Dielectric behaviours of bio ‐ derived epoxy resins from cashew nutshell liquid

In this study, four commercially available bio ‐ derived epoxy systems (extracted from cashew nutshell liquid) were prepared and characterised. The glass transition temperature ( T g ), dielectric spectroscopy, DC conductivity and breakdown properties of these epoxy resins were studied. Differential scanning calorimetry (DSC) demonstrated that the T g of the investigated systems ranged from 67 to 122°C. The DC conductivity was very low ( < 10 (cid:0) 16 S cm (cid:0) 1 ) and comparable to the conventional dielectrics at room temperature (RT). However, all systems showed a strong temperature dependence of the electrical conductivity and exhibited sharp increase around their respective T g . Arrhenius analysis led to activation energy, E a , values around 1 eV; higher E a values were observed in systems with a lower T g . Dielectric spectroscopy revealed a flat and low response at temperature below T g . However, both the real and imaginary permittivity increased with decreasing frequency at mid to low frequencies as the temperatures approached T g . The variations of AC breakdown strength of all samples were not statistically significant, but the DC breakdown strength of sample 2503A þ 2002B was higher than the others, which might be due to reduced charge transport in this system. The results indicate that novel bio ‐ derived epoxy systems from renewable sources are potential alternatives for traditional petroleum ‐ based epoxy systems in certain insulation


| INTRODUCTION
Epoxy resins have been developed for a wide range of industrial applications, including the matrix of structural composites, coatings, binders and electrical insulators, owing to their outstanding mechanical properties, good adhesive properties, excellent performance as electrical insulation etc. However, most commercially important epoxy resins, such as bisphenol A (DGEBA) and bisphenol F (DGEBF) based epoxy resins, are derived from finite petroleum products, which are not renewable. Furthermore, concerns have been raised relating to their environmental and health impacts on living organisms which, for example, resulted in 2016 in bisphenol A being added to the EU's REACH candidate list, after agreement that it should be identified as a substance of very high concern. This compound is a known endocrine disruptor [1]. As such, the replacement of petroleum-based epoxy resins with more renewable counterparts has much to recommend it and, consequently, the development of such systems has received considerable recent attention [2] as has potential applications. For example, the recent review by Kumar et al. [3] contrasted the synthesis and characterisation of petroleum and bio-based epoxy resins, while Ramon et al. [4] addressed the potential of bio-based epoxy systems derived from resources including natural oils, polyphenols and saccharides, natural rubber and rosin for use in the aviation sector.
Evidently, diverse plant resources have been exploited as raw materials from which to prepare biologically based epoxy monomers, including: palm oil [5]; vanillin [6]; soybean oil [7]; isosorbide [8]; lignin [9]. Zhan and Wool [10], for example, studied two bio-derived epoxy systems based on acrylated epoxidised soybean oil (AESO) and phthalated acrylated epoxidised soybean oil (PAESO). Both were cross-linked with divinylbenzene (DVB). The dielectric constant and loss tangent of these two newly developed systems were found to be lower than those of traditional dielectric materials utilised for fabricating printed circuit boards (PCBs). Moreover, they also exhibited elevated glass transition temperature (T g ) values and Young's moduli, which were favoured for PCB applications. Wan et al. [11] developed a new type of epoxy monomer (TEU-EP) based on eugenol with a forked and branching structure. TEU-EP was mixed with 3,3-diaminodiphenyl sulfone (33DDS) and cured. The resulting TEU-EP/33DDS epoxy system exhibited sufficient reactiveness under elevated curing temperatures. Interestingly, T g of the TEU-EP/33DDS system is 33°C higher than that of DGEBA/33DDS system. Furthermore, the Young's modulus and hardness of the new system increased by 39% and 55%, respectively. The dimensional stability and creep resistance were also enhanced. TEU-EP/33DDS had a significantly lower permittivity, dielectric loss factor, and improved thermal resistance with a good amount of pyrolytic residue. Li et al. [12] further reported a method to synthesise a family of multi-functional epoxy monomers based on eugenol that could be upscaled to industrial levels. The viscosity of the new epoxy monomers is substantially lower (<2.5 Pa s) than that of conventional epoxies based on DGEBA (10.7 Pa s), which was favoured for fabricating composites and pre-pregs. The cured samples exhibited a very low dielectric permittivity (2.8) and improved fire resistance with a limiting oxygen index (LOI) value greater than 31: much better than DGEBA-based epoxy.
Fibre reinforced bio-derived epoxy composites have been investigated by many researchers. Niedermann et al. [13] studied a sugar-based epoxy resin (glucopyranoside-GPTE) with improved T g . They fabricated a series of structural composites by incorporating this epoxy with jute and carbon fibres. The mechanical properties of the composites were comparable with those observed in systems based upon conventional DGEBA epoxy resins. Di Landro and Janszen [14] prepared two hemp fabric reinforced bio-derived epoxy systems using a resin transfer moulding (RTM) technique assisted with a vacuum pump. The mechanical and damping properties of the composites showed noteworthy potential applications in non-structures, semi-structures and insulation boards. However, due to the nature of the hemp fabric reinforcement, the moisture absorption of the composites was very high and fire resistance was inferior, which greatly inhibited their commercial applications. Lincoln et al. [15] developed a PCB utilising woven flax fibre reinforced epoxy resin. The resin was derived from an epoxidised linseed oil. In addition, melamine polyphosphate was added into the composite to enhance the fire resistance properties. The flame retardance, heat resistance, mechanical properties, resistivity and permittivity of the PCB laminates met the requirements for proper performance. However, additional improvements in dielectric breakdown, arc resistance and water absorption characteristics were needed for applications in real life. Deng et al. [16] carried out a life cycle analysis of flax fibre strengthened epoxy (based on epoxidised linseed oil) composites for electronics applications. They concluded that the bio-based substrate reduced the fuel consumption for combustion, offering a more efficient energy consumption during incineration.
There are also a range of studies on bio-derived epoxy nanocomposites for use where electrical conduction is required. Rus et al. [17,18] manufactured and characterised composite films of bio-derived epoxy (∼0.1 mm thick) with varied graphite mass content (5, 10, 15, 20, 25 and 30 wt%) using a solution mixing method. The bio-based epoxy was derived from cooking oil, which is an environment-friendly and renewable resource. They found that the conductivity of the composites reached the percolation threshold when the mass fraction of the graphite was higher than 15 wt%, for example, the conductivity of the blends with 20, 25 and 30 wt% graphite increased to 10 3 -10 4 S m À 1 . The blends also exhibited tensile stress and Young's modulus values at the percolation threshold enhanced by ∼200% and ∼300% respectively, in comparison to those of the unfilled counterpart. In the meantime, the elongation at break of the blends declined with increasing mass fraction of the graphite, leading to a decrease in the stiffness of the blends. Varghai et al. [19] further prepared nanocomposites based on unprocessed multi-walled carbon nanotubes (MWCNT) and a bio-derived epoxy resin: diglycidyl ether of diphenolate n-butyl ester (DGEDP-Bu), which was extracted from diphenolic acid. The characteristics of the bio-derived epoxy and commercial DGEBA epoxy composites were studied and compared in this work. The results indicated that the electrical conductivity of the DGEDP-Bu composites outperformed the commercial DGEBA epoxy composites. In addition, the bio-based epoxy blends exhibited comparable mechanical properties to those of DGEBA composites.
A technologically important application area for epoxybased systems relates to their use in high voltage power engineering applications, including, generators, bushings, spacers etc., where the ability to withstand high electric fields for decades is critically important. While the above overview demonstrates that significant advances have been achieved in developing bioderived epoxy resins and their composites, little research on dielectric and insulation properties have been undertaken [10][11][12], particularly in connection with the use of such materials under high field conditions, which is an essential prerequisite for their use as insulating materials in power engineering plant.
The investigation reported here set out to examine the electrical properties of a related series of epoxy-based systems, where both the resin and the hardener had been derived using renewable, biological sources. These systems consist of amended epoxy resins (part A) and amine hardeners (part B). Both parts are formulated using cashew nutshell liquid (CNSL), which is an annually renewable resource that does not interrupt the food chain. The primary building block for these bio-derived epoxy resins is cardanol, which is a distinctive nature-based phenolic resin, extracted by the process of distillation of CNSL. The molecule contains an aromatic ring with a hydroxyl group and an outstretched aliphatic side chain, which provides substantial inherent merits to the ultimate properties of the composites ( Figure 1). The curing agents (part B) are phenalkamines. These bio-derived epoxy systems are designed to offer better characteristics for composite producers and also to contribute to reducing greenhouse emissions and enhancing sustainability in terms of insulation materials in electrical applications.
The primary aims of this work were to examine the influence of changes in composition and hence network topology and retained functional groups on permittivity, conductivity and breakdown performance, which represent critical parameters when considering the suitability of a material for use in high voltage insulation applications. This is not to say that additional parameters, such as mechanical properties, thermal conductivity, thermal expansion coefficient, thermal decomposition temperature etc. are not of great technological importance, but a full consideration of every parameter that may be important in different applications exceeds the scope of this study.

| Materials
All the epoxy resins and hardeners utilised in this study were supplied by Cardolite Specialty Chemicals Europe N.V. Two epoxy resins were used in this work: Cardolite FormuLITE 2501 and 2503A. Both are based on DGEBA and their epoxy equivalent weight (EEW) is 198 and 182-200 g eq. À 1 , respectively. The hardeners used for curing were Cardolite FormuLITE 2401 and 2002B amine-based hardeners. According to the manufacturer, all of these components are made using high amounts of renewable resources, including CSNL, such that typically, FormuLITE resin/hardener systems contain a bio-content of about 35%. The compositions and chemical structures for the resins and hardeners are listed in Table 1.

| Sample preparation
The resin and the hardener were put in a beaker and mixed utilising a magnetic stirrer for 5 min at ambient temperature (20 � 2°C), with the mass ratio recommended for each formulation conforming to the manufacturer's recommendations. The product was subsequently degasified at room temperature (RT) for 20 min and then cast into a steel mould for curing. The mass ratios of all the epoxy systems and curing conditions are listed in Table 2. Each sample was fabricated into two thicknesses of 70 � 5 μm for AC and DC breakdown test and 200 � 10 μm for dielectric spectroscopy and DC conductivity analysis. All specimens were kept in a vacuum desiccator at RT to avoid uptake of moisture from the air.

| Characterisation
The glass transition temperature (T g ) of cured epoxy specimens was determined using a Perkin Elmer DSC7 differential scanning calorimeter (DSC), which was calibrated using high purity indium. A specimen 5-10 mg in mass was sealed in an aluminium pan and heated from RT to 200°C at a heating rate of 20°C min À 1 . Then, it was cooled down to RT and heated to 200°C again. The first heating scan aimed to erase the thermal history of the sample; T g was determined as the temperature at the maximum rate of change of the specificheat-capacity from the second scan. Three specimens were tested for each system in order statistically to analyse the collected results. Dielectric spectra for all samples (nominally 0.2 mm thick) were obtained via a Solartron 1296 dielectric interface coupled with a Schlumberger SI 1260 impedance-phase collecting processor. The test chamber contained two parallel circular electrodes, 32 mm in diameter. Samples of comparable size were placed between these two electrodes for measurement and an AC voltage of 1 V was applied over a frequency range from 0.1 Hz to 1 MHz. Dielectric data were acquired from all the samples at three temperatures (RT, 60 and 90°C). The software employed was Solartron Materials Research and Test Software (SMaRT).
The time dependence of electrical conductivity of all the bio-derived epoxy systems was evaluated using a Keithley 6487 picoammeter. The samples (nominally 0.2 mm thick) were placed between two 20 mm diameter polished gold coated electrodes. A DC voltage of 2 kV was applied (10 kV mm À 1 ) and the current data were collected for 30 min. These measurements were conducted at RT, 60 and 90°C.
AC and DC breakdown analyses were performed according to ASTM D149 and ASTM D3755-14, respectively. The samples were placed between two steel ball bearing electrodes, 6.3 mm diameter. The steel balls were changed after every 20 measurements to prevent surface pitting from influencing the obtained data. The test cell was immersed in silicone oil (Dow Corning 200/20CS) to avoid flashover. The increasing ramp rate of the voltage was 300 V s À 1 for AC breakdown measurements and 350 V s À 1 for DC breakdown measurements: all these tests were implemented at RT, due to instrumental limitations. Two-parameter Weibull distributions were generated for analysis of the resulting data.

| Glass transition temperature
DSC curves of all the bio-derived epoxy samples are shown in Figure 2. The determined T g values range from 67 to 122°C (see Table 3), which is comparable to many conventional resins. The T g of the systems cured with the 2002B hardener is significantly lower than that of the systems cured with 2401B hardener. This  is probably associated with the stoichiometry of each system. To cure a certain amount of resin, more 2002B was needed than 2401B, which indicates that the amine hydrogen equivalent weight of 2002B is lower than that of 2401B. Therefore, the cured 2002B systems contain a lower crosslinking density and fewer network nodes than the cured 2401B systems, resulting in a reduction in T g [20]. The T g values of 2501A systems were lower than those of 2503A systems, which might be attributed to the compositions of these two resins. It can be seen from Table 1 that 2501A contains 10-20 wt% of CNSL with a long aliphatic structure, which can provide flexibility and resilience to the material such that molecular mobility within the cured samples would be enhanced. Figure 3 shows the time dependence of conductivity under a constant DC voltage for the bio-derived epoxy samples, at three temperatures. This figure indicates that all the samples exhibited similar low conductivity values at RT. This is because all the samples were well within the glassy state at RT and the segmental dynamics were therefore constrained at this temperature. While the conductivity of all the samples increased with increasing temperature (see Figure 3(b)), this is especially the case for the 2401A þ 2002B system, which was characterised by the greatest increment on increasing the temperature from RT to 60°C. This can be explained as follows: as the testing temperature (60°C) approached the T g (67°C) of this sample, segmental molecular motion increased and the mobility of retained ionic moieties within the material increased significantly [21]. Similar conclusions were obtained for measurements made at 90°C (see Figure 3(c)). The conductivity of all the samples increased markedly on increasing the temperature from 60 to 90°C.

| DC conductivity
To examine the above behaviour in more detail, the data presented in Figure 3 were analysed quantitatively, assuming Arrhenius behaviour. For this, it was initially necessary to estimate the steady-state DC conductivity from the measured time-dependent conduction. Behaviour First, from Figure 3, it is evident that, while the conductivity varies initially in these systems, at the longest times considered here, it appears close to constant. As such, DC conductivity values were first estimated from each data set by averaging the final 10 data points obtained (σ DC1 ). Second, the time dependent conductivity, σ(t), data shown were fitted to a modified form of the Curie-Von Schweidler law [22], which incordsporates a steady state, DC conductivity term, σ DC2 : In the above equation, a and b are constant and t represents the time after voltage application. Comparison of σ DC1 and σ DC2 values revealed excellent agreement but, due to convergence problems with two of the data sets, the subsequent Arrhenius analysis was undertaken using σ DC1 as a measure of the DC conductivity of each system. Table 4 lists the activation energy values derived for each material formulation together with the uncertainty, as represented by the SE. In all cases, conformity to the Arrhenius relationship was good, with values of the coefficient of determination falling in the range 0.984 to 1.000.
Comparison of Tables 3 and 4 reveals a strong anticorrelation between T g and E a : increasing T g results in a reduction in E a . Similar behaviour and activation energy values have recently been observed by one of us in a very different series of epoxy-based systems [23] and elsewhere [24].

| Dielectric spectroscopy
The variations of dielectric properties with temperature and frequency are shown in Figure 4. At RT, the value of the real part of the relative is permittivity, ε r ', is low and varies little with frequency over the range shown. The peaks in the imaginary part of the relative permittivity, ε r ', (indicated with an arrow in Figure 4(b)), at high frequency (>10 3 Hz) for all the samples are ascribed to the β-relaxation [25]. It was reported that β-relaxation is attributed to the mobility and frequency of the hydroxyl ether groups within the resin system [26,27]. These groups are formed by the chemical reaction of the amine hydrogen of the hardener and the epoxide groups within the system. The real permittivity of these samples was in the range of 2.8 to 3.4 at RT and was less than that of traditional epoxy systems (3.6-4.4) [28]. The imaginary permittivity of the polymers was less than 0.1 and is favourable for electronic insulation use.
A significant uplift at low to mid frequencies is evident in the real and imaginary permittivity of sample 2501A þ 2002B at 60°C. Increases in permittivity with decreasing frequency have been discussed in terms of both DC and quasi-DC conduction processes [29] and therefore, although the dielectric data were acquired using a very much lower applied field than the conductivity data presented in Figure 3 and are therefore likely to relate to somewhat different phenomena (c.f. the activation energy values presented in Table 4 and references [23,24] with that reported in reference [30]), the variation in the permittivity of 2501A þ 2002B at 60°C evinced by Figure 4 aligns well with the pronounced increase in the conductivity of 2501A þ 2002B shown in Figure 3.

-
In addition, the α process in epoxy resins, has been observed and characterised since the early 1960s via both mechanical and dielectric techniques [31,32] and is typically attributed to segmental chain motions triggered by the glass transition. As such, the dielectric behaviour of 2501A þ 2002B at 60°C will also contain a contribution from the dielectric α relaxation and, indeed, it is likely that the increased segmental motion is furthermore a contributory factor leading to increased charge transport [23]. At 90°C, samples of 2501A þ 2002B, 2503A þ 2002B and 2503A þ 2401 B exhibit large uplifts across the entire frequencies range studied, due to that the testing temperature (90°C) being comparable to or higher than the T g of these samples. That is, due to increased charge transport and the influence of the dielectric α relaxation, as discussed above. However, sample 2503A þ 2401 B, with a T g ¼ 122°C, is characterised by a flat real permittivity and low imaginary permittivity with only a small uplift at low frequencies in Figure 4(e) and (f), respectively.

| AC and DC breakdown strength
The Weibull distribution plots of AC breakdown at RT for all the bio-derived epoxy systems are shown in Figure 5. The derived Weibull scale (α) and shape (ß) factors for all the bioderived epoxy systems at 63.2 percentile with error bars that correspond to the 95 % confidence bounds are listed in Table 5.
Unexpectedly, no distinguishable variations of the Weibull scale parameters (α) were observed for all the investigated specimens, even though the T g values of samples cured with 2002B are much lower than those of samples cured with 2401B hardener. Therefore, we infer that the RT AC breakdown behaviour of these systems was not directly correlated with T g . It was reported that AC breakdown strength is sensitive to varied factors, such as ionic impurities, moisture absorption, defects or deficiencies within the structure of the materials [28,33]. It is thus difficult to provide an unambiguous interpretation of the AC breakdown behaviour reported in this work. Figure 6 demonstrates the Weibull distribution of DC breakdown strength at RT for all the samples and Table 6 lists the derived α and ß values. Sample 2503 A þ 2002B had the highest DC breakdown strength, which is around 581 kV mm À 1 . This might be a result of fewer dipolar and ionic impurities being presented in the 2002B systems, which could avoid deteriorating breakdown strength. In addition, the dielectric spectroscopy measurements at RT revealed slightly lower real permittivity values of 2002B systems compared to 2401B systems, which we suspect is due to more impurities in the 2401B systems.
Finally, from an applications perspective, the effect of increasing temperature on breakdown performance is an important consideration. While experimental limitations have precluded the examination of such factors in this study, the tendency for conduction to be enhanced as the temperature increases, particularly in systems characterised by lower glass transition temperatures, implies that such factors are likely to be of considerable technological importance.

| CONCLUSIONS
In this study, the thermal and dielectric properties of four bio-derived epoxy resins derived from CNSL were studied in detail using DSC, dielectric spectroscopy, DC conductivity and breakdown analysis. These newly developed resin systems had a wide range of T g values, from 67 to 122°C; the electrical conductivity is very low (10 À 16 ∼10 À 17 S cm À 1 ) in all systems when in the glass state. However, the conductivity increased sharply at temperature above T g , when segmental motion, dipolar orientation and carrier transport were all activated. The permittivity exhibited a similar temperaturedependence trend, in which all the systems showed a flat and low dielectric response when the temperature was lower than the respective T g . However, the imaginary part of the complex permittivity increased markedly at temperatures above T g . There is no significant difference in the RT AC breakdown strength for all the systems, but the system 2503A þ 2002B revealed a RT higher DC breakdown strength than that of the other three. These results suggest that CNSL-derived epoxy can exhibit appropriate combinations of properties to render them appropriate for use in targeted electrical applications. However, the deterioration in electrical performance revealed at temperatures above T g represents an important consideration.