Atomic-Scale Structure and Catalytic Reactivity of the RuO 2 (110) Surface H. Over, 1 * Y. D. Kim, 1 A. P. Seitsonen, 1,2 S. Wendt, 1 E. Lundgren, 3 M. Schmid, 3 P. Varga, 3 A. Morgante, 4 G. Ertl 1 The structure of RuO 2 (110) and the mechanism for catalytic carbon monoxide oxidation on this surface were studied by low-energy electron diffraction, scanning tunneling microscopy, and density-functional calculations. The RuO 2 (110) surface exposes bridging oxygen atoms and ruthenium atoms not capped by oxygen. The latter act as coordinatively unsaturated sites—a hy- pothesis introduced long ago to account for the catalytic activity of oxide surfaces— onto which carbon monoxide can chemisorb and from where it can react with neighboring lattice-oxygen to carbon dioxide. Under steady-state conditions, the consumed lattice-oxygen is continuously restored by oxygen uptake from the gas phase. The results provide atomic-scale verification of a general mechanism originally proposed by Mars and van Krevelen in 1954 and are likely to be of general relevance for the mechanism of catalytic reactions at oxide surfaces. Information about the elementary processes involved in heterogeneous catalysis may be obtained from studies with well-defined single crystal surfaces by using techniques from surface physics. In contrast to the conditions of “real” catalysis, such investi- gations are, however, usually restricted to low gas pressures. A striking example of the problems introduced by this “pressure gap” is presented by the seemingly simple oxidation of CO on some of the Pt group metals. Ru is a poor catalyst at low-pres- sure conditions (1), but turns out to be superior to Pt and Pd if operated in excess O 2 at atmospheric pressure (2). The original suggestion that this enhanced activity has to be attributed to the presence of a complete monolayer (ML) of O atoms chemisorbed on the Ru surface (3) could not be confirmed. Instead it was found that the latter phase [forming a (1  1)-O overlayer on a Ru(0001) single crystal surface (4 )] is quite inactive, and that the high reactivity is only reached if the total O 2 concentration exceeds the equivalent of about 3 ML (5, 6 ). It will be demonstrated that the active part of this “O- rich” ruthenium phase is RuO 2 , which grows epitaxially with its (110) plane parallel to the Ru(0001) surface at high O 2 exposures and elevated temperatures. Despite the enormous significance of transition metal oxides as catalysts (7, 8), little is known about the microscopic prop- erties of their surfaces. With the present system we combine structural information [derived from scanning tunneling micros- copy (STM) and quantitative low energy electron diffraction (LEED) in conjunction with density-functional theory (DFT) cal- culations] with data for CO adsorption and oxidation to obtain an atomic-scale picture of the reactivity of an oxide surface. The data allow the long-standing concept of coordinatively unsaturated sites (cus) in heterogeneous catalysis (9) to be verified and demonstrate the participation of O at- oms as constituents of the oxide lattice in the catalytic reaction. We prepared the “O-rich” phase by ex- posing a well-defined Ru(0001) single crystal surface to high doses of O 2 (typi- cally about 10 -2 mbar for several minutes) at 700 K. In this way, a total O 2 uptake equivalent to about 10 ML was achieved, as determined by analyzing subsequently tak- en thermal desorption spectra (TDS). The LEED pattern resulting from such a treat- ment consists of a superposition of the hexagonal array of diffraction spots with the (1  1) periodicity of the Ru(0001) substrate, and additional sharp spots arising from three domain orientations of a lat- tice with a rectangular mesh with dimen- sions (6.4  0.3 Å by 3.1  0.2 Å). Within the limits of accuracy, the latter agree with the lattice parameters, (6.38 Å by 3.11 Å), of the (110) plane of bulk RuO 2 ; the fol- lowing quantitative analysis confirms this assignment. Exposure to CO at room temperature re- sulted in changes in the intensities of the LEED spots from the RuO 2 structure but not of those with (1  1) periodicity of the Ru(0001) substrate. This is a strong indica- tion that the superposition of diffraction spots in the LEED pattern arises from the co- existence of patches of Ru(0001)-(1  1)-O (which does not adsorb CO at 300 K) and of RuO 2 (110) (to which CO is strongly bound as outlined below). This view is confirmed by STM data (Fig. 1). The LEED and STM data show that a RuO 2 (110) film is growing epitaxially on the Ru(0001) surface as an incommen- surate overlayer. A film thickness of about 10 to 20 Å is derived from the STM corrugation when crossing the boundary to an adjacent Ru(0001)-(1  1)-O island, in agreement with x-ray reflectivity measurements (10). The RuO 2 (110) domains are well ordered and ex- hibit typical dimensions of several hundred angstroms. Structural analysis was performed by quantitative LEED [the standard technique for surface crystallography (11)] in conjunc- tion with DFT calculations. LEED intensity versus energy data from beams arising from the hexagonal (1  1) phase revealed a struc- ture identical to that of the previously ana- lyzed Ru(0001)-(1  1)-O phase (4 ). The structure for the bare RuO 2 (110) surface is identical to the ideal termination and is re- produced in Fig. 2A. The electronic structure of RuO 2 (110) was examined by pseudovalence charge 1 Fritz-Haber-Institut der Max-Planck-Gesellschaft, De- partment of Physical Chemistry, Faradayweg 4-6, D-14195 Berlin, Germany. 2 Istituto Nazionale per la Fisica della Materia, Unita ` di Roma, Dipartimento di Fisica, Universita ` La Sapienza, Piazzale A. Moro 2, I-00185 Rome, Italy. 3 Institut fu ¨r Allgemeine Physik, Technische Universita ¨t Wien, Wiedner Hauptstr. 8-10, A-1040 Vienna, Austria. 4 Dipartimento di Fisica, Uni- versita ` di Trieste, Via Valerio 2, I-34127 Trieste and TASC-INFM Laboratory, Padriciano 99, I-34012 Tri- este, Italy. *To whom correspondence should be addressed. E- mail: over@fhi-berlin.mpg.de Fig. 1. Large-scale STM (1000 Å by 1000 Å) image together with enlargements (50 Å by 50 Å) as insets. The right side represents a domain of the Ru(0001)-(1  1)-O surface; the magnified inset reveals a hexagonal ar- rangement of dots corresponding to the Ru(0001) lattice where the dark spots mark the locations of the chemisorbed O atoms in the (1  1) overlayer (21). The bright spots represent Ru atoms coordinated to three O atoms. On the left side, RuO 2 (110) domains are visible. The magnified inset shows the internal structure of this phase (parallel rows along the [001] direction) and its rectangular unit cell. The latter is identical to that de- rived from the LEED pattern. Tunneling pa- rameters are U =-1.21 V and 0.46 nA. R EPORTS 25 FEBRUARY 2000 VOL 287 SCIENCE www.sciencemag.org 1474