Commercial multi-walled CNTs (Nanocyl - 3100) have been used as received (sample CNTs). Further details on this material can be found elsewhere 10. CNTs sample was functionalized by oxidation under reflux with HNO₃ (7 M) for 3 h at 130°C, followed by washing with distilled water until neutral pH, and drying overnight at 120°C (sample CNTox was obtained). The CNTox material was heat treated under inert atmosphere (N₂) at 400°C for 1 h (sample CNTox400) and at 900°C for 1 h (sample CNTox900), to selectively remove surface groups. The obtained samples were characterized by adsorption of N₂ at -196°C, temperature-programmed desorption (TPD) and determination of pH at the point of zero charge (pH_PZC) from acid-base titration according to the method of the literature 11. The total amounts of CO and CO₂ evolved from the samples were obtained by integration of the TPD spectra.

Polymer films with thicknesses between 40 and 50 μm were produced by mixing different amounts of CNT (from 0.1 to 1.0%) with N, N-dimethylformamide (DMF, Merck 99.5%) and PVDF (Solef 1010, supplied by Solvay Inc., molecular weight = 352 × 10³ g/mol) according to the procedure described previously 9. Solvent evaporation, and consequent crystallization, was performed inside an oven at controlled temperature. The samples were crystallized for 60 min at 120°C to ensure the evaporation of all DMF solvents. After the crystallization process, the samples were heated until 230°C and maintained at that temperature for 15 min to melt and erase all polymer memory. This procedure produced α-PVDF crystalline phase samples 12.

Topography of the samples and CNT distribution was performed by scanning electron microscopy (SEM, FEI - NOVA NanoSEM 200). The dielectric response of the nanocomposites was evaluated by dielectric measurements with a Quadtech 1920. Circular gold electrodes of 5-mm diameter were evaporated by sputtering onto both sides of each sample. The complex permittivity was obtained by measuring the capacity and tan δ in the frequency range of 100 Hz to 100 kHz at room temperature. The volume resistivity of the samples was obtained by measuring the characteristic I-V curves at room temperature using a Keithley 6487 picoammeter/Voltage source.

Oxidations with HNO₃ originate materials with large amounts of surface acidic groups, mainly carboxylic acids and, to a smaller extent, lactones, anhydrides, and phenol groups 101314. These oxygenated groups (Figure 1) are formed at the edges/ends and defects of graphitic sheets 15. The different surface-oxygenated groups created upon oxidizing treatments decompose by heating, releasing CO and/or CO₂, during a TPD experiment. As this release occurs at specific temperatures, identification of the surface groups is possible 101314. It is well known that CO₂ formation results from the decomposition of carboxylic acids at low temperature, and lactones at higher temperature; carboxylic anhydrides originate both CO and CO₂; phenols and carbonyl/quinone groups produce CO 101314.

Figure 2 shows the TPD spectra of the CNT before and after the different treatments. It is clear that the treatment with HNO₃ produces a large amount of acidic oxygen groups, such as carboxylic acids, anhydrides, and lactones, which decompose to release CO₂. Part of these groups (carboxylic acids) is removed by heating at 400°C. A treatment at 900°C removes all the groups, so that the obtained sample is similar to the original. The total amounts of CO and CO₂ evolved from the samples, obtained by integration of the TPD spectra, are presented in Table 1.

All the samples release higher amounts of CO than CO₂ groups (Table 1). The CNTox sample has the highest amount of surface oxygen. This sample also presents the lowest ratio CO/CO₂ and the lowest value of pH_PZC, indicating that this is the most acidic sample. CNTox900 presents the highest CO/CO₂ ratio, suggesting the less-acidic characteristics, which matches well with the pH_PZC results (Table 1). The acidic character of the samples decreases by increasing the thermal treatment temperature, since the acidic groups are removed at lower temperatures than neutral and basic groups, as seen in previous studies 101314.

The CNT samples have N₂ adsorption isotherms of type II (not shown), as expected for non-porous materials 16. The surface areas of the samples, calculated by the BET method (S_BET), are included in Table 1. It can be observed that the oxidation treatments lead to an increase of the specific surface area. This occurs because the process opens the endcaps of CNTs and creates sidewall openings 17. The specific surface areas of the samples slightly increase as the thermal treatment temperature increases, since carboxylic acids and other groups, introduced during oxidation, are removed.

The morphology and fiber distribution of the composite samples were analyzed by SEM to evaluate the CNT dispersion in the polymeric matrix and determine how the composites influence the polymer crystallization microstructure. Figure 3 shows the SEM images for the PVDF/CNT composites. The main relevant microstructural feature of the composite is that the CNT are randomly distributed into the polymeric matrix. The spherulitic structure characteristic of the pure PVDF is still present in all the composites samples 1218.

CNT agglomerates are nevertheless more often observed for the CNTox composites samples, especially for the ones treated at the highest temperatures. With respect to the electrical properties, oxidation reduces the composite conductivity for a given concentration and shifts the percolation threshold to higher concentrations (Figure 4). This behavior is mainly due to the reduction of the surface conductivity of the CNTs due to the oxidation process 8, and is similar for all the functionalized composites. Further, the increase of surface area due to the functionalization treatment certainly causes surface defects on the CNTs that also reduced electrical conductivity. The increase of agglomerations for the treated samples should not have, on the other hand, a large influence in the electrical response 8. A change of several orders of magnitude of the electrical resistivity with increasing CNTs concentration was observed for all samples, indicating a percolative behavior of the nanocomposites. In general, both in surface (not shown) and in bulk resistivity (Figure 4a), the percolation threshold appears between 0.2 wt.% for the original CNT samples and shifts to 0.5 wt.% CNTs for the functionalized nanocomposites.

Dielectric measurements show that the incorporation of the CNT in the PVDF matrix but leads to a gradual increase of the dielectric constant (ε') as the amount of the filler is increased (Figure 4b). The increase of the ε' is larger for the pristine CNT. A maximum for the 0.5% pristine CNT sample with ε' 22 at a frequency of 10 kHz at room temperature was found, whereas for the functionalized nanocomposites the value is 16. The frequency behavior of the dielectric permittivity is similar to the one obtained for the pure polymer, except for an increase of the low frequency dielectric constant and dielectric loss (not shown) with increasing CNT loading due to interfacial polarization effects (Figure 4b). No noticeable differences have been observed for the different oxidation treatments in terms of the dielectric response. In a previous study 19, it was demonstrated that an increase in the dielectric constant is related with the formation of a capacitor network.