Initial Stages of Oxidation on Graphitic Surfaces: Photoemission Study

Apr 29, 2009 - (9) Identifying the oxygenated functional groups introduced in the C network during ... Our approach is to study the initial stages of ...
4 downloads 0 Views 747KB Size
Download PDF
9009

2009, 113, 9009–9013 Published on Web 04/29/2009

Initial Stages of Oxidation on Graphitic Surfaces: Photoemission Study and Density Functional Theory Calculations Alexei Barinov,*,† O. Baris¸ Malciogˇlu,‡ Stefano Fabris,‡ Tao Sun,‡ Luca Gregoratti,† Matteo Dalmiglio,† and Maya Kiskinova† Sincrotrone Trieste, s.s. 14-km. 163.5 in AREA Science Park, 34012 BasoVizza, Trieste, Italy, and INFM-CNR DEMOCRITOS Theory@Elettra Group and SISSA-Scuola Internazionale Superiore di Studi AVanzati, Via Beirut 4, 34014 Trieste, Italy ReceiVed: March 6, 2009; ReVised Manuscript ReceiVed: April 9, 2009

The initial oxidation stages of perfect and defective graphitic surfaces exposed to atomic oxygen have been studied with a combined high-resolution photoemission spectroscopy (HR-PES) and density functional theory (DFT) computational approach. The resulting oxygen-containing surface functional groups are identified by analyzing the multicomponent C 1s and O 1s core level spectra that are then interpreted on the basis of DFT calculations. In the initial oxidation stage, epoxy groups are formed on perfect graphene, whereas the preferential adsorption of the O atoms on the vacancies of the defective surfaces results in structures containing pairs of oxygen atoms in ether and carbonyl (semiquinone) configurations. The formation of these functional groups is preceded by metastable structures consisting of single O atoms occupying single C vacancies. Remarkable physical and mechanical properties of carbon nanostructures are determined by the two-dimensional honeycomb C-C network, which constitutes the building block of nanotubes, graphite, and graphene sheets. These materials are hydrophobic and chemically inert, in particular when the perfect graphene structure is preserved during fabrication and purification. Many technological applications aim at tailoring the physical, chemical, and electronic properties of these carbonbased materials by reshaping their honeycomb structure. Oxidation is considered to be the most promising and effective approach for selectively functionalizing the surface of these materials in order to modify their properties in a desired manner.1-8 As an example, functionalized graphene sheets can be produced by oxidation of graphite to graphite oxide.9 Identifying the oxygenated functional groups introduced in the C network during oxidation is a real challenge and there is a lot of controversy in this respect even for the simplest case of graphite oxide.10 The major difficulty is that the oxidative functionalization results in the formation of several different O-containing functional groups often in conjunction with hydroxyl groups. There is no consensus in the literature regarding the type and relative abundance of the great variety of possible O-functional groups (epoxide, carbonyl, quinone, carboxylic, ether, ketonic, and others) that can form on the surfaces during the oxidation process. Another important aspect is that carbon materials contain various intrinsic structural defects,11 which besides controlling the electrical,12 transport,13 and chemical14,15 properties, have significant influence on the oxidation process. * To whom correspondence should be addressed. E-mail: alexey.barinov@ elettra.trieste.it. Tel.: +39 040 375 8032. Fax: +39 040 938 8565. † Sincrotrone Trieste. ‡ INFM-CNR DEMOCRITOS Theory@Elettra Group and SISSA-Scuola Internazionale Superiore di Studi Avanzati.

10.1021/jp902051d CCC: $40.75

This study aims at establishing the correlation between surface defects and oxygenated functional groups by identifying the bonding configurations of oxygen on the surface of “perfect” and defective highly ordered pyrolitic graphite (HOPG), formed in the initial oxidation stages. The formation of C-O bonds on perfect and defective graphite was investigated under wellcontrolled oxidation conditions for HOPG surfaces with various concentrations of defects using high-resolution photoemission spectroscopy (HR-PES) in combination with density functional theory calculations (DFT). We show that C-O-C epoxy groups are dominant species formed on the perfect graphene surfaces, whereas on defective surfaces, the O atoms are trapped by the single and di- vacancies leading to the formation of both CdO carbonyl (semiquinone, referred more generically as a ketone hereafter) and C-O-C ether groups. Photoemission spectroscopy is widely used for monitoring the chemical state of oxidized C materials also in the context of complex catalytic applications such as oxidative dehydrogenation.16 In our recent HR-PES study we demonstrated that the C 1s core level spectra are an excellent fingerprint of the structural quality of C-based materials.13 DFT calculations have already predicted formation of epoxy groups in the initial stage of graphite oxidation, and linear chain structures of such epoxy groups have been proposed to mediate the breakdown of the graphene network in oxidizing environment.17 However, we would like to note that in this previous study the C-network was chemically oxidized by acids in liquid solution and was modeled as isolated coronene molecule or nanotubes. Our approach is to study the initial stages of oxidation under highly controlled ultra high vacuum (UHV) conditions on well characterized samples that are then modeled as periodic graphite and graphene surfaces with DFT numerical simulations. The experiments were performed on three types of samples, differing in the density of surface defects: (A) “perfect” HOPG,  2009 American Chemical Society

9010

J. Phys. Chem. C, Vol. 113, No. 21, 2009

(B) weakly defective HOPG with vacancy density of ∼0.015 ML, and (C) highly defective HOPG with vacancy density of ∼0.06 ML. The perfect HOPG was obtained by cleaving graphite in air followed by heating in UHV at temperatures higher than 800 °C until the characteristic C 1s spectrum18 reported for atomically clean graphite was obtained. The defective HOPG samples were created by sputtering with 60 eV Ar+ ions, producing mostly single vacancies as shown by atomic force microscopy-scanning tunneling microscopy (AFMSTM) studies.4,19 After sputtering, the defective graphite was heated up to 800 °C until the Ar 2p photoemission line due to trapped Ar atoms in the graphite cage vanished completely. This annealing did not heal the created defects in the HOPG cage, as evidenced by C 1s spectra (see later). The following oxidation process was carried out by exposing these samples to a flux of atomic oxygen. The advantage of using oxygen plasma or UV photo-oxidation instead of molecular oxygen is that it is more efficient, and particularly in the initial stages the oxidation can be followed by controlling the atomic O dose. The oxygen flux was provided by a TECTRA radio frequency plasma source with the sample placed in front of a Mo plate with small pinholes separating the discharge region. Under these conditions mostly neutral thermalized O and O2 (O2 pressure 10-5 mbar during plasma processing) species can reach the sample surface.20 The O flux of 9 × 1014 atoms cm-2 min-1 at the sample surface was estimated from the Auger and X-ray photoelectron spectroscopy (XPS) analysis of O accumulated in silver and graphite test samples. Perfect and defective HOPG samples were transferred without vacuum loss into the UHV HR-PES measurement chamber. The evolution of the C 1s and O 1s spectra is used to follow the formation and density of the oxygen functional groups. The C 1s spectra were fitted with Doniach-Sunjic components, with parameters determined from high resolution photoemission spectra of HOPG with fixed Lorentzian width and asymmetry of 0.16 eV and 0.06, respectively.18 The broadening of the O 1s line was determined to be 0.9 eV from the spectra of the epoxy functional group acquired on perfect HOPG, while the Gaussian broadening related to the instrumental resolution was 0.25 eV. The numerical simulations were based on the density functional theory with the Perdew-Burke-Ernzerhof generalized gradient approximation (PBE-GGA) for the exchange and correlation energy functional.21 The spin-polarized Kohn-Sham equations were solved in the plane-waves pseudopotential framework by means of the Quantum-ESPRESSO computer package.22 The use of ultrasoft pseudopotentials allowed us to limit the plane-wave basis set to 30 Ry, while the charge density Fourier representation was limited by a cutoff of 300 Ry. Convergence of the results have been checked by increasing these cutoffs to 35 and 350 Ry, respectively. The calculated lattice parameter for graphite was 2.46 Å, in good agreement with previous theoretical studies reporting 2.46 Å.23 The oxidized surface of perfect and defective graphite was modeled with atomic oxygen adatoms on (7 × 7) and (5 × 5) periodic supercell slabs separated by 14 Å of vacuum. Spin polarization was considered in all cases and the ground state always resulted to be non magnetic (total spin equal to zero), also when starting from a spin polarized initial density. Integrations in the Brillouin zone were done by means of regular Monkhorst-Pack grids whose density was equivalent (or larger) to the one of the 14 × 14 mesh for the (1 × 1) supercell. The results obtained for a free-standing graphene layer were checked against those obtained with a three-layer thick graphite slab, in which the

Letters

Figure 1. Experimental C 1s spectra from “perfect” (A), weakly defective with 0.015 ML of vacancy defects (B) and strongly defective with 0.06 ML of vacancy defects in HOPG (C). In the inset of the panel C the atoms around a single vacancy in the graphene are colored in correspondence to their photoemission line shifts predicted by the DFT simulations. Three C atoms surrounding the vacancy are not equivalent due to Jahn-Teller symmetry breaking distortion,25 schematically shown by the green dotted line. In the inset of the panel B a zoom to the region of defect related components of weakly defective graphite is shown.

experimental interlayer spacing was used. The difference in the relaxed geometry of the functional groups calculated in the two supercells were negligible and the difference in core level shifts were always smaller than 0.01 eV. The O and C 1s core level shifts were calculated in the pseudopotential formalism by using the final state theory.24 Binding energy (BE) shifts of the core 1s level were approximated by pseudopotential differences in the total energies calculated in supercells containing an excited atom at different positions. The numerical error of the calculated BE shift is below 5 meV. The comparison between simulation and experiment for the epoxy group on perfect HOPG allow us to estimate the overall accuracy of the final state approximation to be better than 0.2 eV. In the present work, the excited atoms where represented by specific pseudopotentials generated in the following core-excited electron configurations: O* (1s1 2s2 2p4) and C*(1s1 2s2 2p2). Figure 1 shows the C 1s spectra of perfect (A), weakly defective (B) and highly defective (C) HOPG measured before oxygen exposure. Due to the small photoelectron escape depth of 5 Å in our experimental setup the spectra reflect exclusively the changes occurring at the surface. The C 1s spectrum of the perfect HOPG consists of a single feature centered at 284.5 eV, while in the spectra of the defective HOPG additional components appear at lower binding energies. These are interpreted on the basis of the calculated C 1s binding energy. According to the DFT calculations the C 1s core-level BE of all C atoms in the hexagons surrounding a single C vacancy undergo chemical shifts toward lower BE with respect to the C 1s peak of a perfect HOPG, i.e., each missing C atom causes chemical shift of 12 near-by atoms ranging the first to the third in-plane nearest neighbors.13 The resulting four distinct C 1s components, peaked at 284.25, 284, 283.7, and 283.4 eV and with relative intensities 6:3:2:1, respectively, are shown in Figure 1B and C. The color of the components corresponds to the color of atoms

Letters

J. Phys. Chem. C, Vol. 113, No. 21, 2009 9011

Figure 2. Experimental O 1s spectra corresponding to the initial oxidation of “perfect” (A), weakly (B), and strongly (C) defective HOPG samples. O dose is 4 × 1014 atoms cm-2 accompanied by 0.6 kL of O2 during plasma source operation. T ) 20 °C.

around the vacancy shown in the inset in Figure 3C. The deconvolution of the defective HOPG is in agreement with the calculations within 0.15 eV of BE. The surface of the highdefective sample C is almost saturated with defects and the majority of the C atoms at the surface undergo a chemical shift, which results in a dramatic change of the C 1s line shape (Figure 1C). The additional small component at 285 eV, needed for deconvolution, can be attributed to the presence of C adatoms or sp3 carbon.26 The O 1s spectra measured on the perfect and defective samples after exposure to O provide evidence that there are only three relevant surface species involving oxygen atoms bonded on the HOPG surfaces. This is illustrated by the O 1s spectra in Figure 2, measured for the three HOPG samples after the same low dose (atomic oxygen and molecular oxygen from the background). These spectra provide ideal fingerprints for the different C-O groups formed at the three surfaces in the initial oxidation stage. Since the perfect HOPG is completely inert to molecular O2 the species formed are result of bonding of atomic oxygen. Although the defective HOPG is less inert (the vacancies increase the O2 dissociation propensity4) the O 1s peak growth and visible changes in the C 1s spectra were detected at O2 exposures much higher (tens of kL) than the background one during plasma operation. The binding of the first oxygen atoms to the perfect graphene cage is characterized by a single O 1s component at 532.3 eV (Figure 2A). In agreement with previous studies,9,17 our DFT calculations predict that in the initial oxidation stage the O atoms bind by bridging two C atoms of the nondefective graphite surface and form surface epoxy groups (see inset in Figure 3A) that are strongly bound to the hexagonal network of the surface. The binding energy of the adsorbate calculated with respect to an isolated O atom is 1.80 eV (in the 7 × 7 supercell) and 1.92 eV (in the 5 × 5 supercell), in excellent agreement with previous studies reporting 1.92 eV.27 The calculated electron ground state for this surface epoxy group is spin unpolarized. We therefore attribute the O 1s component at 532.3 eV to the epoxy groups that are present on both ‘perfect’ and defective samples. We will show in the following that the two additional components measured on defective samples, one at higher BE, 533.5 eV,

Figure 3. Simulated C 1s core level shift for the binding of atomic O to perfect and defective graphite surfaces. (A) Epoxy group forming on undefective graphite, (B) one O atom in a C vacancy resulting in a planar 3-fold C-O configuration, (C) two O atoms in a C vacancy resulting in a ketonic and a ether group, and (D) two O atoms in a C divacancy, resulting in two ether groups. Small gray circles represent the O atoms, while the other circles represent the surface C atoms. Their color indicates the corresponding calculated C 1s shift.

and the other at lower BE, 531.4 eV (Figure 2, B and C), can be associated to ether and carbonyl (ketonic) C-O groups, respectively, forming in the presence of surface C vacancies. In the case of defective graphite surfaces, there is a preferential binding of the O atoms to the C vacancies. We have simulated the interaction of atomic O with the defective samples by considering three model cases, i.e., one O atom in single C vacancy, two O atoms in single C vacancy, and two O atoms in a double C vacancy. The corresponding lowest energy configurations are displayed in the insets of Figure 3, B, C, and D. These simulations allow us to conclude that the initial oxidation of the defective HOPG surfaces involves attachment of a pair of O atoms to single and double C vacancies, yielding C-O-C ether and CdO ketonic functional groups. The ether group is oriented in the plane of the C network, while the ketonic quinone points outward the surface with an angle of 32.11° with respect to the normal (inset in Figure 3C). The calculations for the (7 × 7) supercell predict that the O 1s components of the two surface groups are shifted in opposite directions with respect to the epoxy O 1s BE, the ether toward higher BE (+2.38 eV) and the ketonic toward lower BE by (-0.74 eV). We note that the presence of epoxy groups neighboring the ketonic and ether

9012

J. Phys. Chem. C, Vol. 113, No. 21, 2009

groups is predicted to be energetically favored by more than 0.6 eV, and the O 1s BE shifts calculated for this configuration (+1.85 and -0.87 eV) are closer to the experimental values. Thus we attribute the components at 533.5 and at 531.4 eV to the ether and ketonic groups, respectively. These assignments are in fair agreement with the near edge X-ray absorption fine structure spectroscopy (NEXAFS) and XPS studies of graphite subjected to high energy O ions where the structure at ∼533 eV was attributed to ether and ∼531 eV to quinone, i.e., O with carbonyl type bond.28 Finally, thermodynamic arguments rule out single vacancies saturated by single O atoms. The equilibrium structure predicted for this case show that the O atom should occupy a substitutional site of the C network (even when starting from highly asymmetric configurations) and display a 3-fold planar coordination with the neighboring C atoms. Adding another O atom to this configuration from the gas phase or from a neighboring surface epoxy group lowers the total energy by 5.7 and 3.95 eV, respectively, and results in the C-O-C ether and CdO carbonyl (ketone) groups, described above. Moreover, an O atom from an epoxy group neighboring the substitutional O atom in the graphene surface can diffuse into the vacancy and form the ether/ketone pair described above with a calculated activation energy of 0.7 eV. With respect to the epoxy configuration on the defect-free regions of the sample, the O 1s shift calculated for this metastable 3-fold O atom is very large (>4 eV), and no relevant peak is detected in that region of the experimental spectrum. However, we cannot rule out the existence of such configurations at earlier stages of oxidation (very low O coverage) that are below the photoemission detection limit. The peak assignments were also supported by the observed variations in the intensity ratio of the ether to ketone components with the probing depth of photoelectrons, which decreases by changing the acceptance angle of the electron analyzer from normal emission to grazing. These measurements showed that with decreasing the photoelectrons probing depth (grazing angle photoemission) the intensity ratio between ether and ketone O 1s components decreases. This is expected since the ether oxygen groups are positioned almost in the graphene surface plane, while the ketone O is above the surface. Close inspection of the O 1s spectra in Figure 2 reveals that for such small exposures the ether component is a factor of 2 larger than the ketone one, in agreement with the theoretical prediction that the C-O-C species are present in both single and double vacancies, while the CdO groups are present only in the case of single vacancies. The conclusions drawn on the basis of the O 1s spectra are supported by the corresponding C 1s spectra, presented in Figure 4. The adsorption of O on perfect HOPG leads to the appearance of two new features in the C 1s BE spectrum (Figure 4 A). They can be deconvoluted with two additional components with respect to the C 1s BE of the clean HOPG, one at larger BE, 286.3 eV (+1.8 eV relative shift), and one at smaller BE, 284.2 eV (-0.3 eV relative BE shift). The width of the latter has to be slightly increased to account for the nearest and next nearest neighbors to the O bonded C atoms. The C 1s BEs calculated for an epoxy group in a (7 × 7) supercell with respect to a C atom in nondefective graphite (Figure 3A) are in good agreement with the photoemission spectra yielding a large positive chemical shift of +1.6 eV and a negative shift of -0.4 eV. The former is associated with the two C atoms bound to O forming the epoxy group, while the latter with the four C atoms second neighbors to the O adatom. The remaining atoms also experience smaller BE relative shifts, in accordance with broadening of

Letters

Figure 4. Experimental C 1s spectra corresponding to the initial oxidation of “perfect” (A), weakly (B), and strongly (C) defective HOPG samples (same as reported in Figure 2). O dose is 4 × 1014 atoms cm-2 at 20 °C, the same as in Figure 2.

low BE component in photoemission spectrum, but the shift rapidly drops to 0 for C atoms beyond the third nearest neighbor. At higher O coverage, yielding second-neighboring epoxy groups, the calculated peak positions of C atoms directly bound to O shift to higher BE (from 1.6 to 1.8 eV), in even better agreement with the experimental data. The initial oxidation of defective HOPG samples yields more structured C 1s photoemission spectra, shown in Figure 4, B and C. The deconvolution of these spectra requires many closely spaced components to account for the surface functional groups that are formed upon O adsorption, which we identify on the basis of our DFT calculations. Overall, the experimental C 1s spectra of the oxidized defective HOPG samples show the appearance of new features at large positive BE and clear shoulders at small negative and positive BE with respect to the main component of perfect graphite set to 0 eV in the Figure 3. In addition to the components identified for the epoxy surface groups, the deconvolution of the experimental C 1s spectra of the oxidized defective HOPG sample (Figure 4 B) at BEs more positive with respect to main C1s HOPG peak require the inclusion of new components at 285.9 eV (+1.4 eV relative shift) in excellent agreement with the C 1s BE for the ethers29 and at 284.9 eV (+0.4 eV relative shift). The region of negative BE side of the spectrum is clearly formed by many closely spaced components due to the different local environments present in these defective samples. Since we cannot resolve so many peaks within relatively narrow energy range we used only two broad components: a necessary minimum to deconvolute the line shape of oxidized defective HOPG samples. These components are rationalized on the basis of the simulated C 1s BE for the C atoms in the supercells modeling

Letters the different O binding configurations. The results, reported in Figure 3, B-D, are in overall good agreement with the experimental spectra. A large positive chemical shift is predicted for the 1s levels of the C atoms neighboring the O atoms in the single vacancy (+1.4 eV shift for both the ether and ketonic groups) and in the divacancy (+1.2 eV shift for the ether group). The surface C atoms second neighbors to the O atom in the ketonic group display instead a negative core-level shift of -0.6 eV, thus larger than the negative component present in the epoxy group (-0.4 eV), while the C 1s level of the atoms second neighbor to the ether O atoms shift to small positive (+0.2 eV) BE. According to these calculations, the presence of O atoms in single and double C vacancies of defective HOPG result in an overall broadening of the reference C 1s peak, in a new component with BE shifts of ∼-0.6 eV, and in new components at large BE close to the epoxy feature but with smaller positive shifts (1.2-1.4 eV). This agreement with the experimental data allows us to associate the majority of the O atoms oxidizing the defective HOPG surfaces to the formation of ketonic and ether functional groups. In summary, by combining the high resolution photoemission spectroscopy with density functional theory calculations we have confirmed the formation of epoxy groups in the initial stages of “perfect” HOPG oxidation. This conclusion is in good agreement with recent experimental evidence based on electron energy loss spectroscopy of the C and O K edges.30 Carbon vacancies present on defective samples at these oxidation stages promote the formation of ether and carbonyl (ketone) surface functional groups. The reported results are the starting point of an ambitious program to unravel complex mechanisms of oxidation occurring in the advanced stages and/or at high temperature, when the formation of oxygen functional groups yields also gasification reactions with volatile CO and CO2 products. Advances into the atomic level insight of these processes have recently appeared in the literature.31 ¨ stu¨nel for preparing the Acknowledgment. We thank H. U pseudopotentials used in the simulations of the core-level binding energy shift, as well as for critical reading of the manuscript. References and Notes (1) Ebbesen, T. W.; Ajayan, P. M.; Hiura, H.; Tanigaki, K. Nature 1994, 367, 519. (2) Tsang, S. C.; Harris, P. F. J.; Green, M. L. H. Nature 1993, 362, 520–522. (3) Ajayan, P. M.; Ebbesen, T. W.; Ichihashi, T.; Iijima, S.; Tanigaki, K.; Hiura, H. Nature 1993, 362, 522–525.

J. Phys. Chem. C, Vol. 113, No. 21, 2009 9013 (4) Lee, S. M.; Lee, Y. H.; Hwang, Y. G.; Hahn, J. R.; Kang, H. Phys. ReV. Lett. 1999, 82, 217–220. (5) Cui, J. B.; Sordan, R.; Burghard, M.; Kern, K. Appl. Phys. Lett. 2002, 81, 3260–3262. (6) Liu, L.; Ryu, S.; Tomasik, M. R.; Stolyarova, E.; Jung, N.; Hybertsen, M.; Steigerwald, M. L.; Brus, L. E.; Flynn, G. W. Nano Lett. 2008, 8, 1965–1970. (7) Schniepp, H. C.; Li, J-.L.; McAllister, M. J.; Sai, H.; HerreraAlonso, M.; Adamson, D. H.; Prud’homme, R. K.; Car, R.; Saville, D. A.; Aksay, I. A. J. Phys. Chem B 2006, 110, 8535–8539. (8) Ionescu, R.; Espinosa, E. H.; Sotter, E.; Llobet, E.; Vilanova, X.; Correig, X.; Felten, A.; Bittencourt, C.; Van Lier, G.; Charlier, J.-C.; Pireaux, J. J. Sens. Actuators B 2006, 113, 36–46. (9) McAllister, M. J.; Li, J.-L.; Adamson, D. H.; Schniepp, H. C.; Abdala, A. A.; Lui, J.; Herrera-Alonso, M.; Milius, D. L.; Car, R.; Prud’homme, R. K.; Aksay, I. A. Chem. Mater. 2007, 19, 4396–4404. (10) Jeong, H.-K.; Lee, Y. P.; Lahaye, R. J. W. E.; Park, M.-H.; An, K. H.; Kim, I. J.; Yang, Ch.-W.; Park, Ch. Y.; Ruoff, R. S.; Lee, Y. H. J. Am. Chem. Soc. 2008, 130, 1362–1366. (11) Hashimoto, A.; Suenaga, K.; Gloter, A.; Urita, K.; Iijima, S. Nature 2004, 430, 870–873. (12) Ishigami, M.; Choi, H. J.; Aloni, S.; Loui, S. G.; Cohen, M. L.; Zettl, A. Phys. ReV. Lett. 2004, 93, 196803. ¨ stu¨nel, H.; Fabris, S.; Gregoratti, L.; Aballe, L.; (13) Barinov, A.; U Dudin, P.; Baroni, S.; Kiskinova, M. Phys. ReV. Lett. 2007, 99, 046803. (14) Brukh, R.; Mitra, S. J. Mater. Chem. 2007, 17, 619–623. (15) Raghuveer, M. S.; Kumar, A.; Frederick, M. J.; Louie, G. P.; Ganasan, P. G.; Ramanath, G. AdV. Mater. 2006, 18, 547–552. (16) Zhang, J.; Liu, X.; Blume, R.; Zhang, A.; Schlogl, R.; Su, D. S. Science 2008, 322, 73–77. (17) Li, J.-L.; Kudin, K. N.; McAllister, M. J.; Prud’homme, R. K.; Aksay, I. A.; Car, R. Phys. ReV. Lett. 2006, 96, 176101. (18) Prince, K. C.; Ulrich, I.; Peloi, M.; Ressel, B.; Cha´b, V.; Crotti, C.; Comicioli, C. Phys. ReV. B 2000, 62, 6866–6868. (19) Hahn, J. R.; Kang, H.; Song, S.; Jeon, I. C. Phys. ReV. B 1996, 53, R1725-R1728. (20) http://www.tectra.de/plasma-source.htm/. (21) Perdew, J. P.; Burke, K.; Ernzerhof, M. Phys. ReV. Lett. 1996, 77, 3865–3868. We have used pseudopotentials C.pbe-rrkjus and O.pbe-rrkjus, prepared by Andrea Dal Corso from http://www.quantum-espresso.org. (22) Quantum Espresso is a community project for high-quality quantumsimulation software, based on density-functional theory, and coordinated by Paolo Giannozzi. See http://www.quantum-espresso.org, and http:// www.pwscf.org. (23) Wirz, L.; Rubio, A. Solid State Commun. 2004, 131, 141–152. (24) Pehlke, E.; Scheffler, M. Phys. ReV. Lett. 1993, 71, 2338–2341. (25) El-Barbary, A. A.; Telling, R. H.; Ewels, C. P.; Heggie, M. I.; Briddon, P. R. Phys. ReV. B 2003, 68, 144107. (26) Dı´az, J.; Paolicelli, G.; Ferrer, S.; Comin, F. Phys. ReV. B 1996, 54, 8064–8068. (27) Sorescu, D. J.; Jordan, K. D. J. Phys. Chem. B 2001, 105, 11227– 11232. (28) Shimoyama, I.; Sekiguchi, T.; Baba, Y. Photon Factory ActiVity Rep. 2001, 22. (29) Jordan, J. L.; Sanda, P. N.; Morar, J. F.; Kovac, C. A.; Himpsel, F. J.; Pollack, R. A. J. Vac. Sci. Technol. A 1986, 4, 1046–1048. (30) Mkhoyan, K. A.; Contryman, A. W.; Silcox, A,; Stewart D. A.; Mattevi, C.; Miller, S.; Chhowalla, M. Nano Lett. 2009, article ASAP. (31) Paci, J. T.; Upadhyaya, H. P.; Zhang, J.; Schatz, G. C.; Minton, T. K. J. Phys. Chem. A. article ASAP.

JP902051D