Letter pubs.acs.org/NanoLett
Evidence for Ambient-Temperature Reversible Catalytic Hydrogenation in Pt-doped Carbons Xiao Ming Liu,† Youjian Tang,‡ En Shi Xu,‡ Thomas C. Fitzgibbons,§ Gregory S. Larsen,† Humberto R. Gutierrez,‡,∇ Huan-Hsiung Tseng,# Ming-Sheng Yu,# Cheng-Si Tsao,# John V. Badding,§ Vincent H. Crespi,‡,∥ and Angela D. Lueking*,†,⊥ †
Materials Research Laboratory, ‡Department of Physics, §Department of Chemistry, ∥Department of Materials Science and Engineering and Materials Research Institute, and ⊥Department of Energy and Mineral Engineering, Department of Chemical Engineering, and EMS Energy Institute, The Pennsylvania State University, University Park, Pennsylvania 16802, United States # Institute of Nuclear Energy Research, Longtan, Taoyuan 32546, Taiwan S Supporting Information *
ABSTRACT: In situ high-pressure Raman spectroscopy, with corroborating density functional calculations, is used to probe C− H chemical bonds formed when dissociated hydrogen diffuses from a platinum nanocatalyst to three distinct graphenic surfaces. At ambient temperature, hydrogenation and dehydrogenation are reversible in the combined presence of an active catalyst and oxygen heteroatoms. Hydrogenation apparently occurs through surface diffusion in a chemisorbed state, while dehydrogenation requires diffusion of the chemisorbed species back to an active catalyst.
KEYWORDS: Graphene, graphane, Pt nanoparticles, hydrogen, spillover, catalysis
T
defects, and heteroatoms may lead to high surface mobility and reversible carbon−hydrogen interactions. Three carbon supports are used: highly ordered graphene, disordered and curved activated carbon (AC), and AC with enriched oxygen heteroatoms (ACox). The high surface area (1886 m2/g) AC support is a network of defective graphenic sheets, typically 3−5 layers thick and curved on a scale of a few nanometers (Figure 1a). Acid oxidation treatment increases the oxygen content from 1 wt % (in AC) to 9 wt % (in ACox), leading to a number of heterogeneous oxygen functional groups. After impregnation with 0.79% by weight Pt, these terminal oxygen groups anchor well-dispersed ∼2 nm diameter particles (Figure 1a). No atomic-sized Pt species were found by high-angle annular darkfield scanning transmission electron microscopy (probe size 2 Å, Figure 1b). Surface oxygen groups are expected to stabilize and anchor deposited Pt as well as influence π-complexation between Pt and the carbon basal plane,14 and this is supported by a mode at 2050 cm−1 in FTIR for Pt/ACox samples (Figure S1 of the Supporting Information) which has been associated with Pt-CO complexes.15 Pt/ACox chemisorbs 2 orders of magnitude more hydrogen (1 wt % at ∼0.67 bar13) than can be
ransition metal catalysts are often dispersed on supports to maximize the catalytically active surface area while minimizing the use of the transition metal, which is usually an expensive precious metal. Often, a reactive species may “spillover” (diffuse) from the catalyst to its support, effectively extending the surface area involved in catalysis.1 Recent studies suggest that catalyst−support interactions can significantly influence the energetics and/or kinetics of hydrogen interactions with both catalysts and supports.2,3 Similar interactions may play a role in reversible adsorption of significant amounts of hydrogen to carbon supports via spillover;4,5 collective stabilization of adjacent chemisorbed H atoms provides a thermodynamic driving force for surface hydrogenation.6−8 Several characterization techniques have provided macroscopic evidence for hydrogen spillover to carbon supports,9−13 but these ex situ and/or low temperature techniques have not conclusively identified the nature of the carbon−hydrogen bond, and the mechanism by which (and degree to which) hydrogen diffuses remains contested. Vibrational spectroscopies, such as Raman spectroscopy, are powerful direct probes of chemical bonding and allow for direct in situ characterization under relevant conditions of pressure (e.g., >0.1 MPa) and temperature (e.g., ∼300 K). Here, we use Raman spectroscopy to probe the nature of the carbon− hydrogen bond in the presence of Pt nanoparticles varying the nature of the carbon support, to elucidate how curvature, © 2012 American Chemical Society
Received: October 3, 2012 Revised: November 26, 2012 Published: November 29, 2012 137
dx.doi.org/10.1021/nl303673z | Nano Lett. 2013, 13, 137−141
Nano Letters
Letter
attributed to the catalyst or support separately. Inelastic neutron scattering reveals that the intensity of the H2 molecular rotor is reduced after the sample is heated to induce spillover.13 Raman spectra of Pt/ACox in ∼100 bar of He at 298 K are characteristic of a sp2-bonded carbonaceous solid, with D (1353 cm−1) and G modes (1599 cm−1) (Figure 1c, He−I). Upon introduction of 100 bar H2, a new Raman mode (Figure 1c, H2−I) arises at 1184 cm−1, which is assigned to the wagging mode of a covalent C−H bond (see the Supporting Information for detailed discussion and consideration of a C−H stretch mode at frequencies above 2500 cm−1). The area of the CH wag is ∼1/seventh that of the G mode, which redshifts to 1590 cm−1 upon exposure to H2. The D mode and its associated second-order features (Figure S2) also redshift in H2. The shift in the D mode can be attributed to a change in the electronic structure and/or disorder that occurs upon hydrogenation of the basal plane of ACox. Room-temperature evacuation of H2 for 72 h reverses these changes: the mode identified with the CH wag nearly disappears, and the D mode shifts back close to its original position in He (Figure 1c, He− II). This pattern of appearance/disappearance of the CH wag and red/blue shift of the D mode is repeated for several cycles (Figure S3), suggesting that hydrogenation of the ACox support is almost entirely reversible. In contrast to the reversible shifts in the D mode, the initial red-shift in the G mode is irreversible (Table S1), which suggests the sp2 carbon is less doped.16 The shift in the doping level is likely brought about by hydrogenation of terminal oxygen groups, which irreversibly bind hydrogen.17 Studies of graphane showed a similar (but unassigned) feature at ∼1180 cm−1 with no shift in the G mode position after annealing in argon to 723 K18 (see also Supporting Information). If the new mode at 1184 cm−1 is really associated with hydrogen motion, then it should downshift by a factor of ∼1.4 when H is replaced by D due to the isotopic shift (see Supporting Information). Exposure of Pt/ACox to ∼30 bar D2 after evacuation of hydrogen downshifts the wag mode to 830 cm−1 (Figure 1d, D2−I), which is very close to the factor of 1.4 expected for an ideal pure-hydrogen vibrational mode. Deuteration is also reversible (Figure S3). C:H (or C:D) thin films produced by hydrocarbon ion beam deposition onto Pt in ultrahigh vacuum also show a mode at 1186 (846) cm−1 which has been assigned to a C−H (C−D) wag.19 The reversibility of hydrogenation at ambient temperature places strict constraints on bond energetics. The H−H and H− C bond energetics must be sufficiently finely balanced that this process occurs readily both forward and reverse.7 We performed density functional calculations for hydrogen atoms bound to one side of planar sp2 carbon in a range of hydrogenation patterns (Table 1). One-sided attachment models a Pt nanoparticle that presumably can access only one side of the sheet. The simplest such geometry is an isolated chemisorbed H atom within a 2 × 2 or 3 × 3 graphene supercell. These C8H (C18H) systems yield a binding energy of 0.67 (0.68) eV relative to a gas-phase H atom and bare graphene, in agreement with previous results, such as a coronene model.20 Pairing the H removes the energetically unfavorable unpaired electron, so that C 8 H 2 (of C 6 V symmetry) with H atoms located at the para (1,4) positions of a hexagonal ring binds hydrogen at 1.71 eV per H. C8H4 (in a 2 × 2 supercell) has C2V symmetry and an adsorption energy of 1.53 eV per H. Chemisorbed hydrogen clusters form as in previous reports.6,7,21
Figure 1. Characterization of Pt supported on oxidized activated carbon (Pt/ACox). (A) Transmission electron microscopy and (B) high-angle annular dark-field scanning images showing size and distribution of Pt nanoparticles as well as carbon morphology. (C) Raman spectra at 514.4 nm, 0.2 mW showing the carbon D (∼1350 cm−1) and G bands (∼1590 cm−1) at 298 K in a capillary exposed sequentially to He, H2, and He. A new C−H mode appears reversibly at 1184 cm−1 under hydrogen exposure. Relative area and Lorentzian fitting parameters are shown. (D) Raman spectra (enlarged 3 × ) of a second H2 exposure, a D2 exposure exhibiting a C−D wagging mode at 830 cm−1 in D2, and a CO poisoning that eliminates the reversibility of this feature, with the CH wag remaining after a 72 h degas at 298 K. The sharp features at 817 cm−1 and 1037 cm−1 in the H2 spectra are rotational transitions of free H2. The remaining features at 753 and 859 cm−1 are the free D2 rotational transitions S(5) and S(6). In each case, samples were initially evacuated at 498 K at 10−7 Torr for 72 h, after which ∼100 bar of He or H2 or ∼2 bar CO was introduced. Prior to each subsequent gas exposure, samples were evacuated at 298 K for 72 h, with the exception of CO (see Methods in the Supporting Information). 138
dx.doi.org/10.1021/nl303673z | Nano Lett. 2013, 13, 137−141
Nano Letters
Letter
Table 1. Adsorption Energies and Wagging Mode Frequencies of Chemisorbed H on Graphene
a
Red dots indicate chemisorption sites for H. Additional structures reported in Table S2. bAdsorption energy is calculated relative to the total energies of gas-phase H atoms and bare graphene. cDouble degenerate (*) and calculated for deuterium (D). dRatio of wag mode intensity to Gmode intensity.
Vibrational spectra at the Γ point were calculated for each of these structures, with Raman active modes identified and mode intensities calculated by linear response theory. A C−H wagging mode is found whose frequency tends to increase with increasing hydrogen coverage. For example, the C8H2 state of C6V symmetry yields two degenerate Raman-active E2 wagging modes at 1144 cm−1 for H and 909 cm−1 for D. In density functional perturbation theory (DFPT), the intensity of this mode is about 1/12 that of the G band, comparable to the 1/seventh intensity observed experimentally for Pt/ACox and the 1/11th intensity observed for Pt/graphene, as described below. The high-density C4H2 pattern has two nondegenerate A1 wagging modes at 1300 cm−1 and 1164 cm−1 (973 cm−1 and 860 cm−1 for deuterium) with an intensity about 1/ninth that of the G band. Both the frequencies and the relative Raman intensities of these modes are consistent (within the limits of computational accuracy and our limited knowledge of the precise hydrogenation pattern) with the identification of the experimentally observed mode at 1184 cm−1 as a vibration of atomic hydrogen bound directly to the carbon basal plane. Exposure of Pt/ACox to CO (after two cycles of reversible hydrogenation) poisons the Pt catalyst and eliminates reversibility: The 1184 cm−1 CH wag remains even after 72 h of evacuation (Figure 1d, H2 + CO). CO does not significantly alter the electronic structure of the carbon surface, as no shift in the G mode is observed (Table S1). This observation suggests the H atom must travel back to the catalyst to recombine and desorb, supporting a reverse-spillover mechanism for dehydrogenation.1 Graphene is planar and atomically pure relative to ACox, allowing us to probe the influence of carbon morphology on spillover. Graphene generally lacks the oxygen groups that anchor Pt nanoparticles, and one would thus expect higher mobility of Pt on a graphene surface. However, the delocalized electrons of graphene/graphite do participate in π-complexation interactions in which π-electrons act as a ligand for the Pt, donating electrons to an unfilled orbital in Pt; Pt in turn backfills unfilled electrons to the LUMO of the carbon.14 With these weakened interactions, Pt particles agglomerate on graphene, forming larger clusters 2−20 nm in diameter (Figure 2a, inset). Despite this distinction, Pt/graphene also exhibits a
Figure 2. Analysis of Pt/C with variations in carbon morphology and catalyst dispersion. (A) Pt/graphene has 3−5 layers of planar graphene, with (inset) large clusters of 2−20 nm Pt particles. (B) Pt/AC shows curved carbon structures and Pt particles of 3−5 nm, which tend to cluster at the periphery of the carbon particle (inset). For both materials, a Raman mode at 1181 cm−1 (Lorentzian fits found in Table S1) appears in H2 but does not disappear with evacuation at 298 K (corresponding right panel). Gas was introduced as described in Figure 1.
similar mode at 1181 cm−1 when exposed to H2 (Figure 2a, H2−I) suggesting that the differences in catalyst particle size and carbon morphology do not prevent diffusion of H from the catalyst to the carbon support. However, dehydrogenation is not observed from Pt/graphene (Figure 2a, He−II), even when the sample is heated to 673 K under vacuum. This irreversibility is qualitatively consistent with previous studies that demon139
dx.doi.org/10.1021/nl303673z | Nano Lett. 2013, 13, 137−141
Nano Letters
Letter
mates21,22 of this activation energy vary widely, from 0.3 to 1.25 eV, they cannot confirm or exclude the possibility of room-temperature diffusion of chemisorbed hydrogen. However, the hydrogen adsorption to Pt/AC is 10-fold reduced relative to Pt/ACox (Figure S4), suggesting that significant surface mobility occurs only for the latter. Hole-doping may reduce the activation energy for H diffusion;25 thus the influence of the oxygen groups on the carbon doping level (i.e., the observed G mode shift) might be a mechanism to facilitate diffusion of chemisorbed H species on a carbon support. Dehydrogenation, that is, reverse mobility, is only seen for Pt/ACox. The CO poisoning studies suggest that H must return to an active Pt catalyst to recombine and desorb. Oxygen groups apparently then also account for increased mobility back to the catalyst, suggesting that the Pt−oxygen−carbon interaction is a key to reversible hydrogenation. One possibility is that oxygen groups alter the level of electron donation to/ from the catalyst to the carbon surface.14 To summarize, three graphenic surfaces were hydrogenated at ambient temperature via hydrogen spillover from a Pt catalyst, evidenced by in situ Raman spectroscopy which revealed a new mode identified with the formation of a C−H chemical bond. These results demonstrate that synthesis of a variant of graphane is possible at mild conditions in the presence of a catalyst and have helped to clarify the mechanism for hydrogen spillover from a supported catalyst to a carbon support. Forward mobility, that is, surface diffusion from the catalyst to its support, apparently occurs via a mobile chemisorbed phase, while reverse mobility occurs by diffusion back to the catalyst. Significant forward and reverse mobility are achieved only when the carbon is oxidized, with consequent changes, for example, in the doping level of the carbon support. Heteroatoms such as oxygen influence metal dispersion, mobility, and the catalyst−support interface. Carbon surface chemistry is thus a key factor in molecular design of a catalyst. These results clarify a strategy to achieve significant hydrogenation and reversible dehydrogenation of a graphenic carbon support via hydrogen spillover, while also suggesting routes to design improved catalysts for hydrogenation and fuel cell applications.
strate that desorption from plasma-deuterated graphite requires high temperatures, that is, greater than 490 K.22 The increased thermal stability in our samples may be due to a higher H surface coverage being associated with stronger hydrogen binding, as discussed earlier.21 No shifts in the D or G modes are observed for Pt/graphene under hydrogen exposure (Table S1), suggesting that the extent of hydrogenation was substantially less than that for Pt/ACox and insufficient to change the doping level significantly. Although limited hydrogenation does occur to chemically unmodified planar graphene, these samples do not desorb chemically bound hydrogen at room temperature, even in the proximity of a Pt catalyst. Oxygen functional groups have been previously implicated in the high hydrogen uptake in Pt/ACox materials.23 Omission of the oxidation step in preparation of Pt/AC leads to moderate clustering of the catalyst without significant agglomeration (Figure 2b), suggesting weaker anchoring of the Pt in the absence of oxygen groups. Pt/AC is not completely devoid of oxygen functional groups, and a feature at 2050 cm−1 in FTIR associated with Pt−CO complexes15 is still present. The reduced oxygen content can explain the clustering of the Pt nanoparticles. When submitted to the same He/H2/He cycle, the Raman behavior of Pt/AC closely resembles that for Pt/ graphene, including a new mode at ∼1181 cm−1 identified with the CH wag, a similar G peak frequency (Table S1), and irreversibility of hydrogenation (Figure 2b, H2−I and He−II). Morphological differences between flat-sheet (i.e., Pt/graphene) and more disordered, curved carbon systems (i.e., Pt/ AC) do not contribute to ready dehydrogenation. The transfer of hydrogen from the catalyst to the carbon support requires collective stabilization of multiple nearby hydrogen atoms to produce a thermodynamic driving force for hydrogen migration.6,7 The new mode identified with the C−H wag is seen in all three carbon materials examined, demonstrating ubiquitous forward mobility, irrespective of curvature, heteroatoms, or catalyst dispersion. Plasmonic surface enhancement and electromagnetic effects are expected to be weak for Pt nanoparticles (see Supporting Information); thus the Raman measurements likely do not exclusively favor carbon regions in the immediate vicinity of the Pt and instead provide a more representative sampling of the carbon surface. Considering the dispersion of the Pt catalyst, some hydrogen atoms must travel at least several nanometers from their source Pt nanoparticles to populate the carbon surface sufficiently to account for the observed 1 wt % hydrogen adsorption for Pt/ACox13 (see also Supporting Information). The ejection of atomic H from the surface to a transient weakly bound, mobile physisorbed state has previously been proposed as a means to populate a support via spillover. However, the probability of ejection to a mobile physisorbed state should be similar for all three carbon materials, if not reduced in the presence of oxygen, which strongly binds H and increases the C−H binding on nearby atoms.17 (In addition, such a physisorbed atomic state would be thermodynamically unfavorable and hence would require appeal to a finely tuned quasi-ballistic nonequilibrium process to obtain significant prevalence). Ejection to a mobile H state is thus inconsistent with our experimental results, and instead one infers that population of the surface occurs via diffusion of chemisorbed species, which would require an activation energy for surface diffusion of less than ∼0.7 eV (see Supporting Information). As theoretical20,24,25 and experimental esti-
■
ASSOCIATED CONTENT
S Supporting Information *
Experimental procedures, assignment of the 1184 cm−1 mode to a CH wag, consideration of a high frequency CH stretch, explanation of C−D downshift, consideration of surface enhancement Raman spectroscopy and electromagnetic effects, diffusion distance, activation energy of diffusion chemisorbed H, Figures S1−S10, and Tables S1−S3. This material is available free of charge via the Internet at http://pubs.acs.org.
■
AUTHOR INFORMATION
Corresponding Author
*E-mail:
[email protected]. Present Address ∇
Department of Physics and Astronomy, University of Louisville, Louisville, Kentucky 40292, United States.
Notes
The authors declare no competing financial interest. 140
dx.doi.org/10.1021/nl303673z | Nano Lett. 2013, 13, 137−141
Nano Letters
■
Letter
(16) Das, A.; Pisana, S.; Chakraborty, B.; Piscanec, S.; Saha, S. K.; Waghmare, U. V.; Novoselov, K. S.; Krishnamurthy, H. R.; Geim, A. K.; Ferrari, A. C.; Sood, A. K. Nat. Nanotechnol. 2008, 3, 210−215. (17) Psofogiannakis, G. M.; Froudakis, G. E. J. Am. Chem. Soc. 2009, 131, 15133−15135. (18) Elias, D. C.; Nair, R. R.; Mohiuddin, T. M. G.; Morozov, S. V.; Blake, P.; Halsall, M. P.; Ferrari, A. C.; Boukhvalov, D. W.; Katsnelson, M. I.; Geim, A. K.; Novoselov, K. S. Science 2009, 323, 610−613. (19) Biener, J.; Chenk, A.; Winder, B.; Schubert, U. A.; Lutterloh, C.; Kupers, J. Phys. Rev. B 1994, 49, 17307−17318. (20) Psofogiannakis, G. M.; Froudakis, G. E. J. Phys. Chem. C 2009, 113, 14908−14915. (21) Zecho, T.; Güttler, A.; Sha, X.; Jackson, B.; Küppers, J. J. Chem. Phys. 2002, 117, 8486−8492. (22) Hornekær, L.; Šljivančanin, Ž .; Xu, W.; Otero, R.; Rauls, E.; Stensgaard, I.; Lægsgaard, E.; Hammer, B.; Besenbacher, F. Phys. Rev. Lett. 2006, 96, 156104. (23) Li, Q.; Lueking, A. D. J. Phys. Chem. C 2011, 115, 4273−4282. (24) Boukhvalov, D. W.; Katsnelson, M. I. J. Phys. D: Appl. Phys. 2010, 43, 175302. (25) Huang, L. F.; Ni, M. Y.; Zhang, G. R.; Zhou, W. H.; Li, Y. G.; Zheng, X. H.; Zeng, Z. J. Chem. Phys. 2011, 135, 064705.
ACKNOWLEDGMENTS We thank Cheng-Yu Wang for collecting the hydrogen adsorption isotherms of Pt/AC and estimation of adsorption time for various energy barriers (see Supporting Information). This work was supported by the U.S. Department of Energy, Basic Energy Sciences. The Raman, TEM, and DFT aspect of the work was supported under the Single Investigator and Small-Group Research (SISGR) program, Awards DE-FG0209ER466556 and DE-SC0002157. T.C.F. (providing Raman support and FTIR and Raman analysis) and J.V.B. were additionally supported as part of Energy Frontier Research in Extreme Environments Center (EFree), an Energy Frontier Research Center under Award No. DE-SC0001057. The adsorption isotherms referenced in this work were supported by the Energy Efficiency and Renewable Energy program, Award DE-FG36-08GO18139. Author contributions to this work are as follows: X.M.L. conducted Raman analysis, prepared figures, analyzed and interpreted these results, and wrote initial drafts of the paper. Y.T. conducted DFT analysis of hydrogenated carbon structures. E.S.X. performed DFT analysis of oxygenated structures. T.C.F. assisted in Raman setup and conducted FTIR analysis. G.L. performed the CO poisoning experiments. H.R.G. conducted the HRTEM and HAADF analysis. H.H.T. and M.S.Y. synthesized the Pt/AC and Pt/ ACox samples, overseen by C.S.T. J.V.B. supervised the Raman analysis and designed experiments. V.H.C. supervised the computational studies. A.D.L. supervised the adsorption isotherms, finalized data interpretation in the paper, and wrote final drafts of the paper, edited by J.V.B. and V.H.C. All PSU authors discussed the results and commented on the paper, which were communicated to C.S.T, H.H.T., and M.S.Y. via e-mail drafts of the paper.
■
REFERENCES
(1) Conner, W. C.; Falconer, J. L. Chem. Rev. 1995, 95, 759−788. (2) Kyriakou, G.; Boucher, M. B.; Jewell, A. D.; Lewis, E. A.; Lawton, T. J.; Baber, A. E.; Tierney, H. L.; Flytzani-Stephanopoulos, M.; Sykes, E. C. H. Science 2012, 335, 1209−12. (3) Ng, M. L.; Balog, R.; Hornekær, L.; Preobrajenski, A. B.; Vinogradov, N. a.; Mårtensson, N.; Schulte, K. J. Phys. Chem. C 2010, 114, 18559−18565. (4) Wang, L.; Yang, R. T. Energy Environ. Sci. 2008, 1, 268−279. (5) Cheng, H.; Chen, L.; Cooper, A. C.; Sha, X.; Pez, G. P. Energy Environ. Sci. 2008, 1, 338−354. (6) Stojkovic, D.; Zhang, P.; Lammert, P.; Crespi, V. Phys. Rev. B 2003, 68, 195406. (7) Lin, Y.; Ding, F.; Yakobson, B. I. Phys. Rev. B 2008, 78, 041402. (8) Singh, A. K.; Ribas, M. A.; Yakobson, B. I. ACS Nano 2009, 3, 1657−1662. (9) Mitchell, P. C. H.; Ramirez-Cuesta, A. J.; Parker, S. F.; Tomkinson, J.; Thompsett, D.; Matthey, J.; Centre, T.; Common, S.; Rg, R. J. Phys. Chem. B 2003, 107, 6838−6845. (10) Contescu, C. I.; Brown, C. M.; Liu, Y.; Bhat, V. V.; Gallego, N. C. J. Phys. Chem. C 2009, 113, 5886−5890. (11) Lin, C.; Yang, Z.; Xu, T.; Zhao, Y. Appl. Phys. Lett. 2008, 93, 233110. (12) Bhowmick, R.; Rajasekaran, S.; Friebel, D.; Beasley, C.; Jiao, L.; Ogasawara, H.; Dai, H.; Clemens, B.; Nilsson, A. J. Am. Chem. Soc. 2011, 133, 5580−5586. (13) Tsao, C.; Liu, Y.; Chuang, H.; Tseng, H.; Chen, T.; Chen, C.; Yu, M.; Li, Q.; Lueking, A.; Chen, S. J. Phys. Chem. Lett. 2011, 2, 2322−2325. (14) Coloma, F.; Sepulveda-Escribano, A.; Fierro, J. L. G.; RodriguezReinoso, F. Langmuir 1994, 10, 750−755. (15) Crossley, A.; King, D. A. Surf. Sci. 1977, 68, 528−538. 141
dx.doi.org/10.1021/nl303673z | Nano Lett. 2013, 13, 137−141