The MWNT/silicone composites were fabricated with two different kinds of MWNT bundles using mass production compatible three roll milling process. Density of MWNT bundles could be controlled by fabrication process of metal precursors. The difference of order of agglomeration of the MWNTs was turned out to be closely related with dispersion of the MWNTs in the composites. Though same composition of catalyst was used, catalyst powder made from gelation of the precursors followed by flame synthesis (FS) consisted of chunk-type particles, while that originated from spray of the precursor solution followed by thermal decomposition (STD) was composed of thin sheet-like particles. After CVD growth of MWNTs, the MWNT bundles were entangled to form large chunks for FS catalyst but they maintained with rod-like morphology for STD catalyst. Furthermore, individual bundle of STD-MWNTs also revealed lower density with more room inside the bundles, which resulted in the composite having higher electrical conductivity due to effective dispersion of STD-MWNTs in the composites. In this study, high electrical conductivity over 1,000 S/m was obtained with the composite of STD-MWNT/silicone. For the first time, direct correlation between morphology of CNT catalysts and physical property of CNT/polymer composite was demonstrated in an experimental manner.
Composite materials with polymer matrix and conducting fillers have long been investigated. The conducting fillers include carbon nanotube, graphene [1–3] and metallic materials [4,5]. Carbon nanotube (CNT) has high electrical conductivity and very low percolation threshold due to its one dimensional rod-type morphology . There are many factors known to affect conductivity and properties of CNT/polymer composite. Type of CNT, order of agglomeration of CNT bundle, catalyst to grow CNT, mixing process of CNT in the matrix, type of matrix and so on [7–27].
Many reports on electrical conductivity of CNT/polymer composite materials, including several review papers [7–9], have been published up to now. They can be categorized into several groups. The CNT network, composed of individual CNT coated with various polymers, showed high electrical conductivity. E. Muñoz et al. reported conductivity of 10,000~20,000 S/m with PLV SWNT/polyethyleneimine-spun fibers . Y. Ru et al. reported high conductivity of 1.7×106 S/m, almost reaching conductivity of pure CNT, with the composite of de-bundled double-wall CNTs by doping with chlorosulfonic acid . Nafion-coated SWNT network showed conductivity of 3200 S/m (SWNT 18wt.%) . Meanwhile, polypyrrole coated CNT networks presented relatively low conductivity of 21.5 S/m  and 1,600 S/m with the composite using in-situ polymerization . Some researchers used one directionally aligned CNTs. T. Souier et al.  acquired 350 S/m of conductivity by infiltrating epoxy into bundles of MWNTs. H. Peng and X. Sun  reported that the composite based on polystyrene and sulfonated poly(ether ether ketones) showed conductivity of 1,330 and 6,670 S/m using dropcasting polymer solutions onto CNT arrays. Use of doped CNTs could also improve electrical conductivities of CNT/polymer composites. The composite with boron doped MWNTs showed high conductivity of 10,000 S/m, 40 times higher than that with undoped MWNTs . In the same way, doping of SOCl2 into SWNTs raised electrical conductivity of the composite from 1,700 to 10,000 S/m .
In general, the composites having CNTs completely embedded in polymer matrix presented lower values of conductivity than polymer coated CNT networks having some pores in between the networks. For example, the composite with MWNTs embedded in polyaniline matrix presented conductivity of 770 S/m, while the network composed of polyaniline coated MWNTs showed 2,540 S/m . Therefore, high conductivity is rarely found for normal composite materials where CNTs are completely embedded in polymer matrix. The MWNT/epoxy composite , the MWNT/SU8 composite  and the graphitized CNT/polyoxymethylene composite  showed conductivities of less than 10 S/m. The poly-(dimethylsiloxane) composite having scaffolds of SWNT based nano-foams presented electrical conductivity over 100 S/m . The MWNT/poly(phenylene sulfide) composite  prepared using compression molding showed conductivity of ~1,000 S/m and the SWNT/poly-(paraphenylene vinylene) composite  also had ~1,000 S/m at SWNT content of 64 wt.%. For this case, maximum reported value of electrical conductivity reached ~3,000 S/m. The value was obtained for the SWNT/polyaniline composite  and the MWNT/poly (methylmethacrylate) composite having 40 wt.% of MWNTs .
II. Results and discussion
A. Properties of metal catalyst and fabricated MWNTs
The catalyst powders synthesized from flame synthesis (FS) and spray and thermal decomposition (STD) are presented in Fig 1(a) and 1(b), respectively. Majority of the former consisted of chunks whose size ranged from several micrometers to several tens of micrometers. On the other hand, majority of the latter was composed of thin sheets whose thickness was less than 100nm. Before thermal treatment, the precursor of FS catalyst was treated using gelation while that of STD catalyst was finely dispersed during spray process. Difference of synthetic process resulted in specific particle shape constituting the catalyst powder produced.
The MWNTs grown from flame-synthesized catalyst (FS-MWNTs) consisted of large chunks. The size of the chunks was measured from several tens of to several hundreds of micrometers. Each chunk was formed by being entangled with a large number of individual MWNT bundle as shown in Figure 2(a) and inset. On the other hand, The MWNTs from thermal decomposition synthesized catalyst (STD-MWNTs) consisted of numerous rod-type bundles whose lateral diameter ranged from several micrometers to several tens of micrometers as shown in Figure 2(b). On a large scale, the order of agglomeration was much severer for FS-MWNTs. Furthermore, MWNT density in individual bundle was also larger for FS-MWNTs than for STD-MWNTs when we compared insets of Figure 2(c) and 2(d). It means that order of agglomeration of the MWNTs used in this study was much severer for FS-MWNT not only on a large scale (collection of the bundles) but also on a small scale (individual bundle). It is noted that morphology of the catalysts was directly related with morphology and agglomeration of the MWNT bundles. In the long run, morphology of the catalysts strongly affected finalized physical property of the composite (electrical conductivity) since the percolation and conduction in the composite were controlled by de-bundling of the MWNT agglomerates.
The characteristics of fabricated MWNTs are analyzed using SEM, TEM, TGA, Raman and resistivity measurement system. Both MWNTs had a diameter of 10–20nm. Number of walls was 8~15 for FS-MWNTs and 9~13 for STD-MWNTs. As was described above paragraph, STD-MWNT bundle showed much lower density than that of FS-MWNT. Measured bulk density of STD-MWNT, 0.015 g/cm3, was about a half of FS-MWNT (0.034 g/cm3). Impurities in the MWNTs were mostly derived from metallic catalysts such as Fe, Co and Al. Content of the impurities was about 3% for both FS-MWNT and STD-MWNT. The STD-MWNT was thermally more stable as shown in Figure 2(e). During thermal heating, FS-MWNT started decomposition earlier and showed lower oxidation peak temperature (606.7°C) than that of STD-MWNT (657.5°C). Crystalline perfection was checked with Raman analysis presented in Figure 2(f). Crystallinity of STD-MWNT was also superior to that of FS-MWNT in that STD-MWNT revealed higher value of IG/ID (1.05) than that of FS-MWNT (0.92). In a consistent manner, conductivity of STD-MWNT powder was a little higher than that of FS-MWNT powder. As a whole, the STD-MWNT showed better physical properties than those of FS-MWNT, and STD-MWNT bundle showed lower population of MWNTs with more room among them, resulting in low bulk density.
B. Fabrication of MWNT/silicone composites
The MWNT/silicone composite materials were fabricated with two-step shearing process using planetary mixer and three roll mill (TRM). For TRM process, the number of passes and the pressure between the rolls were varied to obtain maximum electrical conductivity of the composites by optimizing alignment of the MWNTs in the composites. As an example, electrical conductivities of the composite containing 9.1wt.% MWNT are presented in Figure 3(a) as functions of number of pass and applied pressure. At applied pressure of 1.0 MPa, electrical conductivity increased as number of pass increased and reached maximum value at 6 passes, showing no further increase for more passes. At applied pressure of 2.0 MPa, same trend was observed until 6 passes but electrical conductivity decreased as number of pass increased further up to 15 turns. It was found that over-pressure applied on the composite induced re-aggregation of MWNTs and therefore degraded value of electrical conductivity.
With regard to the types of MWNTs used in this study, the composite having STD-MWNTs showed higher electrical conductivity than the composite with FS-MWNTs. After 6 passes of TRM process, the surface of the former showed large scaled roughness while that of the latter presented small scale of roughness as presented in Figure 3(b) and 3(c). For detailed investigation of the MWNTs in the composite after TRM process, the MWNTs were closely observed after removal of silicone matrix using chemical etching process as shown in Figure 3(d) and 3(e). Note that long range order was observed for STD-MWNT composite with long length of MWNTs while short range order was found for FS-MWNT composite with short length of MWNTs, as was the case with the surface morphologies of the composites in Figure 3(b) and 3(c). Therefore, long range order with longer CNTs seems to be attributed to high electrical conductivity of the composite with STD-MWNTs.
Van der Waals force between carbon nanotubes has been studied by many researchers [28–37]. The empirical approach based on the pairwise summation of interatomic Lennard-Jones (LJ) potentials adapted for graphitic structures has also been widely applied. In this study, we applied the pairwise summation of adapted LJ potentials and the method of the smeared-out approximation suggested by L.A. Girifalco [34,37] to evaluate the potential between two crossed MWNTs. The model potentials for the VDW interaction are based on empirical functions whose parameters are obtained from empirical fits to properties of the relevant CNT systems. In the case of MWNT interaction it was assumed that each pair of layers interacts as SWNTs and use summation over all pairs. In these calculations, we assumed that each MWNT consisted exactly of 11 walls since the MWNTs in this study had 8~15 walls and 11 walls could be chosen as an average value. For the calculations, it was known that only several outer shells of MWNTs play an essential role in the VDW interaction. According to the method of uniform curve, an approximation for the VDW potential can be expressed by multiplying minimum potential energy (left-hand side of equation 1) with uniform curve [31,33]. For MWNTs with 11 shells (average value),
where d, t and γ represent distance between two MWNTs, radius of MWNT and crossed angle between two MWNTs, respectively and C1= −2.161 eV•Å3, C2= 1.036 eV•Å4, d0=2.87Å, b=0.3. Van der Waals force can be delivered by differentiating equation (1) with respect to the distance between two crossed MWNTs (d).
The minimum potential energy (absolute value) increases as the radius of MWNT increases as shown in Figure 4(a). It is 33.1eV for t1=t2=100Å and the value increases to 49.7eV for t1=t2=150Å. Meanwhile, the potential decreases with increasing angle between two MWNTs (Figure 4(b)). For the MWNTs with t1=t2=120Å, it decreases from 116.2 for γ=20° to 39.7 for γ=90°. Van der Waals forces also increase as the radius of MWNT increases and decrease as the crossed angle between two MWNTs increases as presented in Figure 4(c). They increase as the distance between two MWNTs increases and reach zero point at d=~2.9Å, finally reaching their maximum values at d=~3.4Å. For γ=20°, the maximum value increases from 43.3nN for t1=t2=100Å to 54.1nN for t1=t2=150Å. Therefore, the MWNTs with a distance of 3.4Å should jump an activation barrier of 54.1nN to be separated each other. As previously mentioned with Figure 2, FS-MWNT bundles showed higher population of MWNTs with bulk density twice higher than that of STD-MWNT. It means that larger numbers of MWNTs in FS-MWNT/silicone composite have to jump the activation barrier to be separated and well dispersed in the MWNT/silicone composite. Therefore, through the TRM process, more energy should be applied to FS-MWNT/silicone composite to disperse the MWNTs well into silicone matrix. Under the condition of same pressure, STD-MWNTs were dispersed better into the composite and resulted in higher electrical conductivity of the MWNT/silicone composite.
C. Electrical conductivities of MWNT/silicone composites
Electrical conductivities of the MWNT/silicone composites, fabricated with two-step shearing process using planetary mixer and three roll milling, were measured for MWNT/silicone composites and presented in Figure 5(a) and 5(b). The conductivity as a function of MWNT content was measured after 6 passes of TRM process to acquire maximum conductivity for each composition of the composites. For the same MWNT content, the conductivities of STD-MWNT/silicone composites were always higher than those of FS-MWNT/silicone composites. For the conductivities in a direction parallel to MWNT alignment,
where Φ represents MWNT content in the composites. Percolation threshold for STD-MWNT/silicone composite (0.4%) was lower than that for FS-MWNT/silicone composite (0.9%). Well-dispersed STD-MWNT in the composite reached the percolation point with smaller amount of MWNTs. The percolation threshold of STD-MWNT/silicone composite was in a similar range with that of SWNT/polyaniline composite (0.3%) presenting conductivity of ~3,000 S/m . And it was smaller than that of SWNT/poly(paraphenylene vinylene) composite (1.8%) having conductivity of ~1,000 S/m . However, the exponents of MWNT/silicone composites in this study (~0.94) were smaller than that of SWNT/polyaniline composite (2.06~2.13) and that of SWNT/poly-(paraphenylene vinylene) composite (2.0). The exponent is known have a value of 1.1~1.3 for a two dimensional system whereas higher value of 1.6~2.0 was found for a three dimensional system. For the MWNT/silicone system fabricated using TRM process, electrical conductivities in a parallel direction to MWNT alignment, σ‖, was higher than those in a normal direction to MWNT alignment, σ⊥, as presented in Figure 5(b). For STD-MWNT/silicone composite, the ratio of σ‖/σ⊥ was 5.9 for MWNT 1.0wt.% and 1.8~2.0 for more than 8.0wt.% of MWNT content. Meanwhile, for FS-MWNT/silicone composite, the ratio was 3.5 for MWNT 1.0wt.%, 1.8~2.0 for around MWNT 9.0wt.% and finally 1.5~1.7 for more than 17.0wt.% of MWNT content. Therefore, the anisotropy of electrical conductivities was higher for STD-MWNT/silicone composite.
Maximum electrical conductivity of MWNT/silicone composite having specific MWNT content was decided following the conductivities in a direction parallel to MWNT alignment. Though overall arrangement of MWNTs in the composites including all directions of MWNT alignment should be three dimensional, measured conductivity for each direction could be rather two dimensional since it reflected unidirectional characteristics of MWNT alignment in the composite. It could be resulted in the value of the exponent (~0.94) for MWNT/silicone composite in this study. For further confirmation of the anisotropy of MWNT alignment in the composites, polarized Raman spectroscopy was conducted and presented in Figure 5(c) and 5(d). Intensities of measured Raman spectroscopy showed clear difference according to measuring direction. The ratio of Raman intensity parallel to MWNT alignment, IXX, to that normal to MWNT alignment, IYY, was higher for STD-MWNT/silicone composite than for FS-MWNT/silicone composite. The ratio, IXX/IYY, was measured as 2.78 for STD-MWNT/silicone composite while it was determined as 1.97 for FS-MWNT/silicone composite. The trend was in great consistency with measured electrical conductivities of MWNT/silicone composites. The STD-MWNT/silicone composite was fabricated with better dispersion of STD-MWNTs due to less agglomeration of STD-MWNT bundles, which resulted in long range order of MWNT alignment, low percolation threshold and high electrical conductivity. In the long run, the higher order of anisotropy of STD-MWNT alignment was expressed as characteristics such as Raman intensity and electrical conductivity. This study shows that electrical conductivity of CNT/polymer composite can be controlled by fundamental origin such as morphology of CNT catalysts and consequent density of CNT bundles grown from the catalysts.
The MWNT/silicone composites were fabricated with two different kinds of MWNT bundles using mass production compatible three roll milling process. Density of MWNT bundles could be controlled by fabrication process of metal precursors. Though same composition of catalyst was used, catalyst powder made from gelation of the precursors followed by flame synthesis consisted of chunk-type particles, while that originated from spray of the precursor solution followed by thermal decomposition was composed of thin sheet-like particles. After CVD growth of MWNTs, the MWNT bundles were entangled to form large chunks for FS catalyst but they stayed with rod-like morphology for STD catalyst. Furthermore, individual bundle of STD-MWNTs also revealed lower density with more room inside the bundles, which resulted in the composite having higher electrical conductivity due to effective dispersion of STD-MWNTs in the composites.