2.1. Functionalization of BT Particles
The study of BT functionalization was done on commercial particles (BTC), which were reacted with H
2O
2 for different reaction times and then with 3-glycidyloxypropyltrimethoxysilane. To evaluate the effect of the hydroxylation step on functionalization, TGA analyses were carried out on functionalized BTC particles previously subjected to hydroxylation, respectively, for 4, 8, 24 and 48 h (
Figure 1a). For all the samples, the first weight loss observed in the range 100–300 °C is due to adsorbed water and residual hydroxyl groups and appears more pronounced for BTC_8_G (8 h hydroxylation) and BTC_24_G samples (24 h hydroxylation).
Figure 1b presents the TGA curves of the sample hydroxylated for 8 h (BTC_8) and subsequently functionalized (BTC_8_G). The comparison points out the reduction in the amount of residual -OH as a consequence of the reaction with the organosilane. The major weight loss in the range between 300 °C and 600 °C is due to the thermal decomposition of the organic groups introduced by organosilane grafting and depends on the hydroxylation time. The functionalization is not favored by hydroxylation times longer than 8 h; indeed, sample BTC_8_G presents the highest mass loss in the range 300–600 °C. These results can be explained by considering that, under the experimental conditions used for hydroxylation, condensation of -OH groups can be favored by prolonged treatments, thus reducing their availability for functionalization. The measured weight losses in the range 300–600 °C (as reported in
Section 3.5) were used to calculate the content of organosilane grafted on BTC particles hydroxylated for different time (
Table 1).
Solid-state NMR was applied to study the degree of organosilane condensation, check the integrity of the glycidoxypropyl chain and evaluate the functionalization yield.
Figure 2a shows the
13C CPMAS spectra of neat organosilane and the BTC_8_G sample; carbon resonances are assigned according to the scheme reported in the figure. All the signals belonging to the organosilane functional group are present in the spectrum of the BTC_8_G sample. In detail, the resonances of methylene carbons can be identified at 9 (1), 23 (2) and 73.1 (3, 4) ppm, respectively; signals due to methine (6) and methylene (5) carbons of the epoxy group are found at 43.11 and 50.9 ppm, respectively [
22]. The -OCH
3 resonance is not observed at around 50 ppm, indicating the complete organosilane hydrolysis. The broadness and downfield shift (with respect to the position for the pure reagent) of peak 1 indicate the occurrence of condensation, which could be due both to grafting onto the particles’ surface or organosilane self-condensation. Partial epoxy ring opening with the formation of diols is indicated by the reduced intensity of peaks 5 and 6 with respect to the others, the presence of a weak resonance at 64.8 ppm and the intensity of resonances 3 and 4. Similar results were achieved for the other samples (
Figure S1, see
Supplementary Materials).
Organosilane condensation is better proven by the
29Si CPMAS spectrum of the BTC_8_G sample (
Figure 2b), which presents two resonances, respectively, assigned to T
2 (at −56.5 ppm, representing SiCO
3 units with two bridging oxygens) and T
3 (66.0 ppm, SiCO
3 fully condensed units) Si units, with ratio 7/93, as confirmed by the quantitative
29Si MAS experiment (not reported).
The amount of grafted organosilane in the BTC samples was calculated from
13C spectra according to the procedure described in
Section 3.5, and the results are shown in
Table 1.
The quantitative results of 13C NMR and TGA analyses are in good agreement and point out that running the hydroxylation reaction for 8 h leads to the highest functionalization degree of BT nanoparticles.
Figure 3 shows the SEM images of pristine and hydroxylated nanoparticles, namely BTC, BTC_8, BTC_24 and BTC_48. The starting commercial nanoparticles (
Figure 3a) possess an average diameter of 500 nm, with broad size distribution [
21] and a limited degree of aggregation, which is enhanced by the hydroxylation step (
Figure 3b), particularly for longer times (
Figure 3c,d). Aggregation of the particles caused by prolonged hydroxylation justifies the lower functionalization degree detected by TGA and NMR. On the other hand, too short hydroxylation times are not able to introduce sufficient hydroxyl groups for the subsequent reaction. A hydroxylation step of 8 h was therefore selected for studying the functionalization with different organosilanes of sol-gel nanoparticles synthesized in hydrothermal conditions [
21].
Figure 4a shows the
13C CPMAS NMR spectra of the particles obtained by hydrothermal synthesis (BTH), hydroxylated for 8 h and functionalized with 3-glycidyloxypropyltrimethoxysilane (G), 2-[(acetoxy(polyethyleneoxy)propyl]triethoxysilane (A) and triethoxysilylpropoxy (polyethyleneoxy)dodecanoate (T), respectively. The
13C NMR spectra of neat organosilanes are reported in
Figure 4b for comparison, together with molecular structures and carbon labeling. All the signals of the corresponding organosilanes are present in the functionalized BTH nanoparticles, proving effective grafting. The BTH_G spectrum is similar to the one recorded on the BTC_8_G sample, but the peak intensity suggests an improvement in the functionalization favored by the higher surface area displayed by BTH particles, which present an average diameter of 120 nm [
21].
In the case of BTH_A, the signal to noise ratio (S/N) is quite low as a consequence of the reduced amount of anchored silane in comparison with G-functionalized particles. This is likely due to the silane reactivity, which is negatively affected by both steric and electronic inductive effects of the long organic chain. The downfield shift at 10 ppm of resonance 1 indicates condensation, but residual ethoxy groups are still visible at 60 and 19 ppm. The C=O resonance is split into two signals at 183 and 169 ppm with a ratio of 82/18, probably as a consequence of different local environments experienced by the carbonyl groups in open or folded chains.
The spectrum of BTH_T deserves similar comments: the S/N is quite low, accounting for a low amount of anchored silane, and residual ethoxy groups (60 and 15 ppm) suggest incomplete hydrolysis.
The 29Si spectra of BTH_A and BTH_T samples (not shown) suffer from the low amount of Si in the whole sample, leading to a very low S/N. In both samples, a broad resonance due to the overlap** of T2 (at −55 ppm) and T3 (at −65 ppm) peaks, with a ratio T2/T3 of around 30/70, can be argued, indicating that the condensation degree is lower with respect to the case of G-functionalized particles.
Thermogravimetric analyses were carried out on the functionalized BTH particles to quantify the degree of grafting, obtaining TGA curves similar to those of functionalized BTC samples. For the BTH_G sample, the weight loss in the range 300–600 °C attributed to the G silane decomposition was 3.3 wt.%, thus confirming that the functionalization was enhanced with respect to the commercial BT particles (
Table 1).
BTH particles functionalized with the A and T organosilanes present comparable weight losses, namely 3.0 and 2.5 wt.%, respectively. Considering the molecular weight of the three employed organosilanes, A and T lead to a more difficult grafting onto the surface of the particles with respect to G, as suggested by NMR. The degree of functionalization was evaluated according to Maitra et al. [
23] by calculating the grafting density of the organosilanes
, expressed as the number of grafted molecules per square nanometer (
Section 3.5, Equation (1)). Resulting grafting densities are 10.6 molecules nm
−2 for BTH_G, 3.5 molecules nm
−2 for BTH_A and 3.2 molecules nm
−2 for BTH_T. These values are in agreement with NMR considerations. It must be considered also that A and T not only have a longer organic chain with respect to G, but also present different alkoxy groups. The ethoxy group has, in general, lower reaction rates with respect to the methoxy group [
18], and this may also play a role. Nevertheless, grafting of all the considered organosilanes is proven successful.
2.2. BT-Epoxy and BT-PDMS Nanocomposites
Functionalization of hydrothermally synthesized nanoparticles has been shown to produce a higher density of organosilane molecules on the particles’ surface with respect to commercial particles, thus making BTH more attractive for composite production. Accordingly, polymer-based nanocomposites were produced with both bare and G-functionalized BTH particles. PDMS- and epoxy-based composites (
Figure 5) with 3.5% and 21% volumetric BTH_G content were produced to assess the effect of functionalization on the dispersibility of BT particles. To compare the effect of functionalization with different organosilanes, nanocomposites with BTH_A particles were prepared at 3.5 vol.% content. The samples were produced by casting the polymer mixtures containing both bare and modified nanoparticles and applying a combined poling/thermal curing process.
PDMS composites showed generally higher flexibility and a better appearance with respect to the epoxy ones, with smoother surfaces and a lower concentration of surface defects. However, the preparation of the PDMS sample loaded with bare BTH particles at 21 vol.% filler content was not totally successful, as the sample remained sticky even after a post-curing treatment. On the contrary, the same composition with G-functionalized BTH particles was successfully produced, proving the positive effect of particle functionalization for the achievement of high filler loading. Differential scanning calorimetry (DSC) analyses of composites (
Table S1, see
Supplementary Materials) demonstrated that, in general, the addition of the NPs does not lead to relevant changes in the thermal properties of the polymeric matrix. All PDMS-based samples present similar T
g values (120–125 °C). The faultiness of the sample pdms_21_BTH could be due to the non-optimized production process. Furthermore, a low filler concentration does not have remarkable effects on crystallization and melting enthalpies, but at a high filler concentration, we record lower values for melting and crystallization enthalpies. Concerning the epoxy-based samples, the glass transition temperature (T
g) of epoxy samples is in the range 6–9 °C, except for the sample epoxy_21_BTH_G, which presents a T
g = 26 °C. In this case, the high loading of functionalized NPs provides a large number of oxirane rings available for the ring-opening reaction, improving the cross-linking degree.
SEM images of epoxy composites are shown in
Figure 6 to highlight the different particle/matrix interfaces induced by the functionalization. The cross-sections analyzed correspond to the sample regions between the electrodes, subjected to the DC field.
Figure 6a shows the epoxy composite loaded with bare BTH NPs at 21 vol.% filler content. The higher magnification highlights the poor filler–matrix interaction, showing well-defined particles. Using BTH_G particles (
Figure 6b) at the same filler content leads to a more regular and smoother surface, and particles appear embedded in the epoxy matrix, thus confirming the positive effect of functionalization on the polymer-BT interface. Corresponding composites at 3.5 vol.% filler content present similar features. BTH_A particles behave similarly, leading to an improvement in the filler–matrix compatibility of the epoxy composite at 3.5 vol.% loading (
Figure S2, see
Supplementary Materials).
EDXS elemental maps of Ba were recorded on sections subjected to the DC field of epoxy composites loaded at 3.5 vol.% in order to evaluate the dependence of particle dispersion on organosilane functionalization (
Figure 7). With bare BTH NPs (
Figure 7a), the composite presents regions rich in particles aggregated in clusters and others completely lacking fillers. Functionalization of BTH particles with the G silane (
Figure 7b) produces an outcome with particles finely distributed, except for some clusters that appear slightly smaller compared to
Figure 7a. On the contrary, BTH functionalization with A silane leads to a large improvement in particle dispersion in the epoxy matrix. The higher BTH_A dispersibility with respect to BTH_G is probably related to the different reactivity of the end-chain group of the two silanes, which leads consequently to the different strength of the filler–matrix interface. In the case of G, the epoxy ring is expected to react with the epoxy matrix, while the intensity of interactions of A-functionalized particles with the epoxy matrix is lower, resulting in less hindered particles and better dispersibility.
SEM images of the cross-sections of PDMS composites show good matrix-filler continuity without noticeable differences, regardless of the employed particles. On the contrary, EDXS elemental maps of Ba highlight significant variations (particle alignment) due to the application of the DC field on composites produced with particles differing in size and surface modification (
Figure 8). The alignment of BT particles in polymeric matrices has been already reported in the literature, and it was found to depend on several factors, such as particle size, aspect ratio, applied field, etc. [
24]. In this work, with bare BTH particles (
Figure 8a), large clusters (and depletion regions) are formed, giving evidence of a poor matrix-filler compatibility. At lower magnification, large clusters can be identified with partial segregation of the particles at the bottom of the sample, from which macrometric filaments extend vertically in the direction of the electric field, reaching in some cases the top face of the sample (
Figure S3, see
Supplementary Materials). The functionalization with the G silane of these particles (
Figure 8b) further highlights this behavior. Filaments of a few tens of micrometers are observed, aligned along the direction of the applied external electric DC field. Employment of BTH particles functionalized with the A silane (
Figure 8c) produces a result very similar to that of BTH_G particles (
Figure 8b).
From these results, surface modification of the particles with G and A organosilanes clearly affects the distribution of BT particles in both epoxy and PDMS composites: in the first case, homogeneity in the distribution of the particles is enhanced, achieving the best results with the A silane; in the second case, the chemical modification allows one to increase the filler content, leading to the alignment of the particles along the poling direction and avoiding the formation of large, dense clusters and segregation of the particles at the bottom of the sample, with no detectable difference between the two silanes. There are almost no studies in the literature on the effect of silane coupling agents on BT NPs’ alignment. Todd and Shi [
25] suggested that the presence of silane coupling agents could play a crucial role in modifying the molecular polarizability. Accordingly, it is worth noting that the results suggest the dependence of dispersibility and the alignment process on the chemical features of particle–matrix interfaces.
2.3. Dielectric Constant
Figure 9 shows the results of dielectric constant measurements recorded on the different nanocomposites. Epoxy-based samples (
Figure 9a) show a more pronounced decrease in the dielectric constant upon increasing the frequency, with respect to PDMS-based composites (
Figure 9b). The decrease is due to the fact that dipoles are not able to follow fast orientation changes [
26]. The higher polarity of epoxy with respect to silicone is responsible for the different behavior between the matrices, resulting in a decrease over the frequency of permittivity [
26]. Furthermore, in the analyzed frequency range, the addition of barium titanate seems not to have effects on the decrease in dielectric constant over frequency. The dielectric properties of BT particles are demonstrated to be strongly dependent on the particle size. Wada et al. [
27] reported high values of dielectric constant (up to 5000) with a BT particle size of 140 nm; our BTH particles present dimensions around 120 nm and this is highly beneficial to achieve a high dielectric constant. Accordingly, with BTH addition in epoxy-based samples (
Figure 9a), the dielectric constant increases with a dependence on the particle load and functionalization [
28]. Both at high and low filler loadings, a larger improvement is obtained with functionalized nanoparticles with respect to bare ones. Concerning PDMS samples, the situation is slightly different. Functionalization is not shown to improve the dielectric constant as much as in epoxy samples, and this is probably due to the different interaction of the silane coupling agent with the matrix. In fact, covalent bonds can be created between G-functionalized particles and the epoxy matrix through epoxy ring-opening reactions, while only weak interactions can take place between PDMS and NPs, resulting in a weaker particle–matrix interface compared to the epoxy-based system. However, the values of dielectric permittivity for PDMS samples at high BT loading values are particularly interesting when compared to the results reported in the literature [
29], making the produced nanocomposites attractive for applications in the field of dielectric elastomers.