Structural properties and site specific interactions of Ptwith the graphene/Ru„0001… moiré overlayer

Kerstin Donner and Peter Jakoba�

Fachbereich Physik und Wissenschaftliches Zentrum für Materialwissenschaften,Philipps-Universität Marburg, D-35032 Marburg, Germany

�Received 12 August 2009; accepted 21 September 2009; published online 23 October 2009�

The coherence of graphene layers on Ru�0001� over extended distances has been employed toidentify fcc and hcp regions of the associated moiré superstructure. These findings can be used asa straightforward method to discriminate between fcc and hcp hollow sites of Ru�0001�. Ourapproach thereby makes use of the “magnifying lens” characteristics of the graphene/Ru�0001�overlayer and its coherence across several monatomic steps of the substrate. We demonstrate that theindividual regions of the graphene/Ru�0001� overlayer exhibit pronounced variations in interactionstrengths with deposited metal atoms. Specifically, Pt clusters have been grown at 140–180 K andthey are found to organize in a well-ordered periodic array defined by the moiré superlattice. Theirpreferred location within the graphene/Ru�0001� moiré unit cell is identified to be the fcc region.© 2009 American Institute of Physics. �doi:10.1063/1.3246166�

I. INTRODUCTION

The adsorption geometries of surface species are gener-ally characterized in terms of their local binding site, i.e.,on-top, bridge, or hollow site adsorption. In the latter case adistinction between hcp and fcc sites is made depending onwhether a second layer metal atom is present or missingunderneath of the respective threefold coordinated site. Eventhough associated differences in adsorption energies oftenare delicate, adsorbates usually discriminate quite clearlyamong the two similar choices. Except for very low tempera-tures, thermal equilibrium among the various sites is easilyattained, thanks to their short distances.

Theory is, in principle, capable to resolve the differencesin adsorption energies of hcp and fcc sites;1,2 still, experi-mental confirmation of the exact nature of the hollow site isessential.2,3 Such an endeavor, however, constantly fails dueto the faint imaging contrast between hcp and fcc sites inscanning tunnel microscopy �STM� even if atomic resolutionis achieved.2,4 Thanks to a uniform stacking sequence dis-crimination of hcp and fcc sites is less involved for sub-strates with fcc as compared to hcp crystal structure, despitethe similar geometry of their hexagonally close-packed sur-faces. For fcc�111� surfaces a slight misorientation in a welldefined direction uniquely determines the step type �providedthat the upper and lower sides of the polished sample diskare not mixed up�; it is then straightforward to extract thelocations of fcc and hcp sites within the �1�1� unit cell aslong as atomic resolution is attained, i.e., the on-top positionsof substrate atoms are resolved. For the likewise hexagonalhcp�0001� surfaces this procedure would not work due toorientation switching of fcc and hcp sites within the 1�1unit cell when moving one step down/up to a neighboringterrace.

Several methods and approaches exist to achieve such anassignment.

• The weak but nonzero contrast between fcc and hcpsites in STM might be employed and compared withcalculated STM image maps to recover the precise na-ture of the adsorption sites �method A�.

• A straightforward method would be to use adsorbateswith known binding geometries �e.g., derived from adynamical low energy electron diffraction-IVanalysis5,6� as a reference frame �method B�; oxygen onRu�0001�, for example, occupies exclusively hcp hol-low sites at all coverages.

• Yet another approach makes use of the relative coordi-nation of lattices on neighboring terraces, separated bymonatomic steps4,7 �method C�; this analysis assumesnegligible lattice distortions �relaxation� of the stepedges, as well as requires a high linearity of the scan-ning stage in addition to atomic resolution imaging.

• In this paper we suggest to use graphene flakes extend-ing over several steps to uniquely identify the step typesand derive the location of fcc and hcp hollow sites withrespect to the atomic positions of surface atoms�method D�. In principle, this approach resemblesmethod C; however, it benefits from the “magnifyinglens” characteristics8,9 encountered when a weakly in-teracting graphene layer with well defined lattice con-stant, slightly smaller compared to the underlying sub-strate, is overlaid and a moiré pattern emerges in STM.

In order to simplify description of the various regions ofthe graphene overlattice characterized by the coordination ofits carbon atoms with respect to the underlying substrate, wewill label the local regions of the graphene/Ru�0001� super-structure according to the substrate sites located underneathof the center of the graphene hexagons �in accordance witha�Electronic mail: peter.jakob@physik.uni-marburg.de.

THE JOURNAL OF CHEMICAL PHYSICS 131, 164701 �2009�

0021-9606/2009/131�16�/164701/10/$25.00 © 2009 American Institute of Physics131, 164701-1

Downloaded 04 Nov 2009 to 137.248.1.11. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/jcp/copyright.jsp

Refs. 8 and 10�, e.g., the moiré fcc regions define those areaswhere the hexagons are located above fcc sites of Ru�0001�,etc.

For graphene grown on various transition metal sub-strates the observed hexagonal moiré lattices with periodici-ties of 20–35 Å display a vastly enhanced vertical contrastwhich allows an easy discrimination of the inequivalentmoiré hollow regions. According to recent work of Marchiniand co-workers,11,12 these correspond to local on-top/hcp�moiré fcc regions� and on-top/fcc �moiré hcp regions� coor-dinations of the two inequivalent C-atoms of the graphenehexagons; just for completion we mention that the moirémaxima �moiré on-top regions� have been associated withlocal fcc/hcp site occupation of the two C-atoms and withon-top sites underneath of the graphene hexagon centers.

The current high interest in preparing and characterizinggraphene overlayers grown on various solid substrates13,14

may be divided in three main categories.

�i� They may serve as a model system for free standinggraphene15 and allow an investigation of its uniqueelectronic features,16 as well as to explore possibleapplications.

�ii� They can be used as a template to grow or attach othermaterials; the graphene layer thereby effectively re-duces the interaction strength to the substrate, enhanc-ing lateral mobility as well as easing detachment.

�iii� The periodicity of 2.4612 Å for graphite is close tothe lattice constants of prominent close-packed transi-tion metal surfaces and the mismatch of about 5%–10% may lead to the formation of perfectly orderedmoiré superstructures with periodicities of several na-nometers. These overlattices could then be used astemplates to grow uniform islands with nearly perfectlong range order.10 An important ingredient to makesuch a self-assembly work is provided by the varyingadsorption bond strength of the various graphenemoiré regions with deposited material.9,10,12,17

In a very recent overview Wintterlin and Bocquet18 re-viewed the novel field of graphene layers on metal surfacesin terms of their long range order, detailed geometrical andelectronic structure and presented some of the applicationsand perspectives. They suggested that graphene layers maybe divided into essentially two classes depending on the in-teraction strength of graphene with the underlying substrate.Weakly interacting graphene on SiC or Ir�111� yield nearlyperfect Dirac cones �thanks to band gaps at the edges of theBrillouin zone�13,19 and these layers could be characterizedas virtually freestanding graphene sheets. Due to their sub-stantial distance to the substrate these graphene layers areexpected to remain more or less undistorted, which is con-firmed by recent STM data on graphene/Ir�111� with its rela-tively small corrugation amplitude of about 0.27 Å.10

For the more strongly interacting substrate Ru�0001� thisvalue amounts to 1–1.5 Å and, according to surface x-raydiffraction11 �SXRD� and various theoreticalcalculations,12,20 the large apparent corrugation of thegraphene overlayer is indeed mostly geometrical and to alesser extent electronical.21–23 In addition, the graphene lay-

er’s warping is accompanied by a substantial vertical relax-ation of the uppermost Ru layers by about 0.2 Å.24 Thelowest-lying C-atoms are located 2.2 Å above an underlyingRu,12,24 as compared to 3.7–3.9 Å on Ir�111� �Ref. 10� orPt�111�,25 which has been taken as clear evidence for achemical rather than a Van der Waals interaction of grapheneon Ru�0001�.24 This conclusion is supported by x-ray photo-emission spectroscopy data revealing a high degree of orbitalhybridization between graphene and Ru�0001�.26 Not unex-pectedly, the Dirac cones of graphene/Ru are barelyrecognizable.27

Interestingly, graphene on Ru�0001� is forming a �25�25� graphene on �23�23� Ru structure consisting of fourmoiré subcells, as deduced from SXRD data;24 the longrange order as seen by STM, however, does not usually dis-criminate between these subcells leading to a moiré period-icity of about 30 Å.11,21,28,29 Also, the graphene lattice onRu�0001� does not always align perfectly along the Ru atomsrows; rather the graphene and Ru�0001� lattices may be mis-oriented with respect to each other by few degrees.11,21

Remains to mention graphene on Ni�111�,30 a likewisestrongly interacting system, but a rather special case with theNi�111� lattice constant virtually identical to graphene, re-sulting in lattice matched graphene layers with �1�1� longrange order.

II. EXPERIMENTAL

The STM measurements were carried out in an UHVchamber �base pressure p�5�10−11 mbar� using a roomtemperature STM �DME Rasterscope�. The STM is usuallyoperated in the constant current mode; typical settings forSTM data acquisition were a tunneling current of about 0.5nA at a bias voltage of 0.2 V. Pt was evaporated at rates ofabout 5�10−4 ML /s by resistive heating of a 100 �m thickfoil �Goodfellow, purity �99.99%� mounted on a liquid ni-trogen cooled holder. Pressure increase during evaporationtypically was 1�10−10 mbar. During Pt deposition thesample temperature could be set to 100–600 K, as measuredby a thermocouple attached to a sample mounting block. Gasexposures were performed by a capillary ��=2.0 mm� infront of the sample when mounted to the heating stage.Sample cleaning was carried out according to standard pro-cedures by Ar+ ion-sputtering and oxygen dosing/heatingcycles. The T-shaped Ru crystal had a size of 5�5 mm2

�front side�, and was oriented within �0.5° of the �0001�direction, leading to an average step-step distance was about300 Å. It was clamped onto a Ta plate which could be trans-ferred from the STM to either the Pt source, a sample mountin front of the sputter gun, or attached to a heating stagewhere it could be cooled to 150 K and flashed up to 1600 K.Temperatures in this latter stage were measured using anindium-antimonide infrared detector, calibrated with a py-rometer and/or a dummy sample with thermocouple at-tached. Calibration of the scanner �lateral dimensions� isattained by preparing a well-ordered Ru�0001�-�3�3�-�benzene+O� coadsorbate layer31 with its defect-free

164701-2 K. Donner and P. Jakob J. Chem. Phys. 131, 164701 �2009�

Downloaded 04 Nov 2009 to 137.248.1.11. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/jcp/copyright.jsp

domains extending over dozens of nanometers. The verticalscale has been set according to the known step height ofRu�0001�.

III. RESULTS AND DISCUSSION

A. Growth of graphene monolayers on Ru„0001…

Graphene layers may be grown either by surface segre-gation of bulk carbon11,21,32 or by thermal dissociation of�unsaturated� hydrocarbons, e.g., ethylene.10,21,24,27,28,32–35

Our graphene on Ru�0001� layers have been grown by ad-sorbing benzene at low T and subsequent annealing to 1000–1300 K, or by exposing the surface to benzene at this high T.In the former case several cycles were required to obtainextended graphene areas since saturation of chemisorbedbenzene on Ru�0001� is limited to 0.15 ML, correspondingto a C:Ru ratio of 0.9.36 As 10% of chemisorbed benzenedesorbs intact �benzene on a graphene monolayer desorbscompletely� and C:Ru�2,36 for graphene/Ru,24 our obser-vation of about 30% surface coverage after 1 cycle is inaccordance with expectation. The quality of the graphenelayers, especially the long range order was better for benzeneexposures at elevated T and could be further improved byapplying an extra postgrowth annealing step. For Figs. 1 and2 benzene has been adsorbed to saturation at T=170 K andthe sample annealed to successively higher Tann. Our obser-vations regarding the temperature dependency of the growthof graphene flakes largely agree with similar reports on thegrowth of graphene on Ir�111� �Ref. 37� and will be brieflydescribed in the following.

• At low Tann=900–1000 K graphene islands are scat-tered randomly across the surface �Fig. 1�. A novel as-pect in terms of graphene growth is introduced by ourstudy insofar as we detect numerous uniform grapheneplatelets �diameter ��10–12 Å; see line scan of Fig.1�c�� in parallel to the existing more extended grapheneislands. The locations of these hexagonal nanoislands�with a slight tendency toward a triangular shape� onthe Ru�0001� surface is more or less random. It is sug-gested that the graphene platelets represent the stableprecursors of more extended graphene islands or “car-pets.” Direct attachment of carbon atoms to grapheneislands seems to play a subordinate role only, at leastfor benzene decomposition at not too high temperatures.We note that despite their similarity to the dehydroge-nated coronene islands observed on Ir�111�,37 our caseis different since on Ru�0001� these nanoislands musthave formed by synthesis of smaller units, i.e., carbonatoms or benzene cores. As to their uniform size wesuggest that an increasing lattice mismatch at the rim ofgrowing islands limits their size, and/or enhances theirprobability of merging with existing more extendedgraphene islands. Their mere observation, however,suggests that the graphene platelet’s mobility at about1000 K is still relatively low; possibly, an energetic bar-rier exists preventing an easy attachment of grapheneplatelets to existing graphene islands. The energetics

and growth kinetics of graphene islands on Ru�0001�was very recently modeled by Loginova et al.38,39 andindeed such a barrier ��Eb=2 eV� was suggested.

• Tann�1100 K resulted in graphene being preferentiallyattached to Ru steps �not shown�. Graphene platelets arestill observed, however, at a considerably reducednumber.

• Tann�1200 K lead to the formation of wide-stretchedgraphene islands extending over an entire terrace oreven traversing several step edges �Fig. 2�. Thegraphene long range order, however, is relatively poor,most likely due to recombination of neighboring,smaller graphene flakes to form larger units. We notethat the defect density for the small islands is muchbetter than for the larger ones; interestingly, their sizealways tends to be a multiple of the graphene/Ru moiréunit cell. This means that, similar to graphene onIr�111�,36 Smoluchowski ripening40 �mobile islandsmerge� rather than Ostwald ripening41 �C-atoms detachfrom graphene islands to migrate to larger, more stable,islands� prevails.

FIG. 1. �a� STM image �800�600 Å2� of Ru�0001� exposed to benzene at170 K �saturation coverage� and annealed to 1000 K to induce benzenedissociation and graphene formation. Small patches of graphene/Ru�0001��moiré periodicity am�30 Å� are formed in parallel to numerous grapheneplatelets �diameter �=10–12 Å�. �b� Enlarged section of �a� showing aminigraphene island along with several graphene platelets �image size 90�70 Å2�. �c� Line scan across two graphene platelets on Ru�0001�, asindicated by the line in �b�.

164701-3 Pt on graphene/Ru�0001� J. Chem. Phys. 131, 164701 �2009�

Downloaded 04 Nov 2009 to 137.248.1.11. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/jcp/copyright.jsp

• Interestingly, the islands have a tendency to reshape theoriginal Ru�0001� step edges in an attempt to enhancethe boundary zone �→ wetting behavior of two quasi-two-dimensional crystalline solids�. As this processcontinues, the graphene flakes more and more resembleareas embedded in the surface layer rather than repre-senting graphene layers floating a few angstroms abovethe Ru�0001� terraces.

• Perfect graphene carpets extending over many steps andwith the graphene/Ru�0001� moiré aligned along the Ru

atoms rows �sometimes with a slight misalignment� areformed when benzene has been adsorbed and dissoci-ated at T�1000 K and postannealed at Tann

�1200–1250 K �Fig. 3�. In accordance withliterature11 we observe a reshaping �straightening� ofthe Ru�0001� step edges. In addition, interaction be-tween Ru�0001� and graphene leads to kinks the size ofthe moiré unit cell in case that the original Ru step edgeorientation deviates too strongly from the Ru�0001�atom rows �see also Ref. 11�. It is therefore suggested�in accordance with similar findings for graphene onIr�111� �Ref. 37�� that the growth of graphene at el-evated T induces Ru mass transport over “long” dis-tances. Note that the coordination of graphene moirémaxima with respect to kinks as well as to step edges isalways kept alike, which means that the observed struc-tures represent local thermodynamic equilibriumstructures.

• Temperatures in excess of 1400 K resulted in bulk dif-fusion, reducing the surface carbon coverage �C.

Using C2H4 as a carbon source very similar observationsin terms of graphene growth on Ru�0001�, namely, the for-mation of nanographene clusters at T=900–1000 K and thegrowth of extended graphene islands at T�1200 K, prefer-entially located at step edges have been reported veryrecently.34 A particularly interesting aspect is added in thiswork by intercalating oxygen atoms between the graphenelayer and the Ru�0001� substrate, thereby weakening its in-teraction strength as deduced from the substantially reducedcorrugation amplitude of the graphene moiré lattice.

In no case did we observe bilayer graphene, which weattribute to the low bulk carbon content of our Ru�0001�sample; note that bilayers can be grown only by means ofsurface segregation of bulk carbon28,32 since decompositionof hydrocarbons on Ru�0001� at elevated T to form agraphene monolayer is a self-limiting process.

Using k�m=k�C−k�Ru with km=2� /am, kC=2� /aC, and

FIG. 2. �a� STM image �2000�1850 Å2� of Ru�0001� exposed to benzeneat 170 K �i.e., the layer of Fig. 1� and annealed to 1200 K. Extended patchesof rather defective graphene/Ru�0001�, consisting of smaller units withmuch better long range order �moiré periodicity am�30 Å� are formed. Theisland borders adjust to the moiré superstructure and tend to align parallel tosubstrate atom rows. Graphene platelets have converted to small islands ormerged with larger graphene sheets. �b� Enlarged section of �a� displayinggraphene islands located on the plain terraces, attached to the lower stepedge �thereby reshaping it�, or extending across Ru�0001� step edges �imagesize 700�700 Å2�. �c� Several line scans across monatomic steps ofRu�0001� as indicated by the lines in �b�: �i� with and �ii� without a grapheneisland attached to the step �blue and black curves, respectively�, as well as�iii� crossing an isolated graphene island on a plain terrace �green curve�.The corrugation height of the graphene overlayer amounts to �zcorr

�1.1 Å and the average height of graphene above the Ru�0001� surface is�zC=1.6–1.7 Å.

FIG. 3. STM image �400�300 Å2� of Ru�0001� with virtually the entiresurface covered by graphene, produced by dissociating benzene at about1000 K and flashing to 1250 K thereafter. The graphene carpet extends overseveral step edges so that the coherency of the graphene moiré superstruc-ture is maintained on neighboring terraces. The lateral shift �m

�A/B of thegraphene “lattice” when traversing A or B type of steps is due to the lateraloffset of substrate atoms on neighboring terraces and �to a much smallerextent� due to a bending of the graphene sheets.

164701-4 K. Donner and P. Jakob J. Chem. Phys. 131, 164701 �2009�

Downloaded 04 Nov 2009 to 137.248.1.11. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/jcp/copyright.jsp

kRu=2� /aRu=2� /2.706 Å−1, the observed moiré periodicityof am=30�1 Å �for parallel k�C and k�Ru� is in accordancewith aC=2.482�0.006 Å for the periodicity of the graphenehexagons. We conclude that the graphene lattice is about 1%expanded with respect to graphite �dC–C=2.4612 Å�, whichwe attribute to the weakened C–C bond strength as a resultof the Ru–C chemical bonding. The apparent moiré corruga-tion amplitude was about 1�0.3 Å, depending on bias volt-age; the observed distance between graphene and Ru�0001��dC–Ru=1.6–1.7 Å �see line scan in Fig. 2�c�� is in agree-ment with Sutter et al.28 In accordance with Marchini et al.,11

we find that the overall appearance of the graphene/Ru moiréin our STM images depended on the applied bias voltage�sample potential�. Specifically, we have observed either thestandard graphene moiré patterns with one maximum andtwo inequivalent minima per unit cell, or, triangular shapedringlike maxima.

B. Identification of fcc and hcp regionswithin the graphene/Ru„0001… moiré unit cell

A quite common observation encountered for graphenelattices grown at high T on various transition metal surfacesis that the graphene flakes may extend over several stepedges in a carpetlike manner without any sign of disruptionor other type of structural defect.8,9,11 Thereby step edgestend to straighten up along the substrate atom rows and themoiré lattice aligns parallel to these. In a recent study Sutteret al.28 observed that such growth over steps does occur forRu�0001� as well, although only in the downhill direction,i.e., that coherence of graphene flakes across step edges ispossible. Figure 4 �among others, e.g., Fig. 2� clearly cor-roborates this finding and furthermore, demonstrates thatgrowth across step edges occurs for isolated islands as well.We note, however, that islands of the displayed type formonly occasionally under the present conditions, i.e., whenhydrocarbons are adsorbed at low T and slowly annealeduntil graphene islands form; best results in this respect areobtained when graphene is grown at high temperatures T�1000 K by thermal dissociation of hydrocarbons. The roleof steps to act as nucleation centers for the growth of

graphene thereby appears to be much lower for Ru�0001�when compared to Ir�111�,37 as graphene islands are foundequally well on Ru�0001� terraces and at step edges.

In the following we will describe the identification anddiscrimination of the two types of Ru�0001� step edges oc-curring consecutively. Specifically, we will make use of therelative coordination of the graphene lattices on the upperand the lower terraces, as indicated by the lateral offset ofmoiré maxima on both sides of the Ru step edges. Our analy-sis thereby follows the approach presented recently byCoraux et al.8 who analyzed graphene layers extending overstep edges of an Ir�111� substrate.

Before determination and discrimination of moiré-fccand moiré-hcp regions are exemplified, we like to recall thatthe position of fcc and hcp sites within the Ru�0001� unit cellreflects itself in the relative locations of moiré-fcc and moiré-hcp regions within the moiré unit cell �see Fig. 4 of Ref. 9 orFig. 7 of Ref. 11�. Specifically, fcc sites of Ru�0001� andmoiré-fcc regions are located in the center of oppositely ori-ented triangles, defined by on-top sites and moiré-on-top re-gions, respectively. This means that if we succeed to identifymoiré fcc and hcp regions of graphene, it is straightforwardto determine fcc and hcp sites of Ru�0001�.

In Fig. 5 we show a graphene flake extending over sev-eral monatomic steps of the Ru�0001� substrate. Apparently,during its formation the Ru step edges rearranged themselvesso that a well defined coordination of the Ru step edges withrespect to the moiré structure is created, in accordance with

FIG. 4. STM image �250�200 Å2� of Ru�0001� exposed to benzene at 170K and annealed to 1200 K; the shown graphene island, extending across amonatomic �A type� Ru�0001� step edge, has been imaged at another loca-tion of the sample displayed in Fig. 2. Thereby, the edges of the grapheneisland have a pronounced tendency to align along substrate atom rows andthe island size to adjust to multiples of the graphene unit cell.

FIG. 5. STM image �300�570 Å2� of Ru�0001� with virtually the entiresurface covered by graphene, produced by dissociating benzene at about1000 K and flashing to 1250 K thereafter. As in Fig. 3, the graphene carpetextends over several step edges with the coherency of the graphene moirésuperstructure maintained perfectly. We have indicated the max-max dis-tances of the graphene moiré superstructure on the flat terraces, �m

�, andwhen traversing A /B type of steps, �m

�A/B �which come along consecutivelyfor hcp�0001� surfaces with a slight misorientation�. The fcc and hcp regionsof graphene/Ru�0001� as deduced from our analysis are indicated by arrows.

164701-5 Pt on graphene/Ru�0001� J. Chem. Phys. 131, 164701 �2009�

Downloaded 04 Nov 2009 to 137.248.1.11. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/jcp/copyright.jsp

observations of Marchini et al.11 and a suggested strongC–Ru interaction. More precisely, the position of thegraphene/Ru moiré maxima with respect to the Ru step edgeis perfectly reproduced for the two step types; however,while this coordination is maintained to a high degree foreach particular step type, it is different for A and B types ofsteps. There exists a clear preference of the graphene latticeto align along the substrate atom rows and to straighten upstep edges, similar to earlier observations on Ir�111�.8,9 Thecoherence of the graphene/Ru moiré lattice and Ru stepedges thereby seems to be even better than observed forgraphene on Ir�111� with its relatively weak graphene—Irinteraction strength.

When traversing a step edge of the underlying substratethe strictly periodic moiré superstructure experiences an off-set perpendicular to the step edge. This offset is the sum oftwo contributions8 �step edge relaxation would be a thirdone, but we will neglect this effect since it is restricted to theimmediate vicinity �less than 15 Å� of the step edges42�.

�i� The bending of the graphene lattice when traversingthe steps causes the graphene to retract by a certainamount �C

� �few tens of angstrom� perpendicular tothe step edge. As we will show below, �C

� amounts to0.1–0.15 Å for graphene on Ru�0001�, independent ofthe step type.

�ii� The lateral shift of consecutive substrate layers leadsto a displacement of substrate atoms �Ru

� perpendicu-lar to the step edges. For A type of steps �see Fig.6�a�� �Ru

�A= 13aRu

� = 13 ��3aRu�=1.562 Å; for B type of

steps �see Fig. 6�b�� �Ru�B=− 1

3aRu� =− 1

3 ��3aRu�=−1.562 Å.

We have now determined the lateral offset perpendicularto the step edges of the moiré lattice on the upper and lowerterraces of consecutive monatomic steps and compared themto the periodicity am

�=am�3=52 Å of the undistorted

graphene/Ru�0001� moiré overlayer �see Fig. 5�. We find thatthis offset amounts to 85 Å=1.63am

� for the one �A typestep, see below� and 68 Å=1.31am

� for the other �B� steptype. From these offsets �m

�A and �m�B, and in conjunction

with the known lateral offsets for A or B type of steps �Ru�A

and �Ru�B, we can determine the slight lateral offset of the

graphene layer due to the warping as it spreads from oneterrace to the neighboring one.

According to

km��m

�A/B = kC��C

� − kRu� �Ru

�A/B + nA/B2� �1�

�a derivation of this expression is given in the Appendix� forA or B type of steps, the step induced shift of the moirémaxima should equal

�m�A = am

��nA − �Ru�A/aRu

� + �C�/aC

��

= am��nA − 1/3 + �C

�/aC�� �2a�

for A type steps and

�m�B = am

��nB − �Ru�B/aRu

� + �C�/aC

��

= am��nB + 1/3 + �C

�/aC�� �2b�

for B type steps; here, nA/B are integers which depend on theparticular choice of moiré maxima in our analysis. The dif-ference �m

�A−�m�B does not contain the cumbersome shift �C

�

due the graphene bending and we obtain

�m�A − �m

�B = am��nA − nB − 2/3� = am

��n − 2/3� , �3�

which is in fact reproduced perfectly �with the integer n=nA−nB=1�. Quite similarly, we can determine the quantity�m

�A+�m�B for which the lateral offsets of both types of steps

cancel out and we are left with a term containing the shift �C�

�graphene bending�, i.e.,

�m�A + �m

�B = am��nA + nB + 2�C

�/aC�� . �4�

From this expression we derive �C�=−0.029aC

�=−0.125 Å�for nA+nB=3�. The minus sign indicates that the bendingleads to an apparent retraction of the graphene lattice whentraversing a step, in accordance with expectation. We havealso considered a reversed assignment of A and B types ofsteps in our analysis. This would, however, lead to unrealisticand dissimilar values of �C

� for A and B types of steps�1.54 and 3.00 Å, respectively� so that such a reversal �ascompared to Fig. 5� can definitely be ruled out.

As evident from crystallography and depicted in Fig. 6,A type �B type� steps of an hcp�0001� surface have their hcp�fcc� sites located in a triangle �formed by substrate atoms�pointing away from the step edge �on the upper as well as thelower terrace�. The reverse then applies for the fcc and hcpregions of the graphene/Ru�0001� moiré superstructure andwe conclude that the respective regions are as indicated inFig. 5, i.e., the brighter of the two moiré minima denote fccregions of graphene/Ru�0001�.

C. Growth of Pt clusters on the graphene/Ru„0001…moiré overlayer

We have now further investigated whether Pt on the onehand and graphene on the other represent a good combina-tion to grow well-ordered arrays of �sub-� nanometer sizedmetal clusters with the long range order induced by a pre-ferred nucleation within the graphene/Ru�0001� moiré unitcell. In recent work well-ordered arrays of Ir, Pt, W, and Reclusters have been grown on a graphene/Ir�111� moiré

FIG. 6. Schematic view of A and B type of steps of the Ru�0001� samplewith the �1�1� unit cell indicated on both terraces. For A type of steps thelateral offset of the atoms on the lower terraces with respect to those on theupper terrace is �Ru

�A=+ 13aRu

�3; for B type of steps this offset amounts to�Ru

�B=− 13aRu

�3.

164701-6 K. Donner and P. Jakob J. Chem. Phys. 131, 164701 �2009�

Downloaded 04 Nov 2009 to 137.248.1.11. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/jcp/copyright.jsp

lattice.10,43 Thereby the preferred locations of the clusterswere identified as moiré-hcp regions, i.e., with the graphenehexagons located above hcp sites of the underlying Ir�111�substrate.9,10 According to Wintterlin and Bocquet18

graphene/Ir�111� represents a weakly interacting system witha relatively large graphene-Ir interlayer distance �Ref. 19�; inaccordance with expectation, this weakly interactinggraphene layer displays a mere 0.2–0.3 Å apparent corruga-tion of the moiré superstructure in STM images.10 Its originis predominantly geometrical, as density functional theory�DFT� calculations revealed a warping of the graphene layerof the same magnitude.10 It should be mentioned, however,that corrugations in STM images may vary considerably withbias voltage; for graphene on Ir�111� even a contrast inver-sion has been observed.9

Graphene on Ru�0001�, on the other hand, represents amuch stronger interacting system12,24,26,27 with substantialgeometrical distortion of graphene �vertical corrugation/relaxation�, as well as of the Ru�0001� surface.12,24 As to theorigin of the pronounced corrugation of the graphene/Rumoiré in STM, there exists some controversy regarding theextent of the electronic contribution as opposed to a more orless purely geometrical warping of the graphenemonolayer.21–23 The latter supposedly arises due to the spa-tial variation of the relative locations of carbon and Ru atomsleading to periodically varying Ru–C bond strengths andlengths. This fact lead Martoccia et al.24 to conclude that thegraphene/Ru�0001� system might be a good candidate togrow isolated metal nanoislands. More precisely, an en-hanced substrate bond strength of the carbon atoms willcause a local distortion of the otherwise flat graphene lattice,in conjunction with a sp2→sp3 rehybridization.12,20 Forgraphene on Ir�111� Feibelman17 suggested that C-atoms ad-jacent to those carbon atoms forming strong substrate bonds�i.e., those above substrate atoms� exhibit an enhanced reac-tivity toward binding metal atoms on the vacuum side of thegraphene layer; a similar argument probably applies to thegraphene/Ru�0001� system.20

In Fig. 7 increasing amounts of Pt have been depositedat a growth temperature Tg=140–180 K onto our graphene/Ru�0001� layers and warmed up to 295 K for STM imaging.At a deposition rate of 5�10−4 ML /s the growth tempera-ture of about 150 K represents optimum conditions regardingformation and ordering of Pt clusters. Pt growth at lower orhigher substrate temperatures Tg lead to increased amountsof randomly positioned clusters �T120 K�, or to the for-mation of larger and fewer clusters �T�190 K�.

Most importantly, we find that deposited Pt produceswell defined clusters arranged periodically within the moirélattice defined by monolayer graphene on Ru�0001�. In theseries of Fig. 7 the Pt coverage �Pt has been varied by usingidentical settings to heat our Pt foil, while increasing theevaporation time span. The �Pt values have been examinedby imaging �monolayer� Pt islands formed on cleanRu�0001� areas. An estimate of the average number of Ptatoms within the islands in Figs. 7�a�–7�d� is given in TableI, along with the filling factor � of the graphene/Ru�0001�moiré unit cells and a quantification of the relative fractionof islands exceeding monolayer height. In accordance with

previous studies of Ir clusters formed on graphene/Ir�111��Ref. 10� we distinguish three different regimes.

�i� Nucleation regime ��Pt0.05 ML�: islands withsimilar size form at well defined locations within themoiré unit cell. With increasing Pt amount the densityof islands increases steadily while their size remainsabout the same, or increases only slowly. There seemsto exist a minimum size �presumably hexagonal is-lands with seven Pt atoms� and this may be due to ourroom temperature imaging, i.e., annealing of thesamples. The preferred location within the moiré unitcell is the brighter of the two moiré minima, whichaccording to our analysis in Sec. III B are the fccregions.

�ii� Growth regime �0.1 ML�Pt0.5 ML�: virtuallyall moiré unit cells contain a Pt island. While the longrange order clearly signals an identical location withinthe graphene/Ru moiré unit cell, the uniformity andheight of the islands display some variation; the frac-tion of islands with a height exceeding one monolayerincreases steadily with �Pt �see Table I�.

�iii� Coalescence regime ��Pt�0.5 ML, not shown�: dueto their increasing size, neighboring Pt islands mayrecombine to form larger units extending across bor-ders of a moiré unit cell. Due to negligible overallmobility of the merged islands their position is some-what off from the preferred locations found for thesmaller islands.

In parallel to our study Zhang et al.35 performed similarexperiments with Pt deposited onto graphene/Ru�0001� at121 K �and warmed up to room temperature for imaging�.They observed highly dispersed Pt clusters with about uni-

FIG. 7. Series of STM images with increasing amounts of Pt deposited ontographene/Ru�0001�: �a� �Pt=0.06 ML at Tg=145 K, image size 200�165 Å2; �b� �Pt=0.12 ML at Tg=145 K, image size of 200�165 Å2;�c� �Pt=0.18 ML at Tg=180 K, image size of 800�500 Å2; �d� �Pt

=0.24 ML at Tg=145 K, image size of 800�500 Å2. Following Pt depo-sition the sample is warmed up to room temperature for STM imaging. Thedisplayed areas of �c� and �d� have been chosen such that Pt deposited ontoclean Ru�0001� areas are included as well. The graphene layer was producedby dissociating benzene at about 1000 K and flashing to 1250 K thereafter.

164701-7 Pt on graphene/Ru�0001� J. Chem. Phys. 131, 164701 �2009�

Downloaded 04 Nov 2009 to 137.248.1.11. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/jcp/copyright.jsp

form size ���30 Å�, with the preferred location within thegraphene moiré unit cell being the brighter of the two hollowregions, in accordance with our conclusion. At variance toour observation of perfectly periodic arrays of Pt islands,their clusters are randomly distributed which could be due totheir lower deposition temperature, different Pt depositionrates, or total Pt coverage. Just for completion we mentionthat our low temperature deposition experiments �Tg

�100 K� yielded a much higher density of smaller sized Ptcluster.

We have also acquired series of images in order to ana-lyze the structural stability of the Pt islands �not shown�.Briefly, Pt islands observed in our study did never exhibit alateral motion between different regions of the moiré unitcell or across moiré unit cell borders, unless induced by ourSTM tip. This restriction does not apply to motion of Ptatoms within the cluster itself. Specifically, there is always aconsiderable fraction of clusters displaying a frizzy appear-ance, while others seem to be entirely stable; this character-istic is already in evidence in Fig. 7 and we have now en-hanced contrast to better emphasize this property and at thesame time clearly display the graphene overlayer underneathin Fig. 8. We find that stable islands for the most part remainstable while those islands with a frizzy appearance likewisetend to stay this way, even after prolonged observation.

We suggest that frizzy clusters contain a stable core plusone or two extra Pt atoms which are much less stronglybound and therefore mobile at 295 K. In accordance withFeibelman44 we expect stable clusters to contain only 120°corners, i.e., each Pt atom of these clusters is bonded to atleast three other Pt atoms. In case that an additional Pt isattached to such a magic cluster, it will have only two neigh-bors, leading to much reduced diffusion barriers and an ac-cordingly enhanced mobility.

Jumps of the extra atom�s� from one cluster to a neigh-boring one as indicated by a transfer of the island’s frizzyappearance do happen, although quite infrequently. We con-clude that single Pt atoms on graphene are unfavored withrespect to the Pt atoms within a cluster �e.g., compared tothose “satellite” atoms attached to stable cluster cores�. Thismeans that the suboptimal sites at the periphery of a clusterare still favored as compared to detachment. Assuming anattempt frequency of 1013 s−1 and kT�25 meV at roomtemperature, we estimate the relevant detachment activationenergy to be approximately 1.0 eV; about 10–15 events�transfer of frizzyness to a neighboring island� per 400 is-lands within an observation time of 200 s, which gives a rateR��1–2��10−4 s−1�. One example is provided by Figs. 8and 7�b�, displaying identical locations at different times��t=220 s�. Here, the three-layered island at the bottom ofthe images suddenly turns frizzy while all other islands re-mained unchanged.

It is intriguing that many islands of Fig. 8 look likeidentical twins. For some of the stable Pt islands we havetherefore added dots in order to indicate the number andlocation of Pt atoms within the cluster. It is apparent that themajority of monolayer clusters contains 12 Pt atoms and thatthe larger ones tend to be double layer islands.

D. Manipulation of Pt clusters on graphene/Ru„0001…using the STM tip

Finally, we like to address an issue which has puzzled usfor a while when we started imaging Pt clusters on grapheneoverlayers. Namely, the fact that extended empty areas maybe encountered despite ample Pt clusters being present inothers. By varying tunneling conditions, we realized that ourPt islands can easily be picked up by our Pt/Ir STM tip iftunneling resistance is too low, i.e., RT109 � and tunnel-ing currents exceed 100 pA. We therefore used currents be-low 50 pA with bias voltages of about 100 mV, which provedreasonably safe, even during extensive scanning. ReducingRT leads to an occasional pickup of islands which could also

TABLE I. Statistics of Pt islands grown on graphene/Ru�0001� at Tg=140–180 K at a deposition rate of 5�10−4 ML /s.

Averaged Pt coverage�Pt �ML�

Mean number sof Pt atoms per cluster

Filling factor �of graphene moiré

units cells

Percentageof monolayer

Pt islands

Percentageof Pt islands

exceeding monolayerheight

0.06 14 53 75 250.12 20 73 60 400.18 23 97 41 590.24 30 100 15 85

FIG. 8. STM image �200�165 Å2� of 0.12 ML Pt deposited ontographene/Ru�0001� at Tg=145 K and warmed up to 295 K for imaging. Thecolor scheme has been adjusted in order to better display the corrugation ofthe graphene template in parallel to the Pt islands occupying the fcc regionsof the graphene moiré overlayer. The displayed image location is identical toFig. 7�b�, and was taken 220 s earlier, using the same settings for tunneling�Ubias=0.1 V, I=50 pA�. The dots added to some of the clusters indicatethe number and location of Pt atoms within the first layer of the islands; thenumber of Pt layers of the four marked islands are �from left to right� two,three, one, and one.

164701-8 K. Donner and P. Jakob J. Chem. Phys. 131, 164701 �2009�

Downloaded 04 Nov 2009 to 137.248.1.11. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/jcp/copyright.jsp

be induced more selectively by applying voltage pulses atparticular locations. Again comparing Figs. 8 and 7�b�, werealize that an island in the upper right quarter of the imagehas disappeared. Possibly, the frizzy appearance has trig-gered contact with the tip and resulted in pickup whilescanning.

The direct observation of an island being picked upspontaneously by our tip while imaging is displayed in theseries of Figs. 9�a�–9�d�. All four images have been takenconsecutively with identical scanning settings, IT=1.09 nA,Ubias=15 mV �applied to the sample�. It is quite obvious thatpicking up the Pt island has elongated the STM tip in Fig.9�b�, leading to an offset of the z-scale. Larger scale imagesdemonstrating the removal of this particular Pt island areshown in �e�–�g�, with �e� taken before and �f� or �g� after theseries �a�–�d�. We note that in no instance did we ever ob-serve such a tip induced manipulation process for Pt islanddeposited on clean Ru�0001�. At present the pickup processhas been verified by different means:

�i� removal of an island while neighboring islands remainunchanged;

�ii� elongation of the Pt/Ir tip by a few angstroms;�iii� spontaneous deposition of previously acquired Pt

from our STM tip to nongraphene covered cleanRu�0001�. Re deposition onto graphene/Ru�0001� wasnot possible, unless strong voltage pulses �occasion-ally used to clean the tip� were applied.

An alternative process to pickup, again encountered atlow RT, was lateral dragging of Pt islands. In Fig. 9�f� anexample of such tip induced lateral displacement of a Ptisland �without pickup� is demonstrated; interestingly thisdragging occurs in both directions with respect to tip scan-ning, according to a detailed analysis of Fig. 9�f� line by line.Despite the seemingly clear conditions for their activation,both processes depicted in Figs. 9�a�–9�f� occur more or less

statistical which is why, for the time being, we restrainedfrom actually producing artificial structures on the nanometerscale this way.

IV. SUMMARY

The growth, structure, and adsorptive properties ofgraphene on Ru�0001� has been investigated in detail. Basedon the coherence of the graphene lattice extending over sev-eral Ru step edges, we have identified A and B types ofRu�0001� steps and determined fcc as well as hcp regions ofthe graphene/Ru�0001� moiré. This approach represents aconvenient way to discriminate fcc from hcp sites ofRu�0001�. We have further demonstrated the suitability ofgraphene/Ru�0001� to serve as a template for growing peri-odic arrays of uniform platinum nanoislands. Thereby Ptclusters are found to occupy exclusively fcc regions of thegraphene/Ru�0001� moiré unit cell. This finding is in accor-dance with expectations based on theoretical �DFT� modelcalculations performed for Ir clusters grown on graphene/Ir�111�. Pt islands are found to be structurally stable at roomtemperature; they may, however, exhibit some distinct frizzy-ness which is attributed to some of the islands consisting ofstable cores with one or two weakly bound Pt atoms addi-tionally attached. Finally, we have shown that Pt islands canbe readily picked up or laterally displaced by the STM tip attunneling resistances RT109 �.

APPENDIX: ANALYSIS OF THE GRAPHENE MOIRÉLATTICE DISCONTINUITY AT A AND B TYPEOF STEPS

The periodicities of the Ru�0001� substrate, graphenelattice, and graphene/Ru�0001� moiré superstructure are de-fined by the wave vectors k�Ru, k�C, and k�m, respectively, withkRu=2� /aRu=23.22 nm−1, kC=2� /aC=25.315 nm−1, andkm=2� /am=2.095 nm−1. With

k�m = k�C − k�Ru, �A1�

the phases associated with a particular displacement vector d�

will be Ru=k�Ru·d� , C=k�C·d� , and m=k�m ·d� . If d� is chosento represent a graphene moiré unit cell lattice vector, d� =a�m,then the equation

m = k�m · d� = k�C · d� − k�Ru · d� = C − Ru �A2a�

simplifies to

m = 2� = C − Ru = nC2� − nRu2� , �A2b�

i.e., nC=nRu+1. We will now introduce a monatomic stepwhich will lead to extra phase shifts of Ru and C due toeither the lateral displacement of substrate atoms for con-secutive substrate layers or the bending of the graphenesheet, respectively. Since the moiré lattice always tends toalign along the substrate atom rows, as do the step edges ofour Ru�0001� substrate we consider the directions perpen-dicular to the step edges only. All wave vectors ki

� then willbe smaller by a factor �3, e.g., kRu

� =kRu /�3=2� /�3aRu

=13.41 nm−1, while the distances necessary to induce aphase change of 2� increase by �3.

FIG. 9. ��a�–�d�� STM images �100�100 Å2 each� of a Pt island ongraphene/Ru�0001� that gets picked up by our STM tip during scanning.��a�–�d�� Tunneling current was IT=1.09 nA, bias voltage Ubias=15 mV.Larger scale images �350�300 Å2 each� are shown in �e�–�g�, with image�e� taken before, and �f� or �g� after the series �a�–�d�. In �f� lateral draggingof a Pt island along and opposite to the scanning direction is observed.Tunneling settings were IT=0.46 nA, Ubias=150 mV for �e�, andIT=0.65 nA, Ubias=71 mV for �f� or �g�. The horizontal lines in images �c�and �d� displaying the empty graphene/Ru�0001� overlayer are due to spon-taneous changes of the tip leading to slight apparent offsets of the imageplane.

164701-9 Pt on graphene/Ru�0001� J. Chem. Phys. 131, 164701 �2009�

Downloaded 04 Nov 2009 to 137.248.1.11. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/jcp/copyright.jsp

The step induced phase changes in Ru�0001� andgraphene lattices �labeled by a tilde� will be

RuA/B = Ru + � Ru

A/B, �A3�

with Ru=kRu� d� and � Ru

A/B=−kRu� �Ru

�A/B for A or B type ofsteps as well as

C = C + � C, �A4�

with C=kC�d� and � C=−kC

��C�. We define all displace-

ments in the step down direction positive which gives us aminus sign in the expressions for � Ru

A/B and � C. Due toboth effects the moiré lattice will now experience a 2� phasechange for a modified distance d��=am

�+�m�A/B and with

mA/B = C − Ru

A/B = C + � C − Ru − � RuA/B, �A5�

we obtain

mA/B = 2� = kC

��am� + �m

�A/B� − kRu� �am

� + �m�A/B�

− kC��C

� + kRu� �Ru

�A/B. �A6�

Since 2�=kC�am

�−kRu� am

� this expression simplifies to

0 = �kC� − kRu

� ��m�A/B − kC

��C� + kRu

� �Ru�A/B. �A7�

With km�=kC

�−kRu� we obtain our final result

km��m

�A/B = kC��C

� − kRu� �Ru

�A/B, �A8�

which agrees with the equation used by Coraux et al.8 exceptfor a minus sign of �m

� which we attribute to their inversedefinition of positive displacement vectors for �m

�.As the choice of moiré maxima in such an analysis is

kind of arbitrary, this equation needs an extra nA/B2� term onthe right hand side, with nA/B�Z, so that we end up with

km��m

�A/B = kC��C

� − kRu� �Ru

�A/B + nA/B2� . �A9�

1 P. J. Feibelman, Phys. Rev. B 59, 2327 �1999�.2 J. Repp, G. Meyer, K.-H. Rieder, and P. Hyldgaard, Phys. Rev. Lett. 91,206102 �2003�.

3 S. C. Wang and G. Ehrlich, Phys. Rev. Lett. 68, 1160 �1992�.4 B. C. Stipe, M. A. Rezaei, W. Ho, S. Gao, M. Persson, and B. I. Lund-qvist, Phys. Rev. Lett. 78, 4410 �1997�.

5 M. Lindroos, H. Pfnür, G. Held, and D. Menzel, Surf. Sci. 222, 451�1989�.

6 H. Pfnür, G. Held, M. Lindroos, and D. Menzel, Surf. Sci. 220, 43�1989�.

7 P. Jakob, M. Gsell, and D. Menzel, Phys. Rev. B 59, 13285 �1999�.8 J. Coraux, A. T. N’Diaye, C. Busse, and T. Michely, Nano Lett. 8, 565�2008�.

9 A. T. N’Diaye, J. Coraux, T. N. Plasa, C. Busse, and T. Michely, New J.Phys. 10, 043033 �2008�.

10 A. T. N’Diaye, S. Bleikamp, P. J. Feibelman, and T. Michely, Phys. Rev.Lett. 97, 215501 �2006�.

11 S. Marchini, S. Günther, and J. Wintterlin, Phys. Rev. B 76, 075429�2007�.

12 B. Wang, M.-L. Bocquet, S. Marchini, S. Gunther, and J. Wintterlin,

Phys. Chem. Chem. Phys. 10, 3530 �2008�.13 C. Berger, Z. Song, X. Li, X. Wu, N. Brown, C. Naud, D. Mayou, T. Li,

J. Hass, A. N. Marchenkov, E. H. Conrad, P. N. First, and W. A. de Heer,Science 312, 1191 �2006�.

14 A. K. Geim and K. S. Novoselov, Nature Mater. 6, 183 �2007�.15 S. Helveg, C. López-Cartes, J. Sehested, P. L. Hansen, B. S. Clausen, J.

R. Rostrup-Nielsen, F. Abild-Pedersen, and J. K. Nørskov, Nature �Lon-don� 427, 426 �2004�.

16 K. S. Novoselov, A. K. Geim, S. V. Morozov, D. Jiang, M. I. Katsnelson,I. V. Grigorieva, S. V. Dubonos, and A. A. Firsov, Nature �London� 438,197 �2005�.

17 P. J. Feibelman, Phys. Rev. B 77, 165419 �2008�.18 J. Wintterlin and M.-L. Bocquet, Surf. Sci. 603, 1841 �2009�.19 I. Pletikosic, M. Kralj, P. Pervan, R. Brako, J. Coraux, A. T. N’Diaye, C.

Busse, and T. Michely, Phys. Rev. Lett. 102, 056808 �2009�.20 D. E. Jiang, M. H. Du, and S. Dai, J. Chem. Phys. 130, 074705 �2009�.21 A. L. Vázquez de Parga, F. Calleja, B. Borca, M. C. G. Passeggi, Jr., J. J.

Hinarejos, F. Guinea, and R. Miranda, Phys. Rev. Lett. 100, 056807�2008�.

22 B. Wang, M.-L. Bocquet, S. Günther, and J. Wintterlin, Phys. Rev. Lett.101, 099703 �2008�.

23 A. L. Vázquez de Parga, F. Calleja, B. Borca, M. C. G. Passeggi, Jr., J. J.Hinarejos, F. Guinea, and R. Miranda, Phys. Rev. Lett. 101, 099704�2008�.

24 D. Martoccia, P. R. Willmott, T. Brugger, M. Bjorck, S. Gunther, C. M.Schleputz, A. Cervellino, S. A. Pauli, B. D. Patterson, S. Marchini, J.Wintterlin, W. Moritz, and T. Greber, Phys. Rev. Lett. 101, 126102�2008�.

25 Z.-P. Hu, D. F. Ogletree, M. A. Van Hove, and G. A. Somorjai, Surf. Sci.180, 433 �1987�.

26 A. B. Preobrajenski, M. L. Ng, A. S. Vinogradov, and N. Martensson,Phys. Rev. B 78, 073401 �2008�.

27 T. Brugger, S. Günther, B. Wang, J. H. Dil, M.-L. Bocquet, J. Oster-walder, J. Wintterlin, and T. Greber, Phys. Rev. B 79, 045407 �2009�.

28 P. W. Sutter, J.-I. Flege, and E. A. Sutter, Nature Mater. 7, 406 �2008�.29 Y. Pan, H. Zhang, D. Shi, J. Sun, S. Du, F. Liu, and H. J. Gao, Adv.

Mater. 21, 2777 �2009�.30 Y. Gamo, A. Nagashima, M. Wakabayashi, M. Terai, and C. Oshima,

Surf. Sci. 374, 61 �1997�.31 P. Jakob and D. Menzel, Surf. Sci. 235, 197 �1990�.32 E. Sutter, D. P. Acharya, J. T. Sadowski, and P. Sutter, Appl. Phys. Lett.

94, 133101 �2009�.33 T. A. Land, T. Michely, R. J. Behm, J. C. Hemminger, and G. Comsa,

Surf. Sci. 264, 261 �1992�.34 H. Zhang, Q. Fu, Y. Cui, D. L. Tan, and X. H. Bao, J. Phys. Chem. C

113, 8296 �2009�.35 H. Zhang, Q. Fu, Y. Cui, D. L. Tan, and X. H. Bao, Chin. Sci. Bull. 54,

2446 �2009�.36 P. Jakob and D. Menzel, Surf. Sci. 201, 503 �1988�.37 J. Coraux, A. T. N’Diaye, M. Engler, C. Busse, D. Wall, N. Buckanie,

F.-J. Meyer zu Heringdorf, R. van Gastel, B. Poelsema, and T. Michely,New J. Phys. 11, 023006 �2009�.

38 E. Loginova, N. C. Bartelt, P. J. Feibelman, and K. F. McCarty, New J.Phys. 10, 093026 �2008�.

39 E. Loginova, N. C. Bartelt, P. J. Feibelman, and K. F. McCarty, New J.Phys. 11, 063046 �2009�.

40 C. R. Stoldt, C. J. Jenks, P. A. Thiel, A. M. Cadilhe, and J. W. Evans, J.Chem. Phys. 111, 5157 �1999�.

41 M. Zinke-Allmang, L. C. Feldman, and M. H. Grabow, Surf. Sci. Rep.16, 377 �1992�.

42 P. Jakob, M. Gsell, and D. Menzel, J. Chem. Phys. 114, 10075 �2001�.43 A. T. N’Diaye, T. Gerber, C. Busse, J. Myslivecek, J. Coraux, and T.

Michely ”A versatile fabrication method for cluster superlattices” �un-published�.

44 P. J. Feibelman, Phys. Rev. B 80, 085412 �2009�.

164701-10 K. Donner and P. Jakob J. Chem. Phys. 131, 164701 �2009�

Downloaded 04 Nov 2009 to 137.248.1.11. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/jcp/copyright.jsp