Article Cite This: J. Am. Chem. Soc. 2018, 140, 5102−5109
pubs.acs.org/JACS
Formation of Supported Graphene Oxide: Evidence for Enolate Species Zbynek Novotny,†,⊥ Manh-Thuong Nguyen,† Falko P. Netzer,‡ Vassiliki-Alexandra Glezakou,† Roger Rousseau,*,† and Zdenek Dohnálek*,†,§ †
Fundamental and Computational Sciences Directorate and Institute for Integrated Catalysis, Pacific Northwest National Laboratory, P.O. Box 999, Richland, Washington 99352, United States ‡ Surface and Interface Physics, Institute of Physics, Karl-Franzens University, A-8010 Graz, Austria § Voiland School of Chemical Engineering and Bioengineering, Washington State University, Pullman, Washington 99163, United States S Supporting Information *
ABSTRACT: Graphene oxides are promising materials for novel electronic devices or anchoring of the active sites for catalytic applications. Here we focus on understanding the atomic oxygen (AO) binding and mobility on different regions of graphene (Gr) on Ru(0001). Differences in the Gr/Ru lattices result in the superstructure, which offers an array of distinct adsorption sites. We employ scanning tunneling microscopy and density functional theory to map out the chemical identity and stability of prepared AO functionalities in different Gr regions. The AO diffusion is utilized to establish that in the regions that are close to the metal substrate the terminally bonded enolate groups are strongly preferred over bridge-bonded epoxy groups. No oxygen species are observed on the graphene regions that are far from the underlying Ru, indicating their low relative stability. This study provides a clear fundamental basis for understanding the local structural, electronic factors and C−Ru bond strengthening/weakening processes that affect the stability of enolate and epoxy species.
■
INTRODUCTION The engineering of the electronic band structure via doping with atomic or molecular adsorbates,1−4 essential for the use of graphene in novel electronic devices, and the tuning of the chemical reactivity by the incorporation of active chemical groups for catalytic applications5−8 are examples of the versatile ways modified graphene can be employed in novel applications. The functionalization of graphene with oxygen, previously described as graphene oxide (GO),9 has been of particular interest. GO was first synthesized from graphite by exfoliating the individual graphene sheets by Brodie10 in 1859, and the procedure was later modified by Hummers11 and is still used today.12 With modified Hummer’s method, it is possible to synthesize large-scale GO membranes with a very high degree of oxidation and a broad range of oxygen-containing functional groups.13 Membranes of GO produced in such a way show unique properties: while they are impermeable to liquids, vapors, and gases, including helium, they can be readily penetrated by water,14 and very recently GO has been used to purify seawater into drinking water.15 To control covalent functionalization of such GO membranes, selective oxidation of both graphene edges and basal planes and selective etching of GO to create edges and nanopores with unique physical and chemical properties are actively pursued.16 It is generally agreed that the chemical oxidation of graphite forms epoxide and © 2018 American Chemical Society
hydroxyl groups located on the basal plane, while carbonyl and carboxyl groups are present at the edges.5,6,17−19 While the above-mentioned functional groups are most common, other functionalities such as ether, lactone, or ketone groups near carbon vacancies were reported in studies concerning graphene etching using oxygen plasma or ozone.20 The chemistry of graphene is also strongly influenced by defects in the graphene films, as demonstrated for example for graphene zigzag edges and Stone−Wales defects, where nitrogen dopants were identified as active sites for oxygen reduction reaction.21,22 In heterogeneous catalysis, such oxygen-containing functional groups on graphitic carbon, graphene, and carbon nanotubes have been used as anchoring sites for catalytically active particles.23−27 Since the functionalities of GO all contain the carbon−oxygen bonds as a major structural element, understanding their character is of fundamental importance for the properties of GO-related materials. To date, the interaction of atomic oxygen with graphene has been studied mainly by theoretical means with the focus on the free-standing layer, and it is generally accepted that atomic oxygen forms a bridge-bonded (epoxy) configuration on defectfree basal sites of graphene.18,28−30 For graphene supported on Received: December 6, 2017 Published: February 5, 2018 5102
DOI: 10.1021/jacs.7b12791 J. Am. Chem. Soc. 2018, 140, 5102−5109
Article
Journal of the American Chemical Society
Figure 1. (a) STM image of the Gr flake on Ru(0001) obtained at 450 K before dosing atomic oxygen. The hexagonal overlay illustrates the graphene lattice that was superimposed using the current model for this surface using the Mound regions as a reference.47 (b) High-magnification atomically resolved STM image of bare Gr, indicating a large 25 over 23 coincidence unit cell46,47 highlighted by thick white lines. The unit cell is further divided into four regions that can be described by a simplified 12 over 11 model,48 (thin white lines). (b, c) Three distinct areas can be distinguished within each region: Mound, FCC, and HCP. (d) Image of the same area as in (a) after dosing 0.002 monolayers of atomic oxygen at 450 K. Dark AO-related features are observed primarily in FCC regions (dark-blue circles), with a minority of features also found in HCP regions (brown circles). (e) Zoom-in STM image of an area with four AO features, without and with the Gr lattice overlay highlighting the AO features in the FCC regions and their size with respect to the density of C atoms in the Gr lattice. In the latter case, the Gr lattice was aligned to be consistent with the interpretation put forward in this study. Images a, d, and e were acquired at 450 K (Vs = +200 mV, It = 100 pA) and image b at 300 K (Vs = +4.6 mV, It = 1 nA). The color coding of the atoms in (c): C, dark gray; surface Ru, light gray; second layer Ru, blue; third layer Ru, white.
electronic factors that affect the stability of both configurations in different Gr regions.
solid substrates, the situation is less clear and controversial. On one hand, density functional theory (DFT) calculations predicted that oxygen in a terminal, on-top carbon bonded configuration (enolate functional group) is more stable than a bridge-bonded (epoxy group) for single layer graphene (Gr) on the (111) surfaces of Cu and Ni, but no experimental studies exist to verify this prediction.31 On the other hand, the experimental studies on Gr on 6H-SiC(0001),32,33 Gr/ Ir(111),34 and Gr/Pt(111)34 reported the formation of bridge-bonded epoxy functional groups, but the possibility of the on-top bonded enolate configuration was not considered. Here, we employ scanning tunneling microscopy (STM) and provide a direct real-space mapping of atomic oxygen (AO) species binding and stability in different regions of single-layer Gr (simply Gr from here on) supported on Ru(0001). Only defect-free regions are selected in the analysis to allow for a direct comparison with the accompanying theoretical studies. A complex behavior caused by an interplay of the registry and distance between the Gr lattice and the underlying Ru(0001) surface atoms is observed. The experimental and theoretical evidence clearly demonstrates that in the Gr regions that are close to the metal substrate the terminally bonded enolate groups are strongly preferred over bridge-bonded configuration. No AO species are observed on the Gr regions that are far from the underlying Ru, indicating their low relative stability. The DFT calculations further conclude that in these regions the terminal-bonded enolate AO species are also preferred but are close in energy to the epoxy configuration. This study provides a clear fundamental basis for understanding the structural and
■
METHODS
Experimental Details. The Ru(0001) single crystal was cleaned using cycles of Ne+ sputtering and annealing in O2 at 850 K, followed by flash-annealing in ultrahigh vacuum (UHV) to 1600 K in the same experimental setup as described previously.35 Partial Gr layers (∼50% of surface covered with Gr flakes) were grown by chemical vapor deposition at 1100 K using toluene (C7H8) and decane (C10H22) as carbon precursors. The partial Gr layers were chosen because it was found that the apex of the STM tip was easier to modify to give good performance on the metal-covered areas than on Gr-covered areas. Moreover, the edges of Gr were used as reference points for thermal drift compensation during the acquisition of time-lapsed STM movies at elevated temperatures. Atomic oxygen was produced with a commercial thermal gas cracker (Mantis Ltd.) directly attached to a variable-temperature STM (Omicron). The cracker was operated at a cracking power of 75 W with 1 × 10−8 Torr O2 background pressure. The STM tip was retracted several nanometers from the surface during the dose. Both STM heating stage and thermal gas cracker were outgassed for 100+ h prior to the experiments to minimize contamination during the operation and was left running overnight prior the oxygen deposition. The atomic oxygen source was further tested on a bare Ru(0001) surface, where a well-defined (1 × 1)-O/ Ru(0001) overlayer was prepared. As shown previously, only a (2 × 1)-O/Ru(0001) overlayer can be prepared by dosing molecular O2, and atomic oxygen is required to reach the (1 × 1)-O/Ru(0001) overlayer.36 Oxygen coverage in monolayers (ML) on Gr is defined relative to the surface density of carbon atoms (1 ML ≡ 3.71 × 1015 atoms/cm2). Processing of the STM images, acquired with resolution of 512 pixels per line, included background subtraction, noise removal 5103
DOI: 10.1021/jacs.7b12791 J. Am. Chem. Soc. 2018, 140, 5102−5109
Article
Journal of the American Chemical Society
Figure 2. (a) The image shows the first frame from the time-lapsed STM movie (Supporting Information) obtained after dosing a small amount of AO at 500 K. A set of AO trajectories highlighted with blue (FCC region) and orange (HCP region) traces obtained from the STM movie over the period of 3 h (68 frames). Imaging conditions: Vs = +200 mV, It = 100 pA. (b) Possible directions and lengths of single (top) and double (bottom) hops of AO in bridge-bonded epoxy (red arrows) and terminal enolate (green arrows) configurations on Gr monolayer. from frequency domain images, and linear and nonlinear distortion removal using algorithms described in ref 37. Theoretical Calculations. All computations were conducted with the CP2K package.38 First-principles calculations were carried out using the PBE-D3 density functional,39,40 the GTH pseudopotentials,41 and the GPW hybrid basis set scheme42 in which the MOLOPT double-ζ Gaussian basis sets43 were employed to expand the valence electron states and 400 Ry cutoff for computing the electrostatic terms. The electrostatic potential along the surface normal direction was treated by the Martyna−Tuckerman approach.44 We adopted the system of (12 × 11) graphene/Ru(0001) periodicity with the Ru surface consisting of four atomic layers, of which the bottom two were fixed in all geometry optimizations. Because of the large surface unit cell vector sizes (∼30.2 Å), which is 11 times the primitive cell vector of the Ru(0001) surface, only the Γ-point in the Brillouin zone was sampled in the self-consistent calculations.
Figure 1d shows the same area as in Figure 1a, after dosing AO at 450 K. Following the dose, a number (0.002 monolayers) of new dark round features with ∼3 Å full width at half-maximum diameter are observed. These dark features are preferentially found within the FCC region (77%) of the Gr Moiré cell (highlighted with dark-blue circles in Figure 1d), although a smaller number (23%) is also observed within the HCP regions (brown circles, Figure 1d) after dosing AO at 450 K. This difference in stability suggests an energy difference on the order of ∼0.05 eV between the two sites. No AO-related features are observed in the Mound regions. It is important to note that the dark depressions of variable size on some of the Mound regions are unrelated to AO and have also been observed previously on bare Gr.48 In that study, the authors attributed the donut-like appearance of the mound regions to electronic effects. The complete Gr layer exhibited identical behavior, and the AO dose-dependent experiments showed only a linear increase in the AO coverage (see Figure S2). Depending on the tip termination, the AO features appear bright in ∼50% of the acquired images relative to the surrounding Gr substrate. To avoid confusion, we present only the subset where these features appear dark. The same area images with the dark and bright appearances of the AO species are shown in Figure S3. We further focus on determining the chemical identity of the AO species in both FCC and HCP regions. Two possibilities have been previously considered theoretically in the literature: the bridge-bonded, epoxy-like species bound to two neighboring carbons and the terminal; enolate-like species bound on top to a single carbon.31 It should be noted (Figure 1c) that the neighboring carbon atoms in the FCC and HCP regions are not equivalent due to their different coordination by the underlying metal atoms at the interface. As illustrated in the high magnification image in Figure 1e, the AO features are relatively large (∼4 Å), and they do not change with temperature (114−500 K). This large size prevents us from reliably making an accurate assignment of their position, i.e., on-top vs in-between carbon atoms, and consequently of their chemical identity. To determine the AO binding, we take advantage of their motion at elevated temperatures. The AO species became
■
RESULTS AND DISCUSSION The appearance of the bare Gr monolayer on Ru(0001) before the oxidation is illustrated in the STM images shown in Figure 1a,b. The typical Moiré superstructure characteristic of the single Gr layer supported on Ru(0001) substrate can be readily recognized. Small bare Ru regions covered with carbon can be also seen in the top- and bottom-left corners of the image in Figure 1a. These regions are a result of the intentionally incomplete Gr overlayer45 that help us locate the same area before and after the oxidation and reliably prepare the STM tip. The high magnification, atomically resolved image in Figure 1b shows the (25 × 25) Gr Moiré supercell. This (25 × 25) supercell is commensurate with (23 × 23) unit cells of the underlying Ru(0001).46,47 This large Gr supercell can be further divided into four similar (12.5 × 11.5) regions that can be approximated by a slightly simplified (12 × 11) model that is schematically depicted in Figure 1c.48 This model is subsequently used in our DFT calculations.36 Within each of the regions, three distinct areas with different height above the Ru(0001) surface can be identified: The Mound region (brightest, largest separation) at 3.0 Å and the FCC (medium bright) and the HCP (darkest) regions both at approximately 2.2 Å (see Supporting Information Figure S1 for additional computational details). 5104
DOI: 10.1021/jacs.7b12791 J. Am. Chem. Soc. 2018, 140, 5102−5109
Article
Journal of the American Chemical Society
Figure 3. Hoping rate analysis of the AO species obtained from the time-lapsed STM images acquired in the 425−500 K range. (a) Fraction of hops (∼6800 events) observed between consecutive images (acquired 160 s apart) at different temperatures. (b) Distribution of the double hop directions within the FCC region (upper part) and HCP region (lower part) of the Gr Moiré at 500 K. In the double hop analysis (266 events) we included events where the AO moved a distance within the 2.0−2.6 Å range. The analyzed distance range includes both epoxy and enolate diffusion scenarios (see Figure 2b, bottom), but the diffusion directions are convincing: green bars correspond to 60° spaced azimuths expected for enolate species; red bars include additional directions that would have to be equally present if AOs binds as epoxy species.
resulting in hops with azimuths of multiples of 30°. For the terminal AO, all carbon atoms are equivalent in terms of possible azimuths of double hops, and multiples of 60° are expected (green arrows). The lengths of the double hops are 2.13 and 2.46 Å for bridge-bonded AO and 2.46 Å for terminally bonded AO. The large difference in distances between single and double hops allows us to easily distinguish them apart as illustrated in section S4. The evaluation of longer (triple and quadruple) hops is more complex as the range of possible distances is larger. Nonetheless, by selecting relatively low temperatures (
