Evidence for O2-Radical Stabilization at Surface Oxygen Vacancies

Jun 21, 2007 - The Journal of Physical Chemistry C 0 (proofing), ... Massimiliano D'Arienzo , Anna Musinu , Stefano Polizzi , Marco Sommariva , Andrea...
1 downloads 0 Views 127KB Size
Download PDF
10630

J. Phys. Chem. C 2007, 111, 10630-10638

Evidence for O2- Radical Stabilization at Surface Oxygen Vacancies on Polycrystalline TiO2 Emma Carter, Albert F. Carley, and Damien M. Murphy* School of Chemistry, Cardiff UniVersity, Main Building, Park Place, Cardiff CF10 3AT, U.K. ReceiVed: April 16, 2007; In Final Form: May 18, 2007

The characterization and stability of superoxide radicals (O2-) over polycrystalline TiO2 (Degussa P25) was investigated using electron paramagnetic resonance (EPR) spectroscopy. The adsorbed oxygen molecules act as efficient electron scavengers and were therefore used to indirectly probe the sites of electron transfer at the surface of the anatase component of the mixed phase P25 material. A distribution of various stabilization sites on the surface was identified by analysis of the g values and further confirmed by identification of several well-resolved 17O hyperfine patterns. For the first time, on a polycrystalline TiO2 surface evidence for stabilization of superoxide radicals specifically at anion vacancy sites is presented by EPR. These radicals, labeled [Vac...O2-], are characterized by the spin Hamiltonian parameters of gxx ) 2.005, gyy ) 2.011, gzz ) 2.019, and 17OAxx ) 7.64 mT (Ayy ) Azz > 1 mT). The [Vac...O2-] radicals exhibit pronounced reactivity under the influence of thermal, photochemical, and chemical treatment compared to the remaining surface O2- anions bound at nonvacancy sites. The extent of site occupancy was found to be sensitive to the oxygen adsorption temperature and the extent of O2- radical migration on the surface. Thus, the stability and lifetime of the surface O2- anions are directly correlated to the structure of the adsorption site itself at the anatase surface of P25.

1. Introduction The photocatalytic activity of semiconductor metal oxides such as titanium dioxide (TiO2) is of considerable interest because of its potential use in a wide range of applications such as sterilization,1 solar energy conversion,2 and pollution control.3 On absorption of a photon with energy equal to or greater than the band gap of the semiconductor, an electron/hole pair is generated in the bulk. These charge carriers migrate toward the catalyst surface where they can participate in redox reactions with adsorbed molecules. While the bulk properties of TiO2, such as its band gap, are clearly related to its photocatalytic activity, the surface properties are equally important in electron and hole mediated transfer processes. There are several different formulations of TiO2 that are readily available as catalysts. In this study, the mixed-phase Degussa P25 TiO2 (∼80/20 anatase/rutile) is used, which is reported in several instances4,5 to have an increased photocatalytic activity compared to the pure-phase materials. The morphology at the interface of the mixed-phase catalyst is still largely undetermined, although several bulk methods have been used in an attempt to characterize the average properties of the TiO2 samples and elucidate the role of the surface in controlling catalytic activity. One intriguing aspect concerns the nature of the surface oxygen species formed on metal oxides which are known to be dependent on surface morphology. As previous work has shown, anisotropic differences exist in the surface properties of the anatase and rutile polymorphs of TiO26,7 and it was of interest to explore if electron paramagnetic resonance (EPR) could be used to identify the speciation of oxygencentered radicals on the mixed-phase P25 surface. Hurum et al.,6,7 have for example illustrated the complexity of the charge separation processes that can occur in P25 and have clearly * Corresponding author: tel, 00 44 2920 875850; fax, 00 44 2920 874030; e-mail, [email protected].

demonstrated how catalytic “hot spots” can be created at the rutile-anatase interface. If electron transfer occurs at these focal points, the resulting radicals may therefore be expected to be associated (or stabilized) on the anatase and/or rutile phase of the P25 material. In addition to the surface-stabilized radicals, the nature of the intrinsic oxygen anion vacancy defect sites is also of significant interest as it is widely accepted that these sites are important in controlling the surface chemistry of TiO2. For example, in their study of the photo-oxidation of CH3Cl over TiO2(110), Lu et al.8 demonstrated that the photocatalytic sites responsible for the reaction were only present on the vacuum annealed surface possessing surface oxygen vacancies. Numerous further studies9,10 have appeared in the literature describing how CO2 can be used as a probe molecule on vacuum annealed single crystals to quantify the effects of oxygen defects. Henderson et al.11 used temperature-programmed desorption (TPD) to study the adsorption of CO2 at the vacuum annealed TiO2(110) surface. Their data suggested that CO2 preferentially adsorbs nondissociatively onto the vacancy sites and then at five-coordinate Ti4+ sites. On exposure of molecular oxygen to a surface containing CO2 adsorbed only in the vacancy sites, no displacement of the CO2 from the vacancy occurs. Further, on heating the sample to desorb the CO2 from the vacancy sites, the adsorbed O2 does not move into the vacancy sites to reoxidize the surface.11 The vast majority of work related to the characterization of oxygen vacancies has however been confined to well-defined single-crystal surfaces. While vacancy defects are widely implicated as reactive centers in polycrystalline materials, their direct characterization on powders has proven to be extremely difficult. Charge transfer between the perfect or defective surface and adsorbed oxygen molecules leads to several kinds of charged oxygen species including O2+, O2-, O22-.6,12 The use of EPR spectroscopy is therefore fundamental in the identification and

10.1021/jp0729516 CCC: $37.00 © 2007 American Chemical Society Published on Web 06/21/2007

Superoxide Radical Stabilization characterization of these radicals.13-15 Of the many reactions possible with molecular oxygen at the surface, the formation of O2- is thermodynamically most favorable. In this study, the superoxide radical anion has been used to indirectly probe the morphology of the TiO2 polycrystalline surface. In particular multiple adsorption sites for the anion can be identified by analysis of the orthorhombic O2- signal and the 17O hyperfine tensor.16,17 Among the many surface sites available for O2stabilization on TiO2, one site in particular is unusual in terms of the stability and reactivity of the bound O2- radical compared to the radicals at other sites. As will be shown here, this unusual site can be assigned to an oxygen vacancy defect on the anatase surface. This study therefore represents, to the best of our knowledge, the first ever identification of an O2- anion stabilized at an oxygen vacancy defect on polycrystalline TiO2 by EPR and, furthermore, provides evidence of the unusual (and higher) reactivity displayed by the radicals when stabilized at these vacancies. 2. Experimental Section 2.1. Chemicals. Three samples of TiO2 were used in this study, primarily the mixed-phase P25 material and for comparison a single-phase rutile and single-phase anatase material. The mixed-phase P25 was supplied by Degussa (∼80% anatase; ∼20% rutile) with a surface area of ∼50 m2 g-1. The characterization and sample morphology of this material are described elsewhere,18 and it is reported that platelike particles are present with an average particle size of ca. 40 nm. Transmission electron microscopy (TEM) images also show that the (001) and (010) faces of the anatase phase are most abundant, with a lower fraction of the (110) plane.18 The single-phase (high surface area) rutile and anatase materials were kindly supplied by Dr. T. A. Egerton (Newcastle University) and prepared according to the literature.19,20 The 16O2 was supplied by BOC Ltd. and the 17O-labeled dioxygen gas (63% enrichment) by Icon Services Inc., USA. 2.2. Sample Preparation. The polycrystalline TiO2 powder (ca. 10 mg) was slowly heated (over a 5 h period) under dynamic vacuum (10-4 Torr) in an EPR cell up to a maximum temperature of 773 K. This vacuum reduced sample (blue in color) was then held at this temperature for a further 1 h, before exposure to excess oxygen (50 Torr) at this temperature. The sample was cooled under the oxygen atmosphere producing a clean oxidized surface. The residual oxygen was subsequently evacuated at room temperature. It should be noted that exposure of the 773 K thermally reduced TiO2 powder to molecular oxygen at 298 K will produce paramagnetic oxygen radicals. By comparison, oxygen exposure at 773 K results in the formation of diamagnetic surface O2- lattice anions and therefore produces a clean, reoxidized sample; hereafter this sample is referred to as the actiVated sample. This was confirmed by recording the EPR spectrum of the activated sample, which did not display any paramagnetic signal. The nonstoichiometric sample reduced under vacuum at 773 K is blue in color; hereafter this sample is referred to as the reduced sample. The reduced sample was subsequently exposed to 10 Torr O2 at 120 or 298 K and left under the oxygen atmosphere for 10 min before evacuation of the excess molecular oxygen at 120 or 298 K. The activated sample was also exposed to 10 Torr O2 at 298 K followed by UV photolysis at 77 K for a period of 30 min. The excess molecular oxygen was subsequently evacuated at 120 or 298 K. The UV photolysis was performed both in situ and ex situ from the EPR cavity using

J. Phys. Chem. C, Vol. 111, No. 28, 2007 10631

Figure 1. X-band cw-EPR spectra (recorded at 130 K) of P25 after thermal reduction under vacuum at (a) 573 K and (b) 823 K.

a 1000 W Oriel Instruments UV lamp, incorporating a Hg/Xe arc lamp (250 to >2500 nm) in the presence of a water filter. The UV output below 280 nm accounts for only 4-5% of the total lamp output. The EPR spectra were then recorded at variable temperatures in order to study the decay characteristics of the radicals. 2.3. Characterization. The EPR spectra were recorded on an X-band Bruker EMX spectrometer operating at 100 kHz field modulation and 10 mW microwave power and equipped with a high-sensitivity cavity (ER 4119HS). The g values were determined using a DPPH standard. EPR computer simulations were performed using the SimEPR32 program.21 3. Results 3.1. Heterogeneity of Surface O2- Anions. Two samples of Degussa P25 were annealed under vacuum at 573 and 723 K, respectively. This treatment results in the formation of a bluecolored sample and the corresponding EPR signals shown in Figure 1. Both of these spectra can be regarded as composite signals arising from the presence of both bulk and surface Ti3+ cations, formed during thermal reduction. The EPR characteristics of these Ti3+ centers have been thoroughly described elsewhere in the literature.6,7,22-24 The lower reduction temperature of the polycrystalline TiO2 sample is also a common feature of TiO2 powders25 compared to single crystals,26 but in the current case the Ti3+ defect concentration is still estimated to be far less than 1% of the surface. The most important point of relevance to the current study is that more reduced Ti3+ cations are generated at the higher reduction temperatures (compare signal intensities of Figure 1) as expected. At lower reduction temperatures features associated with bulk (g⊥ ) 1.990, g|| ) 1.957) and surface (g ) 1.930) Ti3+ on the anatase phase can be clearly identified, as previously resolved and discussed in detail by Hurum et al.6,7 on photoirradiated P25. At higher reduction temperatures, an additional isotropic signal is also seen centered close to ge. The isotropic signal indicates the presence of an unpaired electron located in a spherically symmetrical environment, and it has been previously assigned to medium-polarized conduction electrons.27,28 Upon exposure of the above reduced samples to 16O2 at 298 K, the EPR signals due to the reduced Ti3+ cations disappear immediately, and simultaneously a new EPR signal shown in Figure 2 appears. At the lower reduction temperature, an orthorhombic EPR signal with the g values of gxx ) 2.005, gyy ) 2.011, and gzz ) 2.019 is observed, which can be easily and

10632 J. Phys. Chem. C, Vol. 111, No. 28, 2007

Carter et al.

Figure 2. EPR spectra (recorded at 130 K) of O2- radicals formed by oxygen exposure at 298 K to a thermally reduced P25 sample annealed at (a) 573 K and (c) 823 K. The corresponding simulated spectra are shown in (b) and (d), respectively.

TABLE 1: The g and 17OA Values for O2- Radicals Formed on Thermally Reduced P25 Surface site

gxx

gyy

gzz

Axx/mT

Ayy/mT

Azz/mT

% cont

I II II′ III

2.005 2.004 2.004 2.001

2.011 2.011 2.011 2.011

2.019 2.023 2.020 2.026

7.64 7.86

II′ > I (Figure 6f). Despite this reversal in distribution, the overall integrated signal intensity of the spectra in Figure 6 remain constant during annealing. This implies that the site I superoxide species is not destroyed but instead must undergo a redistribution from site I to another surface site during annealing. A similar set of results was also obtained following oxygen exposure at 210 K and subsequent annealing to 298 K. One seemingly conflicting result between the 120 K versus 298 K oxygen exposure experiments pertains to the final intensities of the site I superoxide species observed in the roomtemperature EPR spectrum. According to Figure 2, when O2 is exposed to a thermally reduced sample directly at 298 K, the site I species is most intense. However, according to Figure 6, when O2 is exposed to a thermally reduced sample at 120 K and the sample is subsequently “annealed” to 298 K, the site I species now has the lowest abundance (Figure 6f). One simple explanation for these results relates to the accessibility and temperature dependency of the various Ti3+ centers (i.e., surface Ti3+, subsurface Ti3+, conduction electrons) for electron transfer to adsorbed O2. At 298 K more sites are clearly thermally accessible for electron transfer, and this facilitates the overall formation of higher concentrations of O2- radicals in the presence of adsorbed oxygen. To confirm this, the above low-temperature experiment was repeated (i.e., oxygen was adsorbed at 120 K on the thermally reduced sample, evacuated at 210 K, and finally annealed to 298 K under a dynamic vacuum) and a second dose of molecular oxygen was subsequently added to the sample at 298 K. The resulting EPR spectra are shown in Figure 7 before and after exposure to the second oxygen dose. Analysis of the spectra reveals that the overall integrated signal intensity increases from spectrum a to spectrum b in Figure 7, as more O2- radicals are generated at 298 K. Although molecular oxygen had been exposed to the reduced sample at 120 K, the excess oxygen was then evacuated and therefore not available for electron scavenging at 298 K. Furthermore, the overall increased signal

Carter et al.

Figure 7. (a) EPR spectrum (130 K) of the O2- anion formed by O2 exposure at 120 K to a thermally reduced P25 sample, previously annealed at 823 K, followed by evacuation at 120 K. The sample was then subsequently annealed under dynamic vacuum at 298 K. A second dose of oxygen (10 Torr) was then added at 298 K, shown in (b). The relative integrated signal intensities are also given.

Figure 8. (a) EPR spectrum (recorded at 130 K) of O2- radicals formed by photoirradiating P25 under an O2 atmosphere and subsequent evacuation at 298 K. The corresponding simulated spectrum is shown in (b).

intensity in Figure 7b is dominated by the preferential increase in the site I species (at gzz ) 2.019) compared to the site II-III species. This result also demonstrates the peculiar reactivity of site I for electron transfer, since O2- formation occurs specifically at this site under these conditions. 3.4. Thermal versus Photogenerated O2-. Superoxide radicals can also be photogenerated on P25 by UV irradiation of the oxidized sample under an oxygen atmosphere. In this experiment, molecular oxygen (10 Torr) was admitted to the clean, oxidized P25 sample and irradiated at 77 K for a period of 30 min. The resulting spectra are shown in Figure 8a. The gzz region of this sample, which reflects differences in the surface adsorption sites, now shows even greater complexity compared to the thermally generated spectra discussed above. The integrated signal intensities for the superoxide radicals formed on the photogenerated and thermally reduced samples reveal a higher concentration of radicals for the latter sample. Simulation of the EPR spectrum for photogenerated O2- (shown in Figure 8b) also revealed that the site I species at gzz ) 2.019 accounts for only 15% of the overall signal intensity of the photogenerated superoxide radicals. In contrast, this peak contributes over 40% to the overall signal intensity on the thermally generated sample (Figure 2c). However, owing to the

Superoxide Radical Stabilization

J. Phys. Chem. C, Vol. 111, No. 28, 2007 10635

TABLE 2: Comparison of g Values Obtained for O2Radical on Different Polycrystalline TiO2 Surfaces polymorph

g1 (gzz)

g2 (gyy)

g3 (gxx)

ref

anatase

(I) 2.019 (II) 2.023 (III) 2.029 2.019 2.024 2.023 2.029 2.022 2.030

2.011 2.011 2.011 2.009 2.009 a

a a a 2.003 2.003 a

b 29 30 b

2.011 2.008

2.008 2.004

31 24

anatase anatase rutile rutile rutile a

Unresolved. b This study.

photolability of this superoxide species (see Figure 5), this reduced signal intensity is expected in the photogenerated case. 3.5. Morphological Considerations: Anatase versus Rutile. All of the results discussed so far have been recorded on the mixed-phase Degussa P25 (80% anatase, 20% rutile). The photochemistry of the interfacial sites between the anatase and rutile phases of P25 has been studied in detail.6,7 It is known that the interfacial sites can slow the recombination of photogenerated charge carriers by electron transfer from rutile to anatase6,7 and therefore enhance photocatalytic activity. To determine whether the superoxide radicals formed via electron transfer from P25 are stabilized at the anatase or rutile component of the sample, we examined the nature of the O2signals formed on the single polymorphs (i.e., on pure phase anatase and rutile). The EPR spectrum for thermally and photochemically generated O2- on the pure anatase material (Table 2; supplementary Figure S2) was identical to that observed on P25. In contrast the O2- species on rutile are unstable at 298 K but can be observed at lower temperatures (120 K), where the distribution of gzz peaks is different compared to that observed on anatase and on P25. This difference in stability of O2- on polycrystalline anatase versus rutile is not well understood in the literature. Therefore, one may conclude that the observed O2- species on the mixed-phase P25 are actually stabilized at surface sites on the anatase phase of the material. These results confirm the model of electron charge transfer from rutile to anatase as described by Hurum et al. 6,7 4. Discussion 4.1. Nature of the O2- Radical. During thermal treatment of TiO2 under vacuum, both surface and bulk Ti3+ centers are formed from O2 loss from the oxide, according to the reactions

2O(latt)2- f O2(g) + 4e-

(1)

4Ti4+ + 4e- f 4Ti3+

(2)

Diffusion of bulk defects to the surface readily occurs in TiO2,26,35 but at the moderately low temperatures employed here, the majority of the defects are formed at the surface and higher temperatures (1073 K) are required for substantial formation of bulk defects.25 Addition of oxygen at 298 K to the reduced metal oxide results in the formation of O2- in the following exothermic reaction

Ti

3+

-

+ O2 f Ti ‚‚‚O2 4+

(3)

In the case where the excess electrons are stabilized in the conduction band, superoxide formation may also occur via the additional reaction Ti4+ + eCB + O2 f Ti4+‚‚‚O2-.15

The nature of the superoxide radical is generally described using an ionic model, in which electron transfer from the surface to adsorbed oxygen occurs.29 The unpaired electron lies in a π* orbital which results in orthorhombic symmetry (gxx, gyy, and gzz) where the z direction is usually defined along the internuclear axis of the superoxide radical and the y direction is perpendicular to the plane of the surface. The g tensor can be analyzed using the simplified equations gxx ≈ ge, gyy ) ge + 2λ/E, and gzz ) ge + 2λ/∆, originally derived for superoxide impurities in the bulk of alkali halides,30 where λ is the spinorbit coupling constant of oxygen, E is the energy level separation between the σgz and the highest occupied π* orbital, and ∆ is the energy separation between the two 2π* antibonding orbitals due to the electric field generated by the metal cation Mx+ at the adsorption site. From these equations it can be seen that the gzz component is most sensitive to the electric field experienced by the adsorbed anion and can therefore be used as a probe of surface electric field gradients.29 The lower the gzz value the lower the coordination of the metal cation for a given oxide surface.29 This arises from the effects of local coordination on the Madelung constant36,37 which produces a distribution of O2- species as elegantly shown on the MgO surface. In the case of TiO2, any differences in the gzz component will reflect the differences in the electric fields at the adsorption site, as generated for example by slight differences in Ti4+ coordination. Low coordinated Ti4+ cations, being more exposed, will therefore exert stronger electric fields on the π* orbital of O2- and therefore produce lower gzz values, such as 2.019. On the low temperature reduced P25 sample, only a single superoxide species was formed after oxygen adsorption at 298 K (Figure 2a). As the sample becomes more highly reduced at higher temperatures, other superoxide sites with higher gzz values emerge. Therefore the site I O2- species with gzz ) 2.019 must be associated with a surface site possessing Ti4+ in low coordination relative to the site II-III species. 4.2. Information from the 17O Hyperfine Tensor. The g tensor gives limited information on the nature of the oxygen species, and further details can be obtained from the 17O hyperfine tensor. A singly labeled superoxide radical (17O16O)should give a hyperfine pattern of 6 lines along one axis of the hyperfine tensor (the other two components are always very small) while a doubly labeled superoxide radical (17O17O)should give an 11 line pattern along the same axis. The sign and absolute direction of the hyperfine tensor cannot be obtained unambiguously from a powder spectrum, but in the following discussion the largest hyperfine value is defined as Axx, as commonly observed for surface-adsorbed O2-.16,17,29 Analysis of the 17O tensor provides information on the spin density at the oxygen atoms. For an unpaired electron in an axially symmetrical πg orbital, the axial hyperfine tensor can be separated into an isotropic part (aiso) and an anisotropic traceless tensor of the form

(A⊥, A⊥, A||) ) aiso + (-B, -B, 2B)

(5)

aiso ) (2A⊥ + A||)/3

(6)

where

The isotropic and anisotropic terms are then given by aiso ) A0c2s2 and B ) B0c2p2, where c2s2 and c2p2 are the spin densities of the unpaired electron in the 2s and 2p orbitals of the oxygen atoms and A0 and B0 are the theoretical hyperfine constants for a pure 2s and 2p oxygen orbital. Several values for A0 and B0

10636 J. Phys. Chem. C, Vol. 111, No. 28, 2007 have been reported, but here we use the values of A0 ) -165.1 and B0 ) -5.138 mT as described by Morton38 and Chiesa et al.17 The 2s spin density arises from a spin polarization mechanism, and since this effect is small in O2-, only the 2px populations need to be calculated (since Ayy and Azz are assumed to be close to zero). The Axx values for the different 17O2- species on P25 were given in Table 1 (i.e., 7.64 mT for site I, 7.86 mT for site II/II′, and 7.97 mT for site III), and were used to calculate aiso and B. The corresponding total spin density on the O2- anion at the different sites was therefore found to be close to unity in all cases in agreement with other metal oxide systems (Supplementary Table 1). A spin density close to unity indicates that the electron is almost completely localized on the two oxygen atoms and most importantly confirms that all of the observed O2- radicals are symmetrically bound to the surface (i.e., “sideon” rather than “end-on” mode) since the spin densities in both oxygen nuclei are equivalent. 4.3. Nature of the gzz ) 2.019 site: The [Vac...O2-] Center. As discussed in section 4.1, the magnitude of the gzz component gives an indication of the degree of Ti4+ coordination at the stabilization site. However, studies of polycrystalline samples are unable to determine the exact position or identity of the surface sites responsible for O2- stabilization. Nevertheless it is clear from the above results that the site I O2- species displays different thermal and photochemical behavior compared to the other site II-III species. Furthermore in a recent study, this site I species was also shown to undergo preferential reactivity with adsorbed acetone.39 All these collective results indicate that the stabilization site responsible for the 2.019 component is substantially different compared to the other surface sites. During the thermovacuum reduction of TiO2, oxygen atoms are removed from the surface resulting in vacancy sites. The Ti4+ cations at these vacancy sites become reduced to Ti3+, and subsequent electron transfer to adsorbed oxygen will produce the Ti4+‚‚‚O2- radical. It has been shown theoretically on MgO containing surface color centers, that the O2- radicals remain stabilized at the site of electron transfer,40 due to the strong electrostatic attraction between the metal cation and the radical anion, and this is also expected on reduced TiO2 particularly at low temperatures. Therefore the initially formed Ti4+‚‚‚O2adduct should be better represented as [Vac...O2-]. The intrinsic Ti4+ cations within the vacancy are sites of low coordination, and this accounts for the low gzz value of 2.019 at site I. While low gzz values are usually associated with surface sites of higher stability (greater Ti4+‚‚‚O2- interaction), the reason for the greater lability of this site (already proven by theory, as discussed below), despite the low gzz value, is currently unknown and will require further insight from theoretical studies. The lower contribution (or abundance) of this site I signal on the clean, activated surface (Figure 8a) is then expected due to the smaller number of vacancies on an oxidized surface compared to the thermally reduced surface41 (Figure 2c). The remaining site II-III species, with higher gzz values, can then be assigned to Ti4+ sites adjacent to or close to the vacancy. This proposal is also supported by experimental data on single-crystal TiO2 surfaces. For example, Lu et al.42,43 performed a series of photodesorption experiments on a rutile TiO2(110) surface to elucidate the nature of the adsorbed oxygen species on the defective surface. Following oxygen exposure to a sample previously annealed to 900 K under ultrahigh vacuum conditions, the maximum coverage of O2 was 250 K. The thermal conversion process between the two modes of oxygen adsorption was due to an alteration of the O2 adsorption configuration at an anion vacancy defect.42,43 These results indicate that the oxygen adsorbed in the β-O2 state is more strongly bound to the surface. These results are also consistent with the current findings on P25. Oxygen adsorption at 298 K on the thermally reduced surface resulted in a different distribution of superoxide radicals

J. Phys. Chem. C, Vol. 111, No. 28, 2007 10637 (Figure 2c) compared to oxygen adsorption at 120 K followed by annealing to 298 K (Figure 6f). During the annealing of the sample from 120 to 298 K, the overall signal intensity remained constant, indicating migration of the superoxide radicals on the surface from the [Vac...O2-] sites to the more strongly bound nonvacancy sites II and III (analogous to the R f β transition observed by Lu et al.42,43). 5. Conclusion The surface chemistry of TiO2 has been widely studied in the past.25 From studies of single-crystal samples, the structures of the perfect and defective surfaces are well understood, revealing that thermally annealed surfaces contain large numbers of oxygen vacancies associated with higher activity.49 Most of the available experimental evidence from single-crystal studies implies a higher activity of the oxygen species associated with vacancy sites compared to nonvacancy sites. While vacancy and nonvacancy oxygen species are also expected to occur on polycrystalline TiO2, there is little experimental evidence to date to differentiate between these different species on powders. Here we have found using EPR that superoxide radicals are formed and stabilized at the anatase component of the mixed phase P25 material, consistent with the model that electron transfer occurs at the anatase surface in P25. A distribution of O2- species is formed, and these have been classified as [Vac...O2-] species (site I) and the nonvacancy Ti4+‚‚‚O2species (site II-III). The former [Vac...O2-] species can be identified by the spin Hamiltonian parameters of gxx ) 2.005, gyy ) 2.011, gzz ) 2.019, and 17OAxx ) 7.64 mT (Ayy ) Azz > 1 mT). These parameters are consistent with a “side-on” bonded O2- radical adsorbed at a site of low Ti4+ coordination. This [Vac...O2-] site is also distinguished by its lower thermal stability compared to the nonvacancy sites, since it is unstable at temperatures above 333 K. Owing to the instability of the O2- radical at this vacancy site, surface migration can be easily induced by increasing the temperature, consistent with singlecrystal evidence in the literature.42,43 The [Vac...O2-] species were also found to be photolabile (and chemically reactive39) compared to the superoxide species adsorbed at nonvacancy sites. Again all these results are entirely consistent with experiments performed on well-defined single-crystal surfaces of rutile.42,43 While the temperature range of thermal stabilities for the oxygen species observed on P25 may be different compared to those reported in the literature specifically for rutile singlecrystal surfaces, the generic trends are consistent and (i) clearly indicate that [Vac...O2-] species are present on the polycrystalline surfaces and (ii) these species have different thermal, chemical, and photochemical reactivity compared to the nonvacancy sites. Although EPR has been widely used to study defect sites and oxygen species at oxide surfaces, this is the first time that superoxide radicals have been specifically identified at vacancy sites on a polycrystalline TiO2 surface by this technique. Since the oxygen species are exclusively associated with the anatase component of P25, it suggests that this component may be more active for reactions involving reactive oxygen species. Acknowledgment. Funding from the EPSRC is gratefully acknowledged. Supporting Information Available: Additional spectra on aged superoxide samples, investigation of superoxide radicals stabilized over pure-phase samples of anatase and rutile poly-

10638 J. Phys. Chem. C, Vol. 111, No. 28, 2007 morphs at different oxygen adsorption temperatures, and comparison of spin densities for the superoxide radical stabilized over TiO2 and other metal oxide systems. This material is available free of charge via the Internet at http://pubs.acs.org. References and Notes (1) Maness, P. C.; Smolinski, S.; Blake, D. M.; Huang, Z.; Wolfrum, E. J.; Jacoby, W. A. Appl. EnViron. Microbiol. 1999, 65, 4094. (2) Gra¨tzel, M. Nature 2001, 414, 338. (3) Fox, M. A.; Dulay, M. T. Chem. ReV. 1993, 93, 341. (4) Hoffman, M. R.; Martin, S. T.; Choi, W.; Bahnemann, D. W. Chem. ReV. 1995, 95, 69. (5) Linsebigler, A. L.; Lu, G.; Yates, J. T., Jr. Chem. ReV. 1995, 95, 735. (6) Hurum, D. C.; Agrios, A. G.; Gray, K. A.; Rajh, T.; Thurnauer, M. C. J. Phys. Chem. B 2003, 107, 4545. (7) Hurum, D. C.; Agrios, A. G.; Crist, S. E.; Gray, K. A.; Rajh, T.; Thurnauer, M. C. J. Electron Spectrosc. Relat. Phenom. 2006, 150, 155. (8) Lu, G.; Linsebigler, A. L.; Yates, J. T., Jr. J. Phys. Chem. 1995, 99, 7626. (9) Funk, S.; Burghaus, U. Phys. Chem. Chem. Phys. 2006, 8, 4805. (10) Fink, K. Phys. Chem. Chem. Phys. 2006, 8, 1482. (11) Henderson, M.; Epling, W. S.; Perkins, C. L.; Peden, C. H. F.; Diebold, U. J. Phys. Chem. B 1999, 103, 5328. (12) Murata, C.; Yoshida, H.; Kumagai, J.; Hattori, T. J. Phys. Chem. 2003, 107, 4364. (13) Coronado. J. M.; Javier Maira, A.; Conesa, J. C.; Yeung, K. L.; Augugliaro, V.; Soria, J. Langmuir 2001, 17, 5368. (14) Berger, T.; Sterrer, M.; Diwald, O.; Kno¨zinger, E. ChemPhysChem 2005, 6, 2104. (15) Berger, T.; Sterrer, M.; Diwald, O.; Kno¨zinger, E.; Panayotov, D.; Thompson, T. L.; Yates, J. T., Jr. J. Phys. Chem. B 2005, 109, 6061. (16) Tench, A. J.; Holroyd, P. Chem. Commun. 1968, 8, 471. (17) Chiesa, M.; Giamello, E.; Paganini, M. C.; Sojka, Z.; Murphy, D. M. J. Chem. Phys. 2002, 116, 4266. (18) Martra, G. Appl. Catal. A 2000, 200, 275. (19) Egerton, T. A.; Mattinson, J. A. J. Photochem. Photobiol., A 2007, 186, 4266. (20) Jin, C.; Christensen, P. A.; Egerton, T. A.; Lawson, E. J.; White, J. R. Polym. Degrad. Stab. 2006, 91, 1086. (21) Adamski, A.; Spalek, T.; Sojka, Z. Res. Chem. Intermed. 2003, 29, 793.

Carter et al. (22) Howe, R. F.; Gra¨tzel, M. J. Phys. Chem. 1985, 89, 4495. (23) Howe, R. F.; Gra¨tzel, M. J. Phys. Chem. 1987, 91, 3906. (24) Nakaoka, Y.; Nosaka, Y. J. Photochem. Photobiol., A 1997, 110, 299. (25) Jenkins, C. A.; Murphy, D. M. J. Phys. Chem. B 1999, 103, 1019. (26) Diebold, U. Surf. Sci. Rep. 2003, 48, 53. (27) Serwicka, E.; Schlierkamp, M. W.; Schindler, R. N. Z. Naturforsch. 1981, 36, 226. (28) Naccache, C.; Meriaudeau, P.; Che, M.; Tench, A. J. Trans. Faraday Soc. 1971, 67, 506. (29) Che, M.; Tench, A. J. AdV. Catal. 1983, 32, 1. (30) Kanzig, W.; Cohen, M. H. Phys. ReV. Lett. 1959, 3, 509. (31) Wong, N.-B.; Lunsford, J. H. J. Chem. Phys. 1971, 55, 3007. (32) Wong, N.-B.; Lunsford, J. H. J. Chem. Phys. 1972, 56, 2664. (33) Setaka, M.; Fukazawa, S.; Kirino, Y.; Kwan, T. Chem. Pharm. Bull. 1968, 16, 1240. (34) Dufaux, M.; Che, M.; Naccache, C. C. R. Acad. Sci. 1969, 268, 2255. (35) Henderson, M. A. Surf. Sci. 1999, 419, 174. (36) Garrone, E.; Zecchina, A.; Stone, F. S. Philos. Mag. B 1980, 42, 683. (37) Giamello, E.; Ugliengo, P.; Garrone, E. J. Chem. Soc., Faraday Trans. I 1989, 85, 1373. (38) Morton, J. R. Chem. ReV. 1964, 64, 453. (39) Carter, E.; Carley, A. F.; Murphy, D. M. ChemPhysChem 2007, 8, 113. (40) Chiesa, M.; Paganini, M. C.; Giamello, E.; Murphy, D. M.; Di Valentin, C.; Pacchioni, G. Acc. Chem. Res. 2006, 39, 861. (41) Epling, W. S.; Peden, C. H. F.; Henderson, M. A.; Diebold, U. Surf. Sci. 1998, 412, 333. (42) Lu, G.; Linsebigler, A.; Yates, J. T., Jr. J. Chem. Phys. 1995, 102, 3005. (43) Lu, G.; Linsebigler, A.; Yates, J. T., Jr. J. Chem. Phys. 1995, 102, 4657. (44) Perkins, C. L.; Henderson, M. A. J. Phys. Chem. B 2001, 105, 3856. (45) Rusu, C. N.; Yates, J. T., Jr. Langmuir 1997, 13, 4311. (46) Amore Bonapasta, A.; Filippone, F. Surf. Sci. 2005, 577, 59. (47) Funk, S.; Hokkanen, B.; Burghaus, U.; Ghicov, A.; Schmuki, P. Nano Lett. 2007, 7, 1091. (48) Dyrek, K.; Adamski, A.; Sojka, Z. Spectrochim. Acta, Part A 1998, 54, 2337. (49) Lu, G.; Linsebigler, A.; Yates, J. T. J. Phys. Chem. 1994, 98, 11733.