REVIEW

Composites with carbon nanotubesand graphene: An outlookIan A. Kinloch1, Jonghwan Suhr2, Jun Lou3, Robert J. Young1, Pulickel M. Ajayan3*

Composite materials with carbon nanotube and graphene additives have long beenconsidered as exciting prospects among nanotechnology applications. However, afternearly two decades of work in the area, questions remain about the practical impact ofnanotube and graphene composites. This uncertainty stems from factors that include poorload transfer, interfacial engineering, dispersion, and viscosity-related issues that lead toprocessing challenges in such nanocomposites. Moreover, there has been little effortto identify selection rules for the use of nanotubes or graphene in composite matrices forspecific applications. This review is a critical look at the status of composites fordeveloping high-strength, low-density, high-conductivity materials with nanotubes orgraphene. An outlook of the different approaches that can lead to practically usefulnanotube and graphene composites is presented, pointing out the challenges andopportunities that exist in the field.

Carbon nanomaterials have had an unprec-edented impact over the past three dec-ades in defining the reach and applicationsof nanotechnology. Starting with the dis-covery of fullerenes and moving through

the carbon nanotube (CNT) era to graphene andother two-dimensional (2D) materials, the aca-demic world has been flush with new ideas, in-ventions, and innumerable attempts to find thekiller applications for these remarkable nano-structures. Here we will discuss one such ap-plication, which when first introduced seemedclose to embracing these materials but for var-ious reasons has not met expectations: that ofcomposite materials. In particular, the questionof whether structures such as carbon nanotubesand graphene, touted as ideal reinforcements incomposite matrices because of their mechanicalproperties, really are the right choices for me-chanical reinforcement still remains largely un-answered. Moreover, the realm of functionalitiesthat can be accessed by introducing nanotubesversus graphene in a matrix needs careful eval-uation and thought. Although both are sp2

allotropes, the structure, morphology, and di-mensionality of these two profoundly interestingcarbon nanostructures are indeed quite different,and so is the nature of their interactions withthe adjacent matrix (1). Hence, the overall com-posite mechanical behavior provided by thesetwo reinforcement units could be distinct. Itwould indeed be useful to have selection rulesthat choose one over the other in compositeapplications, but no such rational approacheshave emerged in the design of nanocompositeswith CNT or graphene phases.

We will consider three important questionsthat may broadly define the fate of this field.Why should these unique structures be of anyinterest in composites, why have they not pro-duced substantial progress after much effort, andwhat can be done to make them work as goodreinforcements in composites? The discussionwill go beyond just that of mechanical proper-ties and also consider their excellent electricaland thermal properties.Graphene represents the thinnest possible

atomically flat layer, made of a planar hexagonalhoneycomb lattice of strong C-C bonds that buildup graphite when stacked via weak van derWaalsinterlayer forces. Graphene has a Young’smodulusnear 1 TPa in the plane, reflecting the intrinsiccarbon bond stiffness, but being very thin, it canbe flexible in bending, twist, and other deforma-tion modes (2). The in-plane electrical and ther-mal conductivities are the highest among knownmaterials, but the through-thickness properties ofstacked graphene are very poor. CNTs are rolled-up graphenes with the in-plane properties trans-lated to axial properties, making them amongthe stiffest axial fibers ever created. Similar tographene, they can also be easily bent, twisted,and buckled (3). As graphene stacks up to makemultilayer structures, CNTs can also have a nestedstructure of tubes inside tubes [single-walled(SWNTs) to multiwalled nanotubes (MWNTs)],which has a notable effect in determining theirmechanical properties. In general, although thelocal stiffness that can bemeasured is extremelyhigh, owing to a near defect-free structure ofgraphene and nanotubes, a couple of majorissues make reinforcement of these structuresin matrices challenging. First, both grapheneand nanotubes are particulate fillers, with theirlarger dimensions (lateral size of graphene orlengths of nanotubes) reaching several hundredmicrometers at the most or, in exemplary cases,millimeters. Short fibers are generally poor loadcarriers in fiber composites, and this effect isclearly seen when composites are made with

CNT or graphene dispersions. Second, the surfacesof both graphene and nanotube are atomicallysmooth, devoid of any dangling bonds or defects(except at the edges in graphene or tips innanotubes), which means that strong matrix-filler bonds are hard to accomplish, leading topoor interfacial load transfer duringmechanicaldeformation (1) and high electron and phononscatter, compromising electrical and thermalproperties. This interfacial problemwas amajorroadblock in carbon fiber composites beforeindustry figured out the sizing of fibers viachemical modification, but for nanotubes andgraphene, this problem is severe, and attempts tochemically functionalize CNT or graphene surfacesmay substantially compromise their intrinsic prop-erties (4). This compromise also relates to a thirdissue, which is the inhomogeneous dispersion ofnanotubes and graphene in the matrix. Withoutproper surface treatments, CNTs or graphene tendto aggregate easily, owing to strong van derWaalsinteractions between them, to form poorly dis-persed bundles or agglomerates in the matrix,which often leads to poor interfacial connectiv-ity and formation of mechanical stress concen-tration or other functionally singular sites,resulting in severely affected composite perform-ances. Noncovalent functionalization methods(5) could be used to partially overcome the dis-persion challenge, but this approach is ineffectivein solving the interface problem. Therefore, asystematic and careful engineering approach isneeded to designCNTor graphene compositeswithoptimal performances (Fig. 1). We can show thepotential for using CNTor graphene composites forapplications by plotting the hypothetical positionof different CNT or graphene composites in amod-ified Ashby plot (Fig. 1B) and the position of fu-ture continuousCNTor graphene fiber composites.Finally, the ideal dispersion of material through-out the composite is very application dependent;typically for mechanical properties, one wantsas high a loading of the filler as possible that isaligned in the direction of the load, whereas forelectrical percolation, one aims for a randompercolated network with as low a concentrationas possible. For multifunctional applications, therequired microstructures may therefore even becontradictory.

Carbon nanotube and graphene fillers,interfaces, and load transfer

A great deal of effort has been made to de-velop lightweight, strong composite materialswith CNTs and graphene as reinforcement, andalthough these are considered to be discon-tinuous short fillers, they possess outstandingmechanical properties. The extremely high Young’smodulus of CNTs and graphene, their nanoscaledimensions, along with particular geometriesthat offer high specific surface area, present un-precedented opportunities to efficiently tailorthe interface properties between the reinforce-ments and composite matrices. CNT or graphenenanocomposites may not be as strong or stiff asa continuously reinforced composite such astypical carbon fiber laminates that are currently

COMPOSITE MATERIALS

Kinloch et al., Science 362, 547–553 (2018) 2 November 2018 1 of 7

1National Graphene Institute and School of Materials,University of Manchester, Manchester M13 9PL, UK. 2Schoolof Mechanical Engineering and Department of EnergyScience, Sungkyunkwan University, Suwon, 16419, SouthKorea. 3Department of Materials Science and NanoEngineering,Rice University, Houston, TX 77005, USA.*Corresponding author: Email: ajayan@rice.edu

on May 28, 2020

http://science.sciencem

ag.org/D

ownloaded from

used in primary load-carrying structural ap-plications. However, if the extraordinary po-tential and multifunctionality of CNTs andgraphene are fully realized and their nano-composites properly designed, they might stillbecome game-changing composite materials. Inaddition to the challenges of dispersion, viscosity

control, and sizing of the CNTs and graphenewithout compromising intrinsic properties, chal-lenges still exist in accurately characterizing thereinforcement’s mechanical properties. Theseproperties include Young’s modulus and strength,and also the interfacial shear strength at the filler-matrix interface, which will determine the critical

length of the fillers required for the most efficientload-transfer capability. Although these factorsare critically important in designing short-fibercomposite systems, unfortunately most atten-tion seems to be focused on improving practicalissues such as dispersion in matrix materialsand increasing the loading fractions of thesefillers without suffering viscosity-related pro-cessing issues.A substantial challenge has been the proper

characterization of fillers and filler-matrix in-terfaces during loading and assessing the effi-ciency of load transfer in composites. Ramanspectroscopy has been the most powerful tech-nique for both the characterization of carbon-based nanomaterials and the assessment of theirmechanical properties (6). Raman spectroscopycan also be used to evaluate their mechanicalproperties by observing the shift of the Ramanbands when the materials are subjected to de-formation (see Box 1 and Fig. 2).In polymer composites, we are normally con-

cerned with the matrix-reinforcement interface.However, in layered reinforcements, it is alsonecessary to consider the van der Waals bond-ing between the walls of carbon nanotubes andthe individual layers in multilayer graphene.This bonding is relatively weak compared withthe strong covalent bonding within the graphenelayers. When MWNTs and multilayer grapheneare used in composites, their ability to reinforceis therefore limited by easy shear between thewalls or layers, respectively. It is possible to trackinternal stress transfer between the walls of car-bon nanotubes and between the layers of graphenefrom stress-induced Raman band shifts. Im-perfect stress transfer is manifested as Ramanband broadening during deformation and a lowerRaman band shift rate compared with the single-walled or monolayer material. Comparing Ramanband shifts under stress for SWNTs and MWNTsin epoxy nanocomposites shows that interwallstress transfer efficiency is only around 70% forMWNTs (7). Lower Raman band shift rate inmultilayer graphene obtained with increasingnumber of graphene layers shows a similar trendto that in MWNTs. Raman spectroscopy has beeneffective in evaluating the efficiency of stresstransfer at the interface before and after chem-ical modification of CNT or graphene surfaces.Several chemical modification schemes havebeen proposed (1, 8, 9) (Fig. 3), which improvesthe interfacial strength between CNT or graphenefillers and the polymer matrices.

Properties of carbon nanotube andgraphene composites

The simplest way to look at the effect of nano-filler addition to a matrix is from the resultingmechanical properties of the nanocomposites,most commonly assessed through stress-straincurves (Fig. 4A). A telling example is naturalrubber, which is a relatively strong elastomerwith an extension to failure of ~1000%, wherethe addition of both the MWNTs and grapheneleads to a major change in the stress-straincurves and a large increase in the stiffness even

Kinloch et al., Science 362, 547–553 (2018) 2 November 2018 2 of 7

A

B

Youn

g’s

Mod

ulus

[GPa

]

Tensile strength [MPa]

CNT/Graphene as a matrix modifier

Continuous CNT/Graphene preform (fabric)

Polymeric foams

Elastomers

Polymers

Graphene polymer composites

Continuous CNT fiber polymer composites

CNT polymer composites

GFRPs

CFRPs

Carbon fibers

CNT fibers (dry-spun)

Graphenes

CNTs

0.1

0.1

0.011

1

10

10

100

100

1,000

1,000

10,000

10,000

100,000

raphene as a matrix mri

T/

Fig. 1. (A) A schematic illustration of a futuristic CNT or graphene polymer composite thatconsists of continuous CNT fiber preform (fabric) in a polymer matrix and chemically modifiedCNT or graphene as matrix modifiers. The continuous CNT or graphene spun fibers can exceedthe mechanical properties of carbon fibers that are used for state-of-the-art structural reinforce-ments in composites. The best load-carrying capability ever reached yet could be achieved byengineering the continuous CNT or graphene fiber preforms in composite design. It can beenvisioned that the multifunctional CNT or graphene composites are realized by marrying the bestfeatures of the continuous CNT or graphene fibers and the modifiable polymer matrix withchemically functionalized CNT or graphene particulates. (B) Ashby plot of Young’s modulus plottedagainst tensile strength comparing the mechanical properties of conventional polymer composites,including glass fiber–reinforced plastic (GFRP) and carbon fiber–reinforced plastic (CFRP), withCNT or graphene–based polymer composites. Thick outlines represent families of each material. Inthis projected Ashby plot, both mechanical properties of CNT or graphene polymer composites,which are considered to be particulate composites, are scattered around in between those of GFRPsand polymers. Those properties are shown to be below the ones of CFRPs that are continuouslyreinforced composites conventionally used for load-carrying structures. However, the mechanicalproperties of CNTspun fibers can indicate that CNTor graphene fibers can replace carbon fiber and,if properly designed and manufactured, the continuously reinforced CNT or graphene spun fibercomposites might be used in making ultralight yet superstrong structures in the near future. TheAshby plot was made by taking many reported values from the literature and also our estimatedscenario (e.g., for the CNT fiber composites).IL

LUSTRATIO

N:V

.ALT

OUNIAN/S

CIENCE

on May 28, 2020

http://science.sciencem

ag.org/D

ownloaded from

at relatively low loading of the fillers (10, 11). Inboth cases, however, the strength of the rubberis not increased, and the strain to failure is evenreduced. The effectiveness of the reinforcementscan be best assessed by plotting Young’s mod-ulus of the nanocomposite, Ec (as the stress at100% strain for rubbers with nonlinear stress-strain curves), against the filler loading, and anapproximately linear increase in Ec with fillervolume fraction is seen in both cases. The effectof adding carbon black (commonly used nano-filler) shows a much smaller increase in the mod-ulus of rubber (about a factor of 3 lower than forgraphene).The mechanics of deformation of fiber-reinforced

composites are now well established, but this isnot the case for nanocomposites. An obviousway forward is to adapt the continuum me-chanics theories developed for macroscopiccomposites, such as those reinforced by carbonor glass fibers, to composites at the nanoscale.Some researchers have taken an opposing view-point and have suggested that polymer nano-composites are actually quasi-homogeneousmolecular blends that should be regarded asmolecular composites or self-reinforced compo-

sites (12, 13) and that the properties are controlledby interactions on the molecular scale betweenthe nanoparticles and the polymer matrix.One issue that arises is the difficulty of real-

izing the promised high stiffness of the nanotubesor graphene in polymer-based nanocomposites.The effective stiffness of the fillers can be assessedby using the rule of mixtures, and the experimen-tally derived stiffness of CNTor graphene fillers incomposites is almost always observed to be farbelow the anticipated ∼1 TPa modulus. In thecase of CNTs, realization of the high stiffnessonly occurs when the CNTs are extended andaligned in polymer fibers or tapes (14). Valuesapproaching ∼500 GPa have been found, andthis can be rationalized by using amodified ruleof mixtures that takes into account filler size(length of CNTs and diameter for grapheneflakes) and orientation (15):

Ec ¼ EeffhohlVp þ Emð1� VpÞ ð1Þ

where ho is the Krenchel orientation factor andhl is the length factor. The parameter Eeff is theeffective Young’s modulus of the filler that de-pends only on its structure. For example, it will

Kinloch et al., Science 362, 547–553 (2018) 2 November 2018 3 of 7

Box 1. Raman spectroscopy andCNTor graphene composites.

sp2 allotropes of carbon have electronicstructures that lead them to undergo verystrong resonance Raman scattering sothat well-defined-spectra can be obtainedeven from individual CNTs or single layersof graphene (16, 38).These structures showcharacteristic G and 2D bands in theirRaman spectra, and where there are inter-nal or edge defects, a D band may alsobe seen. Subtle differences between thespectra of SWNTs, DWNTs (double-walledcarbon nanotubes), and MWNTs and be-tween those of mono-, bi- and multilayergraphene enable the spectra to be used asfingerprints of the materials. The Ramanbands shift during deformation for bothCNTs and graphene (6). Such downshiftsduring tensile deformation are large andrelatively easy to measure and enable one tomonitor the deformation of CNTor graphenein nanocomposites inside a polymer matrix(6). The shift rates (in cm−1/% strain) arefound to scale with the Young’s modulus ofthe nanocarbons, which enables such bandshifts to be used as a universal stress sen-sor for composite mechanics (39).

2D b

and

posi

tion

(cm

–1)

Axial strain (%)0.0 0.2

SWNTs

MWNTs

SWNT/Epoxy composite - 2D band

0.4 0.6 0.82605

2610

2615Raman wavenumber (cm–1)

2660

2665

2670

2500 2550 2600 2650 2700 2750

Inte

nsity

1.5% strain 0% strain

Fig. 2. Dependence of the 2D Raman bandposition upon strain during the deformationof CNTs in an epoxy resin matrix composite(0.1 wt %). The higher slope for the SWNTsreflects their higher Young’s modulus.(Inset) The shift of the 2D band for the SWNTs.[Reprinted from (7) with permission fromSpringer Nature]

CNT/Graphene composites

π - π interactions

Electrostatic interactionsVan der Waals force

Chemical bondingi t ti

V d W l f

+

Ch i l b di

–+++

–––––––––––––––––––––––––––––––––––––– –

+ + +

+

– – –

Fig. 3. A schematic representation of four different chemical modification schemes forCNTor graphene interacting with polymer matrix. Shown are p-p interactions (i.e., noncovalentinteractions), chemical bonding, van der Waals force, and electrostatic interactions. (Inset)A scanning electron micrograph (SEM) image of the fracture surface of SWNT epoxy compositesshowing good dispersion and pullout of the nanotube bundles from the matrix (9). CNTor graphenesurfaces with high specific surface area are favorable for chemical interactions with other molecules,which can allow for several chemical modification schemes. The ends of CNTs and edges ofgraphene facilitate additional chemical interactions with the polymer matrix.The chemically modifiedCNT or graphene surfaces can promote more uniform dispersion, control loading fractions ofCNTor graphene without suffering high viscosity, and improve interfacial strength between the fillersand polymer matrix in the composites. Although the nanoscale dimensions present a formidablechange, it is vital to understand the interactions between the CNT or graphene fillers and matrixtoward optimum performance in composite design. In particular, the characterization andquantification of the interfacial shear strength will be essential.

COMPOSITE MATERIALS on M

ay 28, 2020

http://science.sciencemag.org/

Dow

nloaded from

be ~ 1 TPa for monolayer graphene but is lowerfor few-layer graphene and graphene oxide (16).A similar reduction in Eeff occurs when goingfrom SWNTs to MWNTs as the result of inter-wall sliding. The orientation factor ho is 1 foraligned CNTs and flakes but much smaller forrandomly oriented CNTs. Hence, the mechanicalproperties of CNTs are more sensitive to tubeorientation, and randomly oriented grapheneflakes offer better reinforcement than randomlyoriented CNTs. The length factor hl (0 ≤hl ≤ 1)reflects the efficiency of stress transfer fromthe matrix to the filler that is controlled by boththe geometry (e.g., aspect ratio) of the fillerand the strength of the filler-matrix interface. Itapproaches unity for long CNTs or large-diameterflakes with strong filler-matrix interfaces. Thehigh values of Eeff found for aligned CNTs infibers and tapes are therefore the result of long,extended CNTs with a high degree of alignment.The very high strengths reported for CNTs

and graphene are also not realized in nano-composites. The celebrated experiment (17) thatmeasured a strength of 130 GPa for graphene(“200 times stronger than steel”!) has been

confirmed theoretically, but the experimentinvolved the nano-indentation of a monolayergraphene membrane and the deformation ofless than a few thousand carbon atoms. Single-nanotube deformation experiments involve sim-ilar numbers of atoms. The strength of allmaterials falls as their size increases owing tothe presence of defects, which is one reasonwhyhigh strengths are not found in macroscopicnanocomposite specimens. Backing out the ef-fective strength of the filler from the fracturestrength of the nanocomposite is possible, usinga rule ofmixtures equation for fracture analogousto that for stiffness. Modest values of strength aregenerally found in randomly aligned nanocompo-sites, but values as large as 88 GPa have been re-ported for SWNTs in highly aligned polymer tapes(18). Strength values of ~10 GPa extracted fromfracture data for graphene nanocomposites (19),although being well below the quoted 130 GPagraphene strength value, compare favorably withthat of carbon fibers. Dispersion of the nano-fillers also plays a major role in their effectivestrengths in nanocomposites, with high valuesonly found for uniform dispersions.

In addition to being processed into extendedstructures for potential high-strength and high-stiffness composites, CNTs and graphene maybest be applied in their particulate state as dis-persed matrix modifiers or fiber modifiers incomposites. The CNTs or graphene dispersed poly-mer composites exhibit time-dependent viscoelasticbehavior under mechanical loadings, owing tothe nature of polymers experiencing molecularrearrangements to release induced local stressor strain as a time-dependent behavior. The pres-ence of CNTs or graphene can affect the polymerchains’mobility particularly in the vicinity of thereinforcements in composites. The viscoelastic be-havior of nanotube and graphene composites hasbeen reported, leading to strong matrix-modifyingeffects of these fillers. With only 0.1 weight %(wt %) of graphene platelets (GPLs), epoxy-GPLcomposites creep substantially less than SWNT-epoxy and MWNT-epoxy composites at thesame loading fraction, mainly because of stron-ger interaction between graphene and epoxy.Functionalization of graphene (e.g., oxidation)can further promote better interaction with poly-mermatrix (20), leading to ahigher glass transition

Kinloch et al., Science 362, 547–553 (2018) 2 November 2018 4 of 7

Stre

ss (σ

)

Strain (ε) Interfacial volume fraction (%)

TS

A B

TP

ES

Strength

Elongation at break

Youn

g’s

mod

ulus Lo

ss m

odul

us

Electrical conductivity

Percolation threshold

Saturation

Linear increase

Fig. 4. Schematic illustrations of representative mechanical and elec-trical properties of CNTor graphene nanocomposites. (A) A schematicrepresentation of tensile stress-strain behaviors of three different poly-mers as CNT or graphene loading fraction increases: thermoset polymer(TS, dark blue region), thermoplastic polymer (TP, blue region), andelastomer (ES, light blue region). Schematic of representative tensilestress-strain curves are compared for pristine thermoset, thermoplasticpolymers and elastomer, and its nanocomposite with CNT or graphenereinforcements, respectively. For the pristine polymers, the stress-straincurves are depicted with yellow lines, while for the corresponding CNT orgraphene nanocomposites, the curves are illustrated with black lines.As the CNT or graphene loading increase (depicted as horizontal whitearrows), the stress-strain curve of each pristine polymer shifts fromthe yellow line to the black line.This shift indicates that with the increase ofCNT or graphene loading fraction, Young’s modulus of the polymer canincrease proportionally; by contrast, the elongation at break can besubstantially decreased. A slight decrease in tensile strength is also shown.Both elastomer and thermoplastic polymer exhibit a similar behavior ofductile-to-brittle transition at high loading fractions of CNTs or graphene,whereas thermoset nanocomposites become increasingly brittle over the

pristine polymer.The shift of the stress-strain curves with CNTor graphenereinforcements clearly shows that there is a compromise between Young’smodulus and toughness, which can be quantified by characterizing thearea of the stress-strain curve. (B) Loss modulus and electrical conduc-tivity with respect to interfacial volume fraction in CNT or graphenenanocomposites. With increasing interfacial volume and strength betweenCNT or graphene fillers and polymer matrix, the loss modulus, whichrepresents viscoelastic damping properties, can linearly increase as a resultof filler-matrix slippage until the dispersion quality can be maintained.However, at higher loading fraction of CNTs or graphene, the mobility ofpolymer chains gradually decreases, eventually leading to no furtherincrease in loss modulus. As for electrical conductivity, as the interfacialvolume fraction of CNTs or graphene increases, the conductivity of thenanocomposites changes from insulating to percolative conductiveregimes, displaying an S-shape curve. Once the random conductivenetwork is formed, the conductivity of the composites rises sharply,which is characterized as the percolation threshold. Beyond that, thecomposites become conductive. Typical low percolation thresholdsof CNTs or graphene in polymer composites are highly advantageous formany electrostatic and electromagnetic applications.

on May 28, 2020

http://science.sciencem

ag.org/D

ownloaded from

temperature. As a consequence of the strongerinteraction at the interface, relaxation processand storagemoduli of the functionalized grapheneepoxy composites decay more slowly over thepristine graphene composites. It hasbeenobservedthatwithout compromisingmodulus and strength,an order-of-magnitude enhancement in dampingproperties can be achieved in CNT-polycarbonatenanocomposites, mainly resulting from the inter-facial slippage between CNTs and the polymermatrix (20). In general for CNTor graphene polymercomposites, the intertube, interlayer interactionand sliding, and the sliding betweenpolymer chains,are all contributing factors to the strong visco-elastic behavior observed in composites (Fig. 4B).The large interfacial volume available can be

used to design interface-directed dynamic com-posites. A few recent studies have shown highlyintriguing morphology evolution of rubbery poly-mer of polydimethylsiloxane (PDMS) to becomemore ordered around CNTs under repeated load-ing, resulting in improved bulk composite stiff-ness for locally crystallized polymer composites(21). Discernable stiffening was observed for thecomposite during cyclic compressive stressing, aphenomenon not observed for the neat polymer.Such self-stiffening behavior was also observedin graphene-based PDMS composites. Because ofits relatively higher chain mobility compared tochemical cross-links, the physically cross-linkednetwork allows polymer chains to undergo re-

alignment and reorientation along graphene sur-faces under cyclic compression loading, resultingin increased interface stiffness (22). The engi-neering of unique interface-related propertiesshould allow us to build materials that arestimuli-responsive, dynamically reconfigurable,adaptable, and self-strengthening under loading.CNT or graphene particles seem to be ideal fillersfor creating such dynamic interfaces in polymers,which can be further tuned through chemicalmodification of the filler surfaces. In this re-gard, quantitative interface evaluation betweenCNT or graphene and matrices enabled by nano-mechanical measurement (see Box 2 andFig. 5) is very important for more effective in-terface engineering for improving overall com-posite performances.In addition to their mechanical properties,

nanotubes and graphene have several other func-tionalities that make them attractive for use incomposites. Important examples are their thermalstability and high electrical and thermal con-ductivities. Nanotubes are used in dispersionsin thermoplastic polymers to improve their elec-trostatic discharge or electromagnetic interfer-ence shielding properties (23). Graphene exhibitsan electrical conductivity comparable to that ofCNTs (105 to 106 S cm−1). However, the per-colation threshold for electrical conductivityof graphene in a polymer composite is muchhigher than for CNTs in composites. The per-

colation threshold of CNTs in high-density poly-ethylene matrix is 0.15 wt %, compared to 1 wt %for graphene in the same polymer (24). Thisis because 1D CNTs form a better conductivenetwork than 2D graphene in composite ma-terials (Fig. 4B). In contrast to the electricalconductivity of CNT or graphene composites, theimprovement in thermal conductivity appearsto be marginal for both. Even though individualCNTs or graphene have thermal conductivitiesas high as ~3000 W m−1 K−1, such particulatecomposites have a more efficient network, in-cluding intertube or interlayer scattering andheat transfer resistance at the filler-matrix inter-face (25). Attempts can still be made to enhancethe thermal conductivity of the composites bysubstantially increasing the weight fraction ofgraphene or CNTs in a matrix. For the sameloading fraction, graphene composites exhibitgreater interaction at the interfaces betweenfiller and matrix, which requires a higher acti-vation energy for the glass transition, resultingin better thermal stability of polymer composites.Compared to the vast number of studies on

polymer matrix composites reinforced by CNTsand graphene, ceramic or metal-based compo-sites have received much less attention, possiblybecause of the fabrication challenges. In addition,mechanical reinforcing effects are more obviousin polymer-based composites. The primary mo-tivation for the development of CNTs or graphene-based ceramic composites is enhancing toughnessor resistance to crack growth because ceramicsare already stiff and strong. For graphene-reinforced ceramic composites, the primarytoughening mechanism is the increased energydissipation due to pullout of graphene nano-sheets (26). Other toughening mechanismsobserved include crack deflection at the matrix-reinforcement interface and crack bridging. Sim-ilar toughening mechanisms were also observedin CNT-reinforced ceramic composites. Addition-ally, improved thermal and electrical conduct-ance are often attractive properties of ceramiccomposites. On the other hand, light metals suchas Al, Mg, and Cu are commonly used as matri-ces for CNTs or graphene-reinforced metal com-posites for mechanical property enhancement(27). It is worth noting that the interfacial strengthbetween CNT or graphene fillers and metalmatrices is substantially affected by the natureof interactions that include chemisorption (e.g.,Ni or Ru) and physisorption (e.g., Pt or Ir) (28),and even reactions leading to carbide formation(e.g., Al) (27). One important lesson in both ceramicand metal composites is that the processing-induced changes in the matrix, such as grain re-finement, are as important in determining thefinal properties of the composites as the amountof CNTs and graphene reinforcements present.Other functional properties, such as anticorrosion,antioxidation, piezoelectric properties, and bio-compatibility, are being explored in both ceramicandmetal composites with CNTs or graphene. Thelarge reduction in weight, in addition to propertyenhancements, is a key reason for adding CNT orgraphenedispersants inmetal and ceramicmatrices.

Kinloch et al., Science 362, 547–553 (2018) 2 November 2018 5 of 7

Fig. 5. Nanomechanical charac-terization of CNT-epoxy inter-faces. Shear-lag fits give a value forinterfacial shear stress (IFSS) of~30 ± 7 MPa for pristine carbonnanotubes and epoxy matrix (25).(Inset) SEM observation of a singlenanotube pullout from epoxyusing an AFM probe (43). It wasobserved that average IFSS in-creases as the embedded lengthdecreases, which is consistent withthe shear-lag theory and demon-strates the development of interfacialshear stresses at the ends of thesefibers. [Reprinted from (43) withpermission from The Royal Society]

20µm

80

70

60

50

40

30

20

10

00 500 1000 1500 2000 2500

Embedded length, L (nm)

IFSS

, τav

erag

e (M

Pa)

Box. 2. Interfacial strength and nanomechanics in CNTor graphene composites.

Single-fiber pullout tests have been used since the early development of composite materialstechnology to measure the shear strength of ductile interfaces capable of developing aconstant shear stress prior to fiber pullout, and the fracture energy of brittle interfaces thatexperience crack initiation and propagation prior to fiber pullout. Accordingly, attempts havebeen made to perform similar experiments on CNTor graphene composites. A successful attemptwas made (40) to attach a CNT to an atomic force microscopy (AFM) probe that was thendipped into a liquid epoxy, and a pullout test was carried out after the epoxy was cured. Morerecently, microfabricated devices have been developed to perform the pullout experiments withbetter control (41).The average interfacial shear strength reported by these tests corresponds tothe range of ~6 to 200 MPa for CNT-epoxy interfaces (42), although such quantitative nano-mechanical measurements remain a big challenge for graphene/polymer interfaces.

COMPOSITE MATERIALS on M

ay 28, 2020

http://science.sciencemag.org/

Dow

nloaded from

As discussed above, the Achilles’ heel forstructural nanocomposites is the difficulty inprocessing them at meaningful filler loadings.This issue arises from the high interfacial volumebetween the matrix and the filler, leading to alow rheological percolation threshold. Conse-quently, the matrix viscosity is found to increasesubstantially at very low concentrations of CNTsor graphene content, albeit with thixotropic be-havior, meaning that the viscosity will drop uponshear (29). Furthermore, the dimensionality of theparticle has an important role. Indeed, the 1Dnature of nanotubes, which makes them idealfor electrical percolation, means that nanotubesbecome nearly unprocessable in a conventionalpolymeric dispersion at loadings of a few volume%, whereas the 2D nature of graphene meansthat the platelets can slide over each other inshear. This allows systems with ~20 wt % GNPsto be processed, giving graphene a competitiveadvantage over CNTs for higher filler load appli-cations. One successful route for addressing thisviscosity increase at moderate concentrations hasbeen to produce liquid crystalline phases, typi-cally nematic, that require near-monodispersedparticle size distribution and an excellent degreeof dispersion (30).A key approach to achieving high loading

fractions is to preform the nanoparticles into

an organized architecture (e.g., fibers) beforeintroducing the matrix so that traditional com-posite production techniques can be subsequentlyused. In a direct analogy to polyaramid coagu-lation spinning, graphene, graphene oxide, andnanotubes have been spun from the liquid crys-talline phases, by using either superacid, or-ganic solvent, or aqueous/surfactant dispersions(31). Alternatively, polymer composite fibers canbe produced where the polymer [e.g., poly(vinylalcohol) or polyacrylonitrile] is used as a binderto improve the interparticle stress transfer, withthe binder optionally being pyrolyzed (32). Fi-nally, specifically for CNTs, fibers can be drawnfrom their growth substrate or zone (33), whichbypasses issues with the liquid-phase rheology,allowing extremely long and continuous CNTs tobe used. The 1D nature of CNTs means that theyare more appropriate for fibers than graphene;in particular, very–high–aspect ratio CNTs canbe used in the fibers, leading to high intertubestress transfer. To date, the mechanical proper-ties of CNT fibers have been considerably betterthan for graphene fibers. Furthermore, grapheneand nanotube fibers could have exceptional multi-functionality, including in applications such asenergy harvesting, energy storage, and sens-ing, that will be essential for future demands ofsmart fabrics and structures.

Another route to using fiber layup productiontechniques is hierarchical composites in whichCNTs or graphene are grafted onto macroscalecarbon or glass fiber by either direct growth ordeposition (34). These “hairy fibers” show in-creased out-of-plane improvement in interlam-inar shear stress and electrical and thermalconductivity. Alternatively, hybrid systems canbe produced where two or more from graphene,nanotubes, and fibers are used within a systemthrough mixing, interacting, and avoiding someof the viscosity and aggregation issues discussedabove. For example, nanotubes could combinewith graphene to give the low electrical percolationof the former with the high electrical conduc-tivity of the latter. Similarly, additions of grapheneor nanotubes to fiber-polymer composites canchange the glass transition temperature, and givefire retardancy properties, barrier properties, struc-tural damage sensing, and increased thermal andelectrical conductivity. A recent study (35) demon-strated that the addition of 2 wt % graphene tothe matrix of a carbon fiber epoxy composite canlead to an increase in axial stiffness of ~10 GPa(∼9%) as the result of the alignment and con-straint of the graphene flakes between the fi-bers. This technology has been reported (36) asbeing used in the case of the ultralightweight“graphene watch” produced by Richard Mille,

Kinloch et al., Science 362, 547–553 (2018) 2 November 2018 6 of 7

Fig. 6. A schematic summarizing variousCNTand graphene additions to compo-sites with anticipated enhanced proper-ties for the prospective applicationsalong with six key parameters in CNTand graphene composites.There has beensubstantial progress in synthesizing contin-uous fiber spinning with CNTand graphene,and also various yet controllable CNT andgraphene structures. The implementation ofthese unique structures of CNTand grapheneas reinforcements in composite materialscould provide the opportunity to translatethe exceptional properties of the individualCNTand graphene into a variety of engineer-ing applications. It can be envisioned that, ifcarefully designed and optimized, CNTandgraphene composites could not only replacecontinuously reinforced carbon fibercomposites as state-of-the-art structuralcomposites but also provide new func-tionalities such as stimuli-responsiveness,energy storage and conversion elements,real-time structural health monitoring,sensing, and vibrational damping in com-posites. A range of perspective applicationsincluding structural applications, multi-functional composites, electrical andthermal management, energy-absorbingcomposites, and self-stiffening compositeswith the required properties are summa-rized along with six key parametersincluding strength/stiffness, lightweight,interface, energy transfer, conductivity,and dispersion in the compositedesign and manufacturing.

ILLU

STRATIO

N:V

.ALT

OUNIAN/S

CIENCE

on May 28, 2020

http://science.sciencem

ag.org/D

ownloaded from

the Swiss luxury watchmaker, in conjunctionwith McLaren Formula 1 motor racing.

Future of CNTor graphene composites

After more than two decades of efforts to createnanotube and graphene composites in variousconcentrations and with various matrix combi-nations, the question still stands: How can thespectacular properties of CNT or graphene struc-tures be accessed in their composite structures?The answers have come mainly from trial-and-error–based attempts to control dispersion, fillerfraction, and, to a limited extent, interface con-trol. The results thus far do not point to com-petitive advantages of CNT or graphene particulateadditives to composites as real structural rein-forcements leading to high strength or increasedmodulus when compared to carbon fibers. Suchan outcome would require continuous fibers andbetter interfacial engineering, which could stillbe possible with spun fibers from nanotubesand chemically modified extended graphene struc-tures. Light weight of CNT or graphene fiberscould be the best advantage provided by thesestructures in composites. In addition, in largemarkets such as the automotive industry, wherethere is a need for cheap reinforcements thatbridge the properties of glass and carbon fibersand to lower production costs, these nanocarbonstructures may provide a solution. For all com-posites, including high-performance ones, thematrix- or fiber-modifying capabilities of thesefillers might prove an important advantage, throughincreasing thermal stability, electrical and ther-mal conductivities, viscoelastic properties, tough-ness, and creation of dynamic interfaces towardbuilding multifunctional composites. Indeed,graphene or nanotubes are already used com-mercially on a large scale as a substitute for tra-ditional conductive additives (e.g., carbon black)for applications such as static dissipative poly-mers. Figure 6 shows a summary of the impor-tant parameters that might determine CNT orgraphene additions to composite materials andpossible properties that could be enhanced for arange of more advanced applications. The prog-ress in fiber spinning with CNTs and continuous

filaments, sheets, and 3D scaffolds with CNT orgraphene elements could see a resurgence of CNTor graphene composites for structural applications.If this possibility is realized, then CNT or graphenefiber structures could not only eventually replacetoday’s carbon fibers but also provide new func-tionalities such as energy storage and conversionelements (e.g., CNT or graphene fibers acting asreinforcements and electrodes) (37), real-timestructure monitoring, sensing, and vibrationaland acoustic damping in composites. These at-tributes would also be ideally suited for thincomposite structures (for which micrometer-sizecarbon fibers may be too thick) or for providingflexibility and resilience in superflexible struc-tures. Unexplored properties of CNT or graphenecomposites, such as radiation resistance, couldalso prove vital in determining whether theseunique carbon nanostructures prove serious con-tenders in composites. Finally, beyond the in-trinsic scientific and technological issues, thewidespread availability of these nanomaterialsin large volumes and at relatively low cost willdictate the ultimate success of nanotube andgraphene composites.

REFERENCES AND NOTES

1. A. K. Geim, Science 324, 1530–1534 (2009).2. J. Zang et al., Nat. Mater. 12, 321–325 (2013).3. O. Lourie, D. Cox, H. Wagner, Phys. Rev. Lett. 81, 1638–1641

(1998).4. T. Ramanathan et al., Nat. Nanotechnol. 3, 327–331 (2008).5. S. H. Song et al., Adv. Mater. 25, 732–737 (2013).6. D. G. Papageorgiou, I. A. Kinloch, R. J. Young, Prog. Mater. Sci.

90, 75–127 (2017).7. R. J. Young, L. Deng, T. Z. Wafy, I. A. Kinloch, J. Mater. Sci. 51,

344–352 (2016).8. M. Terrones et al., Adv. Mater. 23, 5302–5310 (2011).9. P. M. Ajayan, L. S. Schadler, C. Giannaris, A. Rubio, Adv. Mater.

12, 750–753 (2000).10. S. Bhattacharyya et al., Carbon 46, 1037–1045 (2008).11. M. Hernández, M. del Mar Bernal, R. Verdejo, T. A. Ezquerra,

M. A. López-Manchado, Compos. Sci. Technol. 73, 40–46(2012).

12. P. Calvert, Nature 357, 365–366 (1992).13. G. Marom, H. D. Wagner, J. Mater. Sci. 52, 8357–8361 (2017).14. F. M. Blighe et al., Adv. Funct. Mater. 21, 364–371 (2011).15. R. J. Young et al., Compos. Sci. Technol. 154, 110–116

(2018).16. L. Malard, M. Pimenta, G. Dresselhaus, M. Dresselhaus,

Phys. Rep. 473, 51–87 (2009).

17. C. Lee, X. Wei, J. W. Kysar, J. Hone, Science 321, 385–388(2008).

18. Z. Wang, P. Ciselli, T. Peijs, Nanotechnology 18, 455709 (2007).19. I. Vlassiouk et al., ACS Appl. Mater. Interfaces 7, 10702–10709

(2015).20. N. A. Koratkar et al., Appl. Phys. Lett. 87, 063102 (2005).21. B. J. Carey, P. K. Patra, L. Ci, G. G. Silva, P. M. Ajayan, ACS

Nano 5, 2715–2722 (2011).22. L. Cao et al., Small 12, 3723–3731 (2016).23. M. F. De Volder, S. H. Tawfick, R. H. Baughman, A. J. Hart,

Science 339, 535–539 (2013).24. J. Du et al., Carbon 49, 1094–1100 (2011).25. Y. Liu, S. Kumar, ACS Appl. Mater. Interfaces 6, 6069–6087

(2014).26. W. A. Curtin, B. W. Sheldon, Mater. Today 7, 44–49 (2004).27. A. Nieto, A. Bisht, D. Lahiri, C. Zhang, A. Agarwal, Int. Mater.

Rev. 62, 241–302 (2017).28. S. M. Kozlov, F. Vines, A. Gorling, J. Phys. Chem. C 166,

7360–7366 (2016).29. P.-H. Wang, S. Sarkar, P. Gulgunje, N. Verghese, S. Kumar,

Polymer (Guildf.) 150, 10–25 (2018).30. P. K. Rai et al., J. Am. Chem. Soc. 128, 591–595 (2006).31. N. Behabtu et al., Science 339, 182–186 (2013).32. B. Vigolo et al., Science 290, 1331–1334 (2000).33. Y.-L. Li, I. A. Kinloch, A. H. Windle, Science 304, 276–278 (2004).34. M. J. Arlen et al., Macromolecules 41, 8053–8062 (2008).35. J. W. Chu et al., Compos., Part A Appl. Sci. Manuf. 113, 311–317

(2018).36. B. Robinson, https://www.manchester.ac.uk/discover/news/

science-behind-worlds-lightest-graphene-watch-revealed/ (2018).37. X. Mao, G. C. Rutledge, T. A. Hatton, Nano Today 9, 405–432

(2014).38. M. S. Dresselhaus, G. Dresselhaus, R. Saito, A. Jorio, Phys. Rep.

409, 47–99 (2005).39. O. Frank et al., Nat. Commun. 2, 255 (2011).40. A. H. Barber, S. R. Cohen, A. Eitan, L. S. Schadler, H. D. Wagner,

Adv. Mater. 18, 83–87 (2006).41. Y. Ganesan et al., ACS Appl. Mater. Interfaces 3, 129–134 (2011).42. X. Chen et al., Carbon 82, 214–228 (2015).43. W. Wang, P. Ciselli, E. Kuznetsov, T. Peijs, A. H. Barber,

Philos. Trans. A Math. Phys. Eng. Sci. 366, 1613–1626 (2008).

ACKNOWLEDGMENTS

We thank S. K. Lee for her help in creating some of the figures.Funding: J.L. acknowledges support from the U.S. Department ofEnergy, Office of Basic Energy Sciences, under grant DE-SC0018193. P.M.A. and J.L. acknowledge the NSF NanosystemsEngineering Research Center for Nanotechnology-Enabled WaterTreatment (ERC-1449500) for support. J.S. acknowledges supportfrom the National Research Foundation of Korea (NRF) grantfunded by the Korea government (MSIP) (2018R1A2B2001565).I.A.K. and R.J.Y. acknowledge funding from the European Union’sHorizon 2020 research and innovation program under grantagreement no. 785219. Competing interests: The authors declareno competing interests.

10.1126/science.aat7439

Kinloch et al., Science 362, 547–553 (2018) 2 November 2018 7 of 7

COMPOSITE MATERIALS on M

ay 28, 2020

http://science.sciencemag.org/

Dow

nloaded from

Composites with carbon nanotubes and graphene: An outlookIan A. Kinloch, Jonghwan Suhr, Jun Lou, Robert J. Young and Pulickel M. Ajayan

DOI: 10.1126/science.aat7439 (6414), 547-553.362Science 

ARTICLE TOOLS http://science.sciencemag.org/content/362/6414/547

CONTENTRELATED

http://science.sciencemag.org/content/sci/362/6414/543.fullhttp://science.sciencemag.org/content/sci/362/6414/536.fullhttp://science.sciencemag.org/content/sci/362/6414/534.full

REFERENCES

http://science.sciencemag.org/content/362/6414/547#BIBLThis article cites 42 articles, 6 of which you can access for free

PERMISSIONS http://www.sciencemag.org/help/reprints-and-permissions

Terms of ServiceUse of this article is subject to the

is a registered trademark of AAAS.ScienceScience, 1200 New York Avenue NW, Washington, DC 20005. The title (print ISSN 0036-8075; online ISSN 1095-9203) is published by the American Association for the Advancement ofScience

Science. No claim to original U.S. Government WorksCopyright © 2018 The Authors, some rights reserved; exclusive licensee American Association for the Advancement of

on May 28, 2020

http://science.sciencem

ag.org/D

ownloaded from