Mechanical and dielectric properties of epoxy-clay nanocomposites A. Guevara-Morales1,2* and A.C. Taylor1 1. Department of Mechanical Engineering, Imperial College London, South Kensington Campus, London SW7 2AZ, UK. 2. Currently at Department of Mechatronics Engineering, ITESM-CEM, Atizapán de Zaragoza, Estado de México 52926, México. * Corresponding author. A. Guevara-Morales. Tel.: +52 5558645555. Email: a.guevaram@itesm.mx A.C. Taylor email: a.c.taylor@imperial.ac.uk Mechanical and dielectric properties of epoxy-clay nanocomposites Abstract Epoxy-clay nanocomposites were prepared using two types of surface-treated montmorillonite (Closite 30B and Nanomer I28E). Wide angle X-ray scattering showed that all the nanocomposites had an intercalated structure. Improvements in tensile and fracture properties were found. The pure epoxy polymer was very brittle with a fracture energy, Gc, of 131 Jm-2. The addition of the nanoclays significantly increased the value of Gc, up to 240 Jm-2 for 5wt% C30B. The toughening mechanisms acting in the nanocomposites were identified using scanning electron microscopy as crack deflection and plastic deformation of the epoxy matrix around the clay platelets following debonding. From electrical testing, the permittivity and loss angle of the nanocomposites decreased and their breakdown strength increased as desired for insulation applications. The breakdown strength of the pure epoxy was found to be 11.7 kVmm-1, while for a 2 wt% C30B nanocomposite it increased to 14.7 kVmm-1. It was concluded that the restriction of chain mobility inhibited electrical polarisation and thus decreased the permittivity and loss angle. The electrical damage zone was analysed using scanning electron microscopy. It was found that the higher resistance to surface degradation by partial discharges and the creation of a tortuous electrical path, which delayed the propagation of the electrical tree, were the main factors which improved the breakdown strength of the nanocomposites. Keywords: Epoxy-clay nanocomposites; montmorillonite; toughening; dielectric properties; breakdown strength; electrical tree; partial discharges. 1. Introduction Thermosetting epoxy resins are widely used for many applications, including as high strength adhesives, surface coatings, durable laminates and lightweight foams [1]. They are also indispensable materials for insulation systems in power equipment and heavy apparatus such as dry type transformers, rotating electrical machines and switchgear [2]. During the operation of such power equipment, the insulator is subjected not only to electrical stress, but must withstand thermal and mechanical stresses as well as tribological loads [3]. In recent years, polymer nanocomposites reinforced with surface-treated montmorillonite nanoclays have received significant attention due to the substantial improvements they can achieve in mechanical, thermal and barrier properties when compared to the unmodified polymer [4-6]. Furthermore, studies have reported [7,8] that the addition of small amounts of nanofillers can improve the resistance to partial discharges and electrical treeing growth of the polymer matrix. The selection of layered silicate nanoclays as reinforcements is highly attractive because of their relatively low cost, high thermal inertness, ready availability and environmentally friendly characteristics [9]. Significant decreases in permittivity and loss angle have been reported [10,11] with the addition of small amounts of nanoparticles (~5wt%), which are attributed to the restriction of polymer chain mobility by the nanoparticles. This restriction inhibits electrical polarisation because the material loses its freedom to relax under an applied voltage, and thus the permittivity and loss angle decrease. It has also been reported [11] that at higher clay contents the values of permittivity increase. It has been suggested [11] that this is due to the presence of accidental inclusions, imperfections or inhomogeneities in the nanocomposite that are technically difficult to avoid. It has also been reported [12] that epoxy-clay nanocomposites show better AC and DC breakdown strength. These improvements are usually related to the electrical tree growth behaviour in the nanocomposites. It has been suggested [9] that the nanoparticles act as scattering sites that decelerate the electrons and carriers and hence increase the breakdown voltage. Other authors [11] suggest that the presence of nanoparticles changes the charge distribution of the material, acting as trapping sites and decelerating the propagation rate of the electrical tree. This study investigates the effect of the addition of commercially available nanoclays has on the mechanical, fracture and dielectric properties of an epoxy polymer. Scanning electron microscopy is used to investigate the toughening and failure mechanisms. 2. Experimental 2.1 Materials The epoxy used was a digylcidylether of bisphenol A (DGEBA) (Araldite MY750, Huntsman), with an anhydride hardener (Aradur HY917, Huntsman) and imidazole accelerator (DY070, Huntsman), which were mixed in proportions of 100:90:1 by weight. Two organo-modified montmorillonite (MMT) clays were used, as shown in Table 1; the octadecyltrimethylammonium-treated Nanomer I28E supplied by Nanocor, and the methyl, tallow, bis-2-hydroxyethyl, quaternary ammonium-treated Cloisite 30B supplied by Southern Clay Products. These were selected by performing compatibility tests on a range of commercially-available clays, which showed the relative stability of the dispersion of the clays in the DGEBA. Table 1. Organo-modified montmorillonite clays used. Clay Particle size Supplier Modifier Cloisite 30B <10µm Southern Clay Products MT2EtOH: methyl, tallow, bis-2-hydroxyethyl, quaternary ammonium; Nanomer I28E 15-20 µm Nanocor ODTMA: Octadecyltrimethylammonium 2.2 Preparation of Nanocomposites Nanocomposites were prepared by incorporating 1-5 wt% of clay into the epoxy system described above. The required amount of epoxy resin was poured into a beaker and heated to 60 °C to reduce its viscosity. The dried clay was sieved using a 300 m mesh to remove any agglomerates, and gently added to the warm epoxy. The mixture was vigorously hand-stirred using a spatula for a few minutes. An opaque, foamy and viscous mixture was formed. The mixture was then degassed in a vacuum oven for 10 minutes at 60 °C. The hardener and accelerator were added, and the mixture was again hand-stirred with the spatula. The mixture was then placed in the vacuum oven at 60 °C for 1 hour to degas, producing a final transparent solution. The degassed mixture was poured into a steel mould coated with release agent (Frekote 700NC, Henkel), and cured at for 2 hours at 60 °C, followed by 10 hours at 110 °C. The mould was left in the oven to cool slowly to room temperature, and the cured plate removed. 2.3 Characterisation Wide angle X-ray scattering (WAXS) was performed using a Philips PW1700 automated powder diffractometer equipped with CuK radiation (wavelength 1.54 Å) operating at 40 kV and 40 mA with a secondary graphite crystal monochromater. Diffraction patterns were collected between 3° and 25° with a step size of 0.04° and scanning speed of 0.48°/min. Clay samples were prepared by mounting and pressing the clays into an aluminium holder with a glass back support. Nanocomposite samples were prepared in moulds of 50 x 45 x 2 mm for X-ray scattering. The glass transition temperature, Tg, was measured using a Triton TT Dynamic Mechanical Analyser. Moulded prismatic samples of 45 x 3 x 2 mm were loaded under dual cantilever bending mode. Scans were performed over a temperature range from 40 to 180 °C under controlled sinusoidal strain, at a frequency of 1 Hz and a heating rate of 2.5 °C/min. The Tg was determined from the position of the peak of the tan curve, and the typical standard error was ± 1%. 2.4 Mechanical Testing Tensile dumbbell specimens were machined from the moulded plates and tested at room temperature according to the BS ISO 527-2 standard (test specimen 1A). The tests were performed at a constant displacement rate of 1 mm/min. Three replicate samples per formulation were used. The elastic modulus, E, and tensile strength were calculated. Fracture specimens were machined from moulded samples and tested according to the ISO 13586 standard at room temperature of dimensions 70 x 15 x 5 mm. Single-edge notch bend (SENB) tests were performed at a constant displacement rate of 1 mm/min. An initial crack was created by tapping a razor blade into the machined notch tip to generate a natural crack which was approximately half the height of the specimens before testing. Six replicate samples per formulation were used. The fracture toughness, Kc, was calculated and the fracture energy, Gc, was obtained from Kc according to the standard. 2.5 Electrical Testing Electrical tests were performed according to the IEC 60243-1 (1998) standard. Values of permittivity, loss angle, resistivity and breakdown strength were obtained. A plate-plate electrode geometry was used. Tests were carried out at a power frequency of 50 Hz, and at ambient conditions (19 °C; 55% R.H.). 2.6 Scanning Electron Microscopy A JEOL JSM-5300 scanning electron microscope was used to examine the fracture surfaces and to characterise the electrical trees. The samples were sputter-coated with a thin layer of gold before observation to eliminate charging. For the visualization of the electrical trees, a technique proposed by Vouyovitch [13] was used. This consists of making a straight saw cut on each side of the damage zone, purposely avoiding the damage, and then bending the plate at room temperature to break it so that the crack propagates through the damage zone. 3. Results and Discussion 3.1 X-Ray Diffraction The (001) basal reflections measured by WAXS were used to obtain the interlayer spacing of the clays. The results are summarised in Table 2. The (002) basal reflections of the 5 wt% nanocomposites were used to obtain the interlayer spacing of the clays in the epoxy matrix. The nanocomposites showed a further increase in interlayer spacing, as expected due to the intercalation of the epoxy between the clay platelets [14]. Both nanocomposites containing 5 wt% of organo-modified montmorillonite showed similar d-spacings. Table 2. Interlayer spacings of clays and nanocomposites, and glass transition temperatures of epoxy and nanocomposites. As-received clay 5 wt% Nanocomposite Tg d-spacing (Å) d-spacing (Å) [°C] Epoxy N/A N/A 139.3 C30B 17.7 32.5 139.8 I28E 22.6 34.5 145.0 Sample 3.2 Dynamic Mechanical Analysis The glass transition temperature, Tg was determined from the peak of the tan curve at a frequency of 1 Hz, and the results are presented in Table 2. A Tg of 139.3 °C was measured for the unmodified epoxy, values of 139.8 °C and 145.0 °C were found for the C30B and I28E nanocomposites respectively. The glass transition temperature can vary with various factors such as the molecular weight, crosslink density, free volume, and the properties of the interface layer [6]. Some authors [9] have reported an increase in Tg with the addition of clays, which is attributed to the good adhesion of the polymer matrix and the particles acting as restrictions for chain mobility. Others [14,15] have reported an initial increase in Tg up to a certain amount of clay when Tg starts decreasing. This decrease in Tg was related to a change in chain kinetics in the regions surrounding the nanoparticles, which leads to a lower crosslink density [6]. It is observed that Tg increases with increasing d-spacing, as the Tg for the 5 wt% I28E nanocomposite (dspacing: 22.6 Å) is significantly higher than that for the C30B (d-spacing: 17.7 Å), see Table 2. This is an indication that the clay platelets may be acting as a restriction to polymer chain mobility, and that the more intercalated or exfoliated the clays are, the more restriction they generate. The increase in Tg for both nanocomposites indicates that there is good adhesion between both clays and the epoxy matrix. 3.3 Mechanical Properties Tensile tests results are summarised in Table 3. It was found that the modulus of the epoxy resin (3.53 GPa) increased slightly when nanocomposites were formed (3.58-3.66 GPa). The exception was the 5 wt% I28E nanocomposite, where the modulus was decreased, which can be attributed to voids or defects in the material. The standard deviations of these data vary from 2 to 6 % of the mean value. It has been shown that the elastic modulus of a polymer is greatly improved when nanocomposites are formed based on layered silicates [16]. This improvement in elastic modulus is attributed mostly to the much higher stiffness of the clay particles, but also to the restriction the mobility of the polymer chains under load and by the clay platelets [17-19]. Orientation of the clay platelets along the loading direction will also contribute to the increase in elastic modulus [19]. However, Kinloch and Taylor have shown that the orientation of the platelets in such cast plates is not strong [20], and so a random orientation can be assumed. Using the Halpin-Tsai model [20], considering a random orientation of the clay platelets, an elastic modulus of 3.62 GPa was predicted, which agrees well with the measured values. For the calculations a clay modulus of 172 GPa [20] and an aspect ratio of 15 [21] were used. Table 3. Elastic modulus, tensile strength, fracture toughness and fracture energy of epoxy and its nanocomposites. Mean and standard deviation (SD) of the data are shown. Sample Epoxy 2wt% C30B 5wt% C30B 5wt% I28E E [GPa] Mean SD 3.53 E 0.14 3.58 0.03 3.66 0.03 3.37 0.09 Tensile Strength [MPa] Mean SD Tensile strength 46.5 6.1 54.2 5.6 54.3 8.3 50.3 7.4 Kc [MPam1/2] Mean SD 0.68 Kc 0.05 0.92 0.95 0.83 0.08 0.05 0.04 Gc [Jm-2] Mean 131 207 240 179 Similar results as for the elastic modulus were found for the tensile strength. As shown in Table 3, all the nanocomposites have a higher tensile strength (50.3-54.3 MPa) than the epoxy matrix (46.5 MPa). Note that for brittle materials such as these, the tensile strength is strongly dependent upon the surface finish of the specimens [20], and that the standard deviations of these data were approximately 10-15% of the mean value. The load versus displacement behaviour was typically linear elastic, with no evidence of yielding. The fracture toughness, Kc, of the epoxy and its nanocomposites were obtained from the SENB tests. A fracture toughness of 0.68 MPam1/2 was measured for the epoxy, which is typical for such a crosslinked polymer [22]; and an increase was found for all its nanocomposites, with fracture toughness from 0.83 to 0.92 MPam1/2, as shown in Table 3. The epoxy resin has a fracture energy, Gc, of 131 Jm-2, which is typical of a brittle thermosetting polymer [23]. An increase in Gc was found for all the nanocomposites. The addition of 5 wt% of C30B almost doubled the fracture energy of the unmodified epoxy to 240 Jm-2. It has been reported that the addition of nanoclays has a toughening effect on the epoxy resin, which is attributed to crack deflection and plastic deformation around the nanoparticles [20,24]. Scanning electron microscopy (SEM) was used to confirm whether these mechanisms were present for the nanocomposites, and the obtained micrographs are shown in Fig. 1. The fracture surface of the epoxy is shown in Fig. 1a. It is a smooth and glassy surface, which is typical of the fracture surfaces of crosslinked thermosets, and shows that little plastic energy dissipation accompanied the fracture process [25]. The fracture surfaces of the nanocomposites are shown in Fig. 1b-d. These surfaces are rougher than the epoxy matrix, and as the clay content increases, the roughness of the surface also increases, which is in agreement with the increased Gc values reported in Table 3, as rougher fracture surfaces indicate more plastic energy dissipation. One possible toughening mechanism is explained by the crack deflection model proposed by Faber and Evans [26], which assumes that as the crack approaches a particle in a material it is deflected by the particle. Crack deflection will increase the roughness of the surfaces through tilting and twisting of the crack front, increasing the fracture surface area and thus increasing the fracture energy of the material. In the crack deflection process modes I and II occur, and in the crack twisting process modes I and III are present. From fracture mechanics it is known that more energy is required to cause fracture in modes II and III than in mode I [21]. It is important to mention that the crack deflection theory cannot fully explain the toughening effect in the nanocomposites unless the particles are larger than the crack opening displacement [26], which is the case for these nanocomposites [27]. It can also be seen from Fig. 1 that the roughness of the surfaces depends on the type of clay, and that the 5wt% C30B nanocomposite has a rougher surface when compared to the 5 wt% I28E nanocomposite, which is also in agreement with the obtained Gc values. Faber and Evans [26] also concluded that the aspect ratio of the particles has a significant effect on the toughening mechanisms of particulate composites, which may explain the difference in fracture energy for the C30B and I28E nanocomposites. Note that according to the manufacturers’ data (Table 1), the particle size of the C30B clay (10 µm) is smaller than the I28E clay (15-20 µm), however, the I28E interlayer spacing is greater (Table 2), which will reduce the aspect effective ratio of the intercalated particles. The other toughening mechanism is debonding followed by plastic deformation of the epoxy around the nanoparticles (i.e. void growth) [20]. Debonding of the clay platelets from the epoxy leads to the formation of holes. These holes then grow plastically due to the triaxial stress state in the plastic zone at the crack tip, absorbing energy and increasing the toughness. In Fig. 1b-d, these holes are visible as black regions. Thus, it is concluded that the plastic deformation of the epoxy matrix around the debonded particles is another toughening mechanism in these nanocomposites. 3.4 Electrical Properties The permittivity, loss angle and resistivity values of the epoxy and its nanocomposites are presented in Table 4. The permittivity and loss angle measurements of the 5 wt% I28E nanocomposite were very unstable and reliable values could not be recorded. Table 4. Dielectric properties and breakdown strength of epoxy and its nanocomposites. Permittivity Resistivity [.m]x1014 Mean SD Mean SD Epoxy 1.87 0.57 5.30 0.57 1wt% C30B 1.87 0.13 5.25 0.07 2wt% C30B 2.83 1.36 5.35 0.07 5wt% C30B 7.13 2.28 5.27 0.18 5wt% I28E 7.29 0.51 N/D SD: Standard Deviation; N/D: Not determined Sample Loss angle Mean SD 0.00371 0.0001 0.00295 0.0002 0.00309 0.0013 0.00471 0.0031 N/D - Breakdown Strength [kV/mm] Mean SD 11.7 0.8 13.4 2.5 14.7 0.5 13.0 0.3 13.5 1.3 Although inorganic fillers have a higher permittivity values than the polymer matrix, in the range of 5 to 8 [10,28] compared with the 5.30 value for the epoxy (Table 4), the permittivity usually decreases in nanocomposites when compared to the unmodified polymer [9,11,29]. This reduction is related to the restriction of polymer chain movement by the nanoparticles. Here the material loses its freedom to relax under an applied voltage, and the polymer chains are not able to contribute to the electrical polarization [30]. However, the higher values of permittivity which were measured, such as for the 2 wt% C30B, are related to inhomogeneities or defects in the nanocomposites (e.g. voids, air bubbles) or agglomeration of the nanoparticles in the polymer matrix [11]. The loss angle also decreased in the nanocomposites. The loss angle is sensitive to the existence of mobile ionic impurities and polar radicals with dipole moments [11], and its reduction is related to the restriction in polymer chain motion, like the decrease in permittivity. From the electrical tests it was found that the breakdown strength of all the nanocomposites was higher than that of the epoxy matrix. The breakdown strength is defined as the maximum electric field strength that an insulating material can withstand intrinsically without breaking down [31]. The term breakdown is used to describe processes in which a considerable current increase results from a small voltage change, and in the case of polymeric insulators it is always catastrophic, irreversible and destructive, creating a breakdown channel between the electrodes [7]. This increase in breakdown strength is attributed to a broadening of the electrical tree and a decrease in the tree propagation rate that may arise from the resistance to partial discharges, which will be discussed in the following section. 3.5 Resistance to Partial Discharges To explain the increases in performance, it is necessary to consider the mechanisms due to the nanoclays within the breakdown process. It has been found that the addition of organo-modified clays improves the resistance to surface degradation by partial discharges [32]. This improvement can be explained with the multi-core model proposed by Tanaka [28], which is based on the proposal that the interface is divided in three layers: the bonded layer, the bound layer and the loose layer (Fig. 2); and that an electric double layer is superimposed on all of them. The bonded layer is in contact with the particles and it consists of immobile polymer chains bonded to the particle surface. The bound layer is adjacent to the bound layer and consists of less ordered polymer chains. Finally, there is the loose layer, which is considered to be an amorphous region with polymer chains which are relatively free to move. The model suggests that the electrical tree is attracted to the particle due to its higher permittivity when compared to the epoxy matrix. When the electrical tree reaches the loose layer of the interface it propagates through it. However, when it reaches the stronger bound layer it cannot degrade this bound layer so easily and so the electrical tree prefers to crawl along the loose layer until it reaches the epoxy matrix again. Due to the close packing of the particles and the small interfiller distance, the electrical tree is rapidly attracted by another particle, repeating the process and creating a tortuous or zigzag path that will increase the breakdown strength (see Fig. 2). Thus the role of the nanoclays is not exactly as a barrier, as they attract the partial discharge due to their higher permittivity, but due to the higher partial discharge resistance the partial discharge prefers to crawl along the interface, increasing the length of the tree and delaying its propagation. 3.6 Electrical Treeing Electrical treeing is an electrical pre-breakdown phenomenon in which fine erosion channels propagate through the insulator material, in a path resembling tree branches, due to partial discharges. The process can be divided into three stages: inception, propagation, and runaway or final breakdown, which are explained in detail by Dissado and Forthergill [7]. The inception of electrical trees involves bond breaking and relies on electrons or carriers whose kinetic energies are sufficient to cause bond scission (3 to 4 eV), but lower than that required for ionization (8 eV), that can cause electrical breakdown. According to the partial discharge mechanism, the main step in tree inception is the establishment of a path consisting of voids or defects aligned between the electrodes. After the inception of the tree (the formation of the first branch), the electrical tree propagation exhibits a decelerating growth rate that is associated with the branch tip splitting (branching of the tree), extinction of channel discharges and the lack of a vent to release the gases produced by the discharges. Note that breakdown is not always an instantaneous process after the first branch reaches the opposite electrode; it usually requires that the main branch or channel reaches a certain diameter. During the creation of the first branch, electrons or carriers are extracted from the matrix, leading to the formation of positively charged bands along the channel length. When these positive bands reach a certain charge density, they produce a local field which is sufficiently high to attract and neutralise the electrons. At this stage there is not forward erosion, but a lateral one that raises the possibility of branching and widening of the channel. At the runaway stage the main channel becomes so wide that charge repulsion will force the cations in the walls to run through it acting as a conducting projection, higher kinetic energies will be reached, causing the irreversible failure of the insulator. It has been found that cracking is almost instantaneous once large discharges occur. The faster the inception and the propagation, the faster the final breakdown will be, and hence the lower the breakdown strength of the material. Scanning electron microscopy (SEM) was used to examine the electrical trees in the tested samples, and images are presented in Fig. 3. The electrical tree branches are clearly observed in the nanocomposites, extending all over the damage zone (Fig. 3b-e). In the case of the epoxy sample (Fig. 3a), the main channel, indicated by the black arrow, occupies a great part of the damage zone with a diameter of approximately 400 m at the top and 1000 m at its centre. The diameter of the main channel in the nanocomposites is smaller, being approximately 250 m in all cases. It is important to notice that the electrical trees in these samples do not have the common “tree” shape with branches radiating from a point towards the opposite side (see schematic in Fig. 3). Some of the branches are oriented towards the top surface while others are oriented towards the bottom surface. This is an effect of the electrode configuration. A needle-plate electrode configuration is commonly used when the electrical tree is to be analysed. When the applied AC voltage generates a current from the top to the bottom surface, it propagates from the tip of the needle to the opposite plate. During the other half of the cycle, the current is reversed, but the path will be the same because the electrons are attracted by the tip of the needle. This generates the classic tree shape. When a plate-plate electrode configuration is used, as in the present work, the electrons are not attracted by a single point (Fig. 3c-d) but to a larger area, and therefore the tree propagates towards both sides. It is important to note that with the needle-plate electrode configuration the tree will be initiated at the tip of the needle, whereas in the case of plate-plate electrode configuration it will initiate in the vicinity of the major flaw [33], and hence plate-plate electrodes give a more realistic value for the breakdown strength. From Table 4 it can be seen that the addition of the organo-modified clays improved the breakdown strength of the epoxy, which may indicate that both the inception and propagation stages of treeing were delayed with the addition of the clays. The inception stage is determined by the yield strength of the material and by the free volume [7]. According to the results of tensile strength presented in Table 3, all the nanocomposites have a higher tensile strength than the epoxy, indicating that the inception stage may be delayed and hence the breakdown strength of the material is increased. As the behaviour of the epoxy resin and its nanocomposites is purely linear elastic, the tensile strength was considered as equivalent of the yield strength. Although the free volume in the epoxy and its nanocomposites were not analysed in this study, the literature reports an increase of between 2 and 3 % in the free volume for nanocomposites [11,34], which may indicate a decrease in breakdown strength due to the larger space where the electrons will be able to move and acquire energy for bond scission. However, the increase in Tg with the addition of nanoclays (Section 3.2) may be considered as a reduction in free volume, as they act as a restriction to the mobility of the electrons. An important consideration here is that the increase in free volume is mainly localised at the interface of the nanoparticles [28] and when the accelerated electrons or carriers move through the interface they can be decelerated via collision with the inner layers of the interface, resulting in an increase in breakdown strength. The electrons will not be able to accelerate again due to the small interparticle distance. As mentioned above, during the propagation stage the positively charged main channel attracts the electrons or carriers to its walls, producing a lateral erosion that increases the possibility of branching and widening of the channel [7]. In the case of the epoxy, this lateral erosion increases the diameter of the main channel, as observed in Fig. 3, due to the material’s low resistance to surface degradation caused by partial discharges. When the main channel becomes so wide, charge repulsion will force the cations in the walls to run through it acting as a conduction projection, reaching kinetic energies high enough to cause bond scission, degrade the polymer wall and cause the breakdown of the material. Figure 4 shows the inner surfaces of the electrical tree main channel for the epoxy and its 5 wt% I28E nanocomposite. Note the significant difference between the rough channel surface for the nanocomposite and the smooth surface for the epoxy sample. This can be explained by the surface degradation process occurring in the epoxy matrix and its nanocomposites. Erosion due to partial discharges is considered to occur at the same time in both specimens (pure epoxy and its nanocomposite), but in the nanocomposites the nanoparticles offer more resistance to partial discharges, as explained above. Therefore, the nanoparticles are exposed on the surface, which gives the rougher and whiter appearance in the SEM images. This layer of exposed nanoparticles will impede further erosion of the surface due to the higher erosion resistance of the clay platelets [32] and decelerate the widening of the tree channels [8], slowing down the speed of widening of the channels that lead to the final breakdown. Some cracks in the walls of the main channel are observed in Fig. 4a. It is thought that due to the lower fracture toughness of the epoxy resin (see Table 3) the lateral impacts of the electrons in the walls of the epoxy matrix initiated cracks in the surface of the channel. These cracks then acted as free volume sites [7] and allowed the branching of the tree. Note that the breakdown strength is not determined by the number of branches, but by the ease with which those branches propagate. In the case of the epoxy, the low resistance to surface degradation will facilitate the propagation of the tree. In Fig. 5a the electrical tree branches of the 1 wt% C30B nanocomposite are shown at a higher magnification. Although these branches may be seen as an effect of the fracture surface rather than branches of an electrical tree, the damage zone (where the branches are) is actually part of a crater or depression, as can be observed in Fig. 3c. Therefore, the branches have not been affected by the way the samples were prepared for SEM observation, as cutting the damage zone was carefully avoided (Section 2.6). In Fig. 5b a depression is clearly observed, which is the boundary between the crater of the electrical damage zone and the fracture surface created when the samples were prepared for microscopy. It has been suggested that other factors such as the state of stress and the thermal properties of the nanocomposites can contribute to the enhancement of breakdown strength [33,35]. Due to the smaller thermal expansion coefficient and lower linear shrinkage ratio of the nanocomposites when compared to the unmodified epoxy, their state of stress is lower as shown by Imai et al. [35]. This lower state of stress may also contribute to the enhancement of the breakdown strength [7]. In the case of the thermal properties, it has been reported that the addition of organo-modified clays allows a better dissipation of the heat generated by the applied electric field [33], decreasing the possibility of thermal breakdown. In future, a more detailed investigation of the damage zone is required, especially at the inception and propagation stages (i.e. when the final breakdown has not occurred yet). This can be done by testing the specimens at different voltages and for different times in order to be able to determine a sequence of failure mechanisms that are present in the nanocomposite electrical degradation. The samples presented in this study were tested until the final breakdown occurred, and therefore the damage zones shown in Figs. 3-5 present multiple failure mechanisms such as electrical breakdown and fracture. This multiple-failure damage zone complicates the visualisation of the electrical tree. If the main interest is on the electrical treeing process, it is recommended to change the configuration of the electrical tests from a plate-plate electrode configuration to a needle-plate configuration. This will produce a divergent field instead of a quasi-homogeneous one and the tree will grow more clearly. 4. Conclusions Epoxy-clay nanocomposites were prepared using two organo-modified montmorillonites. The elastic modulus of the nanocomposites was higher than that of the epoxy matrix, due to the much higher stiffness of the clay when compared to that of the epoxy, and to the restriction of polymer chain mobility due to the presence of the clay. Results showed that the fracture energy of the epoxy (Gc = 131 Jm-2) was increased for the nanocomposites (to 180-240 Jm-2). The toughening mechanisms acting in the nanocomposites were identified as crack deflection and debonding followed by plastic deformation around the clay platelets. The permittivity and loss angle of the epoxy, measured as 5.30 and 0.0037 respectively, decreased in the nanocomposites, reaching 5.25 and 0.0029 for the 1 wt% C30B nanocomposite. As these two properties are related to the electrical polarisation, it is likely that the restriction of chain mobility inhibited this polarisation and thus decreased the permittivity and loss angle as desired for insulator materials. An increased breakdown strength for all the nanocomposites was found, with the epoxy having a breakdown strength of 11.7 kVmm-1, and its nanocomposites having values from 13.0 to 14.7 kVmm-1. This improvement was due to the nanocomposites’ higher resistance to surface degradation by partial discharges, which delayed the propagation of the electrical tree. The clay particles attract the electrical tree due to their higher permittivity, but due to their higher degradation resistance the electrical tree prefers to crawl along the weaker layer of the interface (loose layer) until it reaches the epoxy matrix that is easier to erode. As the interparticle distance in nanocomposites is very small, the electrical field is attracted by another particle very soon and the process is repeated. This zigzag path increases the length of the tree and delays its propagation. Acknowledgements Dr. Guevara-Morales gratefully acknowledges the financial support from the Mexican National Council for Science and Technology (CONACyT). Some of the equipment used in this work was provided by Dr. Taylor’s Royal Society Mercer Junior Award for Innovation. References [1] Liu HY, Wang GT, Mai YW and Zeng Y (2011) On fracture toughness of nano-particle modified epoxy, Compos Part B 42(8):2170-2175. [2] Lee H and Neville K (1957) Epoxy Resins: Their Applications and Technology, McGraw-Hill Book Company Inc, USA. [3] Veena MG, Renukappa NM, Suresha B and Shivakumar KN (2011) Tribological and Electrical Properties of Silica-Filled Epoxy Nanocomposites, Polym Composite 32(12):2038-2050. [4] Azeez AA, Rhee KY, Park SJ and Hui D (2013) Epoxy clay nanocomposites – processing, properties and applications: A review, Compos Part B 45(1):308-320. [5] Jumahat A, Soutis C, Abdullah SA and Kasolang S (2012) Tensile properties of nanosilica/epoxy nanocomposites, Procedia Eng 41:1634-1640. [6] Yasmin A, Luo JJ, Abot JL and Daniel IM (2006) Mechanical and thermal behavior of clay/epoxy nanocomposites, Compos Sci Technol 66(14):2415-2422. [7] Dissado LA and Fothergill JC (1992) Electrical Degradation and Breakdown in Polymers, IEE Materials and Devices Series 9, Peter Peregrinus Ltd, UK. [8] Tanaka T, Matsunawa A, Ohki Y, Kozako M, Kohtoh M and Okabe S (2006) Treeing phenomena in epoxy/alumina nanocomposite and interpretation by a multi-core model, IEEJ Trans 126(11):1128-1135. [9] Sarathi R, Sahu RK and Rajeshkumar P (2007) Understanding the thermal, mechanical and electrical properties of epoxy nanocomposites, Mat Sci Eng A-Struct 445-446:567-578. [10] Tanaka T, Montanari GC and Mülhaupt R (2004) Polymer Nanocomposites as Dielectrics and Electrical Insulation-perspectives for Processing Technologies, Material Characterization and Future Applications, IEEE Transaction on Dielectrics and Electrical Insulation 11(5):763-784. [11] Tanaka T (2005) Dielectric Nanocomposites with Insulating Properties, IEEE Transaction on Dielectrics and Electrical Insulation 12(5):914-928. [12] Nelson JK and Fothergill JC (2004) Internal charge behaviour of nanocomposites, Nanotechnology 15(5):586-595. [13] Vouyovitch L, Alberola ND, Flandin L, Beroual A and Bessede JL (2006) Dielectric Breakdown of Epoxy-Based Composites: Relative Influence of Physical and Chemical Aging, IEEE Transaction on Dielectrics and Electrical Insulation, 13(2):282-292. [14] Triantafillidis CS, LeBaron PC and Pinnavaia TJ (2002) Thermoset Epoxy-Clay Nanocomposites: The dual role of , -diamines as clay surface modifiers and polymer curing agents, J Solid State Chem 167(2):354-362. [15] Sun Y, Zhang Z, Moon K and Wong CP (2004) Glass transition and relaxation behavior of epoxy nanocomposites, J Polym Sci 42(21):3849-3858. [16] Ray SS and Okamoto M (2003) Polymer/layered silicate nanocomposites: a review from preparation to processing, Prog Polym Sci 28(11):1539-1641. [17] Yasmin A, Abot JL and Daniel IM (2003) Processing of clay/epoxy nanocomposites by shear mixing, Scripta Mater 49(1):81-86. [18] Velmurugan R and Mohan TP (2004) Room temperature processing of epoxy-clay nanocomposites, J Mater Sci 39(24):7333-7339. [19] Sarathi R, Sahu R, Kumar P and Tanaka T (2006) Understanding the performance of epoxy nanocomposites – a physico-chemical approach, IEEJ Trans 126(11):1112-1120. [20] Kinloch AJ and Taylor AC (2006) The mechanical properties and fracture behaviour of epoxy-inorganic micro- and nano-composites, J Mater Sci 41(11):3271-3297. [21] Tarrant A (2004) Themoset-Acrylic/Layered-silicate Nanocomposites: Synthesis and Structure Property Relationships. PhD Dissertation, Imperial College London. [22] Williams JG (1984) Fracture Mechanics of Polymers, Ellis Horwood Series in Engineering Science, Ellis Horwood Limited, UK. [23] Kinloch AJ and Young RJ (1983) Fracture Behaviour of Polymers, Elsevier Applied Science Publishers, UK. [24] Akbari B and Bagheri R (2007) Deformation mechanism of epoxy/clay nanocomposite, Eur Polym J 43(3):782-788. [25] Kinloch AJ and Taylor AC (2003) The toughening of cyanate-ester polymers. Part II Chemical modification, J Mater Sci 38(1):65-79. [26] Faber, KT and Evans AG (1983) Crack Deflection Processes - I. Theory, Acta Metallurgica 31(4):565576. [27] Hsieh TH, Kinloch AJ, Masania K, Sohn Lee J, Taylor AC and Sprenger S (2010) The toughness of epoxy polymers and fibre composites modified with rubber microparticles and silica nanoparticles, J Mater Sci 45:1193-1210. [28] Tanaka T, Kozako M, Fuse N and Ohki Y (2005) Proposal of a multi-core model for polymer nanocomposite dielectrics, IEEE Transaction on Dielectrics and Electrical Insulation 12(4):669-681. [29] Nelson JK, Fothergill JC, Dissado LA and Peasgood W (2002) Towards an understanding of nanometric dielectrics, 2002 Annual Report Conference on Electrical Insulation and Dielectric Phenomena 295-298. [30] Tuncer E, Sauers I, James DR, Ellis AR, Paranthaman MP, Aytug T, Sathyamurthy S, More KL, Li J and Goyal A (2007) Electrical properties of epoxy resin based nano-composites, Nanotechnology 18(2):025703. [31] National Communications System Technology & Standards Division (1996) Telecommunications: glossary of telecommunication terms, General Service Administration Information Technology Service. [32] Kozako M, Fuse N, Yoshimichi O, Okamoto T and Tanaka T (2004) Surface degradation of polyamide nanocomposites caused by partial discharges using IEC (b) electrodes, IEEE Transactions on Dielectrics and Electrical Insulation 11(5):833-839. [33] Flandin L, Vouyovitch L, Beroual A, Bessede JL and Alberola ND (2005) Influences of degree of curing and presence of inorganic fillers on the ultimate electrical properties of epoxy-based composites: experiment and simulation, J Phys D Appl Phys 38(1):144-155. [34] Nelson JK and Hu Y (2005) Nanocomposite dielectrics – properties and implications, J Phys D Appl Phys 38(2):213-222. [35] Imai T, Sawa F, Ozaki T, Shimizu T, Kido R, Kozako M and Tanaka T (2006) Influence of temperature on mechanical and insulation properties of epoxy-layered silicate nanocomposites, IEEE Transactions on Dielectrics and Electrical Insulation 13(1):445-452. Figures Fig. 1 SEM images of the fracture surfaces of (a) epoxy, (b) 2wt% C30B, (c) 5wt% C30B and (d) 5wt% I28E Fig. 2 Partial discharges (PD) degradation path in nanocomposites, drawn based on [28] Fig. 3 SEM images of the electrical trees in (a) epoxy, (b) 5 wt% I28E, (c) 1 wt% C30B, (d) 2 wt% C30B and (e) 5 wt% C30B. (Arrow indicates the main channel or branch of the electrical tree) Fig. 4 SEM images of the electrical tree main channel in the (a) epoxy and (b) 5wt% I28E samples Fig. 5 Electrical damage zone in the 1 wt% C30B nanocomposite, (a) electrical tree branches and (b) crater