3.1. Selection of Deposition Conditions
The process pressure as a function of RF power was characterized for argon (10.0 sccm), oxygen (10.0 sccm), and TVS (4.0 sccm) plasma in
Figure 1. The basic process pressure, when the plasma was turned off, was set to 5.7 Pa using a throttle valve [
21] and the resulting process pressure was independent of RF power during the argon discharge, as expected. While, the process pressure for oxygen plasma increased slightly to 5.5 Pa (5 W) relative to a basic process pressure of 5.0 Pa, due to dissociation of oxygen molecules to reactive oxygen atoms. This oxygen plasma at an RF power of 30 or 100 W was used for substrate pretreatment. The basic process pressure for the TVS precursor at 3.8 Pa was significantly reduced to 1.5 Pa, which is characteristic of the precursor-deficient mode [
24]. This means that the TVS molecules were dissociated into smaller radicals that were highly reactive and resulted in their recombination on the surface of the growing coating. The largest decrease in the process pressure can be observed between 0 and 20 W.
The deposition rate along the axis of tubular reactor was evaluated in
Figure 2 for a TVS flow rate of 4.0 sccm, but a different RF power of 2.0–100 W. The TVS vapors enter the reactor from one end of the tube and exit at the other end connected to the pumping system [
21]. It can be seen in
Figure 2 that the deposition rate is not constant along the tubular reactor, especially in the case of higher power. A deposition rate of 570 nm min
−1 at the precursor input dropped to a minimum of 31 nm min
−1 at a distance of 75 cm from the precursor input for 100 W. In the case of 2.0 W, the deposition rate only slightly varied between 16–49 nm min
−1 along the tube. It has been found that plasma nanocoatings deposited from a mixture of a TVS precursor with oxygen gas at low power efficiently affect the interfacial properties of polymer composites [
25], since the low power resulted in reduced network crosslinking, supported by increased oxygen fraction, which caused a decline of nanocoating mechanical properties favorable to reduce the shear stress across the composite interphase [
16]. For this reason, pulsed plasma at 2.0 W and various oxygen fractions (0–0.71) in TVS/O2 mixture were tested for thin film deposition (
Figure 3). The deposition rate along the tubular reactor varied from 21 to 81 nm min
−1 for an oxygen fraction between 0 and 0.71with a minimum close to half of the reactor. The minimum deposition rate is important to ensure full coverage of all fibers in the bundle by plasma coating to effectively control the interphase in the polymer composite. This parameter was characterized depending on the RF power for pure TVS precursor and depending on the oxygen fraction for an effective power of 2.0 W in
Figure 4. The minimum deposition rate ranged broadly from 16 to 112 nm min
−1, depending on the RF at the top of
Figure 4, but was similar (21–34 nm min
−1) for different oxygen fractions in the lower part of
Figure 4. For TVS/O2 mixture, the TVS flow rate was constant at 4.0 sccm and the oxygen flow rate increased as follows: 0 sccm (zero oxygen fraction), 2.0 sccm (0.33), 2.9 sccm (0.42), 4.3 sccm (0.52), 6.2 sccm (0.61), and 10.0 sccm (0.71).
3.2. Mechanical, Optical, and Chemical Properties of Plasma Nanocoatings
A strong correlation was demonstrated between the shear strength of the polymer composite reinforced with plasma-coated GFs and the nanocoating adhesion [
25], because the shear failure of the composite was shown to be controlled by interfacial adhesion at the nanocoating/glass interface. Importantly, the adhesion of the nanocoating measured on GF is consistent with the adhesion measured on the glass plate using the nanoscratch test [
22]. The critical load required to remove the nanocoating from the glass surface is used to characterize its level of adhesion. This means that the adhesion of the nanocoating increases when the critical load increases. Adhesion of plasma nanocoatings deposited on silicon wafers at selected deposition conditions (2.0 W, 0.71 oxygen fraction) was characterized by a nanoscratch test for samples distributed along the tubular reactor (
Figure 5). The silicon wafer covered with the native silicon dioxide layer is fully comparable to the glass plate in terms of nanocoating adhesion [
25]. The critical load depends on the nanocoating thickness [
26,
27] and, therefore, the deposited nanocoatings had a similar thickness of about 0.10 µm as shown in
Figure 5. The critical load fluctuated only slightly 1.2–1.4 mN along the tubular reactor (
Figure 5), indicating the same adhesion of nanocoatings independent of the deposition rate, see the blue triangles in
Figure 3. FTIR spectra of the same samples are given in
Figure 6 to compare the chemical structure of the deposited nanocoatings along the tubular reactor. Assignment of absorption bands to chemical species [
28,
29] is indicated directly in the spectra. It can be seen that the FTIR spectra are very similar, demonstrating the same chemical character of the deposited nanocoatings, regardless of position along the tube, and thus independent of the deposition rate similar as nanocoating adhesion. The polar groups, hydroxyl (–OH) and carbonyl (–C=O), improve wettability of the nanocoatings with polyester resin and the vinyl groups on the surface of the nanocoatings are responsible for covalent bonding with the polyester resin during the curing process [
30]. It is known that a high concentration of Si–O–C species is responsible for the adhesion of the nanocoating to the glass surface [
16,
17].
Nanocoating adhesion as a function of RF power for plasma nanocoatings deposited on silicon wafers from pure TVS precursor is characterized in
Figure 7. The critical load increased slightly from 1.5 to 2.1 mN with enhanced RF power (2.0–100 W), but higher critical load values may be affected by thicker coating. FTIR spectra corresponding to the samples of
Figure 7 are shown in
Figure 8. It is evident that the concentration of the vinyl groups (1590, 1402, 1007, and 951 cm
−1) decreased considerably with enhanced RF power and no hydroxyl and carbonyl groups were present in the plasma nanocoatings, which could mean reduced interfacial adhesion at the polyester/nanocoating interface, when such a nanocoating is used for surface modification of GFs in GF/polyester composite. The absence of Si–O–C species in nanocoatings should also result in lower adhesion at the nanocoating/glass interface. However, we can expect the formation of the Si–O–C species only at this interface due to the hydroxyl groups typically present on any silicon dioxide surface. This could be a reason for good adhesion even for plasma nanocoatings deposited from a pure TVS precursor. According to Ref. [
15], the reduced concentration of silicon atoms from 9 to 5 at.% and hydrogen atoms from 55 to 53 at.% at the expense of increased carbon concentration from 36 to 42 at.% with enhanced RF power should be responsible for the decreasing intensity of absorption band assigned to the Si-H group in
Figure 8.
Oxidized plasma nanocoatings prepared from the TVS/O2 mixture could therefore be advantageous for nanocoating adhesion. However, the critical load decreased slightly from 1.5 to 1.3 mN with increased oxygen fraction (0–0.71) for plasma nanocoatings deposited at an effective power of 2.0 W (
Figure 9). The corresponding FTIR spectra are shown in
Figure 10. We can see that all spectra are similar. Oxygen atoms were incorporated into nanocoating during the plasma process and form a Si–O–C network with side-hydroxyl and -carbonyl groups bonded to this network. The concentration of all these species, related to the integral intensity of the absorption band, was partially increased with increased oxygen fraction, which is advantageous for improving the interfacial adhesion at both nanocoating interfaces, since the concentration of the vinyl groups (1590, 1402, 1007, and 951 cm
−1) was not affected by the oxygen fraction. The expected improved adhesion at the nanocoating/glass interface for oxidized nanocoatings does not correspond to the trend of the critical load in
Figure 9.
Dispersion curves for the refractive index and the extinction coefficient were determined from ellipsometric spectra for plasma nanocoatings deposited on silicon wafers at selected powers (10–100 W) from pure TVS precursor (
Figure 11). The refractive index was shifted to higher values at enhanced RF power (left in
Figure 11), which corresponds to increased nanocoating density of and also to an increase in mechanical properties (Young’s modulus and hardness) [
19]. As the RF power increases, the TVS molecules are fragmented into increasing number of smaller radicals that recombine forming a more crosslinked and denser network resulting in higher mechanical properties of plasma nanocoatings [
19]. It was found that the Young’s modulus of plasma nanocoating correlates with its refractive index for a selected wavelength of 633 nm [
20] and therefore, the refractive index range (1.59–1.63) in
Figure 11 can be assigned to the Young’s modulus range (9.1–14.8 GPa) with enhanced power from 10 to 100 W. These values correspond to the mechanical properties of the polymer-like materials and are consistent with those determined by nanoindentation measurements [
21]. Polymer-like nanocoatings synthetized by plasma nanotechnology are characterized by a high yield strength at a sufficiently low Young’s modulus and as such are significant for surface modification of GFs used in GF/polyester composites [
17]. The extinction coefficient is simply related to the absorption coefficient and this means that the plasma nanocoatings deposited at lower powers (10 and 30 W) were transparent for visible light (extinction coefficient k = 0), because the absorption edge was below 400 nm (right in
Figure 11). Dispersion curves for the refractive index and the extinction coefficient corresponding to the plasma nanocoatings deposited from the TVS/O2 mixture were similar to those for 10 W in
Figure 11.
3.3. Shear Properties of Glass-Fiber (GF)/Polyester Composites
The bundles of unsized GFs were plasma coated in a tubular PECVD reactor. During the plasma process, the fragments of precursor molecules not only form the plasma nanocoating on the surface of fibers at the edge of the bundle, but also diffuse into the bundle to form the plasma nanocoating on the inner GFs at the center of the bundle. The deposition rate decreases in the direction towards the center of the bundle due to shielding of adjacent fibers. Multiple shielding is controlled by
tn =
ts . qn−1, where
ts is the film thickness on the fiber at the edge of the bundle,
tn is the film thickness on the n-th fiber in the direction toward the center of the bundle shielded by n–1 fibers, n = 22 for a bundle with 1600 single filaments, and
q is the shielding factor of 0.9 [
21]. The shielding effect analysis for a bundle with 1600 filaments predicts that the thickness of the plasma nanocoating on the central fiber is only one tenth of the thickness of the nanocoating on the fiber at the edge of the bundle. The GF bundle was plasma coated using a pure TVS precursor at 4.0 sccm and an effective power of 25 W for a different deposition time of 0–20 min. The corresponding coating thickness for the fiber at the edge of the bundle at a deposition rate of 53.2 nm min
−1 was 0–1064 nm. Untreated and plasma-coated GFs were used to make the GF/polyester composite in a form of short beams. These composite short beams were characterized by the SBS test to evaluate the short-beam strength (Equation (1)), which corresponds to the maximum shear stress of composite failure in interlaminar shear mode [
23]. The short-beam strength of GF/polyester composite as a function of deposition time and coating thickness is given in
Figure 12. We can see that the short-beam strength increased from 13.8 MPa for untreated GFs to a maximum of 30.8 MPa, which corresponds to a coating of 133 nm (2.5 min) on the surface of fibers at the edge of the bundle, followed by a slight decrease to 27.7 MPa, which is within the error bars and, therefore, not statistically significant. Even 1 min deposition leading to a 53.2 nm-thick coating would appear to be sufficient for the proper functioning of the plasma nanocoating in GF/polyester composite. However, previous analysis of the correct coating thickness was performed using the microindentation test [
31], where individual filaments were tested, and the analysis showed that the coating thickness below 100 nm could be insufficient, which would result in lower interfacial shear strength [
15,
16]. Surface chemical analysis of plasma-coated GFs with a coating thickness below 100 nm revealed that some internal filaments in the bundle were not covered with the continuous coating [
16]. Therefore, we decided that a coating thickness of 150 nm on fibers at the edge of the bundle would be sufficient for proper surface modification of GFs and this thickness was used for other plasma coatings.
Plasma nanocoatings deposited from pure TVS precursor at 4.0 sccm and an enhanced RF power of 2.0–100 W were used for surface modification of GFs, which were embedded in polyester resin to produce short composite beams. The short-beam strength of these composites depends on the RF power as shown in
Figure 13. First, the strength increased from 32.0 MPa (2.0 W) to a maximum of 34.9 MPa (5 and 10 W), but then significantly decreased to 28.1 and 26.2 MPa for 25 and 100 W, respectively. This decrease is surprising because the interfacial adhesion at the nanocoating/glass interface (
Figure 7) does not suggest any such decrease, although an increased Young’s modulus (9.1–14.8 GPa) obtained from the refractive index change is considered (
Figure 11). The reason could be a failure at the other interface between the nanocoating and the polyester resin because the concentration of the vinyl groups, responsible for the chemical bond at the polyester/nanocoating interface, decreased significantly with enhanced power (
Figure 8). The short-beam strength for plasma-coated GFs was compared with the strength for oxygen-plasma pretreated GFs, where no coating was deposited, and the corresponding value of 22.4 MPa was indicated in
Figure 13. Model simulations have clarified that an appropriate interlayer (nanocoating) is required for proper interphase functionality between the fiber reinforcement and the polymer matrix in polymer–matrix composites [
16].
Oxidized plasma nanocoatings prepared from the TVS/O2 mixture at an effective power 2.0 W were also deposited on GFs to analyze the effect of oxide species at both nanocoating interfaces on the shear properties of GF/polyester composite. The short-beam strength of these composites was plotted against the oxygen fraction (0–0.71) in TVS/O2 mixture in
Figure 14. The shear strength of GF/polyester composites with GFs coated with oxidized plasma nanocoatings was significantly increased compared to GFs coated with non-oxidized nanocoatings (
Figure 13). This means that a shear strength of 31.1 MPa (zero oxygen fraction) was increased sharply to a higher level of 43.5–44.1 MPa for a wide range of oxygen fraction of 0.33–0.61 and then decreased to 39.4 MPa for the highest oxygen fraction of 0.71. The steep increase was not predicted by the critical load (
Figure 9) for the corresponding oxidized nanocoatings, but could be explained by decreasing the Young’s modulus with enhanced oxygen fraction due to reduced crosslinking of Si–O–C network as found in a recent study [
19]. This assumption of increased adhesion at the nanocoating/glass interface is supported by an increased concentration of Si–O–C species as described above (
Figure 10). In contrast, the increased concentration of carbonyl groups for the highest oxygen fraction (0.71 in
Figure 10) could be responsible for some decrease in shear strength due to reduced interfacial adhesion, where the formation of carbonyl groups prevents covalent bonding to the glass surface as justified in recent studies [
16,
17,
25]. The maximum shear strength for oxidized plasma nanocoatings was 13% higher than 39.2 MPa for industrially sized GFs indicated in
Figure 14. Typical load-displacement curves measured during the SBS test for the unsized (untreated), industrially sized, and plasma-coated GFs were previously shown [
21]. Short composite beams that failed in interlaminar shear mode were delaminated and the spot, where the failure occurred in the plane of the reinforcement, was observed by SEM. SEM micrographs of fractured composite beams corresponding to unsized (untreated) GFs or poor adhesion at polyester/glass interface (oxygen-plasma pretreated GFs) are given in
Figure 15.
Figure 15a shows that even with weak interfacial adhesion between the GFs and the polyester resin some fibers were broken, but the fibers were bare and the polyester resin was almost completely removed. Only a few resin residues can be found on the fiber surface as seen in detailed view (
Figure 15b). In the case of stronger interfacial adhesion (2.0 W, 0.71 oxygen fraction), the fibers were fully covered with a polyester resin that formed the so-called hackles between the fibers in
Figure 16a and in detailed view (
Figure 16b). These hackles are typical of interlaminar mode II shear loading.