北京化工大学：《材料导论》课程阅读材料（复合材料）Polymer_nanotechnology-_Nanocomposites.pdf_大学文库

Feature Article Polymer nanotechnology: Nanocomposites D.R. Paul a,1 , L.M. Robeson b,* aDepartment of Chemical Engineering and Texas Materials Institute, University of Texas at Austin, Austin, TX 78712, United States b Lehigh University, 1801 Mill Creek Road, Macungie, PA 18062, United States article info Article history: Received 19 February 2008 Received in revised form 2 April 2008 Accepted 4 April 2008 Available online xxx Keywords: Nanotechnology Nanocomposites Exfoliated clay abstract In the large field of nanotechnology, polymer matrix based nanocomposites have become a prominent area of current research and development. Exfoliated clay-based nanocomposites have dominated the polymer literature but there are a large number of other significant areas of current and emerging interest. This review will detail the technology involved with exfoliated clay-based nanocomposites and also include other important areas including barrier properties, flammability resistance, biomedical applications, electrical/electronic/optoelectronic applications and fuel cell interests. The important question of the ‘‘nano-effect’’ of nanoparticle or fiber inclusion relative to their larger scale counterparts is addressed relative to crystallization and glass transition behavior. Of course, other polymer (and composite)-based properties derive benefits from nanoscale filler or fiber addition and these are addressed. 2008 Elsevier Ltd. All rights reserved. 1. Introduction The field of nanotechnology is one of the most popular areas for current research and development in basically all technical disciplines. This obviously includes polymer science and technology and even in this field the investigations cover a broad range of topics. This would include microelectronics (which could now be referred to as nanoelectronics) as the critical dimension scale for modern devices is now below 100 nm. Other areas include polymer-based biomaterials, nanoparticle drug delivery, miniemulsion particles, fuel cell electrode polymer bound catalysts, layer-by-layer self-assembled polymer films, electrospun nanofibers, imprint lithography, polymer blends and nanocomposites. Even in the field of nanocomposites, many diverse topics exist including composite reinforcement, barrier properties, flame resistance, electro-optical properties, cosmetic applications, bactericidal properties. Nanotechnology is not new to polymer science as prior studies before the age of nanotechnology involved nanoscale dimensions but were not specifically referred to as nanotechnology until recently. Phase separated polymer blends often achieve nanoscale phase dimensions; block copolymer domain morphology is usually at the nanoscale level; asymmetric membranes often have nanoscale void structure, miniemulsion particles are below 100 nm; and interfacial phenomena in blends and composites involve nanoscale dimensions. Even with nanocomposites, carbon black reinforcement of elastomers, colloidal silica modification and even naturally occurring fiber (e.g., asbestos-nanoscale fiber diameter) reinforcement are subjects that have been investigated for decades. Almost lost in the present nanocomposite discussions are the organic–inorganic nanocomposites based on sol–gel chemistry which have been investigated for several decades [1–3]. In essence, the nanoscale of dimensions is the transition zone between the macrolevel and the molecular level. Recent interest in polymer matrix based nanocomposites has emerged initially with interesting observations involving exfoliated clay and more recent studies with carbon nanotubes, carbon nanofibers, exfoliated graphite (graphene), nanocrystalline metals and a host of additional nanoscale inorganic filler or fiber modifications. This review will discuss polymer matrix based nanocomposites with exfoliated clay being one of the key modifications. While the reinforcement aspects of nanocomposites are the primary area of interest, a number of other properties and potential applications are important including barrier properties, flammability resistance, electrical/electronic properties, membrane properties, polymer blend compatibilization. An important consideration in this review involves the comparison of properties of nanoscale dimensions relative to larger scale dimensions. The synergistic advantage of nanoscale dimensions (‘‘nano-effect’’) relative to larger scale modification is an important consideration. Understanding the property changes as the particle (or fiber) dimensions decrease to the nanoscale level is important to optimize the resultant nanocomposite. As will be noted, many nanocomposite systems noted in the literature can still be modeled using continuum models where absolute size is not important since only shape and volume fraction * Corresponding author. Tel.: þ1 610 481 0117. E-mail addresses: drp@che.utexas.edu (D.R. Paul), lesrob2@verizon.net (L.M. Robeson). 1 Tel.: þ1 512 471 5392. Contents lists available at ScienceDirect Polymer journal homepage: www.elsevier.com/locate/polymer ARTICLE IN PRESS 0032-3861/$ – see front matter 2008 Elsevier Ltd. All rights reserved. doi:10.1016/j.polymer.2008.04.017 Polymer xxx (2008) 1–18 Please cite this article in press as: Paul DR, Robeson LM, Polymer (2008), doi:10.1016/j.polymer.2008.04.017

loading are necessary to predict properties. Nanoscale is considered where the dimensions of the particle, platelet or fiber modification are in the range of 1–100 nm. With the platelet or fiber, the smallest dimension is considered for that range (platelet thickness or fiber diameter). 2. Fundamental considerations In the area of nanotechnology, polymer matrix based nanocomposites have generated a significant amount of attention in the recent literature. This area emerged with the recognition that exfoliated clays could yield significant mechanical property advantages as a modification of polymeric systems [4–6]. The achieved results were at least initially viewed as unexpected (‘‘nano-effect’’) offering improved properties over that expected from continuum mechanics predictions. More recent results have, however, indicated that while the property profile is interesting, the clay-based nanocomposites often obey continuum mechanics predictions. There are situations where nanocomposites can exhibit properties not expected with larger scale particulate reinforcements. It is now well-recognized that the crystallization rate and degree of crystallinity can be influenced by crystallization in confined spaces. In these cases, the dimensions available for spherulitic growth are confined such that primary nuclei are not present for heterogeneous crystallization and homogeneous nucleation thus results. This results in the value of n in the Avrami equation approaching one and often leads to reduced crystallization rate, degree of crystallinity and melting point. This has been observed in phase separated block copolymers [7,8] and has also been observed in polymer blends [9]. Confined crystallization of linear polyethylene in nanoporous alumina showed homogeneous nucleation with pore diameters of 62–110 nm but heterogeneous nucleation for 15–48 nm pores [10]. Linear polyethylene [11] and syndiotactic polystyrene [12] in nanoporous alumina both showed decreased crystallinity versus bulk crystallization. With nanoparticle incorporation in a polymer matrix, similarities to confined crystallinity (as noted above for crystallization in nanopores) exist as well as nucleation effects and disruption of attainable spherulite size. With inorganic particle and nanoparticle inclusions, nucleation of crystallization can occur. At the nanodimension scale, the nanoparticle can substitute for the absence of primary nuclei thus competing with the confined crystallization. At higher nanoparticle content, the increased viscosity (decreased chain diffusion rate) can lead to decreased crystallization kinetics. Thus, the crystallization process is complex and influenced by several competing factors. Nucleation of crystallization (at low levels of addition) evidenced by the onset temperature of crystallization (Tc) and crystallization half-time has been observed in various nanocomposites (poly- (3-caprolactone)–nanoclay [13], polyamide 66–nanoclay [14,15], polylactide–nanoclay [16], polyamide 6–nanoclay [17], polyamide 66–multi-walled carbon nanotube [18], polyester–nanoclay [19], poly(butylene terephthalate)–nanoclay [20], polypropylene–nanoclay (sepiolite) [21], polypropylene/multi-walled carbon nanotube [22]). At higher levels of nanoparticle addition, retardation of the crystallization rate has been observed even in those systems where nucleation was observed at low levels of nanoparticle incorporation [15,18,20,22–24]. The higher level of nanoparticle inclusion was noted to yield retardation of crystallization due to diffusion constraints. This was also apparent in a study where unmodified and organically modified clay were incorporated in maleic anhydride grafted polypropylene [25]. Nucleation was observed with unmodified clay, whereas the exfoliated clay yielded a reduced crystallization rate. A recent review of the crystallization behavior of layered silicate clay nanocomposites noted that while nucleation is observed in many systems the overall crystallization rate is generally reduced particularly at higher levels of nanoclay addition [26]. Another ‘‘nano-effect’’ noted in the literature has been the change in the Tg of the polymer matrix with the addition of nanosized particles. Both increases and decreases in the Tg have been reported dependant upon the interaction between the matrix and the particle. In essence, if the addition of a particle to an amorphous polymer leads to a change in the Tg, the resultant effect on the composite properties would be considered a ‘‘nano-effect’’ and not predictable employing continuum mechanics relationships unless the Tg changes were properly accounted for or were quite minor. The glass transition of a polymer will be affected by its environment when the chain is within several nanometers of another phase. An extreme case of this is where the other environment is air (or vacuum). It has been well-recognized in the literature that the Tg of a polymer at the air–polymer surface or thin films (<100 nm) may be lower than that in bulk [27]. This can also be considered a con- finement effect. A specific experimental example was reported where poly(2-vinyl pyridine) showed an increase in Tg, poly(methyl methacrylate) (PMMA) showed a decrease in Tg and polystyrene showed no change with silica nanosphere incorporation. These differences were ascribed to surface wetting [28]. The Tg decrease for PMMA was ascribed to free volume existing at the polymer surface interface due to poor wetting. In most literature examples where Tg values have been obtained, usually only modest changes are reported (<10 C) as noted in various examples tabulated in Table 1. In some cases the organic modification of clay can result in a decrease in Tg due to plasticization [29]. It should be noted that the values noted in Table 1 involved relatively low levels of nanoparticle incorporation (<0.10 wt fraction and even lower volume fraction) and larger changes in Tg could be expected at much higher volume fraction loadings. For crosslinked polymers, another consideration is necessary as the presence of nanoparticles could yield a crosslink density change over the unmodified composite. This could be due to preferential interactions of the crosslinking agent with the nanoparticle surface or interruption of the crosslink density due to confinement effects. A theoretical model has been Table 1 Glass transition changes with nanofiller incorporation Polymer Nanofiller Tg change (C) Reference Polystyrene SWCNT 3 [34] Polycarbonate SiC (0.5–1.5 wt%) (20–60 nm particles) No change [35] Poly(vinyl chloride) Exfoliated clay (MMT) (<10 wt%) 1 to 3 [36] Poly(dimethyl siloxane) Silica (2–3 nm) 10 [37] Poly(propylene carbonate) Nanoclay (4 wt%) 13 [38] Poly(methyl methacrylate) Nanoclay (2.5–15.1 wt%) 4–13 [39] Polyimide MWCNT (0.25–6.98 wt%) 4 to 8 [40] Polystyrene Nanoclay (5 wt%) 6.7 [41] Natural rubber Nanoclay (5 wt%) 3 [42] Poly(butylene terephthalate) Mica (3 wt%) 6 [43] Polylactide Nanoclay (3 wt%) 1 to 4 [29] SWCNT ¼single-walled carbon nanotubes; MMT ¼ montmorillonite; MWCNT ¼ multi-walled carbon nanotubes. 2 D.R. Paul, L.M. Robeson / Polymer xxx (2008) 1–18 ARTICLE IN PRESS Please cite this article in press as: Paul DR, Robeson LM, Polymer (2008), doi:10.1016/j.polymer.2008.04.017

developed to predict the glass transition temperature of nanocomposites [30]. The model predicts both increases and decreases in Tg dependant upon specific interactions and shows good agreement with the experimental data noted above [28]. A situation does exist where significant increases in the glass transition temperature have been noted involving polyhedral oligomeric silsesquioxane (POSS) cage structures chemically reacted into the polymeric network [31–34]. These cage structures with a particle diameter in the range of 1–3 nm can be functionalized to provide chemical reactivity with various polymer systems. Examples include octavinyl (R ¼ vinyl group) incorporation for copolymerization with PMMA [31], amine groups for incorporation into polyamides [32] and polyimides [33]. This parallels the glass transition increase often noted in the sol–gel inorganic–organic networks. While the glass transition temperature and crystallinity are the major property changes of interest of the nanocomposite polymer matrix, other ‘‘nano-effects’’ or property improvements over larger scale dimensions can be observed. Disruption of packing of rigid chain polymers resulting in higher free volume has been observed in permeability studies [44], surface area effects in photovoltaic applications involving conjugated polymers, surface area effects for catalysts incorporated in polymers, polymer chain dimensions where the radius of gyration is greater than the distance between adjacent nanoparticles, optical properties, nanofiber scaffolds for tissue engineering are additional areas. The ‘‘aging’’ of polymers is a thickness dependant property with rapid change at nanoscale dimensions [45,46]. This property is due to the ability of free volume to diffuse out of the sample and the diffusion coefficient (although very low) becomes important in the time scale associated with polymer utility (days to years) at nanoscale thicknesses. Surface area effects including catalysts, bioactivity, often require nanolevel dimensions to achieve optimum performance. This review of polymer matrix based nanocomposites is divided into two major sections: clay-based nanocomposites with emphasis on mechanical reinforcement and other property modifications. Mechanical enhancement is usually associated with polymer-based composites, however, a number of other areas have emerged where additional property enhancements can be realized by incorporation of nanoscale particles, platelets or fibers. 3. Clay-based polymer nanocomposites 3.1. Structure of montmorillonite The clay known as montmorillonite consists of platelets with an inner octahedral layer sandwiched between two silicate tetrahedral layers [47] as illustrated in Fig. 1. The octahedral layer may be thought of as an aluminum oxide sheet where some of the aluminum atoms have been replaced with magnesium; the difference in valences of Al and Mg creates negative charges distributed within the plane of the platelets that are balanced by positive counterions, typically sodium ions, located between the platelets or in the galleries as shown in Fig. 1. In its natural state, this clay exists as stacks of many platelets. Hydration of the sodium ions causes the galleries to expand and the clay to swell; indeed, these platelets can be fully dispersed in water. The sodium ions can be exchanged with organic cations, such as those from an ammonium salt, to form an organoclay [48–57]. The ammonium cation may have hydrocarbon tails and other groups attached and is referred to as a ‘‘surfactant’’ owing to its amphiphilic nature. The extent of the negative charge of the clay is characterized by the cation exchange capacity, i.e., CEC. The X-ray d-spacing of completely dry sodium montmorillonite is 0.96 nm while the platelet itself is about 0.94 nm thick [47,58]. When the sodium is replaced with much larger organic surfactants, the gallery expands and the X-ray d-spacing may increase by as POSS Si O Si O Si O Si O Si O O O O Si O O Si O Si O R R R R R R R R R = alkyl, aryl, cycloaliphatic, vinyl, amino, nitrile halogen. alcohol, ester, isocyanate, glycidyl etc. Tetrahedral sheet Na+ 2:1 Layer Octahedral sheet Tetrahedral sheet Interlayer or gallery Fig. 1. Structure of sodium montmorillonite. Courtesy of Southern Clay Products, Inc. D.R. Paul, L.M. Robeson / Polymer xxx (2008) 1–18 3 ARTICLE IN PRESS Please cite this article in press as: Paul DR, Robeson LM, Polymer (2008), doi:10.1016/j.polymer.2008.04.017

much as 2 to 3-fold [59,60]. While the thickness of montmorillonite platelets is a well-defined crystallographic dimension, the lateral dimensions of the platelets are not. They depend on how the platelets grew from solution in the geological process that formed them. Many authors grossly exaggerate the lateral size with dimensions quoted of the order of microns or even tens of microns. A commonly used montmorillonite was accurately characterized recently by depositing platelets on a mica surface from a very dilute suspension and then measuring the lateral dimensions by atomic force microscopy [61]. Since the platelets are not uniform or regular in lateral size or shape, the platelet area, A, was measured and its square-root was normalized by platelet thickness, t, to calculate an ‘‘aspect ratio’’. The distribution of aspect ratios found is shown in Fig. 2. If each platelet were circular with diameter D, then ffiffiffi A p =t would be ffiffiffiffiffiffiffiffiffi p=4 p ðD=tÞ ¼ 0:89ðD=tÞ. Since t is approximately 1 nm, Fig. 2 shows that the most probable lateral dimension is in the range of 100–200 nm. 3.2. Nanocomposite formation: exfoliation Nanocomposites can, in principle, be formed from clays and organoclays in a number of ways including various in situ polymerization [4,6,62–68], solution [51,53], and latex [69,70] methods. However, the greatest interest has involved melt processing [71– 139] because this is generally considered more economical, more flexible for formulation, and involves compounding and fabrication facilities commonly used in commercial practice. For most purposes, complete exfoliation of the clay platelets, i.e., separation of platelets from one another and dispersed individually in the polymer matrix, is the desired goal of the formation process. However, this ideal morphology is frequently not achieved and varying degrees of dispersion are more common. While far from a completely accurate or descriptive nomenclature, the literature commonly refers to three types of morphology: immiscible (conventional or microcomposite), intercalated, and miscible or exfoliated. These are illustrated schematically in Fig. 3 along with example transmission electron microscopic, TEM, images and the expected wide angle X-ray scans [48–53,83]. For the case called ‘‘immiscible’’ in Fig. 3, the organoclay platelets exist in particles comprised of tactoids or aggregates of tactoids more or less as they were in the organoclay powder, i.e., no separation of platelets. Thus, the wide angle X-ray scan of the polymer composite is expected to look essentially the same as that obtained for the organoclay powder; there is no shifting of the Xray d-spacing. Generally, such scans are made over a low range of angles, 2q, such that any peaks from a crystalline polymer matrix are not seen since they occur at higher angles. For completely exfoliated organoclay, no wide angle X-ray peak is expected for the nanocomposite since there is no regular spacing of the platelets and the distances between platelets would, in any case, be larger than what wide angle X-ray scattering can detect. Often X-ray scans of polymer nanocomposites show a peak reminiscent of the organoclay peak but shifted to lower 2q or larger d-spacing. The fact that there is a peak indicates that the platelets are not exfoliated. The peak shift indicates that the gallery has expanded, and it is usually assumed that polymer chains have entered or have been intercalated in the gallery. Placing polymer chains in such a confined space would involve a significant entropy penalty that presumably must be driven by an energetic attraction between the polymer and the organoclay [76–79]. It is possible that the gallery expansion may in some cases be caused by intercalation of oligomers or low molecular weight polymer chains. The early literature seemed to suggest that ‘‘intercalation’’ would be useful and perhaps a precursor to exfoliation. Subsequent research has suggested alternative ideas about how the exfoliation process may occur in melt processing and how the details of the mixing equipment and conditions alter the state of dispersion achieved [54,82,84,140]. These ideas are summarized in the cartoon shown in Fig. 4 [84]. As made commercially, the particles of an organoclay powder are about 8 mm in size and consist of aggregates of tactoids, or stacks of platelets; the stresses imposed during melt mixing break up aggregates and can shear the stack into smaller ones as suggested in Fig. 4. However, there evidently is a limit to how finely the clay can be dispersed just by mechanical forces. If the polymers and organoclay have an ‘‘affinity’’ for one another, the contact between polymers and organoclay can be increased by peeling the platelets from these stacks one by one until, given enough time in the mixing device, all the platelets are individually dispersed as suggested in Fig. 4. This notion is supported by many TEM images at various locations in the extruder and is more plausible than imagining the polymer chains diffusing into the galleries, i.e., intercalation, and eventually pushing them further and further apart until an exfoliated state is reached. The nature of the extruder and the screw configuration are important to achieve good organoclay dispersion [83]. Longer residence times in the extruder favor better dispersion [83]. In some cases, having a higher melt viscosity is helpful in achieving dispersion apparently because of the higher stresses that can be imposed on the clay particles [84,126]; however, this effect is not universally observed. The location of where the organoclay is introduced into the extruder has also been shown to be important [120]. However, no matter how well these process considerations are optimized, it is clear that complete exfoliation, or nearly so, cannot be achieved unless there is a good thermodynamic affinity between the organoclay and the polymer matrix. This affinity can be affected to a very significant extent by optimizing the structure of the surfactant used to form the organoclay [85,88,99,100,109,113,119,141] and possibly certain features of the clay itself like its CEC [115], as this affects the density of surfactant molecules over the silicate surface. A key factor in the polymer–organoclay interaction is the af- finity polymer segments have for the silicate surface [84,85,94,113,141,142]. Nylon 6 appears to have good affinity for the silicate surface, perhaps by hydrogen bonding, and as a result very high levels of exfoliation can be achieved in this matrix provided the processing conditions and melt rheology are properly selected [83,84,120]. Surfactants with a single long alkyl tail give the best exfoliation [141]. As more long chain alkyls are added to the surfactant, the extent of exfoliation is decreased [141]. It has been proposed that at least one alkyl tail is needed to reduce the platelet–platelet cohesion while adding more than one tends to block Fig. 2. Aspect ratio distribution of native sodium montmorillonite platelets [61]. Reproduced with permission of the American Chemical Society. 4 D.R. Paul, L.M. Robeson / Polymer xxx (2008) 1–18 ARTICLE IN PRESS Please cite this article in press as: Paul DR, Robeson LM, Polymer (2008), doi:10.1016/j.polymer.2008.04.017

polyurethanes, and some other polar polymers [84,96,101]. It has been found that a small amount of a polyolefin that has been lightly grafted with maleic anhydride, w1% MA by weight is typical, can act as a very effective ‘‘compatibilizer’’ for dispersing the organoclay in the parent polyolefin [84,101,103,106,117,118,143–148]. This does not lead to the high level of exfoliation that can be achieved in polyamides, but this approach has allowed such nanocomposites to move forward in commercial applications, particularly in automotive parts [54,99,106]. In the case of olefin copolymers with polar monomers like vinyl acetate and methacrylic acid (and corresponding ionomers), the degree of exfoliation that can be achieved progressively improves as the polar monomer content increases [113,119]. In all cases, the best exfoliation is achieved when the structure of the surfactant and the process parameters are optimized. 3.3. Characterization of nanocomposite morphology An important issue is to relate the performance of nanocomposites to their morphological structure; experimental evaluation of performance is certainly easier than characterization of their morphology. Wide angle X-ray scattering, WAXS, is frequently used because such analyses are relatively simple to do. However, such analyses can be misleading and are not quantitative [149– 151]. As indicated in Fig. 3, the organoclay has a characteristic peak indicative of the platelet separation or d-spacing; other peaks may be seen resulting from multiple reflections as predicted by Bragg’s law. The presence of the same peak in the nanocomposite is irrefutable evidence that the nanocomposite contains organoclay tactoids as suggested in Fig. 3. However, the absence of such a peak is not conclusive evidence for a highly exfoliated structure as has been repeatedly pointed out in the literature [151]; many factors must be considered to interpret WAXS scans. If the sensitivity, or counting time, of the scan is low, then an existing peak may not be seen. When the tactoids are internally disordered or not well aligned to one another, the peak intensity will be low and may appear to be completely absent. These issues can be well illustrated by analyses of polyolefin nanocomposites, which are never fully exfoliated, that have been injection molded. X-ray scans of the molded surface reveal a peak indicating the presence of tactoids. However, after milling away the surface of these specimens, subsequent scans of the milled surface in the core of the bar may not reveal a peak because the tactoids are more randomly oriented in the interior than near the as-molded surface [103,118]. However, if a more sensitive scan is made, the peak can usually be seen. In some cases, the WAXS scan may reveal a shift in the peak location relative to that of the neat organoclay. The peak may shift to lower angles, or larger d-spacing, and is generally taken as evidence of ‘‘intercalation’’ of polymers (or perhaps other species) into the galleries [48–53,73,79]. However, an opposite shift may also occur, and this is usually attributed to loss of unbound surfactant from the gallery or to surfactant degradation [89,107]. All of these processes may occur simultaneously rendering uncertainty in the interpretation. In any case, intercalation per se does not seem to be a contributor to develop useful nanocomposite performance. Small angle X-ray scattering, SAXS, can be more informative and somewhat quantitative as explained by numerous authors [17,152– 156]. However, this technique has not been widely used except in a few laboratories probably because most laboratories do not have SAXS facilities or experience in interpreting the results. Other techniques like solid-state NMR and neutron scattering have also been used on a limited basis to explore clay dispersion [95,157– 162]. A far more direct way of visualizing nanocomposite morphology is via transmission electron microscopy, TEM; however, this approach requires considerable skill and patience but can be quantitative. Use of TEM is often criticized because it reveals the morphology in such a small region. However, this can be overcome by taking images at different magnifications and from different locations and orientations until a representative picture of the morphology is established. The major obstacle in obtaining good TEM images is not in the operation of the microscope but in microtoming sections that are thin and uniform enough to reveal the morphology. Fortunately, the elemental composition of the clay compared to that of the polymer matrix is such that no staining is required. When exfoliation is essentially complete, as in the case of nylon 6, one can see the w1 nm thick clay platelets as dark lines when the microtome cut is perpendicular to the platelets. Image analysis can be used to quantify the distribution of platelet lengths, but meaningful statistics require analyzing several hundred particles [58,84,93,119]. However, it must be remembered that the dimensions observed reflect a random cut through an irregular platelet and only rarely will the maximum dimension be seen [163,164]. Thus, the aspect ratio distribution seen in this way will lead to smaller values than true dimensions like those given by Fig. 2. Even for the best nylon 6 nanocomposites, exfoliation is generally never complete and one can see particles consisting of two, three or more platelets [58]. In some cases, these platelets may be skewed relative to one another as suggested in Fig. 5 [93]. Thus, some particles may appear to be longer than the platelets really are. These kinds of issues should be kept in mind when interpreting quantitative analyses of particle aspect ratios and in comparison of observed performance with that predicted by composite theory [58]. Nanocomposites made from polyolefins, styrenics, and other polymers that lead to lower degrees of exfoliation reveal particles much thicker than single clay platelets as expected [101,110,111,117,119]. However, the clay particles are also much longer than the individual clay platelets indicated in Fig. 2. As the Length of the whole particle Length of a single platelet a b Length of the whole particle Length of a single platelet ‘Skewed’ agglomerate 50 nm Fig. 5. Examples of skewed platelets such that particles appear longer than platelets of MMT [93]. Reproduced with permission of Elsevier Ltd. 6 D.R. Paul, L.M. Robeson / Polymer xxx (2008) 1–18 ARTICLE IN PRESS Please cite this article in press as: Paul DR, Robeson LM, Polymer (2008), doi:10.1016/j.polymer.2008.04.017

polymer–organoclay affinity is increased by adding a compatibilizer, e.g., PP-g-MA or PE-g-MA, or increasing the content of a polar comonomer, e.g., vinyl acetate, the clay particles not only become thinner (fewer platelets in the stack) but also become shorter [103,106,117–120]. However, generally the particle thickness decreases more rapidly than the length such that the aspect ratio increases; this generally improves performance. The fact that the particles become shorter does not mean that clay platelets are breaking or being attributed during processing, although, this may occur under some extreme conditions [120]. Instead, considerable evidence indicates that the vision of tactoids as usually drawn, see Figs. 3 and 4, where the platelets are all of the same length and in registry with one another is not correct. Fig. 6 shows a more realistic vision of a tactoid where the particle length can be much longer than individual platelets and how these particles evolve as dispersion improves [106,117]. Complications arise when calculating an average aspect ratio of particles when there is a distribution of both length and thickness. First, one can calculate a number average, a weight average, or other weightings of the distribution [58,113,119]. Second, one can average the aspect ratios or average separately the lengths and thickness and calculate an aspect ratio from these averages [113,119]. There is no theoretical guidance on which is the better predictor of performance or for use in composite modeling [113,117,119]. To take full advantage of the reinforcement or tortuosity clay platelets or particles can provide to mechanical and thermal or barrier properties of nanocomposites, they must be oriented in the appropriate direction and not curled or curved. The alignment of particles is affected by the type of processing used to form the test specimen, e.g., extrusion, injection molding, etc. This is a separate issue from the degree of dispersion or exfoliation which is usually determined in the mixing process. Techniques like compression molding usually do not lead to good alignment or straightening of the high aspect ratio particles, and measurements made on such specimens often underestimate the potential performance. TEM can be used to assess and even quantify particle orientation and curvature and this information can, in principle, be factored into appropriate models to ascertain their effect on performance [86,110,131,134,138]. 3.4. Nanocomposite mechanical properties: reinforcement A common reason for adding fillers to polymers is to increase the modulus or stiffness via reinforcement mechanisms described by theories for composites [58,165–185]. Properly dispersed and aligned clay platelets have proven to be very effective for increasing stiffness. This is illustrated in Fig. 7 by comparing the increase in the tensile modulus, E, of injection molded composites based on nylon 6, relative to the modulus of the neat polyamide matrix, Em, when the filler is an organoclay versus glass fibers [58]. In this example, increasing the modulus by a factor of two relative to that of neat nylon 6 requires approximately three times more mass of glass fi- bers than that of montmorillonite, MMT, platelets. Thus, the nanocomposite has a weight advantage over the conventional glass fiber composite. Furthermore, if the platelets are aligned in the plane of the sample, the same reinforcement should be seen in all directions within the plane, whereas fibers reinforce only along a single axis in the direction of their alignment [165]. In addition, the surface finish of the nanocomposite is much better than that of the glass fiber composite owing to nanometer size of the clay platelets versus the 10–15 m diameter of the glass fibers. A central question is whether the greater efficiency of the clay has anything to do with its nanometric dimensions, i.e., a ‘‘nano-effect’’. To answer this requires considering many issues which we will do later in this section; however, the short answer is that we can explain essentially all of the experimental trends using composite theory without invoking any ‘‘nano-effects’’ [58]. Fig. 8 shows an analogous comparison of nanocomposites based on thermoplastic polyolefin or TPO matrix, polypropylene plus an ethylene-based elastomer, with conventional talc-filled TPO [103]. The latter is widely used in automotive applications; however, in some cases, they are being replaced with TPO nanocomposites. In Fig. 6. A more realistic picture of clay tactoids and how they become shorter as the level of dispersion increases. wt % filler 0 10 20 30 40 E / E m 1 2 3 4 Nanocomposites Glass Fibers Fig. 7. Comparison of modulus reinforcement (relative to matrix polymer) increases for nanocomposites based on MMT versus glass fiber (aspect ratio w20) for a nylon 6 matrix [58]. Reproduced by permission of Elsevier Ltd. Filler Content (wt%) 0 5 10 15 20 25 Relative Modulus 1.0 1.4 1.8 2.2 2.6 3.0 TPO with MMT TPO with Talc Fig. 8. Comparison of modulus reinforcement for nanocomposites based on MMT versus talc for a TPO matrix [103]. Reproduced by permission of Elsevier Ltd. D.R. Paul, L.M. Robeson / Polymer xxx (2008) 1–18 7 ARTICLE IN PRESS Please cite this article in press as: Paul DR, Robeson LM, Polymer (2008), doi:10.1016/j.polymer.2008.04.017

this case, doubling the modulus of the TPO requires more than four times more talc than MMT; this presents a weight, and consequently fuel, savings along with the improved surface finish [103,106]. The polyamide nanocomposites in Fig. 7 are very highly exfoliated, whereas the exfoliation of the clay in the TPO of Fig. 8 is not nearly so perfect [103,106,117,118]. However, it should be recognized that the talc particles do not have as high aspect ratio as the glass fibers used in these comparisons [58,186]. Another factor at play here is the lower modulus of TPO than nylon 6. The lower the matrix modulus, the greater is the relative increase in reinforcement caused by adding a filler [94]. Fig. 9 shows dynamic mechanical moduli of the nylon 6 nanocomposites from Fig. 7 versus temperature. The intersection of these curves with the horizontal line shown is a good approximation to the heat distortion temperature, HDT, of these materials [58]. This temperature, used as a benchmark for many applications, can be increased by approximately 100 C by addition of about 7% by weight of MMT. This effect has been explained by simple reinforcement, as predicted by composite theory, without invoking any special ‘‘nano-effects’’; the effect of MMT on the glass transition of these materials is very slight if any at all [58]. Indeed, glass fibers cause an analogous increase in HDT. Addition of fillers, including clay, can also increase strength as well as modulus [84]; however, the opposite may also occur [99]. A main issue is the level of adhesion of the filler to the matrix. For glass fiber composites, chemical bonding at the interface using silane chemistry is used to achieve high strength composites [186]. On the other hand, the modulus of glass fiber composites is not very much affected by the level of interfacial adhesion [186]. Unfortunately, at this time there is no effective way to measure the level of adhesion of clay particles with polymer matrices. In addition, there are no effective methods at this time to create chemical bonds between clay particles and polymer matrices analogous to those used for glass fibers. Generally, addition of organoclays to ductile polymers increases the yield strength; however, for brittle matrices failure strength is typically decreased [84,99–101]. Addition of fillers generally decreases the ductility of polymers, e.g., elongation at break. For glass fibers, talc, etc. this is well known and expected. Similar trends are also seen for nanocomposites [84,100], but this seems to have been unexpected and disappointing to some working in this field. Impact strength is an energy measurement, i.e., a force acting through a distance. A reduced elongation at break often means a reduced energy to break but there are exceptions to this [84,100,116]. Addition of clay may increase the stress levels via reinforcement more than the reduction in deformation as recently demonstrated for some nanocomposites [116]. Generally speaking, the reduction in ductility or energy to break is more severe when the polymer matrix is below its glass transition, whereas the effects of adding clay may not be so dramatic when the matrix is above its glass transition temperature [82,84,100,119]. This involves a shift in fracture mechanisms that is beyond the scope of this review. Melt rheological properties of polymers can be dramatically altered in the low shear rate or frequency region such that these fluids appear to have a yield stress [84,117,118,187–189]. The effects in the high shear rate region are usually much less dramatic [84]. A great deal has been written about these effects and their causes, and this will not be reviewed here. Interestingly, the addition of clay seems to be an effective way to increase ‘‘melt strength’’ which can be useful in some polymer processing operations like film blowing or blow molding [54,96]. To answer the question of whether the large increase in modulus caused by clay platelets or particles relative to conventional fillers, like that illustrated in Figs. 7 and 8, is due to some ‘‘nanoeffect’’, one must first determine whether effects of this magnitude can be predicted by composite theories. That is, by a ‘‘nano-effect’’, we mean some change in the local properties of the matrix caused by the extremely high surface area filler and the small distances between nanofiller particles even at low mass loadings. It is well known that clay particles are effective nucleating agents which greatly change the crystalline morphology and crystal type for polymers like nylon 6 or PP [87]. Potential ‘‘confinement’’ effects are also discussed in the context of nanocomposites. A basic premise of composite theories is that the matrix and filler have the same properties as when the other component is not there. These theories, thus, only predict the effects of simple reinforcement and do not allow for any ‘‘nano-effects’’ of the type mentioned. Clearly, reinforcement does occur and the issue is whether that alone can explain the observations or not. Composite theories consider only the aspect ratio, orientation and volume fraction of filler in the matrix; the absolute filler particle size does not enter into the calculations. We have already mentioned the difficulties of experimentally determining the aspect ratio (sectioning issues, averaging of distributions, etc.). Furthermore, determining what values to assign to the properties of the clay platelets (like its modulus) is not trivial. Finally, the various composite theories differ somewhat in their predictions owing to the assumptions and simplifications used in their mathematical formulation. These and other issues are worth remembering as we proceed with their analysis using data for nylon 6 nanocomposites. We wish to compare composite calculations with experimental data for the modulus of nylon 6 nanocomposites where the degree of exfoliation is very high but not perfect. An image analysis of many TEM photomicrographs was used to construct a platelet length distribution which looks very similar to that in Fig. 2 [58]; the number average platelet length was found to be 91 nm. While the majority of the clay particles was single platelets (thickness w 0.94 nm), there were some doublets, some triplets, and a few quadruplets. It was estimated, by a rather involved analysis, that the number average platelet thickness was 1.61 nm [58]. Thus, an upper bound on the aspect ratio for perfect exfoliation (using number averages of the distribution) would be about 91/0.94 ¼ 97 while a more realistic estimate might be 91/1.61 ¼57. Next, we need the in-plane modulus of a montmorillonite platelet. Information from a variety of sources suggests that a reasonable value is 178 GPa [58,190]; however, some molecular dynamics calculations suggest significantly larger values [191]. A density for MMT of 2.83 g/cm3 was used to convert weight fractions -40 0 40 80 120 160 200 log E' (Pa) 8.0 8.2 8.4 8.6 8.8 9.0 9.2 9.4 9.6 9.8 10.0 Temperature (°C) HMW PA-6 / (HE)2M1R1 Nanocomposites wt % MMT 0 1.6 3.2 4.6 7.2 8.9 / 1.82 MPa Fig. 9. Experimental storage modulus data versus temperature for nylon 6 nanocomposites. The horizontal line is used to estimate the heat distortion temperature (HDT) at an applied stress of 1.82 MPa or 264 psi [58]. Reproduced by permission of Elsevier Ltd. 8 D.R. Paul, L.M. Robeson / Polymer xxx (2008) 1–18 ARTICLE IN PRESS Please cite this article in press as: Paul DR, Robeson LM, Polymer (2008), doi:10.1016/j.polymer.2008.04.017

to volume fractions. The properties of the matrix (modulus, Poisson ratio, density, etc.) were experimentally measured values [58]. The equations of Halpin–Tsai [168] and Mori–Tanaka [167] are frequently used for composite calculations; the former predicts higher levels of reinforcement for the cases of interest here than the latter as seen in Fig. 10. Interestingly, the predictions via these two theories and the two estimates of aspect ratio give results that bracket the experimental data. Thus, we conclude that simple reinforcement considerations adequately explain the observations given all the issues involved in making these calculations. Any ‘‘nano-effect’’ is relatively minor if at all. More needs to be said about the comparisons shown in Figs. 7 and 8. The aspect ratio of the glass fibers in the nylon 6 matrix is about 20, whereas the aspect ratio of MMT platelets is 3–5 times larger than this. However, calculations using composite theory reveal that the larger aspect ratio of the clay versus glass fibers is not enough to explain all of the large differences in modulus reinforcement shown in Fig. 7. A significant part of the difference in modulus enhancement stems from the much higher modulus of MMT than glass fibers, i.e., 178 versus 72.4 GPa [58]. The comparison of MMT versus talc in a TPO matrix shown in Fig. 8 is more complex to explain, but similar factors are at play [103,117]. 3.5. Nanocomposite thermal properties: dimensional stability The high thermal expansion coefficients of neat plastics causes dimensional changes during molding and as the ambient temperature changes that are either undesirable or in some cases unacceptable for certain applications. The latter is a particular concern for automotive parts where plastics must be integrated with metals which have much lower coefficients of thermal expansion, CTE. Fillers are frequently added to plastics to reduce the CTE. For low aspect ratio filler particles, the reduction in CTE follows, more or less, a simple additive rule and is not very large; in these cases, the linear CTE changes are similar in all three coordinate directions. However, when high aspect ratio fillers, like fibers or platelets, are added and well oriented, the effects can be much larger; in these cases, the CTE in the three coordinate directions may be very different. The fibers or platelets typically have a higher modulus and a lower CTE than the matrix polymer. As the temperature of the composite changes, the matrix tries to extend or contract in its usual way; however, the fibers or platelets resist this change creating opposing stresses in the two phases. When the filler to matrix modulus is large, the restraint to dimensional change can be quite significant within the direction of alignment. Platelets can provide their restraint in two directions, when appropriately oriented, while fibers can only do so in one direction. Because of their shape differences, fibers can cause a greater reduction in the direction of their orientation than platelets can [166,170]. The CTE in the direction normal to the fibers or the platelet plane can actually increase when such fillers are added. Theories based on the mechanisms described above are available for quantitatively predicting CTE behavior [166,170]. Montmorillonite platelets are particularly effective for reducing CTE of plastics as shown in Fig. 11 for well-exfoliated nylon 6 nanocomposites [86]. These data were measured in the flow direction of injection molded bars. When the semicrystalline nylon 6 matrix is above its glass transition temperature, the CTE reduction is greater than when below the Tg. Of course, the neat nylon 6 has a higher CTE above Tg than below; however, because of its lower modulus above Tg, the MMT platelets are more effective for reducing CTE. Note that the two curves in Fig. 11 seem to cross at about 7 wt% MMT. For these specimens, the CTE in the transverse direction is also reduced by adding MMT but not quite as efficiently as in the flow direction since platelet orientation is not as great in the former as the latter direction. The CTE in the normal direction actually increases as MMT is added. These trends are quantitatively predicted by the theories mentioned earlier [86]. CTE behavior is also a major consideration for the TPO materials used in automotive applications [103,106,117]. As seen in Fig. 12, MMT is much more efficient at reducing CTE than talc in these materials [106]. Again, composite theories capture these trends reasonably well [117]. 4. Variations and applications of polymer-based nanocomposites: properties other than reinforcement Polymer composites comprising nanoparticles (including nanofibers where the fiber diameter is in the nanodimension range) are often investigated where reinforcement of the polymer matrix is achieved. While the reinforcement aspects are a major part of the nanocomposite investigations reported in the literature, many other variants and property enhancements are under active study and in some cases commercialization. The advantages of nanoscale particle incorporation can lead to a myriad of application vol % MMT 0 1234 E / E m 1 2 3 Ef = 178 GPa Em = 2.75 GPa HMW / (HE)2M1R1 Expt. Data Mori-Tanaka Halpin-Tsai 97 97 57 ν 57 m = 0.35 νf = 0.20 Fig. 10. Experimental and theoretical stiffness data for nylon 6 nanocomposites; model predictions are based on unidirectional reinforcement of pure MMT having a filler modulus of 178 GPa and aspect ratio of 57 (experimentally determined number average value) and 97, corresponding to complete exfoliation. Note that experimental modulus data are plotted versus vol% MMT since MMT is the reinforcing agent [58]. Reproduced by permission of Elsevier Ltd. Wt % MMT 012345678 Expansion Coefficient (10-5 mm/mm °C) 4 6 8 10 12 14 16 Flow Direction (as-molded) HMW Nylon 6 T > Tg T Tg and T < Tg [86]. Reproduced by permission of Elsevier Ltd. D.R. Paul, L.M. Robeson / Polymer xxx (2008) 1–18 9 ARTICLE IN PRESS Please cite this article in press as: Paul DR, Robeson LM, Polymer (2008), doi:10.1016/j.polymer.2008.04.017

possibilities where the analogous larger scale particle incorporation would not yield the sufficient property profile for utilization. These areas include barrier properties, membrane separation, UV screens, flammability resistance, polymer blend compatibilization, electrical conductivity, impact modification, and biomedical applications. Examples of nanoparticle, nanoplatelet and nanofiber incorporation into polymer matrices are listed in Table 2 along with potential utility where properties other than mechanical property reinforcement are relevant. 4.1. Barrier and membrane separation properties The barrier properties of polymers can be significantly altered by inclusion of inorganic platelets with sufficient aspect ratio to alter the diffusion path of penetrant molecules as illustrated in Fig. 13. Various continuum models have been proposed to predict the permeability of platelet filled composites as listed in Table 3. These models are generally based on random, parallel platelets perpendicular to the permeation direction (random in only two directions). The model by Bharadwaj introduces an orientation factor [196]. At high aspect ratio which can be achieved (such as with exfoliated clay) in nanocomposites, significant decreases in permeability are predicted and observed in practice. Four of these models have been applied to polyisobutylene/vermiculite nanocomposites with aspect ratios predicted in the range of expectations [70]. The variability in the models was, however, shown to be substantial. In many cases, the nanocomposites investigated can be approximated by the continuum models, thus the ‘‘nano-effect’’ is not observed. This should not be surprising as the dimensions of permeating gas molecules are still much lower than the nanodimension modification. Differences would be expected in those cases where the Tg of the matrix polymer is changed. However, for practical applications the nanoscale dimensions are still quite important as transparency can be maintained along with surface smoothness for thin films; critical for food packaging applications. Exfoliated clay modified poly(ethylene terephthalate) (PET) is one of the more prevalent nanocomposites investigated by both academic and industrial laboratories for barrier applications [197– 199]. In situ polymerized PET-exfoliated clay composites were noted to show a 2-fold reduction in permeability with only 1 wt% clay versus the control PET [197]. PET-exfoliated clay composites also prepared via in situ polymerization using a clay-supported catalyst exhibited a 10 to 15-fold reduction in O2 permeability with 1–5 wt% clay [198]. The moisture vapor transmission, however, did not show any significant change. Exfoliated clay modified chlorobutyl rubber showed decreased diffusion for several organic chemicals suggesting utility for chemical protective gloves/clothing [200]. Exfoliated clay added to polyamide 6/polyolefin (polyethylene or polypropylene) blends yielded an improved barrier to styrene permeation for melt blown films [201]. It was noted that the polymer blend nanocomposite was a better barrier than the control polyamide nanocomposite. While most of the papers investigating barrier properties incorporate low levels of exfoliated clay, a novel approach employed producing a self-supporting clay fabric film followed by infiltration with an epoxy resin/amine hardener mixture and polymerization [202]. The resultant semitransparent nanocomposite film contained up to 77% volume fraction clay with an oxygen permeability 2–3 orders of magnitude lower than the control epoxy. The concept of mixed matrix membranes involving molecular sieve inclusions in a polymer film to enhance the permselectivity properties for membrane separation was developed by Koros et al. [203] to address the limits imposed by upper bound limits typically observed with polymer membranes [204]. These inclusions (carbon molecular sieves, zeolite structures) need to be at nanolevel dimensions as the dense layer thickness of commercial membranes is in the range of 100 nm. This approach has shown promise in exceeding the noted upper bound in various studies [205,206]. The addition of silica nanoparticles to poorly packing polymer membranes (specifically poly(4-methyl-2-pentyne)) has been shown to yield even poorer packing and higher free volume [44]. This leads to increased permeability for larger organic molecules and a selectivity reversal for mixtures of these molecules with smaller molecules (e.g., n-butane/methane). This indicates that the separation process has changed from molecular sieving expected at low free volume and small void diameters to surface diffusion as the free volume and void diameters increase. The addition of nanoparticle TiO2 to poly(trimethyl silylpropyne) (also a high free volume polymer with poor chain packing) showed a decrease in gas permeability up to 7 vol% TiO2 with increasing permeability and higher free volume observed above 7 vol% loading [207]. Filler Content (wt%) 0 5 10 15 20 25 Expansion coefficient (10-5 mm/mm °C) 4 6 8 10 12 TPO with MMT TPO with talc Fig. 12. Comparison of linear coefficients of thermal expansion (flow directions) as a function of filler content of TPO composites formed from MMT and talc [106]. Reproduced by permission of Elsevier Ltd. Table 2 Examples of nanoscale filler incorporated in polymer composites for property enhancement other than reinforcement Nanofiller Property enhancement(s) Application/utility Exfoliated clay Flame resistance, barrier, compatibilizer for polymer blends SWCNT; MWCNT Electrical conductivity, charge transport, Electrical/electronics/ optoelectronics Nanosilver Antimicrobial ZnO UV adsorption UV screens Silica Viscosity modification Paint, adhesives CdSe, CdTe Charge transport Photovoltaic cells Graphene Electrical conductivity, barrier, charge transport Electrical/electronic POSS Improved stability, flammability resistance Sensors, LEDs Permeation path imposed by nanoplatelet modification of polymer films Fig. 13. Barrier to permeation imposed by nanoparticles imbedded in a polymeric matrix. 10 D.R. Paul, L.M. Robeson / Polymer xxx (2008) 1–18 ARTICLE IN PRESS Please cite this article in press as: Paul DR, Robeson LM, Polymer (2008), doi:10.1016/j.polymer.2008.04.017