Spectroscopy Using Quantum Logic We Excite the Spectroscopy Transition by Applying Coherent Radiation Tuned Near the P

R EPORTS

10. G. F. Bignami, P. A. Caraveo, R. C. Lamb, T. H. 24. S. Corbel et al., Science 298, 196 (2002). Council (PPARC), the IPNP of Charles University, the Markert, J. A. Paul, Astrophys. J. 247, L85 (1981). 25. L. Maraschi, A. Treves, Mon. Not. R. Astron. Soc. 194, South African Department of Science and Technology 11. C. Motch, F. Haberl, K. Dennerl, M. Pakull, E. Janot- 1P (1981). and National Research Foundation, and the University Pacheco, Astron. Astrophys. 323, 853 (1997). 26. M. Beilicke, M. Ouchrif, G. Rowell, S. Schlenker, IAU of Namibia. We appreciate the excellent work of the 12. K. Bernlo¨hr et al., Astropart. Phys. 20, 111 (2003). Circ. 8300, 2 (2004). technical support staff in Berlin, Durham, Hamburg, 13. J. A. Hinton, N. Astron. Rev. 48, 331 (2004). 27. N. J. Tasker, J. J. Condon, A. E. Wright, M. R. Griffith, Heidelberg, Palaiseau, Paris, Saclay, and Namibia in the 14. S. Funk et al., Astropart. Phys. 22, 285 (2004). Astron. J. 107, 2115 (1994). construction and operation of the equipment. L.C., C.M., 15. F. Aharonian et al., Science 307, 1938 (2005). 28. M. Ribo´, P. Reig, J. Martı´, J. M. Paredes, Astron. M.O., and M.T. are also affiliated with the European 16. F. Aharonian et al., Astropart. Phys. 22, 109 (2004). Astrophys. 347, 518 (1999). Associated Laboratory for Gamma-Ray Astronomy, 17. M. de Naurois et al., in 28th International Cosmic Ray 29. J. Martı´, J. M. Paredes, M. Ribo´, Astron. Astrophys. jointly supported by CNRS and the Max Planck Society. Conference,T.Kajitaet al., Eds. (Universal Academy 338, L71 (1998). Press, Tokyo, 2003), p. 2907. 30. J. S. Clark et al., Astron. Astrophys. 376, 476 (2001). Supporting Online Material 18. F. Piron et al., Astron. Astrophys. 374, 895 (2001). 31. The support of the Namibian authorities and of the www.sciencemag.org/cgi/content/full/1113764/DC1 19. M. Ribo´ et al., Astron. Astrophys. 384, 954 (2002). University of Namibia in facilitating the construction Fig. S1 20. J. J. Condon et al., Astron. J. 115, 1693 (1998). and operation of HESS is gratefully acknowledged, as is References 21. A. Martocchia, C. Motch, I. Negueruela, Astron. the support of the German Ministry for Education and Astrophys. 430, 245 (2005). Research (BMBF), the Max Planck Society, the French 19 April 2005; accepted 16 June 2005 22. V. Bosch-Ramon, J. M. Paredes, Astron. Astrophys. Ministry for Research, the CNRS-IN2P3 and the Published online 7 July 2005; 417, 1075 (2004). Astroparticle Interdisciplinary Programme of the CNRS, 10.1126/science.1113764 23. R. J. Gould, G. P. Schre´der, Phys. Rev. 155, 1408 (1967). the UK Particle Physics and Astronomy Research Include this information when citing this paper.

the spectroscopy and logic ion, respectively. Spectroscopy Using Quantum Logic We excite the spectroscopy transition by applying coherent radiation tuned near the P. O. Schmidt,*. T. Rosenband, C. Langer, W. M. Itano, spectroscopy ion_s transition resonance (Fig. J. C. Bergquist, D. J. Wineland 1B), leading to Y 0 k,À kjÀ k,À k À We present a general technique for precision spectroscopy of atoms that lack Y0 Y1 ðÞa S þ b S L 0 m suitable transitions for efficient laser cooling, internal state preparation, and 0 k,À k À kjÀ k À k,À detection. In our implementation with trapped atomic ions, an auxiliary ‘‘logic’’ ðÞa S 0 m þ b S 0 m L ð1Þ ion provides sympathetic laser cooling, state initialization, and detection for a simultaneously trapped ‘‘spectroscopy’’ ion. Detection is achieved by applying a where kak2 þ kbk2 0 1. We now drive a red mapping operation to each ion, which results in a coherent transfer of the sideband (RSB) p pulse (5, 9) on the spec- spectroscopy ion’s internal state onto the logic ion, where it is then measured troscopy ion (Fig. 1C) so that with high efficiency. Experimental realization, by using 9Beþ as the logic ion and 27 þ Al as the spectroscopy ion, indicates the feasibility of applying this technique Y1 Y Y2 0 ðÞak,ÀSk0Àm þ bk,ÀSk1Àm k,ÀL to make accurate optical clocks based on single ions. 0k,ÀSk,ÀLðÞðak0Àm þ bk1Àm 2Þ The continued development of atomic-state describe how two different atomic species can manipulation techniques has led to improve- share these requirements, so that one species thereby mapping the spectroscopy ion_s ments in physical measurements and devices. need only have a good spectroscopy transition, internal state to the transfer mode, which is A good example of this improvement is in the and the other species can fulfill the other three shared by both ions. During the mapping area of atomic spectroscopy, where increases requirements, as outlined in (4). We demon- process, a transient entanglement is created in resolution and accuracy to levels better than strate this technique experimentally with trapped between the internal state of the 27Alþ ion 15 k,À k À 1partin10 (1, 2) have been achieved with atomic ions. and the motional state. The S 0 m compo- atomic clocks. This, in turn, has improved tests To describe the transfer protocol, we con- nent of the wave function is unaffected by kjÀ k À of fundamental theories (3). To achieve these sider two atomic ions of different species lo- this operation, because the state S –1 m levels of performance in spectroscopy, four key cated on the weak axis of a three-dimensional does not exist (5, 9); this is the key element requirements are as follows: (i) selection of an trap. We designate one of the ions as the logic of quantum logic used here (10, 11). We ap- atom with a good reference or Bspectroscopy[ ion and the other as the spectroscopy ion, and ply a final RSB p pulse on the logic ion (Fig. transition, which is suitably narrow and rela- we treat the internal states of the ions as two- 1D), which yields tively immune to environmental perturbations; level systems with eigenstates labeled k,À and kjÀ (ii) cooling to minimize velocity-induced fre- . The Coulomb interaction of the ions Y2 Y Yfinal 0k,ÀSðÞak,ÀL þ bkjÀL k0Àm ð3Þ quency shifts; (iii) reliable initial state prepara- couples their motion, which is best described tion; and (iv) efficient state detection. in a normal-mode basis (5–7). All normal and thereby completes the mapping of the Previously, to satisfy these requirements, modes are assumed to be initially cooled to spectroscopy ion_s state onto the logic ion. an atom was chosen that simultaneously had near their ground states by Doppler laser The logic ion is next measured (not shown k,À kjÀ a good spectroscopy transition and other, cooling on the logic ion, which sympathetical- in Fig. 1), projecting its state to L or L, more strongly allowed transitions to accom- ly cools the spectroscopy ion (8). which can be efficiently distinguished. By plish requirements (ii) to (iv) (1, 2). Here we One of the ions_ normal modes (which has repeating this experiment many times, we can k À k k2 k k2 harmonic oscillator states labeled n m,wheren determine the probabilities a and b as func- Time and Frequency Division, National Institute of denotes the nth level of the mode m) is chosen tions of the spectroscopy probe frequency and Standards and Technology, 325 Broadway, Boulder, as the transfer mode on which the state thereby determine the spectroscopy ion_stran- CO 80305, USA. mapping is implemented. This mode is cooled sition frequency. *Present address: Institut fu¨r Experimentalphysik, to its ground state (7). Figure 1A shows both We implemented the technique with a single Universita¨t Innsbruck, Technikerstrasse 25, A-6020 ions and the transfer mode in their ground 9Beþ logic ion (5, 11)andasingle27Alþ Innsbruck, Austria. 0 .To whom correspondence should be addressed. states, described by the wave function y0 spectroscopy ion (Fig. 2) simultaneously k,À k,À k À E-mail: [email protected] S L 0 m, where the indices S and L denote trapped in a linear Paul trap similar to that de-

www.sciencemag.org SCIENCE VOL 309 29 JULY 2005 749 R EPORTS 27 þ , È scribed in (12). Our choice for Al was moti- 2p 13.8 MHz and wy 2p 14.9 MHz quence took 1msandwasrepeated700 vated by its potential as a high-accuracy optical orthogonal to the weak trap axis. We performed times per data point. The observed linewidth 1 3 j clock based on the S0 $ P0 reference tran- Doppler cooling with a sˆ -polarized laser beam of the transition is Fourier-transform–limited K , 9 þ k,À Yk2 µ 0 0 sition, which has a Q factor of Q f/Df 2 tuned to the Be L P3/2, F 3, mFµ with a fitted full width at half maximum line- 1017,wheref is the frequency of the transition –3À transition. Counting fluorescence photons width of 63 kHz for an excitation pulse 0 and Df is its width (13). Although it was pre- on this cycling transition also allowed us to duration of tp 12.6 ms. The observed contrast viously considered (14, 15), 27Alþ has not yet determine the 9Beþ ion_s final measured state, of 93% (normalized to the contrast of a 9Beþ k,À been used as an optical frequency standard because an ion in L fluoresces strongly, sideband transition) was limited by the sponta- kjÀ 27 þ 3 because it lacks an accessible cooling transition. whereas an ion in L fluoresces negligibly neous decay of the Al P1 state during the As a demonstration of the technique, we for this laser frequency (11, 14). State rotations protocol and by a shot-to-shot variation of the k1 0 9 þ probed the magnetic substates of the S0, F and resolved-sideband cooling on Be were Rabi frequency caused by fluctuations of pho- À Yk3 ¶ 0 À 5/2 P1, F 7/2 transitions, where F and performed by means of two-photon stimulated non number in the Doppler-cooled radial modes ¶ k,À k À kjÀ k ¶À F are the total angular momentum quantum Raman pulses on L n m $ L n m transi- (Debye-Waller reduction factors) (18, 19). numbers of the ground and excited state, tions (5, 11). Figure 3B shows internal state oscilla- respectively, in 27Alþ Ewavelength l , 267 Ramansidebandcoolingofthetwoaxial tions on the aluminum ion obtained by , ^ 9 þ nm, natural linewidth G/2p 520 Hz (16) .We modes, optical pumping of Be (7, 11), and varying the interrogation duration ti (on- drove these transitions with a frequency- state preparation of the spectroscopy ion (see resonance Rabi flopping) (5). The observed doubled dye laser that had been stabilized to below) initialize the ion pair to the state coherence time of 118 ms was limited by the k,À k,À k À , 3 an isolated high-finesse cavity (17), yielding a S L 0 m, which is the starting point for the 305-ms lifetime of the P1 excited state, spectral linewidth of less than 6 Hz. Two spectroscopy protocol described above. State Rabi frequency fluctuations caused by the separately switchable probe laser beams, transfer was carried out using the in-phase radial motion Debye-Waller factors, and incident at 45- with respect to the weak trap weak-axis mode (7). magnetic field fluctuations. axis, were pˆ- and sˆ þ=j -polarized, respectively. Figure 3A shows experimental data for We verified the coherence of the transfer 9 þ k,À K k2 0 The Be system, with L S1/2, F 2, single–laser pulse (Rabi) spectroscopy of the process by performing a Ramsey experiment 0 À kjÀ K k2 ¶ 0 0 À 27 þ k1 0 0 À 0k,À Yk3 mF –2 and L S1/2, F 1, mF¶ –1 , Al S0, F 5/2, mF 5/2 S P1, in which the first (optical) p/2 pulse created a ¶ 0 0 À 0kjÀ where mF is the magnetic quantum number of F 7/2, mF¶ 7/2 S transition. We ap- superposition state in the aluminum ion. This the total angular momentum F, has been de- plied interrogation pulses of constant intensity coherence was then mapped to the beryllium scribedindetail(7, 11). The trap provides and duration, corresponding to approximate p 9 þ , single Be trap frequencies of wz 2p 3.8 pulses on this transition when the laser beam , MHz along the weak axis (z axis) and wx was tuned to resonance. An experimental se-

Fig. 1. Spectroscopy and transfer scheme for spectroscopy (S) and logic (L) ions sharing a common normal mode of motion, the transfer mode, with excitation n. (Only the ground and first excited Fig. 2. Partial 9Beþ and 27Alþ energy level dia- states of the transfer mode are shown.) (A) Initialization to the ground internal and transfer-mode grams (not to scale). Shown are the relevant states. (B) Interrogation of the spectroscopy transition. (C) Coherent transfer of the internal transitions for Doppler and Raman cooling on the superposition state of the spectroscopy ion into a motional superposition state by use of an RSB p 9Beþ ion, the spectroscopy transition, and the pulse on the spectroscopy ion. (D) Coherent transfer of the motional superposition state into an difficult-to-reach Doppler cooling transition at internal superposition state of the logic ion by use of an RSB p pulse on the logic ion. 167 nm on the 27Alþ ion.

k1 0 0 À Yk3 ¶ 0 Fig. 3. (A) Rabi spectroscopy of the S0, F 5/2, mF 5/2 P1, F probe pulse. (B) Rabi flopping at the center frequency of the transition 0 À 27 þ 7/2, mF¶ 7/2 transition in Al , showing a frequency scan across the in (A). The data are fit by an exponentially damped sinusoidal function. resonance. The data (black circles) are fit by the theoretically expected (C) Two-ion Ramsey time scan. The data are fit by a sinusoidal 27 þ probability P,,S of finding Al in the ground state after applying the function.

750 29 JULY 2005 VOL 309 SCIENCE www.sciencemag.org R EPORTS ion, where it was probed with a second (Raman ble in the spectroscopy atom, we can again rely negligible. For this purpose, we replaced the transition corresponding to a radio frequency) on quantum logic methods for state prepara- RSB pulse in all preparation stages with a 27 þ k,À K p/2 pulse applied to the beryllium ion. Be- tion. The process for preparing the Al S wait time of the same duration, thereby k1 0 0 À 27 þ 3 Y 1 tween the two Ramsey pulses, we mapped the S0, F 5/2, mF þ5/2 state is outlined in allowing for Al P1 S0 radiative aluminum ion_s internal state onto the transfer Fig. 4A. If we assume that the transfer mode is decay during the remaining time (G1.5 ms) mode, waited a duration td, and then mapped initialized in its ground state, each stage of the of the stage. Figure 4C shows the contrast of 27 þ 0 Y 0 the transfer mode state onto the beryllium ion. preparation process is composed of three steps: the Al mF þ5/2 mF ¶ þ5/2 transi- 0 0 Y 0 Figure 3C shows a plot of the detected popu- (i) a pˆ-polarized, mF-preserving (mF¶ mF) p tion, normalized to the mF þ5/2 mF ¶ lation P,,L of the beryllium ion versus td.This pulse that does not change the motional state þ7/2 transition, as a function of the number of B [ j signal oscillates with a frequency correspond- ( carrier transition); (ii) a sˆ -polarized, mF- preparation repetitions for both cases. The de- 0 ing to the transfer mode frequency. The optical changing (mF mF¶ þ 1) RSB p pulse; and terministic preparation achieves the maximum phases of the logic ion and spectroscopy ion (iii) a laser cooling stage (on the 9Beþ ion) that contrast after a single preparation sequence, laser beams do not need to be stabilized from puts the transfer mode back into its ground whereas we observed only a slow increase in experiment to experiment, because the phase state. The cooling stage incorporates spontane- contrast for optical pumping (21). This result k,À kjÀ between the states L and L after the last ousemission,whichisrequiredinnormal also demonstrates that initial state prepara- Ramsey pulse depends on the phase differ- optical pumping and makes the RSB p pulse tion of spectroscopy ions can be achieved ences between the two pulses applied to each irreversible, hence insuring irreversibility of without depending on spontaneous emission ion, both of which were held constant over all the overall process. We can Bsweep[ the from their excited state, which is particularly 0 experiments (20). population to the desired state mF þ5/2 by important for ions with a long-lived excited 27 þ 1 3 The spectroscopy protocol does not depend sequentially applying these three steps to the state, as in the case of the Al S0$ P0 0 I on phase coherence between the laser pulses, mF –5/2, –3/2, , þ3/2 ground states (in clock transition. 0 because we measure only the spectroscopy Fig. 4A, only the transfer from the mF þ1/2 We have demonstrated spectroscopy using ion_s state probabilities kak2 and kbk2.However, and þ3/2 states is shown). By adapting this quantum logic, a technique for precision the coherence of the process can be useful; for procedure, we can prepare any of the 27Alþ atomic spectroscopy that removes the require- 0 example, the Ramsey experiment provides a ground states. In Fig. 4B, we show DmF 0 ments of efficient cooling, state preparation, sensitive way to measure and correct for drifts spectroscopy on all Zeeman states of the 27Alþ and state detection from the species upon k1 0 À Yk3 ¶ 0 À in the mode frequency. (For this purpose, it is, S0, F 5/2 P1, F 7/2 transition in a which the spectroscopy is performed. Although in practice, easier to perform both Ramsey and magnetic field of 3 mT (chosen to spectrally we have implemented this technique with mapping pulses on the 9Beþ ion.) resolve all transitions) by using this state atomic ions, it may also be applicable to other In spectroscopy experiments, initial state preparation technique. systems for which quantum logic techniques preparation is often accomplished with optical We demonstrated in a separate experi- are being developed, including neutral atoms pumping on a strongly allowed transition. ment that the effect of optical pumping of (22, 23) and perhaps molecules with simple Because such a transition may not be accessi- the 27Alþ during deterministic preparation is internal state structure. This technique offers a way to investigate new atomic ion species and their potential for providing reference transitions for high-accuracy optical clocks, such 1 3 27 þ 10 þ as the S0 $ P0 transitions in Al and B (4, 15) or optical transitions in Heþ (24, 25). Because coherence of the transfer process is not required for spectroscopy, it should be possible to detect transitions in the spectroscopy ion by detecting the recoil upon absorp- tion (24). On the other hand, the coherence in the transfer process could be beneficial in quantum information experiments, where the information could be distributed between different species, thereby simplifying the laser beam addressing of adjacent quantum bits.

References and Notes 1. P. Gill, Ed., Proceedings of the 6th Symposium on Frequency Standards and Metrology (World Scientific, Singapore, 2002). 2.S.A.Diddams,J.C.Bergquist,S.R.Jefferts,C.W.Oates, Science 306, 1318 (2004). 3. S. G. Karshenboim, E. Peik, Eds., Astrophysics, Clocks and Fundamental Constants, Lecture Notes in Physics 648 (Springer, Berlin, Heidelberg, 2004). 4.D.J.Wineland,J.C.Bergquist,J.J.Bollinger,R.E. Drullinger,W.M.Itano,inProceedings of the 6th Sym- posium on Frequency Standards and Metrology,P.Gill, Ed. (World Scientific, Singapore, 2002), pp. 361–368. Fig. 4. Deterministic state preparation of the spectroscopy ion. (A) The last steps of the deterministic 5. D. J. Wineland et al., J. Res. Nat. Inst. Stand. Technol. 0 103, 259 (1998). preparation of the mF þ5/2 ground state consist of pˆ-polarized carrier (CAR) p pulses followed by sˆ j p 27 þ 9 þ 6. G. Morigi, H. Walther, Eur. Phys. J. D 13, 261 (2000). -polarized RSB pulses on Al and ground state cooling on Be (CL), to ensure irreversibility of 7. M. D. Barrett et al., Phys. Rev. A. 68, 042302 (2003). the sequence. (B) Carrier Rabi spectroscopy of all pˆ-polarized transitions in 27Alþ using deterministic 0 Y 8. D. J. Larson, J. C. Bergquist, J. J. Bollinger, W. M. Itano, preparation. (C)NormalizedcontrastofthemF þ5/2 þ5/2 pˆ-polarized transition versus number D. J. Wineland, Phys. Rev. Lett. 57, 70 (1986). of preparation repetitions for optical pumping and the deterministic preparation. The lines are guides 9. D. M. Meekhof, C. Monroe, W. M. Itano, B. E. King, D. J. to the eye. Wineland, Phys. Rev. Lett. 76, 1796 (1996).

www.sciencemag.org SCIENCE VOL 309 29 JULY 2005 751 R EPORTS

10. J. I. Cirac, P. Zoller, Phys. Rev. Lett. 74, 4091 (1995). coherent transfer step, Fig. 1C), thus eliminating a deg- ulating discussions and comments on the manuscript. 11. C. Monroe, D. M. Meekhof, B. E. King, W. M. Itano, D. J. radation of linewidth and contrast caused by transfer This work was supported by the Office of Naval Re- Wineland, Phys. Rev. Lett. 75, 4714 (1995). mode heating during interrogation. search, the Advanced Research and Development 12. M. A. Rowe et al., Quant. Inform. Comp. 2, 257 (2002). 20. Methods are available as supporting materials on Activity/National Security Agency, and the National 13. W. M. Itano, unpublished calculation. Science Online. Institute of Standards and Technology (NIST). P.O.S. 14. H. G. Dehmelt, IEEE Trans. Instrum. Meas. IM-31,83 21. We did not prepare the ion in any particular state acknowledges support from the Alexander von Hum- (1982). before each experiment. Rather, competition between boldt Foundation. This work is a contribution of NIST, 15. N. Yu, H. Dehmelt, W. Nagourney, Proc. Natl. Acad. state preparation and off-resonant depumping deter- not subject to U.S. copyright. Sci. U.S.A. 89, 7289 (1992). mined the distribution of polarizations before the next 16. E. Tra¨bert, A. Wolf, J. Linkemann, X. Tordoir, J. Phys. repetition of the experiment. Supporting Online Material B At. Mol. Opt. Phys. 32, 537 (1999). 22. K. Eckert et al., Phys. Rev. A. 66, 042317 (2002). www.sciencemag.org/cgi/content/full/309/5735/749/ 17. B. Young, F. Cruz, W. Itano, J. Bergquist, Phys. Rev. 23. E. Charron, E. Tiesinga, F. Mies, C. Williams, Phys. DC1 Lett. 82, 3799 (1999). Rev. Lett. 88, 077901 (2002). Materials and Methods 18. D. J. Wineland, W. M. Itano, Phys. Rev. A. 20, 1521 (1979). 24. B. Roth, U. Fro¨hlich, S. Schiller, Phys. Rev. Lett. 94, 19. For long interrogation times, sideband cooling on the 053001 (2005). logic ion can be applied during and/or after interroga- 25. M. Herrmann et al., in preparation. 3 May 2005; accepted 16 June 2005 tion of the spectroscopy transition (before the first 26. We thank S. Diddams, J. Koelemeij, and J. Ye for stim- 10.1126/science.1114375

ture, defect mobility, and valence of the cerium Electron Localization Determines ions is still missing (3, 11, 12, 14, 17–21). Ceria crystallizes in a cubic fluorite struc- Defect Formation on ture and exposes the thermodynamically most stable (111) surface (3). This surface is the Ceria Substrates oxygen termination of stoichiometric O-Ce-O trilayers stacked along the E111^ direction and Friedrich Esch,1* Stefano Fabris,2 Ling Zhou,3 also represents the major fraction of the active Tiziano Montini,4,5 Cristina Africh,1,5,6 Paolo Fornasiero,4,5 surface in catalytic nanocrystallites (12). Giovanni Comelli,1,5,6 Renzo Rosei1,5,6 We report high-resolution scanning tunneling microscopy (STM) results on CeO2(111), which we interpret in light of density func- The high performance of ceria (CeO2) as an oxygen buffer and active support for noble metals in catalysis relies on an efficient supply of lattice oxygen at tional theory (DFT) calculations. The single- reaction sites governed by oxygen vacancy formation. We used high-resolution crystal sample was cleaned by repeated cycles - scanning tunneling microscopy and density functional calculations to unravel of sputtering and annealing to 900 C. Anneal- the local structure of surface and subsurface oxygen vacancies on the (111) ing at this temperature leads to oxygen release surface. Electrons left behind by released oxygen localize on cerium ions. and is therefore a means to control the degree Clusters of more than two vacancies exclusively expose these reduced cerium of surface reduction of the sample. Although ions, primarily by including subsurface vacancies, which therefore play a crucial ceria is an insulator (band gap 6 eV), its role in the process of vacancy cluster formation. These results have implications reduction and the elevated temperatures used - - for our understanding of oxidation processes on reducible rare-earth oxides. in this study (300 to 400 C) enhance the electron conductivity (3). Furthermore, the

Materials based on ceria (CeO2) are used in the of transition-metal oxide surfaces, a thorough elevated temperatures considerably suppress production and purification of hydrogen, the characterization of these vacancies has led to the adsorption of species from the ultrahigh purification of exhaust gases in three-way auto- greater understanding of the fundamental vacuum chamber background (10j10 mbar,

motive catalytic converters, and other catalytic features of the reactivity (6–9) and to the mainly H2,H2O, and CO) that would interfere applications (1–4). In all such applications, design of efficient supported catalysts (10). with the microscope tip. In this way, atomic highly mobile lattice oxygen is involved in This knowledge has been lacking for the rare- resolution could be obtained with STM in oxidation processes. Over a wide range of earth oxides, but their chemistry is likely large-scale images (Fig. 1). The DFT calcu- working temperatures (from room temperature different because excess electrons left behind lations are based on an advanced implementa- to 1000-C), ceria plays two key roles: (i) by the removal of neutral oxygen localize on tion (22–24) that captures full electron releasing and storing oxygen, and (ii) promot- empty f states. In ceria, this results in the localization on Ce 4f states in a manner not ing noble-metal activity and dispersion (3, 5). valence change Ce4þ Y Ce3þ of two cations directly accessible to standard DFT calcula- Both phenomena are controlled by the type, per vacancy and in an extraordinary efficiency tions (25). Filled-state images are shown to size, and distribution of oxygen vacancies as for reversible oxygen release (3, 11, 12). map the positions of the outermost oxygen the most relevant surface defects. In the case Reduction drastically modifies the reactivity of atoms. Their local relaxation around the de- ceria substrates (13, 14). fects discriminates between different defect 1Laboratorio Nazionale Tecnologie Avanzate e Nano- Oxygen vacancies are also crucial for the geometries. scienza (TASC)–Istituto Nazionale per la Fisica della binding of catalytically active species to ceria Various defects had already been observed Materia (INFM), 34012 Trieste, Italy. 2Scuola Inter- (3). The high activity of Au/ceria catalysts in in earlier STM and atomic force microscopy nazionale Superiore di Studi Avanzati (SISSA) and the water-gas shift reaction has recently been studies of ceria at lower temperatures (17–21), INFM Democritos National Simulation Center, 34014 Trieste, Italy. 3Institute of Physical Chemistry and traced back to highly dispersed, ionic Au spe- but on a small scale and with limited resolution Electrochemistry, University of Hannover, D-30167 cies that form only in the presence of defects that precluded a discussion of their relative dis- Hannover, Germany. 4Chemistry Department and (15, 16). tribution. Moreover, Namai et al. reported de- Consorzio Interuniversitario Nazionale per la Scienza 5 The control of the density and the nature of fect mobility even at room temperature (20, 21). e Tecnologia dei Materiali (INSTM), Center of oxygen vacancies could provide a means for These results are in contrast to our observa- Excellence for Nanostructured Materials (CENMAT), 6Physics Department, University of Trieste, 34127 tailoring the reactivity of ceria-based catalysts. tions: None of the defects shown in Fig. 1 is Trieste, Italy. However, despite extensive spectroscopic, mi- mobile on a time scale of minutes. Direct dif- *To whom correspondence should be addressed. croscopic, and diffraction studies, a detailed fusion of vacancies (i.e., hopping of lattice E-mail: [email protected] atomic-level insight into the local defect struc- oxygen) on this surface requires temperatures

752 29 JULY 2005 VOL 309 SCIENCE www.sciencemag.org