*3.1. Dispersion*

Two mixing modes were investigated in the three-roll mill:


The eight total suspension pass gaps are resumed in Table 1. To avoid an excessive promotion of rupture mechanism that could damage the aspect ratio of the nanoparticles, the speed on the apron roll was always set at 180 rpm.


**Table 1.** Three-roll mill mixing modes investigated.

Resistivity measurements for the two gap modes investigated are plotted in Figure 3. While the di fference in first passes is still not perceivable, it reaches one order of magnitude after eight passes. Using the regressive gap mode seems to promote the lower electrical resistivity of SWCNT/epoxy composites for the same curing process parameters and without increasing the particle concentration. This is expected to result from the di fferent agglomerate dispersion mechanisms [8], where erosion is fostered over rupture. The constant mode (gaps of 15–5 μm for each pass) imposes high shear forces to the bigger initial CNT agglomerates, promoting their rupture and, most likely, rupture of the nanotubes. On the other hand, starting with wider gaps in the regressive mode favors the erosion of the agglomerates over rupture, limiting the breakage of CNTs. A higher CNT aspect ratio for the same concentration seems to help the formation of more e fficient secondary networks for electrons to flow across the bulk composite, by enhancing the chances of contact between adjacent particles.

**Figure 3.** Electrical resistivity of SWCNT/epoxy 0.1 wt.% bar-shaped samples produced with two gap-modes in the three-roll mill and measured with the four-point method.

Additionally, Figure 4 depicts the through-plane electrical resistivity of SWCNT/epoxy nanocomposites at different concentrations. For the lowest amount of SWCNTs (0.001 wt.%) the electrical resistivity was too high to be measured by the multimeter in resistance mode (limited to 50 MΩ) and hence the value presented corresponds to the multimeter range limit. From the figure it can be concluded that the percolation concentration lies between 0.001 wt.% and 0.005 wt.%. It can be observed that the through-plane resistivity for 0.1 wt.% SWCNT/epoxy disk-shaped samples in Figure 4 is substantially higher than the value in Figure 3, which corresponds to the in-plane resistivity measured with the four-point method.

**Figure 4.** Through-plane electrical resistivity of SWCNT/epoxy as a function of particle concentration for eight passes.

The influence of the curing temperature of the SWCNT/epoxy composites on the final electrical resistivity was investigated for samples collected after different numbers of passes in the three-roll mill. Specimens of SWCNT/epoxy at a concentration of 0.1 wt.% (mixed in constant gap mode) and at four different passes number were cured both at room temperature and heated up at 80 ◦C, accordingly also varying the overall curing time. The resistivity measurements are depicted in Figure 5. The beginning of a plateau is noticeable at pass number six, after which no improvements come from additional mixing in the three-roll mill. Moreover, by increasing the curing temperature, the electrical resistivity of the SWCNT/epoxy 0.1 wt.% composites appears to decrease around 50%. This is probably due to the fact that, at 80 ◦C, the resin matrix is less viscous than at room temperature, as it emerges from the rheological tests performed in [22]. A more liquid suspension provides less viscous resistance for the secondary agglomerates' network formation, under Van der Waals forces or intentional applied dielectrophoresis (useful consideration for alignment experiments described further on in this paper, which in fact were performed at 80 ◦C).

**Figure 5.** SWCNT/epoxy 0.1 wt.%—influence of curing temperature on electrical resistivity at different passes.

## *3.2. Electric Fields*

Dielectrophoresis experiments were performed with SWCNT/epoxy 0.01 wt.% composites. The main goal of this set-up was to produce multiple plate specimens simultaneously and with a shape defined by the silicone mold, which is a scalable process with potential application in industry. With every experiment, six disk samples were produced (20 mm diameter and 1.5 mm thickness), as can be seen in Figure 6.

**Figure 6.** Silicon mold (corresponding to point 3 in Figure 2), which allows for the production of six discs of 20 mm diameter per 1.5 mm thickness.

Furthermore, the influence of the use of distinct mold release agents was investigated. Figure 7 shows the resistivity values measured between the electrodes after the curing of SWCNT/epoxy 0.01 wt.% at 100 ◦C, with and without electric field. The first measurements presented here were done before opening the mold by connecting the electrodes to a multimeter after sample cool down. Therefore, the values represent the total resistivity of the material connecting the electrodes (six plates and the interconnections). A one order of magnitude decrease in the resistivity was generally measured.

**Figure 7.** Impact of mold release agen<sup>t</sup> (MR1,2,3) on plate resistivity measured at the electrodes (after cure) for SWCNT/epoxy 0.01 wt.% composites.

However, a clear difference was noticeable when different mold release agents were used. With the most conductive one among the three, MR3 (silver paint), both the resistivity with and without electric field were two orders of magnitude lower than that of MR1. MR2 (intermediate conductivity) appeared to be in between the two and showed a minor difference between field and no field application. This is likely related to the contact resistance of the interface between the electrodes and the SWCNT/epoxy composite, which is presumably higher for MR1 and MR2.

After specimen retrieval from the mold and thorough cleaning and gold sputtering of the individual disk samples for through-plane resistivity measurements, this property was evaluated. Figure 8 displays the average resistivity of individual samples corresponding to the results in Figure 7. This not only shows the difference between electric field and no field to be less pronounced than in Figure 7, but the resistivity values also show the opposite trend—MR3 leads to the highest average resistivity and MR1 to the lowest. This behavior can be explained by considering the properties of the individual samples after curing under the electric field, depicted in Figures 9 and 10. In these graphs, the electrical resistivity of samples is plotted according to their fixed position in the mold, numbered from 1 to 6. For samples produced without electric field, the electrical resistivity of six samples for each mold release agen<sup>t</sup> does not vary considerably (Figure 10). However, when an electric field is applied, as depicted in Figure 9, it is possible to observe that MR3 and MR2 show a grea<sup>t</sup> disparity of results between the six samples. MR3, for example, has specimen 4 with a resistivity of only 1.7 kΩ·cm, being the most conductive sample produced. However, samples 1, 2, 5 and 6 are all above 10 kΩ·cm.

**Figure 8.** Impact of mold release agen<sup>t</sup> on single disc average resistivity of SWCNT/epoxy 0.01 wt.%.

**Figure 9.** With electric field: Impact of mold release agen<sup>t</sup> on resistivity of SWCNT/epoxy 0.01 wt.% individual samples cured at 100 ◦C.

**Figure 10.** Without electric field: Impact of mold release agen<sup>t</sup> on resistivity of SWCNT/epoxy 0.01 wt.% individual samples cured at 100 ◦C.

A possible explanation for the measurements of total resistivity in Figure 7 is the following phenomenon: given the very low contact resistance of silver paint (MR3), more electrical current can flow between the electrodes. Due to this micro-current, the electric field propagates in the material across the sample thickness as new conductive paths are generated in the material due to CNT dielectrophoresis. However, as soon as one of the six samples becomes more conductive (e.g., sample 4 of MR3 in Figure 9), more and more current starts flowing only through it, since it offers less resistivity, and less current through the other samples (such as 1, 2, 5 and 6 of MR3 in Figure 9). This would explain the scatter in the resistivity values of samples MR2 and MR3 (Figure 9) and the resistivity measured in the closed mold presented in Figure 7. These results highlight the importance of the contact interface (more specifically its resistivity) between the electrodes and the composite for influencing the homogeneous electrical resistivity of SWCNT nanocomposites with electric fields.

#### **4. Discussion and Conclusions**

An optimized dispersion process that limits rupture mechanisms, hence presumably ensuring a high aspect ratio and well dispersed particles, proved to be fundamental when the aim is to decrease the electrical resistivity of thermoset nanocomposites. Once a good level of matrix infiltration, dispersion and distribution in the CNT agglomerates is obtained, it was found that heat application during the curing process further decreased the electrical resistivity of the nanocomposites. This is expected to be due to a reduction in the matrix viscosity at higher temperatures, allowing more mobility of the

particles and hence leading to more contacts between them, as investigated in [22]. This proved to be beneficial when combined with the application of an electric field between the electrodes, leading to a reduction in resistivity of up to one order of magnitude compared to the samples cured without the field. As explored in [22], the clear alignment of CNTs due to the electric field was not observed in scanning electron microscope (SEM) images of sample fracture surfaces. Instead, the reduction in electrical resistivity was attributed to an electric-field-induced increase in contact points between the nanoparticles. This possibility was investigated in [22] with finite element simulations of CNT networks in epoxy at di fferent concentrations.

Furthermore, an RTM set-up for applying electric fields to nanocomposites during curing has been developed and successfully tested. When applying electric fields to the nanocomposite during curing, the conductivity of the interface between the electrodes and the nanocomposite resin was shown to play a crucial role in determining the final resistivity of the bulk materials produced. A highly conductive interface (MR3) promoted more inhomogeneous samples, leading to pronounced changes in the resistivity at a local level (significant di fferences between six samples produced simultaneously). On the other hand, a highly resistive interface (MR1) prevented such local di fferences and resulted in a smaller average resistivity in the samples produced. The use of a nonconductive mold release agen<sup>t</sup> therefore seems more suitable for a scalable industrial process since it leads to composites with more homogeneous electrical properties.

Overall, the use of an electric field during processing shows promising results for enhancing the through-plane electrical conductivity of polymer composites with carbon-based nanoparticles by influencing the structure of the conductive network during the processing step. Such technology would reduce the amount of CNTs necessary to achieve similar electrical properties, hence allowing a cheaper production of composites with a broad range of industrial applications (e.g., sensors, batteries, electrical heating technologies). The aim of further research in this topic should be to understand the impact of the type of carbon nanoparticles used (or a combination of particles with di fferent size scales), the e ffect of stronger electric fields with lower particle concentrations and limitations arising from matrix viscosity.

**Author Contributions:** Conceptualization, M.V.C.M., M.M., C.H. and F.H.; Data curation, M.V.C.M., M.M. and N.S.; Funding acquisition, C.H.; Investigation, M.V.C.M., M.M. and N.S.; Methodology, M.V.C.M. and C.H.; Supervision, C.H. and F.H.; Writing—original draft, M.V.C.M. and M.M.; Writing—review and editing, M.V.C.M., M.M., N.S., C.H. and F.H. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research has received funding from the European Union's Horizon 2020 research and innovation program under the Marie Skłodowska-Curie Grant Agreement no. 642890 (TheLink, www.thelink-project.eu).

**Conflicts of Interest:** The authors declare no conflict of interest.
