pubs.acs.org/JPCL
Direct Observation of Site-Specific Molecular Chemisorption of O2 on TiO2(110)
Zhi-Tao Wang,† Yingge Du,† Zdenek Dohn alek,‡ and Igor Lyubinetsky*,† †
Environmental Molecular Sciences Laboratory, Institute for Interfacial Catalysis, and Pacific Northwest National Laboratory, Richland, Washington 99352, United States, and ‡Fundamental and Computational Sciences Directorate, Institute for Interfacial Catalysis, and Pacific Northwest National Laboratory, Richland, Washington 99352, United States
ABSTRACT Molecularly chemisorbed O2 species were directly imaged on reduced TiO2(110) at 50 K with high-resolution scanning tunneling microscopy (STM). Two different O2 adsorption channels, one at bridging oxygen vacancies (VO) and another at 5-fold coordinated terminal titanium atoms (Ti5c), have been identified. While O2 species at the Ti5c site appears as a single protrusion centered on the Ti5c row, the O2 at VO manifests itself by a disappearance of the VO feature. It is found that the STM tip can easily dissociate O2 species, unless extremely low magnitude of the tunneling parameters are used. The O2 molecules chemisorbed at low temperatures at these two distinct sites are the most likely precursors for the two O2 dissociation channels, observed at temperatures above 150 and 230 K at the VO and Ti5c sites, respectively. SECTION Surfaces, Interfaces, Catalysis
(DFT) results have indicated that chemisorbed anionic O2 species bind rather strongly at VO's with adsorption energy on the order of 2 eV.18-20 Photostimulated desorption studies have demonstrated the existence of two types of chemisorbed O2 species,21-23 while the maximum amount of chemisorbed O2 has been experimentally determined to be about twice the VO coverage at temperatures below ∼100 K.10 However, detailed information regarding the O2 chemisorption sites is limited. The chemistry of O2 on TiO2(110) at elevated temperatures is not well understood either. Theoretical studies have shown that the activation energy for O2 dissociation at VO's is ∼1 eV,20,24 which is smaller than the corresponding O2 chemisorption energy (∼ 2 eV).18-20 Therefore for O2 at a VO, dissociation rather than desorption is energetically favorable at elevated temperatures.19,20,25,26 Scanning tunneling microscopy (STM) investigations have directly demonstrated that each O2 dissociation event causes the healing of a single VO and deposition of an oxygen adatom (Oa) at a nearby Ti5c site.26-28 In addition, STM studies have recently revealed a minor (though potentially important) second O2 dissociation channel occurring on the 5-fold coordinated terminal titanium sites (Ti5c), which leads to the formation of secondnearest neighbor Oa pairs at 300 K.14 It has been suggested previously that extra charges provided by Ti interstitials mediate this process.14 Our recent results have revealed that the delocalized unpaired electrons associated with the VO's could be utilized in this channel as well, whereas the
T
he chemistry of oxygen on TiO2 surfaces is an important component in many catalytic and photocatalytic processes, such as water splitting and waste remediation,1-3 and has been extensively studied.4-6 In particular, O2 is considered as one of the oxidizing reagents in photooxidation processes and also as an electron scavenger facilitating these reactions, although the exact role of O2 is still unclear.1,2,7 In addition, surface-bound oxygen species may considerably affect TiO2 catalytic propoerties.2,7 The majority of fundamental research on the O2 interaction with TiO2 has been carried out on the rutile (110)-(11), which has in fact became the model transition-metal oxide surface, suitable for the studies under ultrahigh vacuum (UHV) conditions. Conventional sample preparation in the UHV (ion sputtering and annealing) leads to its partial reduction as manifested by the creation of point defects, such as bridging oxygen vacancies (VO's) and Ti interstitials.8 Since reoxidation of metal oxide surfaces has considerable consequences for their catalytic activity, atomic-level studies of the oxygen adsorption, as a first step in healing the VO defects and reoxidation of the reduced TiO2, are quite important but rather scarce so far. The investigation of molecular adsorption of O2 can be considered a natural first step providing information about possible O2 surface chemistry on TiO2(110). At sufficiently low temperatures, O2 only physisorbs on stoichiometric TiO2(110) (at T < 60 K),9,10 but on reduced surfaces, it also molecularly chemisorbs at VO sites (at T < 150 K).10-12 It is well understood that charge transfer to adsorbed O2 (e.g., from VO's and Ti interstitials) is required for its chemisorption.11,13-17 Upon O2 adsorption at 120 K, Henderson et al. observed a new Ti-O2- feature in the electron energy loss spectra that replaced the defect Ti3þ state.11 Density functional theory
r 2010 American Chemical Society
Received Date: November 11, 2010 Accepted Date: December 3, 2010 Published on Web Date: December 07, 2010
3524
DOI: 10.1021/jz101535f |J. Phys. Chem. Lett. 2010, 1, 3524–3529
pubs.acs.org/JPCL
Figure 2. STM images of the same (3.0 3.2) nm2 area (a) before and (b) after O2 exposure, taken at V = 0.8 V, I = 3 pA, showing the O2 chemisorption at VO sites, and then (c) imaged at V = 1.5 V, I = 3 pA. Corresponding ball-models of the rectangular region marked in a-c illustrate the observed events.
Figure 1. STM image (17 14) nm2 of a TiO2(110) surface at 50 K after O2 exposure of 5 1013 cm-2, taken at V = 0.3 V, I = 1 pA. Inset shows a clean surface before O2 exposure (VO density ∼0.08 ML).
total surface charge is a limiting factor.16 Furthermore, recent STM studies have indicated that the O2 dissociation channel at the Ti5c turns on at temperatures higher than 150-180 K,29 while the O2 dissociation at VO sites has been observed at temperatures as low as ∼80 K30 and 120 K.14,29 However, the latter observations seems to be in contrast with the majority of the ensemble-averaging technique studies that have detected the O2 dissociation at VO sites only at temperatures above 150-200 K.10-12,21 In this study, we investigate some of the unexplored aspects of O2 chemistry on TiO2(110) with high-resolution STM at low temperatures, focusing on the initial stages of O2 adsorption at 50 K. We successfully imaged molecularly chemisorbed O2 species residing at two distinctive adsorption sites (VO and Ti5c), which can be observed intact only at very low magnitude of the tunneling parameters to avoid the STM tip-induced O2 dissociation. At higher temperatures, the O2 species at both VO's and Ti5c's dissociate spontaneously above 150 and 230 K, respectively. An STM image of a clean reduced TiO2(110) surface taken at 50 K is shown in the inset of Figure 1. The bright rows in the empty-state STM image correspond to the terminal Ti5c's, while the dark rows are the bridging oxygen atoms (Ob).8 Beyond the periodic surface structure, bright features of VO's can be recognized on the dark Ob rows (marked by squares).8 A few larger and brighter spots on Ti5c rows (marked by triangles) are due to minor background adsorption (likely, CO and H2O). In the current study, we have focused on the initial stages of O2 adsorption (O2 coverage less than VO density). The effect of the O2 exposure (5 1013 cm-2) on TiO2(110) at 50 K is illustrated in Figure 1. Comparison with a clean TiO2(110) surface reveals that a number of VO's (initially ∼0.08 monolayer, ML) have disappeared (∼70% in this particular case), while the rest of the VO's are left intact. In addition, isolated bright round spots centered on Ti5c rows (marked by hexagons) can also be seen in the STM image (at the level of ∼0.01 ML). The disappearance of VO's is attributed to the chemisorption of O2 molecules inside of them, while the new bright features are assigned to a molecularly chemisorbed O2 on the Ti5c sites. The comprehensive explanation
r 2010 American Chemical Society
of our assignments is provided below along with the discussion of Figures 2 and 3. For a detailed investigation of the O2 adsorption on TiO2(110), we have followed a small surface region. Figure 2a,b shows STM images of the same surface area before and after O2 exposure. The observed changes are marked in the images, and the interpretation of corresponding surface processes is provided in ball models of the highlighted regions (dotted rectangles) below the images. From the comparison of the images, it is clear that several VO's have virtually disappeared as a result of O2 exposure. Since no additional features are detected, this is attributed to the molecular adsorption of O2 at a VO site O2 ðgÞ þ VO f O2 ðaÞ=VO
ð1Þ
While images in Figure 2a,b were taken at a bias voltage, V, of 0.8 V and tunneling current, I, of 3 pA, scanning the same area at a higher V of 1.5 V results in the appearance of small round bright features situated on the terminal Ti5c sites adjacent to the original VO positions (while VO features are still absent), as shown in Figure 2c. These new species are recognized as Oa's, whose appearance in STM images is well documented in the literature.26-28 Evidently, Oa is created as a result of the dissociation of adsorbed O2 by the STM tip at a higher bias, whereas the original VO is healed by a second O atom, similar to O2 dissociative adsorption at elevated temperatures discussed earlier26-28 tip
O2 ðaÞ=VO sf Ob þ Oa
ð2Þ
The products of the tip-induced process provide conclusive evidence of the preadsorption of a single O2 molecule at a VO according to reaction 1. While the existence of molecular chemisorption of O2 at a VO at low temperature has been postulated in a number of previous studies (as a corresponding precursor state of the VO-mediated channel of O2 dissociation),10-12,21 here we present a direct observation of the process. Interestingly, the virtual disappearance of the VO from the STM image is apparently caused by a particular appearance (at low bias voltage) of O2 adsorbed at a VO, which closely resembles that of
3525
DOI: 10.1021/jz101535f |J. Phys. Chem. Lett. 2010, 1, 3524–3529
pubs.acs.org/JPCL
illustrated in Figure 3c. One can see that scanning of the same area with slightly elevated tunneling parameters (V = 0.6 V, I = 3 pA) converts a single species into two Oa species occupying two neighboring Ti5c sites tip
O2 ðaÞ=Ti5c sf Oa þ Oa
Upon O2 dissociation, one Oa stays at the original Ti5c site previously occupied by adsorbed O2, while the other Oa is deposited onto the nearest-neighbor Ti5c. We denote an Oa pair separated by one lattice constant as Oa-Oa(1). Note that a nearest-neighbor Oa-Oa(1) separation at 50 K is different than that of a second-nearest neighbor configuration, Oa-Oa(2), reported previously at 300 K.14,16 In our recent report, the large Oa-Oa(2) separation at 300 K was partially attributed to the Coulombic repulsion between the two negatively charged Oa's.16 However, the combined action of Coulombic repulsion and thermal activation at 50 K is apparently not enough to overcome the activation barrier for Oa diffusion along the Ti5c trough (∼1.1-1.3 eV).16,26,27 Also note that the Oa-Oa(1) pairs are not well resolved due to their close proximity at 50 K, and instead of two single protrusions, the feature is rather dumbbell-like in Figure 3c. A similar feature (likely, also a result of the tip-induced dissociation) has been reported for O2 adsorption on TiO2(110) at 80 K.32 Note that the majority of the STM images throughout our study were taken at rather low values of both bias voltage (0.3-0.8 V) and tunneling current (1-3 pA). Such scanning conditions are crucial in order to minimize STM tip-induced O2 dissociation, as will be discussed below. For rutile TiO2, a wide bandgap semiconductor, generally used tunneling parameters are considerably larger (V ∼ 1.3-1.7 V and I ∼ 50-100 pA).8 However, STM images of TiO2(110) can be acquired even at voltages as low as 0.3 V, probably, through tunneling into the defect states lying within the band gap close to the Fermi level. Because of the poor conductivity of TiO2, scanning at such low voltages would set the tip very close to the surface, resulting in tunneling instabilities due to strong tip-surface interactions. To mitigate this effect, we employed very low tunneling current on the order of 1 pA, which moves the tip away from the surface, and in addition, minimizes tunneling current-induced processes. Figure 4 shows the probability of O2 dissociation by the STM tip at both Ti5c and VO adsorption sites, as a function of sample bias voltage in the constant tunneling current mode. Each data point depicts the statistical average value acquired upon scanning three different (30 30) nm2 areas with typically between 200 and 800 dissociation events detected. Since we have found that O2 dissociation also strongly depends on the particular STM tip condition, all data were acquired using the same tip in the course of a single experimental run. However, wherever possible, the data scattering related to different tip conditions has been also included in the error bars. For both O2 adsorption sites, the obtained plots have similar shapes, with the dissociation probability increasing relatively sharply with increasing V and saturating rapidly at 100% dissociation probability. The dissociation probability also depends on tunneling current, with dependencies shifting toward lower V's with I increasing. Remarkably, Figure 4
Figure 3. STM images of the same (2.3 2.6) nm2 area (a) before and (b) after O2 exposure, taken at V = 0.3 V, I = 1 pA, showing the O2 chemisorption at Ti5c sites, and then (c) imaged at V = 0.6 V, I = 3 pA. Corresponding ball-models of the rectangular region marked in a-c illustrate the observed events.
regular Ob atom. It is well-known that STM images reflect the local density of states and are generally sensitive to the amount of local charge. In turn, as discussed earlier, a charge transfer toward O2 adsorbed at a VO is essential in the chemisorption of a charged (O2- or O22-) species.11,13-17 Hence, a low contrast observed in STM images is likely caused by a close similarity of the empty-state densities near the Fermi level for the negatively charged O2 (whereas donated charge at least partially fills π* states) and Ob2- species (taking into account the formal charge of the latter).19,24,31 This also agrees with DFT studies that have indicated that the reduced TiO2(110) surface is reoxidized, although not completely, by an O2 chemisorption at a VO site.24,25 It should be noted that the appearance of the feature of the chemisorbed O2 at a VO depends on a particular STM tip condition. Occasionally, we observed the dim O2 features, which were less bright than the VO features (shown in Figure S1 of the Supporting Information). The molecular adsorption of O2 on Ti5c rows is illustrated in Figure 3a,b, which displays the identical surface area before and after O2 exposure imaged at V = 0.3 V and I = 1 pA. Comparing these images, it is clear that the emergence of each single bright feature of O2 is not related to the disappearance of any VO but occurs at the regular terminal Ti5c site O2 ðgÞ þ Ti5c f O2 ðaÞ=Ti5c
ð3Þ
Although the molecular chemisorption of O2 at Ti5c was expected as a precursor to a second, non-VO-mediated channel for O2 dissociation,14,16 this is the first direct observation of such species. Repeated scanning of the same region has revealed that O2 molecules chemisorbed at Ti5c's are quite stable and immobile at 50 K. This observation is consistent with DFT calculation results that have indicated that O2 diffusion along Ti5c troughs is hindered by an activation barrier of ∼0.6 eV.24 (As discussed earlier, physisorbed O2's may also be present on Ti5c rows at 50 K,9,10 but they are likely too mobile to be imaged with STM). The key additional evidence of the chemical makeup of the observed species is once more provided by deliberate STM tip-induced dissociation, as
r 2010 American Chemical Society
ð4Þ
3526
DOI: 10.1021/jz101535f |J. Phys. Chem. Lett. 2010, 1, 3524–3529
pubs.acs.org/JPCL
Figure 4. Probability of O2 dissociation by STM tip at both Ti5c (red symbols) and VO sites (blue symbols), as a function of sample bias voltage in the constant tunneling current mode. (Dependencies for O2 at Ti5c's could not be measured at I > 1 pA due to the large tip instabilities at V < 0.3 V). The lines are shown as a guide for the eye only.
demonstrates that the tunneling parameters required to prevent a tip-induced dissociation have quite small values. In fact, even for the lowest I that we can employ (∼ 1 pA), the dissociation V thresholds for Ti5c and VO adsorption sites are just ∼0.3 and 1.2 V, respectively. Note also that generally, the tip-induced O2 dissociation at Ti5c sites occurs at smaller values of V and I than ones found for VO sites. While a detailed study of the mechanism of STM tip-induced O2 dissociation on TiO2(110) was beyond the scope of the work, we speculate that it may be to some extent caused by an inelastic electron tunneling process. Similarly, the tip-induced dissociation of O2 on Pt(111) at 40-150 K has been attributed to the intramolecular vibrational excitations via resonant inelastic electron tunneling mechanism.33 In addition, a strong electronic field in the tunneling gap may also play a substantial role.34-36 In particular, this process may be responsible for a higher V dissociation threshold observed for a VO site, taking into the account that the O2 species “inside” a VO is better screened from an external electric field than O2 at a Ti5c site with a relatively open configuration. Nonetheless, further investigations are required to determine the mechanism of the tip-induced O2 dissociation on TiO2(110). We have also explored the thermally induced dissociation of chemisorbed O2 species. In a series of experiments, we analyzed samples that were either preexposed to O2 at low temperature and subsequently annealed at a certain temperature, and/or directly dosed at elevated temperature. In both cases, the obtained results are consistent with each other. Figure 5 presents STM images that illustrate temperature-dependent results for 130, 180, and 260 K (all images were acquired at low values of tunneling parameters to avoid tip-induced effects). Figure 5a shows a STM image of the TiO2(110) surface exposed to an O2 dose of 3 1014 cm-2 at 130 K. Evidently, this surface looks similar to the one exposed to O2 at 50 K, with a number of chemisorbed O2 at both VO's (determined by disappeared VO's) and Ti5c sites (bright spots on Ti5c rows), as seen from the comparison of Figures 5a and 1. We have not observed O2 dissociation at VO sites at 130 K, although we did observe it at higher tempera-
r 2010 American Chemical Society
Figure 5. STM images after O2 exposure of 3 1014 cm-2 (a) at 130 K, followed by annealing at (b) 180 K and (c) 260 K for 10 min, illustrating the onset of thermally induced O2 dissociation at VO sites (>150 K) and at Ti5c sites (>230 K). All images were acquired at 50 K, and at V = 0.3 V, I = 1 pA.
tures (>150 K). In particular, Figure 5b shows the TiO2(110) surface exposed at 130 K with the same O2 dose (as in Figure 5a), and then annealed at 180 K for 10 min. As a result of annealing, additional bright spots appear on Ti5c rows (marked by circles). The majority of the features can be identified as Oa's, indicating that the thermally induced O2 dissociation at VO sites has turned on. The O2 chemisorbed on Ti5c rows have been found to stay intact at 180 K, indicating that O2 dissociative and molecular states coexist at this temperature. While O2 species at Ti5c's are hardly distinguishable from Oa's in Figure 2b, their existence have been verified by producing the Oa-Oa pairs via tip-induced O2 dissociation according to reaction 4 discussed above (data not shown). We would like to emphasize that our results support the long-standing view based on the results of the ensemble-averaged methods that demonstrated the temperature threshold for O2 dissociation at VO's to be above 150 K.10-12 Finally, we have found that thermally induced O2 dissociation at Ti5c sites occurs at temperatures higher than 230 K. This conclusion is supported by Figure 5c, which displays the TiO2(110) surface pre-exposed to O2 at 130 K, and then annealed at 260 K for 10 min. One can see that in addition to single Oa's, there are a number of paired Oa species at Ti5c rows (marked by ovals), indicating thermally induced O2 dissociation at Ti5c sites. This is also confirmed by the fact that the number of Oa-Oa pairs detected upon annealing at
3527
DOI: 10.1021/jz101535f |J. Phys. Chem. Lett. 2010, 1, 3524–3529
pubs.acs.org/JPCL
and assist in understanding of the surface reactivity of transition-metal oxides.
260 K correlates with the number of O2 chemisorbed at Ti5c rows observed at 130 K (e.g., comparing panels c and a of Figure 5). Furthermore, in addition to Oa-Oa(1) pairs (marked by the solid oval in Figure 5c), we have also observed Oa-Oa(2) (dotted ovals) and even Oa-Oa(1.5) (dashed ovals) pairs (corresponding STM line profiles are shown in Figure S2 of the Supporting Information). Statistically, annealing at 260 K results in the formation of a majority of Oa-Oa(1.5) pairs (∼70%), while Oa-Oa(2) and Oa-Oa(1) pair configurations account for ∼25 and 5%, respectively. It should be noted that such a Oa-Oa(1.5) configuration, where one Oa occupies a bridging position between two Ti5c sites, has been predicted by DFT calculations to be relatively stable.24 Finally, we point out that the observed temperature threshold for O2 dissociation at Ti5c's around 230 K correlates well with a DFT calculated activation energy barrier of ∼1 eV for O2 dissociation at Ti5c sites, reported in our previous work.16 As a final remark, note that after submission of the current manuscript, the observation of molecular O2 adsorption at low temperatures and dissociation by STM tip has been also reported elsewhere,37 albeit only for one adsorption channel (at VO sites). However, the imaging of O2 at VO's (extremely dim features) and easiness of its dissociation by a tip agree well with results presented here. In summary, we have performed a high-resolution STM investigation of the initial stages of oxygen adsorption on a reduced TiO2(110) surface at 50 K. Molecularly chemisorbed O2 species, not directly observed until now on TiO2(110), have been imaged using “extremely mild” tunneling conditions. By tracking the same surface area before and after oxygen exposure, two different O2 chemisorption pathways have been identified. The first one is mediated by VO's, and the second occurs at Ti5c sites. Empty-state STM images allow for direct observation of the O2 species on Ti5c rows, while the O2 features chemisorbed at VO sites reveal themselves only by VO disappearance. The dissociation of chemisorbed O2 can be easily induced at STM tunneling conditions that are normally used for TiO2(110) imaging. The details of the tipinduced dissociation have been found to strongly depend on the type of the O2 adsorption site (VO vs Ti5c), and occurs more readily at Ti5c sites. We have also deliberately used tipinduced dissociation (leading to Oa production) to verify the chemical makeup of the adsorbed species. In addition, accounting for a possible tip-induced O2 dissociation may clarify the reported differences between previous STM and the ensemble-averaging technique results. We emphasize that use of very low values of STM tunneling voltage and current could be a general requirement for the studies of molecular adsorption not only on TiO2(110) but also on other wide-bandgap metal oxides, to avoid the tip-induced effects that can severely distort STM results. We conclude that the negatively charged O2 species chemisorbed at both VO and Ti5c sites are the most likely precursors for the previously established O2 dissociation channels on TiO2(110). Upon temperature increase, O2 dissociation occurs spontaneously above 150 and 230 K at VO and Ti5c sites, respectively. In general, the results reported here provide a molecular level insight into the thermal chemistry of O2 on reduced TiO2,
r 2010 American Chemical Society
EXPERIMENTAL METHODS The experiments were conducted in a UHV system (base pressure
