsorbed 18O2. Figure 7 shows C16O (mass 28) and C18O (mass 30) PSD data resulting from the same experiment conducted in Figure 4, except with 18O2. Despite all the O-isotope exchange observed in TPD between TFA and adsorbed 18O (Figure 3), virtually no 18O ended up in the CO PSD product. The C16O PSD trace (and that of the other PSD products) were not significantly different in shape or intensity from those conducted with 16O2 (Figure 4). Given that O exchange was facile into the TFA-diolate TPD state (360 K) and by extension into the TFA molecular state (250 K), the absence of C18O PSD indicates that a form of TFA other than the diolate must reside on the surface. This second form of photoactive TFA also required the presence of oxygen on the surface (since no photochemistry was seen without coadsorbed oxygen), but does not involve a chemical state through which oxygen scrambling can occur. Adsorption of O2 at RT (where only dissociative oxygen exists34) had no variation on the CO PSD (data not shown). All structural options of active sites present on the reduced surface (steps, vacancies, etc.) can be

excluded since they do not initiate photochemistry on the reduced surface. The important issue now is to quantify the amount of photochemistry occurring in the 3 PSD channels and compare these with the amount of available TFA and TFAdiolate on the surface. 3.3. Photochemistry Mechanism. Previous work in our laboratory6-10 has established a two-step mechanism for the photodecomposition of simple organic carbonyl molecules on the oxidized TiO2(110) surface. This mechanism is depicted in the Introduction with reactions (1-3) for a hypothetical organic carbonyl. The first step appears to be universal for all carbonyls examined to this point, including TFA. The choice of which reaction (2 or 3) in step two is believed to be based on the thermodynamics of the respective C-R and C-R′ bond energies. That is, when charge transfer occurs, the charge transfer state responds with bond breaking events that lower the overall energy of the system the most (i.e., by breaking the weakest bonds). The gas phase product is identified in PSD and the surface product is detected in postirradiation TPD. In the TFA case, several photodecomposition products remained on the TiO2(110) surface after UV irradiation, providing information regarding the mechanism of TFA photodecomposition. Figure 8 depicts TPD traces from ∼0.5 ML TFA adsorbed on the oxidized TiO2(110) surface without UV and after UV irradiation for 5 min. The figure shows the preferential loss of intensity in the TFA feature at 360 K associated with photodecomposition of the TFA-diolate (amounting to ∼0.24 ML TFA), and little or no loss of the molecular TFA feature at 250 K (see the mass 43 traces). Additionally, several new peaks for masses 14, 28, 42, 44, and 50 appeared in TPD at higher temperatures after photolysis assigned to surface photodecomposition products that formed as a result of UV irradiation. New features at 625 K for masses 14 and 42 were consistent with formation of ketene from thermal decomposition of acetate during TPD.11 The acetate photoproduct coverage was estimated to be ∼0.10 ML based on comparison with similar signals obtained from TPD of acetic acid on

16906

J. Phys. Chem. C, Vol. 114, No. 40, 2010

Zehr et al.

TiO2(110). (Acetic acid decomposes on TiO2(110) to acetate, with the acid proton transferred to a bridging oxygen site. At RT, a (2 × 1) overlayer of acetate can be formed constituting 0.50 ML.11,28,29,35,36 Of this 0.50 ML acetate coverage, we estimate that ∼0.35 ML decomposes in TPD to ketene and ∼0.15 ML recombines with surface protons to desorb as the parent. See TPD spectra in ref 11). On the basis of this comparison, we estimate that ∼0.10 ML of the photodecomposed 0.24 ML of TFA-diolate followed reaction 4

(CF3)(CH3)COO(a) + hν f CF3(g) + CH3COO(a)

(4) in which CF3 was ejected and acetate was retained on the surface. From this, we conclude that ∼0.14 ML of the photodecomposed TFA diolate followed a different pathway. Aside from acetate, other surface-bound photodecomposition products were detected by postirradiation TPD. CO2 and CO appeared in peaks at 595 K. (The CO signal at ∼120 K is from background adsorption during cooling.) Minor quantities of fluorinated fragments with mass 31 (not shown) and 50 were also observed as broad TPD features centered at 595 K. Absence of TPD features at mass 19, 20, or 38 channels suggested that F was not deposited on the surface during photodecomposition of the TFA diolate. (Mass 20 was sampled by exposing the irradiated surface to water before TPD in an effort to titrate off any adsorbed F- as HF). These observations suggest that the CO2 and CO signals at 595 K, along with the weak signals at masses 31 and 50 at the same temperature, resulted from thermal decomposition of trifluoroacetate on TiO2(110). TPD experiments for trifluoroacetic acid dosed on TiO2(110) (see Supporting Information) also show CO and CO2 states at ∼600 K, as well as weak features assigned to CF3 signals consistent with the data in Figure 8. Using the TPD signals from desorption of 1 ML of CO and CO2 on TiO2(110) as standards,32,37-39 we estimate the coverage of trifluoroacetate formed from photodecomposition of the 0.24 ML of TFA-diolate. On the basis of these comparisons, roughly 0.07 ML of trifluoroacetate was formed as a result of reaction 5

(CF3)(CH3)COO(a) + hν f CH3(g) + CF3COO(a)

(5) Subtracting the amounts of TFA-diolate that photodecomposed by reactions 4 and 5 (0.10 and 0.07 ML, respectively), from the total 0.24 ML of photodecomposed TFA left ∼0.07 ML unaccounted for in postirradiation TPD. The amount of CO PSD was estimated to be ∼0.06 ML based on comparison of the time-dependent integrals of the CO PSD trace in Figure 4 with that obtained from TPD of 1 ML CO on the clean surface. The agreement between the total amount of TFA photodecomposed and the amounts of various photodecomposition products detected suggests that two avenues of TFA photochemistry occur on TiO2(110), one involving photodecomposition of the proposed TFA-diolate, which is partitioned into two product distributions depicted in reactions 4 and 5. A second avenue does not appear to involve the TFA-diolate (because of no 18O scrambling) and results in complete fragmentation, possibly according reaction 6

TFA(a) + hν f CH3(g) + CF3(g) + CO(g)

(6)

Figure 9. Schematic model for the photodecomposition of TFA-diolate on TiO2(110). Energies were derived from DFT calculations.

Consecutive photoreactions to evolve CO from the surface products of either reaction 4 or 5 can be excluded for two reasons. First, photodecomposition of either acetate to liberate CO would result in 18O in the CO PSD product based on the TPD data presented above, and second, separate photochemical studies of acetate35,40 and trifluoroacetate (see Supporting Information) under UHV conditions do not result in photodecomposition on the TiO2(110) surface. The mechanism of TFA-diolate photodecomposition on TiO2(110) can be expressed as a hole-mediated process. The TFA-diolate, depicted in Figure 9, reacts with a VB hole and becomes “ionized” (or “neutralized” depending on the charge on the ground state species), generating an unstable TFA-diolate. From there, the unstable TFA-diolate can regain an electron and be quenched back to its original state (not depicted) or be decomposed by pathways that best minimize the available energy. DFT calculations were conducted to estimate the heat of reaction associated with removal of the CH3 group versus the CF3 group from the ground state TFA-diolate in an attempt to characterize which pathway is most favorable. Starting with an oxygen adatom on TiO2(110), DFT estimates a heat of adsorption of TFA of -132 kJ/mol to form the TFA-diolate species that bridge bonds to surface Ti4+ sites. This value is somewhat higher than what is estimated from the reverse reaction (decomposition of TFA-diolate to liberate gas phase TFA) obtained from TPD (+96 kJ/mol), however, both estimates illustrate the strong binding of the TFA-diolate to the TiO2(110) surface. Starting with the TFA-diolate as a reference point, DFT indicates that the thermodynamics of C-CF3 bond cleavage, resulting in gas phase CF3 and a surface acetate, should be favored by ∼50 kJ/mol over that of C-CH3 bond cleavage, resulting in gas phase CH3 and a surface trifluoroacetate. Comparison of the amounts of photoinduced C-CH3 versus C-CF3 bond cleavage based on TPD (0.07 versus 0.10 ML, respectively; see above) indicates that both channels were accessed. This suggests that in the TFA-diolate case thermodynamics (alone) does not dictate the partitioning between these two pathways. Cleavage of the C-CF3 bond is highly favored, but not exclusively observed. One possibility is that dynamics

Photochemistry of TFA on Rutile TiO2(110)

J. Phys. Chem. C, Vol. 114, No. 40, 2010 16907 incrementally exclude higher energy (lower wavelength) photons from the incident light emitted from the Hg arc lamp used in this study. The integrated CH3 and CF3 PSD signals were monitored and expressed in Figure 10 as ratios relative to that of the CO PSD signal. (Ratios are employed because the overall yields decreased as more of the UV spectrum was eliminated to the point that no signals were detected with a 400 nm cutoff filter.) A portion of the CH3 and CF3 signals likely came from the same channel responsible for CO, but a significant portion came from the diolate channel, providing a reference in which to compare the wavelength dependent yield of CO. The data for the CO-to-CF3 ratio is scattered mostly because of the weak signals associated with the CF3. In contrast, the CO-to-CH3 ratio increased slightly (from ∼1.5 using a 385 nm filter to ∼1.8 using a 235 nm filter). However, considerable CO signal was seen with photon energies restricted to near the TiO2 band gap energy, and the “no filter” case did not follow the trend. On the basis of these data, there is no strong wavelength dependence in the CO PSD signal yield. These observations provided additional support for at least two distinct photochemical pathways for TFA on TiO2(110), one that followed the TFAdiolate species and another that involves an unknown TFA species whose photodecomposition to CO (and other fragments), is, as yet, not well understood.

Figure 10. Changes in the CO-to-CH3 and CO-to-CF3 PSD ratios as a function of cutoff filter energies.

in the “excited” charge transfer state plays a role in directing the photodecomposition reaction. For example, the CF3 radical is 4.5 times “heavier” than the CH3 radical, suggesting that a lower kinetic energy (relative to that of CH3) on the repulsive side of the charge transfer event may contribute to its lower yield. An issue in describing TFA-diolate photochemistry on TiO2(110) then becomes accurately describing the energy of the ionized diolate and how energy is subsequently partitioned between C-CH3 and C-CF3 bond breaking reaction coordinates. The charges of an adsorbed TFA-diolate were examined through Bader analysis.41,42 The electronegative CF3 group had a net charge of -0.02 e-, while the CH3 group had a charge of +0.22 e-. This indicates CF3 and CH3 groups should presumably interact with nearby holes differently. Transfer of a hole, or partial hole “density”, to CH3 may be more accessible than otherwise expected, due to its partial positive charge. This may explain why CH3 fragmentation is observed despite the thermodynamic preference for CF3 fragmentation. Further studies however are warranted in order to further define the photoexcited diolate state. While the details of TFA-diolate photodecomposition dynamics are unclear, it is obvious from the depiction in Figure 9 that complete fragmentation from the TFA-diolate state to all gas phase products (CH3, CF3, and CO) and a surface O adatom is highly unlikely from a thermodynamic perspective. Starting again with the TFA-diolate as a reference point, DFT estimates that the energy required for complete fragmentation is greater than 460 kJ/mol (>4.8 eV). Even if one were to assume that all of the energy resulting from an e-/h+ pair neutralization event were concentrated into the TFA-diolate, only e-/h+ pairs generated with g4.8 eV would cause complete fragmentation. Given that thermalization of charge carriers to the band edges is rapid in TiO2 (on the picosecond to subpicosecond time scale43-46), it is extremely unlikely that “hot” charge carriers are responsible for the observed CO PSD. Nevertheless, Figure 10 explores in a simple way the possibility that nonthermalized charges might contribute to the observed CO PSD signal. In this figure, a series of “cutoff” filters were employed to

4. Conclusions Photooxidation of trifluoroacetone on rutile TiO2(110) occurred by two processes. The first involved formation and photodecomposition of a TFA-diolate species into two sets of products, gas phase CF3 + surface acetate and gas phase CH3 + surface trifluoroacetate. The partitioning between these two sets of products did not match the thermodynamics of the two reactions, suggesting that energy partitioning in the “excited” state was key to describing how the TFA-diolate photodecomposed. The second photodecomposition process involved compete fragmentation to CO (and most likely CH3 and CF3). This process did not involve molecularly adsorbed TFA and did not pass through the TFA-diolate species based on oxygen isotope studies. Acknowledgment. The authors thank Michel Dupuis, Dave Dixon, and Bob Hamers for their insights. Work reported here was supported by the U.S. Department of Energy, Office of Basic Energy Sciences, Division of Chemical Sciences, Geosciences, and Biosciences. Pacific Northwest National Laboratory is a multiprogram national laboratory operated for the U.S. Department of Energy by the Battelle Memorial Institute under contract DEAC06-76RLO1830. The experimental studies reported here were performed in the William R. Wiley Environmental Molecular Science Laboratory (EMSL), a Department of Energy user facility funded by the Office of Biological and Environmental Research. Computational resources were provided by the Molecular Science Computing Facility located in EMSL and the National Energy Research Scientific Computing Center in Berkeley, CA. Supporting Information Available: This material is available free of charge via the Internet at http://pubs.acs.org. References and Notes (1) Fujishima, A.; Zhang, X.; Tryk, D. A. Surf. Sci. Rep. 2008, 63, 515. (2) Shiraishi, Y.; Hirai, T. J. Photochem. Photobiol., C 2008, 9, 157. (3) Fujishima, A.; Rao, T. N.; Tryk, D. A. J. Photochem. Photobiol., C 2000, 1, 1.

16908

J. Phys. Chem. C, Vol. 114, No. 40, 2010

(4) Thompson, T. L.; Yates, J. T., Jr. Chem. ReV. 2006, 106, 4428. (5) Linsebigler, A. L.; Lu, G.; Yates, J. T., Jr. Chem. ReV. 1995, 95, 735. (6) Henderson, M. A. J. Phys. Chem. B 2005, 109, 12062. (7) Henderson, M. A. J. Catal. 2008, 256, 287. (8) Henderson, M. A. J. Phys. Chem. C 2008, 112, 11433. (9) Henderson, M. A. Surf. Sci. 2008, 602, 3188. (10) Zehr, R. T.; Henderson, M. A. Surf. Sci. 2008, 602, 2238. (11) Henderson, M. A. J. Phys. Chem. B 2004, 108, 18932. (12) Epling, W. S.; Peden, C. H. F.; Henderson, M. A.; Diebold, U. Surf. Sci. 1998, 413, 333. (13) Henderson, M. A.; Epling, W. S.; Perkins, C. L.; Peden, C. H. F.; Diebold, U. J. Phys. Chem. B 1999, 103, 5328. (14) Afeefy, H. Y.; Liebman, J. F.; Stein, S. E. Neutral Thermochemical Data. In NIST Chemistry WebBook, NIST Standard Reference Database Number 69; Linstrom, P. J., Mallard, W. G., Eds.; National Institute of Standards and Technology: Gaithersburg MD, 2008; http://webbook.nist.gov. (15) Henderson, M. A. Surf. Sci. 1994, 319, 315. (16) Van deVondele, J.; Krack, M.; Mohamed, F.; Parrinello, M.; Chassaing, T.; Hutter, J. Comput. Phys. Commun. 2005, 167, 103. (17) CP2K deVelopment home page; 2009; http://cp2k.berlios.de (18) Lippert, G.; Hutter, J.; Parrinello, M. Theor. Chem. Acc. 1999, 103, 124. (19) Goedecker, S.; Teter, M.; Hutter, J. Phys. ReV. B 1996, 54, 1703. (20) Krack, M. Theor. Chem. Acc. 2005, 114, 145. (21) Perdew, J. P.; Burke, K.; Ernzerhof, M. Phys. ReV. Lett. 1996, 77, 3865. (22) Martin, I.; Langer, J.; Stano, M.; Illenberger, E. Int. J. Mass Spectrom. 2009, 280, 107. (23) Handbook of Chemistry and Physics, 82nd ed.; Lide, D. R., Ed.; CRC Press: New York, 2002. (24) Redhead, P. A. Vacuum 1962, 12, 203. (25) King, D. A. Surf. Sci. 1975, 47, 384. (26) Kim, K. S.; Barteau, M. A. Langmuir 1988, 4, 945.

Zehr et al. (27) Kim, K. S.; Barteau, M. A. J. Catal. 1990, 125, 353. (28) Fukui, K.-I.; Iwasawa, Y. Surf. Sci. 2000, 464, L719. (29) Guo, Q.; Cocks, I.; Williams, E. M. J. Chem. Phys. 1997, 106, 2924. (30) Ikuta, S.; Yoshihara, K.; Shiokawa, T. Chem. Lett. 1975, 4, 289. (31) Yamashita, S. Bull. Chem. Soc. Jpn. 1974, 47, 1373. (32) Linsebigler, A.; Lu, G. Q.; Yates, J. T. J. Chem. Phys. 1995, 103, 9438. (33) Tarnovsky, V.; Kurunczi, P.; Rogozhnikov, D.; Becker, K. Int. J. Mass Spectrom. Ion Processes 1993, 128, 181. (34) Du, Y.; Dohnalek, Z.; Lyubinetsky, I. J. Phys. Chem. C 2008, 112, 2649. (35) Idriss, H.; Legare, P.; Maire, G. Surf. Sci. 2002, 515, 413. (36) Fukui, K.-i.; Onishi, H.; Iwasawa, Y. Appl. Surf. Sci. 1999, 140, 259. (37) Henderson, M. A. Surf. Sci. 1998, 400, 203. (38) Linsebigler, A.; Lu, G. Q.; Yates, J. T. J. Phys. Chem. 1996, 100, 6631. (39) Thompson, T. L.; Diwald, O.; Yates, J. T., Jr. J. Phys. Chem. B 2003, 107, 11700. (40) Henderson, M. A., unpublished results. (41) Bader, R. F. W. Acc. Chem. Res. 1985, 18, 9. (42) Henkelman, G.; Arnaldsson, A.; Jonsson, H. Comput. Mater. Sci. 2006, 36, 354. (43) Tamaki, Y.; Furube, A.; Murai, M.; Hara, K.; Katoh, R.; Tachiya, M. J. Am. Chem. Soc. 2006, 128, 416. (44) Tamaki, Y.; Furube, A.; Katoh, R.; Murai, M.; Hara, K.; Arakawa, H.; Tachiya, M. C. R. Chim. 2006, 9, 268. (45) Tamaki, Y.; Furube, A.; Murai, M.; Hara, K.; Katoh, R.; Tachiya, M. Phys. Chem. Chem. Phys. 2007, 9, 1453. (46) Turner, G. M.; Beard, M. C.; Schmuttenmaer, C. A. J. Phys. Chem. B 2002, 106, 11716.

JP910507K