DOI: 10.1126/science.1258544, 1092 (2014);346 Science

et al.Jae-Hwang Leeprojectile penetrationDynamic mechanical behavior of multilayer graphene via supersonic

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

  ): November 27, 2014 www.sciencemag.org (this information is current as of

The following resources related to this article are available online at

http://www.sciencemag.org/content/346/6213/1092.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/2014/11/25/346.6213.1092.DC1.html can be found at: Supporting Online Material

http://www.sciencemag.org/content/346/6213/1092.full.html#ref-list-1, 4 of which can be accessed free:cites 27 articlesThis article

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

subject collections:This article appears in the following

registered trademark of AAAS. is aScience2014 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

Nov

embe

r 28

, 201

4w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

o

n N

ovem

ber

28, 2

014

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

on

Nov

embe

r 28

, 201

4w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

o

n N

ovem

ber

28, 2

014

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

on

Nov

embe

r 28

, 201

4w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

o

n N

ovem

ber

28, 2

014

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

chondrule formation took place in the weaklyionized “dead zone,” which contains gas poorlycoupled to local magnetic fields and occurs within~3 gas scale heights of the midplane (35). Ourmeasurements therefore indicate that a substan-tial magnetic field [ yet still well below the ~400 mTequipartition field strength (3)] existed in the deadzone, potentially as a result of fields inherited fromthe collapse of the solar system’s parent molecularcloud. Given our measured field strengths, mass ac-cretion driven by the MRI or magnetic braking at2.5 AU would have been <0.04 × 10−8 to 3.5 × 10−8

Msun year−1, where Msun is the Sun’s mass (sup-

plementary text). Meanwhile, the MCW modelwould predict mass accretion rates of 0.3 × 10−7

to 30 × 10−7Msun year−1 or less. The inferred age of

Semarkona chondrules is 2 to 3 My after the firstcalcium alunimum–rich inclusions (36). Given thatprotoplanetary disks are observed to have ac-cretion rates of 10−9-10−7 Msun year

−1 at 2 to 3 Myafter collapse of their parent molecular clouds (2),both magnetic mechanism could fully account forthe expected accretion rates. This suggests thatmagnetic fields govern the observed rapid trans-formation of protoplanetary disks into planetarysystems around Sun-like stars.

REFERENCES AND NOTES

1. K. E. Haisch Jr., E. A. Lada, C. J. Lada, Astrophys. J. 553,L153–L156 (2001).

2. L. Hartmann, N. Calvet, E. Gullbring, P. D'Alessio,Astrophys. J. 495, 385–400 (1998).

3. M. Wardle, Astrophys. Space Sci. 311, 35–45 (2007).4. X.-N. Bai, J. Goodman, Astrophys. J. 701, 737–755 (2009).5. N. J. Turner, S. Fromang, C. Gammie, H. Klahr, G. Lesur,

M. Wardle, X.-N. Bai, in Protostars and Planets VI (Univ. ofArizona Press, Tucson, AZ, 2014).

6. R. M. Crutcher, Annu. Rev. Astron. Astrophys. 50, 29–63(2012).

7. I. W. Stephens et al., Nature 514, 597–599 (2014).8. J. N. Cuzzi, R. C. Hogan, K. Shariff, Astrophys. J. 687,

1432–1447 (2008).9. T. Nakamura et al., Science 333, 1113–1116 (2011).10. F. H. Shu, H. Shang, T. Lee, Science 271, 1545–1552 (1996).11. S. J. Desch, H. C. Connolly Jr., Meteorit. Planet. Sci. 37,

183–207 (2002).12. S. J. Desch, T. C. Mouschovias, Astrophys. J. 550, 314–333

(2001).13. N. Sugiura, D. W. Strangway, Proc. Lunar Planet. Sci. Conf. 15th,

C729 (1985).14. R. R. Fu, E. A. Lima, B. P. Weiss, Earth Planet. Sci. Lett. 404,

54–66 (2014).15. H. Leroux, G. Libourel, L. Lemelle, F. Guyot, Meteorit. Planet.

Sci. 38, 81–94 (2003).16. M. Uehara, N. Nakamura, Earth Planet. Sci. Lett. 250,

292–305 (2006).17. S.-C. L. L. Lappe et al., Geochem. Geophys. Geosyst. 12,

Q12Z35 (2011).18. M. Winklhofer, K. Fabian, F. Heider, J. Geophys. Res. 102,

22695 (1997).19. I. Garrick-Bethell, B. P. Weiss, Earth Planet. Sci. Lett. 294,

1–7 (2010).20. R. J. Reisener, J. I. Goldstein, in Lunar Planet. Sci. Conf. XXX

(Lunar and Planetary Institute, Houston, TX, 1999), pp. 1868.21. D. Stöffler, K. Keil, E. R. D. Scott, Geochim. Cosmochim. Acta

55, 3845 (1991).22. R. R. Fu et al., Science 338, 238–241 (2012).23. J. Gattacceca et al., Meteorit. Planet. Sci. 49, 652–676 (2014).24. C. M. O. D. Alexander, D. J. Barber, R. Hutchison, Geochim.

Cosmochim. Acta 53, 3045–3057 (1989).25. G. D. Cody et al., Earth Planet. Sci. Lett. 272, 446–455 (2008).26. B. P. Weiss, E. A. Lima, L. E. Fong, F. J. Baudenbacher,

J. Geophys. Res. 112, B09105 (2007).27. E. A. Lima, B. P. Weiss, R. R. Fu, A. C. Bruno, in AGU Fall

Meeting (San Francisco, 2013), vol. GP43B-07.28. G. S. Watson, Mon. Not. R. Astron. Soc. 7, 160–161 (1956).

29. C. Suavet, J. Gattacceca, P. Rochette, L. Folco, Geology 39,123–126 (2011).

30. S.-C. L. L. Lappe, J. M. Feinberg, A. Muxworthy, R. J. Harrison,Geochem. Geophys. Geosyst. 14, 2143–2158 (2013).

31. H. Miura, T. Nakamoto, M. Doi, Icarus 197, 269–281 (2008).32. E. H. Levy, S. Araki, Icarus 81, 74–91 (1989).33. C. P. McNally, A. Hubbard, M.-M. Mac Low, D. S. Ebel,

P. D'Alessio, Astrophys. J. 767, L2 (2013).34. S. J. Bus, R. P. Binzel, Icarus 158, 146–177 (2002).35. X.-N. Bai, Astrophys. J. 739, 50 (2011).36. S. Mostefaoui et al., Meteorit. Planet. Sci. 37, 421–438

(2002).

ACKNOWLEDGMENTS

We thank S. A. Balbus, A. J. Brearley, H. C. Connolly, A. M. Hughes,B. C. Johnson, J. L. Kirschvink, M. Mac Low, G. J. MacPherson,M. I. Petaev, D. D. Sasselov, H. E. Schlichting, J. B. Simon,N. Turner, and B. Zanda for discussions that improved themanuscript. We also thank J. Gross, S. Wallace, and Z. I. Balogh forhelp with SEM and STEM sample analyses and acknowledgeS.-C. L. L. Lappe, N. S. Church, S. Russell, M. Uehara, andN. Nakamura for pioneering work on the magnetism of dustyolivines. We thank T. F. Peterson for supporting critical

instrumentation and analysis costs. R.R.F., B.P.W., E.A.L., S.J.D.,and C.S. thank the NASA Origins Program, while R.R.F. and B.P.W.thank the U.S. Rosetta Project, Jet Propulsion Laboratory forsupport. R.R.F. thanks the NSF Graduate Research FellowshipProgram, and C.S. thanks the NASA Lunar Science Institute andthe NASA Solar System Exploration and Research Virtual Institutefor support. R.J.H. and T.K. thank the European Research Councilunder the European Union’s Seventh Framework Programme andthe Leverhulme Trust for support. X.N.B. acknowledges supportfrom NASA through the Hubble Fellowship. D.G., D.L.S., and R.L.W.thank the Defense Advanced Research Projects Agency QuASARprogram and the NSF for support.

SUPPLEMENTARY MATERIALS

www.sciencemag.org/content/346/6213/1089/suppl/DC1Supplementary TextFigs. S1 to S12Tables S1 to S4References (37–125)Database S1

30 June 2014; accepted 31 October 2014Published online 13 November 2014;10.1126/science.1258022

MATERIALS SCIENCE

Dynamic mechanical behavior ofmultilayer graphene via supersonicprojectile penetrationJae-Hwang Lee,1,2* Phillip E. Loya,1 Jun Lou,1 Edwin L. Thomas1*

Multilayer graphene is an exceptional anisotropic material due to its layered structurecomposed of two-dimensional carbon lattices. Although the intrinsic mechanicalproperties of graphene have been investigated at quasi-static conditions, itsbehavior under extreme dynamic conditions has not yet been studied. We report thehigh–strain-rate behavior of multilayer graphene over a range of thicknesses from10 to 100 nanometers by using miniaturized ballistic tests. Tensile stretching of themembrane into a cone shape is followed by initiation of radial cracks that approximatelyfollow crystallographic directions and extend outward well beyond the impact area.The specific penetration energy for multilayer graphene is ~10 times more than literaturevalues for macroscopic steel sheets at 600 meters per second.

Graphene, the atomic monolayer buildingblock of graphite, is known for its excep-tionally high intrinsic strength and stiff-ness arising from the two-dimensional (2D)hexagonal lattice of covalently bonded

carbon atoms. Recently, graphene’s in-planeYoung’s modulus (Y‖) wasmeasured to bemorethan 1.0 TPa using atomic force microscopenanoindentation (1). Because tensile mechani-cal stresses in a material cannot be transmittedfaster than the speed of sound [c~ (Y/r)1/2, wherer is the density of the material], the nonequil-ibrium local stress arising from the inertial effectbecomes important under dynamic conditionsaccompanying high–strain-rate, predominantly

tensile loading (2). In this regard, the relativelylow density (~2200 kg m−3) of graphene (3),along with its high modulus, leads to a superiorin-plane speed of sound (c‖ ~ 22.2 km s−1), im-plying that concentrated stresses applied underextreme conditions can rapidly be delocalized.Nanoindentation has served as an effective

technique to study the tensile mechanical prop-erties of monolayer graphene. It is inherentlya low-speed test (<<1 m s−1), but strain ratescan reach ~105 to 106 s−1 for very thin samples(4), whereas most high-speed, high–strain-ratemechanical characterization techniques, such assplit-Hopkinson pressure bar (5) and ballistictests (6), are inappropriate for testing very thinspecimens. To address high-speed and high–strain-rate tensile-dominated penetration of thinfilms, we improved our laser-induced projectileimpact test (LIPIT) (7). In this advanced LIPIT(or “a-LIPIT”), a single micrometer-size solidsilica sphere (or “m-bullet”) is fired at a high speed

1092 28 NOVEMBER 2014 • VOL 346 ISSUE 6213 sciencemag.org SCIENCE

1Department of Materials Science and NanoEngineering, RiceUniversity, Houston, TX 77005, USA. 2Department of Mechanicaland Industrial Engineering, University of Massachusetts,Amherst, MA 01003, USA.*Corresponding author. E-mail: leejh@umass.edu (J.-H.L.); elt@rice.edu (E.L.T.)

RESEARCH | REPORTS

(<3 km s−1) with a high aiming accuracy (<1.1°deflection) toward a thin film. The velocity ofthe m-bullet is measured before and after pen-etration to determine the energy lost during thetest. We employed multilayer graphene (MLG)membranes in a range of thicknesses (10 to100 nm, equivalent to 30 to 300 graphene layers)to apply localized, very–high–strain-rate tensiledeformation (~107 s−1) at a specific area. As thethickness of the MLG membranes (h) is alwaysconsiderably less than the diameter (D) of am-bullet (D/h ≥ 40), the high–strain-rate in-planetensile behavior of MLG can be assessed fromthe thickness-dependent characteristics of theenergy required for projectile penetration throughthe membrane.The MLG membranes are prepared by me-

chanical exfoliation of highly ordered pyrolyticgraphite (grade SPI-1, SPI Supplies; fig. S1), asshown inFig. 1A. A silicam-bullet [D=3.7 T 0.02mm,based on imaging by scanning electron micros-copy (SEM)] is propelled by expanding gases cre-ated by the laser ablation of a gold film (~50 nm

thick). A 20-mm-thick elastomeric layer of cross-linked polydimethylsiloxane is used to confinethe ablation products, eliminates the temperaturerise of the m-bullet, and diminishes the strengthof the shock waves propagating through air.The impact speed (vi) is measured (with the re-producibility of vi within T 2%) using a triple-exposure photograph of the moving m-bulletwith T 10 m s−1 error, where the time gap be-tween the three exposures is achieved by em-ploying different travel distances for each laserpulse (fig. S2). The residual speed (vr) of them-bullet is similarly measured after the m-bullethas traveled a certain distance (d) beyond themembrane (Fig. 1B). An average thickness (have)of the impact area was determined using athickness-dependent optical transmittance mea-surement (fig. S3) given by an analytical formula(8) and the optical parameters of graphite (9).An additional postpenetration image allowedidentification of the penetration area (Fig. 1C).The schematic in Fig. 1D depicts the series of

events during penetration accompanying the ki-

netic energy loss of a m-bullet (DEk). As them-bullet impacts a strike face area (As = pD2/4),an elastic wave radially propagates at c‖ and aconic deformation of the MLG membrane fol-lows with a radial speed of its base, vc, which isgenerally slower than c‖ (step ii). The primaryforce is axisymmetric tension with a strong ra-dial gradient and results in the cone shape.Typically, three to six cracks are initiated nearthe center of As and propagate outward in theradial direction (step iii), resulting in the crea-tion of the same number of petals. The trans-ferredmomentum to theMLGmembrane inducescreasing and folding of each triangular-shapedpetal at its base while the elastic extension of themembrane is rapidly relaxed along the radialdirection via snap-back (step iv). The longest crackfrom the impact center is defined as the maxi-mum crack distance, Lmax, which we use as theestimation of the final radius of the conic de-formation due to the reduction of the tangentialstress. DEk is composed of two terms, the net en-ergy to penetrate amembrane (Ep) and the energyloss due to air drag (Eair). Ep depends on variousenergy dissipation mechanisms including elasticstretching of the membrane, fracture, and heat-ing, as well as the kinetic energy transfer to themembrane petals and membrane debris.

DEk ¼ 1

2mðv2i − v2r Þ ¼ Ep þ EairðdÞ ð1Þ

As the mass of a m-bullet, m is calculated to be5.0 T 0.1 × 10−14 kg, based on the measured di-ameter D and the density of silica (1900 kg m−3)provided by its vendor (microParticles GmbH).The incident Ek is 9 nJ at 600 m s−1 and 21 nJ at900 m s−1 while DEk is in the range of 1 to 5 nJ.The measured deceleration by air drag (aair) is~0.6 × 108 m s−2, assuming constant deceler-ation, which yields Eair = maaird ~1.07 nJ for d ~350 mm (the travel distance of a m-bullet afterpenetration when we take the triple exposure).Therefore, the primary contribution to the ki-netic energy loss is the net energy to penetratea MLGmembrane (Ep) under our experimentalconditions.The MLG membranes had a typical lateral

grain size of ~10 mm. Due to the stress concen-tration at the impact site, the grain boundaryeffects (10, 11) (if any; see the supplementarymaterials) would occur only if the grain bounda-ries exist withinAs. As the ratio,D/h is quite large(40 to 350), the mechanical response of the MLGfilm depends primarily on its in-plane tensilestrength under a high strain rate. A typical pe-netration hole features a set of petals (Fig. 2). Thearea directly beneath the m-bullet impact showsextensive damage through complex, fine-scalefractures, folding, delamination, and loss of partsof the membrane (indicated by the yellow ar-rowheads). As the initiation of the radial cracksmay not be exactly at the impact center, asymme-tric shaped petals are often observed (e.g., Fig. 2, AandD). The damage area is thusmuchwider thanAs, in strong contrast to the observed behavior ofpolycrystalline gold and amorphous, glassypoly(methylmethacrylate) (PMMA)membranes,

SCIENCE sciencemag.org 28 NOVEMBER 2014 • VOL 346 ISSUE 6213 1093

Sample holder

µ-bullet (1000 m s-1)

ExpandingPDMS film

t1

t2

t3

µ-bullet(D = 3.7 µm) PDMS film

(20 µm thick)Gold film

(50 nm thick)

Glass substrate(210 µm thick)

Ablation area

Sample holder

Target sample

Glue

500 µm

MLG

vi

m, D

(i)

vc c||

h

(ii)

(iii)

vr

(iv)Petal

Lmax

Highly -deformed tipFolding

Radiallypropagating

cracks

Rounded cone

Radial tension

Tangential tension

Strike face area, As

d

Snapback

Crease

Debris

c||

θA

Fig. 1. Themicroballistic experiment. (A) Scheme of the experiment. PDMS, polydimethylsiloxane. (B)Side-view image of a moving m-bullet taken by triple exposure at time steps t1 to t3. (C) MLG membraneon a sample holder after a-LIPIT. Three separate impact test regions are highlighted by green backlight.(D) Schematic illustration of penetration steps: (i) prepenetration stage; (ii) conic deformation stage; (iii)fracture stage; and (iv) postpenetration stage, showing the film morphology after penetration andrelaxation. Scale bars in (B) and (C), 50 mm.

RESEARCH | REPORTS

inwhich penetration results in a circular hole withanareaof ~As (see figs. S6 andS7). Correspondingly,a much smaller penetration energy is measured.Many independent penetration experiments

(47 events for vi = 600m s−1 and 43 events for vi =

900m s−1) were carried out for statistical analysis(figs. S4 and S5). The average apex angles (qA inFig. 1D) of petals for the 600– and 900–m s−1

projectile velocities are 83 T 21° (for 107 cracks)and 70 T 23°(for 93 cracks), respectively, indicat-

ing that the higher tangential stresses inducedat 900 m s−1 were relaxed through more radialcracks. Despite the in-plane isotropic elastic na-ture due to the approximate sixfold symmetry ofgraphene (12), the preferential crack propagation

1094 28 NOVEMBER 2014 • VOL 346 ISSUE 6213 sciencemag.org SCIENCE

have=29 nm have=93 nm

have=99 nmhave=30 nm

58˚

93˚

114˚

95˚

113˚

94˚89˚

64˚

42˚

86˚

64˚86˚

74˚

65˚

58˚77˚

0 30 60 90 120 1500

5

10

Total 107 cracks

Num

ber

of p

etal

s

Angle between adjacent cracks, θA (°)

vi=600 m s-1<θ

A>=83°

0 30 60 90 120 1500

5

10

Total 93 cracks

<θA>=70° v

i=900 m s-1

Num

ber

of p

etal

s

Angle between adjacent cracks, θA (°)

Fig. 2. Representative penetration features of MLGmembranes. (A and B) SEM images of petals, radial cracks, folds, and snap-back damage to the petaltips and (C) the adjacent crack-pair angle distribution for vi = 600ms−1. (D andE) SEM images and (F) the adjacent crack-pair angle distribution for vi = 900ms−1.The inset in (F) shows the armchair (red) and zigzag (green) directions.The circles in the SEM images show As. Scale bars in (A), (B), (D), and (E), 5 mm.

Fig. 3. Damagefeatures of a thinMLG membrane. (A)Bright-field TEMmicrograph of theimpact region(have ~ 10 nm) forvi = 900 m s−1. (B)Higher magnification ofthe petal apex areaindicated by the redarrowhead in (A).(C) Three dark-fieldTEM micrographs areoverlaid to show bendcontours, rotation-tiltmoiré fringes, and afine-scale mosaictexture resulting fromsnap-back andmembrane folding.The two diffractionpatterns show (C1) atypical hexagonalspot pattern of theundeformed film and(C2) the altered, multireflection-satellite spot pattern from near the impact region. The red and blue regions correspond to imaging with g ¼ ð0110Þ, andthe green region corresponds to g ¼ ð2110Þ. Scale bars, (A) 5 mm; (B) and (C), 0.5 mm.The voidlike features at the upper left in (A) are a result of residualwater-soluble polymer from film preparation.

RESEARCH | REPORTS

directions ⟨1100⟩ (the armchair direction) and⟨2110⟩ (the zigzag direction) cause preferred an-gles between adjacent cracks to be a multiple of30° (13). Correlation to the underlying crystallo-graphic orientation of the membrane is noted inthe distribution of the angle between adjacentcracks displaying preferences for small multiplesof 30° (Fig. 2, C and F).Transmission electron microscopy (TEM) was

carried out on snap-back portions of the petalregions of thin (have ~ 10 nm) fractured mem-branes (Fig. 3). The bright-field TEM image showscomplicated local folding of themembrane nearthe penetration (Fig. 3, A and B). The higher-magnification dark-field image shows bendcontours and moiré fringes (upper left and leftcenter regions in Fig. 3C, resulting from theinterference of electrons scattered from super-posed folded MLG regions; see diffraction pat-tern insets in Fig. 3C). In the extensively foldedregion nearest the impact origin, a fine-scalemosaic structure is evident in both the electrondiffraction pattern and dark-field images due tothe deformation resulting from the rapid elasticrelaxation (i.e., petal snap-back).Despite a relatively wide fluctuation in Lmax,

its lower limit is well fit by 0.1have +D/2 (Fig. 4A).The film thickness inhomogeneity is representedby the coefficient of variation (CV) of the localthicknessesmeasured over a circular area (radiusr = Lmax). MLG membranes that have CV > 10%in the impact area clearly lead to a greater fluctua-tionofLmax. The radial speedof the circumferentialbase of the expanding cone-shape deformationregion of a membrane can be approximated byvc ≅ 1.23c‖[vi/2

1/2c‖)]2/3 and thus scales with the

cube root of the in-plane speed of sound in thematerial (14). Values for vc correspond to 1950and 2560m s−1 for impacts of 600 and 900m s−1,

respectively. An empirical estimation of deforma-tion parameters is then possible by setting the lowerlimit of Lmax to the maximum radius of the cone,because the tangential tensile stress (Fig. 1D) isthe origin of the radial cracks. For example, assum-ing a simple 1Dmodel for a 900–m s−1 impact to a50-nm-thick MLG membrane, we estimate thepenetration time tp ≅ Lmax/vc ~ 3 ns; the 1D-approximate average maximum tensile strainemax ≅ (vitp/Lmax)

2/2 = ~6%, close to the lowerboundary of the reported failure strain range, 5to 25% (1, 15, 16); and the 1D-approximate av-erage tensile strain rate as given by the maxi-mum strain divided by the penetration time ore:max ¼ ðvi=LmaxÞ2tp=2 ∼ 107 s−1. A further dis-cussion of the estimation of strain and strain rateis available in the supplementary materials.From Eq. 1, DEk(have) can be fit with a linear

function (Fig. 4B), where the y-intercept valuecorresponds to Eair. Therefore, the intrinsic en-ergy dissipation of a MLGmembrane is given byEp(have) = 0.026have and 0.030have for the twovelocities we employed, and a similar trend isalso found in macroscopic ballistic tests (17, 18).ForD/have >> 1, Ep can be expressed by two terms,Ep ¼ ðrAshaveÞv2i =2þ Ed, where the first termrepresents the minimum inelastic energy trans-fer to targetmaterial withinAs and Ed representsall of the other energy dissipation mechanisms.The specific energy dissipation, E∗

p≡EpðrAshaveÞ−1,which is insensitive to material density by takingaccount of the mass within Ashave, is given byE∗p ¼ v2i =2þ E∗

d , where E∗d is the specific delo-

calized penetration energy. E∗d is thus a figure

of merit to evaluate the impact energy delocal-ization ability of a material as more samplemassbeyond As contributes to the energy dissipa-tion, whereas the material-independent energydissipation term, v2i =2, serves as a baseline.

Statistical values of Eph−1ave for MLG, PMMA,and polycrystalline gold were determined fromthe fitted slopes of Ep versus have (Fig. 4B andfigs. S6E and S7C). From this data (see squaredata points in Fig. 4C), MLG exhibits the highestE*p (or E∗

d), namely 1.26MJ kg−1 (or 0.86MJ kg−1)at 900 m s−1, compared with 0.58 MJ kg−1 (or0.19 MJ kg−1) for gold and 0.52 MJ kg−1 (or0.08 MJ kg−1) for PMMA. We also calculatedE∗p from previous macroscopic ballistic tests of

several materials: PMMA (19), aluminum (20, 21),steel (22, 23), and Kevlar KM2–polyvinyl butyral(PVB) composite fabric (24). These macroscopictests used a millimeter-scale spherical steel pro-jectile to penetrate a thin sheet (0.4 < h < 6 and1 <D/h < 20) without appreciable deformation ofthe projectile (table S1). The overall trend of E∗

p isquite similar, despite the large differences in themicroscopic a-LIPIT and traditional macroscopictestsandthehugerange in tensilemodulus, strength,and density among PMMA, aluminum, and steel.This implies that the principal energy dissipationmechanism for the three macroscopic materials isthe kinetic energy transfer from the projectile tothe target material within As (i.e., to the massrAsh), which results in a dishing process (25). Thisis why the E∗

p values of PMMA and gold froma-LIPIT also follow this trend and indicate a goodcorrespondence between the micro- and macro-scopic high–strain-rate evaluation methods forthesematerials, which also display localized pen-etration via a dishing process.However, the two highly anisotropic layered

materials—the Kevlar KM2-PVB composite fab-ric and the MLG membrane—deviate stronglyfrom this trend. The Kevlar KM2-PVB compositeis an armor-grade laminate made of strong, stiffpolyaramid fibers (strength ~ 4 GPa, Y ~ 84GPa,c ~ 7.6 km s−1) embedded in PVB resin (26). The

SCIENCE sciencemag.org 28 NOVEMBER 2014 • VOL 346 ISSUE 6213 1095

0 50 100

0

1

2

3

4

5

(at 900 m s-1)

0.030 have + 0.96

0.026 have

+ 1.18

Cha

nge

of k

inet

ic e

nerg

y, ∆

Ek (

nJ)

Thickness, have

(nm)

CV (%)

010203040

900 m s-1

600 m s-1

(at 600 m s-1)

0 50 100

0

10

20

900 m s-1

600 m s-1

Max

imum

cra

ck d

ista

nce,

Lm

ax (

µm)

Thickness, have

(nm)

0.10 have + D/2

0 500 1000

0.0

0.5

1.0

1.5

Spe

cific

pen

etra

tion

ener

gy,

E* p(M

J kg

-1)

Impact velocity, vi (m s-1)

ρ (g cm-3)

1.0

1.8

3.3

6.0

11.0

20.0

MacroscopicSteelAluminumPMMAKevlar armor

MicroscopicGoldPMMAMLG

v i2/2

Fig. 4. Analyses of microballistic results. (A) Maximum crack distances in MLG membranes of various thicknesses for the two different penetratorvelocities. (B) Kinetic energy changes of a m-bullet versus thickness after penetration of MLG membranes. Colors in (A) and (B) represent the thicknessinhomogeneity (via CV) of the film in the impact area. (C) Specific penetration energy of MLG, PMMA, and gold membranes compared with macroscopicmaterials at various impact velocities. The density of each material is represented by a logarithmic color scale. Error bars denote SD.

RESEARCH | REPORTS

planar fourfold symmetric high-stiffness wovenpolyaramid multilayer fabric shows an extensiveconelike deformation under ballistic impact, andthe higher E∗

p (or E∗d) values (see blue triangles

in Fig. 4C) can be understood by the contribu-tion of large sample mass well beyond the im-pact area to energy absorption (27). Similarly,the superior value of E∗

p for MLG (see greensquares in Fig. 4C) can be explained by its abil-ity to simultaneously be stiff, strong, and elas-tic, stretching into a cone shape due to the forceimparted by the forward moving projectile. Asa result, the E∗

d values of the MLG membrane(0.92 and 0.86 MJ kg−1 for 600 and 900 m s−1)substantially surpass those of steel (0.08 and0.11 MJ kg−1) at the same vi. Therefore, the MLGdemonstrates specific delocalized penetrationenergy 8 to 12 times higher than that of steeldue to the strong delocalization behavior at theseimpact speeds. As the higher delocalization effectresults in a wider penetration hole, this tendencycould be disadvantageous in certain aspects suchas multi-hit capability. However, this potentialweakness of MLG will be substantially relievedwhen the crack propagation is deflected by form-ing a composite.Our microscopic ballistic results reveal that the

superior in-plane speed of sound, high strength,stiffness, and structural anisotropy make MLG anextraordinary armor material exhibiting excellentimpact energy delocalization under a supersonicpenetration event. Because material far beyondthe strike face area can also consume kinetic en-ergy from the m-bullet while theMLGmembranesustains high dynamic tensile stress, the m-bulleteffectively experiences a higher areal density ma-terial. As large-scale production of graphene-basedcomposite materials is becoming possible (28),other graphene-like materials are being studied(29), and the results suggest opportunities for theuse of ordered anisotropic nanocomposites for sur-prising mechanical applications. The good corre-spondence between the micro- and macroscopicprojectile penetration tests, especially in themea-sured specific energy absorption, suggests thatthe microballistic method with its high-energyresolution may offer an effective means for theexploration of high–strain-rate physics of vari-ous materials, as well as practical advantages inrapid, high-throughput testing.

REFERENCES AND NOTES

1. C. Lee, X. Wei, J. W. Kysar, J. Hone, Science 321, 385–388 (2008).2. J. E. Field, S. M. Walley, W. G. Proud, H. T. Goldrein,

C. R. Siviour, Int. J. Impact Eng. 30, 725–775 (2004).3. J. W. Anthony, R. Bideaux, K. Bladh, M. C. Nichols,

Handbook of Mineralogy (Mineralogical Society of America,Chantilly, VA, 1990).

4. B. Gotsmann, H. Rothuizen, U. Duerig, Appl. Phys. Lett. 93,093116 (2008).

5. B. A. Gama, S. L. Lopatnikov, J. W. Gillespie, Appl. Mech. Rev.57, 223–250 (2004).

6. J. A. Zukas, High Velocity Impact Dynamics (Wiley-Interscience,New York, 1990).

7. J.-H. Lee et al., Nat. Commun. 3, 1164 (2012).8. S. G. Tomlin, J. Phys. D 1, 1667–1671 (1968).9. E. D. Palik, Handbook of Optical Constants of Solids (Academic

Press, San Diego, 1998), vol. 2.10. H. I. Rasool, C. Ophus, W. S. Klug, A. Zettl, J. K. Gimzewski,

Nat. Commun. 4, 2811 (2013).

11. G. H. Lee et al., Science 340, 1073–1076 (2013).12. M. S. Dresselhaus, G. Dresselhaus, J. Mater. Res. 6, 1114–1118

(1991).13. K. Kim et al., Nano Lett. 12, 293–297 (2012).14. S. L. Phoenix, P. K. Porwal, Int. J. Solids Struct. 40, 6723–6765

(2003).15. R. Grantab, V. B. Shenoy, R. S. Ruoff, Science 330, 946–948 (2010).16. J. Zhang, J. Zhao, J. Lu, ACS Nano 6, 2704–2711 (2012).17. M. A. Iqbal, N. K. Gupta, Lat. Am. J. Solids Struct. 5, 259–287 (2008).18. S. K. Garcia-Castillo, S. Sanchez-Saez, E. Barbero, J. Strain

Anal. Eng. Des. 47, 444–452 (2012).19. L. H. Abbud, A. R. A. Talib, F. Mustapha, H. Tawfique,

F. A. Najim, Int. J. Mech. Mater. Eng. 5, 123–128 (2010).20. W. Goldsmith, S. A. Finnegan, Int. J. Mech. Sci. 13, 843–866 (1971).21. N. K. Gupta, M. A. Iqbal, G. S. Sekhon, Int. J. Impact Eng. 32,

1921–1944 (2006).22. J. Dean, C. S. Dunleavy, P. M. Brown, T. W. Clyne, Int. J. Impact

Eng. 36, 1250–1258 (2009).23. C. J. Yungwirth, J. O’Connor, A. Zakraysek, V. S. Deshpande,

H. N. G. Wadley, J. Am. Ceram. Soc. 94, S62–s75 (2011).24. B. L. Lee et al., J. Compos. Mater. 35, 1605–1633 (2001).25. S. C. Wright, N. A. Fleck, W. J. Stronge, Int. J. Impact Eng. 13,

1–20 (1993).

26. M. Cheng, W. Chen, T. Weerasooriya, J. Eng. Mater. Technol.127, 197–203 (2005).

27. R. J. Rabb, E. P. Fahrenthold, Int. J. Numer. Methods Eng. 83,537–557 (2010).

28. D. Li, M. B. Müller, S. Gilje, R. B. Kaner, G. G. Wallace,Nat. Nanotechnol. 3, 101–105 (2008).

29. M. Xu, T. Liang, M. Shi, H. Chen, Chem. Rev. 113, 3766–3798(2013).

ACKNOWLEDGMENTS

This work was funded by the Defense Threat Reduction Agencyunder contract 1-12-10008 and the Welch Foundation grant C-1716.

SUPPLEMENTARY MATERIALS

www.sciencemag.org/content/346/6213/1092/suppl/DC1Materials and MethodsSupplementary TextFigs. S1 to S7Table S1Reference (30)

9 July 2014; accepted 23 October 201410.1126/science.1258544

REPELLENT SURFACES

Turning a surface superrepellenteven to completely wetting liquidsTingyi “Leo” Liu1 and Chang-Jin “CJ” Kim1,2*

Superhydrophobic and superoleophobic surfaces have so far been made by roughening ahydrophobic material. However, no surfaces were able to repel extremely-low-energyliquids such as fluorinated solvents, which completely wet even the most hydrophobicmaterial. We show how roughness alone, if made of a specific doubly reentrant structurethat enables very low liquid-solid contact fraction, can render the surface of any materialsuperrepellent. Starting from a completely wettable material (silica), we micro- andnanostructure its surface to make it superomniphobic and bounce off all available liquids,including perfluorohexane. The same superomniphobicity is further confirmed with identicalsurfaces of a metal and a polymer. Free of any hydrophobic coating, the superomniphobicsilica surface also withstands temperatures over 1000°C and resists biofouling.

The ability to understand the extraordinaryliquid repellency of natural surfaces (1, 2) hasaffected a wide range of scientific and tech-nological areas, from coatings (3), heat trans-fer (4), and drag reduction (5) to biomimetics

(6). Whereas the wetting-resistant surfaces devel-oped since the 1960s (7–10) used only surface rough-ness to trap gas with no interest in the apparentcontact angles, superhydrophobic surfaces createdsince the late 1990s (1, 11, 12) combined the rough-ness with a hydrophobic material to superrepelwater—that is, to display a very large apparent con-tact angle (q* > 150°) and a very small roll-off angle(qroll-off < 10°). For low-energy liquids such as oilsor organic solvents, a roughness with overhangingtopology was necessary to make the hydrophobicmaterial superoleophobic (13, 14), omniphobic(15), or superomniphobic (16, 17). Despite the use ofthe prefix “omni-” (6, 15–18), however, no natural

or man-made surface has been reported to repelliquids of extremely low surface tension or energy(i.e., g < 15 mJ/m2), such as fluorinated solvents,which completelywet existingmaterials (10, 19–21).Departing from the prevailing approach of rough-ening a hydrophobic material, we first proposethat the material’s inherentwettability, depicted bythe intrinsic contact angle qY, is irrelevant whendealing with a completely wetting liquid (qY = 0°).Focusing instead solely on the roughness details,we develop a surface that superrepels all availableliquids, including fluorinated solvents—for instance,perfluorohexane (C6F14; namely, 3M FluorinertFC-72), whose surface energy (g = 10 mJ/m2) isthe lowest known and which has never been ob-served to bead up, let alone roll off, on any surface.To avoid the confusion with the petal effect

(22), in which a droplet with large contact anglessticks to the surface, it helps to first clarify that“repelling”means that droplets not only bead onbut also roll off the surface. To repel (i.e., q* > 90°with a small qroll-off) or superrepel (q* > 150° withqroll-off < 10°) a wetting liquid (qY < 90°) on astructured surface, two conditions must be met:

1096 28 NOVEMBER 2014 • VOL 346 ISSUE 6213 sciencemag.org SCIENCE

1Department of Mechanical and Aerospace Engineering,University of California, Los Angeles (UCLA), Los Angeles, CA90095, USA. 2Department of Bioengineering, UCLA, LosAngeles, CA 90095, USA.*Corresponding author. E-mail: cjkim@ucla.edu

RESEARCH | REPORTS