Direct Imaging of Site-Specific Photocatalytical Reactions of O2 on

Dec 12, 2011 - Photostimulated reactions of single O2 molecules on reduced TiO2(110) surfaces were directly observed at an atomic level with high-reso...
0 downloads 0 Views 3MB Size
Download PDF
Letter pubs.acs.org/JPCL

Direct Imaging of Site-Specific Photocatalytical Reactions of O2 on TiO2(110) Zhi-Tao Wang,† N. Aaron Deskins,‡ and Igor Lyubinetsky*,† †

Environmental Molecular Sciences Laboratory and Institute for Integrated Catalysis, Pacific Northwest National Laboratory, Richland, Washington 99352, United States ‡ Department of Chemical Engineering, Worcester Polytechnic Institute, Worcester, Massachusetts 01609, United States ABSTRACT: Photostimulated reactions of single O2 molecules on reduced TiO2(110) surfaces were directly observed at an atomic level with high-resolution scanning tunneling microscopy at 50 K. Two distinct reactions of O2 desorption and dissociation occur at different active sites of terminal Ti atoms and bridging O vacancies, respectively. Two reaction channels follow very different kinetics. While hole-mediated O2 desorption is promptly and fully completed, electronmediated O2 dissociation is much slower and is quenched above some critical O2 coverage. Evidently, the O2 photochemistry on TiO2(110) is quite more complex than thought previously. Density functional theory calculations indicate that both coordination and charge state of an O2 molecule chemisorbed at the specific site largely determine a particular reaction pathway. SECTION: Surfaces, Interfaces, Catalysis

P

Petrik and Kimmel,26−28 resulting in a healing of the original VO with a new bridging oxygen atom (Ob) and deposition of an oxygen adatom (Oa) at adjacent terminal five-fold coordinated Ti (Ti5c) site29

hotostimulated reactions on TiO2 have attracted much attention due to the variety of potential applications ranging from a hydrogen production by water splitting to environmental remediation through organic pollutant oxidation.1,2 In the majority of these processes, the oxygen plays a crucial role, serving as a simplest oxidizing reagent or as an electron scavenger.3,4 Hence, the physicochemical properties of O2 adsorbed on rutile TiO2(110) (model oxide surface) have been extensively investigated,5−14 and, in particular, the chemisorbed O2 molecules have been recently imaged by scanning tunneling microscopy (STM).15−17 Whereas the O2 desorption from rutile TiO2(110) is the most comprehensively studied photoreaction on TiO2 (by traditional ensembleaveraging techniques), details of its mechanism are still far from being understood. On a basis of extensive research of photostimulated desorption (PSD) of O2 from TiO2(110) by ultraviolet (UV) light, Yates and coworkers have developed a hole-mediated desorption model.3,18,19 This model incorporates several processes, starting with band gap excitation of electron−hole pairs in TiO2 by UV photon absorption, hole diffusion to surface, and recombination with electron or trapping at surface, with the latter followed by hole transfer to chemisorbed O2 anion and neutral O2 desorption

O2− + h+ → O2 (g)↑

O2 2 −/VO + e− → Ob2 − + Oa−/Ti5c

They also demonstrated the existence of nondissociated, also called “photoblind” O2 species, yet the underlying mechanism has not been elaborated.26,27 In addition, whereas prevailing models assume that only sites associated with VO’s are involved in the O2 photochemistry,3,18,19 the role of different sites is still unknown. Here we report the first spatially resolved observation at an atomic level of two site-specific photoinduced reactions of O2 on TiO2(110), imaged by STM at 50 K. We demonstrate that additional active sites should be taken into account to describe properly the O2 photochemistry on TiO2(110). Two reactions follow very different kinetics, whereas both coordination and charge state of an O2 chemisorbed at specific active site largely determine a particular reaction pathway. The inset of Figure 1a shows an STM image of a clean reduced TiO2(110) surface (VO coverage, Θ(VO), ∼0.09 monolayer, ML) recorded at 50 K. The periodic bright and dark rows in the empty-state image correspond to Ti5c and Ob atoms, respectively, whereas the bright features on the dark Ob rows are VO’s (marked by squares).30 A typical image of the TiO2(110) surface after O2 exposure (Θ(O2) ∼0.08 ML) is presented in Figure 1a. It has been shown recently that the major O2 chemisorption channel at the VO’s manifests itself in the STM image by the disappearance of the VO features,15−17

(1)

It is recognized that O2 chemisorption on reduced TiO2(110) is enabled by an electron transfer from pre-existing defects such as bridging O vacancies (VO) and Ti interstitials, resulting in a formation of nominally O2− species,7,20−23 although the exact extent of electron transfer is not well-understood and other species such as O22− have been also suggested.18,20,24,25 In the prevailing model, solely the hole-induced O 2 desorption is considered.3,18,19 However, an electron-mediated O2 dissociation at the VO site has been recently invoked by © 2011 American Chemical Society

(2)

Received: October 20, 2011 Accepted: December 12, 2011 Published: December 12, 2011 102

dx.doi.org/10.1021/jz2014055 | J. Phys. Chem.Lett. 2012, 3, 102−106

The Journal of Physical Chemistry Letters

Letter

Figure 2. STM images of the same (1.8 × 2.2) nm2 area (a) before and (b) after O2 exposure (Θ(O2) ∼0.08 ML) and (c) after UV irradiation for 15 min, showing the O2 desorption from the Ti5c site. The ball models of the rectangular region, marked in panels a−c, illustrate the observed events. (d) DFT calculated bonding configuration, adsorption energy, and Bader charge for O 2 chemisorbed at Ti5c site.

Figure 1. STM images (19 × 15) nm2 of TiO2(110) surface (a) after O2 exposure (Θ(O2) ∼0.08 ML) and (b) following UV irradiation for 15 min (all carried out at 50 K). Inset shows a clean surface before O2 exposure (VO density ∼0.09 ML).

and thus only some of the unfilled VO’s (∼0.01 ML) can be seen in Figure 1a. In addition, a smaller number of O2 molecules chemisorb at the regular Ti5c sites16 and appear as a single protrusions on the Ti5c rows (marked by octagons). Following UV irradiation, a substantial number (∼0.05 ML) of isolated round spots appears on the Ti5c rows, as marked by circles in Figure 1b. They are attributed to Oa species, produced upon photoinduced dissociation of O2 chemisorbed at VO sites (reaction 2), whereas the rationalization of such assignment is provided below. Lack of observation of the Oa pairs reveals that the O2 dissociation does not occur at the Ti5c sites.16 In addition, the concentration of the empty VO’s is found to be unchanged after UV irradiation, indicating an absence of O2 desorption from VO sites. Evidently, this observation does not support a description of an O2 PSD as occurring at the VO sites, which has been invoked in the literature.26,27 However, our results indicate that O2 desorption does take place from different sites. In Figure 1b, the O2 desorption reveals itself by the disappearance of O2 molecules from Ti5c rows after the UV irradiation. Whereas the O2/Ti5c species have somewhat similar appearance with Oa species,16 seen in Figure 1b, the O2 desorption from Ti5c sites will be validated below in Figure 2. The STM images in Figure 2a−c display the same surface area before and after O2 exposure and following UV irradiation. The observed changes in the highlighted rectangle region are marked in the images, and their interpretation is provided by corresponding ball models. After O2 exposure, a round bright feature can be seen on Ti5c row in Figure 2b, which is recognized as O2 molecule chemisorbed at Ti5c site.16 After UV irradiation, the O2 molecule on Ti5c row disappears without an emergence of any new feature, evidently indicating the desorption of O2 species chemisorbed at Ti5c site

the recently reported O2 PSD at similar Θ(O2) ∼0.08 ML26 has apparently originated from these sites. Note also that the detected O2 desorption from Ti5c sites is consistent with a holemediated model, considering a modified reaction 1, O2−/Ti5c + h+ → Ti5c + O2(g).31 Figure 3a−c shows the photochemistry of the O2/VO species, displaying the evolution of the same area after O2 adsorption

Figure 3. STM images of the same (2.6 × 2.6) nm2 area (a) before and (b) after O2 exposure and (c) after UV irradiation for 15 min, showing the O2 dissociation at the VO site. The ball models illustrate the observed events in the rectangular region, marked in panels a−c. DFT calculated bonding configurations, adsorption energies, and Bader charges for (d) O2/VO, and (e) Oa/Ti5c species.

and UV irradiation. Two VO’s in the chosen highlighted region in Figure 3a become invisible in Figure 3b due to the molecular adsorption of O2 at VO sites.16,17 Following UV irradiation, a bright round feature appears on the Ti5c site adjacent to the original VO position, as shown in Figure 3c. It is recognized as Oa,22,32 which appeared as a result of O2 dissociation at the VO site (reaction 2). Whereas the possible existence of O2 dissociation has been recently suggested by Petrik and Kimmel,26 here we present the first spatially resolved evidence of this process. Figure 3d,e shows the calculated bonding

hν

O2 /Ti5c → Ti5c + O2 (g)↑

(3)

Figure 2d shows the most stable bonding configuration calculated by DFT for O2/Ti5c species, and its adsorption energy of −1.99 eV and the Bader charge of 0.84 e− are comparable to 1.6−2.0 eV and 0.9 e− reported in the literature.22,24 However, whereas O2 desorption from the Ti5c sites (not adjacent to VO’s) has not been previously considered, 103

dx.doi.org/10.1021/jz2014055 | J. Phys. Chem.Lett. 2012, 3, 102−106

The Journal of Physical Chemistry Letters

Letter

dissociated and 50% nondissociated in both cases. However, the current observations that O2 desorption occurs only from Ti5c sites and O2 dissociation can be fully completed for Θ(O2) < Θcrit indicates that these processes are even more complex. The photoinduced processes of electron-mediated dissociation and hole-mediated desorption of O2 can be considered as complementary oxidative and reductive surface reactions, respectively. For photochemistry to be effective, both reactions should occur at substantial and balanced rates.4 However, because of a considerable concentration difference between the chemisorbed O2/VO and O2/Ti5c species, a larger total number of electrons is consumed in the dissociation compared with the number of holes consumed in the desorption process, and the balance is eventually breached. Consequently, an increasing number of “excess” holes are accumulated, likely trapped at the undercoordinated surface oxygen sites,4,35 as dissociation continues. Because electrons upon reaching the surface effectively recombine with trapped holes, the electron-mediated O2 dissociation becomes hindered when a concentration of excess holes exceeds some critical level. The minuscule increase in dissociated O2 for Θ(O2) > Θcrit in Figure 4 is likely caused by the continuing O2 desorption, which consumes a small but increasing amount of holes, and it is consistent with similar slopes of the dissociation and desorption plots for higher Θ(O2). Figure 5 displays the concentrations of O2/Ti5c (red circles) and O2/VO species (blue circles for Θ < Θcrit and white

configurations and adsorption energies for O 2 species chemisorbed and dissociated at VO. In agreement with literature,24,25,33 the O2 adsorption at VO is exothermic by −2.63 eV, whereas the dissociation is exothermic by −5.79 eV. One can also notice in Figure 3a−c that the bottom O2/VO species appears intact after UV irradiation. Furthermore, analysis of the larger areas confirms that the detected number of Oa’s is less (by ∼35%) than the number of O2-filled VO’s. Apparently, this indicates that some O2/VO species remain nondissociated. Comparison of the photoinduced dissociation and desorption of O2 clearly shows that these two reactions originate from the distinctive, differently coordinated adsorption sites of VO and Ti5c, respectively. Note that such relation between photoreactivity and an adsorption site on TiO2 (for O2 in this case) has not been demonstrated until now. It should also be noted that O2 molecules chemisorbed at VO and Ti5c sites are in different charge states, with the calculated Bader charges of 0.99 and 0.84 e−, respectively.34 Hence, it is conceivable that both specific coordination and charge state of chemisorbed O2 largely determine particular reaction pathway. The amount of the reacted (and nonreacted) O2/VO and O2/Ti5c species as a function of the initial O2 coverage is shown in Figure 4. Initially, the amount of both desorbed (red circles)

Figure 4. Concentration of photodesorbed O2/Ti5c, photodissociated O2/VO, and nondissociated O2/VO species after 15 min UV irradiation as a function of the initial O2 coverage. The dashed lines are linear regressions.

Figure 5. Concentration of O2 species chemisorbed at Ti5c and VO sites as a function of the UV irradiation time, normalized to the initial O2 concentrations of 0.01, 0.07 (Θ > Θcrit) and 0.04 (Θ < Θcrit) ML, respectively. Insert shows the initial regions of the plots in a semilog scale (the lines are linear regressions).

and dissociated (blue circles) O2 species increases linearly, and there are no nonreacted species (white circles). Therefore, a higher slope of the dissociation plot basically reflects a larger number of O2/VO vs O2/Ti5c species, which also can be seen in Figure 1a. However, the slope for the O2 dissociation sharply decreases above some critical coverage, Θcrit, of ∼0.05 ML, apparently indicating that only a certain amount of O2 (∼0.04 ML) can be dissociated and the process is nearly quenched thereafter. This is also reflected in the simultaneous appearance and a linear increase in the amount of nondissociated O2/VO species in Figure 4. The existence of such “photoblind” O2 species has been also reported by Kimmel and Petrik.26 It should be noted that obtained results agree even quantitatively with reported PSD results for similar Θ(O2) of ∼1 O2/VO and long irradiation times.26 For instance, the detected amount of desorbed O2 (∼10% of initial Θ(O2)) is close to ∼14% previously reported, with ∼50% of the remaining O2 photo-

triangles for Θ > Θcrit coverages) as a function of the UV irradiation time. Whereas the number of O2/Ti5c species promptly decays to zero, the O2/VO concentration decreases considerably slower, reaching zero only for Θ(O2) < Θcrit. In addition, the inset in Figure 5 shows that neither plot can be described by a simple exponential dependence. However, the difference in the initial slopes indicates that the initial rate constant is about 16 times larger for O2 desorption in comparison with O2 dissociation. Despite the fact that it is generally known that holes diffuse slower than electrons,3,4 the considerable disparity in the reaction rates may perhaps be attributed to a substantial difference in the dynamics of electron and hole transfer to chemisorbed O2/VO and O2/Ti5c species, 104

dx.doi.org/10.1021/jz2014055 | J. Phys. Chem.Lett. 2012, 3, 102−106

The Journal of Physical Chemistry Letters

Letter

0.8 V, which are sufficiently low to avoid the tip-induced O2 dissociation.16 Computational Methods. For the DFT simulations, we employed a similar approach to our previous work.23,38 The calculations were performed with the CP2K program,39,40 which incorporates the Gaussian and plane waves (GPW) method.41 Core electrons were described by Goedecker− Teter−Hutter pseudopotentials,42 whereas the valence electrons were expanded as a double-ζ Gaussian basis set. We utilized a (6 × 2) periodic slab that was four O−Ti−O trilayers thick (total of 192 atoms) and had two VO’s, giving a Θ(VO) of 0.17. All trilayers were relaxed except of the bottom one. Atomic charges were calculated using the Bader method.43 The calculations were performed with the PBE exchange correlation functional as spin-polarized.44 The Γ-point was used to sample reciprocal space. In the calculations, one O2 molecule was adsorbed in various states (adsorbed at Ti5c, filling VO, or dissociated - healing VO and producing Oa). In all cases, the O2 molecule in the gas phase was taken as the reference so that energy changes were calculated according to: ΔE = Esurf+O2(a) − Ebare surf − EO2(g). A more negative ΔE value thus indicates more stable adsorption.

respectively. (The latter apparently has a much higher rate.) As a side comment, note that the initial rates of O2 dissociation are practically the same for Θ(O2) < Θcrit and Θ(O2) > Θcrit, Figure 5 inset, indicating that dissociation, starts similarly but then asymptotically saturates to a nonzero level for Θ(O2) > Θcrit. Considering the nondissociated O2/VO species, which appeared as a result of the reaction quenching for Θ(O2) > Θcrit, it seems they are different from the original, as-adsorbed O2/VO’s. UV irradiation converts them to the species that either does not react with electrons, or these reactions do not result in dissociation.36 We speculate that they are species with a reduced negative charge, created upon the interaction with discussed above “excess” holes: O22−/VO + h+ → O2−/VO. Interestingly, the difference between the nondissociated, presumably O2−/VO and as-adsorbed O22−/VO species is also reflected in the observation that the former is considerably more stable when deliberately scanned at the elevated tunneling parameters (not shown), which otherwise would easily dissociate the latter (e.g., at 1.8 V and 150 pA).15−17 However, a comprehensive evaluation of the O2/VO species that are not susceptible to photodissociation requires more research. In conclusion, photocatalytic reactions of single O2 molecules on reduced TiO2(110) surfaces were imaged for the first time with STM. We demonstrate that additional active sites should be taken into account in order to properly describe two complementary redox channels (O2 desorption and dissociation). DFT calculations suggest that both coordination and charge state of chemisorbed O2 at the specific site largely determine a particular reaction pathway. Two O2 reactions follow very different kinetics, whereas the desorption is much faster than the dissociation, and later is also hindered above some critical O2 coverage. The results, obtained here for the O2 coverages up to ∼ VO concentration, should be considered as a first step in the examination of O2 photochemistry on TiO2(110) with STM. However, the conclusions obtained here likely can be projected to higher O2 coverages. It is reasonable to expect that comparable, site-specific photoreactive channels also operate there and, possibly, through the mechanisms containing similar aspects, which will be corroborated in forthcoming studies. In general, our results facilitate a molecular level insight into the photochemistry of O2 on TiO2. In particular, they should have a considerable effect on the interpretation of ensemble-averaged experimental data, where the site-specific chemistry cannot be easily attained.

■

AUTHOR INFORMATION

Corresponding Author *E-mail: [email protected].

■

ACKNOWLEDGMENTS We thank M. A. Henderson, Z. Dohnalek, G. A. Kimmel, and N. G. Petrik for stimulating discussions. This work was supported by the U.S. Department of Energy (DOE) Office of Basic Energy Sciences, Division of Chemical Sciences, and performed at EMSL, a DOE User Facility sponsored by the Office of Biological and Environmental Research and located at Pacific Northwest National Laboratory.

■

REFERENCES

(1) Hoffmann, M. R.; Martin, S. T.; Choi, W.; Bahnemann, D. W. Environmental Applications of Semiconductor Photocatalysis. Chem. Rev. 1995, 95, 69−96. (2) Linsebigler, A. L.; Lu, G. Q.; Yates, J. T. Jr. Photocatalysis on TiO2 Surfaces: Principles, Mechanisms, and Selected Results. Chem. Rev. 1995, 95, 735−58. (3) Thompson, T. L.; Yates, J. T. Jr. Surface Science Studies of the Photoactivation of TiO2-New Photochemical Processes. Chem. Rev. 2006, 106, 4428−53. (4) Henderson, M. A. A Surface Science Perspective on Photocatalysis. Surf. Sci. Rep. 2011, 66, 185. (5) Dohnalek, Z.; Lyubinetsky, I.; Rousseau, R. Thermally-Driven Processes on Rutile TiO2(110)-(1 × 1): A Direct View at the Atomic Scale. Prog. Surf. Sci. 2010, 85, 161−205. (6) Pang, C. L.; Lindsay, R.; Thornton, G. Chemical Reactions on Rutile TiO2(110). Chem. Soc. Rev. 2008, 37, 2328−53. (7) Henderson, M. A.; Epling, W. S.; Perkins, C. L.; Peden, C. H. F. Interaction of Molecular Oxygen with the Vacuum-Annealed TiO2(110) Surface: Molecular and Dissociative Channels. J. Phys. Chem. B 1999, 103, 5328−37. (8) Dohnálek, Z.; Kim, J.; Bondarchuk, O.; White, J. M.; Kay, B. D. Physisorption of N2, O2, and Co on Fully Oxidized TiO2(110). J. Phys. Chem. B 2006, 110, 6229−35. (9) Epling, W. S.; Peden, C. H. F.; Henderson, M. A.; Diebold, U. Evidence for Oxygen Adatoms on TiO2(110) Resulting from O2 Dissociation at Vacancy Sites. Surf. Sci. 1998, 412/413, 333−43.

■

EXPERIMENTAL SECTION The STM experiments were conducted in an ultrahigh vacuum system (base pressure ∼8 × 10−12 Torr at 50 K). The partially reduced TiO2(110) surface was prepared by multiple cycles of Ar+ ion sputtering and annealing at 800−900 K. O2 was introduced through a directional doser coupled to the STM stage. The detailed experimental setup was previously described.23,37 The UV light source was a Hg lamp coupled to the STM stage via fiber optic cable. The infrared portion of the spectrum was blocked by a water filter, with typical fluxes of ∼2 × 1015 photons/cm2s (for energies >3 eV), which did not caused a noteworthy sample temperature increase (