Surface Review and Letters, Vol. 7, Nos. 5 & 6 (2000) 613–617 c World Scientific Publishing Company
THE RELATIONSHIP BETWEEN BULK AND SURFACE PROPERTIES OF RUTILE TiO2 (110) ULRIKE DIEBOLD,∗ MIN LI, OLGA DULUB, ELEONORE L. D. HEBENSTREIT and WILHELM HEBENSTREIT Department of Physics, Tulane University, New Orleans, La 70118, USA Received 7 July 2000 We report scanning tunneling microscopy and complementary spectroscopic measurements on TiO2 (110) surfaces. We show data on (i) a surface restructuring process that results from annealing in oxygen; (ii) Pt clusters, grown at room temperature and encapsulated upon high temperature annealing; and (iii) adsorption of sulfur. In each case, heavily reduced, dark crystals show a very different behavior than more stoichiometric, light blue ones.
The bulk properties of a single-crystalline sample are usually of little interest to surface investigations provided the crystal is well ordered and clean enough not to cause any impurity segregation. We report recent results on a system where the state of the bulk has vast and rich influences on mesoscopic and nanoscopic surface properties. Titanium dioxide is a reducible metal oxide, i.e. it exists in a variety of oxidation states and crystal structures.1 It is used in many technical areas where high temperature reduction and oxidation processes play a role, for example as promoter in heterogeneous catalysis,2 in protective coatings, and as the active material in gas sensors.3 Heating bulk single crystals in vacuum or in a reducing atmosphere leads to visible color changes from transparent to light and, eventually, dark blue (see Fig. 1). While stoichiometric TiO2 is a wide gap semiconductor (Egap = 3 eV), reduced crystals exhibit enough n-type doping to allow detailed structural investigations with surface science techniques. In the following, we give three examples of how the reduction state of TiO2 crystals (which is closely related to the crystal color) affects surface phenomena. The experimental details are described elsewhere.4–8
(i) The first example illustrates the relationship between bulk defects and surface structure. A systematic investigation was performed using five small samples (2 × 2 × 2 mm3 ), which were cut from a polished TiO2 (110) single crystal and reduced in a furnace (Fig. 1). The crystal color is correlated with heating temperature rather than time. Before reduction, all cubes were transparent, similar to cube 2 (which was reoxidized in air from a color similar to cube 3). The resistivity of the crystals correlates with their color and ranges from 10 1 to 103 Ω cm at 300 K.9 Electron paramagnetic resonance (EPR) reveals that darker rutile crystals exhibit higher concentrations of extended Ti3+ -related bulk defects such as crystallographic shear planes (CSP’s), with a decrease in substitutional defects (related to O vacancies) and Ti interstitials as compared to lighter crystals. After sputtering and annealing in ultrahigh vacuum (UHV) to moderate temperatures, the surface exhibits a (1 ×1) structure (Fig. 2). Such a surface is essentially bulk-terminated with some relaxations.10 When the cubes in Fig. 1 were sputtered and UHVannealed for 20 min at 973 K, darker crystals were rougher than lighter ones.9 In addition, both dark cubes (3 and 4) exhibited scattered bright rows (typically 40 ˚ A long) along the [001] direction which
∗ Corresponding
Tel: 504-862-8279
author. E-mail:
[email protected]
613
Fax: 504-862-8702
614 U. Diebold et al.
Fig. 1. The bulk defects that are formed upon reduction of TiO2 single crystals generate color centers and cause a change in crystal color. Photograph of rutile single crystals heated in a furnace to various temperatures: cube 1 — 19 h at 1273 K; cube 2 — 21 h 40 min at 1450 K (was like cube 3), then reoxidized in air at 1450 K; cube 4 — 35 min at 1450 K; cube 5 — 1 h 10 min at 1350 K. ˚ × 300 ˚ Fig. 3. STM image (300 A A) after a flat, (1 × 1)-terminated TiO2 (110) surface was reannealed in O2 (1 × 10−6 mbar, 550 K, 20 min). Reoxidation of the reduced bulk causes formation of irregular networks of pseudohexagonal rosettes (“R”), added small (1 × 1) islands, and strands.5
Fig. 2. STM image (400 ˚ A × 400 ˚ A, top) and model (bottom) of the TiO2 (110)(1 × 1) surface. The surface of a light blue crystal was prepared by sputtering and annealing in ultrahigh vacuum.4 Bright and dark rows in the STM image correspond to positions of fivefoldcoordinated Ti and twofold-coordinated O atoms (residing ∼ 1 ˚ A above the surface), respectively.12 The surface is flat, with step edges running parallel to the h1¯ 11i and h001i directions.
were not observable on the light cubes (1 and 2). These rows are commonly named (1 × 2) strands. In a recent paper5 we showed that they consist of Ti2 O3 as originally suggested by Onishi et al.11 The heavily reduced sample (dark blue cube 3) exhibits a slightly higher work function (increase by 0.2 eV) than the
lighter samples, which might indicate more oxygen vacancies12 on its surface. Evidence for CSP’s extending to the surface was seen for samples even darker than the ones used in this study.13,14 Strong changes in surface structure were observed after annealing at moderate temperatures dark crystals in oxygen.6,15,16 The surfaces restructure and are quite rough. Their appearance is dominated by three small scale features: irregular networks of connected, (6×6.5 ˚ A)-wide rosettes (“R”); one-unit-cellwide [001]-oriented strands [with an appearance in scanning tunneling microscopy (STM) images similar to the (1 × 2) reconstruction]; and small (tens of ˚ A) (1 × 1) islands (see Fig. 3). Under the same preparation conditions, i.e. heating preannealed, flat samples in 1 × 10−6 mbar 18 O2 for 10 min at 573 K, dark crystals exhibit a restructured surface similar to the one shown in Fig. 3. Lighter crystals (cubes 1 and 2) show only a (1 × 1) structure.9 The formation and stability range of these structures depend on a variety of other preparation parameters, such as oxygen pressure, annealing temperature and annealing time.6 The oxygen-induced restructuring occurs through reaction of interstitial Ti atoms (diffusing from the
The Relationship Between Bulk and Surface Properties of Rutile TiO2 (110) 615
reduced bulk to the surface) with oxygen offered from the gas phase. This reoxidation mechanism results in the growth of additional TiO2 on the surface, as was confirmed by low energy ion scattering (LEIS) and static secondary ion mass spectrometry (SIMS) experiments using isotopically labeled 18 O2 .6,16 Depending on the kinetic parameters (the flux of Ti interstitials, which is related to crystal color, gas pressure, temperature, etc.), the growth can either occur in a layer-by-layer fashion, rendering stoichiometric (1 × 1)-terminated surfaces, or lead to incomplete structures with missing TiO2 units.15 In extreme cases it can result in a complete reoxidation of the whole crystal as seen in cube 2 in Fig. 1. The formation of the rosette structure is very interesting from a surface reactivity point of view. It consists of an incomplete TiO2 layer with atoms missing in a regular fashion; see the model in Fig. 3. First-principles calculations have shown that rosettes are stable structures.6 In contrast to (1 × 1)-terminated surfaces, they exhibit fourfoldcoordinated Ti atoms, which are expected to drastically influence the surface chemistry. (ii) The presence of bulk defects also affects the high temperature behavior of metal overlayers on TiO2 . When clusters of group VIII metals, supported on TiO2 , are annealed under reducing conditions, they may encapsulate with a thin “crust” of reduced titania. This so-called strong-metal–support interaction causes a drastic change in the activity and selectivity of the titania-promoted catalysts.17 There is disagreement in the literature on whether encapsulation is also achieved on a “flat model catalyst,” i.e. nanoscopic Pt clusters grown on a TiO2 (110) single crystal. LEIS measurements by the group of Madey have shown that the signal from the metal overlayer (Fe18 and Pt19 ) is suppressed after annealing under UHV conditions, and is regained by mild sputtering. X-ray photoemission spectroscopy (XPS) under grazing exit angle indicated a TiOx (x ≈ 1) layer. In contrast, Schierbaum et al., using nearly the same experimental conditions and methods, observed no evidence for encapsulation.20 Gao et al.21 compared the thermal stability of Pt films vapor-deposited on both a Nb-doped TiO2 (100) film [grown by oxygen plasma-assisted molecular beam epitaxy] and a reduced TiO2 (110) crystal. The TiO2 (100) films, which were essentially defectfree, showed no evidence for encapsulation, while Pt
Fig. 4. STM images of a Pt overlayer, evaporated at room temperature and heated to high temperature, causing encapsulation with a TiOx overlayer.7 The top image (500 ˚ A × 500 ˚ A) was filtered to show the structure of the encapsulation layer. The small scale image at the bottom was taken on top of an encapsulated cluster.
layers on the reduced single crystal encapsulated under similar experimental conditions. Figure 4 shows STM results from 25 ML Pt, grown at room temperature on a (blue) TiO2 (110) crystal, and then annealed in UHV as described in Ref. 7. Only O and Ti signals appear in LEIS. STM further confirms that the clusters are encapsulated, and are covered with a strikingly beautiful TiOx film, shown in Fig. 4(B). A model for the encapsulation layer is discussed in Ref. 22. This self-organized structure is very different from the Ti4 O7 layers observed for ultrathin titanium oxide films grown on a flat Pt(111) single crystal.23 When taking into account the deposited amount of Pt (25 ML), the surface coverage after encapsulation (∼ 40%), and the height of the clusters (∼ 40 ˚ A), and assuming no Pt left the surface during heating, a simple calculation shows that the clusters in Fig. 4(A) are “icebergs” reaching several tenths of angstroms deep into the substrate. Such a partial burying of supported clusters has been observed for a similar system, Pd/TiO2 (110),24 during annealing in an oxygen atmosphere. Oxygen adsorbed dissociatively on the Pd surface and spilled over onto the TiO2 substrate, where it stimulated regrowth of TiO2 around and over the clusters (similar to the
616 U. Diebold et al.
oxygen-induced restructuring discussed above). In the case of Pt, encapsulation takes place upon annealing in UHV, and annealing in high pressures of oxygen can even reverse the strong-metal support interaction (SMSI) effect.17 While bulk defects clearly play a role, the encapsulation process itself is poorly understood at this point. (iii) Lastly, we turn to crystal-color-dependent adsorption processes on TiO2 (110). The adsorption site of elemental sulfur shows a strong temperature dependence.8 Sulfur sits on the fivefold-coordinated Ti atoms (see Fig. 2) when adsorbed at room temperature. On stoichiometric samples, it desorbs almost completely upon heating (Fig. 5). On dark crystals, some S remains at the surface up to much higher temperatures (Fig. 5). The XPS S2p peak shifts to lower binding energies. STM shows that S is adsorbed at the position of bridging oxygen atoms (Fig. 2) after heating the S-covered sample or after adsorbing onto a hot (> 150◦ C) surface.25 On dark crystals, a (3 × 1) superstructure forms at 300◦ C.8 At 400◦C more complicated structures form which involve a third adsorption site at the position of in-plane oxygen atoms.25 We have proposed a model of the (3 × 1) superstructure8 where every third bridging oxygen atom is replaced by a S atom, and the rest of the bridging
oxygens are removed. The oxygen either desorbs into the gas phase or migrates into the bulk. Because of the low temperatures involved, the latter process appears more likely. Oxygen diffusion in TiO2 crystals occurs via a site exchange mechanism. The diffusing species are effectively O vacancies. Under UHV conditions the onset for diffusion of O vacancies as well as Ti interstitials occurs at 400 K.26 In more stoichiometric crystals the flux of O vacancies to the surface is small and the competitive desorption process takes over. Similar defect-mediated site exchange processes appear to happen for Cl adsorption as well.25 Summarizing, we have shown that a variety of phenomena on TiO2 (110) surfaces, namely the nanoand mesoscopic surface structure, the high temperature encapsulation of group VIII metals, and the surface sulfidation process, are affected by the reduction state of bulk single crystals. In each case, diffusion of bulk defects at elevated temperatures plays a role. Investigations of such processes may clarify the mechanisms underlying technical applications, for example the change of conductivity in high temperature oxgyen gas sensors based on TiO2 .3 While the TiO2 surface is the first system where the influence of the bulk has been studied with surface science techniques, it is conceivable that similar effects may occur in other reducible metal compound systems.
Acknowledgments This work was supported by NSF-CAREER. We thank M. A. Henderson of the Pacific Northwest National Laboratories and D. R. Jennison of the Sandia National Laboratories for useful discussions.
Reference
Fig. 5. Temperature-dependent XPS measurements of TiO2 (110) surfaces after adsorption of S (to saturation coverage) at room temperature. On stoichiometric, light crystals, S mainly desorbs. On reduced, dark crystals, some S remains on the surface and switches the adsorption site. Initially it binds to surface Ti atoms; at higher temperatures it replaces surface oxygen atoms.8 This change in bonding is accompanied by a strong shift in the S2p binding energy.
1. G. V. Samsonov, The Oxide Handbook (IFI/Plenum, New York, 1982). 2. C. N. Satterfield, Heterogeneous Catalysis in Industrial Practice, 2nd edition (McGraw-Hill, New York, 1991). 3. U. Kirner, K. D. Schierbaum, W. Go¨ pel, B. Leibold, N. Nicoloso, W. Weppner, D. Fischer and W. F. Chu, Sensors and Actuators B1, 103 (1990). 4. U. Diebold, J. Lehman, T. Mahmoud, M. Kuhn, G. Leonardelli, W. Hebenstreit, M. Schmid and P. Varga, Surf. Sci. 411, 137 (1998). 5. M. Li, W. Hebenstreit and U. Diebold, Phys. Rev. B61, 4926 (2000).
The Relationship Between Bulk and Surface Properties of Rutile TiO2 (110) 617 6. M. Li, W. Hebenstreit, L. Gross, U. Diebold, M. A. Henderson, D. R. Jennison, P. A. Schultz and M. P. Sears, Surf. Sci. 437, 173 (1999). 7. O. Dulub, W. Hebenstreit and U. Diebold, Phys. Rev. Lett. 84, 3646 (2000). 8. E. L. D. Hebenstreit, W. Hebenstreit and U. Diebold, Surf. Sci. 461, 87 (2000). 9. M. Li, W. Hebenstreit, U. Diebold, A. M. Tyryshkin, M. K. Bowman, G. G. Dunham and M. A. Henderson, J. Phys. Chem. B104, 4944 (2000). 10. G. Charlton, P. B. Hoowes, C. L. Nicklin, P. Steadman, J. S. G. Taylor, C. A. Muryn, S. P. Harte, J. Mercer, R. McGrath, D. Norman, T. S. Turner and G. Thornton, Phys. Rev. Lett. 78, 495 (1997). 11. H. Onishi, K.-C. Fukui and Y. Iwasawa, Bull. Chem. Soc. Jpn. 68, 2447 (1995). 12. U. Diebold, J. F. Anderson, K. O. Ng and D. Vanderbilt, Phys. Rev. Lett. 77, 1322 (1996). 13. H. Norenberg, R. E. Tanner, K. D. Schierbaum, S. Fischer and G. A. D. Briggs, Surf. Sci. 396, 52 (1998). 14. R. A. Bennett, S. Poulston, P. Stone and M. Bowker, Phys. Rev. B59, 10341 (1999). 15. M. Li, W. Hebenstreit and U. Diebold, Surf. Sci. 414, L951 (1998).
16. M. Li, W. Hebenstreit, U. Diebold, M. A. Henderson and D. R. Jennison, Faraday Discuss. 114, 245 (1999). 17. G. L. Haller and D. E. Resasco, Adv. Catal. 36, 173 (1989). 18. J. M. Pan and T. E. Madey, Catal. Lett. 20, 269 (1993). 19. F. Pesty, H. P. Steinrueck and T. E. Madey, Surf. Sci. 339, 83 (1995). 20. K. D. Schierbaum, S. Fischer, M. C. Torquemada, J. L. d. Segovia, E. Roma` n and J. A. Martin-Gato, Surf. Sci. 345, 261 (1996). 21. Y. Gao, Y. Liang and S. A. Chambers, Surf. Sci. 365, 638 (1996). 22. D. R. Jennison, O. Dulub, W. Hebenstreit and U. Diebold, Surf. Sci. (2000), in preparation. 23. A. B. Boffa, H. C. Galloway, P. W. Jacobs, J. J. Benitez, J. D. Batteas, M. Salmeron, A. T. Bell and G. A. Somorjai, Surf. Sci. 326, 80 (1995). 24. R. A. Bennett, P. Stone and M. Bowker, Catal. Lett. 59, 99 (1999). 25. E. L. D. Hebenstreit, W. Hebenstreit and U. Diebold, submitted to Surf. Sci. (2000). 26. M. A. Henderson, Surf. Sci. 419, 174 (1999).
