# Fileset

[s41467-018-06524-3.pdf](https://mdr.nims.go.jp/filesets/9bd3c704-a3ff-453f-af73-89e969a1995a/download)

## Creator

Jangyup Son, Junyoung Kwon, SunPhil Kim, Yinchuan Lv, Jaehyung Yu, Jong-Young Lee, Huije Ryu, [Kenji Watanabe](https://orcid.org/0000-0003-3701-8119), [Takashi Taniguchi](https://orcid.org/0000-0002-1467-3105), Rita Garrido-Menacho, Nadya Mason, Elif Ertekin, Pinshane Y. Huang, Gwan-Hyoung Lee, Arend van der Zande

## Rights

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

## Other metadata

[Atomically precise graphene etch stops for three dimensional integrated systems from two dimensional material heterostructures](https://mdr.nims.go.jp/datasets/8c6a3823-7bf0-45bb-aacc-1e5294857c3a)

## Fulltext

Atomically precise graphene etch stops for three dimensional integrated systems from two dimensional material heterostructuresARTICLEAtomically precise graphene etch stops for threedimensional integrated systems from twodimensional material heterostructuresJangyup Son1, Junyoung Kwon2, SunPhil Kim1, Yinchuan Lv3, Jaehyung Yu1, Jong-Young Lee2, Huije Ryu2,Kenji Watanabe 4, Takashi Taniguchi4, Rita Garrido-Menacho3,5, Nadya Mason3,5, Elif Ertekin1,5,Pinshane Y. Huang5,6, Gwan-Hyoung Lee 2 & Arend M. van der Zande1,5Atomically precise fabrication methods are critical for the development of next-generationtechnologies. For example, in nanoelectronics based on van der Waals heterostructures,where two-dimensional materials are stacked to form devices with nanometer thicknesses, amajor challenge is patterning with atomic precision and individually addressing each mole-cular layer. Here we demonstrate an atomically thin graphene etch stop for patterning vander Waals heterostructures through the selective etch of two-dimensional materials withxenon difluoride gas. Graphene etch stops enable one-step patterning of sophisticateddevices from heterostructures by accessing buried layers and forming one-dimensionalcontacts. Graphene transistors with fluorinated graphene contacts show a room temperaturemobility of 40,000 cm2 V−1 s−1 at carrier density of 4 × 1012 cm−2 and contact resistivity of80Ω·μm. We demonstrate the versatility of graphene etch stops with three-dimensionallyintegrated nanoelectronics with multiple active layers and nanoelectromechanical deviceswith performance comparable to the state-of-the-art.DOI: 10.1038/s41467-018-06524-3 OPEN1 Department of Mechanical Science and Engineering, University of Illinois at Urbana-Champaign, 1206 W Green Street, Urbana, IL 61801, USA. 2Departmentof Materials Science and Engineering, Yonsei University, 50 Yonsei-ro, Seodaemun-gu, Seoul 03722, Korea. 3 Department of Physics, University of Illinois atUrbana-Champaign, 1110 W Green Street, Urbana, IL 61801, USA. 4National Institute for Materials Science, 1-1 Namiki, Tsukuba, Ibaraki 305-0044, Japan.5 Frederick Seitz Materials Research Laboratory, University of Illinois at Urbana-Champaign, 104 S Goodwin Avenue MC-230, Urbana, IL 61801, USA.6Department of Materials Science and Engineering, University of Illinois at Urbana-Champaign, 1304 W Green Street, Urbana, IL 61801, USA. These authorscontributed equally: Jangyup Son, Junyoung Kwon. Correspondence and requests for materials should be addressed to G.-H.L. (email: gwanlee@yonsei.ac.kr)or to A.M.V.D.Z. (email: arendv@illinois.edu)NATURE COMMUNICATIONS |  (2018) 9:3988 | DOI: 10.1038/s41467-018-06524-3 |www.nature.com/naturecommunications 11234567890():,;http://orcid.org/0000-0003-3701-8119http://orcid.org/0000-0003-3701-8119http://orcid.org/0000-0003-3701-8119http://orcid.org/0000-0003-3701-8119http://orcid.org/0000-0003-3701-8119http://orcid.org/0000-0002-3028-867Xhttp://orcid.org/0000-0002-3028-867Xhttp://orcid.org/0000-0002-3028-867Xhttp://orcid.org/0000-0002-3028-867Xhttp://orcid.org/0000-0002-3028-867Xmailto:gwanlee@yonsei.ac.krmailto:arendv@illinois.eduwww.nature.com/naturecommunicationswww.nature.com/naturecommunicationsAs next-generation technologies of electronic, photonic,and mechanical devices approach the atomic scale, it isimportant to develop atomically precise fabricationmethods. Among them, etch stops, critical for the vertical inte-gration of nanoelectronic and nanomechanical devices, are cre-ated by layering materials with drastically different etchproperties and embedded into a structure, allowing for patterningfeature sizes, accessing buried layers, or undercutting to createsuspended structures. Accordingly, improved fabrication techni-ques are especially needed in nanoelectronics based on van derWaals (vdW) heterostructures, where two-dimensional (2D)materials are stacked to form electronic devices with nanometerthicknesses1–5. Many of the applications for 2D material het-erostructure devices demand out-of-plane integration, contactingmultiple active layers, and creating interconnects between thedifferent layers1–5. Examples include 2D material-based inte-grated circuitry like NAND gates6 or ring oscillator7; devicesbased on interlayer tunneling like light-emitting diodes (LEDs)8or tunnel transistors9; and nanoelectromechanical systems(NEMS) like resonators based on atomic membranes10.In the current state of the art, the relative ease of assembly ofvdW heterostructures, which occurs through the sequential pick-up and stamped release of individual atomic layers, contrastsstarkly with the difficulty of patterning and electrically addressingeach layer in a heterostructure device. As a result, the majority ofstudies in this field are performed on monolayers or on hetero-structures where each layer has been carefully offset so they areaccessible for electrical contacts through direct deposition ofmetal on top or on the exposed edges of individual 2D layers11–14.The current state-of-the-art method is to use edge contacts whereheterostructures are etched through to expose the edges of buriedlayers of graphene encapsulated in insulating hexagonal boronnitride (hBN), and then metals are evaporated onto the edge tomake one-dimensional (1D) contacts13. This method has led to adramatic improvement in the mobility and quality of electronicdevices because it allows contact to electronic layers that are fullyencapsulated and thus have a minimum of disorder13,15. How-ever, edge contacts still require careful offsetting of active layerbecause the etching is not selective so all vertically aligned layersin the heterostructure are exposed simultaneously. A method thatcombines the superior device behavior of the edge contacts butthat simultaneously allows ready patterning of 2D hetero-structures from large area continuous sheets and individuallyaddressing of each layer are critical for translating many of therecent demonstrations of this class of devices into scalabletechnologies.Here we show methods to fabricate nanostructures and accessburied interfaces with the precision of a single atomic layer byusing graphene as impermeable etch masks and etch stops16,17.These techniques, which we call GES (graphene etch stops),represent a straightforward method to selectively expose andcontact embedded graphene layers within 2D heterostructures.This concept takes advantage of the high chemical selectivity ofXeF2, a vapor phase, strong fluorinating agent commonly used asan isotropic etchant for silicon in the microelectromechanicalsystems (MEMS) industry18. Several 2D materials including hBNand transition metal dichalcogenides, are quickly etched whenexposed to XeF216,19,20. In contrast, graphene reacts with XeF2 toform fluorographene (FG)21–24, a wide band-gap semiconductingmonolayer21,22 with composition C4F, in the case that only oneside is exposed22. There have been several demonstrations thattake advantage of this selectivity to use graphene as an etch maskfor shaping MoS216, as a mask to etch underlying silicon25–27, andto create a sacrificial release layer to suspend graphene mem-branes on silicon on insulator17,22. Our innovation has been toapply this etch selectivity to access buried graphene layersembedded within the heterostructures and as masks for pat-terning the underlying layers. Surprisingly, the embedded con-tacts, which is composed of FG–metal contacts, lead to roomtemperature carrier mobilities of 40,000 cm2 V−1 s−1 at carrierdensity n= 4.0 × 1012 cm−2 and behave as 1D contacts with lowcontact resistivity of 80Ω μm, approaching theoretical limits11,28.This capability enables simple and scalable methods to verticallyintegrate 2D devices through contacting multiple active layers,interlayer vias, and suspended nanostructures, yet maintains thestate-of-the-art performance of fully encapsulated 2D devices.ResultsSelective etching by graphene etch stop. Figure 1a illustrates theuse of GES to pattern a heterostructure of 2D materials, andFig. 1b, c are optical images of the same heterostructure beforeand after exposure to XeF2. The heterostructure is fabricated bystacking individual materials using established polymer-free,aligned transfer techniques11,29. Specifically, the heterostructure iscomposed of two monolayer graphene flakes, set in a cross-alignment and embedded between hBN layers, then placed on topof a Si/SiO2 (285 nm) substrate. The color corresponds with thethickness of the heterostructure. In Fig. 1b, two graphene layersare not visible because their contrast is completely overwhelmedby the much thicker hBN, but the inset Raman map confirmstheir positions. As seen in Fig. 1c, after exposure to XeF2 (3 Torrfor 30 s at room temperature), the exposed hBN is completelyetched, while the graphene layers and the hBN underneath themremain. In Supplementary Note 1, Supplementary Figure 1, andSupplementary Table 1, we show that the same process can beapplied to many other 2D material heterostructures as well,including hBN, MoS2, WSe2, and black phosphorus (BP).To examine the selectivity and resolution limits of GES, weobtained cross-sectional images of the etched heterostructureswith a scanning transmission electron microscope (STEM), asshown in Fig. 1d, e. Importantly, the hBN layers under the FGshow no etching, indicating that they are protected from theXeF2. Moreover, Fig. 1e shows that the buried layer of graphene(G1) is unaffected by the etch process and there are atomicallysharp and clean interfaces between stacked graphene and hBNlayers (see also Raman data in Supplementary Figure 2). Theseresults demonstrate that FG maintains the impermeable nature ofgraphene30,31 through the chemical modification process16,32. InFig. 1d, at the edge of the etch mask, the underlying hBN has asub-nanometer slope. These images demonstrate that GES is aself-arresting etch process that enables atomic precision out ofplane and nanometer-scale feature sizes without requiring precisecontrol in timing or conditions.Characteristics of fluorinated graphene etch stops. The self-arresting nature of GES means that it is scalable as well as beingatomically precise. Fig. 2a demonstrates this scalability byapplying GES to a large area heterostructure array. We patternedlarge area graphene as etch masks for patterning large area WS2,both grown by chemical vapor deposition (CVD). First, large areacontinuous graphene was patterned into lines with lithographyand oxygen plasma. Two sets of the patterned lines were thensequentially transferred onto large area continuous monolayerWS2 grown by CVD. The second transfer is set perpendicular tothe first to form a cross-hatch pattern. After exposure to XeF2,WS2 uncovered by graphene was completely etched, while theregion covered by graphene remained under FG (process flow inSupplementary Figure 3 and photoluminescence maps in Sup-plementary Figure 4). This process of combining prepatternedgraphene masks with the selective etch can be repeated to scalablyachieve arbitrarily complex heterostructures of layered FG,ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-06524-32 NATURE COMMUNICATIONS |  (2018) 9:3988 | DOI: 10.1038/s41467-018-06524-3 | www.nature.com/naturecommunicationswww.nature.com/naturecommunicationsgraphene, and other materials on the wafer-scale patterns, whichcould not be realized with conventional patterning and etchingtechniques.Before examining the application of GES to 2D heterostructuredevices, we confirm the structure and electrical properties of FG.Figure 2b shows the Raman spectra of graphene on hBN underincreasing exposure to XeF2. Initially, only the G and 2D peak arevisible, indicating clean graphene with no defects, as well as oneadditional peak from the underlying hBN. After exposure to XeF2,the D peak appeared and the 2D peak was suppressed. Bothphenomena are a result of the breaking of hexagonal symmetrywithin the graphene lattice due to the formation of sp3 bonds bybonding of fluorine atoms onto the graphene surface22.Supplementary Figures 5 and 6 show additional structuralanalyses of X-ray photoelectron spectroscopy (XPS) and TEMand demonstrate that the fluorination condition results in only asp3-type lattice transition without formation of voids, consistentwith the observation that FG acts as an impermeable barrier16.Figure 2c shows the electrical transport through a prefabricatedgraphene transistor on hBN (device shown in the inset andadditional details in Supplementary Figure 7). Before fluorination,the graphene shows mS conductance and gate dependence typicalof the linear dispersion in graphene band structure. The deviceconductance drastically decreased as a function of fluorinationtime. After 10 s, the device conductance decreased by a factor of10, while after 30 s, the graphene became insulating withresistance exceeding 60 GΩ. As shown in Supplementary Figure 8,when graphene on hBN is functionalized for longer periods (720s), it maintained its structure and high resistance for over2 months in ambient conditions.Electrical properties of fluorinated graphene contacts. In therest of the paper, we will explore the application of GES to fab-ricating electronic and mechanical devices from 2D hetero-structures. Two persistent challenges in nanoscale device researchare how to minimize the impact of environment on limiting thepotentially outstanding electronic mobility of nanomaterials andhow to engineer low resistance contacts to nanomaterials. Pre-vious studies have shown that achieving the theoretical limits ofperformance in graphene devices requires graphene to nevercome into contact with solvents or polymers and chargedimpurity scattering to be suppressed by fully encapsulating thesamples in hBN13. However, doing this brings a challenge of howto electrically contact the encapsulated graphene layers. In Fig. 3a,we demonstrate the application of GES to electrically contact aburied graphene layer encapsulated in hBN. Using e-beamlithography, electrodes were patterned on top of hBN/G/hBNheterostructure. Then the structure was exposed to XeF2 beforehBN1hBN2hBN3G1G2XeF2 gasFG1FG2hBN1hBN2G1ab cG1G2G1G2SiO2SiO2d eFG1FG2FG2hBN2hBN3hBN1hBN1hBN2hBN1hBN2FG2FG1hBN1hBN2G1SiO2G1Fig. 1 Selective etching of a vdW heterostructure with XeF2 gas. a Schematic of the XeF2 etching process for a vdW heterostructure of stacked hBN andgraphene layers. b, c Optical micrographs of a corresponding heterostructure fabricated from stacked exfoliated flakes before and after exposure to XeF2.The scale bar is 10 μm. Before etching, the heterostructure is composed (from bottom to top) of silicon oxide substrate, 5 nm hBN, 1 L graphene, 8 nm hBN,1 L graphene, and 10 nm hBN. The inset in b shows a Raman map of the 2D graphene peak, indicating the positions of the two graphene layers (G1 and G2).c Optical micrograph of the sample after XeF2 etching, with an inset illustration indicating the cross-sectional structure in the region indicated. The changesin color between b and c represent changes in film thickness as determined by thin-film interferometry. The substrate is brown, while the thinnest hBN isblue and increasing thickness, and the changes in color represent changes in the hBN thickness from dark blue (thinnest) to light blue to green to yellow(thickest). d False-color cross-sectional bright-field STEM image of the etched heterostructure. The scale bar is 10 nm. The hBN layers (hBN1 and hBN2)covered with graphene masks (FG1 and FG2) were protected from XeF2 etching. e Annular dark-field STEM image taken from the white-dashed area of (c)shows atomically sharp and clear heterointerfaces. The scale bar is 5 nmNATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-06524-3 ARTICLENATURE COMMUNICATIONS |  (2018) 9:3988 | DOI: 10.1038/s41467-018-06524-3 |www.nature.com/naturecommunications 3www.nature.com/naturecommunicationswww.nature.com/naturecommunicationsmetallization. As it is well known, the XeF2 does not attack thepolymer. However, within the patterned regions, the top hBN isetched away, locally exposing and fluorinating the buried gra-phene layer. Electrodes were then deposited through the samepolymer mask directly on to the fluorographene regions (1 nmCr, 30 nm Pd, and 40 nm Au; see Supplementary Figure 9 andMethods for details). Figure 3a shows a cross-sectional high-resolution transmission electron microscope (HR-TEM) image ofthe FG electrical contact. The lithographic pattern is transferredinto the hBN, and the evaporated metal is deposited only on theexposed FG, while the graphene channel under the hBN is neverexposed.Figure 3b shows the field-effect characteristics of a grapheneHall bar device encapsulated by hBN with FG contacts. Themobility was calculated by the Drude model, μ= ne/σ where μ, n,e, and σ are the carrier mobility, carrier density, electron charge,and sheet conductivity, respectively. At high carrier concentrationof n= 4.0 × 1012 cm−2, the sheet resistance was 45Ω per squareat room temperature, corresponding to a carrier mobility of40,000 cm2 V−1 s−1, close to the theoretical limit28. As shown inthe inset of Fig. 3b, the mobility drastically increases withdecreasing carrier concentrations, as expected from the acoustic-phonon-limited model13,28. On a Hall bar device measured at lowtemperature T= 1.7 K (Supplementary Figure 10), the low carrierconcentration mobility increased to 460,000 cm2 V−1 s−1. Thismobility corresponds with a mean free path of 4.6 μm, similar tothe channel width of the device so the mobility is likely limited bydevice dimensions rather than material properties. The deviceconductance (Ids−Vds) is linear and displayed no hysteresis(Supplementary Figure 11). The contact resistance of the FG tothe buried graphene channel was quantified by performingtransfer length measurements on the device shown in the inset ofFig. 3c. Figure 3c shows the resistance vs. channel length atdifferent charge concentrations. The contact resistances areextracted from the extrapolated zero-length intercepts to get 21Ω for holes and 49Ω for electrons for a 4 μm wide channel.Figure 3d shows the contact resistance vs. carrier concentrationfrom 2.4 K to room temperature. The contact resistance vs.temperature is shown in the Fig. 3d inset. The resistivity is notsignificantly affected by temperature and can reach a value of 80Ω∙μm at n= 4.0 × 1012 cm−2, which means absence of thepotential barrier at the contact. This is distinct from the graphenebIntensity (a. u.)1200 1600 2000 2400 2800Raman shift (cm–1)hBNG2DD0 sDGG2D2D10 s30 s D+D′D+D′10–210–310–410–1010–910–11–40 –20 0 20 40Vg (V)Conductance (S)hBN G10 s30 s0 schBNhBNGFG SiO2hBNGMetal XeF2hBNhBNGWS2FGFGWS2GaG/WS2WS2G/G/WS2FG/WS2SiO2FG/G/WS2GFig. 2 Raman and conductance measurements of fluorinated graphene. a Optical images (bottom) of CVD WS2 sheet covered with prepatterned CVD GESbefore and after XeF2 etching (the scale bar is 50 μm). Schematic illustrations (top) show cross-section of graphene-covered WS2 along white dashed linesfor each step. After etching, uncovered WS2 is etched away, meanwhile the covered WS2 remains unchanged under protection of fluorinated GES. b Ramanspectra of graphene on hBN under increasing exposure to XeF2. As XeF2 exposure time increases, the graphene D peak becomes prominent while the 2Dpeak is damped, indicating the formation of sp3-type defects. These results are consistent with other studies of graphene fluorination. c Electricalconductance of graphene on hBN fluorinated by XeF2 treatment (the scale bar in the inset is 10 μm). After a 30 s XeF2 exposure, FG becomes fullyinsulatingARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-06524-34 NATURE COMMUNICATIONS |  (2018) 9:3988 | DOI: 10.1038/s41467-018-06524-3 | www.nature.com/naturecommunicationswww.nature.com/naturecommunicationsdevices with surface-contacted metal electrodes, which showtemperature dependence of contact resistance due to the potentialbarrier formed at the contact13. In addition, all the devices displayrobust chemical and electrical stability. The devices showed nosignificant change after 1 month, when stored in air (Supple-mentary Figure 12). Taken together, the outstanding mobility,low contact resistance, and stability from encapsulation makethese devices comparable to the state of the art across all metrics(Supplementary Figure 13 and Supplementary Table 2 contain acomparison), while simultaneously being much easier to fabricatedue to the self-arresting mechanism of GES. However, these lowcontact resistances are surprising, especially given the high in-plane resistance of FG measured in Fig. 2b.To explain the low contact resistance, density functional theorywas applied to simulate the interlayer distance and chargedistribution at a FG and metal heterointerface. Figure 3e showsthe equilibrated structure and local density of states in fourdifferent interfaces: either graphene or FG and either Cr or Pdmetal (see Supplementary Figure 14 and Methods for simulationdetails). The 1 nm-thick Cr adhesion layer forms islands, not acontinuous film (Supplementary Figure 15), so both Cr and Pdwill make direct contact to the FG surface. From a Landauerframework, the factors governing contact resistance are related tothe carrier transmission probability T and the number M ofconduction modes available12. Transport must occur both fromthe metal to the graphene under the metal and from the grapheneunder the metal to the channel region, which have differenttransmission probabilities. Additionally, the number of grapheneconduction modes under the metal is reduced in some cases dueto charge transfer doping by the metal. The contact resistance canhBNGFGMetal contactshBN1hBN2 MetalG FGadbcRC (Ω·μm)102–3 –2 –1 0 1 2 3n (×1012 cm–2)4–41030 100 200 300Temperature (K)0100200RC (Ω·μm)n = −4×1012 cm–2n = 4×1012 cm–22.4 K10 K50 K100 K200 K300 K060080010001200–4 –3 –2 0 2 3 4–1 1n (×1012 cm–2)3.02 ÅCPdCPdF1.6 Å2.5 Å3.62 ÅCCr2.9 Å1.6 ÅCCrFe0 1 2 3 4Channel length (μm)100200300R (Ω)n = −3.0 × 1012 cm–2n = +3.0 × 1012 cm–2400300 K05 6hBN-encapsulated G� (Ω per sq)–5 5400200300 KhBN-encapsulated G051015μ (×104  cm2 V–1 s–1) 300 Kn (×1012 cm–2)0 1 52 3 4G-Cr FG-Cr G-Pd FG-PdFig. 3 Electrical performance of hBN-encapsulated graphene device with FG via contacts. a Schematic of a hBN-encapsulated graphene device with FG viacontacts and false-color cross-sectional HRTEM image of the FG via contact region (the scale bar is 5 nm). b Four-probe resistivity as a function of carrierdensity at room temperature. The right inset shows a Hall bar graphene device used for carrier mobility measurement (the scale bar is 5 μm). Right insetshows the electron (red) and hole (blue) mobilities extrapolated by applying the Drude model to the measured conductivity (σ= neμ, where σ, n, e, and μare the sheet conductivity, carrier density, electron charge, and carrier mobility, respectively). Black dashed line shows the predicted intrinsic phononlimited mobility of graphene at room temperature28. c Plot of total resistances of the graphene TLM device as a function of channel length, at fixed electronand hole carrier densities. The inset shows optical micrograph of the TLM device (the scale bar is 10 μm). d Contact resistances of the device as a functionof carrier density and temperature. The inset shows contact resistances as a function of temperature and indicates no significant change in contactresistance. e Isosurfaces of the total charge densities at the interfaces between G–Cr, FG–Cr, G–Pd, and FG–Pd, calculated with DFT. The shortened atomicdistances of C–F–Cr and C–F–Pd at the interfaces of FG–metal contacts lead to small contact resistance. This results because orbital overlap through abridge of F facilitates charge transfer from metals to FGNATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-06524-3 ARTICLENATURE COMMUNICATIONS |  (2018) 9:3988 | DOI: 10.1038/s41467-018-06524-3 |www.nature.com/naturecommunications 5www.nature.com/naturecommunicationswww.nature.com/naturecommunicationsbe improved by achieving a smaller effective metal/graphenecoupling length to increase T or by finding metals that lead tohigh metal-induced doping concentrations to increase M. Fromthe isosurfaces, without fluorine the G–Cr interface shows weakvdW bonding33, and the small orbital overlap leads to low T dueto the presence of a tunneling barrier and a large effectivecoupling distance34. The G–Pd interface shows somewhat moreorbital hybridization between the graphene pz and metal d states,consistent with prior results33. Interestingly, when fluorinated thedegree of hybridization increases for both metals, particularly forFG–Pd, indicating a reduced coupling length that renders thecharge transfer more ballistic and lowering the contact resis-tance34. Additionally, the fluorine hybridization opens a band gapin graphene, reducing its work function, and inducing a largeeffective n-type charge transfer doping. The combination of thestrong coupling and the enhanced charge transfer doping suggestthat the contact resistance is limited by the FG–grapheneinterface rather than the metal–FG interface, resulting in 1Dedge contacts rather than 2D surface contacts. To test thishypothesis, a variable width channel device was fabricated. Thecontact resistance was linearly proportional to the reciprocal ofchannel width (Supplementary Figure 16) as expected in 1Dcontacts. These simulations and measurements show that thedominant contact resistance in the FG-buried devices is at the 1Dgraphene–FG interface. Similar results have been seen inelectrically contacting the sides of etched heterostructures whereonly the 1D edge of graphene is exposed13. Unlike in 2D surfacecontacts, this result indicates that the contact resistance of themetallized FG is independent of the contact length and there is nointrinsic lower limit of size. Hence, it should be possible to scaledown the size of embedded contacts and vias to nanoscaledimensions without impacting functionality.Fabrication of three-dimensional (3D) integrated systems from2D materials. In addition to offering a simpler fabrication pro-cess and state-of-the-art device properties, the selective etch stopalso enables capabilities that cannot be easily or scalably realizedusing other techniques. For example, interlayer vias and inde-pendently contacting multiple active layers in vertically alignedheterostructures are critical to integrated circuits like NANDgates6, where logic operations are computed by coupling the gatesand channels of several transistors in series, and graphene-basedmultilayered printed circuit boards (PCBs). Similarly, manydevice applications of 2D heterostructures that rely out-of-planetransport, like vertical PN junctions35, tunnel junctions9, or lightemitters based on 2D materials8,36 require the same ability tocontact vertically aligned layers separately.Figure 4 outlines proof-of-concept demonstrations of usingGES to fabricate interlayer vias and vertically integrate multipleactive layers. First, the GES is a self-arresting process that allowsaccess to multiple buried layers set at different depths within asingle etch step. This allows the creation of interlayer vias, whichare critical components for integrated circuits, where wiring anddevices can operate on more than one plane. Figure 4a is anoptical image of a multilayer graphene–hBN heterostructure withinterlayer vias fabricated in a single lithography and etch stepidentical to the one used in Fig. 3 on a single layer. Theheterostructure is formed by sequentially stacking three graphenelayers, each separated by few-layer hBN. Each graphene layeroperates as a separate transistor, and the out-of-plane vias areformed by exposing two layers within a single opening beforemetallization. Figure 4b is the corresponding transfer curveswithin each graphene device. In this particular geometry, thelayers are offset, so they may all be controlled with the globalbackgate. The inset of Fig. 4b is the interlayer transport currentthrough the vias vs. interlayer bias, which shows a lineardependence. For example, the total resistance measured fromthe labeled electrodes, B1–M1, is 1 kΩ at Vg= 0 V, equivalent tothe in-plane channel resistance of the graphene. The contactresistance is negligibly small compared with the correspondingchannel resistances, demonstrating that GES enables efficient,simple, and selective contacts to vertically offset layers to createlow resistance interlayer vias, which could be integrated to thecomplicated devices such as multilayer PCBs or light emitterbased on 2D materials.The second demonstration takes advantage of the combinedhigh in-plane resistance of FG with the low contact resistancewhen the FG is metallized. Through sequential patterning andetching steps, GES allows independent contact of multiple activelayers that interact to generate device functionality. This allowsthe creation of 3D integrated circuitry from 2D materials, where,for example, vertically offset encapsulated 2D layers act as boththe gate and channel in a transistor, which has been difficult torealize with conventional patterning techniques or 1D edgecontacting13 (Supplementary Figure 17). Figure 4c shows theoptical image and schematic illustration of a hBN-encapsulatedgraphene transistor channel with graphene backgate (i.e. twocoupled active layers). In the heterostructure, the two graphenesheets are separated by hBN. Key to this demonstration is that thetop graphene layer is larger than the bottom layer and fully coversit. In order to access the buried bottom layer, the selective etchprocess is repeated twice with an oxygen plasma shaping step inbetween, then lastly depositing electrodes contacting all layers(see Supplementary Figure 18 for the full fabrication process).The high in-plane and vertical resistance of both the top FG andthe dielectric hBN allow access to the buried bottom layer withoutshorting the two layers together. Figure 4d is the transfer curve ofthe resulting embedded all 2D material field effect transistors,showing that the top graphene channel (G2) can be effectivelymodulated by the bottom graphene gate (G1). Like the singlelayer demonstration in Fig. 3, all 2D layers are encapsulated,protecting them from extrinsic disorder and resulting in highcarrier mobilities. The two demonstrations above prove that theselective etch process enables interlayer vias and 3D integration ofmultiple active device layers made entirely of 2D materials, bothcapabilities that are critical to the development of integratedcircuitry from 2D materials.As a final demonstration of a different kind of verticalintegration, a common application of selective etches in MEMS isto suspend mechanically responsive structures. In studies on this,it was shown that graphene can be used as an etch mask forunderlying silicon to generate suspended FG membranes17,22.These atomic membranes behave as tensioned mechanicalresonators37, useful as low mass chemical sensors or tunableradio frequency filters or oscillators. In Fig. 4e, we show that thesame concept can be applied to fabricate suspended grapheneNEMS from 2D heterostructures. Figure 4e is an angled scanningelectron microscopic (SEM) image of a suspended few-layergraphene membrane clamped by graphite supports. Thismembrane was fabricated by first creating a heterostructure ofnarrow few-layer graphene ribbon on 70-nm-thick blackphosphorus, with 100 nm thick graphite at either end. The blackBP etches far more quickly than other 2D materials when exposedto XeF2 (Supplementary Figure 1 and Supplementary Table 1),allowing it to act as a sacrificial release layer that undercuts thegraphene (see fabrication process in Supplementary Figure 19).The resulting FG membrane was fully suspended withoutwrinkles and contamination in a dry vapor phase process.Figure 4f is the mechanical resonance of the membrane measuredusing modulated laser optomechanical actuation and dynamicreflection contrast detection10,37 (Supplementary Figures 20 andARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-06524-36 NATURE COMMUNICATIONS |  (2018) 9:3988 | DOI: 10.1038/s41467-018-06524-3 | www.nature.com/naturecommunicationswww.nature.com/naturecommunications21 and Methods). The graphene membrane has a resonantfrequency of f1= 5.24 MHz and quality factor of 255 at roomtemperature, comparable to state-of-the-art graphene resonatorswith similar dimensions produced via a wet process ormechanical exfoliation over trenches10,38. Just as XeF2 has foundwide applicability in MEMS or NEMS industry as a selectivesilicon etch, using GES to produce graphene-based resonators hasa great potential since the whole process is liquid free and clean.Moreover, this route can produce suspended graphene with muchhigher aspect ratios and gap depths than with conventionaltransfer or wet etching techniques.DiscussionTaken together, the demonstrations in Fig. 4 show that GESproposed in this work enable advanced fabrication of 3D-eAmplitude (a.u.)4.8 5.0 5.2 5.4 5.6Frequency (MHz)Q = 255fa200R (Ω)bG1 (bottom)G2 (middle)G3 (top)hBNB2B1via1via2M1M2T1T24006008001000G1 (B1–B2)G2 (M1–M2)G3 (T1–T2)–40 –20 0 20 40Vds (mV)–80–4004080I ds (μA)B1–M1T1–M2B1–T1Vg = 0 V–4 –2 0 2 4n (×1012 cm–2)–2 –1 0 1 2Vg (V)3456Vds = 10 mVI ds (μA) c dB2 B1 via1 M1 M2 via2 T1 T2G1G2G3hBNhBNS DGateSDGatehBNG1 gate electrodehBNG2 graphenechannelMultilayer GSuspendedfew-layer GBP undermultilayer GBP undermultilayer GFG hBNFig. 4 Fabrication of 3D integrated devices from 2D materials and suspended graphene mechanical resonators with a dry and one-step etching process. aOptical microscopic image and schematic illustration of the multi-stacked graphene devices connected with via contacts (the scale bar is 10 μm). Sourceand drain contacts were denoted as B1 and B2 for bottom graphene device (G1), M1 and M2 for middle graphene device (G2), and T1 and T2 for topgraphene device (G3). All the graphene devices are connected with two via contacts (via1 and via2). b Plots of resistance vs. carrier density of multi-stacked three graphene devices in a. The inset shows Ids− Vds curves obtained from two graphene devices connected with via1 or via2. Linear curves andsmall resistance indicate that these three graphene devices embedded in hBN are electrically connected with low resistance via contacts. c Opticalmicroscopic image and schematic illustration of the hBN-encapsulated graphene device with graphene backgate (the scale bar is 5 μm). d Ids− Vg curve ofthe graphene device (G2) in c. Gate voltage was applied with bottom graphene (G1). e False-color scanning electron micrograph of the suspendedgraphene membrane (the scale bar is 2 μm). The inset is a magnified image of the suspended membrane (the scale bar is 1 μm). f Plot of normalizedamplitude vs. frequency of the few-layer graphene resonator. Black circles and red solid line are the optomechanical response and Lorentzian fit,respectivelyNATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-06524-3 ARTICLENATURE COMMUNICATIONS |  (2018) 9:3988 | DOI: 10.1038/s41467-018-06524-3 |www.nature.com/naturecommunications 7www.nature.com/naturecommunicationswww.nature.com/naturecommunicationsintegrated electronic and mechanical devices based on 2Dmaterials. Figure 3 shows that the structures will maintain thehigh mobility and low contact resistances that are currently thestate of the art. Figure 1 shows that nanometer scale in-planefeatures and devices should be possible. Most of the demonstra-tions above use exfoliated materials, but as shown in Fig. 2a, thescalability of this technique means that all demonstrations willalso work on arrays of devices patterned from continuous, largearea heterostructures. The self-arresting nature of this processmeans that precision is not needed to achieve uniformity ofdevices, a huge benefit in atomically precise electronics. All ofthese components suggest that the selective etch process is amajor capability necessary for the realization of atomically pre-cise, all-2D nanoelectronics as a viable technology, in applicationswhere vertical integration is critical, like integrated circuit logiccomponents (e.g., NAND gates), devices operating through out-of-plane transport (e.g., 2D material tunnel junctions and LEDs),and in 2D nanoelectromechanical systems. Finally, many othermaterials such as transition metals, silicon, and MBE grown III–Vmaterials are also etched by XeF2, so GES may find broadapplication for the fabrication of atomically precise devicesbeyond just 2D materials.MethodsFabrication of vdW heterostructures. To fabricate the heterostructures, we use a2D material pick-up technique with similar established methods13,29. Beforecreating the heterostructure, it is necessary to fabricate a sacrificial transfer sub-strate. First, a 0.5-mm-thick polydimethylsiloxane (PDMS) droplet is deposited ona microscope glass slide, then cured overnight at 60 °C. At the same time, poly(bisphenol A carbonate, Sigma Aldrich) (PC) dissolved in chloroform is depositedonto a microscope slide glass. The chloroform is allowed to evaporate in air atroom temperature, then the remaining PC film is manually peeled off by hand. Thepeeled-off PC film is placed onto the PDMS, then the entire structure is baked at170 °C for 15 min to form conformal contact between PC film and PDMS. Theresulting transfer substrate is then fixed to a micromanipulator. In parallel, all 2Dflakes used for the vdW heterostructures were separately exfoliated onto the SiO2(285 nm)/Si substrates with the scotch tape method. The thickness or layer numberof each material is separately confirmed using a combination of Raman spectro-scopy, atomic force microscopy, and optical microscopy. For the first pick-up, it isnecessary to start with an extra thick layer of hBN (~20 nm). The PC/PDMS stampis placed onto the target hBN flake at 70 °C. To increase adhesion strength betweenPC and hBN, the temperature is then raised to 130 °C. Then PC/PDMS stamp isgradually lifted up during cooling to 70 °C. This process is then repeated to pick upother 2D flakes subsequently at 90 °C. Each 2D piece must be smaller than the toplayer of hBN. After stacking, the stacked heterostructure was transferred onto aclean SiO2/Si substrate by releasing the PC film from the PDMS at a highertemperature above 190 °C. Lastly, the PC film was removed by rinsing the samplein chloroform.Xenon difluoride etching. The XeF2 etcher (Xactix etching system) was used forthe selective etching of 2D materials in pulse mode with PXeF2= 3 Torr at roomtemperature. The pulse time for etching, i.e., exposure time was set according to thethickness of top layer of hBN, between 30 s, to 2 min. However, it should be notedthat the exposure time is not proportional to etch rate because etching stops at thegraphene layer.Device fabrication. The e-beam lithography (EBL, TESCAN) was performed togenerate patterns to selectively etch the vdW heterostructures. The exposure to theXeF2 gas did not affect the ability to remove the poly (methyl methacrylate)(PMMA) used as an e-beam resist using normal solvents. For fabrication of thedevices in Fig. 3, the vdW heterostructure was etched by first patterning the PMMAon top of the heterostructure, then exposing the entire structure to XeF2. The toplayers of the heterostructure were etched away exposing the contact area of theembedded graphene, which fluorinated during etching (Supplementary Figure 9).Then metals of Cr/Pd/Au (1 nm/30 nm/40 nm) were deposited using e-beamevaporator (Temescal six pocket e-beam evaporation systems). Finally, lift-offprocess was performed simply by soaking the samples in acetone.Sample preparation for TEM. In the TEM images of Supplementary Figures 6 and15, graphene was grown by CVD following standard recipes described in a previouspaper39. The CVD graphene was then transferred onto a TEM grid. PMMA wasspin-coated on the as-grown CVD graphene on a copper foil, followed by etchingof copper in ammonium persulfate solution. After rinsing in multiple baths of de-ionized (DI) water, the graphene/PMMA film floating on DI water was scoopedwith the TEM grid. PMMA film was removed by dipping it in acetone. The cross-section TEM specimens in Fig. 1d, e in the main text were prepared using FEIHelios 600i Dualbeam focused ion beam (FIB), using standard lift-out procedureswith a final milling step of 2 kV to reduce surface damage. For Fig. 3a, cross-sectionTEM sample of the encapsulated graphene device was prepared with FIB (JIB-4601F, JEOL).High-resolution TEM. STEM images in Fig. 1d, e were acquired with a 200 kVaberration-corrected JEOL 2200FS STEM. HR-TEM images in SupplementaryFigures 6 and 14 were acquired on a Cs-corrected TEM (JEM-ARM200F, JEOL).The acceleration voltage was fixed at 80 kV to minimize damage of graphene byelectron beam irradiation.Scanning electron microscopy. The SEM images in Fig. 4e of the heterostructurestack on an SiO2 substrate were acquired on a Hitachi S-4700 field-emission gunSEM with 2 kV accelerating voltage. The sample is tilted by 45° with respect to thebeam direction. False coloring was added after data acquisition.Raman spectroscopy. Raman measurements in Fig. 2b and Supplementary Fig-ure 8 were acquired on a Renishaw using a 633 nm laser and an 1800 mm−1grating. To minimize damage of graphene by irradiation of the laser, a power of< 5 mW was used with an acquisition of 60 s.XPS analysis. XPS measurement in Supplementary Figure 5 was acquired usinga K-alpha XPS system (Thermo VG, UK). For this measurement, graphenegrown by CVD was used. To prevent peak shift by charging effect of the sub-strate, the CVD graphene was transferred onto Au-coated SiO2 substrate. Weutilized monochromated Al as X-ray sources (Al Kα line: 1486.5 eV) and X-raypower of 12 kV and 3 mA. All measurements were carried out in vacuum (P <5 × 10−9 mbar).Electrical measurements. For the temperature-dependent electrical measure-ments in Fig. 3, the devices were placed on a commercial chip carrier with 32 leadsand electrically contacted using aluminum wires with a wedge-wire bonder. Thenthe devices were loaded into cryostat, with a base temperature of 1 K. Conventionaltwo-point and four-point lock-in measurements were performed using an SR830.For measurement of via and graphene gated devices in Fig. 4, measurements wereperformed in air at room temperature with a semiconductor parameter analyzer(Keithley 4200).Resonator measurements. Two lasers of different wavelengths were focused onthe center of the graphene membrane and used to actuate and detect themechanical resonance. To actuate the membrane, a 623 nm diode laser wasmodulated electrically. The reflected light of a second 520 nm laser was monitoredthrough the Si-based avalanche photodetector. The modulation frequency wastuned and monitored using a spectrum analyzer to find the resonance frequency.The measurements were performed in an optical cryostat at < 5 μTorr to reducedamping of the membrane.Simulation of the FG–metal interface. The atomic-scale structure and chargedistribution of the interface between metal and FG shown in Fig. 3e and Sup-plementary Figure 14 was simulated using density functional theory40,41implemented in VASP42 in conjunction with projected augmented wave43. Thegeneralized gradient approximation of Perdew–Burke–Ernzerhof44 was appliedto describe the exchange-correlation functional. An energy cutoff of 350 eV waschosen for the plane wave basis and achieves convergence of the total energy ofboth Pd and Cr and graphene sheet to within 0.01 eV of the total energy.Geometry optimization allowed relaxation until the forces on each atom were< 0.1 eV/Å. A 64-atom supercell of graphene and 7 layers of Pd (111) (or Cr(111)) were used to describe the metal electrode, which is sufficient to recoverbulk-like properties in the interior of the slab. Within GGA-PBE, the graphenelattice constant was 2.47 Å, and in our simulations, the Pd metal electrodes areunder 3% compressive strain due to lattice mismatch with the graphene mem-brane. In each supercell, 20 Å of vacuum is included to avoid interaction betweenadjacent images in the z-direction. A 4 × 8 × 1 mesh was used to sample thesystem k-space. Furthermore, graphene or fluorinated graphene was introducedon both sides of the metal to maintain symmetry in the supercells and to avoidspurious introduction of electric fields and dipole moments across the supercell.Data availabilityThe authors declare that all data supporting the findings of this study are available withinthe paper and its supplementary information files.Received: 13 February 2018 Accepted: 29 August 2018ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-06524-38 NATURE COMMUNICATIONS |  (2018) 9:3988 | DOI: 10.1038/s41467-018-06524-3 | www.nature.com/naturecommunicationswww.nature.com/naturecommunicationsReferences1. Cui, X. et al. Multi-terminal transport measurements of MoS2 using a van derWaals heterostructure device platform. Nat. Nanotechnol. 10, 534–540 (2015).2. Novoselov, K. S., Mishchenko, A., Carvalho, A. & Castro Neto, A. H. 2Dmaterials and van der Waals heterostructures. Science 353, aac9439 (2016).3. Duong, D. L., Yun, S. J. & Lee, Y. H. van der Waals layered materials:opportunities and challenges. ACS Nano 11, 11803–11830 (2017).4. Jariwala, D., Marks, T. J. & Hersam, M. C. Mixed-dimensional van der Waalsheterostructures. Nat. Mater. 16, 170–181 (2017).5. Frisenda, R. et al. Recent progress in the assembly of nanodevices and van derWaals heterostructures by deterministic placement of 2D materials. Chem.Soc. Rev. 47, 53–68 (2018).6. Wang, H. et al. Integrated circuits based on bilayer MoS2 transistors. NanoLett. 12, 4674–4680 (2012).7. Guerriero, E. et al. Gigahertz integrated graphene ring oscillators. ACS Nano 7,5588–5594 (2012).8. Withers, F. et al. Light-emitting diodes by band-structure engineering in vander Waals heterostructures. Nat. Mater. 14, 301–306 (2015).9. Britnell, L. et al. Resonant tunnelling and negative differential conductance ingraphene transistors. Nat. Commun. 4, 1794 (2013).10. van der Zande, A. M. et al. Large-scale arrays of single-layer grapheneresonators. Nano Lett. 10, 4869–4873 (2010).11. Nagashio, K., Nishimura, T., Kita, K. & Toriumi, A. Contact resistivity andcurrent flow path at metal/graphene contact. Appl. Phys. Lett. 97, 143514 (2010).12. Xia, F., Perebeinos, V., Lin, Y., Wu, Y. & Avouris, P. The origins and limits ofmetal-graphene junction resistance. Nat. Nanotechnol. 6, 179–184 (2011).13. Wang, L. et al. One-dimensional electrical contact to a two-dimensionalmaterial. Science 342, 614–617 (2013).14. Giubileo, F. & Di Bartolomeo, A. The role of contact resistance in graphenefield-effect devices. Prog. Surf. Sci. 92, 143–175 (2017).15. Gao, Q. & Guo, J. Role of chemical termination in edge contact to graphene.APL Mater. 2, 056105 (2014).16. Huang, Y. et al. An innovative way of etching MoS2: characterization andmechanistic investigation. Nano Res. 6, 200–207 (2013).17. Robinson, J. T. et al. Graphene strained by defects. ACS Nano 11, 4745–4752(2017).18. Williams, K. R. et al. Etch rates for micromachining processing.J. Microelectromech. S. 5, 256–269 (1996).19. Roy, T. et al. Field-effect transistors built from all two-dimensional materialcomponents. ACS Nano 8, 6259–6264 (2014).20. Zhang, R., Drysdale, D., Koutsos, V. & Cheung, R. Controlled layer thinning andp-type doping of WSe2 by vapor XeF2. Adv. Funct. Mater. 27, 1702455 (2017).21. Nair, R. R. et al. Fluorographene: a two-dimensional counterpart of teflon.Small 6, 2877–2884 (2010).22. Robinson, J. T. et al. Properties of fluorinated graphene films. Nano Lett. 10,3001–3005 (2010).23. Stine, R., Lee, W. K., Whitener, K. E., Robinson, J. T. & Sheehan, P. E.Chemical stability of graphene fluoride produced by exposure to XeF2. NanoLett. 13, 4311–4316 (2013).24. Kashtiban, R. J. et al. Atomically resolved imaging of highly orderedalternating fluorinated graphene. Nat. Commun. 5, 4902 (2014).25. Rangarajan, A., Wood, J., Koepke, J. & Lyding, J. Graphene as an etch maskfor silicon. http://hdl.handle.net/2142/72832 (2014).26. Fang, C., Shapter, J. G., Voelcker, N. H. & Ellis, A. V. Graphene masks aspassivation layers in the electrochemical etching of silicon. J. Mater. Sci. 49,7819–7823 (2014).27. Kim, J., Lee, D. H., Kim, J. H. & Choi, S. H. Graphene-assisted chemicaletching of silicon using anodic aluminum oxides as patterning templates. ACSAppl. Mater. Interfaces 7, 24242–24246 (2015).28. Hwang, E. H. & Das Sarma, S. Acoustic phonon scattering limited carriermobility in two-dimensional extrinsic graphene. Phys. Rev. B 77, 115449(2008).29. Pizzocchero, F. et al. The hot pick-up technique for batch assembly of van derWaals heterostructures. Nat. Commun. 7, 11894 (2016).30. Bunch, J. S. et al. Impermeable atomic membranes from graphene sheets.Nano Lett. 8, 2458–2462 (2008).31. Berry, V. Impermeability of graphene and its applications. Carbon 62, 1–10(2013).32. Liu, Y. et al. Basal plane fluorination of graphene by XeF2 via a radical cationmechanism. J. Phys. Chem. Lett. 6, 3645–3649 (2015).33. Ji, X., Zhang, J., Wang, Y., Qian, H. & Yu, Z. A theoretical model formetal–graphene contact resistance using a DFT–NEGF method. Phys. Chem.Chem. Phys. 15, 17883–17886 (2013).34. Giovannetti, G. et al. Doping graphene with metal contacts. Phys. Rev. Lett.101, 026803 (2008).35. Lee, C. H. et al. Atomically thin p–n junctions with van der Waalsheterointerfaces. Nat. Nanotechnol. 9, 676–681 (2014).36. Kim, Y. D. et al. Ultrafast graphene light emitters. Nano Lett. 18, 934–940(2018).37. Chen, C. & Hone, J. Graphene nanoelectromechanical systems. Proc. IEEE101, 1766–1779 (2013).38. Chen, C. et al. Performance of monolayer graphene nanomechanicalresonators with electrical readout. Nat. Nanotechnol. 4, 861–867(2009).39. Lee, G.-H. et al. High-strength chemical-vapor-deposited graphene and grainboundaries. Science 340, 1073–1076 (2013).40. Hohenberg, P. & Kohn, W. Inhomogeneous electron gas. Phys. Rev. 136,B864–B871 (1962).41. Kohn, W. & Sham, L. J. Self-consistent equations including exchange andcorrelation effects. Phys. Rev. 140, A1133–A1138 (1951).42. Kresse, G. & Furthmüller, J. Efficiency of ab-initio total energy calculations formetals and semiconductors using a plane-wave basis set. Comput. Mater. Sci.6, 15–50 (1996).43. Blöchl, P. E. Projector augmented-wave method. Phys. Rev. B 50, 17953–17979(1994).44. Perdew, J. P., Burke, K. & Ernzerhof, M. Generalized gradient approximationmade simple. Phys. Rev. Lett. 77, 3865–3868 (1996).AcknowledgementsThis research was primarily supported by the National Science Foundation MRSECprogram under NSF Award Number DMR-1720633. Work at Yonsei was funded bySamsung Research Funding & Incubation Center of Samsung Electronics under ProjectNumber SRFC-MA1502-12. Y.L. and P.Y.H. were supported by the Air Force Office ofScientific Research under award number FA9550-7-1-0213; R.G.M acknowledges fund-ing by DOE Basic Energy Sciences under DESC0012649. This work was carried out inpart in the Fredrick-Seitz Material Research Laboratory Central Facilities and the Microand Nano Technology Laboratory at UIUC. K.W. and T.T. acknowledge support fromthe Elemental Strategy Initiative conducted by the MEXT, Japan and JSPS KAKENHIGrant Numbers JP15K21722.Author contributionsJ.S., J.K., G.H.L., and A.M.v.d.Z. conceived and designed the study. J.S. and J.K. fabricatedsamples and carried out experiments under the guidance of G.H.L. and A.M.v.d.Z. andwith the help of the other authors. S.P.K. performed optomechanical measurement. Y.L.and P.Y.H. executed the STEM and SEM experiments. J.Y. and E.E. carried out simu-lation. J.Y.L. and H.R. assisted in the preparation of heterostructure samples. R.G.M. andN.M. carried out the low-temperature electrical measurements. K.W. and T.T. preparedhigh-quality hBN. All authors contributed to the discussion of this work. J.S., J.K., P.Y.H.,G.H.L., and A.M.v.d.Z. wrote the manuscript.Additional informationSupplementary Information accompanies this paper at https://doi.org/10.1038/s41467-018-06524-3.Competing interests: The authors declare no competing interests.Reprints and permission information is available online at http://npg.nature.com/reprintsandpermissions/Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims inpublished maps and institutional affiliations.Open Access This article is licensed under a Creative CommonsAttribution 4.0 International License, which permits use, sharing,adaptation, distribution and reproduction in any medium or format, as long as you giveappropriate credit to the original author(s) and the source, provide a link to the CreativeCommons license, and indicate if changes were made. The images or other third partymaterial in this article are included in the article’s Creative Commons license, unlessindicated otherwise in a credit line to the material. If material is not included in thearticle’s Creative Commons license and your intended use is not permitted by statutoryregulation or exceeds the permitted use, you will need to obtain permission directly fromthe copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.© The Author(s) 2018NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-06524-3 ARTICLENATURE COMMUNICATIONS |  (2018) 9:3988 | DOI: 10.1038/s41467-018-06524-3 |www.nature.com/naturecommunications 9http://hdl.handle.net/2142/72832https://doi.org/10.1038/s41467-018-06524-3https://doi.org/10.1038/s41467-018-06524-3http://npg.nature.com/reprintsandpermissions/http://npg.nature.com/reprintsandpermissions/http://creativecommons.org/licenses/by/4.0/http://creativecommons.org/licenses/by/4.0/www.nature.com/naturecommunicationswww.nature.com/naturecommunications Atomically precise graphene etch stops for three dimensional integrated systems from two dimensional material heterostructures Results Selective etching by graphene etch stop Characteristics of fluorinated graphene etch stops Electrical properties of fluorinated graphene contacts Fabrication of three-dimensional (3D) integrated systems from 2D materials Discussion Methods Fabrication of vdW heterostructures Xenon difluoride etching Device fabrication Sample preparation for TEM High-resolution TEM Scanning electron microscopy Raman spectroscopy XPS analysis Electrical measurements Resonator measurements Simulation of the FG–nobreakmetal interface References References Acknowledgements Author contributions Competing interests ACKNOWLEDGEMENTS