Published: January 19, 2011 r 2011 American Chemical Society 424 dx.doi.org/10.1021/nl103053t | Nano Lett. 2011, 11, 424���430 LETTER pubs.acs.org/NanoLett Size-Selective Carbon Nanoclusters as Precursors to the Growth of Epitaxial Graphene Bo Wang,���,�� Xiufang Ma,���,�� Marco Ca���o,��� Renald Schaub,*,��� and Wei-Xue Li*,��� ��� Scottish Centre for Interdisciplinary Surface Spectroscopy, School of Chemistry, University of St Andrews, St Andrews, KY16 9ST, United Kingdom ��� State Key Laboratory of Catalysis and Center for Theoretical and Computational Chemistry, Dalian Institute of Chemical Physics, Chinese Academy of Sciences, Dalian 116023, China ABSTRACT: The nucleation and growth mechanisms of graphene on Rh(111) via temperature-programmed growth of C2H4 are studied by scanning tunneling microscopy and spectroscopy, and by density func- tional theory calculations. By combining our experimental and first- principles approaches, we show that carbon nanoislands form in the initial stages of graphene growth, possessing an exclusive size of seven honeycomb carbon units (hereafter labeled as 7C6). These clusters adopt a domelike hexagonal shape indicating that bonding to the substrate is localized on the peripheral C atoms. Smoluchowski ripening is identified as the dominant mechanism leading to the formation of graphene, with the size-selective carbon islands as precursors. Control experiments and calculations, whereby coronene molecules, the hydrogenated analogues of 7C6, are deposited on Rh(111), provide an unambiguous structural and chemical identification of the 7C6 building blocks. KEYWORDS: Graphene, rhodium, coronene, carbon, STM, DFT, nucleation, growth Emillennium, ver since its experimental isolation at the turn of the new graphene has fascinated physicists due to its unique electronic structure.1 But more recently, the chemistry community has also realized its potential significance: During a typical heterogeneous catalytic cycle, the reaction of a carbon source (carbon oxides or hydrocarbons) at the surface of an active transition metal (TM) can lead to the formation of various condensed carbonaceous phases. Carbidic or graphitic in nature (as for the one-atom-thick graphene), depending on the inter- action strength with the metal catalyst, these carbon forms can be either beneficial or detrimental, by playing an active role in the chemical conversion or by leading to the deactivation (poisoning or coking) of the catalytic sites.2 Understanding the formation and development of these various phases on the surfaces of relevant materials, and assessing their stability and reactivity, proves hence to be of utmost importance.3 The synthesis of graphene can be accomplished by two complementary approaches based on the high-temperature pyrolysis of small hydrocarbons on TM surfaces: temperature- programmed growth (TPG) and chemical vapor deposition (CVD), as extensively demonstrated for (we restrict our discus- sion below to the 4d and 5d TMs): Ru(0001),4-6 Rh(111),7,8 Pd(111),9 Ir(111),10-12 and Pt(111).13,14 Both synthesis meth- ods lead to single-layer graphene characterized by a Moire superstructure, irrespective of the carbon source (ethene, pro- pene, or benzene) or of the TM support.12,15,16 By means of the TPG method, in which the sample is first exposed to the carbon source (e.g., at 300K) and then subsequently annealed to the desired temperature (above 900 K), an incomplete overlayer composed of graphene islands is obtained, the maximum cover- age of which is defined by the initial carbon uptake at saturation. The size of the carbon islands, ranging from a few to hundreds of nanometers, can be tuned by controlling the growth temperature, and Smoluchowski ripening of small, yet unidentified, carbon nanoislands was reported to govern graphene growth at about 900 K.12 A recent photoelectron spectroscopy study showed that a strong C-TM interaction exists for the intermediate carbidic species on Ir(111) by TPG but disappears when these condense into graphitic islands at higher temperature.17 The authors explained with the help of DFT calculations that the carbidic clusters bind strongly to the metal substrate through their peripheral atoms, forcing the clusters to adopt a domelike shape. When the carbon source is dosed onto a sample kept at a high temperature, as in the CVD method, the synthesis of an extended, single-layer graphene covering up to 100% of the substrate is achieved. No direct evidence of carbon nanoislands has been reported for this approach. Nevertheless, the observa- tion of nonlinear growth kinetics on Ru(0001), extracted from low energy electron microscopy (LEEM) data,18 strongly sug- gests that the formation of graphene proceeds by incorporation of clusters of approximately 5 C atoms rather than monomers. The attachment energy of an isolated C atom to an edge of graphene is prohibitively high, whereas it decreases significantly for larger C clusters.18 Received: August 29, 2010 Revised: November 25, 2010

425 dx.doi.org/10.1021/nl103053t |Nano Lett. 2011, 11, 424���430 Nano Letters LETTER A consensus emerges from these previous studies: Graphene growth involves a series of complicated reactions, whereby dehydrogenation of the carbon source occurs at relatively low temperatures (800 K), followed and/or paralleled by a transi- tion from carbidic species into a graphitic film at higher tem- peratures (900 K). Yet, many aspects of the growth mech- anism(s) are not fully understood. Although observed by several research groups, the fundamental carbidic building blocks have not been formally identified and little is known about their atomic-scale structure, thermal stability, and electronic proper- ties. The atomistic processes conducive to the formation of a weakly bonded graphitic overlayer based on the merger of strongly interacting carbon species still remain elusive and incomplete. Addressing these issues constitutes a pivotal require- ment for our ability to create graphene in a well-controlled and reproducible manner, to tailor the physical and chemical proper- ties of graphene-based nanoscale devices (as of interest to physicists), and to devise strategies either promoting the stability or suppressing the formation of the various carbon phases on TM-based catalysts (as of interest to chemists). In this report, the decomposition process of carbon-contain- ing molecules (ethene) on Rh(111) following the TPG method (the CVD method is inaccessible to our experimental facility) was investigated by low-temperature scanning tunneling micro- scopy (STM) and density functional theory (DFT) calculations. Our microscopy images reveal, in agreement with studies on other TM substrates (Ru, Ir, Pt), that TPG leads to the synthesis of graphene nanoislands with a size of several hundreds of nanometers. We experimentally identify, prior and during the initial stages of graphene island nucleation, carbon nanoclusters of a perfectly monodispersed size. The coarsening of these clusters results in the growth of graphene (a mechanism known as Smoluchowski ripening). STM and STS imaging establish that the carbon nanoislands adopt a domelike shape with an exclusive honeycomb structure composed of exactly seven fused benzene units, hereafter labeled 7C6. These observations are fully sup- ported by DFT calculations, which further highlight that the enhanced stability of 7C6 arises from a subtle cluster-size-depen- dent balance between C-C and C-metal bonding. Control experiments and calculations, whereby the hydrogenated analo- gues of 7C6 (i.e., coronene molecules) are deposited onto Rh(111), not only provide strong support to the identification of 7C6 but allow us to further our understanding of the chemical bonding of carbon clusters with a TM substrate. We finally discuss the formation of 7C6 as a result of the agglomeration of C2 hydrocarbon units. Methodology. All experiments were performed in a surface analysis system (in an ultrahigh vacuum environment, UHV, with a base pressure below 1 10-10 mbar) consisting of a prepara- tion chamber allowing for standard sample preparation and characterization by AES, and a microscope chamber housing a CreaTec low-temperature STM. The Rh(111) crystal was cleaned by repeated cycles of Ar�� sputtering and annealing in oxygen (3 10-7 mbar) at 1100 K and finalized by flash- annealing in vacuum at 1200 K. STM and STS were performed at liquid helium temperature using the constant-current mode and homemade W and PtIr tips. The cleanliness of the surface was monitored by STM and AES. Graphene was prepared by exposure of the Rh surface to C2H4 (99.995% purity) at room temperature (RT) followed by sequential annealing up to 973 K in UHV. Coronene molecules were deposited at RT on Rh(111) by a homemade evaporator. During the acquisition of differential conductance maps of the surface (or dI/dV images), standard lock-in detection techniques were utilized,19 whereby the feedback loop was kept active in order to maintain the tunneling current at a constant value. Topographic and dI/dV signals were simultaneously acquired at each pixel (the frequency of the voltage modulation is set to be higher than the bandwidth of the feedback system, hence the tip-sample distance does not react to the modulation when the dI/dV signal is acquired at each pixel of the topographic image). To ensure that STM tips were clean and suitable for measure- ments on carbon nanostructures, STS spectra were first recorded on bare Rh(111). Only those tips capable of yielding electro- nically featureless and smooth signatures corresponding to the clean metal surface were utilized. The DFT calculations were performed with the Vienna Ab-initio Simulation Package (VASP) code,20,21 using projector augmented wave (PAW) potentials22 and the generalized gradient approximation (GGA-PBE) for the exchange-correlation functional.23 Anenergycutoff of400eVwasusedtoexpandthewavefunctionsinto the plane-wave basis. The surfaces were modeled by employing the repeated slab geometry composed of five metal layers in the slab and a vacuum region of about 15 ��, on which single C atoms (or C clusters) were adsorbed on one side of the slab. The adsorbate and the topmost two layers were relaxed during structure optimization until the resi- dual forces on the atoms were less than 0.02 eV/�� (or 0.05 eV/��). For C clusters, we used (4 4), (6 6), or (7 7) unit cells according to the cluster size, and used (3 3 1), (2 2 1), (1 1 1) gamma centered K points, respectively, for the surface Brillouin zone sampling. All parameters defining the numerical accuracy of the calculations were carefully tested. STM images were simulated by using the Tersoff-Hamann approximation,24 in which the tunneling current is considered to be proportional to the integrated local density of states (LDOS) within a given energy window determined by the applied bias on the sample. The positive (negative) bias indicates that empty (occupied) states are imaged, in line with what is adopted in our STM measurements. Results and Discussion. Figure 1a displays STM images of a Rh(111) surface saturated with 18 Langmuirs of C2H4 at RT and annealed to 473 K. It can be seen that molecular adsorption has induced the well-known formation of the two coexisting (2 2) and c(4 2) superstructures.25 Upon deposition at RT, ethene readily deprotonates to ethylidyne (C2H3) and adsorbs in the 3-fold hexagonal close packed (hcp) hollow site with its C-C axis perpendicular to the surface.26 Annealing the system to 773 K in UHV (Figure 1b) results in the appearance of small protrusions located on the substrate terraces with sizes ranging from 1 to 2 nm and an apparent height of less than 0.2 nm. Further successive annealings up to 973 K (Figure 1c,d) lead to a remarkable narrowing of the particle size distribution and a decrease in the particle density. As can be seen from the high- magnification STM image presented in the inset of Figure 1c, the size distribution collapses into a single cluster size of 1 nm in diameter. This decay coincides with the emergence of graphene islands not only attached to the step edges but also occasionally found on the terraces of the substrate (especially on large terraces). The graphene islands are easily recognizable from their typical Moire patterns.6,10-12,14 Figure 1e shows the density variation of the 1 nm sized islands (which we will later identify as 7C6) with annealing temperature. The results presented in Figure 1 indicate that thermal decomposition of ethene at temperatures below 770 K leads to