492 E.T. Thostenson et al. / Composites Science and Technology 65 (2005) 491–516 8. Precursors for macroscopic composites . . . . . . . . . . . . . . . . . . . . . . . . ............................. . . . 509 8.1. Long nanotube fibers and strands. . . . . . . . . . . . . . . . . . . . . . . ............................ . . . 509 8.2. Multi-scale hybrid composites . . . . . . . . . . . . . . . . . . . . . . . . . ............................ . . . 510 8.2.1. Controlled growth of carbon nanotubes on fiber surfaces . ............................. . . . 510 8.2.2. Nanoclay-enhanced matrix . . . . . . . . . . . . . . . . . . . . . . ............................. . . . 510 8.2.3. Nanotube or nanofiber-reinforced matrix . . . . . . . . . . . . ............................. . . . 511 8.3. Fibers and films . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ............................ . . . 511 8.3.1. Nanocomposite fibrils . . . . . . . . . . . . . . . . . . . . . . . . . . ............................. . . . 511 8.3.2. Nanocomposite films . . . . . . . . . . . . . . . . . . . . . . . . . . ............................. . . . 512 9. Concluding remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 513 Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 513 References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5131. Introduction from the bottom-up, that is, by building them from atoms, molecules, and the nanoscale powders, ﬁbers In 1985 Professor A. Kelly authored an article in and other small structural components made from them.Composites Science and Technology titled ‘‘Composites This diﬀers from all previous manufacturing, in whichin Context’’ . It was stated that, ‘‘The large scale so- raw materials. . . get pressed, cut, molded and otherwisecial changes which inﬂuence the development of new coerced into parts and products’’ .materials are reviewed and the new materials and pro- Scientists and engineers working with ﬁber-reinforcedcessing methods being developed in response to these composites have practiced this bottom-up approach inare described and contrasted with some recent advances processing and manufacturing for decades. Whenin composite materials science.’’ Emerging technologies designing a composite the material properties are tai-at the time included in situ metal–matrix composites, lored for the desired performance across various lengthcarbon-ﬁber-reinforced thermoplastic composites, SiC- scales. From selection and processing of matrix and ﬁberreinforced aluminum as well as toughening of ceramics materials, and design and optimization of the ﬁber/ma-through the use of ﬁber-reinforcement. Tremendous trix interface/interphase at the sub-micron scale to thedevelopments have been made  in many aspects of manipulation of yarn bundles in 2-D and 3-D textilescomposites research and technology during the two dec- to the lay-up of laminae in laminated composites and ﬁ-ades since the publication of KellyÕs paper. Recent ad- nally the net-shape forming of the macroscopic compos-vances in producing nanostructured materials with ite part, the integrated approach used in compositesnovel material properties have stimulated research to processing is a remarkable example in the successfulcreate multi-functional macroscopic engineering materi- use of the ‘‘bottom-up’’ approach.als by designing structures at the nanometer scale. Moti- The expansion of length scales from meters (ﬁnishedvated by the recent enthusiasm in nanotechnology, woven composite parts), micrometers (ﬁber diameter),development of nanocomposites is one of the rapidly sub-micrometers (ﬁber/matrix interphase) to nanome-evolving areas of composites research. ters (nanotube diameter) presents tremendous opportu- Nanotechnology can be broadly deﬁned as, ‘‘The cre- nities for innovative approaches in the processing,ation, processing, characterization, and utilization of characterization, and analysis/modeling of this new gen-materials, devices, and systems with dimensions on the eration of composite materials. As scientists and engi-order of 0.1–100 nm, exhibiting novel and signiﬁcantly neers seek to make practical materials and devicesenhanced physical, chemical, and biological properties, from nanostructures, understanding material behaviorfunctions, phenomena, and processes due to their nano- across length scales from the atomistic to macroscopicscale size’’ . Current interests in nanotechnology levels is required. Knowledge of how the nanoscaleencompass nano-biotechnology, nano-systems, nano- structure inﬂuences the bulk properties will enable de-electronics, and nano-structured materials, of which sign of the nanostructure to create multi-functionalnanocomposites are a signiﬁcant part. composites. Through nanotechnology, it is envisioned that nano- The challenges in nanocomposites research perhapsstructured materials will be developed using a bottom-up can be best illustrated by the electron micrographsapproach. ‘‘More materials and products will be made shown in Fig. 1 [5–7], where multi-walled carbon nano-
E.T. Thostenson et al. / Composites Science and Technology 65 (2005) 491–516 493tubes (MWCNTs, 10–20 nm in diameter) have beendeposited on the surface of carbon ﬁbers (7 lm in diam-eter) in yarn bundles (measured in millimeters). Whenconsolidated into a composite, the reinforcement scalesspan seven orders of magnitude. Fig. 2 shows a trans-mission electron microscope (TEM) image of the nano-composite structure near the ﬁber/matrix interface,where the diﬀerence in reinforcement scale is readilyapparent. A morphological characteristic that is of fundamentalimportance in the understanding of the structure–prop-erty relationship of nanocomposites is the surface area/volume ratio of the reinforcement materials. This paper Fig. 2. TEM micrograph showing the nanotube composite structurediscusses nanocomposites based upon the three catego- directly adjacent to the carbon ﬁber/polymer matrix interface .ries of reinforcement materials: particles (silica, metal,and other organic and inorganic particles), layeredmaterials (graphite, layered silicate, and other layered compare the properties and performance of nanocom-minerals), and ﬁbrous materials (nanoﬁbers and nano- posites with traditional ﬁber composites.tubes). As illustrated in Fig. 3, the change in particlediameter, layer thickness, or ﬁbrous material diameterfrom micrometer to nanometer, changes the ratio bythree orders in magnitude. At this scale, there is often 2. Nanoparticle-reinforced compositesdistinct size dependence of the material properties. Inaddition, with the drastic increase in interfacial area, Particulate composites reinforced with micron-sizedthe properties of the composite become dominated more particles of various materials are perhaps the mostby the properties of the interface or interphase. widely utilized composites in everyday materials. Parti- In this paper, we address the state of knowledge in cles are typically added to enhance the matrix elasticprocessing, characterization, and analysis/modeling of modulus and yield strength. By scaling the particle sizenanocomposites with a particular emphasis on identify- down to the nanometer scale, it has been shown that no-ing fundamental structure/property relationships and vel material properties can be obtained. A few systemsFig. 1. Variation in reinforcement scales from millimeters to nanometers: (from left) from woven fabric of yarn bundles, to a single carbon ﬁber withentangled carbon nanotubes grown on the surface [5,6], to the nanometer diameter and wall structure of the carbon nanotube .
494 E.T. Thostenson et al. / Composites Science and Technology 65 (2005) 491–516 Fig. 3. Surface area/volume relations for varying reinforcement geometries.are reviewed below for illustrating the resulting modiﬁ- the composite modulus increases monotonically withcation in matrix properties. particle weight fraction. However, the strengths of com- Micron-scale particles typically scatter light making posites are all below the strength of neat resin due tootherwise transparent matrix materials appear opaque. non-uniform particle size distribution and particleNaganuma and Kagawa  showed in their study of aggregation. The work of Thompson et al.  on metalSiO2/epoxy composites that decreasing the particle size oxide/polyimide nanocomposite ﬁlms also noted similarresulted in signiﬁcantly improved transmittance of visi- diﬃculties in processing. Their study utilized antimonyble light. Singh et al.  studied the variation of fracture tin oxide (11–29 nm), indium tin oxide (17–30 nm) andtoughness of polyester resin due to the addition of alu- yttrium oxide (11–44 nm) in two space-durable polyi-minum particles of 20, 3.5 and 100 nm in diameter. mides: TOR-NC and LaRC TMCP-2. The nanoscaleFig. 4 shows that the initial enhancement in fracture additives resulted in higher stiﬀness, comparable or low-toughness is followed by decreases at higher particle vol- er strengths and elongation, and lower dynamic stiﬀnessume fraction. This phenomenon is attributed to the (storage modulus). The dispersion of metal oxides on aagglomeration of nanoparticles at higher particle vol- nanometer scale was not achieved.ume content. Ash et al.  studied the mechanical behavior of alu- Lopez and co-workers  examined the elastic mod- mina particulate/poly(methyl methacrylate) composites.ulus and strength of vinyl ester composites with the They concluded that when a weak particle/matrix inter-addition of 1, 2 and 3 wt% of alumina particles in the face exists, the mode of yielding for glassy, amorphoussizes of 40 nm, 1 lm and 3 lm. For all the particle sizes, polymers changes from cavitational to shear, which leads to a brittle-to-ductile transition. This behavior is attributed to increased polymer chain mobility, due to the presence of smaller particles, and also the capability to relieve tri-axial stress because of poorly bonded larger particles. An extensive review of the structure–property relationships in nanoparticle/semi-crystalline thermo- plastic composites has been made by Karger-Kocsis and Zhang . 3. Nanoplatelet-reinforced composites Two types of nanoplatet-reinforced composites areFig. 4. Normalized fracture toughness with respect to volume fraction reviewed: clay and graphite. In their bulk state, bothfor various sized particles . Reprinted with permission from . clay and graphite exist as layered materials. In orderCopyright (2002) Kluwer Academic Publishers. to utilize these materials most eﬃciently, the layers must
E.T. Thostenson et al. / Composites Science and Technology 65 (2005) 491–516 495be separated and dispersed throughout the matrixphase. The morphology of clay/polymer nanocompos-ites is illustrated in Fig. 5 . In the conventional mis-cible state, the interlayer spacing in a clay particle is atits minimum. When polymer resin is inserted into thegallery between the adjacent layers, the spacing expands,and it is known as the intercalated state. When the layersare fully separated, the clay is considered to be exfoli-ated. Fig. 6 shows the TEM image of a montmorillonitepoly (L -lactic acid) (PLLA) matrix nanocomposite, dem-onstrating intercalated and exfoliated clay layers . Montmorillonite, saponite, and synthetic mica arecommonly used clay materials, and developments inclay-based nanocomposites have been recently reviewed[16–18]. The advantages of polymer-based clay nano-composites include improved stiﬀness, strength, tough-ness, and thermal stability as well as reduced gaspermeability and coeﬃcient of thermal expansion. Table1 shows some properties of two commercially availableclay particles with surface modiﬁcations [19,20]. Thelack of aﬃnity between hydrophilic silicate and hydro-phobic polymer causes agglomeration of the mineral inthe polymer matrix. Surface modiﬁcation of clay parti- Fig. 6. TEM micrograph of a montmorillonite poly (L -lactic acid)cles facilitates the compatibility. nanocomposite, showing both intercalated and exfoliated states . Table 2 illustrates the unique performance of Nylon- Reprinted with permission from . Copyright (2003) American6/clay hybrid over a wide range of mechanical and ther- Chemical Society.mal properties, as summarized by Okada and Usuki. The pioneering work at Toyota Research Lab has Table 1clearly demonstrated that the addition of small amounts Properties of clay platelets [19,20]of montmorillonite clay material signiﬁcantly enhances Physical properties ClositeÒ 30B Nanomer 1.28Ethe tensile strength, tensile modulus, and heat degrada- Color Oﬀ white Whitetion temperature (HDT), and reduces the rate of water Density (g/cm3) 1.98 1.90 ˚ D-spacing (D0 0 1), A 18.5 >20absorption and CTE in the ﬂow direction. Aspect ratio 200–1000 200–500 Table 3 summarizes the constituent properties of Surface area (m2/g) 750 750exfoliated clay [14,21,22]. Luo and Daniel  have Mean particle size (lm) 6 8–10 modeled the YoungÕs modulus of clay nanocomposites using a three-phase model: epoxy matrix, exfoliated clay nanolayer, and intercalated clay cluster (parallel platelet system). Fig. 7 shows that the experimental data lie within the upper (Voigt) and lower (Reuss) bond predic- tions and coincide fairly well with the Mori–Tanaka and Eshelby model predictions. The modeling work of Tsai and Sun  demonstrated that well dispersed platelets in the polymer matrix could signiﬁcantly enhance the load transfer eﬃciency in these composites. Fig. 8 compares the fracture toughness, KIC, of epoxy matrix (DGEBA) composites reinforced with interca- lated (ODTMA) and exfoliated (MT2EtOH) clay up to 10 vol%. Miyagawa and Drzal  attribute the high fracture toughness of intercalated clay composites to crack bridging by clay particles as well as crack deﬂec-Fig. 5. Morphologies of polymer/clay nanocomposites: (a) conven- tion due to excellent adhesion of clay/epoxy interfacetional miscible, (b) partially intercalated and exfoliated, (c) fully and clay aggregate strength. On the other hand, the frac-intercalated and dispersed and (d) fully exfoliated and dispersed . ture of individual clay platelets in exfoliated clay
496 E.T. Thostenson et al. / Composites Science and Technology 65 (2005) 491–516Table 2Properties of Nylon-6/clay nanocomposites Sample Wt% clay Montmorillonite Tensile strength (MPa) Tensile modulus (GPa) Charpy impact strength (kJ/m3) HDT at 18.5 kg cmÀ2 (°C)NCH-5 4.2 107 2.1 2.1 152NCC-5 5.0 61 1.0 1.0 89Nylon-6 0 69 1.1 1.1 65Sample Wt% clay Montmorillonite Rate of water absorption 25 °C, 1 day CTE (°C · 10À5) Flow direction Perpendicular directionNCH-5 4.2 0.51 6.3 13.1NCC-5 5.0 0.90 10.3 13.4Nylon-6 0 0.87 11.7 11.8Table 3Properties of exfoliated graphite platelets as compared to exfoliatedclay platelets Graphene sheeta Clay plateletb,cPhysical structure $1 nm · 100 nm $1 nm · 1000 nmTensile modulus (GPa) 1,000 170Tensile strength (GPa) 10–20 1Resistivity (X cm) $50 · 10À6 1010–1016Thermal conductivity (W/m K) 3000 0.67CTE $1 · 10À6 8–16 · 10À6Density (g/cm3) 2.0 2.5–3.0D-spacing (nm) 0.34 1.85 a Fukushima and Drzal . b Miyagawa and Drzal . c Luo and Daniel . Fig. 8. Inﬂuence of clay content and exfoliation on fracture toughness . gas diﬀusivity as compared with unreinforced epoxy. The results are consistent with the Hatta-Taya theory and it was revealed that the dispersion of platelets is more eﬀective than spherical or ﬁber-like reinforcement in improving the nanocomposite barrier properties. Regarding the other layered material, the exfoliated graphite or graphene sheet has about the same thickness as exfoliated clay. Table 3 shows their high tensile mod- ulus, tensile strength, thermal conductivity, and low electrical resistivity, comparing to clay platelets. The low electrical resistivity of exfoliated graphite facilitates the conductivity of polymer composites when a thresh- old percolation weight content of the conductive phaseFig. 7. Comparison of theoretical models with experimental data for is reached. Fig. 9 shows the results of Fukushima andnanocomposite elastic modulus . Drzal  on the resistivity of epoxy matrix (EPON 828) composites with the addition of vapor grown car-composites leads to less rough fracture surface and bon ﬁber (VGCF), carbon black, PAN carbon ﬁber,lower fracture toughness. exfoliated graphite and milled graphite. The graphite In addition to mechanical properties, the thermal sta- nanoplatelets are treated in O2 plasma for initiating rad-bility, ﬁre resistance and gas barrier properties of poly- ical polymerization. The percolation threshold for exfo-mer/clay nanocomposites can be enhanced through the liated graphite is around 1 wt%. Zheng and co-workersaddition of nanometer-scale reinforcement. For exam- [25,26] have also reported reduced percolation thresh-ple, Ogasawara et al.  have investigated the helium olds in exfoliated graphite nanoplatelet/thermoplasticpermeability of nanoclay for potential applications in li- composites.quid hydrogen tanks and fuel cells. The addition of Song et al.  have modeled the von Mises stress dis-nanoclay to the epoxy resin substantially decreased the tribution in graphite/PAN nanocomposites with respect
E.T. Thostenson et al. / Composites Science and Technology 65 (2005) 491–516 497 Fig. 9. Comparison of various carbon reinforcement materials on the composite bulk resistivity .to the level of exfoliation. They demonstrated the reduc- tinuum elastic theory and molecular dynamics simula-tion in the magnitude of stress concentration with layer tions. The axial YoungÕs modulus of the nanoﬁber isthickness at the tip of graphite layers, aligned with the particularly sensitive to the shell tilt angle, where ﬁbersaxial tensile load. Such conclusion is similar to those ob- that have small tilt angles from the axial direction showtained for short ﬁber composites where the dispersion of much higher YoungÕs modulus than ﬁbers with large tilta ﬁber bundle has the beneﬁt of reducing the stress con- angles. The wide-ranging morphology of the carboncentration in the matrix material at the bundle ends and nanoﬁbers and their associated properties results in athus, the chance of matrix cracking or plastic broad range of scatter for experimental results on pro-deformation. cessing and characterization of nanoﬁber composites. The greatly enhanced electrical conductivity of poly- Finegan and co-workers [32,33] have investigated themeric material with the addition of small amount of processing and properties of carbon nanoﬁber/polypro-graphite platelets has found many practical applications. pylene nanocomposites. In their work, they used a vari-These include EMI shielding and heat management of ety of as-grown nanoﬁbers. Carbon nanoﬁbers that wereelectronic and computer devices or equipment, electro- produced with longer gas phase feedstock residencestatic paint for automobiles and polymer sheath of elec- times were less graphitic but adhered better to the poly-tric cables . propylene matrix, with composites showing improved Nanoclay/polypropylene composites are being used tensile strength and YoungÕs modulus. Oxidation ofas functional parts in automobiles. For instance, Gen- the carbon nanoﬁber was found to increase adhesioneral Motors (GM) is using about 660,000 lbs of nano- to the matrix and increase composite tensile strength,composite material per year. but extended oxidation reduced the properties of the ﬁbers and their composites. In their investigation on the nanoﬁber composite damping properties, Finegan4. Nanoﬁber-reinforced composites et al.  concluded that the trend of stiﬀness varia- tion with ﬁber volume content is opposite to the trend Vapor grown carbon nanoﬁbers (CNF) have been of loss factor and damping in the composite isused to reinforce a variety of polymers, including poly- matrix-dominated.propylene, polycarbonate, nylon, poly(ether sulfone), Ma and co-workers  and Sandler et al.  havepoly(ethylene terephthalate), poly(phenylene sulﬁde), spun polymer ﬁbers with carbon nanoﬁbers as reinforce-acrylonitrile-butadiene-styrene (ABS), and epoxy. ment. Ma et al. utilized a variety of techniques toCarbon nanoﬁbers are known to have wide-ranging achieve dispersion of carbon nanoﬁbers in a poly(ethyl-morphologies, from structures with a disordered ene terephthalate) (PET) matrix and subsequently melt-bamboo-like structure  (Fig. 10(a)) to highly graph- spun ﬁbers. The compressive strength and torsionalitized ‘‘cup stacked’’ structures [29,30] (Figs. 10(b) and moduli of the nanocomposite ﬁbers were considerably(c)), where the conical shells of the nanoﬁber are nested higher than that for the unreinforced PET ﬁber. Sandlerwithin each other. Carbon nanoﬁbers typically have et al.  produced ﬁbers from semicrystalline high-diameters on the order of 50–200 nm. Wei and Srivast- performance poly(ether ether ketone) (PEEK) contain-ava  have modeled the mechanical properties of ing up to 10 wt% vapor-grown carbon nanoﬁbers. Theircarbon nanoﬁbers with varying morphology using con- experimental results highlight the need to characterize
498 E.T. Thostenson et al. / Composites Science and Technology 65 (2005) 491–516 ﬁber composites, such as dispersion and adhesion, are similar to those for nanotube-reinforced composites and are discussed in the following section. 5. Carbon nanotube-reinforced composites As reviewed by Thostenson et al. , the morphol- ogy of a carbon nanotube is deﬁned by the orientation and magnitude of the chiral vector in a grephene sheet, which is ‘‘wrapped up’’ to form the single-walled carbon nanotube (SWCNT). The two limiting conﬁgurations are armchair and zigzag nanotubes (Fig. 11). Since their observation over a decade ago , numerous investigators have reported remarkable phys- ical and mechanical properties of carbon nanotube. Be- low is a summary of these exceptional properties excerpted from Collins and Avouris . The density of a SWCNT is about 1.33–1.40 g/cm3, which is just one-half of the density of aluminum. The elastic modulus of SWCNT is comparable to that of diamond (1.2 TPa). The reported tensile strength of SWCNT is much higher than that of high-strength steel (2 GPa). The tremendous resilience of SWCNT in sustaining bending to large angles and restraightening without damage is distinctively diﬀerent from the plastic defor- mation of metals and brittle fracture of carbon ﬁbers at much lower strain when subjected to the same type of deformation. The electric current carrying capability is estimated to be 1 · 109 amp/cm2, whereas copper wires burn out at about 1 · 106 amp/cm2. The thermal conductivity of SWCNT is predicted to be 6000 W/ m K at room temperature; this is nearly double the ther- mal conductivity of diamond of 3320 W/m K. SWCNTs are stable up to 2800 °C in vacuum and 750 °C in air, whereas metal wires in microchips melt at 600–1000 °C. SWCNTs have great potential in ﬁeld emission applications because they can activate phosphors at 1– 3 V if electrodes are spaced 1 lm apart. Traditional Mo tips require ﬁelds of 10–100 V/lm and have very limited lifetimes.Fig. 10. TEM micrographs of the nanoscale structure of carbonnanoﬁbers showing: (a) disordered bamboo-like structures ,(b) highly graphitized sidewall of a cup-stacked (molecular modelsinset) nanoﬁbers showing the shell tilt angle  and (c) a nesting of thestacked layers . Reprinted with permission from . Copyright(2003) American Chemical Society.both the crystalline matrix morphology and nanocom-posite structure when evaluating performance. This iscrucial to understanding the intrinsic properties of thenanoscale reinforcement. Many of the key challenges associated with the pro- Fig. 11. Atomic structures of (a) armchair and (b) zig-zag carboncessing, characterization, and modeling of carbon nano- nanotubes .
E.T. Thostenson et al. / Composites Science and Technology 65 (2005) 491–516 499 The outstanding thermal and electric properties com- structure. To obtain statistically meaningful data nearlybined with their high speciﬁc stiﬀness and strength, and 700 nanotubes of the same batch were examined. Fig. 12very large aspect ratios have stimulated the development shows a TEM micrograph of a MWCNT, indicating theof nanotube-reinforced composites for both structural outer (d) and inner (di) diameter as well as a histogramand functional applications [36,39,40]. for the nanotube outside diameter distribution, which is a bimodal distribution with peaks near 18 and 30 lm. In order to obtain a probability density function for5.1. Carbon nanotube morphology the nanotube diameter distribution, Thostenson and Chou  ﬁtted the data of Fig. 12 to double Lorentzian Carbon nanotubes have been fabricated by a variety distribution and double Gaussian distributions. Forof techniques . The morphology of nanotubes exhib- small-diameter nanotubes, the Gaussian curve mostits a great degree of variability. Furthermore, in spite of accurately ﬁt the data, but for large-diameter nanotubes,our knowledge in general regarding the exceptional the Gaussian curve underestimates the amount of nano-properties measured and predicted to date, our ability tubes. The volume distribution of the nanotubes is ob-in applying carbon nanotubes to structural and func- tained from the diameter distribution functions.tional composites is handicapped by the lack of a com- Although the large-diameter nanotubes are of a rela-prehensive grasp of the basic and precise knowledge of tively small percentage of the total number of nano-their properties. The brief review of the morphology tubes, they occupy a signiﬁcant percentage of volumeand properties given in this sub-section serves to exem- within the composite. The more accurate modeling ofplify such variability. the volume distribution by the Lorentzian curve at large First, regarding the morphology of SWCNTs, the nanotube diameter is crucial because the volume occu-variability could include the nanotube length, diameter pied by a given nanotube in the composite varies withand chirality as well as the tube-end conﬁguration d2.(end-caps). The variability in morphology is much more The wall thickness of the same MWCNTs as in Fig.pronounced in a MWCNT, which can be considered as 12 are presented in Fig. 13(a) as a function of nanotubecomposed of nested SWCNTs. The major additional diameter. A strong linear relationship between the nano-structural parameters include nanotube outer and inner tube diameter and wall thickness is observed. The nano-diameter, number of nested SWCNTs (wall thickness), scale tubular structure of the MWCNTs also results in aand growth-induced conﬁguration, such as bamboo distribution of nanotube density. From the measure-structures. ment of inside and outside diameter, the nanotube den- To demonstrate the structural variability of sity per unit length can be calculated by assuming thatMWCNTs, Thostenson and Chou  have utilized the graphite layers of the tube shell have the density ofhigh-resolution TEM micrographs taken of the CVD- fully dense graphite (qg = 2.25 g/cm3). The nanotubegrown tubes and image analysis software to measure density as a function of diameter is shown in Fig.the structural dimensions for quantifying both the distri- 13(b), where the curved line is obtained directly frombution of nanotube diameter and the nanotube wall the straight line in Fig. 13(a). Fig. 12. (a) Diameter distribution of CVD-grown multi-walled carbon nanotubes taken from measurements of (b) TEM micrographs .
500 E.T. Thostenson et al. / Composites Science and Technology 65 (2005) 491–516 Fig. 13. Variation in: (a) nanotube wall thickness with nanotube diameter and (b) density with nanotube diameter .Fig. 14. Diameter sensitivity of elastic modulus: (a) predicted by molecular structural mechanics and Li and Chou  and (b) other methods(Hernandez et al. ; Lu ; Krishnan ).5.2. Elastic and strength properties of carbon nanotubes ulus is about 0.5 TPa for diameters >1.25 nm. Both the YoungÕs modulus and shear modulus values ap- The elastic and strength properties of SWCNT and proach to the corresponding values of graphite at largeMWCNTs have been extensively studied both analyti- tube diameter. Fig. 14(b) shows the nanotube diametercally and experimentally. Review articles on these sub- dependence of axial modulus predicted by Hernandezjects can be found, for example, in Thostenson et al. et al.  using tight-binding molecular dynamics, and, Qian et al.  and Srivastava et al. . The follow- the shear modulus predictions of Lu  based on latticeing brief summary not only presents some recent results dynamics. Besides the axial YoungÕs modulus and shearbut also illustrates the variability of results obtained modulus, the transverse YoungÕs moduli of SWCNTsfrom analytical predictions as well as experimental mea- and double-walled carbon nanotubes (DWCNT) havesurements. Such variability poses considerable uncer- been studied using the molecular structural mechanicstainty in utilizing the elastic modulus and strength data method . The variations of PoissonÕs ratio ofin micromechanical models for composites. SWCNTs with the tube diameter predicted by the Fig. 14(a) gives the axial tensile YoungÕs modulus and molecular structural mechanics approach of Li andshear modulus as functions of SWCNT diameter, which Chou are shown to depend strongly on both nanotubewere obtained by Li and Chou  using the molecular diameter and chirality (Fig. 15). The results of Hernan-structural mechanics approach. The results are sensitive dez et al. , Lu , and Sanchez-Portal et al.  onto nanotube diameter and nanotube structure at small the other hand, are nearly insensitive to tube diameter.diameter. The axial YoungÕs modulus approaches to The recent results of the axial YoungÕs modulus of car-1.03 TPa for diameters P1.0 nm, while the shear mod- bon nanotubes are summarized in Table 4 [43,45,49–55].
E.T. Thostenson et al. / Composites Science and Technology 65 (2005) 491–516 501 Table 5 Experimental and theoretical results for nanotube axial strength Axial strength (GPa) Method of measurement Experimental measurements 13–52 Dual AFM cantilivers, SWCNT bundles  11–63 Dual AFM cantilivers, MWCNTs  45 (±7) AFM – lateral force mode, MWCNTs  150 (±30%) TEM – direct tension, MWCNTs  Method of calculation Theoretical calculations (SWCNT) 150 Molecular dynamics simulation  93–112 Molecular mechanics simulation Fig. 15. PoissonÕs ratio predicted by molecular structural mechanics ofLi and Chou and other methods (Hernandez et al. ; Lu ;Sanchez-Portal et al. ). MWCNTs often leads to the separation and pullout of the individual layers .Table 4 The tensile strength data of carbon nanotubes areExperimental and theoretical results for nanotube axial YoungÕs summarized in Table 5 [53,57–60]. A variety of tech-modulus niques have been adopted for measuring the strengthElastic modulus (TPa) Method of measurement of SWCNT bundles and MWCNTs. Again, there existsExperimental measurements large variability in the measured and predicted strength1.26 (±20%) TEM – thermal vibration of a beam  values.1.28 (±40%) AFM – 1 end clamped 0.81 (±50%) AFM – 2 ends clamped  By employing the molecular structural mechanics, Li0.1–1.0 (±30%) TEM – electrostatic deﬂection  and Chou [61,62] studied the free vibrations of carbon0.27–0.95 Dual AFM cantilevers  nanotubes and predicted that the fundamental frequen-0.91 (±20%) TEM – direct tension  cies of cantilevered or bridged SWCNTs as nanomechan- Method of calculation ical resonators could reach the level of 10 GHz–1.5 THz.Theoretical calculations The eﬀects of tube diameter, length, and end constraints0.97 Empirical lattice dynamics  on the fundamental frequency have been discerned. Fur-1.0 (±15%) Ab initio  thermore, the fundamental frequencies of DWCNTs are1.05 (±5%) Molecular structural mechanics  about 10% lower than those of SWCNTs of the same0.68 Pin-jointed Truss model  outer diameter. The non-coaxial vibration of double- walled nanotubes begins at the third resonant frequency. The potential of carbon nanotubes for mass sensing,Large variability exists in both experimental and analyt- strain sensing and pressure sensing has been exploredical ﬁndings. by Li and Chou [63,64]. To illustrate the load-sharing characteristics ofMWCNTs, the modeling works of Li and Chou  5.3. Carbon nanotube/polymer compositesfor a DWCNT is recapitulated below. To delineate theeﬀect of van der Waals forces in the deformation of Both SWCNTs and MWCNTs have been utilized forMWCNT, a two-layer MWCNT under two diﬀerent reinforcing thermoset polymers (epoxy, polyimide, andtensile loading conditions is examined in their work. phenolic), as well as thermoplastic polymers (polypro-One loading condition assumes that uniform forces are pylene, polystyrene, poly methyl methacrylateapplied only on atoms at the end of the outer layer, (PMMA), nylon 12, and poly ether ether ketonewhile in the other loading condition, uniform forces (PEEK)). Several composite systems are reviewedare applied on atoms at the end of both inner and outer below.layers. In both loading conditions, the atoms at the end Tai et al.  have processed a phenolic-based nano-of the outer layer reach nearly the same displacement. composite using MWCNTs, which were synthesizedBut the displacements of the atoms at the end of the in- through the ﬂoating catalyst chemical vapor depositionner layer show a large diﬀerence in the two loading con- process with tube diameter < 50 nm and length > 10ditions. When the forces are applied only on the outer lm. SEM images of brittle tensile fracture surfaces showlayer, the inner layer exhibits small axial displacements. fairly uniform nanotube distribution and nanotube pull-This means that the force acting on the outer layer is out. Enhancement in YoungÕs modulus and strength duebarely transferred to the inner layer through van der to the addition of nanotubes was reported.Waals interactions. This ﬁnding is consistent with the Gojny and co-workers  fabricated nanocompositesexperimental observation that deformation of consisting of DWCNTs with a high degree of dispersion.
502 E.T. Thostenson et al. / Composites Science and Technology 65 (2005) 491–516The resulting composites showed increase of strength, as-processed 5-wt% nanocomposite ﬁlm showing large-YoungÕs modulus and strain to failure at a nanotube con- scale dispersion and alignment of carbon nanotubes intent of only 0.1 wt%. In addition, the nanocomposites the polymer matrix. The arrow in Fig. 16(a) indicatesshowed signiﬁcantly enhanced fracture toughness as the direction of alignment taken as the principal mate-compared to the unreinforced epoxy. rial direction with a nanotube orientation of 0°. The Ogasawara et al.  reinforced a phenylethyl termi- grey lines perpendicular to the arrow in the TEM micro-nated polyimide with MWCNTs, which are a few hun- graph are artifacts from microtome cutting process. Fig.dred lm in length and 20–100 nm in diameter. The 16(b) shows the distribution of nanotube alignmentcomposites were made by mechanical blending of the from the image analysis. Based on the data, the standardnanotubes in the matrix, melting at 320 °C, and curing deviation of nanotube alignment from the principalat 370 °C under 0.2 MPa pressure. The resulting elastic material direction is less than ±15°.and mechanical properties are shown in Table 6. The The axial elastic properties of the MWCNT/polysty-addition of MWCNTs enhances the tensile YoungÕs rene system have been modeled by Thostenson andmodulus, and reduces the tensile strength and ultimate Chou  using a ‘‘micromechanics’’ approach throughstrain. deﬁning an equivalent eﬀective ﬁber property for the Thostenson and Chou [7,68] have characterized the diameter-sensitive carbon nanotube elastic modulus.nanotube structure and elastic properties of a model To accurately model the elastic properties of the com-composite system of aligned MWCNTs embedded in a posite, the contribution to the overall elastic moduluspolystyrene matrix. In this work, a micro-scale twin- of each nanotube diameter, and the volume fractionscrew extruder was utilized to obtain high shear mixing that tubes of a speciﬁc diameter occupy within thenecessary for disentangling the CVD-grown MWCNTs, composite have been taken into account. Using theand disperse them uniformly in a polystyrene thermo- knowledge of the tube diameter distribution functionsplastic matrix. The polymer melt was then extruded as well as the nanotube density and volume distribu-through a rectangular die and drawn under tension be- tion as functions of tube diameter, the micromechanicsfore solidiﬁcation. The process of extruding the nano- approach identiﬁes the correlation between axialcomposite through the die and subsequent drawing YoungÕs modulus, and the diameter, volume fractionresults in a continuous ribbon of aligned nanocompos- and length of nanotubes of the aligned nanocompositeites. Fig. 16(a) shows a TEM micrograph of the model system.Table 6Properties of CNT/polyimide nanocomposites CNT (wt%) CNT (vol%) Tga (°C) Eb (GPa) ruts (MPa) emax (%) r0.2 (MPa) 0 0 335 2.84 115.6 7.6 69.8 3.3 2.3 339 3.07 99.5 4.0 80.5 7.7 5.4 350 3.28 97.6 3.6 84.614.3 10.3 357 3.90 95.2 2.6 92.6 a Onset temperature of decrease in storage modulus (DMA). b Strain range of 0.5–1.0% (tensile testing).Fig. 16. (a) TEM micrograph of process-induced orientation in nanocomposite ribbons  and (b) image analysis of orientation distribution .
E.T. Thostenson et al. / Composites Science and Technology 65 (2005) 491–516 503 Experiments by Frogley et al.  on silicone-based A recent example of functionalization at nanotubeelastomers reinforced with SWCNTs have shown signif- ends for the preparation of SWCNT-reinforced poly-icant increases in the initial modulus of the composites, mer composites can be found in the work of Senaccompanied by a reduction in the ultimate properties. et al.  and Hamon et al. . In their study ofRaman spectroscopy experiments show a loss of stress the eﬀect of interfacial reaction between SWCNTstransfer to the nanotubes at around 10–20% strain, sug- and the polyurethane matrix, ester-functionalizedgesting the break-down of the eﬀective interface between SWCNTs, SWCNT-COO(CH2)11CH3, were synthe-the phases. On the other hand, the reorientation of the sized and electrospun with polyurethane. Accordingnanotubes under strain in the samples may be responsi- to Sen et al. , the chemical functionalization isble for the initial increase in modulus enhancement un- an eﬀective approach to exfoliate the SWCNT bundlesder strain. and improve the processability of SWCNTs. The ester More recently, experiments on silicone elastomer- form of SWCNTs has been shown to be easily dis-reinforced with SWCNTs have shown signiﬁcant in- persed in organic solvent as both individual nanotubescrease in stiﬀness and strength. However, the relative and small bundles of 2–5 nanotubes. The improvedmagnitudes of the improvement decreased with higher chemical compatibility and dispersion of the function-nanotube volume loading because the composite became alized nanotubes within the polyurethane matrix en-more brittle . ables a signiﬁcant enhancement in the tensile strength and tangent modulus of the composite mem-5.4. Interfaces in carbon nanotube/polymer composites brane fabricated by electrospinning when compared to the pure polymer membrane. In the processing of nanocomposites, carbon nano- Covalent chemical functionalization of the nanotubetubes need to be separated from bundles and dispersed sidewalls has also been achieved. Nanotube walls canuniformly in a polymer matrix for maximizing their be modiﬁed by reactive elements, such as ﬂuorine.contact surface area with the matrix. Modiﬁcation of Fluorine chemically binds to, or functionalizes, thenanotube surfaces, for example, the creation of cova- nanotube walls at room temperature and produces se-lent chemical bonds between nanotubes and the poly- vere modiﬁcation of the nanotubeÕs tubular structuremer matrix, enhances their interactions and gives rise at temperatures of 500–600 °C . Nanotube sidewallsto higher interfacial shear strength than van der Waals can also be functionalized through non-covalent inter-bonds [36,71]. Sinnott  has provided an in-depth action through the irreversible adsorption of a bifunc-review of the chemical functionalization of carbon tional molecule onto the surface of a SWCNT in annanotubes, where the chemical bonds are used to tai- organic solvent [72,76]. Frankland et al.  usedlor the interaction of the nanotube with the other enti- molecular dynamics simulations to show that theties, such as a solvent, a polymer matrix or other nanotube/polymer interfacial shear strength can be en-nanotubes. The two major approaches to functional- hanced by over an order of magnitude with the forma-ization are chemical methods and irradiation with tion of cross-links involving less than 1% of theelectrons or ions. Brief excerpts from Sinnott  nanotube carbon atoms with negligible changes in theare given below, and a few examples concerning mod- elastic modulus.iﬁcation of nanotube/polymer interface in composites In their tailoring the interface of MWCNT/polymerare also given. composites, nanotube chemical modiﬁcation, Eitan The chemical methods involve the attachment of and co-workers  ﬁrst attached carboxylic acid tochemical bonds to either the nanotube ends or sidewalls. the tube surface. This was followed by further reactionsFirst, nanotube caps, which are more reactive because of to attach di-glycidyl ether of bisphenol-A-based epoxidetheir high degree of curvature can react with a strong resin. It is then possible to further react the epoxideacid. The open ends can be stabilized by carboxylic acid functional group to enable better interaction betweenand hydroxide groups. In the work of Liu et al. , the polymer matrix chains of the composite and the sur-sonication of puriﬁed SWCNTs in a mixture of concen- face of the nanotube. Fig. 17 depicts the interface tailor-trated sulfuric and nitric acid cuts the tube into short ing process and the resulting improvement in elasticsegments with open ends, about 300 nm in length, with modulus of unreinforced polycarbonate .carboxylic acid (–COOH) groups covalently attached to Gojny and co-workers [80,81] functionalized thethe openings. The mechanism involves the oxidation of nanotubes with amino acids by heating oxidized nano-the nanotube walls at defect sites. As reviewed by Sin- tubes with an excess triethylenetetramine. Their experi-nott , functionalized short SWCNTs could be made mental results indicate that the introduced functionalsoluble in water or organic solvents. Longer, microme- groups lead to covalent bonding of the nanotube surfaceter-length SWCNTs can be made soluble in organic sol- with the epoxy resin. Upon expansion of matrix cracksvents through ionic functionalization of the carboxylic through heating by the electron beam in the TEM, itacid groups. is seen (Fig. 18) that the functionalized outer layer of
504 E.T. Thostenson et al. / Composites Science and Technology 65 (2005) 491–516Fig. 17. Interface tailoring in polycarbonate/MWCNT composites and the resulting improvement in elastic modulus of unreinforced polycarbonate. weak van der Waals bonds, pull out in a sword- in-sheath mechanism. Chemical functionalization of nanotubes can also be accomplished through irradiation with electrons or ions. Electron irradiation of carbon nanotubes causes their collapse in an anisotropic manner due to the knockout of atoms in the nanotube walls [72,82]. Ion deposition can induce cross-links between nanotubes in the bundle and between shells in MWNTs, which could lead to eﬃ- cient load transfer among the tube layers. Hu et al.  and Ni et al.  approached the mod- iﬁcation of carbon nanotube/polystyrene composites through polyatomic ion beam deposition. Molecular dy- namic simulations have demonstrated the modiﬁcation of a composite of (10,10) SWCNT/polystyrene through the deposition of a beam of C3 Fþ polyatomic ions. 5 Covalent cross-links were induced between otherwise un-functionalized SWCNT and a polystyrene matrix. 5.5. Modeling of transport and constitutive properties The modeling of transport properties is illustrated by the work of Qunaies et al.  for the electrical behav- iors of SWCNT/polyimide composites. The insulating nature of polyimides may cause signiﬁcant accumulation of electrostatic charge on their surface, resulting in localFig. 18. Telescoping fracture of an amino acid functionalized multi- heating and premature degradation to electronic compo-walled nanotube where the outer layer of the nanotube is still nents or space structures. Polyimide reinforced withembedded in the matrix after fracture . SWCNT provides a level of electrical conductivity suﬃ- cient to permit electrostatic discharge as well asthe multi-walled nanotube fractures and remains embed- enhancement in thermal and mechanical properties.ded in the epoxy matrix while the inner shells of the The percolation transition in CP2 polyimide/SWCNTnanotube, which are bound together by the relatively composites was noticed between 0.02 and 0.1 vol% of
E.T. Thostenson et al. / Composites Science and Technology 65 (2005) 491–516 505SWCNT; DC conductivity changed from 3 · 10À17 to SWCNTs is assembled; and twisting the nanoarray1.6 · 10À8 s/cm. The modeling predicts percolation con- (helical angle, 10°) provides an additional load transfercentrations of SWCNT in the vicinity of measured per- mechanism. Next, the nanoarrays are surrounded by acolation threshold (0–0.2 vol%) for diﬀerent SWCNT polymeric matrix and assembled into a second twistedbundle size. Both analytical and characterization results array, nanowire with nanotube volume fraction oflie between the single tube and seven tube arrangements, 54.6%. Lastly, the nanowires are impregnated with aindicating that SWCNTs are dispersed in CP2 polyimide polymer matrix and assembled into the ﬁnal helical ar-matrix as very thin bundles. ray, micro-ﬁber with nanotube volume fraction of In the constitutive modeling of SWCNT/polymer 33.2%. The self-similar scheme, ranging from individualcomposites [86–88], constitutive relationships for nano- SWCNTs, nanoarray, nanowire to nanoﬁber is shown incomposites were developed as a function of the molecu- Fig. 19. The self-similar approach and the equivalent-lar structure of the polymer and nanotubes, and continuum modeling can predict elastic properties ofpolymer/nanotube interface. In this approach, the the SWCNT/polymer composites in a combined rangemolecular dynamics simulation was ﬁrst used to obtain spanning from dilute to hyper-concentrated SWCNTthe equilibrium molecular structure, which consisted of volume fraction.a (6,6) SWCNT and ﬁve PmPV [poly(m-phenylenevinyl-ene) substituted with octyloxy chains] oligomer, each 10repeating units in length. Then a suitable representativevolume element (RVE) of the nano-structure material ischosen. An equivalent-truss model of the RVE is devel-oped as an intermediate step. The total strain energies inthe molecular and equivalent-truss models, under identi-cal loading conditions, are set equal. The nanotube, thelocal polymer near the nanotube, and the nanotube/polymer interface are then modeled as an eﬀective con-tinuum ﬁber by using an equivalent-continuum model-ing. The eﬀective ﬁbers then serve as a means forincorporating micromechanical analyses for the predic-tion of bulk mechanical properties of SWCNT/polymercomposites as functions of nanotube lengths, concentra-tions and orientations . Pipes and Hubert [89,90] and Odegard et al. adopted a self-similar approach to constitutive proper-ties of SWCNT/polymer composites. There are three Fig. 19. Self-similar approach to constitutive properties of SWCNT/major steps in the modeling. First, a helical array of polymer .Fig. 20. (a) The photoelastic stress pattern of a short ﬁber/polymer composite  and (b) the shear stress distribution of a nanotube/polymercomposite under axial tension using a multi-scale simulation .
506 E.T. Thostenson et al. / Composites Science and Technology 65 (2005) 491–5166. Comparison of properties and performance of nanoplatelets are smaller than those of traditional composites by about three orders in magnitude, there Although the length scales of reinforcements in nano- are considerable diﬀerences and similarities in the prop-composites, i.e., diameter of nanoparticles, diameter and erties and performance of nanocomposites and tradi-length of nanoﬁbers and nanotubes, as well as thickness tional ﬁber composites. It is highly desirable to presentFig. 21. (a) Slip band formation in a composite with Al2O3 ﬁber in an aluminum matrix  and (b) nanoscale buckling of carbon nanotubes in analigned nanocomposite .Fig. 22. Key mechanisms of energy dissipation have been identiﬁed in the fracture of short as well as continuous ﬁber-reinforced composites . Fig. 23. Fracture mechanisms in carbon nanotube-reinforced composites .
E.T. Thostenson et al. / Composites Science and Technology 65 (2005) 491–516 507some key properties of nanocomposites in contrast to ture, ﬁber pullout, ﬁber/matrix debonding/crackthe same properties in traditional composites, only a bridging and matrix cracking. Fig. 22 shows schemati-few examples are selected below. cally these mechanisms operating at a crack tip  as It is well known in short-ﬁber composites that the well as the micrographs demonstrating the individualpresence of ﬁber-ends induces stress concentrations in failure modes. It is interesting to note that all these fail-the matrix materials when the composite is subjected ure modes have also been observed in nanotube rein-to loading. The nature of stress concentration and the forced polymer composites as demonstrated byassociated singularities have been studied in terms of Thostenson and Chou  in Fig. 23.the ﬁber bundle-end shape and bundle aspect ratio One of the reasons of the recent enthusiasm toward. Fig. 20 compares (a) the photoelastic stress pattern carbon nanotubes as reinforcements for composite mate-of a short ﬁber/polymer composite  with the shear rials is their reported high elastic modulus and strengthstress distribution of a nanotube/polymer composite un- comparing to those of existing continuous ﬁbers. Theder tension using a multi-scale simulation . Two comparison is further demonstrated below. Fig. 24types of nanotube/matrix interfacial bonding conditionsare considered. The length of nanotube is less than 1 nmwhereas the short ﬁbers are of millimeters in length. Thegeneral similarity in local stress concentration isunmistakable. In the case of uniaxial compression of continuous ﬁ-ber composite, it is well known that ﬁber defects or ﬁbermisalignment may activate ﬁber bending and subsequentﬁber buckling. Fig. 21(a) shows slip band formation in acomposite with Al2O3 ﬁber in an aluminum matrix .The activation and subsequent observation of nanotubebuckling in a composite under compressive loading aremuch more diﬃcult because of the small size and align-ment of nanotubes. Thostenson and Chou  have suc-ceeded in such experiments. The multiple buckling ofindividual MWCNTS in a polymer composite can beseen in Fig. 21(b). An atomistic modeling of elasticbuckling of carbon nanotubes has been performed byLi and Chou . Several key mechanisms of energy dissipation havebeen identiﬁed in the fracture of short as well as contin- Fig. 24. Tensile modulus and strength of several major commercialuous ﬁber-reinforced composites. These are ﬁber frac- ﬁbers (data from Shindo ).Fig. 25. Comparison of SWCNT properties with the properties of commercial ﬁbers (Hernandez et al. ; Li and Chou ; Zhang et al. ; Yuet al. ; Yakobson et al. ).
508 E.T. Thostenson et al. / Composites Science and Technology 65 (2005) 491–516Fig. 26. Comparison of electrical resistivity and thermal conductivity . Shaded region indicates the predicted carbon nanotube properties from[101–104].summarizes the tensile modulus and strength of several 7. Critical issues in nanocompositesmajor commercial ﬁbers: PAN-based and pitched-basedcarbon ﬁbers, Kevlar ﬁber and glass ﬁber . A compar- Just as in traditional ﬁber composites, the major chal-ison of the ﬁber properties with those of SWCNTs is lenges in the research of nanocomposites can be catego-made in Fig. 25. Here, the experimental results of Yu rized in terms of the structures from nano to micro toet al.  are presented along with the modulus predic- macro levels. There is still considerable uncertainty intions of Hernandez et al.  using the tight-binding theoretical modeling and experimental characterizationmolecular dynamics, Li and Chou  using the molecu- of the nano-scale reinforcement materials, particularlylar structural mechanics, and Zhang et al.  using a nanotubes. Then, there is a lack of understanding ofcontinuum approach. Because of the lack of correspond- the interfacial bonding between the reinforcements anding strength data, these results of modulus predictions the matrix material from both analytical and experimen-are indicated by horizontal lines. The range of measured tal viewpoints. Lastly, the challenges at the level ofstrength data of Yu et al., denoted by the open circles, is nanocomposites have mainly to do with the following is-indicated by two vertical lines. In addition, the strength sues related to composites processing:predictions of Yakobson et al.  using moleculardynamics are indicated by a vertical line at 150 GPa. 7.1. Dispersion Lastly, Fig. 26 shows the electrical resistivity and ther-mal conductivity of several bulk metallic materials (Cu, Uniform dispersion of nanoparticles, and nanotubesAl, Mg, brass, bronze, and Ti) as well as pitch carbon ﬁ- against their agglomeration due to van der Waals bond-bers (ranging from P-25 to P-140), PAN carbon ﬁbers ing is the ﬁrst step in the processing of nanocomposites.(T-300), and highly oriented pyrolytic graphite (HOPG) Beside the problems of agglomeration of nanoparticles,are given. The ranges of predicted electrical resistivity exfoliation of clays and graphitic layers are essential.and thermal conductivity for SWCNTs are shown as the SWCNTs tend to cluster into ropes and MWCNTs pro-rectangular area in the top of Fig. 26 [100–104]. Because duced by chemical vapor deposition are often tangledof the scattering in data of predicted electrical and thermal together like spaghettis. The separation of nanotubesproperties of SWCNTs, the results shown in Fig. 26 are in a solvent or a matrix material is a prerequisite forunderstood not to be inclusive. aligning them. The addition of carbon nanotubes has been shown tosigniﬁcantly reduce the resistivity and percolation 7.2. Alignmentthreshold in both polymer matrix materials, such asepoxy [105,106] and poly(butylene terephthalate)  Because of their small sizes, it is exceedingly diﬃcultas well as ceramic matrix materials, such as silicon car- to align the nanotubes in a polymeric matrix material inbide  and aluminum oxide . a manner accomplished in traditional short ﬁber com-
E.T. Thostenson et al. / Composites Science and Technology 65 (2005) 491–516 509posites. The lack of control of their orientation dimin- Zhu et al.  reported the direct synthesis of longishes the eﬀectiveness of nanotube reinforcement in strands of ordered SWCNTs by an optimized catalyticcomposites, whether for structural or functional chemical vapor deposition with a ﬂoating catalyst methodperformance. in a vertical furnace. A salient feature of this synthesis method is in the use of n-hexane in combination with thiophene and hydrogen, so large portions of long7.3. Volume and rate SWCNTs are formed and assembeled into macroscopic strands. SWCNT strands with a length of 20 cm and a High volume and high rate fabrication is fundamen- diameter on the order of 0.3–0.5 mm have been accom-tal to manufacturing of nanocomposites as a commer- plished. The tensile modulus of the strands is in thecially viable product. The lessons learned in the range of 49–77 GPa. The approximate volume fractionfabrication of traditional ﬁber composites have clearly of nanotube in the strands, determined by analyzingdemonstrated that the development of a science base the spacing between the nanotube ropes in the strands,for manufacturing is indispensable. Eﬃciency in manu- is less than 48%. Hence, the tensile modulus based uponfacturing is pivotal to the future development of the net cross-section of the strand would be in the rangenanocomposites. of 100–150 GPa, consistent with the modulus of large SWCNT bundles.7.4. Cost eﬀectiveness Ericson et al.  recently reported the synthesis of macroscopic, neat SWCNT ﬁbers. Because of the high- Besides high volume and high rate production, the cost temperature stability of SWCNTs, wet spinning is theof nanocomposites also hinges on that of the nano- only viable approach, as is the case for conventionalreinforcement material, particularly, nanotubes. It is rod-like polymers such a PBO, PPTA and PBZT. Theanticipated that as applications for nanotubes and their main challenge to the production of neat SWCNT ﬁ-composites increase the cost will be dramatically reduced. bers, according to Ericson et al. is dispersing the SWCNTs at high enough concentrations suitable for eﬃcient alignment and coagulation. Davis et al. 8. Precursors for macroscopic composites have shown that SWCNTs can be dispersed at high con- centrations in superacids. The functionalization of For substantially realizing their exciting potentials, ro- SWCNT sidewalls eliminated wall–wall van der Waalsbust processing and manufacturing methods are required interactions and promotes their ordering into an alignedto incorporate nano-reinforcements into macroscopic phase of individual mobile SWCNTs surrounded byfunctional and structural composites. To utilize these acid anions. Then, using conventional ﬁber-spinningnanostructured materials in engineering applications, it techniques, this ordered SWCNT dispersion could beis crucial to develop processing techniques that are both extruded and coagulated in a controlled fashion to pro-scalable for producing macroscopic structures and capa- duce continuous lengths of macroscopic neat SWCNTble of eﬃciently utilizing nanoscale reinforcement in the ﬁbers. Davis et al.  reported that the neat SWCNTas-manufactured composite. Some promising techniques ﬁbers possess good mechanical properties, with afor processing precursors for macroscopic composites are YoungÕs modulus of 120 ± 10 GPa and a tensile strengthbrieﬂy outlined in the following. of 116 ± 10 MPa. The electrical resistivity of the ﬁbers is around 0.2 mX cm. The thermal conductivity of the8.1. Long nanotube ﬁbers and strands either-coagulated ﬁber is 21 W/m K. The process reviewed above for assembling carbon Just as in traditional ﬁber composites, there has been nanotubes into continuous ﬁbers [111–113] have beenstrong impetus in producing ‘‘continuous’’ carbon nano- achieved through post-processing methods. Li et al.tubes that will undoubtedly facilitate their structural and  have reported the spinning of ﬁbers and ribbons offunctional applications. Recent advances in processing carbon nanotubes directly from the chemical vapor depo-long carbon nanotubes in the form of individual nano- sition synthesis zone of a furnace using a liquid source oftubes and nanotube strands are brieﬂy reviewed below. carbon and an iron nanocatalyst. The liquid feedstock is First, Zheng et al.  have reported the synthesis of mixed with hydrogen and injected into the hot zone where4-cm-long individual SWCNTs by Fe-catalyzed decom- an aerogel of nanotubes forms. This aerogel is capturedposition of ethanol. In this process, a FeCl3 solution was and wound out of the hot zone continuously as a ﬁberapplied with a dip-pen to one end of the Si substrate, or ﬁlm (Fig. 27(a)). The wind-up assembly could operatewhich was then placed in a horizontal quartz tube fur- at lower temperature outside the furnace hot zone (Fig.nace with the catalyst end directed toward the gas ﬂow, 27(b)). Fig. 27(c) shows that a permanent twist is intro-and ethanol vapor was then introduced into the furnace. duced into a ﬁber that consists of well-aligned MWCNTsThe growth rate is 11 lm/s. after its removal from the furnace.
510 E.T. Thostenson et al. / Composites Science and Technology 65 (2005) 491–516 1.8 GPa, respectively. Large strain to failure was also recorded. The concept of assembling SWCNTs into nano arrays, nanowires and ﬁnally, microﬁbers for modeling the con- stitutive properties of SWCNT/polymer composites [89–91] may have signiﬁcant technological implications. If SWCNTs can be produced and controlled in a contin- uous manner, the resulting microﬁber would represent the most eﬃcient translation of the properties of nano- reinforcements to the microscopic and macroscopic level. 8.2. Multi-scale hybrid composites 8.2.1. Controlled growth of carbon nanotubes on ﬁber surfaces The controlled surface growth of carbon nanotube is best illustrated by the work of Thostenson et al.  using carbon ﬁbers. Fig. 1 demonstrates the reinforcement hierarchy in the multi-scale nanotube/ﬁber composites in which the woven fabric is deposited with nanotubes. Here, the length scales of structures encompass the mac- roscopic woven fabric (in meters), the microscopic car- bon ﬁbers (in microns) with nanotube growth, and individual nanotubes (in nanometers). It was concluded that carbon nanotubes at the ﬁber/matrix interface im- proved the interfacial shear strength because of local stiﬀening of the polymer matrix.Fig. 27. Schematic of the direct spinning process where: (a) the wind-up is by an oﬀset rotating spindle, (b) the wind-up assembly thatoperates at a lower temperature, outside the furnace hot zone and 8.2.2. Nanoclay-enhanced matrix(c) SEM micrograph showing permanent twist introduced into a Multi-scale hybrid composites have also been pro-nanotube ﬁber from direct CVD spinning . Reprinted with duced using nanoclay as reinforcement for the matrixpermission from . Copyright (2004) AAAS. material. The motivation of adding nanoclay to a resin matrix is for enhancing the resin stiﬀness. The beneﬁts The approach of coagulation-based carbon nanotube of such improvement have been demonstrated in thespinning was ﬁrst developed by Vigolo et al. . In compressive strength of ﬁber composites, which is inﬂu-this process, SWCNTs dispersed by a surfactant were in- enced by the matrix shear modulus. Subramaniyan andjected through a needle or capillary tube using a syringe Sun  measured the oﬀ-axis compressive strength ofpump into co-ﬂowing stream of polymer solution that S2 glass ﬁber (Vf = 35%) reinforced vinyl ester resin withcontained 5 wt% of polyvinyl alcohol. Flow-induced the addition of nanoclay particles. The improvement inalignment of the nanotubes took place at the tip of thecapillary. The nanotube ribbons so formed could bedrawn in the third dimension when the polymer solutionwas slowly pumped out from the bottom of the con-tainer. The needle or the capillary tube was orientedso that the SWCNT injection was tangential to the cir-cular trajectory of the polymer solution in the container.Vigolo et al.  reported that the elastic modulus ofthe nanotube ﬁbers is 10 times higher than the modulusof high-quality buckypaper. By modifying the process of Vigolo et al. [115,116]and Poulin et al. , Dalton et al.  reported theconversion of gel ﬁbers into long solid nanotube com-posite ﬁber in a continuous process at a rate of morethan 70 cm/min. The resulting composite ﬁbers are50 lm in diameter, containing 60% SWCNTs. The ﬁber Fig. 28. Inﬂuence of nanoclay on the static compressive strength of S2YoungÕs modulus and tensile strength are 80 and glass/vinyl ester composites .