DOI: 10.1126/science.1199183, 1168 (2011);331 Science, et al.Ayrat Dimiev

Layer-by-Layer Removal of Graphene for Device Patterning

This copy is for your personal, non-commercial use only.

clicking here.colleagues, clients, or customers by , you can order high-quality copies for yourIf you wish to distribute this article to others

  here.following the guidelines

can be obtained byPermission to republish or repurpose articles or portions of articles

  ): March 7, 2011 www.sciencemag.org (this infomation is current as of

The following resources related to this article are available online at

http://www.sciencemag.org/content/331/6021/1168.full.htmlversion of this article at:

including high-resolution figures, can be found in the onlineUpdated information and services,

http://www.sciencemag.org/content/suppl/2011/03/03/331.6021.1168.DC2.html http://www.sciencemag.org/content/suppl/2011/03/01/331.6021.1168.DC1.html

can be found at: Supporting Online Material

http://www.sciencemag.org/content/331/6021/1168.full.html#relatedfound at:

can berelated to this article A list of selected additional articles on the Science Web sites

http://www.sciencemag.org/content/331/6021/1168.full.html#ref-list-1, 2 of which can be accessed free:cites 30 articlesThis article

http://www.sciencemag.org/content/331/6021/1168.full.html#related-urls1 articles hosted by HighWire Press; see:cited by This article has been

http://www.sciencemag.org/cgi/collection/mat_sciMaterials Science

subject collections:This article appears in the following

registered trademark of AAAS. is aScience2011 by the American Association for the Advancement of Science; all rights reserved. The title

CopyrightAmerican Association for the Advancement of Science, 1200 New York Avenue NW, Washington, DC 20005. (print ISSN 0036-8075; online ISSN 1095-9203) is published weekly, except the last week in December, by theScience

on

Mar

ch 7

, 201

1w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

used to plot the solid curve in Fig. 2C, whichshows good agreement between simulation andexperimental data. At zero time delay, thecombined EI(t) and Eb(t) equals E(t), the fieldof the synthesized sawtooth. This field can beFourier reconstructed from the CEP and the am-plitudes (Fig. 2D). A comparison of Fig. 2, C andD, indicates that the sawtooth field is nearly com-pletely retrieved, thus confirming the efficacy ofthe shaper-assisted linear cross-correlation tech-nique. Given the many sources that lead to ampli-tude fluctuations and phase distortions, this resultdemonstrates the robustness of the synthesizedpulses, permitting successful field retrieval.

The instantaneous field of a Fourier-synthesizedwaveform is CEP dependent, whereas the pulseenvelope is not [for an illustration, see figure 3 in(20)]. This result can now be experimentally vali-dated here. Using the same sawtooth as an exam-ple, we recorded the I(t) of waveforms that haveidentical field amplitudes but different CEPs. Theresults are shown in Fig. 3, where the waveformshave the same normalized amplitudes as thosein Fig. 2 but three different values of the CEP (p/2,p, and 3p/2). The effect of the CEP on the shapeof the electric field as predicted by Fourier anal-ysis is clearly displayed.

The procedure we have described can beextended to synthesize and retrieve various re-petitive fields. For example, transform-limited sub-cycle sine and cosine pulse trains are formed withnormalized amplitudes of {1,1,1,1,1} for five har-monics and a CEP of p/2 and 0, respectively. Asquare wave has amplitudes of {1, 0, −1/3, 0, 1/5}and a CEP of 0. Cross-correlation of these syn-thesized fields is shown in Fig. 4.

The average power of the beam of synthe-sized pulses in this experiment was a few milli-watts. This power is limited mainly by the powerof the generated fifth harmonic and the repetitionrate of the pump laser. One desirable improve-ment is to substantially increase the duty cycle ofthe laser. Raising the pump laser pulse energyand repetition rate, as well as using a larger beamsize, could easily increase the average power tothe watt level. Starting with a continuous-wave(CW) or quasi-CW laser source could make thesynthesized waveforms inherently more stable.Several groups are making substantial progressin this direction (21, 22).

Despite the limited number of waveforms thatcan be synthesized with five harmonics, we have

described a simple solution to an optical wave-form synthesizer that is analogous to RF andmicrowave function generators and showed thatseveral common waveforms are accessible. Wefurther showed an all-optical scheme to retrievethe instantaneous field. For broad application, itis desirable to have many more spectral compo-nents available for the synthesis. Other researchgroups have obtained >200 sidebands by drivingthe rotational Raman and vibrational Raman co-herence simultaneously (23).

References and Notes1. E. Goulielmakis et al., Science 317, 769 (2007).2. G. Sansone et al., Science 314, 443 (2006).3. Z. Jiang, C.-B. Huang, D. E. Leaird, A. M. Weiner,

Nat. Photonics 1, 463 (2007).4. T. W. Hansch, Opt. Commun. 80, 71 (1990).5. L. Matos et al., Opt. Lett. 29, 1683 (2004).6. R. K. Shelton et al., Science 293, 1286 (2001).7. S. E. Harris, A. V. Sokolov, Phys. Rev. Lett. 81, 2894

(1998).8. A. V. Sokolov, D. R. Walker, D. D. Yavuz, G. Y. Yin,

S. E. Harris, Phys. Rev. Lett. 85, 562 (2000).9. J. Q. Liang, M. Katsuragawa, F. L. Kien, K. Hakuta,

Phys. Rev. Lett. 85, 2474 (2000).10. S. W. Huang, W.-J. Chen, A. H. Kung, Phys. Rev. A 74,

063825 (2006).11. T. Suzuki, M. Hirai, M. Katsuragawa, Phys. Rev. Lett. 101,

243602 (2008).12. W.-J. Chen et al., Phys. Rev. Lett. 100, 163906 (2008).13. Z.-M. Hsieh et al., Phys. Rev. Lett. 102, 213902 (2009).14. D. Keusters et al., J. Opt. Soc. Am. B 20, 2226 (2003).

15. Methods and experimental details are available assupporting material on Science Online.

16. See, for example, www.falstad.com/fourier/.17. A. M. Weiner, in Ultrafast Optics (Wiley, New York,

2009), p. 92.18. A. M. Weiner, Rev. Sci. Instrum. 71, 1929 (2000).19. A. Galler, T. Feurer, Appl. Phys. B 90, 427 (2008).20. S. N. Goda, M. Y. Shverdin, D. R. Walker, S. E. Harris,

Opt. Lett. 30, 1222 (2005).21. J. T. Green, J. J. Weber, D. D. Yavuz, Phys. Rev. A 82,

011805(R) (2010).22. F. Couny, F. Benabid, P. J. Roberts, P. S. Light,

M. G. Raymer, Science 318, 1118 (2007).23. D. D. Yavuz, D. R. Walker, M. Y. Shverdin, G. Y. Yin,

S. E. Harris, Phys. Rev. Lett. 91, 233602 (2003).24. We are indebted to S. E. Harris for inspiring us to work on

this project and for his comments during manuscriptpreparation. We thank S.-Y. Wu, W.-J. Chen, andS.-L. Wang for helpful assistance and S. Goda,F.-G. Yee, C.-B. Huang, and C.-L. Pan for discussions.This work was supported by the Academia Sinica(AS-98-TP-A10) and the National Science Council(NSC 97-2120-M-001-002 and NSC 98-2112-M-001-008-MY3) of Taiwan. R.-P.P. was supported by theNational Science Council (NSC 96-2221-E-009-131-MY3)of Taiwan.

Supporting Online Materialwww.sciencemag.org/cgi/content/full/science.1198397/DC1Materials and MethodsFigs. S1 to S3Table S1Movie S1

28 September 2010; accepted 11 January 2011Published online 20 January 2011;10.1126/science.1198397

Layer-by-Layer Removal of Graphenefor Device PatterningAyrat Dimiev, Dmitry V. Kosynkin, Alexander Sinitskii, Alexander Slesarev,Zhengzong Sun, James M. Tour*

The patterning of graphene is useful in fabricating electronic devices, but existing methods donot allow control of the number of layers of graphene that are removed. We show that sputter-coatinggraphene and graphene-like materials with zinc and dissolving the latter with dilute acid removes onegraphene layer and leaves the lower layers intact. The method works with the four different types ofgraphene and graphene-like materials: graphene oxide, chemically converted graphene, chemicalvapor–deposited graphene, and micromechanically cleaved (“clear-tape”) graphene. On the basis ofour data, the top graphene layer is damaged by the sputtering process, and the acid treatment removesthe damaged layer of carbon. When used with predesigned zinc patterns, this method can be viewedas lithography that etches the sample with single-atomic-layer resolution.

The electronic properties of graphene (1), atwo-dimensional (2D) network of sp2 car-bon atoms, can vary as a function of the

number of carbon layers in the sample (2, 3).

Existing methods for patterning graphene do notallow one to control the number of layers re-moved (4–7). Some of them are designed to workwith a single existing layer of carbonmaterial (4, 5),

Fig. 4. Cross-correlation sig-nals obtained for a few of thepulse trains shown in Fig. 1B.The amplitudes required for thesyntheses are given in table S1:(A) transform-limited subcyclecosine (CEP = 0); (B) transform-limited subcycle sine (CEP= p/2);(C) square pulses with CEP = 0.The corresponding reconstructedfields are shown in the insets.

4 MARCH 2011 VOL 331 SCIENCE www.sciencemag.org1168

REPORTS

on

Mar

ch 7

, 201

1w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

whereas others that were applied toward few-layer graphene (FLG) indiscriminately etch all ofthe layers present in the sample down to thesupporting substrate (6, 7).

In a typical single-atomic-layer lithographyprocedure (Fig. 1A), we coated graphene thatwas supported on a Si/SiO2 wafer with 5 nm ofzincmetal by sputtering (8). The sample was thenplaced in dilute HCl solution (typically 0.02 to0.1 M) for 3 to 5 min to dissolve the zinc. Lastly,the sample was rinsed with water and dried in astream of nitrogen. This series of steps removesone layer of carbon, and the procedure can berepeated to remove additional carbon layers. Forpatterning of FLG, zinc was deposited on thetargeted areas only, thus removing a carbon layerstrictly from those zinc-coated areas.

The method was used initially on grapheneoxide (GO) and chemically converted graphene(CCG). The GO was prepared according to theHummers method (9), modified as we describedpreviously (8, 10). The effect of treatment of thesame GO flake in the course of two consecutivetreatments is shown in Fig. 1, B to D. The scan-ning electron microscopy (SEM) image of theoriginal flake (Fig. 1B) is dark; the features arebarely visible. The flake appears semitransparentafter the first treatment (Fig. 1C) and almostcompletely transparent after the second treatment(Fig. 1D). The flake’s wrinkles and folded areas,which survived after the second treatment (Fig.1D), show the outline of the original flake.

A similar set of SEM images is shown forpatterned GO samples (Fig. 1, E to G). The owlimage on the GO flake that was electron-beam(e-beam)–patterned and Zn/HCl-treated (Fig. 1E)shows that one GO layer can be removed in onetreatment. Figure 1, F and G, shows different frag-ments of a continuous few-layered GO film atop aSi/SiO2 substrate. The samples were first exposedto one Zn/HCl treatment in the shape of horizontalstripes and then to one more treatment in theshape of vertical stripes, as shown in schematicFig. 1A. The treated areas appear lighter in theSEM when compared with the untreated areas,and the GO film acquires a transparent appear-ance. The underlying flakes are not removedfrom the treated areas. The lower-layer flakes thatextend from the treated areas through the un-treated areas are visible. Note that even in theareas exposed to two treatments where horizontaland vertical stripes overlap (marked with “n-2”on Fig. 2, F and G), the number of remaininglayers is not zero. The lower layers that remainare visible in the areas where the original filmcontained three and more GO layers.

In order to determine the number of carbonlayers removed, we supplemented SEM images

with atomic force microscopy (AFM) analysis,which in this case is a sensitive tool because ofthe large interlayer distance in the GO samples.Figure 2 shows a GO flake during the course oftwo consecutive Zn/HCl treatments. The SEMimage of the original flake (Fig. 2A) was dark,whereas the AFM image (Fig. 2B) was wellpronounced. The thickness of the flake’s mainbody (Fig. 2C) was 2.2 to 2.3 nm, which corre-sponds to two carbon layers of highly oxidizedGO (11–13). After the first Zn/HCl treatment, theflake became semitransparent in the SEM image(Fig. 2D), and the wrinkles and folded areas werevisible. The AFM image (Fig. 2E) indicated asmaller height difference. The thickness was 0.8to 0.9 nm (Fig. 2F), which corresponds to onelayer of reduced GO or, namely, CCG (13–15).After the second Zn/HCl treatment (Fig. 2, G and

H), the flake disappeared, and the wrinkles andfolded areas were the only parts still observable.The height profiles (Fig. 2I) were flat, suggestingthat no layered carbon material was left on thesurface except at the former folded regions. Thesharp peaks in the height profiles 5 and 6 arecaused by particulate contamination. These SEMand AFM images along with the height profilesshow that two layers of the bilayer GO flake aresequentially and controllably removed duringtwo consecutive Zn/HCl treatments.

Next we applied the method toward chemicalvapor–deposited (CVD) graphene (Fig. 3), whichwas grown on copper and transferred to a Si/SiO2

substrate (16). Bilayer CVD graphene was fab-ricated by stacking two independently grownmono-layer graphene films. As has been shown, Ramanspectroscopy can be used as a sensitive tool to

Department of Chemistry, Department of Mechanical Engi-neering and Materials Science, Department of Computer Sci-ence, and the Smalley Institute for Nanoscale Science andTechnology, Rice University, MS-222, 6100 Main Street,Houston, TX 77005, USA.

*To whom correspondence should be addressed. E-mail:tour@rice.edu

Fig. 1. Controlled layer-by-layer removal of graphene. (A) Schematic illustration of the method: (Step 1)The bilayer graphene on top of a Si/SiO2 substrate. (Step 2) A patterned layer of zinc metal is sputteredatop the graphene. (Step 3) The zinc is removed by aqueous HCl (0.02 M) in 3 to 5 min with simultaneousremoval of one graphene layer. (Step 4) Patterning of a second zinc stripe. (Step 5) HCl treatment removesthe second stripe of zinc plus the underlying carbon layer. (B to D) SEM image of the same bilayer GOflake: (B) original, (C) after the first, and (D) after the second Zn/HCl treatment. (E) SEM image of amonolayer GO flake patterned in the image of an owl. (F and G) SEM images of a continuous GO filmpatterned with horizontal and vertical stripes in two consecutive Zn/HCl treatments. The lightest squares(an example is marked with “n-2”), where the horizontal and vertical stripes overlap, represent areasexposed to two treatments. Areas exposed to one treatment (examples are marked with “n-1”) are with ashade between the lightest and darkest squares. The darkest squares (examples are marked with “n”)represent the areas with the original untreated GO film.

www.sciencemag.org SCIENCE VOL 331 4 MARCH 2011 1169

REPORTS

on

Mar

ch 7

, 201

1w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

determine the number of graphene layers (17, 18).Figure 3A is the Raman spectrum of the bilayergraphene film fabricated by stacking two mono-layer sheets. The spectrum for the monolayer istaken from the areas where the two precursorsheets do not overlap, and thus it keeps itsmonolayer identity. The two spectra in Fig. 3Awere normalized to bring the 2D peaks to thesame height. The following changes occurred inthe Raman spectrum as the second graphene lay-er was added: the G band dramatically increasedin intensity and slightly red-shifted. A blue shiftof 10 cm−1 was observed for the 2D band, withsome increase of full width at half maximumvalue. These characteristics are indicative of thechange in the electron bands caused by the inter-actions of the graphene layers (17, 18) and areconsistent with the presence of misoriented bilay-er graphene (19–22). The relative intensities oftheG and 2D peaks are distinct for the two spectraand consistent with earlier reports (20–22).

The as-fabricated bilayer graphene film waspatterned with horizontal stripes in one Zn/HCltreatment (Fig. 1B). In the areas where the toplayer was selectively removed, the Raman spec-trum shows the presence of monolayer grapheneonly. The increase in D-band intensity (Fig. 3Cversus Fig. 3A bottom) indicates slight disruptionof the remaining layer, or it might be caused byresidual stripped carbon impurities.

In addition to Raman spectral analysis, theelectrical properties of devices fabricated on abilayer CVD-graphene film were measuredbefore and after the Zn/HCl treatment. The chan-nel width was 3 mm, and the channel length was10 mm. For 60 devices measured, the conductiv-ity decreased 2 to 10 times after removing of onelayer from an original bilayer device. The current-voltage [I(V)] curves before (black) and after thetreatment (red) are shown in Fig. 3D for one ofthe devices where conductivity decreased by afactor of 6 after the treatment, which is an aver-age for the series of experiments. The conduc-tivity of bilayer graphene has been shown to be 4to 10 times higher than that formonolayer graphenebecause of the interaction of the bottom layerwith the SiO2 substrate (13, 23). Moreover, the 2-to-10 range that we recorded is not likely to beassociated with the reproducibility of the Zn/HCltreatment method but with inherent differencesin the original graphene samples, such as mis-matched interaction between graphene layers andcorrugations between the layers. We find similarinconsistencies in electrical properties of as-grownmonolayer CVD graphene itself, wherein devicesfabricated on the same monolayer graphene filmbut separated by just 5 mm show differences inconductivities in this same range.

Finally, the graphene layer removal methodwas effective when applied to few-layer micro-

mechanically cleaved graphene and highly orderedpyrolytic graphite (HOPG). As shown in figs. S1and S2, the Zn/HCl protocol can be used to se-quentially remove layer after layer of the micro-mechanically cleaved graphene.

In order to probe the mechanism of thereaction, in a control experiment we depositedzinc metal on CVD graphene by thermal evap-oration instead of sputtering. No change in thegraphene structure was observed after treatmentwith HCl. However, a carbon layer was success-fully removed from GO and CVD graphene bysputter-coating 5 nm of aluminum and dissolvingthe latter in either 0.1 M HCl or in concentratedaqueous NaOH. Thus, a key factor in removingthe carbon layer in graphene is the sputteringprocess, which forms defects in the top layer,allowing its further removal. Sputtered metalatoms have about 100 times more energy thanthose produced by thermal evaporation (24). Ithas been shown that irradiation of graphite bylow energy ion beams produces atomic-sizedefects that do not extend further than the toplayer (25, 26). The threshold energy for carbonatom displacement has theoretical estimates of 18to 22 eVand experimental values of 18 to 20 eVfor e-beam irradiation (27). At the same time, thecorresponding experimental energy values forion irradiation are >30 eV, with the most fre-quently reported being 31 to 34.5 eV (28–30).

Fig. 2. Bilayer GO flake in thecourse of two consecutive treat-ments. (A) SEM image, (B) AFMimage, and (C) height profiles ofthe original bilayer GO flake. (Dto F) Corresponding images andheight profiles for the same flakeafter the first Zn/HCl treatment.(G to I) Images and height pro-files of the same flake after thesecond treatment. Height pro-files 1 to 6 on (C), (F), and (I) aretaken across the correspondinglynumbered lines indicated on (B),(E), and (H). The blue arrows onthe height profiles indicate theflake’s boundaries, and the redtriangles indicate the locationsbetween which the heights weredetermined. All the horizontalscale bars indicate 10 mm.

4 MARCH 2011 VOL 331 SCIENCE www.sciencemag.org1170

REPORTS

on

Mar

ch 7

, 201

1w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

These values are ion dependent, and the valuesfor Ar+ and Kr+ are 47.3 eVand 80.8 eV, respec-tively, to yield one-carbon atom vacancy defects(26). There are no reports of threshold energiesfor metal atoms needed to form similar defects.The average energy for sputtered metal atoms,including zinc, is 0.5 to 3 eV (24, 31). For zinc,the percentage of atoms having 20 eVand 34 eVis only 0.8% and 0.5%, respectively (31); henceonly a small number of the sputtered atoms willhave sufficient energy for carbon-atom displace-ment in the graphene. However, the ability toform defects is not solely a function of the in-cident particle energy, but it also depends on itschemical identity (26, 30).

To further investigate the effect of sputteringon graphene, we separately sputter-coated CVDgraphene and GO samples with as little as 0.6 nmof zinc, aluminum, copper, or gold, followed byanalysis of the sputtered products before dissolv-ing the metals. Assuming a densely packed anduniform coating, 0.6 nm should provide a three-to four-atom-thick layer ofmetal atop of graphene.Figure 4A is a transmission electron microscopy(TEM) image of CVD graphene suspended on agrid after sputter-coating with 0.6 nm of zinc.Although sputtering forms some holes in thegraphene, the larger part of the carbon network isnot broken. The graphene sheet not only re-mained suspended on the grid but also supportedthe zinc clusters. However, the effect of sputter-ing onto graphene atop an SiO2 surface might bedifferent from graphene suspended on a grid. Zincatoms sputtered atop GO samples on a Si/SiO2

substrate under the same conditions do not formdiscernible clusters in the SEM (Fig. 4B). Thezinc nanoparticles are visible on the Si/SiO2

substrate surface but not on the GO flake. Thezinc atoms likely form chemical bonds with theoxygen functionalities of the GO, thus preventingthe agglomeration of the zinc.

Raman spectra of the sputter-coated CVDgraphene samples (Fig. 4C) have the features ofamorphous carbon (18, 19), suggesting that thegraphene conjugation has been severely disrupted.The carbon 1s (C1s) x-ray photoelectron spec-troscopy (XPS) data (Fig. 4, D and E) also exhibitnoticeable changes in comparison with untreatedgraphene. The main carbon peak at 284.8 eVbroadens, and a small carbonyl group signalappears at 288.0 to 289.0 eV. The main peakbroadening is consistent with the presence of sp3-hybridized carbon in addition to sp2-hybridizedcarbon (32), suggesting a substantial transfor-mation in the structure. The presence of carbonylgroups provides evidence for the formation ofvacancy defects because carbonyl groups canexist only at the edges of graphene. The effect ofdifferent metals on the XPS spectrum is apparent.Gold and copper cause less main peak broaden-ing, but they introduce oxygen functionalities, asevidenced by the shoulder in the 286.0 to 289.0eVregion for gold and the well-pronounced C=Opeak at 288.6 eV for copper. Aluminum deposi-tion leads to broadening of the main carbon peak,

Fig. 3. Removing a carbon layer from CVD graphene. (A) Raman spectra of a bilayer CVD graphene film(top) as fabricated by stacking two monolayer (bottom spectrum) graphene films. The two spectra arescaled to have the same 2D peak height. a.u., arbitrary units. The Raman spectra here and in (C) weretaken with 514-nm laser excitation. (B) SEM image of a bilayer CVD graphene film patterned by horizontalstripes in one Zn/HCl treatment. The yellow numbers indicate the number of carbon layers present afterthe treatment. (C) Raman spectra of the bilayer graphene film exposed to one Zn/HCl treatment [markedwith “1” in (B)]. (D) Two I(V) curves measured for the same device: the as-fabricated bilayer graphene(black line) and after exposing it to one Zn/HCl treatment (red line).

Fig. 4. Analysis of metal-sputtered graphene samples. (A) TEM image of monolayer CVD graphenesputter-coated with 0.6 nm zinc. Dark, 1- to 3-nm-sized zinc clusters are visible on the graphene surface.(B) SEM image of a GO flake on a Si/SiO2 wafer sputter-coated with 0.6 nm zinc. Zinc clusters appear on thewafer surface, but no clustering is observed on the GO surface. (C) Raman spectra and (D and E) C1s XPSspectra of monolayer CVD graphene on a Si/SiO2 wafer sputter-coated with 0.6 nm of four different metals:zinc, aluminum, gold, and copper. In (D), black line, original graphene; blue, aluminum-coated; and red,zinc-coated. In (E) black line, original graphene; dark cyan, copper-coated; and magenta, gold-coated.

www.sciencemag.org SCIENCE VOL 331 4 MARCH 2011 1171

REPORTS

on

Mar

ch 7

, 201

1w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

as does zinc, but aluminum does not afford asmany carbonyls. Lastly, zinc sputtering causesboth the main peak to broaden and oxidizesgraphene to produce carbonyl groups. The oxida-tion can be explained by reaction of the carbonradicals at the newly formed defect sites withtraces of oxygen present in the sputtering chamber(5 × 10−5 torr) or upon subsequent exposure to air.The XPS data indicate that zinc sputtering, amongthe four testedmetals, most affects the graphene asseen by the broadening of the main C1s signal.Interestingly, GO samples sputter-coated with me-tals behave somewhat differently than do CVDgraphene samples, but both form similar materialscharacterized by defective carbon layers with asignificant content of carbonyl groups; the resultsof those experiments are in (8) and fig. S3.

Sputtering damages graphene but does notremove it. Dissolving the zinc in acid is anessential step in removing the carbon layer. Toinvestigate the role of the dissolving agent, wetreated sputtered metals with different solutions.It was found that the carbon layer was removedwhen dissolving sputtered zinc using HCl orwhen dissolving sputtered aluminum using eitherHCl or concentrated NaOH. In all of those mix-tures, H2 gas is evolved. Conversely, no carbonlayer was removed when dissolving sputteredcopper in HCl/CuCl2 or when dissolving sput-tered gold in KI/I2; no gas is evolved in thosemetal dissolutions. Although gas evolution prob-ably plays a role in the carbon layer removal, afurther difference of the two carbon-strippingmetals, zinc and aluminum, versus the non–carbon-stripping metals, copper and gold, is in

their oxidation potentials. The combination oftop-layer hole formation in the graphene, thelarge oxidation potentials of zinc and aluminum,as well as the evolved gas during their dissolu-tion, all likely contribute to the mechanistic ef-ficacy of the process.

The resolution that should be obtainable fromthis technique is being primarily dictated by theresolution of themetal patterningmethod. A 100-nm-wide zinc line was formed atop graphenewith e-beam lithography, and a commensuratelysized removal of the carbon layer was affordedby treatment with HCl (fig. S4). Furthermore, onthe basis of AFM data, no delamination on thefabricated pattern corners or edges was observedwhile removing carbon layers fromGO or HOPG.Over larger areas, scrolling of the edges wasnever observed by SEM.

We have demonstrated layer-by-layer removalof carbon from four types of graphene. This workdemonstrates single-atomic-layer lithography inmultilayer graphene structures and could provideimpetus for the design of new graphene deviceembodiments.

References and Notes1. A. K. Geim, K. S. Novoselov, Nat. Mater. 6, 183 (2007).2. K. S. Novoselov et al., Science 306, 666 (2004).3. K. S. Novoselov et al., Nat. Phys. 2, 177 (2006).4. G. Lu et al., Langmuir 26, 6164 (2010).5. L. C. Campos, V. R. Manfrinato, J. D. Sanchez-Yamagishi,

J. Kong, P. Jarillo-Herrero, Nano Lett. 9, 2600 (2009).6. L. B. Gao et al., J. Am. Chem. Soc. 131, 13934 (2009).7. S. S. Datta, D. R. Strachan, S. M. Khamis, A. T. C. Johnson,

Nano Lett. 8, 1912 (2008).8. Materials and methods are available as supporting

material on Science Online.

9. W. S. Hummers, R. E. Offeman, J. Am. Chem. Soc. 80,1339 (1958).

10. D. C. Marcano et al., ACS Nano 4, 4806 (2010).11. N. I. Kovtyukhova et al., Chem. Mater. 11, 771 (1999).12. S. Stankovich et al., Carbon 45, 1558 (2007).13. C. Gómez-Navarro et al., Nano Lett. 7, 3499 (2007).14. V. C. Tung, M. J. Allen, Y. Yang, R. B. Kaner, Nat.

Nanotechnol. 4, 25 (2009).15. C. Mattevi et al., Adv. Funct. Mater. 19, 2577 (2009).16. X. Li et al., Science 324, 1312 (2009); 10.1126/

science.1171245.17. A. C. Ferrari et al., Phys. Rev. Lett. 97, 187401 (2006).18. A. C. Ferrari, Solid State Commun. 143, 47 (2007).19. P. Lespade, A. Marchand, M. Couzi, F. Cruege, Carbon

22, 375 (1984).20. P. Poncharal, A. Ayari, T. Michel, J.-L. Sauvajol, Phys. Rev.

B 78, 113407 (2008).21. Y. Hao et al., Small 6, 195 (2010).22. S. Unarunotai et al., ACS Nano 4, 5591 (2010).23. J.-H. Chen et al., Solid State Commun. 149, 1080 (2009).24. K. Wasa, S. Hayakawa, Handbook of Sputter Deposition

Technology: Principles, Technology, and Applications(Noyes, Park Ridge, NJ, 1991).

25. J. R. Hahn, H. Kang, S. Song, I. C. Jeon, Phys. Rev. B 53,R1725 (1996).

26. J. R. Hahn, H. Kang, Phys. Rev. B 60, 6007 (1999).27. F. Banhart, J. Kotakoski, A. V. Krasheninnikov, ACS Nano

5, 26 (2011).28. F. Banhart, Rep. Prog. Phys. 62, 1181 (1999).29. G. L. Montet, G. E. Myers, Carbon 9, 179 (1971).30. D. Marton, K. J. Boyd, T. Lytle, J. W. Rabalais, Phys. Rev. B

48, 6757 (1993).31. M. Szymoński, Appl. Phys. (Berl.) 23, 89 (1980).32. F. Larachi, S. Dehkissia, A. Adnot, E. Chornet, Energy

Fuels 18, 1744 (2004).

Supporting Online Materialwww.sciencemag.org/cgi/content/full/331/6021/1168/DC1Materials and MethodsFigs. S1 to S4

18 October 2010; accepted 31 January 201110.1126/science.1199183

Helix-Rod Host-Guest Complexeswith Shuttling Rates Much Fasterthan DisassemblyQuan Gan,1,2,3 Yann Ferrand,2 Chunyan Bao,2 Brice Kauffmann,2 Axelle Grélard,2

Hua Jiang,1* Ivan Huc2*

Dynamic assembly is a powerful fabrication method of complex, functionally diverse moleculararchitectures, but its use in synthetic nanomachines has been hampered by the difficulty of avoidingreversible attachments that result in the premature breaking apart of loosely held moving parts. Weshow that molecular motion can be controlled in dynamically assembled systems through segregationof the disassembly process and internal translation to time scales that differ by four orders ofmagnitude. Helical molecular tapes were designed to slowly wind around rod-like guests and then torapidly slide along them. The winding process requires helix unfolding and refolding, as well as a strictmatch between helix length and anchor points on the rods. This modular design and dynamic assemblyopen up promising capabilities in molecular machinery.

Controllingmotion at themolecular scale iskey to the engineering of synthetic nano-machines (1–7). Suchcontrol canbeexerted

upon introducing irreversible bonding betweendifferent moving parts of a molecular structure sothatmotion only takes place in defined directions.In rotaxanes (8–11), rings are irreversibly locked

around rods alongwhich theymay slide. In catenanes(12), interlocked rings may rotate around one an-other. An important improvement to these designswould entail not forbidding undesired motionsbut rather segregating them to much longer timescales, just as biological molecular machines self-assemble and disassemble slowlywhile their work

regime is rapid (13). We now have implementedthis concept in molecular shuttles that slowly bindto rods and then rapidly slide along them withoutrequiring an irreversible attachment. The shuttlesconsist of oligomeric aromatic amide moleculartapes that fold into very stable helices possessinga cylindrical cavity with high affinities for com-plementary rod-like guests. A guest terminatedwith bulky ends cannot thread itself in the cavitybut nevertheless forms a complex via a slow helixunfolding-refolding process, thereby allowing thehelix to rapidly shuttle along the guest withoutdissociating.

Aromatic oligoamides 1,2, and3 (Fig. 1A)weredesigned according towell-established rules (14,15)to fold into helical structures stabilized by localpreferential conformations at aryl-amide linkagesand intramolecular p-p interactions between aro-matic groups (fig. S1) (16). Based on studies of

1Beijing National Laboratory for Molecular Sciences, CAS KeyLaboratory of Photochemistry, Institute of Chemistry, ChineseAcademy of Sciences, Beijing, 100190, China. 2Institut Européende Chimie et Biologie, Université de Bordeaux–CNRS UMR5248and UMS3033, 2 rue Robert Escarpit, F-33607 Pessac, France.3Graduate School of Chinese Academy of Sciences, Beijing,100049, China.

*To whom correspondence should be addressed. E-mail:hjiang@iccas.ac.cn (H.J.); i.huc@iecb.u-bordeaux.fr (I.H.)

4 MARCH 2011 VOL 331 SCIENCE www.sciencemag.org1172

REPORTS

on

Mar

ch 7

, 201

1w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from