Influence of material and processing parameters on carbon nanotube dispersion in polymer melts
Melt processing of polymer–carbon nanotube (CNT) composites is of great industrial importance for the large-scale production of composite materials and desired products. The properties of these composites greatly depend on the quality of CNT dispersion. In this chapter, a broad overview of the influence of material and processing parameters on the dispersion of multi-walled CNTs in thermoplastic polymer matrices during melt processing is provided. The results discussed here are based on small-scale batch mixers as well as on laboratory-scale continuous extruders. To analyse and quantify the state of dispersion, microscopic methods were applied.
In the past few decades, extensive research and development have been carried out in the field of polymer composite production. On an industrial scale, melt processing has become the method of choice to incorporate colour pigments or reinforcement particles, for instance. Nowadays, a wide variety of melt compounding technology is available which is applicable to polymer-CNT composite production. In comparison to other composite production methods such as solvent casting, melt mixing is fast, economic, and relatively environmentally friendly. Melt processing extruders can produce composites from a few kilograms to a few tons per hour depending on their size and demand. For this purpose, single (direct incorporation) and multi-step (masterbatch dilution) processing routes mostly on continuously working compounders are employed. However, for smaller batches or special matrix systems, discontinuous processes are also used. The melt mixed composites can be directly processed to the desired products by means of injection or compression moulding, profile extrusion, blow moulding, and fibre spinning.
On an industrial scale, mostly multi-walled CNTs (MWCNTs) are used and therefore, this chapter focuses on MWCNTs which are introduced, for instance, to obtain electrically and thermally conductive and/or mechanically reinforced materials. However, achievement of the desired properties, especially mechanical enhancement of these composites, greatly depends on the extent of CNT dispersion and distribution in the polymer matrix. Agglomerates above a certain size may act as stress concentrators and induce cracks and fracture.
The term ‘dispersion’ describes the process during which the smallest dispersible unit of an agglomerate is separated from it and is mixed in the host matrix system but is also used to characterize the achieved state of CNT separation. In our case, the smallest dispersible unit is an individual carbon nanotube in a polymer matrix. A good distribution is achieved when all available CNTs are uniformly arranged in the host matrix. During melt processing, the shear forces generated inside a mixing unit are supposed to cause dispersion and distribution of the CNTs in the polymer matrix. Practically, however, the aim of good dispersion and distribution cannot be achieved in all cases and in several scientific reports it is reported that complete dispersion of CNT agglomerates in polymer melts is difficult. Quite often the presence of undispersed agglomerates in composite materials is reported (Du et al., 2004; Takase, 2007; Baets et al., 2008; Masuda and Torkelson, 2008; Kasaliwal et al., 2009; Micusik et al., 2009; Prashantha et al., 2009). The choice of processing conditions and base materials seems to have a big impact on dispersion. For example, in Fig. 4.1, several undispersed agglomerates can be seen in optical micrographs of polycarbonate with 1 wt% MWCNTs which are processed at different mixing speeds.
The difficulties in dispersing CNT in polymer melts are usually caused by strong van der Waals interactions, often in combination with physical entanglements between neighbouring tubes. The strength of van der Waals interactions in turn depends on the distance between the nanotubes and the physical or chemical nature of the surrounding medium. The entanglements and thus the agglomerate structure among other things depend on the structure of the support material and the catalyst arrangement. Especially in the case of fluidized bed materials for large-scale production, curved, entangled, and intertwined tubes can be generated due to imperfect growth. Such kinds of ‘as-produced’ CNT materials with primary agglomerates of high cohesive strength can make the preparation of composites a very interesting and challenging technological task.
This chapter aims to provide an overview of the factors that can influence dispersion of MWCNTs during melt processing and is divided into five main sections. After the introduction, the second section deals with fundamental aspects of filler dispersion, as it is necessary to have an insight into the dispersion process. The state of the art of melt-processed polymer–MWCNT composites, along with the different methods that are employed to access filler dispersion, is discussed in the third section. The fourth section is based on small-scale lab studies on melt-processed composites where the influences of variation in material properties and in technological parameters such as mixing speed, temperature, and time on CNT dispersion are discussed, mainly for polycarbonate-based composites. Beside direct incorporation of MWCNTs in polymers for composite manufacturing, the masterbatch dilution method is also briefly considered. The fifth section deals with continuous melt mixing using a co-rotating twin-screw extruder, in which the influence of processing parameters such as screw profile, residence time, mixing speed, and throughput on CNT dispersion during masterbatch production and its dilution is reported on the examples of poly(caprolactone) and poly(carbonate).
Generally, the problem of polymer–CNT composite production can be considered as a homogenization task where filler particles have to be distributed within a viscous medium. In the case of weak interactions between individual CNTs, this process can be classified as distributive mixing. However, as mentioned in the introduction, most CNT materials are delivered as primary agglomerates with remarkable strength. To achieve homogeneous CNT dispersion in polymer composites, the mixing process has to involve a dispersive component. More precisely, this means the mechanical agitation of the MWCNTs agglomerates has to enable their breakdown in size. Principally, the disintegration or dispersion of primary agglomerates can proceed by removal of individual MWCNTs from the agglomerate surface (erosion) or agglomerate breakage (rupture).
A simple theoretical model which is widely used in the literature considers the balance of external stress and agglomerate strength. Agglomerate breakage is assumed to take place if the stress generated by the polymer flow exceeds the agglomerate strength. To estimate the cohesive strength of agglomerates made up of convex-shaped particles, equations like the one proposed by Rumpf (Rumpf, 1970) can be used:
The inter-particle adhesive forces (F) along with porosity (ε), the coordination number of particles forming the agglomerate (k) and the surface area of particles forming the agglomerates (A) are responsible for the cohesive strength of agglomerates. The inter-particle adhesive forces which hold particles together result in agglomerate formation and its growth and can be classified into three main categories. They are, namely, surface and field forces, material bridges, and interlocking (Rumpf, 1970; Schubert, 1979), schematically shown in Fig. 4.2.
4.2 Illustration of different categories of particle adhesion. (Reprinted from Tomas, 2007, with permission from Elsevier.)
A high cohesive strength of agglomerates results in higher resistance against particle separation and dispersion in the matrix system. In the case of dry fillers, surface and field forces are predominantly responsible for the agglomerate strength, whereas for partially wetted or infiltrated agglomerates, liquid bridges are the main influence. Although the forces like van der Waals are considered very weak intermolecular forces, they become quite significant at the nanoscale due to the very high specific surface area of the material. For instance, as the particle size decreases, van der Waals forces exceed the gravitational forces by several orders of magnitude (Tomas and Kleinschmidt, 2009). Further, if liquid bridges are formed between incomplete wet primary particles in partially infiltrated agglomerates, the agglomerate strength increases tremendously. However, if the agglomerates are fully infiltrated, then agglomerate strength becomes negligible (Schubert, 1979).
To achieve uniform dispersion of filler agglomerates in a polymer melt, it is necessary to overcome their agglomerate strength. During melt processing, shear and elongational stresses generated inside a mixing unit provide the required energy for the size reduction of filler agglomerates. In Fig. 4.3, a scheme describing the reduction in the particle size versus the required energy is shown. Large agglomerates are broken down into smaller ones and then to primary particles as the dispersion energy increases. If the applied energy is too high, the particles can be damaged (as e.g. breakage of nanotubes).
4.3 Schematic showing reduction in particle size with increasing dispersion energy. (Reprinted from Wang, 2003, with permission from Elsevier.)
The process of size reduction of filler agglomerates as shown in Fig. 4.3 consists of several stages during dispersive melt mixing operations. These stages are, namely, filler incorporation (in melt), wetting and infiltration of filler (by polymer melt), followed by dispersion, distribution, and flocculation (of filler in the polymer melt). In a melt mixing operation, all these stages, briefly described in the following, run parallel to each other.
As the filler is incorporated in a polymer melt, the filler surface comes into contact with the melt and the melt wets the filler surface. Wetting of a surface is defined as the replacement of a solid–air interface with a solid–liquid interface (Parfitt, 1973). Wetting can be further classified into three types, namely, adhesional, immersional, and spreading wetting. A solid is regarded as completely wet with a liquid if the contact angle is zero (Jaycock, 1973).
As soon as the filler surface is wetted, the melt infiltrates into the filler agglomerate. It is reported in the literature that material parameters like porosity or packing density of agglomerates, melt viscosity, melt molecular weight, and branching of the polymer can significantly affect melt infiltration into agglomerates and thereby the dispersion of agglomerates. Yamada et al. (1997, 1998) found that a matrix with a lower melt viscosity infiltrates faster into agglomerates of carbon black (CB) than a high melt viscosity matrix, and, below a critical packing density, agglomerates exhibited high erosion rates. Similar findings have been reported for CaCO3 agglomerates by Levresse et al. (1999). While studying the intercalation of nano-clay in polystyrene by annealing, Vaia et al. (1995) reported faster intercalation when using a matrix with lower molecular weight and melt viscosity. Furthermore, Roland et al. (2004) reported that dispersion of CB was worse in branched poly(isobutylene) than in linear poly(isobutylene) even when applying high shear stresses.
Following the stage of melt infiltration, the particles are subjected to dispersion. In this stage, the size of filler agglomerates is reduced which is attributed to the two main dispersion mechanisms of rupture and erosion. Rupture mechanism is a fast process during which the large agglomerates are broken down into smaller ones in a short time. In erosion, large agglomerates are reduced to smaller ones by the removal of single particles, aggregates, or fragments and clusters from the surface which needs a comparatively much longer time. The critical shear stress required for dispersion by the erosion process is much lower than the one required for the rupture process. The ratio of applied shear stress and cohesive strength of agglomerates could be decisive in governing the speed of agglomerate dispersion. Thus, the step of filler dispersion is considered to be very important as it determines the rate at which size reduction of the filler in the polymer melt occurs (Manas-Zloczower, 1994).
These dispersion mechanisms have been extensively investigated for fillers such as carbon black (CB), silica, calcium carbonate (CaCO3), etc. and are reported in the literature (e.g. Bolen and Colwell, 1958; Kao and Mason, 1975; Powell and Mason, 1982; Shiga and Furuta, 1985; Rwei et al., 1990; Rwei et al., 1991; Rwei et al., 1992; Hansen et al., 1998). Furthermore, several models are proposed to describe the dispersion mechanism as in e.g. (Tadmor, 1976; Manas-Zloczower et al., 1982; Manas-Zloczower and Feke, 1989; Coran and Donnet, 1992; Seyvet and Navard, 2001; Potente et al., 2002; Lozano et al., 2003; Kasaliwal et al., 2010a).
As the mixing process continues, the particles separated from agglomerates are distributed in the polymer matrix. If the melt state is maintained for a long time, it is possible that the well-distributed particles can flocculate to form clusters; this process is also known as secondary agglomeration. This effect is quite well known for epoxy and elastomer systems, but seldom addressed for much higher viscous thermoplastic melts (Lellinger et al., 2008; Pegel et al., 2008, see also Chapter 7).
Due to the high relevance to industrial applications, melt processing of polymer–MWCNT composites is being extensively investigated and reported in the literature. Nevertheless, comprehensive and reliable studies on relationships between processing and structure of polymer–CNT composites can rarely be found. Many studies simply focus on reporting about the physical properties of the manufactured composites and interpret the results with qualitative statements about CNT dispersion.
Morphological investigations are usually based on microscopic methods such as light microscopy (LM), scanning electron microscopy (SEM), atomic force microscopy (AFM), and transmission electron microscopy (TEM), to characterize the state of dispersion. Nevertheless, apart from a few examples available in the literature, these tools are seldom used to systematically investigate the influence of processing conditions on the state of dispersion. Accessing the quality of filler dispersion is very important to differentiate the impact of different processing conditions on dispersion; but in most cases this is only done qualitatively and not by using quantitative techniques. In the following, some examples from the literature are discussed where such tools were used to draw useful conclusions on parameters influencing the state of dispersion.
For instance, Leer et al. (2006) discussed the state of MWCNT dispersion in systems based on polycarbonate produced using a micro-single screw extruder. Based on optical micrographs analysed using a normalized grey scale distribution, they found that longer compounding time achieved by up to three processing cycles yielded better MWCNT dispersion. Andrews et al. (2002) characterized the state of dispersion of MWCNTs in polypropylene by applying LM and introducing a dispersion index based on an assignment of appearance on a scale between 1 and 10. They reported that the state of dispersion improved drastically as the mixing energy was increased during melt compounding in a batch mixer by increasing either mixing speed or mixing time. Further, from TEM investigations on nanotubes extracted from the composite, they concluded progressive tube shortening with increasing mixing energy. Lin et al. (2006) investigated the state of MWCNT dispersion in polycarbonate for composites produced by a masterbatch dilution step in different lab-scale mixers. They applied TEM investigations to qualitatively characterize the state of MWCNT dispersion and recommended a mixer with the best mixing efficiency. In addition, tubes extracted from the composites were observed using AFM and their length distribution was evaluated in order to quantify the effect of breakage, which was found to be small and similar for all mixers.
A detailed study was reported by Kasaliwal et al. (2009) where mixing conditions were varied for polycarbonate with MWCNT using a factorial design. The effects of melt temperature and mixing speed on dispersion were assessed by a dispersion index obtained from LM images.
Villmow et al. (2008b) studied the influence of MWCNT content, screw profile, temperature profile, and rotation speed during twin-screw extrusion using a Berstorff ZE 25 extruder on the dispersion of poly(lactic acid) (PLA) masterbatches containing 15 and 7.5 wt% MWCNTs and samples diluted from these. From LM, a dispersion index and the number of agglomerates per mm2 were assessed and TEM was used to study MWCNT dispersion and network formation at the submicron scale. High rotation speed (500 rpm) that still ensures a certain residence time of the melt, combined with a screw profile containing mainly mixing elements was found to be highly convenient to disperse and distribute the MWCNT in the PLA matrix as well as during masterbatch production as the dilution step. The temperature profile showed less influence. Nonetheless, an increasing temperature profile resulted in slightly better nanotube dispersion.
The morphological characterization of composites using microscopy techniques gives a direct visual impression of the state of CNT dispersion, even if a 3D structure is visualized using 2D images. Nevertheless, these methods have certain advantages and drawbacks. LM is the easiest method and gives an excellent overview of the micron-scale state of dispersion but has limitations in accessing submicron-size particles. SEM and AFM give general overviews of the state of dispersion on both micron and submicron scale. TEM gives an excellent insight into the state of filler dispersion and distribution on the submicron scale. However, due to time-consuming sample preparation, TEM is only rarely performed. Nevertheless, with the help of proper image-processing techniques and different statistical approaches, valuable quantitative analysis, even deducing on the 3D structure, can be performed to characterize the state of dispersion at all length scales. In this regard, Chapter 9 of this book provides a useful overview of the application of image analysis methods for the quantitative description of dispersion, distribution, and orientation of nanotubes.
Apart from morphological investigation, changes in composite properties are often used to characterize the quality of CNT dispersion. Of these indirect methods, especially melt rheological and electrical resistivity measurements are quite commonly employed. Melt rheology reflects the state of network formation between the nanotubes and polymer chains which enhance elastic properties and induces changes in the melt viscosity of the matrix polymer especially at low frequencies/shear rates as described in detail in Chapter 15 of this book. In addition, several examples (Pötschke et al., 2002; Du et al., 2004; Huang et al., 2006; Fan and Advani, 2007) can be referred to. As well as oscillatory and steady state measurements in shear, creep recovery experiments are found to give sensitive information about the state of dispersion. At longer recovery times, higher storage compliance is observed for systems which are better dispersed (Triebel et al., 2010).
Quite often also, electrical percolation thresholds or achieved conductivity values are used as a measure of the state of dispersion. At a given CNT content, as soon as enough separated tubes become available for electrical network formation, percolation occurs. As more nanotubes are involved in the network, conductivity of the composites further increases before levelling off. Thus, an increase in electrical conductivity can be correlated with an improvement in dispersion, at least in a certain range. An example using dielectric spectroscopy is given by Pötschke et al. (2003), where polycarbonate-based composites were processed in a small-scale mixer under different mixing speeds and times and, depending on these conditions, the MWCNTs were either percolated or not. The influence of mixing conditions was great near the percolation threshold and diminished with increasing nanotube concentration. Also the study by Kasaliwal et al. (2009) considers the effect of mixing conditions on the electrical behaviour and relates the resistivity to the macro dispersion index assessed by quantitative analysis of LM micrographs. Interestingly, even at quite low dispersion indices, the composites with only 1 wt% MWCNT were already conductive. This indicates that also bad dispersion states despite big agglomerates may lead to composites with high conductivities. Electrically conductive networks needed for electrical percolation can be formed at lower CNT amounts at better dispersion but can also be formed from connected agglomerates. Different possible scenarios are illustrated in Fig. 4.4. Thus, the use of electrical percolation or electrical values is equivocal for assessing the state of dispersion.
4.4 Sketches hypothesizing different arrangements of nanotubes and their agglomerates illustrating percolated (b, c, d) and not percolated (a, e) structures: (a) highly agglomerated structure; (b) cluster–cluster percolation; (c) network consisting of small agglomerates and dispersed nanotubes; (d) network of well-dispersed nanotubes; (e) completely dispersed nanotubes not forming a network.
Based on former investigations with CB, Le et al. (2009) presented an online technique to characterize the state of CNT dispersion during compounding with rubber (see also Le et al. 2004). Here the electrical conductance is directly measured in a batch mixer. The trends in on-line conductance followed the values measured off-line and were closely related to the state of dispersion observed microscopically.
Another method used in the literature is Raman imaging, where the intensity of the G-band was recorded for 40 × 40 μm spots and the standard deviation of that map was correlated to the quality of the CNT dispersion (Du et al., 2004). Thereby, differences in dispersion of wet and dried SWCNT in PMMA could be assessed. Some authors have employed wide angle X-ray scattering (WAXS) (McNally et al., 2005) and small angle X-ray scattering (SAXS) (Pujari et al., 2009) investigations to characterize the state of MWCNT dispersion. Brühwiler et al. (2010) recently presented another example using SAXS, where on injection-moulded plates position-resolved average SAXS intensities were obtained for different scattering vectors q. The homogeneity of the intensity distribution sensitive to nanotubes was assigned to the nanotube dispersion, showing agglomerates of MWCNT in PC at higher loadings. Recently, in some reports, differential scanning calorimetry (DSC) was used to qualitatively differentiate states of dispersion. The nucleation action of nanotubes in a crystallizable polymer matrix is proportional to the amount of nanotube surface. Thus, at better dispersions, a higher surface is available for crystallization and the melting enthalpy was found to be higher (Masuda and Torkelson, 2008; Villmow et al., 2008b; Pujari et al., 2009).
Batch compounding in small-scale mixers is very helpful in determining the factors that can influence CNT dispersion. The main advantage of using small-scale mixers is that a small material amount is needed and these mixers can be easily handled. Thus, a broad variety of mixtures and processing parameters can be investigated over short time scales. In this section some important relationships between different melt processing conditions as well as the choice of base materials on the CNT dispersion are summarized. The results obtained are helpful for understanding the scientific background about ongoing dispersion processes and to have guidelines for larger mixing scales.
The results presented originate from experiments using a DACA microcompounder. The assembly represents a conical twin screw extruder but has a bypass and a chamber volume of only 4.5 cm3. The design of this kind of compounder allows the variation of the technological parameters’ mixing time, speed and temperature.
To obtain empirical relationships or to rate the quality of different composites, the state of CNT dispersion has been investigated by means of LM and TEM. As described in Chapter 9, thin sections were prepared from extruded strands and at least six light micrographs have been used to evaluate the area ratio of undispersed agglomerates AA (area of agglomerates related to the image area) and the degree of dispersion DLM. From at least five TEM images the degree of dispersion DTEM as well as the distribution coefficient QP were derived. For some sets, the particle size distribution (number of agglomerates in different size classes) was evaluated and the mean circle equivalent diameter x50 was determined.
The properties of MWCNT materials produced by different manufacturers can vary significantly. This variation is mainly concerning their purity, packing density or porosity of agglomerates, length and diameter distribution of tubes, size distribution of produced agglomerates, or structure of the agglomerates themselves. In Table 4.1, some of these characteristic features of MWCNT materials from different manufacturers are summarized. In Fig. 4.5, SEM micrographs of two of these as-received materials are shown which differ in agglomerate size, surface roughness, and substructure. When looking inside the fractured surface of agglomerates, in both materials, a substructure was observed revealing smaller agglomerates. The size of these small agglomerates could be dependent on the length of the nanotubes. At higher magnifications, the agglomerates can appear to have a combed yarn structure (as in Fig. 4.6 (a) for Nanocyl NC 7000) or they can indicate a bird’s nest structure (as in case of Fig. 4.6 (b) for Baytubes® C150 HP).
‡Not measurable. Blank space: data unavailable. Ts-Na = Tsinghua-Nafine Nano-Powder Commercialization Engineering Center, Beijing, China.
4.5 SEM images of MWCNT materials from different manufacturers shown at different magnifications: (a, b, c) Nanocyl NC 7000; (d, e, f) Baytubes C150 HP (from Nanocyl S.A., Sambreville, Belgium) (from Bayer MaterialScience AG, Leverkusen, Germany) (reproduced from Villmow et al., 2010).
4.6 Optical micrographs of PC composites containing 1 wt% MWCNT melt compounded in a DACA microcompounder at 280 °C, 50 rpm, 15 min.: (a) Nanocyl NC 7000; (b) Baytubes C150 HP; (c) Ts-Na (1st lot); (d) Ts-Na (2nd lot), section thickness 15 μm.
MWCNT materials, namely, Nanocyl NC7000, Baytubes® C150 HP, and Ts-Na (two lots, delivered at different dates), were melt compounded with polycarbonate under identical conditions using the DACA microcompounder. Representative LM and TEM micrographs of these composites are shown in Figs 4.6 and 4.7, respectively. Such images were analysed to obtain the degrees of dispersion DLM and DTEM as well as the distribution coefficient QP. The corresponding results are summarized in Table 4.2 (for details, see Chapter 9).
4.7 TEM images of PC composites containing 1 wt% MWNT: (a) Nanocyl NC 7000; (b) Baytubes C150 HP; (c) Ts-Na (1st lot); (d) Ts-Na (2nd lot); (a, c, and d reproduced from Pegel et al., 2008).
The microscopy images and the analysis indicate that Nanocyl NC7000 can be better dispersed than the other CNT materials, whereas the second lot of Ts-Na showed the worst dispersion. The reason for better dispersability of a nanotube material cannot be attributed to one specific factor but it is collectively based on several factors such as differences in internal morphology (see Fig. 4.5), packing density, and tube dimensions which depend on synthesis conditions. Slight deviations in manufacturing conditions during synthesis can significantly affect the dispersability as evidenced in the two lots of Ts-Na where identical morphological features of the single nanotubes (length and diameter) were found (Pegel et al. 2008). Possibly, small differences in the content of impurities and available surface groups due to changed nanotube growth conditions may also be contributory factors.
Interestingly, this sensitivity of dispersability and dispersion obtained on the characteristic features of the starting primary nanotube agglomerates is rarely discussed in the literature. Even the variation in diameter or the length of nanotubes might influence their dispersability. To address this issue in more detail, polycarbonate was melt compounded with 1 wt% of different MWCNT materials from the same producer, having approximately the same tube length but varying in the tube diameter (from ca. 5 to 10 nm), using the DACA microcompounder. As illustrated in Fig. 4.8, the dispersion of the MWCNT improves as the MWCNT diameter (measured using TEM investigations) is increased, indicating better dispersability of the thicker tubes. As thinner MWCNTs are less stiff than thicker ones, they are able to be much more entangled with neighbouring tubes and develop stronger van der Waals interactions. In addition, as per the Rumpf equation (Rumpf, 1970), the strength of agglomerates is inversely proportional to the radius of the constituting tube, thinner tubes form agglomerates with higher strength.
4.8 Optical micrographs of PC with 1 wt% MWCNT having different external diameter distributions (material and data supplied by Nanocyl S.A.); (a) 5.4 ± 1.6 nm, C purity 89%, AA = 8.03%; (b) 6.9 ± 1.9 nm, C purity 98.7%, AA = 1.8%; (c) 9.9 ± 4.1 nm, C purity 90%, AA = 0.3%. Composites produced at 280 °C, 50 rpm, 15 minutes, section thickness 15 μm. AA, area ratio.
The lengths of MWCNTs can be varied either during synthesis or during post-treatment. For the latter case, ball milling is quite often used. To study the effect of variation in MWCNT length on dispersion, MWCNT NC 7000 as delivered was subjected to ball milling for different times and these materials were melt compounded with PC. Optical micrographs of composites with 4 wt% nanotubes are shown in Fig. 4.9. The agglomerate area and size significantly decreased with shortening the tubes, whereas at the highest milling time a huge number of very small agglomerates were observed. It has to be taken into account that during ball milling also a compaction of agglomerates can occur, so that possibly the remaining smaller agglomerates are much more compacted and more difficult to disperse into single tubes on the sub-micron scale. This is indicated by the fact that the better dispersed ball milled nanotubes showed worse electrical properties than the untreated ones. An additional effect may arise from surface groups generated by the ball milling shortening.
4.9 Optical micrographs of PC with 4 wt% MWCNT based on Nanocyl N7000; (a) Nanocyl NC 7000, length x50 1300 nm, AA = 5.87%; (b) ball milling for 3 hours, length x50 360 nm, AA = 0.856%; (c) ball milling for 10 hours, length x50 200 nm, AA = 1.32%. Composites produced at 280 °C, 50 rpm, 15 minutes, section thickness 5 μm.
A qualitative improvement in dispersion was also reported when using shear pulverized versus as produced MWCNTs in melt compounding with polypropylene (Masuda and Torkelson, 2008; Pujari et al., 2009).
Polymers differ in several ways, such as chemical structure of the backbone chain, side groups, their amorphous or crystalline nature, surface energies, etc. Even for a given polymer, there could be variations in matrix molecular weight and therefore in melt viscosity at a given temperature. Collectively, these factors have an impact on CNT dispersion in different matrices. It can be assumed that the state of macrodispersion of nanotubes does not change upon solidification, even if in partially crystalline polymers changes in the nanoarrangement may occur. In this study, different polymers such as polypropylene (PP), poly methyl methacrylate (PMMA), polycarbonate, polystyrene (PS), polyamide 66 (PA66), polyamide 12 (PA12), low density polyethylene (LDPE), and poly(ethylene terephthalate) (PET) were melt compounded with 1 vol.% MWCNT (Nanocyl NC 7000, sieved agglomerate size fraction of 100–200 μm). To achieve similar applied shear stresses during compounding, the mixing temperature of each polymer was adjusted in a way that, at a given shear rate (of 300 rad/s), the melt viscosity was the same (500 Pas). Optical micrographs illustrating the state of MWCNT agglomerate dispersion in these polymers are shown in Fig. 4.10.
4.10 Optical micrographs of 1 vol% MWCNT composites produced by using (a) PA66 (265 °C); (b) PET (290 °C); (c) PC (300 °C); (d) PA12 (220 °C); (e) PP (200 °C); and (f) LDPE (170 °C). composites produced at 100 rpm, 5 min., processing temperature indicated beside the name of the polymer, section thickness 20 μm.
Although the applied shear stress was the same in all cases, tremendous differences in the agglomerate dispersion were observed. Apart from melt viscosity, interfacial energy between polymer and nanotubes is a property that may influence MWCNT dispersion. In order to verify that possible influence, the agglomerate area ratio is plotted versus interfacial energy (taken from the literature1) in Fig. 4.11. Interestingly, no simple relationship is found, even if there is a general tendency for the area ratio to increase with interfacial energy. Excluding PMMA and PS, the tendency is clearer.
So far, the deviation of PS and PMMA is not easily explainable but may be attributed to the high stiffness of PS chains or the bulky methyl methacrylate side groups in PMMA, which hinder the infiltration of these polymer melts into the primary agglomerates. These results indicate that interfacial energy may not be the only influencing factor (next to viscosity) on macro scale dispersion. Possibly, in addition, chemical affinity and steric aspects of molecular structure have to be considered.
During melt compounding, the shear stresses applied to the primary agglomerates are important for their dispersion. Increasing melt viscosity results in higher shear stresses on agglomerates and can lead to faster size reduction of fillers as compared to low viscosity matrices. Several studies concerning the influence of polymer matrix viscosity on the dispersion of layered silicates (nanoclay) support this argument (Fornes et al., 2001; Gianelli et al., 2005; Chu et al., 2007).
In composites with carbon nanotubes, in addition, the internal porous structure of the primary agglomerates has to be considered, in which the stage of melt infiltration is very important. Infiltration is enhanced when using low viscosity polymer melts. If the melt infiltration is suitable, the agglomerates can be significantly weakened and thus lower shear stresses are needed for their dispersion. Infiltrated agglomerate surface layers enhance the peripheral erosion of agglomerates. Thus, the effects of high and low viscosity counteract on the generation of high stresses and enabling good infiltration, and that balance is also dependent on the nanotube agglomerate structure.
In order to investigate the influence of matrix viscosity on dispersion, three polycarbonates of different molecular weight and thereby melt viscosity at a given temperature were melt mixed with 1 wt% Baytubes® C150 HP. Different mixing speeds were employed to generate different applied shear stresses on filler agglomerates. In Fig. 4.12, optical micrographs of some of these composites are shown.
4.12 Optical micrographs of polycarbonate with 1 wt% MWCNT (Baytubes C150 HP) where the polycarbonate differs in melt viscosity; (a) low melt viscosity (Makrolon2205); (b) medium melt viscosity (Makrolon 2600); (c) high melt viscosity (Makrolon 3108). Composites prepared at 280 °C, 50 rpm, 5 min., section thickness 20 μm.
In Figs 4.13 and 4.14, the agglomerate area ratio of these composites (prepared at different mixing speeds) is plotted versus mixing speed and applied shear stress, respectively. At lower mixing speeds up to 150 rpm, dispersion is better in the high viscous matrix than in lower viscosity matrices as applied shear stresses are higher. However, as the mixing speed increases, differences in the state of dispersion narrow and vanish at high mixing speeds, although big differences in applied shear stress exist. When plotting the area ratio versus the applied shear stress, it becomes obvious that (above 400 kPa) to generate a certain agglomerate area higher stresses were applied for the high viscous matrix. To generate the same (low) agglomerate ratio, much less shear stress was needed in the case of low viscous matrix, indicating the important role of the infiltration process in these composites.
4.13 Area ratio AA vs. mixing speed for composites of PCs with different viscosity and 1 wt% Baytubes C150 HP (adapted from Kasaliwal et al., 2010b).
4.14 Area ratio AA vs. applied shear stress for composites of PCs with different viscosity and 1 wt% Baytubes C150 HP (adapted from Kasaliwal et al., 2010b).
In a rubber MWCNT system, Le et al. (2009) mentioned similar observations whereas lowering of the melt viscosity of natural rubber till an optimum level was found to improve the nanotube dispersion.
Even if the matrix molecular weight is related to the melt viscosity of polymers, it was important to know in context with the important agglomerate infiltration step whether there is an additional influence of molecular weight besides viscosity on dispersion. To observe that effect on MWCNT dispersion, it is necessary to adjust the temperature of the given polymer to have similar melt viscosities. Therefore, the three polycarbonates used previously were now melt compounded at temperatures adjusted to generate the same melt viscosity. Thereby, comparable shear stresses were applied from the different polycarbonates on the primary agglomerates. As illustrated in Figs 4.15 and 4.16, with increasing molecular weight also the size of undispersed agglomerates increases.
4.15 Optical micrographs of polycarbonate with 1 wt% MWCNT (Baytubes C150 HP) where the polycarbonate differs in molecular weight but shows the same melt viscosity: (a) low molecular weight (Makrolon 2205, 260 °C); (b) medium molecular weight (Makrolon 2600, 290 °C); (c) high molecular weight (Makrolon 3108, 310 °C). Composites prepared at 100 rpm, 5 min., temperatures adjusted to obtain a zero shear viscosity of about 500 Pas, section thickness 20 μm.
4.16 Particle size distribution in composites of polycarbonate with 1 wt% MWCNT (Baytubes C150 HP) where the polycarbonates differ in molecular weight but have the same melt viscosity (adapted from Kasaliwal et al., 2010b).
The results show that the molecular weight of matrix clearly has an additional influence on the size of undispersed agglomerates whereby low molecular weight results in smaller and less numerous particles. This can be attributed to the much faster infiltration of polymer chains with lower molecular weight into the filler agglomerates thereby making them weaker and facilitating the dispersion process.
In small-scale batch mixing, the important process parameters that can be varied are rotation speed, mixing time, melt temperature, and degree of filling. These parameters affect the shear rate, residence time, melt viscosity, and by that shear stresses or mixing energy provided to the system. For a given system, varying mixing conditions therefore is expected to influence the dispersion kinetics and the final dispersion state of primary nanotube agglomerates.
To observe the effects of variation in technological parameters, polycarbonate (medium viscosity, Makrolon 2600) was melt compounded with 1 wt% MWCNTs (Baytubes® C150 HP) by varying melt temperature and mixing speed at a constant mixing time of 5 min (Kasaliwal et al., 2009). In Fig. 4.17, the remaining agglomerate area ratio AA (based on sections with 20 μm thickness) is plotted versus mixing speed, for composites prepared at three melt temperatures. In composites prepared at low mixing speed better dispersion is observed when mixing temperature is low or melt viscosity is high. In composites prepared at high speeds, dispersion is the same regardless of applied mixing temperature or melt viscosity. In Fig. 4.18, the area ratio AA is plotted versus mixing energy indicating that a common dependency is received for all three temperatures. With increasing mixing energy, the area ratio first decreases significantly but above about 4000 J/cm3 the curve levels off, and further increase in mixing energy does not seem to enhance dispersion further. The results correspond to those mentioned in Section 4.4.4: at low speeds, agglomerate dispersion is shear stress-driven and at high speeds similar dispersions are achieved irrespective of the melt temperature or viscosity.
4.17 Area ratio AA vs. mixing speed for composites of PC with 1 wt% Baytubes C150 HP produced at different temperatures (adapted from Kasaliwal et al., 2009).
4.18 Area ratio AA vs. mixing energy for composites of PC with 1 wt% Baytubes C150 HP produced at different temperatures (adapted from Kasaliwal et al., 2009).
In any melt mixing process the agglomerates can undergo dispersion by both rupture and erosion mechanisms, whereas the extent of these mechanisms might vary depending upon the mixing conditions employed. To investigate the share of both mechanisms a kinetic study (14 mixing times between 0.5 and 40 min.) was performed using a low viscous PC (Makrolon 2205) and 1 wt% Baytubes® C 150 HP, applying three mixing speeds (50, 100, and 300 rpm) at a melt temperature of 280 °C (Kasaliwal et al., 2010a). The area ratio of undispersed agglomerates AA was plotted versus mixing time (Fig. 4.19) and shows a dramatic increase with mixing speed and time. The dispersion kinetics as assessed by plotting the relative change of area ratio related to the area ratio at 0.5 min., versus mixing time (Fig. 4.20) reveals faster and higher relative changes at higher mixing speed. However, with progressing MWCNT dispersion the rate slows down and under the conditions selected it was not possible to get complete dispersion at macro scale. At high mixing speeds and prolonged mixing time, usually small agglomerates are observed, which are difficult to disperse. From theoretical calculations it can be concluded that smaller agglomerates have higher agglomerate strength than the larger ones (Kasaliwal et al., 2010a) and therefore are more difficult to disperse. The remaining agglomerates may be related to the internal structure of the primary agglomerates as illustrated in Fig. 4.5, where denser small agglomerates in the size range of 1–10 μm were observed within the big agglomerates.
4.19 Area ratio AA vs. mixing time at different mixing speeds for composites of PC with 1 wt% Baytubes C150 HP (reproduced from Kasaliwal et al., 2010a).
4.20 Relative change in the area ratio AA vs. mixing time for composites of PC with 1 wt% Baytubes Cl50 HP (reproduced from Kasaliwal et al., 2010a).
Further mathematical treatment of these values using a model developed by Kasaliwal et al. (2010a) resulted in the conclusion that, at high speed (300 rpm), the fast dispersion is governed by the rupture-dominant mechanism, whereas at low speeds (50 and 100 rpm), the slower dispersion is governed by both rupture-and erosion-dominant mechanisms.
In addition, electrical resistivity of all samples was measured. When plotting resistivity versus the remaining agglomerate ratio (see Fig. 4.21), two plateaus and a transition range between low and high resistivity were found, indicating dispersion-dependent percolation behaviour.
4.21 Volume resistivity vs. area ratio AA for composites of PC with 1 wt% Baytubes C150 HP (reproduced from Kasaliwal et al., 2010a).
At agglomerate area ratios above 7%, corresponding to low mixing energies, the amount of dispersed nanotubes is not high enough for percolation and electrical resistivity of the composites is altogether dominated by the matrix system. When a sufficient amount of dispersed nanotubes is generated, in our case between 7% and 4% of the remaining agglomerates, percolation occurs and in that small concentration range the resistivity decreases by more than 12 decades depending on the state of dispersion. Below approximately 4%, no additional changes in resistivity were observed, even if the area ratio goes down to nearly 0.2%; the electrical resistivity is dominated by the nanotubes. This is a very interesting finding since it is commonly assumed that the electrical resistivity decreases with increasing filler dispersion (see discussion in Section 4.3). However, these results clearly indicate that this assumption holds only in a limited transition range.
When using masterbatches for the production of composite materials, the quality of dispersion within the masterbatch is an important prerequisite for achieving good dispersion in the diluted composites. In a good masterbatch it is expected that all of the primary nanotube agglomerates will be dispersed into single tubes and wetted by polymer. When a masterbatch contains uninfiltrated, large remaining agglomerates, it is quite difficult to disperse them in the dilution step (compare, e.g. Miušík et al., 2009), whereas infiltrated smaller agglomerates may be dispersed at that stage. As agglomerates are subjected to shearing twice, first, during masterbatch production and, second, during dilution, in most cases better dispersion is observed in diluted composites.
During masterbatch dilution, an already existing percolated nanotube network can be extended by the incorporation of additional polymer chains or by rupture and erosion processes, as previously described. For this, miscibility between the polymer chains in the masterbatch and the diluting polymer is required; otherwise a phase-separated structure can be formed between a masterbatch phase and a non-filled polymer phase. As in many cases the nature of the polymer used in the masterbatch is not known, the addition of non-matching dilution polymer may lead to non-homogeneous materials.
In principle, the effects of mixing conditions on dispersion are similar to the ones observed during direct incorporation of MWCNTs in melt compounding. Distinct from direct incorporation, the ‘primary agglomerates’ which should be dispersed are now the masterbatch with already infiltrated polymer chains. Thus, the effects discussed in context with wetting and infiltration of the polymer melt into primary agglomerates do not play a major role. As a general effect, it is found that higher shear forces (achieved by lower temperature or higher rotation speed) improve dispersion.
For instance, in Fig. 4.22, TEM images of polycarbonate–MWCNT composites produced from a masterbatch (from Hyperion Cat. Inc., USA) with 15 wt% MWCNTs at two different temperatures and low mixing speed are shown. In the sample produced at 250 °C, better dispersion is observed as compared to that diluted at 300 °C, which may be assigned to higher shear stresses at the lower mixing temperature. It should be emphasized that both samples were completely free of primary agglomerates as checked by LM investigations.
4.22 Polycarbonate with 0.875 wt% CNT composites diluted from a masterbatch containing 15 wt% MWCNT under different dilution conditions: (a) 250 °C, 50 rpm, 15 min.; (b) 300 °C, 50 rpm, 15 min. (reproduced from Pegel et al., 2008).
In the dispersion at the nanoscale, another effect has to be considered which is well known for the formation of electrical pathways in epoxy systems, but not much regarded in thermoplastic melts. This is the effect of secondary agglomeration, also called flocculation or clustering. It occurs between well-separated nanotubes due to long-range interactions between the nanotubes in the liquid and strongly depends on the matrix viscosity. For melts with well-dispersed nanotubes, a shear-induced insulator–conductor transition was described by Skipa et al. (2010), using time-resolved measurements of electrical conductivity under steady shear and in the quiescent melt. These effects were assigned to the agglomeration of nanotubes under steady shear and the formation of an electrical conductive network of interconnected agglomerates (Alig et al., 2008; Skipa et al., 2010), however, without morphological investigations.
The effect of secondary agglomeration could be visualized using TEM after subjecting the composite shown in Fig. 4.22 to a second mixing cycle at 300 °C and different mixing speeds (Pegel et al., 2008). At 300 °C, the previously nicely dispersed composite (Fig. 4.22 (a)) now develops small secondary agglomerates, the size of which is bigger in the case of a lower mixing speed (Fig. 4.23 (a)) than for a higher mixing speed (Fig. 4.23 (b)).
4.23 Secondary agglomeration of MWNT in the polycarbonate melt processed from the state shown in Fig. 4.22 (a) in a second step at (a) 300 °C, 50 rpm, 5 min; (b) 300 °C, 250 rpm, 5 min (reproduced from Pegel et al., 2008).
The effect of secondary agglomeration can also be seen when looking at the morphology of injection-moulded parts produced under different injection-moulding conditions (Villmow et al., 2008a). This was illustrated by TEM images of injection-moulded plates of polycarbonate with 2 and 5 wt% MWCNTs when investigating the morphology in different depths from the surface. In the inner part, especially when high melt temperatures had been used during moulding, secondary agglomerates could be seen.
In the previous section, results obtained by small-scale batch mixing of polymer–MWCNT composites using microcompounders were discussed. These processing–property relationships can be applied to a certain amount also to melt processing using continuous melt mixing processes like twin-screw extrusion, which is highly relevant to industrial manufacturing of composite materials and parts. As CNT producers increase their manufacturing capacities and the prices of CNT drop, industrial applications of melt mixed composites gain more importance. In that context, investigations on influencing factors in laboratory-scale twin-screw extrusion (Villmow et al., 2008b) and injection moulding (Lellinger et al., 2008; Villmow et al., 2008a) have started to attract attention.
From small-scale mixing it was concluded that high screw speed and long residence time result in better dispersion of MWCNTs within polymer matrices. Previous investigations on PLA (Villmow et al., 2008b), using a twin-screw extruder, also indicated that higher mixing speeds (100 rpm vs. 500 rpm) were favourable in masterbatch production and dilution. However, in comparison to small-scale mixing, the melt mixing using continuous processes provides new challenges. The residence time of polymer melt during twin-screw extrusion is a complex function of rotation speed, throughput, and screw configuration and cannot be adjusted independently. The interaction between different processing parameters leads to complex functions of shear conditions and residence time, thus affecting CNT dispersion. In this section, the influence of extrusion conditions using a co-rotating intermeshing twin-screw extruder (ZE25, from Berstorff, Germany) on the macroscopic MWCNT dispersion is discussed for masterbatches based on poly (caprolactone) (PCL) and their diluted composites.
4.5.1 Preparation of masterbatches: influence of screw configuration on residence time and MWCNT dispersion
Twin-screw extruders provide a modular assembly, and especially the extrusion screw can be varied in nearly endless configurations. This machine consists of different numbers of conveying, kneading, and mixing elements, resulting in different flow and shear conditions. For this study, masterbatches containing 7.5 wt% MWCNTs (Nanocyl NC 7000, from Nanocyl S.A. Belgium) based on PCL (CAPA 6800, from Perstorp, UK) were produced, using five different screw configurations. Two classes of screw configurations were used; whereas dispersive screws favour the break-up of cohesive agglomerates, distributive screws increase the homogeneity of the mixture by continuously generating new interfacial area.
The two dispersive screw configurations (SC1 and SC2, see Fig. 4.24) used in the present study contain kneading and conveying elements. For these screws, the number of kneading blocks is the same, but the number of right-handed conveying elements was varied to control the minimum residence time during extrusion. The kneading blocks consisted of five kneading discs with a positive (45°) staggering angle. Both screws were designed to have a length to diameter ratio (L/D) of 36.
4.24 Dispersive screw configurations SC1 and SC2 (adapted from Villmow et al., 2010).
Three distributive screw configurations were designed based on the dispersive ones. In SC3, the number and location of conveying elements of screw configuration SC1 were maintained but the kneading elements were substituted by mixing elements. Screw configuration SC4 was designed to have additional right-handed conveying elements compared to SC3 (Fig. 4.25).
4.25 Distributive screw configurations SC3 and SC4 (adapted from Villmow et al., 2010).
The fifth screw profile was based on SC4, but having an extended L/D of 48. Additionally, this screw configuration contains two extra mixing zones to have five in total (Fig. 4.26). Each mixing zone consisted of mixing elements having five discs, whereas each had 10 cogs. All experiments were performed at a mean barrel temperature of 200 °C.
4.26 Distributive screw configuration SC5 (adapted from Villmow et al., 2010).
The use of kneading elements results in longer minimum residence times, as these elements exhibit an active (conveying or stuffing) effect in comparison to mixing elements that act neutrally regarding their conveying behaviour. This effect is independent from rotation speed and throughput. The further addition of back conveying elements (SC2 and SC4) leads to a significant increase in tR,min for both types of mixing. The gain in tR,min is thereby in the range of 10 to 30 seconds depending on the mixing speed. Besides the number of back conveying elements, tR,min strongly depends on L/D of the processing unit. Thus, SC5 exhibits the longest tR,mindetermined in the frame of this study.
The influence of mixing speed and throughput on tR,min using different screw configurations is shown in Fig. 4.27. As an increase of mixing speed leads to higher conveying velocity and decreased filling degree of the screws, shorter tR,min were observed. Increasing the rotation speed from 100 to 500 rpm results in a decrease in tR,min in the range of 30 to 50 seconds. Furthermore, the use of an extended screw (SC5) increased tR,min by about 50% for both investigated mixing speeds. The influence of throughput on the tR,min was investigated in the range between 5 and 15 kg/h, by maintaining constant rotation speed of 500 rpm. The tR,min decreases exponentially with increasing throughput and was approximately halved when comparing the values determined for 5 and 15 kg/h.
4.27 Minimum residence time tR,min of PCL depending on processing parameters and screw configurations: (a) versus rotation speed; (b) versus throughput (adapted from Villmow et al., 2010).
The investigation of macro dispersion from LM investigations (sample thickness 2.5 μm, particles > 5 μm regarded) reveals an exponential decrease of the area ratio AA of undispersed MWCNT agglomerates with increasing residence time tR,min for both types of mixing (dispersive or distributive) as illustrated in Fig. 4.28. For comparable residence times, better MWCNTs dispersions were observed for distributive mixing in comparison to dispersive mixing.
4.28 Dependence of the agglomerate area ratio AA on the minimum residence time tR,min in PCL masterbatches with 7.5 wt% MWCNT (adapted from Villmow et al., 2010).
The exponential decrease of AA with residence time results in a plateau level (at about 2%) when using distributive screw configurations, which was not observed for the dispersive screw configurations. Interestingly, even when applying long residence times (low throughputs), remaining primary agglomerates in the range of 70 to 100 μm were observed, as shown in Fig. 4.29 (b) for SC5. The particle size distributions for the composites also indicate that these large particles were present under all conditions investigated as shown in Fig. 4.29 for composites produced with SC2 and SC5.
In addition, investigation of the morphology development along the length of the screw during extrusion was performed, giving additional information concerning the impact of different mixing and kneading elements on the MWCNT dispersion. The extruder was stopped after reaching steady state extrusion conditions and the screws were pulled out from the processing unit within a few seconds. Several samples of approximately 1 gram were taken from the screws (using screw configurations SC2 and SC5, rotation speed 500 rpm and throughput 5 kg/h). The area ratio AA of undispersed MWCNTs within the PCL masterbatches decreases exponentially along the extruder as illustrated in Fig. 4.30. The numbers 1 to 6 indicate some of the sampling positions, which will be referred to in Fig. 4.31.
4.30 Development of MWCNT macro dispersion along the extrusion length for screw profiles SC2 and SC5 (adapted from Villmow et al., 2010).
4.31 Development of MWCNT agglomerate size distribution along the extrusion length for screw configurations SC2 (a) and SC5 (b) (adapted from Villmow et al., 2010).
The agglomerate size distributions as shown in Fig. 4.31 illustrate that their width was similar for all samples taken along the extruder. The maximum agglomerate size of 120 μm observed directly after the melting zone was also found in the material taken close to the extrusion die. This finding indicates, again, that the influence of extrusion length or residence time on the dispersion of large primary agglomerates is limited. However, it is the particle size distribution in the range between 10 and 60 μm, which changes significantly along the extruder. It has to be assumed that the breaking up of very big agglomerates into fragments of 100 μm or smaller happens in the melting zone of the twin-screw extruder at the very beginning of the mixing process and is thereby mainly influenced by the rotation speed, which controls the shear forces acting on the agglomerates.
The shear conditions during extrusion can be quantified using the specific mechanical energy input (SME), where the torque during extrusion is taken into account. Fig. 4.32 shows the area ratio AA depending on S ME and reveals a power law dependency for both types of mixing. Although the overall energy input is lower when using distributive screw configurations, they result in better MWCNT dispersion. This has to be attributed to the high mixing efficiency of mixing elements as compared to kneading elements. Generally, it can be concluded from these results that low throughputs and high rotation speeds could further improve the MWCNT’s dispersion.
The final stage in the production of nanocomposites was done by dilution of the masterbatch exhibiting the best MWCNT macro dispersion, which was processed with an extended distributive screw under application of 500 rpm and 5 kg/h. The dilution process was done under the same conditions in order to get composites with MWCNT contents between 0.125 and 4.0 wt%.
Figure 4.33 shows light microscopic images of two different PCL composites with 0.5 and 1.0 wt% content and demonstrates that the primary MWCNT agglomerates were entirely dispersed within the matrix. The resulting composites showed a very low percolation threshold of 0.244 vol.%, as measured on compression moulded samples (Fig. 4.34).
4.33 Light microscopic images of PCL composites with different MWCNT contents: (a) 0.5 wt%; (b) 1.0 wt% (reproduced from Villmow et al., 2010).
The mixing conditions found to be optimal for the preparation of the PCL-based masterbatches (7.5 wt% MWCNT, extended distributive screw (SC5), 500 rpm, 5 kg/h) were also applied to the production of masterbatches based on poly(carbonate) (Lexan 141R, from Sabic Innovative Plastics) with the same type of MWCNTs (NC7000, from Nanocyl S.A., Belgium). The mean barrel temperature was adapted to the poly(carbonate) and was in the range of 260 °C. The dilution of the masterbatch was repeated under the same conditions. The MWCNT dispersion obtained was excellent, almost showing nearly no remaining primary agglomerates. Correspondingly, the percolation threshold as measured on compression-moulded plates was as low as 0.110 vol.% (Fig. 4.34). This outcome indicates that the results obtained for the optimization of extrusion processing for PCL/MWCNT composites can be transferred successfully to other polymer/CNT systems.
To make polymer–MWCNT composites a commercial success, the production of these composites free of undispersed primary MWCNT agglomerates is very important. Next to reducing the amount of nanotubes available for network formation, undispersed primary agglomerates above certain sizes act as crack initiators in mechanical tests and reduce surface appearance of parts made from such composites.
The factors influencing the dispersion of MWCNTs in a polymer matrix are numerous and can be divided into material and processing parameters. Different synthesis methods employed by CNT manufacturers result in MWCNT materials varying in certain characteristic features such as agglomerate structure, packing density, lengths and diameters, purity etc. These differences affect the dispersability of the as-produced primary MWCNT agglomerates. Thus, MWCNTs produced by different manufacturers disperse to different extents in a given polymer matrix. Furthermore, polymer matrices differing in their chemical structure were found to influence the quality of MWCNT dispersion, whereby the exact reason behind these differences is still under discussion. The interfacial energy between MWCNTs and polymers is one of the factors that could influence the dispersion of MWCNTs, where lower interfacial energy was found to enhance dispersion. Other factors like chain stiffness and mobility could have determining effects. In addition, melt viscosity was shown to have counteracting effects on the step of melt infiltration of primary agglomerates and the generation of high shear stresses, which has to be balanced in order to get good nanotube dispersion. Interestingly, it could also be shown that next to melt viscosity, the molecular weight of the matrix polymer has an additional influence on dispersion, which can be attributed to the faster infiltration of lower molecular weight polymer chains, even if the polymer shows the same melt viscosity as higher molecular chains.
Concerning the processing parameters, the investigations of dispersion kinetics in small-scale mixing showed that two important parameters drive the size reduction of MWCNT agglomerates: polymer melt infiltration and applied shear stress. If the applied shear stresses are high enough, agglomerates can be broken up and quickly dispersed. When the agglomerates are well infiltrated, they can disperse by the application of relatively lower shear stresses. Thus, an optimum balance between melt infiltration and applied shear stresses is desirable for fast agglomerate dispersion. Such balanced optimized mixing conditions could be realized by using low melt viscosities, long mixing/residence time, and high mixing speeds.
Furthermore, masterbatch dilution technology is a promising route to produce composites with no or only small remaining agglomerates, as the materials undergo the mixing procedure twice, thus enhancing mixing energy input and residence time. However, the quality of the starting masterbatch is also important for the quality of the composites prepared from it by dilution.
In the case of continuous extruders, parameters such as screw configuration, mixing speeds, throughput, temperature profile, etc. have significant influences on the shear stresses and residence time and thereby on the quality of MWCNT dispersion. Based on investigations on PLA-based masterbatches, it was found that higher L/D ratios of screws with distributive screw configurations can yield better MWCNT dispersion.
Systematic studies on influencing factors on the dispersion of MWCNT in polymers by melt compounding have just started and are still in the state of development. So far, complete dispersion of MWCNT agglomerates via melt compounding has rarely been reported in the literature, let alone investigated seriously. To solve this problem, the application of additives and surfactant as dispersion aids could be promising. Such additives can tremendously influence the interfacial energy, thereby easing the agglomerate dispersion; however, further scientific studies are required on that topic, addressing other additive influences on other composite properties, like electrical and mechanical properties.
We thank Nanocyl S.A. (Sambreville, Belgium) for providing specially synthesized and prepared nanotubes as well as for supplying commercially available products. We also thank Dr Beate Krause for help in the analysis of MWCNT length based on TEM images generated by Regine Boldt (both IPF). Parts of the research were funded by the German Federal Ministry of Education and Research (BMBF) within the Framework Concept ‘Research for Tomorrow’s Production’ (funding number 02PU2392), managed by the Project Management Agency Karlsruhe (PTKA). We also acknowledge financial support from the project INTELTEX (Intelligent multi-reactive textiles integrating nanofiller based CPC-fibres) – a European Integrated Project supported through the Sixth Framework Programme for Research and Technological Development.
Baets, J., Godara, A., Devaux, J., Verpoest, I. Toughening of polymerized cyclic butylene terephthalate with carbon nanotubes for use in composites. Composites Part A-Applied Science and Manufacturing. 2008; 39:1756–1761.
Brühwiler, P.A., Barbezat, M., Necola, A., Kohls, D.J., Bunk, O., Schaefer, D.W., Pötschke, P. Comparison of quasistatic to impact mechanical properties of multiwall carbon nanotube/polycarbonate composites. Journal of Materials Research. 2010; 25:1118–1130.
Kasaliwal, G., Göldel, A., Pötschke, P. Influence of processing conditions in small-scale melt mixing and compression molding on the resistivity and morphology of polycarbonate-MWNT composites. Journal of Applied Polymer Science. 2009; 112:3494–3509.
Kasaliwal, G., Göldel, A., Pötschke, P., Pegel, S., Heinrich, G. Analysis of agglomerate dispersion mechanisms of multiwalled carbon nanotubes during melt mixing in polycarbonate. Polymer. 2010; 51:2708–2720.
Krause, B., Mende, M., Pötschke, P., Petzold, G. Dispersability and particle size distribution of CNTs in an aqueous surfactant dispersion as a function of ultrasonic treatment time. Carbon. 2010; 48:2746–2754.
Krause, B., Pötschke, P., Häussler, L. Influence of small scale melt mixing conditions on electrical resistivity of carbon nanotube-polyamide composites. Composites Science and Technology. 2009; 69:1505–1515.
Le, H.H., Ilisch, S., Prodanova, I., Radusch, H.J. Online electrical conductivity as a measure to characterize the rheological and thermodynamic effects on the carbon black dispersion in rubber compounds. Kautschuk Gummi Kunststoffe. 2004; 57:355–362.
Le, H.H., Kasaliwal, G., Ilisch, S., Radusch, H.J. Online measured electrical conductance in-process characterization of the dispersion behaviour of carbon nanotubes in rubber compounds. Kautschuk Gummi Kunststoffe. 2009; 62:326–333.
Leer, C., Carneiro, O.S., Covas, J.A., Maia, J.M., Van Hattum, F.W.J., Bernardo, C.A., Biro, L.P., Horvath, Z.E., Kiricsi, I. Dispersion of carbon nanotubes in polycarbonate and its effect on the composite properties. In: Vilarinho P.M., ed. Advanced Materials Forum III, Pts 1 and 2. Zurich-Uetikon: Trans Tech Publications Ltd; 2006:514–516. [1125–1130].
Lellinger, D., Xu, D.H., Ohneiser, A., Skipa, T., Alig, I. Influence of the injection moulding conditions on the in-line measured electrical conductivity of polymer-carbon nanotube composites. Physica Status Solidi B:Basic Solid State Physics. 2008; 245:2268–2271.
Levresse, P., Manas-Zloczower, I., Feke, D.L., Bomal, Y., Bortzmeyer, D. Observation and analysis of the infiltration of liquid polymers into calcium carbonate agglomerates. Powder Technology. 1999; 106:62–70.
Masuda, J., Torkelson, J.M. Dispersion and major property enhancements in polymer/multiwall carbon nanotube nanocomposites via solid-state shear pulverization followed by melt mixing. Macromolecules. 2008; 41:5974–5977.
McNally, T., Pötschke, P., Halley, P., Murphy, M., Martin, D., Bell, S.E.J., Brennan, G.P., Bein, D., Lemoine, P., Quinn, J.P. Polyethylene multiwalled carbon nanotube composites. Polymer. 2005; 46:8222–8232.
Miušík, M., Omastova, M., Krupa, I., Prokes, J., Pissis, P., Logakis, E., Pandis, C., Pötschke, P., Pionteck, J. A comparative study on the electrical and mechanical behaviour of multi-walled carbon nanotube composites prepared by diluting a masterbatch with various types of polypropylenes. Journal of Applied Polymer Science. 2009; 113:2536–2551.
Pegel, S., Pötschke, P., Petzold, G., Alig, I., Dudkin, S.M., Lellinger, D. Dispersion, agglomeration, and network formation of multiwalled carbon nanotubes in polycarbonate melts. Polymer. 2008; 49:974–984.
Prashantha, K., Soulestin, J., Lacrampe, M.F., Krawczak, P., Dupin, G., Claes, M. Masterbatch-based multi-walled carbon nanotube filled polypropylene nanocomposites: assessment of rheological and mechanical properties. Composites Science and Technology. 2009; 69:1756–1763.
Pujari, S., Ramanathan, T., Kasimatis, K., Masuda, J., Andrews, R., Torkelson, J.M., Brinson, L.C., Burghardt, W.R. Preparation and characterization of multiwalled carbon nanotube dispersions in polypropylene: melt mixing versus solidstate shear pulverization. Journal of Polymer Science Part B:Polymer Physics. 2009; 47:1426–1436.
Shiga, S., Furuta, M. Processability of electron-paramagnetic-res in an internal mixer. 2. Morphological-changes of carbon-black agglomerates during mixing. Rubber Chemistry and Technology. 1985; 58:1–22.
Skipa, T., Lellinger, D., Böhm, W., Saphiannikova, Alig, M., I. Influence of shear deformation on carbon nanotube networks in polycarbonate melts: interplay between build-up and destruction of agglomerates. Polymer. 2010; 51:201–210.
Takase, H. Evaluation and applications of dispersing carbon nanotube in the polymers. In: Hosokawa M., Nogi K., Naito M., Yokoyama T., eds. Nanoparticle Technology Handbook. 1st edn. Oxford: Elsevier; 2007:588–593.
Triebel, C., Katsikis, N., Stará, H., Münstedt, H. Investigations on the quality of dispersion of nanofillers in poly(methyl methacrylate) composites by creep-recovery experiments. Journal of Rheology. 2010; 54:407–420.
Villmow, T., Kretzschmar, B., Pötschke, P. Influence of screw configuration, residence time, and specific mechanical energy in twin-screw extrusion of polycaprolactone/multi-walled carbon nanotube composites. Composites Science and Technology. 2010; 70:2045–2055.
Villmow, T., Pegel, S., Pötschke, P., Wagenknecht, U. Influence of injection molding parameters on the electrical resistivity of polycarbonate filled with multi-walled carbon nanotubes. Composites Science and Technology. 2008; 68:777–789.
Villmow, T., Pötschke, P., Pegel, S., Häussler, L., Kretzschmar, B. Influence of twin-screw extrusion conditions on the dispersion of multi-walled carbon nanotubes in a poly(lactic acid) matrix. Polymer. 2008; 49:3500–3509.
Yamada, H., Manas-Zloczower, I., Feke, D.L. The influence of matrix viscosity and interfacial properties on the dispersion kinetics of carbon black agglomerates. Rubber Chemistry and Technology. 1998; 71:1–16.
1The interfacial energies were calculated by harmonic mean equation. Surface energies and their polar and dispersive parts were taken from http://www.surfacetension.de/solid-surface-energy.htm. The surface energy values of CNT are taken from Barber et al. (2004) and were assumed to be independent of temperature.