Composites Science and Technology 100 (2014) 128–135

Contents lists available at ScienceDirect

Composites Science and Technology

journal homepage: www.elsevier .com/ locate /compsci tech

Multi-scale interlaminar fracture mechanisms in woven compositelaminates reinforced with aligned carbon nanotubes

http://dx.doi.org/10.1016/j.compscitech.2014.06.0030266-3538/� 2014 Elsevier Ltd. All rights reserved.

⇑ Corresponding author. Tel.: +1 617 252 1539.E-mail address: wardle@mit.edu (B.L. Wardle).

1 Current address: University of California, Santa Barbara, United States.

Sunny S. Wicks, Wennie Wang 1, Marcel R. Williams, Brian L. Wardle ⇑Massachusetts Institute of Technology, 77 Massachusetts Ave., Cambridge, MA 02139, United States

a r t i c l e i n f o a b s t r a c t

Article history:Received 16 March 2014Received in revised form 19 May 2014Accepted 5 June 2014Available online 13 June 2014

Keywords:A. Carbon nanotubesA. LaminateA. Hybrid compositesB. Fracture toughnessD. Fractography

Aligned nanoscale fibers have been shown to provide inter and intralaminar reinforcement of fiber-rein-forced polymer composites. In one hierarchical architecture, aligned carbon nanotubes (CNTs) grown onadvanced microfibers in a woven ply creates a ‘fuzzy fiber’ reinforced plastic (FFRP) laminate. Here, themechanisms of Mode I fracture toughness enhancement for such laminates are elucidated experimentallyby varying the type of epoxy and aligned CNT length. Reinforcement effectiveness is found to vary fromreduced initiation toughness to over 100% increase in steady-state fracture toughness, depending uponthe interlaminar fracture mechanisms. Fracture-surface morphology investigations reveal that interlam-inar toughness enhancement is significantly less for an aerospace infusion resin than that for a muchtougher hand lay-up marine epoxy. Long (�20 micron) aligned CNTs add significant toughness in steadystate (>1 kJ/m2 increase for marine epoxy) via CNT pullout and by driving the crack through tortuouspaths around and through microfiber bundles/tows, whereas shorter CNTs produce less toughening (oreven reduced toughness in aerospace epoxy), which is attributed to increased microfiber–matrix cohesivefailure. These findings reveal the multi-scale nature of the aligned-CNT reinforced composite ply inter-face, and the mechanisms at work at the micro and nanoscales that influence the macroscopic behavior.These new insights provide impetus for using aligned CNTs to tailor microfiber polymer composites byadjusting microfiber orientation, steering cracks through tortuous paths, and maximizing fracture surfacearea through both microfiber and nanofiber pullout.

� 2014 Elsevier Ltd. All rights reserved.

1. Introduction

Aerospace structural components are increasingly being madeof (micro)fiber reinforced plastics (FRPs) due to their excellentmass-specific mechanical properties. While properties in the planeof the microfibers are strong, stiff, and tailorable, the interlaminarregion is relatively weak and compliant, leading to failure due todelamination and other damage modes. To address these deficien-cies at the interlaminar interface, several architectures have beendeveloped to enhance strength and/or toughness in the through-thickness direction of the laminate, including Z-pinning, stitching,and 3-D weaving [1–3]. These processes result in a significantreduction in in-plane properties of the laminate due to damagefrom insertion of through-thickness direction reinforcements thatare hundreds of microns in diameter, and loss of microfiber volumefraction in the in-plane direction [1,4]. Alternative routes toincreasing interlaminar toughness include modification of matrix

properties through added toughening agents, compliant or rein-forced interlayers [5,6] or particles [7–9] and nanofibers [10,11],including in particular carbon nanotubes (CNTs) that possessadvantageous specific stiffness and strength [12]. Their excellentelectrical and thermal conductivity make them attractive as ananofiller for aerospace composites due to their ability to addmultifunctional properties and capabilities [13–16]. While CNTspossess significant potential for composite reinforcement, onlysmall improvements in composite properties have been realizeddue to difficulties in processing and controlling CNT orientationduring manufacturing.

Many approaches to CNT incorporation have been explored inindustry and in the literature, dominated by mixing CNTs in poly-mer matrices before infiltration [17–23] in low loading to mini-mize influence on matrix viscosity. CNTs tend to agglomerateinto bundles, and techniques to separate the CNTs by sonication[24], surface modification [25,26] and/or shear mixing [27] caneasily damage the CNTs. Alignment of the CNTs in the matrix isalso difficult to obtain even with severe processing techniques likeextrusion [28] that are not compatible with laminated composites,though some orientation has been controlled through directional

S.S. Wicks et al. / Composites Science and Technology 100 (2014) 128–135 129

infusion [29]. Grafting CNTs onto microfiber surfaces prior to poly-mer infiltration allows precise control over CNT organization anddistribution, though care must be taken to avoid new manufactur-ing challenges that accompany the grafting process by preventingdamage to the microscale fibers [30] and understanding changesto microfiber packing through decreased compressibility [31,32].CNT-grafted microfibers show potential for mechanical reinforce-ment and have been demonstrated to provide interlaminarreinforcement [33,34], increase interfacial shear strength[35–37], and locally reinforce the matrix [38]. These mainapproaches to reinforce composites with CNTs are reviewed inthe literature [39].

In this paper, aligned CNTs are implemented in a three-dimen-sionally reinforced nano-engineered composite. The micron-scalediameter fibers are alumina, serving as a model system for thisstudy, because the microfiber tensile strength and microfiber–polymer interfacial shear strength are preserved following theCNT growth process [40,41], unlike (tensile) strength lossesobserved by many researchers for the industrially more relevantcarbon fibers due to the carbon fiber surface sensitivity to hightemperatures and/or metal catalysts [30,35,36]. Other researchershave noted interfacial shear strength enhancement for nanostruc-ture-modified microfibers [42,43], usually with concomitantmicrofiber strength loss in the carbon fibers. The aligned carbonnanotubes in this work are grown in situ on the surfaces of microf-ibers, termed ‘fuzzy fibers’ as shown in Fig. 1a, and extend acrossinterlaminar and intralaminar regions to provide mechanicalreinforcement [44–46]. By controlling the CNT alignment and dis-persion in fuzzy fiber reinforced plastics (FFRP), facile manufactur-ing with polymer matrix systems is enabled that is compatiblewith existing manufacturing processes [31,47]. Preliminary workon the same alumina fiber system in a thick coarse weave haddemonstrated a 76% improvement in steady-state toughness for

Fig. 1. Fuzzy fiber reinforced plastics (FFRP) (a) Illustration of FFRP architecture. (b) FFRchopped-fiber plate bending arm reinforcement. (c) Fibers in long CNT fuzzy architecture(d) Fibers in short CNT fuzzy architecture before polymer infiltration with �6 lm CNTinterpretation of the references to colour in this figure legend, the reader is referred to

laminates selectively reinforced along the midplane [45], a geom-etry that can lead to jumping of the crack from the tough midplaneto adjacent ply interfaces that do not have CNT reinforcement. Thework herein remedies this situation with laminates reinforcedthroughout the thickness with aligned CNTs. Mode I double canti-lever beam (DCB) laminates were manufactured and tested with abrittle aerospace epoxy and a tough marine epoxy to understandepoxy interaction with both microfibers and CNTs as it relates totoughening behavior.

2. Materials and methods

FFRP laminates are manufactured by producing ‘fuzzy fiber’plies and then infiltrating with an epoxy matrix either by vacuumassisted resin infusion (VARI) or hand lay-up. Alumina cloth(Cotronics Ultra-Temp 391, Dutch weave tape, areal density of371 g/m2) is first dipped in a 50 mM solution of iron nitrate in iso-propanol, distributing catalyst to all alumina fiber surfaces evenwithin the cloth, and then dried in an oven for 8 h at 30 �C. Thecloth is cut into 200 � 700 plies onto which aligned CNTs are grownby chemical vapor deposition in a 200 tube furnace (Lindberg BlueM). The cloth is heated to a furnace setpoint of 650 �C under1040 sccm of hydrogen, then stabilized at temperature for 5 minfor the hydrogen pre-treatment step to create catalyst particlesof iron on the microfiber surface [45,46]. Ethylene is introducedat 350 sccm to grow aligned CNTs, the lengths of which are con-trolled by varying the duration of ethylene exposure. For example,the bulk of the CNTs in this set are made with approximately 3 minof ethylene exposure, yielding CNTs that are �19 lm long, whichcause the annular aligned CNT forests to split into ‘mohawks’ alongthe microfiber length (see Fig. 1c). The short CNT plies are madewith 1 min of ethylene exposure to yield 6 lm long annular A-CNTs around the microfiber as shown in Fig. 1d [46]. Tuning the

P DCB test specimen, with dimensions of 20 � 180 mm (width � length), with redbefore polymer infiltration with �20 lm long aligned CNTs on the surface of fibers.

fuzz on fiber surface. Insets show arrangement of aligned CNTs around fibers. (Forthe web version of this article.)

Fig. 2. Manufacturing routes for FFRP production. (a) Side schematic of infusion setup showing primary epoxy (Hexcel RTM6) flow path. (b) Hand lay-up for viscous epoxysystem (West Systems 105/206).

130 S.S. Wicks et al. / Composites Science and Technology 100 (2014) 128–135

A-CNT length also affects the microfiber geometrical arrangementwhile also reinforcing the matrix. A-CNTs on the microfiber sur-faces effectively push apart the microfibers in the tows, with thelong CNTs inducing the greatest degree of tow swelling that canalter the local tow geometry and, importantly for this work, howthe neighboring plies nest together at the interlaminar interfaces.The A-CNTs are not functionalized after growth, i.e., they are usedas-grown, and the A-CNTs do not affect the degree of cure of thematrix, as determined by comparing the energy to cure via Differ-ential Scanning Calorimetry (DSC) on epoxy and A-CNT reinforcedepoxy.

Laminates are manufactured 8 plies thick by infusion or handlay-up depending on polymer type. Both matrix systems are epoxy,though they have very different characteristics: the aerospaceepoxy is based on a tetrafunctional resin, tetraglycidylether of4,40-diaminodiphenyl methane (TGDDM), while the marine epoxyis based on a difunctional resin, bisphenol-A epoxy resin (BPA).The aerospace resin is a highly crosslinked polymer compared tothe marine epoxy due to the chemical structure, and as suggestedby the high Tg for the aerospace resin; the differing chemical struc-ture leads to different mechanical behavior, where the aerospaceepoxy has a tensile strength of 75 MPa with a strain to failure of3.4%, while the marine epoxy has a tensile strength of 50 MPaand a strain to failure of 4.5%. VARI is used to manufacture bothFFRP and baseline (without CNTs) laminates with the unmodifiedaerospace infusion resin (Hexcel RTM6, 33 cPs at processing tem-perature), while hand lay-up is used with the marine epoxy.Fig. 2a shows the VARI setup, where spiral tubes and distributionmesh sheets draw the epoxy along and into the laminate. The lam-inate is layered with porous Teflon peel ply and flow media on bothsurfaces, before being sandwiched between steel and glass platesheld apart by spacers to ensure a flat and uniform part with somecontrol over volume fraction. A 50.8 mm (200) wide steel plate that

Fig. 3. Mode I fracture toughness of double cantilever beam laminates. Strain energy reaerospace and marine epoxy. Initiation toughness based on visual onset of propagation,

matches the width of the laminate is placed on top of the stack tominimize ‘racetracking’ of the epoxy along the edges during infu-sion. This allows the vacuum bag to conform to the top plate andthe edges of the laminate. Prior to infusion, the plates and laminatestack are heated under vacuum to 150 �C (300 �F) for 2 h to dis-place any residual moisture. Then, the aerospace epoxy is degassedat 80 �C, and slowly drawn into the laminate at 90 �C at a vacuumof 29 in Hg (740 mmHg). The inlet and outlet of the system is thenclosed off and cured at 160 �C and post cured at 180 �C. Themicrofiber volume fractions of laminates made by infusion are allcomparable: 41–48% for baseline, 40–51% for the short CNT, and40–48% for the long CNT FFRP laminates, with resulting CNTweight fractions in the total laminate at 0.5% and 2% for the shortand long CNT systems, respectively. Successful testing was per-formed on 11, 3, and 14 specimens of baseline, short CNT, and longCNT laminates, respectively. Due to challenges in growing consis-tent short CNT plies, only an exploration into short CNT laminatesis performed. Hand lay-up is used with viscous epoxy systems toproduce both baseline and FFRP laminates as shown in Fig. 2b. Inhand lay-up, a 2-part marine epoxy (West Systems Resin 105and Hardener 206, 725cPs) is mixed and poured onto a flat surface.As shown in Fig. 2b, a ply is then laid down in the epoxy pool,allowing the epoxy to wick into the interior of the ply. More epoxyis poured on top, and the process is repeated until the entire lam-inate stack is wet. The stack is then compressed between twoplates wrapped in vacuum bagging film held apart by spacers,and left to cure for 24 h. The microfiber volume fractions of lami-nates made by hand lay-up were 35–43% for baseline, 44–50% forthe short CNT, and 33–41% for the long CNT FFRP laminates, withresulting CNT weight fractions in the total laminate at 0.5% and2% for the short and long CNT systems, respectively. The rangesof microfiber volume fractions are evidence of the difficulty to con-trol volume fraction even when using spacers due to the variations

lease rate, GIc, of baseline, short CNT, and long CNT curves of laminates made withand percentages are changes from baseline of each set.

Fig. 4. Mode I fracture of aerospace epoxy laminates. (a) Representative R-curves of baseline and FFRP tests. (b) Smooth epoxy cleavage in epoxy of baseline laminates insteady-state portion of R-curve. (c) Fiber–matrix adhesive failure and resin-rich regions in short CNT FFRP laminates in steady-state portion of R-curve. (d and e) Rough epoxyfracture along fiber tows in long CNT FFRP laminates in steady-state portion of R-curve. (f) CNTs pulled out of fracture surface in FFRP laminates.

Fig. 5. Example fractography of aerospace epoxy laminate with long A-CNTs. (a) Fiber–matrix interface adhesive failure in initiation region of crack propagation. (b) Evolutionof strain energy release rate along a long CNT aerospace laminate. (c) Rough matrix cohesive failure in peak toughness regions of FFRP.

S.S. Wicks et al. / Composites Science and Technology 100 (2014) 128–135 131

in cloth compressibility, and large scale implementation of FFRPlaminates should employ a closed mold manufacturing system thatwould allow microfiber volume fraction to be more tightly con-trolled. For both hand lay-up and infusion, aligned CNTs have beenpreviously demonstrated to remain attached to the microfiberswithin the FFRP laminate [47,48]. Fracture testing was performedon 10, 5, and 17 specimens of baseline, short CNT, and long CNTlaminates, respectively, and in all cases the extracted values forfracture show sufficiently low variability to allow comparison.

To fabricate double cantilever beam (DCB) laminates, a 25 lmfluoropolymer release film is placed between the middle two plieson the last 5 cm on one end of the laminate during lay-up, to form astarter precrack at the laminate centerline. The resulting laminatesare cut into 20 mm-wide specimens. To prevent excessive bendingduring DCB testing, 3.2 mm thick chopped fiberglass plate tabs (seeFig. 1b), cut to the same width of the laminate, are used to sand-wich the laminate [45]. The surfaces of the laminate and fiberglassare roughened with sandpaper and cleaned with acetone, isopro-panol, and methanol to prepare the surfaces for bonding. A30 min epoxy is used to glue the fiberglass plates and hinges fortest grips to the samples. The tabbed DCB specimens are testedin a Zwick Roell mechanical tester with a 10 kN load cell followingASTM D5528 [49]. The samples are loaded at 0.5 mm/min and load,stroke, and crack length are recorded. The strain energy releaserate, GIc, for each laminate type is calculated using the Compliance

Calibration Method, as outlined in ASTM D5528 [49]. Initiationtoughness, steady-state toughness, and the maximum observedtoughness are computed.

3. Results and discussion

Mode I fracture toughness is characterized by a resistance (R-curve) behavior that includes an initiation value as well as a regionof increasing toughness (resistance) as the crack extends until asteady-state value is reached. Exemplary R-curves are shown inFigs. 4a and 6a. The initiation toughness, GIc, initiation, was assessedby three different methods in following the ASTM standard: onsetof visual propagation (VIS GIc, initiation), deviation from linearity(NL GIc, initiation), and a 5% offset from linearity (NL GIc, initiation). Thesteady-state toughness was taken as the average of the entire R-curveexcluding at least the first 5 points (where toughness is increasing).Because the standard provides no method to calculate steady-statetoughness when the R-curves do not plateau, the comparison ofthe maximum GIc value (GIc,max) is also presented. Computed tough-ness values are summarized in Table 1, and the results for visual ini-tiation onset and steady-state are included in Fig. 3. The trendsobserved for each epoxy type will be discussed in detail next.

Representative GIc R-curves for the aerospace epoxy laminatesshowing the evolution of toughness with increased crack length

Fig. 6. Mode I fracture of marine epoxy laminates. (a) Representative R-curves of baseline and FFRP tests. (b) Individual microfibers pulled out of fracture surface in baselinelaminates. (c) microfiber and tow pullout across crack tip, where vertical lines demarcate millimeters of crack growth. (d and e) Broken tow surface revealing broken fibersand holes from microfiber fracture and pullout. (f) Rough CNT-filled fracture of epoxy between microfibers showing long CNT pullout.

Table 1Summary data for Mode I fracture tests.

5% GIc, initiation (kJ/m2) NL GIc, initiation (kJ/m2) VIS GIc, initiation (kJ/m2) GIc, max (kJ/m2) GIc, steady-state (kJ/m2)

Aerospace Epoxy Baseline 0.32 ± 0.08 0.21 ± 0.08 0.36 ± 0.08 0.47 ± 0.08 0.38 ± 0.07Short CNT 0.14 ± 0.02 {�56%} 0.095 ± 0.035 {�55%} 0.14 ± 0.01 {�61%} 0.23 ± 0.01 {�51%} 0.19 ± 0.003 {�50%}Long CNT 0.29 ± 0.17 {�9%} 0.14 ± 0.06 {�33%} 0.25 ± 0.10 {�29%) 0.66 ± 0.15 {+40%} 0.50 ± 0.11 {+32%}Marine Epoxy Baseline 0.57 ± 0.26 0.27 ± 0.09 0.58 ± 0.21 1.33 ± 0.12 1.19 ± 0.09Short CNT 0.74 ± 0.33 {+31%} 0.52 ± 0.15 {+90%} 0.71 ± 0.25 {+22%} 2.65 ± 0.33 {+99%} 2.25 ± 0.19 {+89%}Long CNT 0.77 ± 0.25 {+37%} 0.31 ± 0.06 {+15%} 0.83 ± 0.32 {+43%} 2.67 ± 0.42 {+101%} 2.38 ± 0.32 {+100%}

132 S.S. Wicks et al. / Composites Science and Technology 100 (2014) 128–135

are summarized in Fig. 4a, and differences in the shape of theR-curves can be seen, notably between the baseline and long CNTFFRP. The baseline curves exhibit minimal toughening after crackinitiation while the long CNT FFRP curve exhibits a lower initiationtoughness followed by clear toughening behavior. A moderatereduction (12–36% across all three measures) in initiation tough-ness is observed for these long CNT FFRP laminates, whereas thereis a moderate, but larger, increase in both steady-state and maxi-mum toughness of 32.9% and 39.0%, respectively. Laminates madewith short CNTs in the same aerospace epoxy lead to significantlyreduced toughness, with all three measures of initiation as well assteady-state toughness dropping by over 50%.

SEM fractography of the fractured laminates shows that theCNTs in the FFRP induce microscale changes to the fracture surface.In aerospace epoxy baseline laminates, as shown in Fig. 4b, fracturesurfaces are dominated by smooth epoxy regions with someexposed microfibers. The drop in toughness in the short CNT FFRPsis counterintuitive, but can be understood by considering the frac-ture surface in Fig. 4c where exposed microfiber surfaces are adja-cent to resin-rich regions with no CNTs. In contrast, the fracturesurfaces of long FFRP laminates, shown in Fig. 4d and e, show avery different morphology comprised of a rougher fracture surfaceas the crack follows the tows, exposing some microfiber surfaceswith a mixture of fiber–epoxy interface adhesive failure and matrixcohesive failure. Exposed in the rough surfaces of the matrix cohe-sive failure are CNTs that were pulled out, up to a few microns,during crack propagation as shown in the SEM image in Fig. 4f, sug-gesting CNT pullout as a mechanism for toughening the matrix[50]. Thus, the drop in toughness for the short CNTs is primarily

attributed to the short CNTs leaving resin rich regions betweenplies combined with the crack path being driven to the fiber–matrix interface. The short CNTs are not long enough to effectivelybridge the interlaminar region. Comparatively, longer A-CNTs fullybridge the interlaminar region and contribute toughness from CNTpullout and increased cloth nesting. Fracture morphology evolu-tion within a given laminate, as shown in Fig. 5, provides furthersupport for these observations. The fracture surfaces correspond-ing to the toughest regions of the laminate are dominated bymatrix cohesive failure, creating rough, jagged surfaces as seen inFig. 5c. Conversely, bare exposed microfibers are observed in theinitiation fracture surface in Fig. 5a where toughness is lowest.The causes for evolution of fracture morphology that governwhether the fracture remains in the matrix or follows the microfi-ber surface might include nesting prevention of the woven cloth atinitiation by the release film precrack, or variation in the CNTmorphology along the laminate, among others.

Representative R-curves of Mode I fracture of marine epoxy lam-inates are shown in Fig. 6a. Unlike the R-curves in the aerospaceresin laminates above, both the baseline and FFRP laminates forma typical R-curve, i.e., classic toughening behavior. Changes in initi-ation, steady-state, and maximum fracture toughnesses are summa-rized in Table 1, where large gains in all measures are seen for bothshort and long CNTs. The baseline laminates consistently fracture,reaching steady-state within 10 mm of crack advance. The initiationtoughness for long CNT laminates increase above the baseline 15–43% across all three initiation measures, while the steady-stateand max toughnesses increase by 100% and 101%, respectively.The long CNT FFRP laminates behave in two ways, either toughening

Fig. 7. Example fractography of a marine epoxy laminate with long A-CNTs. (a) Minimal fiber pullout in fracture surface of initiation region. (b) Evolution of strain energyrelease rate along a long-CNT laminate with gradual increasing toughening (c) Tow breakage and microfiber pull-out in FFRP laminates.

S.S. Wicks et al. / Composites Science and Technology 100 (2014) 128–135 133

for 15 mm of crack advance until peaking, or toughening over theentire length of the sample. Samples in the latter category wereexcluded from the steady-state average as those laminates did notreach steady-state before the end of the test, so GIc,ss is conserva-tively based on specimens that did not have an increasing toughnesstrend to the end of the test. If the steady-state values were evaluatedto include the peak value reached by these specimens at the end ofthe specified crack length of 50 mm, the GIc,ss would have been2.41 kJ/m2, corresponding to a 102% increase over the baseline.The short CNT laminates are nearly as tough as the long CNT lami-nates, with the initiation toughness at 22–90% above baseline forthe 3 initiation measures, and the steady-state and max toughnessvalues increase by 89% and 99%, respectively.

The fracture surfaces of the marine epoxy laminates highlight amajor difference between the epoxy systems. All marine epoxy lam-inate fracture surfaces show widespread microfiber–matrix inter-face failure and microfiber pullout (see Fig. 6b and d, contrasted tothe morphology of the toughest regions of aerospace epoxy FFRPas discussed above in Fig. 5c). Fiber–matrix interface failure leavesportions of the fracture surface exposed and smooth, indicating thatthe prevention of debonding at the microfiber surface is not requiredfor interlaminar reinforcement. Instead, increased toughness inmarine epoxy laminates manifests in increased microfiber and towbreakage which can be observed during crack propagation as thesefibers are pulled free from the surrounding matrix and bridge across

Fig. 8. Example fractography of marine epoxy laminate with short A-CNTs. (a) Tow breaway. (b) Minimal fiber pullout and flat fracture surface one ply interface away from the cinitiation region. (d) Strain energy release rate along a short CNT laminate.

the crack tip, as seen in the side view of the crack wake in Fig. 6c. Inaddition to pull-out and bridging of microscale fibers as seen inFig. 6d and e, the broken epoxy between microfibers reveals pulloutand bridging of nanoscale CNTs, indicating toughening at multiplelength scales. The CNTs pulled out of the marine epoxy in Fig. 6fare up to a few microns in length, similar to the CNTs in the aero-space epoxy surfaces.

Increasing toughness in the FFRP laminates as seen in Fig. 7 cor-responds directly to increased microfiber breaks in the fracturesurface. This fracture morphology suggests that the CNTs increasethe toughness of the interlaminar region by creating a tortuouspath, even to the extreme such that the crack propagates throughthe surface of the weave rather than around each tow. The CNTsgrown on the microfiber surfaces within each tow help push apartthe microfibers causing the tow to swell, allowing for tows onopposite plies to nest and create a tortuous path for the crack tofollow. The monotonic increase from initiation to steady-statetoughness (limited by the laminate specimen test length inFig. 7b) may be caused by variations in weave nesting along thecrack length, which is prevented at initiation by the release filmin all baseline and FFRP laminates, but could vary along the lengthof the sample depending on how adjacent plies register to oneanother. Example fractography on short CNT marine epoxy isshown in Fig. 8, which demonstrates that while the fracturesurfaces show limited microfiber fracture, other toughening

akage and fiber pull-out as crack jumps from main crack path to one ply interfaceenterline main DCB crack path. (c) Flat fracture and fiber–matrix interface failure at

134 S.S. Wicks et al. / Composites Science and Technology 100 (2014) 128–135

mechanisms come into play such as crack jumping across plies toanother interface. Therefore, the grown A-CNTs change the micro-scale fracture behavior not only via pullout and altering mechani-cal properties of the matrix (simple nanofiber reinforcementchanging, e.g., moduli), but also by altering the failure path andthe fracture surface texture, adding surface area through bothmicrofiber and CNT pullout and failure.

Other mechanisms of interlaminar toughening also need to beconsidered in the context of the above findings. The A-CNTs donot quantitatively alter the interlaminar region (thickness) in thewoven laminate interface, nor do the A-CNTs cause any measur-able change in the polymer matrix degree of cure. Hard nanostruc-tures such as CNTs have been shown to change local crystallinity oralter polymer orientation in some cases, particularly for thermo-plastics [51,52]. However, this is not expected in these epoxies asconfirmed through prior studies on the aerospace epoxy showingno change in polymer orientation for a wide range of CNT volumefractions [53] and by preliminary DSC investigations which againshow no measurable change in the either epoxy’s characteristicswith and without CNTs. Thus, the two reinforcement mechanismsof either increased interlaminar thickness (adding a compliantinterlayer) or polymer morphology alteration are not contributingsubstantially to toughness trends observed in this work.

4. Conclusions and recommendations

The addition of aligned CNTs has been shown to improve thesteady-state Mode I fracture toughness of laminated composites.Fractography reveals toughening mechanisms for both aerospaceand marine epoxy laminates at several length scales, from thepull-out of carbon nanotubes to inducing tow breakage. The tough-ening behavior and magnitude of steady-state toughness improve-ment, however, is dependent on the type of epoxy. In the morebrittle aerospace epoxy laminate, modest improvement insteady-state toughness with long CNTs is made possible by main-taining cohesive interlaminar matrix failure around woven towfeatures while increasing fracture surface area through CNT pulloutand rough epoxy fracture. The tougher marine epoxy allows largeimprovements with CNTs by significantly adding instances ofmicrofiber breakage and pullout along with CNT pullout and epoxyfracture. Varying the CNT length begins to reveal how the geomet-rical arrangement of microfibers through tow swelling and changesin nesting affect the crack propagation path and subsequent inter-laminar toughness. When applying the fuzzy fiber architecture to achosen fiber and matrix system, the dependence of toughness onthe CNT length and polymer type will need to be established fortailoring interlaminar toughness.

The clarification of multi-scale toughening modes in this workwill be supplemented through future studies establishing theextent to which nanoscale reinforcement toughens the matrix only(i.e. no microfibers) and isolating the role and mechanisms of theCNT-matrix interface in resisting crack growth, leading to anunderstanding of how the choice of epoxy and nanofiber type dic-tate the magnitude of improvement possible. The alumina FFRPsystem studied here is a useful model system to understand fea-tures of aligned CNT property tailoring that are isolated frommicrofiber damage due to CNT growth. However, a more engineer-ing relevant system is carbon FFRP, particularly for aerospaceapplications. Future work will include study of carbon FFRP lami-nates, based on recent work demonstrating synthesis of CNTs oncarbon fibers that avoids deleterious damage to the fibers [30].

Acknowledgments

This work was supported by Airbus Group, Boeing, Embraer,Lockheed Martin, Saab AB, ANSYS Inc., Hexcel, and TohoTenax

through MIT’s Nano-Engineered Composite aerospace STructures(NECST) Consortium. Sunny is supported by the NASA Space Tech-nology Research Fellowships (NSTRF) and MIT’s Linda and Richard(1958) Hardy Fellowship. The authors thank Dale Lidston, SorayaKalamoun, and Henna Jethani for fabrication and testing assis-tance, and Hexcel for materials donation. This work made use ofthe MRSEC Shared Experimental Facilities at MIT, supported bythe NSF under award number DMR-08-19762, and core facilitiesat the Institute for Soldier Nanotechnologies at MIT supported bythe U.S. Army Research Office under contract W911NF-07-D-0004.

References

[1] Tong L, Mouritz AP, Bannister MK. 3D Fibre reinforced polymer composites.Oxford: Elsevier; 2002.

[2] Mouritz AP. Review of z-pinned composite laminates. Compos A2010;38(12):709–28.

[3] Dransfield KA, Jain LK, Mai Y-W. On the effects of stitching in CFRPs—1. Mode Idelamination toughness. Compos Sci Technol 1998;58:815–27.

[4] Mouritz AP, Cox BN. A mechanistic interpretation of the comparative in-planemechanical properties of 3D woven, stitched, and pinned composites. ComposA 2010;41(6):709–28.

[5] Yasaee M, Bond IP, Trask RS, Greenhalgh ES. Mode I interfacial tougheningthrough discontinuous interleaves for damage suppression and control.Compos A 2012;43(1):198–207.

[6] Sager TJ, Klein PJ, Davis DC, et al. Interlaminar fracture toughness of wovenfabric composite laminates with carbon nanotube/epoxy interleaf films. J ApplPolym Sci 2011;121(4):2394–405.

[7] Böger L, Sumfleth J, Hedemann H, Schulte K. Improvement of fatigue life byincorporation of nanoparticles in glass fibre reinforced epoxy. Compos A2010;41(10):1419–24.

[8] Hayes BS, Seferis JC. Modification of thermosetting resins and compositesthrough preformed polymer particles: a review. Polym Compos2001;22(4):451–67.

[9] Siddiqui NA, Woo RSC, Kim J-K, Leung CCK, Munir A. Mode I interlaminarfracture behavior and mechanical properties of CFRPs with nanoclay-filledepoxy matrix. Compos A 2007;38(2):449–60.

[10] Zhou Y, Jeelani S, Lacy T. Experimental study on the mechanical behavior ofcarbon/epoxy composites with a carbon nanofiber-modified matrix. J ComposMater 2013. 0021998313512348.

[11] Bortz DR, Merino C, Martin-Gullon I. Mechanical characterization ofhierarchical carbon fiber/nanofiber composite laminates. Compos A2011;42(11):1584–91.

[12] Coleman JN, Khan U, Blau WJ, Gun’ko YK. Small but strong: a review of themechanical properties of carbon nanotube–polymer composites. Carbon2006;44(9):1624–52.

[13] Guzmán de Villoria R, Yamamoto N, Miravete A, et al. Multi-physics damagesensing in nano-engineered structural composites. Nanotechnology2011;22(18):185502.

[14] Gibson RF. A review of recent research on mechanics of multifunctionalcomposite materials and structures. Compos Struct 2010;92(12):2793–810.

[15] Wu AS, Na W-J, Yu W-R, et al. Carbon nanotube film interlayer for strain anddamage sensing in composites during dynamic compressive loading. ApplPhys Lett 2012;101(22):221909.

[16] Gao L, Chou T-W, Thostenson ET, et al. In situ sensing of impact damage inepoxy/glass fiber composites using percolating carbon nanotube network.Carbon 2011;49(10):3382–5.

[17] Seyhan AT, Tanoglu M, Schulte K. Mode I and mode II fracture toughness of E-glass non-crimp fabric/carbon nanotube (CNT) modified polymer basedcomposites. Eng Fract Mech 2008;75(18):5151–62.

[18] Gojny FH, Wichmann MHG, Köpke U, et al. Carbon nanotube-reinforced epoxy-composites: enhanced stiffness and fracture toughness at low nanotubecontent. Compos Sci Technol 2008;64(15):2363–71.

[19] Thakre PR, Lagoudas DC, Riddick, et al. Investigation of the effect of single wallcarbon nanotubes on interlaminar fracture toughness of woven carbon fiber–epoxy composites. J Compos Mater 2011;45(10):1091–108.

[20] Godara A, Mezzo L, Luizi F, Warrier A, et al. Influence of carbon nanotubereinforcement on the processing and the mechanical behaviour of carbonfiber/epoxy composites. Carbon 2009;47(12):2914–23.

[21] Chou T-W, Gao L, Thostenson ET, Zhang Z, Byun J-H. An assessment of thescience and technology of carbon nanotube-based fibers and composites.Compos Sci Technol 2010;70(1):1–19.

[22] Bekyarova E, Thostenson ET, Yu A, et al. Functionalized single-walled carbonnanotubes for carbon fiber epoxy composites. J Phys Chem C2007;111(48):17865–71.

[23] Lee S-B, Choi O, Lee W, et al. Processing and characterization of multi-scalehybrid composites reinforced with nanoscale carbon reinforcements andcarbon fibers. Compos A 2011;42(4):337–44.

[24] Chandrasekaran V, Advani S, Santare M. Role of processing on interlaminarshear strength enhancement of epoxy/glass fiber/multi-walled carbonnanotube hybrid composites. Carbon 2010;48(13):3692–9.

S.S. Wicks et al. / Composites Science and Technology 100 (2014) 128–135 135

[25] Fan Z, Santare MH, Advani SG. Interlaminar shear strength of glass fiberreinforced epoxy composites enhanced with multi-walled carbon nanotubes.Compos A 2008;9(3):540–54.

[26] Ashrafi B, Guan J, Mirjalili V, et al. Enhancement of mechanical performance ofepoxy/carbon fiber laminate composites using single-walled carbonnanotubes. Compos Sci Technol 2011;71(13):1569–78.

[27] Gojny FH, Wichmann MHG, Fiedler B, et al. Influence of nano-modification onthe mechanical and electrical properties of conventional fibre-reinforcedcomposites. Compos A 2005;36(11):1525–35.

[28] Isayev AI, Kumar R, Lewis TM. Ultrasound assisted twin screw extrusion ofpolymer-nanocomposites containing carbon nanotubes. Polymer2009;50(1):250–60.

[29] Qiu J, Zhang C, Wang B, Liang R. Carbon nanotube integrated multifunctionalmultiscale composites. Nanotechnology 2007;18(27):275708.

[30] Steiner SA, Li R, Wardle BL. Circumventing the mechanochemical origins ofstrength loss in the synthesis of hierarchical carbon fibers. ACS Appl MaterInterfaces 2013;4(11):4892–903.

[31] Lomov SV, Gorbatikh L, Kotanjac Z, et al. Compressibility of carbon wovenfabrics with carbon nanotubes/nanofibres grown on the fibres. Compos SciTechnol 2011;71(3):315–25.

[32] Lomov SV, Wicks S, Gorbatikh L, et al. Compressibility of nanofibre-graftedalumina fabric and yarns: Aligned carbon nanotube forests. Compos SciTechnol 2014;90:57–66.

[33] Veedu VP, Cao A, Li X, et al. Multifunctional composites using reinforcedlaminae with carbon-nanotube forests. Nat Mater 2006;5(6):457–62.

[34] Kepple KL, Sandborn GP, Lacasse PA, et al. Improved fracture toughness ofcarbon fiber composite functionalized with multi walled carbon nanotube.Carbon 2008;46(15):2026–33.

[35] Qian H, Bismarck A, Greenhalgh ES, Shaffer MSP. Carbon nanotube graftedcarbon fibres: a study of wetting and fibre fragmentation. Compos A2010;41(9):1107–14.

[36] Sager RJ, Klein PJ, Lagoudas DC, Zhang Q. Effect of carbon nanotubes on theinterfacial shear strength of T650 carbon fiber in an epoxy matrix. Compos SciTechnol 2009;69(7–8):898–904.

[37] Zhang FH, Wang RG, He XD, Wang C, Ren LN. Interfacial shearing strength andreinforcing mechanisms of an epoxy composite reinforced using a carbonnanotube/carbon fiber hybrid. J Mater Sci 2009;44(13):3574–7.

[38] Qian H, Kalinka G, Chan KL, et al. Mapping local microstructure andmechanical performance around carbon nanotube grafted silica fibres:methodologies for hierarchical composites. Nanoscale 2011;3(11):4759–67.

[39] Qian H, Greenhalgh ES, Shaffer MSP, Bismarck A. Carbon nanotube-basedhierarchical composites: a review. J Mater Chem 2010;20(23):4751–62.

[40] Garcia EJ, Hart J, Wardle BL. Long carbon nanotubes grown on the surface offibers for hybrid composites. AIAA J 2008;46(6):1405–12.

[41] Lachman N, Wiesel E, Guzmán de Villoria R. Interfacial load transfer in carbonnanotube/ceramic microfiber hybrid composites. Compos Sci Technol2012;72(12):1416–22.

[42] Jin S-Y, Young RJ, Eichhorn SJ. Hybrid carbon fibre–carbon nanotube compositeintefaces. Compos Sci Technol 2014;95:114–20.

[43] Allen RJ, Ghita O, Farmer B, et al. Mechanical testing and modelling of avertically aligned carbon nanotube composite structure. Compos Sci Technol2013;77:1–7.

[44] Garcia EJ, Wardle BL, Hart AJ, Yamamoto N. Fabrication and multifunctionalproperties of a hybrid laminate with aligned carbon nanotubes grown In Situ.Compos Sci Technol 2008;68(9):2034–41.

[45] Wicks SS, Guzmán de Villoria R, Wardle BL. Interlaminar and intralaminarreinforcement of composite laminates with aligned carbon nanotubes.Compos Sci Technol 2010;70(1):20–8.

[46] Yamamoto N, Hart AJ, Garcia EJ, et al. High-yield growth andmorphology control of aligned carbon nanotubes on ceramic fibers formultifunctional enhancement of structural composites. Carbon 2009;47(3):551–60.

[47] Yamamoto N, Guzmán de Villoria R, Wardle BL. Electrical and thermalproperty enhancement of fiber-reinforced polymer laminate compositesthrough controlled implementation of multi-walled carbon nanotubes.Compos Sci Technol 2012;72(16):2009–15.

[48] Wicks SS, Kalamoun S, Williams MR, et al., Effect of manufacturing route onmode I fracture toughness of aligned carbon nanotube reinforced composites.In: 53rd AIAA Structures, Structural Dynamics, and Materials (SDM)conference, Honolulu, HI, April 23–26; 2012.

[49] ASTM D 5528-01. Standard Test Method for Mode I Interlaminar FractureToughness of Unidirectional Fiber-Reinforced Polymer Matrix Composites,ASTM International.

[50] Wagner HD, Ajayan PM, Schulte K. Nanocomposite toughness from a pull-outmechanism. Compos Sci Technol 2013;83:27–31.

[51] Ma Q, Mao B, Cebe P. Chain confinement in electrospun nanocomposites: usingthermal analysis to investigate polymer–filler interactions. Polymer2011;52(14):3190–200.

[52] Meng J, Zhang Y, Song K, Minus M. Forming crystalline polymer-nanointerphase structures for high-modulus and high-tensile/strength compositefibers. Macromol Mater Eng 2014;299(2):144–53.

[53] Wardle BL, Saito DS, Garcia EJ, et al. Fabrication and characterization of ultra-high volume fraction aligned carbon nanotube–polymer composites. AdvMater 2008;20(14):2655–796.