# Fileset

[ncomms10745.pdf](https://mdr.nims.go.jp/filesets/34f35ccc-6740-4830-9dfb-3b663dcb852c/download)

## Creator

Patrick Gallagher, Menyoung Lee, Francois Amet, Petro Maksymovych, Jun Wang, Shuopei Wang, Xiaobo Lu, Guangyu Zhang, [Kenji Watanabe](https://orcid.org/0000-0003-3701-8119), [Takashi Taniguchi](https://orcid.org/0000-0002-1467-3105), David Goldhaber-Gordon

## Rights

[Creative Commons BY Attribution 4.0 International](https://creativecommons.org/licenses/by/4.0/)

## Other metadata

[Switchable friction enabled by nanoscale self-assembly on graphene](https://mdr.nims.go.jp/datasets/2c4273bd-ce81-4c77-bd3d-e457a4f6efe3)

## Fulltext

Switchable friction enabled by nanoscale self-assembly on grapheneARTICLEReceived 22 Dec 2015 | Accepted 13 Jan 2016 | Published 23 Feb 2016Switchable friction enabled by nanoscaleself-assembly on graphenePatrick Gallagher1, Menyoung Lee1, Francois Amet2,3, Petro Maksymovych4, Jun Wang4, Shuopei Wang5,Xiaobo Lu5, Guangyu Zhang5, Kenji Watanabe6, Takashi Taniguchi6 & David Goldhaber-Gordon1Graphene monolayers are known to display domains of anisotropic friction with twofoldsymmetry and anisotropy exceeding 200%. This anisotropy has been thought to originatefrom periodic nanoscale ripples in the graphene sheet, which enhance puckering around asliding asperity to a degree determined by the sliding direction. Here we demonstrate thatthese frictional domains derive not from structural features in the graphene but from self-assembly of environmental adsorbates into a highly regular superlattice of stripes with period4–6 nm. The stripes and resulting frictional domains appear on monolayer and multilayergraphene on a variety of substrates, as well as on exfoliated flakes of hexagonal boron nitride.We show that the stripe-superlattices can be reproducibly and reversibly manipulated withsubmicrometre precision using a scanning probe microscope, allowing us to create arbitraryarrangements of frictional domains within a single flake. Our results suggest a revisedunderstanding of the anisotropic friction observed on graphene and bulk graphite in terms ofadsorbates.DOI: 10.1038/ncomms10745 OPEN1 Department of Physics, Stanford University, Stanford, California 94305, USA. 2 Department of Physics, Duke University, Durham, North Carolina 27708,USA. 3 Department of Physics and Astronomy, Appalachian State University, Boone, North Carolina 28608, USA. 4 Center for Nanophase Materials Sciences,Oak Ridge National Laboratory, Oak Ridge, Tennessee 37831, USA. 5 Institute of Physics, Chinese Academy of Sciences, Beijing 100190, China. 6 NationalInstitute for Materials Science, 1-1 Namiki, Tsukuba 305-0044, Japan. Correspondence and requests for materials should be addressed to D.G.-G. (email:goldhaber-gordon@stanford.edu).NATURE COMMUNICATIONS | 7:10745 | DOI: 10.1038/ncomms10745 | www.nature.com/naturecommunications 1mailto:goldhaber-gordon@stanford.eduhttp://www.nature.com/naturecommunicationsNanometre-scale surface textures with long-range orderoften give rise to pronounced frictional anisotropy. Thesetextures sometimes originate from crystal structures:periodic tetrahedral reversals in the antigorite lattice createnanoscale surface corrugations, which generate the anisotropicfriction that governs certain seismic processes1. A large frictionalanisotropy similarly arises for some quasicrystal intermetallics,whose surfaces are textured by atomic columns2. Rotationallyaligned molecules also form ordered nanotextures with associatedanisotropic friction, as observed in organic crystals3. In adsorbedorganic films4–6, rotational symmetry of the host surface permitsmultiple stable molecular orientations, yielding frictional domainswith anisotropy along different axes.From a technological standpoint, nanometre-scale systemswith such multistability are appealing platforms for switches ormemories. Bistable states in redox centres7, rotaxane molecules8and iron clusters9 can be addressed and switched usingscanned probes, enabling dense information storage. Multistablenanotextures could find application in nanoelectromechanicalsystems if the friction-producing textures could be dynamicallycontrolled, as in biomimetic tapes with magnetically actuatedmicropillars10. Existing schemes for tuning friction atsubmicrometre scales include Fermi level modulation insilicon11 and mechanical oscillation of a sliding contact12—nonhysteretic techniques, which require maintenance of a voltageor oscillation, a disadvantage for circuitry.In this study, we identify a friction-producing nanotexture thatnaturally forms on graphene exposed to laboratory air and exploitits multistability to hysteretically switch friction with submicro-metre precision. Using high-resolution atomic force microscopy(AFM), we directly image a superlattice of nanoscale stripes onexfoliated graphene and we show that this striped nanotextureproduces the anisotropic friction previously observed13–15 ongraphene monolayers. This nanotexture strongly resemblespatterns of adsorbates observed on graphite16–18 and we inducean apparently identical nanotexture on flakes of hexagonal boronnitride (hBN) using a thermal cycling procedure. Consistent withthe adsorbate picture, we can rapidly and predictably reorient thefrictional domains by scanning a probe tip along the flake in achosen direction—a departure from nanoassembly techniques19such as dip-pen nanolithography20 and nanografting21, for whichwriting a different ‘colour’ requires submerging the sample in adifferent ‘ink’.ResultsSuperlattice of nanoscale stripes. To image friction, we measurethe deflection (diving board motion) and torsion (axial twist) of ascanned AFM cantilever in light contact with the sample. Thedeflection signal primarily contains topographic information,while the meaning of the torsion signal depends on scan direc-tion. For lateral scanning (motion perpendicular to cantileveraxis; Fig. 1b lower panel), the torsion measures lateral tip-sampleforces commonly interpreted as friction forces. In this ‘friction-imaging’ mode, tip-sample forces transverse to the scan directionresult in deflection, contributing spurious topographic signals.When the cantilever is scanned longitudinally (Fig. 1c lowerpanel), the torsion signal directly measures tip-sample forcestransverse to the scan direction. For an isotropic surface, this‘transverse force’ signal is zero.As reported previously13,14, the friction signal of exfoliatedmonolayer graphene flakes on silicon dioxide reveals up to threedistinct domains of friction despite a featureless topographysignal (Fig. 1a,b). The domains vary in size from tens ofnanometres to tens of micrometres and produce sharp contrast intransverse force, confirming their anisotropic character (Fig. 1c).Tapping-mode AFM images taken with ultrasharp tips within thedifferent domains (Fig. 1d) reveal periodic stripes along axesrotationally separated by 60� (angular orientation does notmeasurably vary within a given domain; see SupplementaryNote 1). Within experimental error (typically ±0.2 nm), stripeperiod (typically B4 nm) does not change across a sample,although we have observed global changes in stripe period afterthermal cycling (for example, from 4 to 6 nm; see SupplementaryNote 2). Peak-to-trough stripe amplitude ranges between 10 and100 pm, but strongly depends on tip conditions and oscillationparameters.The observed frictional anisotropy of a given domain respectsthe symmetry of the stripe-superlattice. The friction signalapproximately tracks the cosine of the angle between scan axisand stripes (Fig. 1e)—friction is maximized when the two arealigned—whereas the transverse force is zero when the stripes areperpendicular or parallel to the scan axis, as required bysymmetry (Fig. 1f). In between these zeros, the transverse forcechanges sign so as to guide the sliding tip towards the low frictionaxis (lower panel in Fig. 1c). We conclude that the stripes ingraphene produce the observed friction anisotropy, similar tofriction-producing nanotextures in other systems1–6.The stripes are not unique to monolayer graphene on SiO2. Weobserve stripe domains and anisotropic friction on grapheneflakes up to 50 nm thick (the maximum thickness investigated)without change in stripe period and with minor change inmagnitude of frictional anisotropy (Supplementary Note 3), aswell as on graphene flakes on different substrates (SupplementaryNote 4). Stripe domains can also form on exfoliated flakes of hBNon SiO2: single crystals show at most three distinct domains ofanisotropic friction (Fig. 2b,c), each characterized by a differentorientation of stripes, whose typical period is B4 nm (Fig. 2d). Asfor graphene, the friction signal is maximized when scanningalong the stripes. However, whereas we observed stripes on nearlyall graphene flakes as exfoliated, only occasionally did we observestripes on hBN as exfoliated. We found that a cryogenic thermalcycle such as immersion in liquid nitrogen and subsequentremoval to ambient conditions (Methods) would reliably producestripes on hBN. A full understanding of the effect of thermalcycling is beyond the scope of our study; our limited variable-temperature AFM experiments found stripes to form on hBN oncooling from 300 to 250 K, although vacuum conditions probablyinfluenced the evolution with temperature (SupplementaryNote 5).The behaviour of stripes on epitaxial heterostructures ofgraphene and hBN implies that stripes on both materials share acommon origin. The nearly perfect rotational alignment betweenstacked lattices22 results in a moiré pattern with lattice constantB14 nm in regions where graphene has grown on the hBN(Fig. 3a). Despite this additional superstructure, stripes form onexposed layers of both graphene and hBN with no measurabledifference in period and often appear to maintain phase across agraphene/hBN boundary. Furthermore, using the moiré patternto infer lattice orientation23 (Fig. 3b), we find that the stripes runalong the armchair axes of both graphene and hBN in all25 epitaxial heterostructures and 5 mechanically assembledheterostructures that we studied (Supplementary Note 6).Previous studies have ascribed the anisotropic friction inmonolayer graphene to periodic ripples in the graphene sheetinduced by stress from the substrate13–15. Although our dataconfirm the presence of periodic structure, the extreme similarityof the stripes on graphene and hBN—materials with differentbending stiffness and response to stress24—suggests that thestripes are adsorbates rather than features of the crystalsthemselves. The orientation of the stripes further rules outperiodic ripples, which are expected to produce a high frictionARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms107452 NATURE COMMUNICATIONS | 7:10745 | DOI: 10.1038/ncomms10745 | www.nature.com/naturecommunicationshttp://www.nature.com/naturecommunicationsTopography (nm)1 0 –1Friction, relative to SiO20.15 0.25 0.35Transverse force (arb.)–40 0 400.50–0.5Topography (nm)Torsion TorsionTransverse forceFriction force Scan90°090°0ScanEasyI II I IIIII IIIIIIIIIIIIIIIIIIIIISiO2MonolayerBilayerSample rotationSample rotationa b cde fFigure 1 | Stripes on exfoliated graphene. (a) Contact mode topography scan of a graphene flake on silicon oxide, showing monolayer, bilayer and trilayerregions. Scale bar, 3mm. (b) Simultaneously recorded friction signal (upper panel), showing three distinct domains of friction labeled I, II and III. Lowerpanel: cartoon of the friction imaging mode. The cantilever is scanned laterally and friction between the tip and sample produces the measured torsion ofthe cantilever. (c) Transverse force signal (upper panel) from the same region as in b, measured by recording the torsion while scanning the cantileverlongitudinally (lower panel). Surface anisotropy pushes the tip towards the local ‘easy’ axis, creating a transverse force that twists the cantilever.(d) Tapping mode topography scans of the graphene monolayer, taken within each of the three domains. Each domain is characterized by stripes of period4.3±0.2 nm along one of three distinct axes rotationally separated by 60�. Scale bars, 20 nm. (e) Friction relative to SiO2 for each domain as a function ofclockwise sample rotation angle; zero degrees corresponds to the orientation shown in a–c. For each polar plot, the origin and circumference correspond torelative friction values of 0.15 and 0.4, respectively. Dotted lines indicate the sample rotations at which the stripes shown in d are parallel to the scan axis.The friction signal is approximately sinusoidal, with the highest friction produced when stripes are parallel to the scan axis. (f) Transverse force signal foreach domain as a function of clockwise sample rotation angle. Unshaded and grey-shaded regions indicate positive and negative transverse signals,respectively. The origin of each polar plot is zero. The transverse signal for a given domain switches sign as the stripe axis rotates through the lateral axis.Topography (nm)8 4 0 0.11 0.18 0.25Friction, relative to SiO2 Transverse force (arb.)25 0 –250.20–0.2Topography (nm)II IIIII IIII IIIII IIa b c dFigure 2 | Stripes on exfoliated hBN. (a) Contact mode topography scan of a terraced hBN flake, thickness 5–9 nm, after thermal cycling in liquid nitrogen.Scale bar, 5 mm. (b,c) Simultaneously recorded friction signal (b) and separately recorded transverse force signal (c) showing the presence of three distinctdomains (I, II and III). The contrast between I and III is weak in friction, but strong in transverse force. (d) Tapping mode topography scans of the threedomains, taken in the regions indicated in b and c. Each domain is characterized by stripes of period 4.7±0.2 nm along one of three distinct axesrotationally separated by 60�. Scale bars, 20 nm.NATURE COMMUNICATIONS | DOI: 10.1038/ncomms10745 ARTICLENATURE COMMUNICATIONS | 7:10745 | DOI: 10.1038/ncomms10745 | www.nature.com/naturecommunications 3http://www.nature.com/naturecommunicationsaxis perpendicular to the stripes13,14 and a zigzag stripeaxis15,25—both opposite to our findings. Periodic ripples havenever been observed in scanning tunnelling microscopy (STM) ofthe graphene lattice and our STM data are no exception(Supplementary Note 7); on the other hand, adsorbates can bedisturbed by the pressure of the STM tip under standard imagingconditions26, perhaps explaining why the stripes that we observehave not previously been reported.Various organic adsorbates are known to self-assemble intonanoscale stripes on graphite. Surfactant molecules, for instance,form stripes17 whose 4–7 nm period is set by molecular lengthand Debye screening27; anisotropic van der Waals interactionsalign the stripes along the armchair axes17. Alkanes also producearmchair-aligned stripes of 4 nm period on graphite, where theperiod is again determined largely by molecular length16.Self-assembly of inorganic species has been reported as well: Luet al.18,28 observed crystallographically aligned stripes of 4 nmperiod on graphite submerged in water and correlated the stripeswith the presence of dissolved nitrogen gas. Noting that gasenrichment at the interface between a hydrophobic surface andwater is theoretically expected29, Lu et al.18,28 argued that thestripes were self-assembled columns of molecular nitrogenadsorbed to the graphite surface. Stripes of similar period werelater observed in ambient on multilayer epitaxial graphene30,31;following Lu et al.18,28, these stripes were attributed to nitrogenadsorbates trapped at the interface between graphene and anambient water layer. We note that these studies18,28,30,31 do notprovide a direct chemical analysis of the stripes to prove theirnitrogen content. Why stripes should form instead of ahomogeneous layer of nitrogen is also unexplained.We propose that the stripes on graphene and hBN are self-assembled environmental adsorbates, in view of their similarity tostripes formed by adsorbates on graphitic surfaces16–18,27,28,30,31and their aforementioned dissimilarity to structural ripples, aswell as our ability to manipulate the stripes by physical contact(see below). Although direct determination of the chemicalmakeup of the stripes is beyond the scope of our work, our datasuggest that the species that self-assembles is airborne andubiquitous in the laboratory, as stripes of uniform period fullycover our cleaved or annealed crystal surfaces that have not beenexposed to any chemical processing (Methods). From thisperspective, an interpretation in terms of nitrogen andwater18,28,30,31 or other common inorganic molecules isattractive. However, hydrocarbons are also plentiful inlaboratory air (arising from, for instance, outgassing plastics orvacuum pump oil) and certain species could preferentially attachto graphene or hBN due to lattice match16.Recent work resolved nanoscale stripes in the transverse forceresponse of bulk graphite and ascribed them to a novelpuckering-induced stick–slip friction process32. These stripesproduced domains of anisotropic friction33 such as those ongraphene and hBN. We suggest a reinterpretation of these data interms of adsorbates, which would unify our understanding ofanisotropic friction in graphite, graphene and hBN.Manipulation of frictional domains. Adsorbates can sometimesbe mechanically manipulated by AFM19, raising the possibility ofpatterning friction on these materials. For monolayer grapheneon SiO2, scanning at the low normal force used for imaging(1 nN) often minimally affects the frictional domains, butscanning at high normal force (30 nN) reproducibly reorientsthe domains (Fig. 4a). We devised two standard approaches fordomain manipulation (Fig. 4b). The ‘brush stroke’ consists ofraster scanning a rectangular window at high normal force; weretract the tip after every line so that it only scans the sample inone direction. Brush strokes produce reproducible results—oftena domain flop—that depend on the scan angle and the initial‘canvas’ domain. For scan angles near the canvas stripe axis, thecanvas switches to the domain with stripes next closest to the scanaxis (Fig. 4c). Our second approach is to ‘erase’ the canvasdomain within a rectangular scan window by rapid, back-and-forth scanning at high normal force. This mode destabilizes thedomains within the scan window, leaving only the most stabledomain, determined primarily by local strain and partly by scanaxis. Although strain gradually varies across the flake (seediscussion below), erasing still produces deterministic resultswithin a specific region.The brush stroke and eraser allow us to rapidly create patternsof friction with submicrometre precision. Without optimizing ourprocedure, creating a block letter ‘S’ 5 mm tall using the erasertook 16 min, whereas creating a ‘U’ using brush strokes took36 min (Fig. 4d and Supplementary Movie 1). After writing, thepattern gradually decayed: here the ‘S’ widened, while the ‘U’narrowed (Fig. 4e). We wrote the same pattern in different partsof the flake and found that whether a domain grew or shrank withtime, and how rapidly it evolved, depended on position. Theabsence of other obvious symmetry-breaking mechanismssuggests that local strain induced by the substrate determinesStripesMoiré26°abFigure 3 | Orientation of stripes on graphene and hBN. (a) Tapping modetopography image of graphene islands grown by van der Waals epitaxy onexfoliated hBN. The image has been differentiated along the horizontal axisfor clarity. Graphene islands can be distinguished from the hBN surface bythe presence of a moiré pattern, which is partially outlined in black for oneof the grains. The sample surface is covered with stripes of period4.3±0.1 nm, oriented along one of three distinct axes rotationally separatedby 60�. The stripe period is the same on graphene and hBN, and the stripesfrequently appear to cross the graphene/hBN boundary without a phaseslip. Scale bar, 50 nm. (b) Fast Fourier transform (FFT) of the topographysignal used to produce a. The moiré pattern within the graphene grainsappears as a sixfold-symmetric pattern with segments extendingB70mm� 1 from the origin; these protruding segments are parallel to themomentum-space moiré lattice vectors. The dominant stripe domain ongraphene and hBN produces a pair of isolated points in the FFT, one ofwhich is circled in black. The stripe axis is rotated 26±4� from the moirélattice vectors, indicating that the stripe axes are nearly aligned with thearmchair axes of the graphene and hBN. The quoted angular precisionreflects the width of the moiré peaks; we also expect a few-degreesystematic error in the angular estimate, as a misalignment betweengraphene and hBN lattices of 0.1�—a reasonable expectation for van derWaals epitaxial heterostructures23—would rotate the moiré pattern by 4�with respect to the graphene lattice. The small area of nearly vertical stripesin a produces a pair of points, circled in red, which can barely be seen withthis colourscale. Scale bar, 100 mm� 1.ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms107454 NATURE COMMUNICATIONS | 7:10745 | DOI: 10.1038/ncomms10745 | www.nature.com/naturecommunicationshttp://www.nature.com/naturecommunicationsthe relative stability of the domains. Domain stability in turndetermines the effective resolution of our patterning technique:although we can write crisp lines 100 nm wide in some parts of aflake, in other parts these features only persist for minutes beforedecaying to match the canvas domain. In addition, although wecan pattern friction on several different monolayer grapheneflakes, others show only weak response to both patterning modesdescribed; the strain field in these flakes probably strongly favoursthe local canvas domain, making the canvas difficult to switch.Whether patterning friction is possible on thicker crystalsrequires further investigation. Our first attempts indicate thatdomains can be rewritten with the eraser or brush stroke,although the resulting domains are not as sharp as on monolayergraphene. Proximity to the substrate could be stabilizing thestripes, allowing for more flexible control of domain shape. Ourwork underscores the major role played by adsorbates, ratherthan structural deformation, in determining friction on grapheneand hBN—and perhaps on other layered materials, such astransition metal dichalcogenides. The periodic perturbation fromthe adsorbates might open gaps at the superlattice energy ormodify the Fermi velocity in graphene34, with measurableconsequences for electronic properties of ultraclean graphene/hBN heterostructures35.MethodsSample preparation. Flakes of graphene and hBN were prepared by mechanicalexfoliation (3M Scotch 600 Transparent Tape or 3M Scotch 810 Magic Tape) underambient conditions (40–60% relative humidity) on n-doped silicon wafers with 90or 300 nm of thermal oxide. The substrates were not exposed to any chemicalprocessing following thermal oxidation. For graphene exfoliation, we used bulkcrystals of both Kish graphite (Sedgetech, USA) and highly oriented pyroliticgraphite (HOPG ZYA, SPI Supplies, USA) and observed no differences insuperlattice phenomena between samples produced using different graphitesources or tapes. For hBN exfoliation, we used bulk crystals provided by KenjiWatanabe and Takashi Taniguchi. We also prepared graphene flakes on othersubstrates (Supplementary Note 4), including SU-8 epoxy (MicroChem, USA),200 nm of Au(111) on mica (Phasis, Switzerland) and 5 nm of Pt (electron-beamevaporation) on magnesium oxide (MTI, USA).We prepared epitaxial graphene heterostructures on oxidized silicon substratesby mechanical exfoliation of hBN followed by graphene growth at 500 �C bya remote plasma-enhanced chemical vapour deposition process describedpreviously22. We also mechanically assembled heterostructures of graphene onhBN using both wet36 and dry37 transfer methods. Polymer residues from theassembly process were removed by annealing samples in a tube furnace for 4 h at500 �C under continuous flow of oxygen (50 sccm) and argon (500 sccm); beforeremoval to air, we allowed the samples to cool (5–10 �C min� 1), to below 100 �Cunder the same flow of oxygen and argon.Thermal cycling. We found stripes to appear on our samples after thermal cyclingto liquid nitrogen temperatures or below using a variety of methods. Mostcommonly, and specifically for the sample shown in Fig. 2, we immersed thesample in liquid nitrogen for 1–5 min and then removed it to atmosphere, and blewoff the condensation with dry air. This procedure would almost always producestripes on graphene, hBN or graphene/hBN heterostructures. In other cases, weloaded the sample in the vacuum chamber of a cryostat—either a cryogen-freedilution refrigerator or a Quantum Design PPMS—and thermal cycled to a basetemperature between 25 mK and 100 K. Cooling and warming rates varied between1 and 30 K min� 1. We warmed up the samples under various atmospheresincluding moderate vacuum, helium gas or nitrogen gas; in all of these cases (overten different samples cycled in the dilution refrigerator or PPMS) we found stripeson every flake or heterostructure (totaling several tens) that we checked.The epitaxial heterostructure in Fig. 3 was not cycled to low temperature: thesample displayed stripes in AFM with no further processing following removalfrom the growth furnace. Some of our assembled heterostructures (SupplementaryNote 6) required a low-temperature thermal cycle to produce stripes after theoxygen/argon anneal, although in other cases we observed stripes withoutcryogenic treatment.AFM and STM measurements. All images shown in Figs 1–4 were taken with aPark XE-100 AFM under ambient conditions (40–60% relative humidity) exceptfor Fig. 4d,e, which were taken in 10% relative humidity by flooding the chamber ofthe XE-100 with dry air. (We observed no significant difference in domain mut-ability or evolution between 10 and 50% relative humidity.) To resolve the stripesin tapping mode, we used sharp silicon probes (MikroMasch Hi’Res-C15/Cr-Au)with a nominal tip radius of 1 nm, a typical resonant frequency of 265 kHz and aLow normal force High normal force ImagingBrushingErasingAfter 90 minIIIIIIa bc d eFigure 4 | Rewritable friction on monolayer graphene. (a) Cartoon illustrating the response of the striped adsorbates to the scanning tip. At low normalforce, the tip minimally disturbs the stripes as it scans the surface and the stripe structure rapidly heals. At high normal force, the stripe structure is heavilydisturbed, creating a new stripe domain in the wake of the scanning tip. (b) Summary of our scanning modes. For imaging, we rapidly scan the cantileverback and forth at low normal force, while slowly moving it in the direction perpendicular to the fast scan axis. The erasing mode is identical, but at highnormal force. For a brush stroke, we raster-scan the cantilever such that the tip only moves in one direction when in contact with the sample. After scanningeach line, we lift the cantilever, move it to the start of the next line and touch down again. (c) Domain switching as a function of scan angle on themonolayer flake studied in Fig. 1, rotated as in Fig. 1a–c. The image shown is a collage of 12 transverse force images, each taken after executing a single 3 mmby 1mm brush stroke on a canvas composed initially of a single domain. For each canvas domain we show four brush strokes nearly parallel with the canvasstripes, where each brush stroke is directed radially outward from the origin of the semicircle. The brush strokes steer the canvas domain towards thedomain whose stripes are next nearest the brush axis. Scale bar, 3mm. (d) Transverse force image immediately after writing block letters ‘S’ and ‘U’ indomains III and I, respectively, on a canvas of domain II (same flake and orientation as in a). The block letter ‘S’ was written by ‘erasing’, whereas the ‘U’was written using brush strokes. Scale bar, 3 mm. (e) Transverse force image of the same area, taken 90 min later. The ‘S’ (domain III) has expanded into thecanvas, while the ‘U’ (domain I) has decayed.NATURE COMMUNICATIONS | DOI: 10.1038/ncomms10745 ARTICLENATURE COMMUNICATIONS | 7:10745 | DOI: 10.1038/ncomms10745 | www.nature.com/naturecommunications 5http://www.nature.com/naturecommunicationstypical cantilever Q of 400. See Supplementary Note 8 for a detailed interpretationof the tapping mode topography signal.For measurements in contact mode, we used silicon probes (MikroMaschHQ:NSC19/Al BS-15) with a nominal tip radius of 8 nm and a typical resonantfrequency of 65 kHz. We used a normal force setpoint of 1 nN for all friction andtransverse force imaging scans shown, with scan rates B10 mm s� 1. SeeSupplementary Note 8 for a discussion of the friction imaging mechanism. Fordomain manipulation we used a normal force setpoint of 30 nN. For brush strokeswe used scan rates B30 mm s� 1, whereas for erasing we used scan ratesB300mm s� 1.When imaging friction or transverse force, we collected torsion data for bothforward-moving and backward-moving scans. To eliminate offsets in the frictionand transverse force signals for Fig. 1e,f, we subtracted backward images fromforward images and divided by two. All friction or transverse force images shownare just the forward scan, with any torsion offset eliminated by subtracting theaverage of forward and backward torsion values on SiO2.To study stripe formation with changing temperature (Supplementary Note 5),we used an Omicron varible-temperature AFM/STM operating in ultrahighvacuum (UHV; 8� 10� 11 mbar). Samples were not baked in UHV beforeexperiments. The sample stage was cooled by a copper braid attached to a cold sinkheld at low temperature by continuous flow of liquid nitrogen; by this method, weachieved a base temperature of 110 K. We used the same sharp probes as forambient AFM (MikroMasch Hi’Res-C15/Cr-Au). In UHV, the cantilever Qreached 5,000, which significantly restricted scan speed for tapping mode; wetherefore used on-resonance frequency-modulation mode, imaging at a typicalfrequency shift of � 30 Hz. For all images, we applied a DC tip-sample bias tonullify the contact potential difference.STM measurements (Supplementary Note 7) were carried out under ambientconditions using the Park XE-100. We prepared our tip by mechanically cutting aPt/Ir wire and scanning the sample at high bias voltages until we achieved atomicresolution of the graphene lattice.Error bars and lateral calibration. All values quoted for moiré period and angularorientation are extracted from the fast Fourier transform of the AFM images. AFMimages of all heterostructures described in this study are corrected for thermal driftby performing an affine transformation to produce regular moiré hexagons(we used the free software Gwyddion, available at gwyddion.net). All error barsreflect the full width at half maximum of the peaks in the fast Fourier transform;for instance, 12.0±0.5 nm means that the full width at half maximum of the peakmaps to 1 nm in real space. The lateral scale of the Park XE-100 was calibratedby measuring the moiré period of graphene/hBN heterostructures grownby van der Waals epitaxy, in which the graphene and hBN lattices are nearlyperfectly aligned, and defining this period (averaged over several samples) to be13.6 nm. This definition corresponds to the assumption made in SupplementaryNote 6 that the lattice constants for hBN and graphene are ahBN¼ 0.25 nm andagraphene¼ ahBN/1.018. The lateral scale of the Omicron variable-temperature AFMwas calibrated to the lateral scale of the Park XE-100 by measuring the moirépattern of the same sample in both systems.References1. Campione, M. & Capitani, G. C. Subduction-zone earthquake complexityrelated to frictional anisotropy in antigorite. Nat. Geosci. 6, 847–851 (2013).2. Park, J. Y. et al. High frictional anisotropy of periodic and aperiodic directionson a quasicrystal surface. Science 309, 1354–1356 (2005).3. Bluhm, H., Schwarz, U. D., Meyer, K.-P. & Wiesendanger, R. Ansiotropy ofsliding friction on the triglycine sulfate (010) surface. Appl. Phys. A 61, 525–533(1995).4. Overney, R. M., Takano, H., Fujihara, M., Paulus, W. & Ringsdorf, H.Ansiotropy in friction and molecular stick-slip motion. Phys. Rev. Lett. 72,3546–3549 (1994).5. Last, J. A. & Ward, M. D. Electrochemical annealing and friction anisotropy ofdomains in epitaxial molecular films. Adv. Mater. 8, 730–733 (1996).6. Liley, M. et al. Friction anisotropy and asymmetry of a compliant monolayerinduced by a small molecular tilt. Science 280, 273–275 (1998).7. Gittins, D. I., Bethell, D., Schiffrin, D. J. & Nichols, R. J. A nanometre-scaleelectronic switch consisting of a metal cluster and redox-addressable groups.Nature 408, 67–69 (2000).8. Cavallini, M. et al. Information storage using supramolecular surface patterns.Science 299, 531 (2003).9. Loth, S., Baumann, S., Lutz, C. P., Eigler, D. M. & Heinrich, A. J. Bistability inatomic-scale antiferromagnets. Science 335, 196–199 (2012).10. Northen, M. T., Greiner, C., Arzt, E. & Turner, K. L. A gecko-inspiredreversible adhesive. Adv. Mater. 20, 3905–3909 (2008).11. Park, J. Y., Ogletree, D. F., Thiel, P. A. & Salmeron, M. Electronic control offriction in silicon pn junctions. Science 313, 186 (2006).12. Socoliuc, A. et al. Atomic-scale control of friction by actuation of nanometer-sized contacts. Science 313, 207–210 (2006).13. Choi, J. S. et al. Friction anisotropy-driven domain imaging on exfoliatedmonolayer graphene. Science 333, 607–610 (2011).14. Choi, J. S. et al. Facile characterization of ripple domains on exfoliatedgraphene. Rev. Sci. Instr. 83, 073905 (2012).15. Choi, J. S. et al. Correlation between micrometer-scale ripple alignment andatomic-scale crystallographic orientation of monolayer graphene. Sci. Rep. 4,7263 (2014).16. McGonigal, G. C., Bernhardt, R. H. & Thomson, D. J. Imaging alkane layers atthe liquid/graphite interface with the scanning tunneling microscope. App.Phys. Lett. 57, 28–30 (1990).17. Manne, S. & Gaub, H. E. Molecular organization of surfactants at solid-liquidinterfaces. Science 270, 1480–1482 (1995).18. Lu, Y.-H., Yang, C.-W. & Hwang, I.-S. Molecular layer of gaslike domains at ahydrophobic water interface observed by frequency-modulation atomic forcemicroscopy. Langmuir 28, 12691–12695 (2012).19. Tseng, A. Tip-Based Nanofabrication: Fundamentals and Applications(Springer, 2011).20. Piner, R. D., Zhu, J., Xu, F., Hong, S. & Mirkin, C. A. ‘Dip-pen’nanolithography. Science 283, 661–663 (1999).21. Xu, S. & Liu, G.-Y. Nanometer-scale fabrication by simultaneous nanoshavingand molecular self-assembly. Langmuir 13, 127–129 (1997).22. Yang, W. et al. Epitaxial growth of single-domain graphene on hexagonal boronnitride. Nat. Mater. 12, 792–797 (2013).23. Tang, S. et al. Precisely aligned graphene grown on hexagonal boron nitride bycatalyst free chemical vapor deposition. Sci. Rep. 3, 2666 (2013).24. Singh, S. K., Neek-Amal, M., Costamagna, S. & Peeters, F. M.Thermomechanical properties of a single hexagonal boron nitride sheet. Phys.Rev. B 87, 184106 (2013).25. Ma, T., Li, B. & Chang, T. Chirality- and curvature-dependent bending stiffnessof single layer graphene. Appl. Phys. Lett. 99, 201901 (2011).26. Magonov, S. & Whangbo, M.-H. Surface Analysis with STM and AFM:Experimental and Theoretical Aspects of Image Analysis (Wiley, 2008).27. Wanless, E. J. & Ducker, W. A. Organization of sodium dodecyl sulfate at thegraphite-solution interface. J. Phys. Chem. 100, 3207–3214 (1996).28. Lu, Y.-H., Yang, C.-W. & Hwang, I.-S. Atomic force microscopy study ofnitrogen molecule self-assembly at the HOPG-water interface. Appl. Surf. Sci.304, 56–64 (2014).29. Dammer, S. M. & Lohse, D. Gas enrichment at liquid-wall interfaces. Phys. Rev.Lett. 96, 206101 (2006).30. Wastl, D. S. et al. Observation of 4 nm pitch stripe domains formed by exposinggraphene to ambient air. ACS Nano 7, 10032–10037 (2013).31. Wastl, D. S., Weymouth, A. J. & Giessibl, F. J. Atomically resolved graphiticsurfaces in air by atomic force microscopy. ACS Nano 8, 5233–5239 (2014).32. Rastei, M. V., Heinrich, B. & Gallani, J. L. Puckering stick-slip friction inducedby a sliding nanoscale contact. Phys. Rev. Lett. 111, 084301 (2013).33. Rastei, M. V., Guzmán, P. & Gallani, J. L. Sliding speed-induced nanoscalefriction mosaicity at the graphite surface. Phys. Rev. B 90, 041409 (2014).34. Park, C.-H., Yang, L., Son, Y.-W., Cohen, M. L. & Louie, S. G. Anisotropicbehaviours of massless Dirac fermions in graphene under periodic potentials.Nat. Phys. 4, 213–217 (2008).35. Dean, C. R. et al. Boron nitride substrates for high-quality graphene electronics.Nat. Nanotechnol. 5, 722–726 (2010).36. Amet, F., Williams, J. R., Watanabe, K., Taniguchi, T. & Goldhaber-Gordon, D.Insulating behavior at the neutrality point in single-layer graphene. Phys. Rev.Lett. 110, 216601 (2013).37. Wang, L. et al. One-dimensional electrical contact to a two-dimensionalmaterial. Science 342, 614–617 (2013).AcknowledgementsWe gratefully acknowledge Byong-man Kim and Ryan Yoo of Park Systems for verifyingthe presence of stripes in our samples using their Park NX-10 AFM. We thank DanielWastl for carefully reading our manuscript and for encouraging us to re-examinewhether the stripes we observed were caused by periodic structural ripples orself-assembled adsorbates. We thank Trevor Petach and Arthur Barnard for other helpfuldiscussions. Sample fabrication and ambient AFM/STM were performed at the StanfordNano Shared Facilities with support from the Air Force Office of Science Research,Award Number FA9550-12-1-02520. Variable-temperature AFM studies were conductedat the Center for Nanophase Materials Sciences, which is a DOE Office of Science UserFacility; our use of the facility was supported by the Center for Probing the Nanoscale,an NSF NSEC, under grant PHY-0830228. S.W., X.L. and G.Z. acknowledge support fromthe National Basic Research Program of China (Program 973) under grant 2013CB934500,the National Natural Science Foundation of China under grants 61325021 and 91223204,and the Strategic Priority Research Program (B) of the Chinese Academy of Sciencesunder grant XDB07010100. K.W. and T.T. acknowledge support from the ElementalStrategy Initiative conducted by the MEXT (Japan). T.T. acknowledges support fromJSPS Grant-in-Aid for Scientific Research under grants 262480621 and 25106006.Author contributionsP.G. identified the stripes, performed all experiments and wrote the paper. P.G. andF.A. fabricated the assembled heterostructures. M.L., F.A. and D.G.-G. discussed dataARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms107456 NATURE COMMUNICATIONS | 7:10745 | DOI: 10.1038/ncomms10745 | www.nature.com/naturecommunicationshttp://www.nature.com/naturecommunicationsand experimental directions, and assisted in writing the paper. P.M. and J.W.supported the variable-temperature AFM measurements. S.W., X.L. and G.Z.grew the epitaxial graphene/hBN heterostructures. K.W. and T.T. grew the bulkhBN crystals.Additional informationSupplementary Information accompanies this paper at http://www.nature.com/naturecommunicationsCompeting financial interests: The authors declare no competing financialinterests.Reprints and permission information is available online at http://npg.nature.com/reprintsandpermissions.How to cite this article: Gallagher, P. et al. Switchable friction enabled by nanoscaleself-assembly on graphene. Nat. Commun. 7:10745 doi: 10.1038/ncomms10745 (2016).This work is licensed under a Creative Commons Attribution 4.0International License. The images or other third party material in thisarticle are included in the article’s Creative Commons license, unless indicated otherwisein the credit line; if the material is not included under the Creative Commons license,users will need to obtain permission from the license holder to reproduce the material.To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/NATURE COMMUNICATIONS | DOI: 10.1038/ncomms10745 ARTICLENATURE COMMUNICATIONS | 7:10745 | DOI: 10.1038/ncomms10745 | www.nature.com/naturecommunications 7http://www.nature.com/naturecommunicationshttp://www.nature.com/naturecommunicationshttp://npg.nature.com/reprintsandpermissionshttp://npg.nature.com/reprintsandpermissionshttp://creativecommons.org/licenses/by/4.0/http://www.nature.com/naturecommunications title_link Results Superlattice of nanoscale stripes Figure™1Stripes on exfoliated graphene.(a) Contact mode topography scan of a graphene flake on silicon oxide, showing monolayer, bilayer and trilayer regions. Scale bar, 3thinspmgrm. (b) Simultaneously recorded friction signal (upper panel), showing three Figure™2Stripes on exfoliated hBN.(a) Contact mode topography scan of a terraced hBN flake, thickness 5-9thinspnm, after thermal cycling in liquid nitrogen. Scale bar, 5thinspmgrm. (b,c) Simultaneously recorded friction signal (b) and separately recorded  Manipulation of frictional domains Figure™3Orientation of stripes on graphene and hBN.(a) Tapping mode topography image of graphene islands grown by van der Waals epitaxy on exfoliated hBN. The image has been differentiated along the horizontal axis for clarity. Graphene islands can be dis Methods Sample preparation Thermal cycling AFM and STM measurements Figure™4Rewritable friction on monolayer graphene.(a) Cartoon illustrating the response of the striped adsorbates to the scanning tip. At low normal force, the tip minimally disturbs the stripes as it scans the surface and the stripe structure rapidly hea Error bars and lateral calibration CampioneM.CapitaniG. C.Subduction-zone earthquake complexity related to frictional anisotropy in antigoriteNat. Geosci.68478512013ParkJ. Y.High frictional anisotropy of periodic and aperiodic directions on a quasicrystal surfaceScience309135413562005Bluhm We gratefully acknowledge Byong-man Kim and Ryan Yoo of Park Systems for verifying the presence of stripes in our samples using their Park NX-10 AFM. We thank Daniel Wastl for carefully reading our manuscript and for encouraging us to re-examine whether t ACKNOWLEDGEMENTS Author contributions Additional information