2.1.Particle preparation

The ZnO particles were prepared following the aqueous sol-gel synthesis method. Briefly, zinc acetate and oxalic acid were mixed at a molar concentration ratio of 0.1 /0.1 (M). The oxalic acid solution was added dropwise to the zinc acid solution, which was kept under constant agitation for 90 min (formation of the gel). The gel obtained was dried for 12 h at 60^oC and then subjected to a thermal treatment at 600^oC for 3 h.

The alumina particles were prepared by the citrate route. A polymeric citrate precursor derived from a citric acid aluminum nitrate solution at a molar concentration ratio of 0.5/0.1 (M) was used. The precursor thermal decomposition at 60^oC formed a precipitate that was dried at 200^oC for 2 h, followed by thermal treatment at 900^oC for 2 h. The synthesized ZnO and Al₂O₃ particles were characterized in a SmartLab XRD diffractometer from Rigaku in PB/PSA mode Parallel Beam/Parallel Slit Analyzer with a grazing incidence angle (ω=0.5∘) 2θ geometry and Cu Kα radiation(40 kV, 44 Amp) at a step size of 0.04^o. The diffraction patterns were indexed using the Integrated X-ray powder diffraction Software PDXL from Rigaku. The surface morphology of the synthesized powders was determined through a field emission scanning electron microscope (FE-SEM) Jeol JMS-7200F, using an accelerating voltage of 1.5 kV and different amplification factors.

2.2.Composite preparation

The resin compound Epolam 2500, a biphasic epoxy resin composed of primer and hardener was purchased from Sika Advanced Resins®. The primer and hardener were mixed at a 100:22 weight ratio. Zinc oxide or γ-Al₂O₃ particles were incorporated by solution blending in the epoxy resin at concentrations of 0.1, 0.25, and 0.5 wt. %. In a first step, ZnO particles were suspended in ethanol and sonicated for about 5 min to break the agglomerates as much as possible and until a homogeneous dispersion was achieved. Then, the epoxy resin was added to the solution that was kept under constant stirring at 50^oC until ethanol was completely evaporated, and apparent uniform distribution of the ZnO in the resin was attained. The same mixing strategy was followed for the γ-Al₂O₃-epoxy composite. All samples were cured at room temperature for 24 h followed by a dwelling time of 2 h at 100^oC.

2.3.Thermal characterization

2.3.1.Modulated thermal response

Thermal images were acquired through an active infrared thermography setup, consisting of a TTL controllable laser (120 mW LASER LDCU5, at 658 nm); a function generator, for the generation of the TTL control signal; a closed sample chamber; and an IR camera (FLIR, SC5000), suitable to operate in the 2.5 - 5.1 μm wavelength range. Between the IR camera and the sample chamber, a germanium window was placed to avoid undesirable contributions to the thermal images. The spherical lens is used to ensure a highly focused beam impinging on the frontal sample surface. Figure 1 displays a schematic of the active thermography system.

Figure 1 Schematic representation of the active thermography system. 

The capabilities of the IR camera electronics, along with its software, allow the phase-sensitive images processing and synchronization to the reference signal (from the function generator); so, there is no need for an external lock-in amplifier module. The sample chamber was kept at high vacuum (1×10-6 Torr) during the measurement using a rotatory vacuum pump (Varian Vacuum Technologies, DS 102) coupled with a turbo-molecular vacuum pump (Varian Vacuum Technologies, Turbo-V 81-AG) to avoid heat losses due to conductive and convective heat transfer to the surroundings. The modulation frequency of the excitation beam was fixed to 10 mHz.

3.1.Ceramic-particles characterization

The acquired X-ray diffraction patterns of the γ-Al₂O₃ and ZnO synthesized particles are presented in Figs. 2a and 2b, respectively, identifying the characteristic XRD peaks with the Miller indices of their corresponding diffraction planes. While the ZnO particles exhibit sharp (Lorentz-type) peaks (Fig. 2b) corresponding mainly to the wurtzite hexagonal phase of ZnO; the XRD obtained for the γ-Al₂O₃ particles presents broader peaks (Fig. 2a), and their positions correspond predominantly to cubic γ-Al₂O₃ [³⁰]. The observed peaks broadening is, presumably, due to the overlapping of Lorentz-type peaks. An estimate of the crystallite size was calculated using the Rietveld method. The crystallite size of γ-Al₂O₃ particles is about 3.32 nm, while the crystallite size for ZnO particles is 22.5 nm. The apparent morphology of the obtained particles was analyzed by FE-SEM micrographs (Fig. 3). The SEM micrographs show that the ZnO powders are a combination of rod-like structures with an average thickness of 63 nm and small cubic and irregular chunks, which seem to cohere in micrometric aggregates with a maximal averaged size of 20 μm in the powder form (Figs. 3c and 3e). On the other hand, the micrographs of the γ-Al₂O₃ particles show small irregular particles, as small as 50 nm and as large as 0.22 μm, which also present a tendency to form large aggregates in the powder form (Fig. 3d and 3f). Micrographs of the composites at a concentration of 0.5 wt. %, epolam/ZnO, and epolam/γ-Al₂O₃, are presented in Fig. 3a and 3b, respectively. As it can be observed on the micrographs, the particles are not visible on the surface of the composites, and the apparent surface roughness of the epolam /γ-Al₂O₃ is higher than that of the epolam/ZnO at this particle concentration.

Figure 2 X-ray diffraction pattern recorded for the (a) γ -Al₂O₃ and (b) ZnO synthesized particles. 

Figure 3 a) SEM micrographs of Epolam + 0.5 wt. % ZnO, b) Epolam + 0.5 wt. % γ -Al₂O₃, c) ZnO particles at 20kx, d) γ -Al₂O₃ particles at 10kx, e) ZnO particles at 80kx, and f) γ-Al₂O₃ particles at 80kx. 

3.2.Thermophysical Properties

The slope method for thermal diffusivity measurements shows that for a homogenous thermally thin, sample the following relationship exists:

where Δψ is the photothermal phase shift; ΔTac is the (modulated) temperature amplitude; r is the radial coordinate; and mΔψ and mLn[r⋅ΔTac] are the slopes of the linear regions in the Δψ versus r and Ln[r⋅ΔTac] versus r curves, respectively [³¹,³²,³³,³⁴,³⁵]. From Eq. (1) the thermal diffusivity, αeff, can be obtained if the modulation frequency, ω is known. Figure 4 presents the Δψ vs. r and Ln[r⋅ΔTac] vs. r plots for the epoxy matrix. The convective-radiative heat losses at low frequencies are responsible for deviations from linearity [³⁶].

Figure 4 a) Δ ψ vs. r, and b) Lnr∙∆Tacvs. r plots, for the epoxy sample. The solid red line shows the best least-squares linear fit. 

Following the same data processing, the thermal diffusivity of the different epoxy/γ-Al₂O₃ and epoxy/ZnO composites was calculated; Figure 5 shows the results obtained. In any case, the error bars associated with the effective thermal diffusivity are one hundred times smaller than the αeff values (∼10-9 m²⋅s^-1) for the epoxy/Al₂O₃ composites. However, for the epoxy/ZnO composites, the variations of αeff as a function of the ZnO content are of the same order of magnitude of the corresponding error bars; therefore, the addition of ZnO has no real impact on the effective thermal diffusivity of such composite samples. On the other hand, the epoxy/γ-Al₂O₃ composites exhibit a non-linear monotonic decreasing behavior in αeff, with the γ-Al₂O₃ content. Figure 6 presents the transient thermal response of the samples under the experimental conditions described in 2.3.2 for different γ-Al₂O₃ and ZnO contents.

Figure 5 Obtained values for the effective thermal diffusivity, as a function of the particle content, for (a) Epoxy/Al₂O₃, and (b) Epoxy/ZnO composites. 

Figure 6 Thermal response (in amplitude) of a) Epoxy/Al₂O₃, and b) Epoxy/ZnO composites, recorded employing the Hot-Disk measurement system. 

The effective thermal conductivity, keff, of the composite samples was determined from the ΔT vs. t curves by linear fitting procedures according to the model reported by [²⁹]. The obtained values for the neat epoxy matrix, epoxy/γ-Al₂O₃, and epoxy/ZnO composites agree well to similar epoxy-based composites reported elsewhere [³²,³³,²⁶,³⁷]. Figure 7 summarizes these results, indicating that the epoxy/γ-Al₂O₃ composites exhibit an increasing monotonic behavior for keff as a function of the γ-Al₂O₃ content, deviating from linearity at the origin. This supports that the oxide particle loading is well-dispersed within the matrix since dispersion has proven to be one of the most important factors to improve the thermal conductivity [²⁶]. The improvement in thermal conductivity is recorded up to 17.8 % by introducing just 0.5 wt. % of γ-Al₂O₃. Furthermore, the improvement in thermal conductivity at such low oxide particle loading is similar to the improvement obtained by [²⁶,⁴⁷] filler loading up to 10 wt. % was added.

Figure 7 Effective thermal conductivity values of (a) Epoxy/Al₂O₃, and (b) Epoxy/ZnO composites, obtained from Hot-Disk measurement data. 

A quite different thermal behavior occurs for the epoxy/ZnO composites, were a non-monotonic dependency of keff on the ZnO content is evident. Moreover, the sample with 0.25 wt. % content of ZnO shows an important decrease in the keff value compared to the samples with 0.1 and 0.5 wt. % content of ZnO, but still higher than the keff value corresponding to the neat epoxy sample. Lower conductivity at this concentration could be due to agglomerate formation at this concentration.

Considering that keff, αeff are not independent thermophysical parameters, but related through the (effective) volume-specific heat capacity (also known as heat capacity per unit volume), Ceff=keff⋅αeff-1. Figure 8 displays the calculated values of Ceff from this relation for all composite samples.

Figure 8 Calculated C _eff values of a) Epoxy/Al₂O₃, and b) Epoxy/ZnO composites, as a function of the particle content. 

Again, the epoxy/γ-Al₂O₃ samples showed an increasing non-linear monotonic behavior in the Ceff values as a function of the γ-Al₂O₃ content. On the other hand, the epoxy/ZnO samples show that, for a 0.25 wt. % content of ZnO, Ceff has a local minimum mostly due to the effective thermal conductivity, since the effective thermal diffusivity values are practically the same for all the epoxy/ZnO samples here studied. Assuming that the synthesized γ-Al₂O₃ and ZnO particles are homogeneously distributed into the bulk of the sample, the effective volume-specific heat capacity could be estimated by Eq. (2) as follows:

In Eq. (2), C(m), C(d) are the volume-specific heat capacity of the continuous (neat epoxy matrix) and discrete phases (γ-Al₂O₃ and ZnO particles) of the composite, respectively; w(d) is the content fraction of the discrete phase, given by the wt. % particle content of the composite; and ρrel=ρ(m)⋅ρ(d)-1 is the ratio of the mass densities. However, Eq. (2) has a decreasing behavior-owing to the thermophysical properties difference between the discrete and continuous composite phases-, contrary to the experimental observation for all composites. The observed demeanor of the thermophysical properties, αeff, keff, and Ceff- together with the theoretical estimation Cbulk-, suggest that the ballistic heat transfer mechanism competes with the diffusive heat transfer mechanism in response to the possible formation of axial thermal paths inside the composites, being the one exception, the epoxy/ZnO composite with 0.25 wt. %. For such a particular sample, it is plausible that the concentration of ZnO particles induces the agglomeration of the particles in disjoint volume elements, diminishing the formation of thermal paths and, therefore, a greater contribution of the diffusive heat transfer mechanism is observed. However, these hypotheses must be corroborated by quantification of the spatial distribution of the dispersed particles.

3.3.Thermal stability and degradation kinetics

The thermal stability of the epoxy/γ-Al₂O₃ and epoxy/ZnO was studied by TGA at a heating rate of 10^oC/min under a nitrogen atmosphere. Figure 9 displays the thermograms and their derivative (DTG) obtained for epoxy/γ-Al₂O₃ and epoxy/ZnO, respectively. As a reference, neat epoxy without ceramic particle incorporation was also analyzed. From TGA curves (Figs. 9a and 9c) of the neat epoxy, it is possible to observe that the main thermal transition occurs from 250 to 500^oC with a weight loss of about to 70 %, attributed to the degradation of the epoxy network [¹⁵]. A stable zone from 500 to 700^oC follows this thermal transition, and a final weight loss of about 10 % occurs from 700 to 900^oC.

Figure 9 TGA thermograms for ceramic/epoxy and neat epoxy. The respective DTG thermograms for each concentration is displayed on the top right corner of the image. (a) Epoxy/Al₂O₃ and (b) Epoxy/ZnO. 

The total mass loss for all samples was ranged from 79.8 to 76.9 %. The derivative of the thermogravimetric curves helps to determine the temperature at which the maximum rate of thermal decomposition is reached (see Fig. 9). DTG curves displayed in Figs. 9b and 9d show a three-stage degradation for the neat epoxy in a nitrogen atmosphere, which is affected by the increasing content of each oxide. The first degradation stage of the epoxy starts at 314^oC and is associated with uncured epoxy resin and loss of volatiles. The second stage of degradation onset was detected between 318 and 330^oC with a maxima temperature decomposition at about 340^oC; this points out to the decomposition of aromatic functional groups. The last stage observed between 480 and 550^oC corresponds to the breaking of the main chains of the epoxy resin. These polymer degradation stages match with data reported in the literature [¹⁵,³⁸,³⁹].

The DTG curves for all composites showed that the addition of ceramic particles inhibited the first stage of the thermal degradation of epoxy, contributing to the dissipation of its volatile fraction. These findings have not been reported in other studies. The results demonstrate that for temperatures under 250^oC, the addition of either ZnO or γ-Al₂O³ with concentrations <1 wt. % increase the thermal stability of the epoxy matrix; this is in agreement with the analysis of volatile organic compounds in polymers reported elsewhere [⁴⁰]. Similar thermal transitions were observed for the second and third stages of thermal decomposition for the neat epoxy, epoxy/γ-Al₂O₃, and epoxy/ZnO, as the transitions appear at similar temperatures ranges (318-330 and 480-550^oC, respectively). However, for epoxy/γ-Al₂O₃ at a concentration of 0.1 wt. %, it was noted that the third transition shifts towards lower temperatures with an onset at about 455^oC.

The incorporation of γ-Al₂O₃ decreases slightly the temperature at which the maximum rate of thermal decomposition occurs from 340 to 334^oC, suggesting the detriment of the thermal stability of the matrix caused by the oxide. It is believed that γ-Al₂O₃, at low concentrations, acts as a catalyst at high temperatures, accelerating the breaking of the polymeric chains of epoxy resin and decreasing the decomposition temperature. In the case of ZnO incorporation, the DGT curves for all concentrations showed similar thermal behavior to neat epoxy, indicating the apparent conservation of the thermal stability of the matrix.

Thermal stability factors such as initial decomposition temperature (the temperature of 5 wt. % loss, Td5) and char at 900^oC were determined for each composite. Table I lists the thermal stability factors for both composites at different filler content. The Td5 for epoxy/γ-Al₂O₃ at 0.1 and 0.25 wt. % was similar to that of neat epoxy. However, as the γ-Al₂O₃ content increased to 0.5 wt. %, the Td5 decreased by 15^oC. For the epoxy/ZnO, the Td5 decreased by 19^oC compared to neat epoxy, regardless of the particle content. The char at 900^oC decreased for both composites compared to neat epoxy. Regardless of the ZnO particle content, the char at 900^oC remains constant at 22.8 wt. % while a lower concentration of γ-Al₂O₃ displays the lowest char at 900^oC, 20.20 wt. %.

The thermal properties in this study are compared for epoxy resin-based composites produced by the above-mentioned process for particle loadings <1 wt. %. The comparability of the results obtained with other studies is restricted due to differences in the epoxy resin system used; for instance, Jiang et al. [²¹] studied the behavior of epoxy (DGEBA) with contents of nano alumina (α-Al₂O₃) from 1 to 4 wt. %; the results show the Td5 value of the composites similar to that of the neat epoxy resins (362.1^oC) with char contents increasing gradually with increasing concentrations of nano Al₂O₃ [21]. Fan-Long Jin and Soo-Jin Park [41] studied the same epoxy resin system (DGEBA) with Al₂O₃ loadings ranging from 5 up to 15 wt. %; their results show a two-stage degradation process with char increasing as the concentration of Al₂O₃ was increased [⁴¹]. However, in our case, the char yield decreased for lower concentrations of alumina by 2.88 wt. % of the total mass used in the study compared to neat epoxy. The Td5 values remain the same until the alumina concentration reached 0.5 wt. %, at this point, the Td5 value decreased by 15^oC.

Kinetic analysis of thermal degradation was carried out with Multilinear Regression (MLR) method to fit four data curves obtained at a constant heating rate by TGA in an inert atmosphere to the Arrhenius relationship, k(t)=A⋅exp(Ea/RT), determining the pre-exponential factor (A) (min^-1), Ea is the apparent activation energy (kJ⋅mol^-1), R is the universal gas constant, and T is the temperature (K). For the iso-conversion method, the expression for the rate of conversion dα/dT=k(T)⋅[1-α]n can be used, where α is the degree of chemical conversion, n is the order of reaction, and k is the reaction rate. The three kinetic parameters have been obtained using various iso-conversion model-free as well as MRL methods from materials data as a function of temperature and conversion. Non-isothermal TGA is one of the best methods to understand the kinetics degradation process of polymers, therefore adequate to find solutions for the estimation of the kinetic parameters [⁴²]. The degradation kinetics analysis at different heating ramps (5, 10, 20 and 50^oC/min) showed a slight displacement towards higher temperatures with higher heating ramps for the neat epoxy resin and both composites (Fig. 10 and 11). This behavior is markedly observed at temperatures below 550^oC for neat epoxy, 425^oC for epoxy/γ-Al₂O₃ and 475^oC for epoxy/ZnO.

Figure 10 Thermograms of neat epoxy and Epoxy/Al₂O₃ at different heating ramps used to determine the kinetic degradation of the MRL method. a) Neat epoxy, b) 0.1 wt. %, c) 0.25 wt. %, and d) 0.5 wt. %. 

Figure 11 Thermograms of neat epoxy and Epoxy/ZnO at different heating ramps used to determinate kinetic degradation of MRL method. a) Neat epoxy, b) 0.1 wt. %, c) 0.25 wt. %, and d) 0.5 wt. %. 

Figure 12 presents the activation energy (Ea) for all the composites and neat epoxy. Activation energy values were obtained with a determination coefficient (R2) at 95 % and taken conversions of (α) of 5, 10, 15, 20, 25, and 30 %. Ea of neat epoxy (121.53±9.15 kJ/mol) and epoxy/γ-Al₂O₃ @ 0.10 % and @ 0.25 % (118.11±6.47 and 104.63±7.34 kJ/mol, respectively) display similar values (Zone I). However, a significant decrease in Ea is observed for the γ-Al₂O₃ content of 0.50 wt. % (75.42±7.13 kJ/mol) (Zone II). These results are in agreement with the activation energy observed by Omrani and Rostami at a similar concentration of alumina (0.5 phr) [⁴³]. Epoxy/ZnO composites, regardless of the concentration of the filler, show similar behavior to epoxy/γ-Al₂O³ @ 0.5 %. These results suggest that loadings ≤0.5 wt. % of ZnO and only the loading of 0.5 wt. % of γ-Al₂O₃ decreased the thermal stability of neat epoxy resin. Other authors have reported decreasing activation energy values with the addition of oxide particles, mainly attributed to a catalytic effect of these particles [⁴⁴,⁴⁵,⁴⁶]. This effect has been observed for copolymer/TiO₂ [⁴⁴] at a concentration of 0.2 wt. % and polyacrylate/ZnO (45) composites. Therefore, the incorporation of ZnO (for all cases) and Al₂O₃ at 0.5 wt. % had a significant impact on Ea values of neat epoxy.

Figure 12 Activation energy vs. nanoparticle concentration in wt. %.