3.1. Discharge Properties
Figure 4 shows the time dependence of the consumed power for three plasma devices. It can be seen that the consumed power values of the two fluorinated devices were higher than that of the un-fluorinated one at the aging time of 0 h, then, the consumed power of the three devices increased with the aging time; this was mainly because the plasma was situated nearer the grounded electrode with the degradation of the top PI layer, and therefore in a region where the expected electric field strength was higher [
14]. Besides, the power data in case of 1# device can be well fit by a simple linear Function (3), the calculated
A and
B values were 0.88 and 36.54, respectively. A second-order polynomial Function (4) can be used to fit the cases of 2# and 3# devices, the calculated
C,
D and
E values were 0.024, 0.027, 38.17 for 2# device and 0.017, −0.017, 38.72 for 3# device. The fitted results show that the plasma processing had less effects on the consumed power values of fluorinated devices. For instance, after aging time of 20 h, the consumed power of 3# device increased by about 18%, which was about 46% lower than the case of 1# device. In Ref. [
14], a method using a coating of Polydimethylsiloxane oil was proposed to prevent erosion of the PI-based actuator. Similar phenomenon of slower increase rate of the consumed power has also been observed.
where
A and
B are free parameters.
where
C,
D and
E are free parameters.
Lissajous figures corresponding to four different discharge cases marked in
Figure 4 are shown in
Figure 5a,b, for 1# device and 3# device, respectively. It is seen that the shapes of Lissajous figures did not change apparently. However, the enclosed area was expanded for both cases after aging and the case of 1# device seemed more pronounced, which led to higher
Ek and
P. The calculating results of
C0 and
Ceff values at different aging times for the three devices are given in
Figure 5c,d, along with the fits based on Function (3). In case of
C0 fitting, the values of
A were −8 × 10
−4, 0.012 and 0.015, while the values of
B were 4.50, 4.38 and 4.28, for the cases of 1#, 2# and 3# devices, respectively. In case of
Ceff fitting, the values of
A were 0.188, 0.098 and 0.049; while the values of
B were 13.58, 13.75 and 13.89, for the cases of 1#, 2# and 3# devices, respectively. Based on the fitted lines, it can be seen that for 1# device,
C0 values remained almost unchanged with the aging time, while
Ceff increased significantly from 13.5 ± 0.3 to 17.3 ± 0.3 pF (about 28% higher). Compared to 1# device, both
C0 and
Ceff values had slight increase for 2# and 3# devices. Taking 3# device as an example, after 20 h of plasma discharge aging,
C0 values increased from 4.3 ± 0.1 to 4.6 ± 0.1 pF, while
Ceff values increased from 14 ± 0.3 to 14.9 ± 0.3 pF (about 6.4% higher), which was much lower than variation range of 1# device. According to Ref. [
14], the increase range of
Ceff value directly revealed the degradation degree of the dielectric beneath discharge plasma; it can be concluded that when using fluorinated PI as a dielectric, the dielectric degradation was effectively inhibited.
3.3. Surface Temperature and Plasma Diagnosis
Surface temperature of three devices at different aging times (0, 10 and 20 h) was measured using a thermal infrared camera.
Figure 7 quantitatively gives the temperature distribution along the edge of the exposed electrode (
x = 0, 5 and 10 mm). For the three devices at different aging times, the
y direction distribution indicated a maximum temperature rise on the edge of the exposed electrode (
x = 0 mm). A gradual drop in temperature was observed on the downstream side of the electrode (
x = 5 and 10 mm). Also, the temperature perturbation at
x = 0 mm was strongest and the more uniform distribution appeared at a downstream location (
x = 5 and 10 mm). It is further noted that the surface temperature value in the same positon gradually increased with the aging time, and the surface temperature variations of the un-fluorinated PI were more pronounced and affected by long plasma processing. For instance, after 20 h of discharge aging, the maximum temperature rise at the position of
x = 0 mm was higher in the case of the un-fluorinated PI (about 19 °C) than the cases of the 30 and 60 min fluorinated PI dielectrics (about 3.6 and 2.9 °C, respectively). At the position of
x = 20 mm, the maximum temperature of the un-fluorinated PI increases about 9 °C, while no obvious variation shown in
Figure 7c can be observed for the 60 min fluorinated PI.
Optical emission spectroscopy (OES) is used to study discharge plasma. The emission spectrum in the range of 360–400 nm emitted from the plasma device with un-fluorinated PI is shown in
Figure 8. The emission intensity was normalized with the intensity at 380.5 nm. As expected with other atmospheric air non-equilibrium discharge [
32], N
2 peaks were dominant in the emission spectra. The major spectra came from the second positive system (SPS) of N
2 (C
3П
u→B
3П
g) and the first negative system (FNS) of N
2+ (B
2Σ
u+→X
2Σ
g+). Typical peaks were identified according to Refs. [
33,
34,
35], which included N
2 (C
3П
u→B
3П
g, 2–4) at 371.1 nm, N
2 (C
3П
u→B
3П
g, 1–3) at 375.5 nm, N
2 (C
3П
u→B
3П
g, 0–2) at 380.5 nm, N
2+ (B
2Σ
u+→X
2Σ
g+, 0–0) at 391.4 nm, N
2 (C
3П
u→B
3П
g, 2–5) at 394.3 nm, and N
2 (C
3П
u→B
3П
g, 1–4) at 399.5 nm.
The typical parameters such as N
2 (C
3П
u) vibrational temperature (
Tvib), electron temperature (
Te) and N
2 (C
3П
u) rotational temperature (
Trot) are important parameters to evaluate plasma characteristics. The relative intensity ratio between 371.1 and 380.5 nm (
I371.1/
I380.5) can be used as an indicator of
Tvib; the relative intensity ratio between 391.4 and 380.5 nm (
I391.4/
I380.5) is used to obtain the temporal and spatial averaged
Te by the line-ratio technique of OES [
36,
37]. The variations of the two parameters at the discharge aging time of 0 and 20 h are shown in
Figure 9a,b, respectively. For the three devices at the aging time of 0 h, the
I371.1/
I380.5 and
I391.4/
I380.5 values remained almost the same (around 0.19 and 0.11, respectively), suggesting that there were no conspicuous effects of surface fluorination on
Tvib and
Te. When the discharge time increased to 20 h,
I371.1/
I380.5 values of all devices shown in
Figure 9a still had little changes (from around 0.19 to 0.20) during the plasma processing, suggesting that
Tvib had minor dependence on the discharge aging time. Nevertheless, for 1# device shown in
Figure 9b, the
I391.4/
I380.5 value increased apparently from 0.11 (0 h) to 0.164 (20 h), indicating a remarkable change of plasma characteristics—an increase in the number of high energy electrons and the average electron energy [
38]. Compared to 1# device, the changes of the
I391.4/
I380.5 values for 2# and 3# devices were not obvious (from about 0.11 to 0.12). The different changes of
I391.4/
I380.5 values for three devices can be explained by the distribution of surface temperature. As shown in
Figure 9c, at the aging time of 0 h, the temperature distribution was uniform along the edge of the exposed electrode. However, at the aging time of 20 h, intense temperature perturbation can be clearly seen from
Figure 9d, which was caused by the severe dielectric degradation of the un-fluorinated PI. As a result, the electrical field distortion became worse and more direct heat injected from the irregular glow spots to the dielectric surface, where the intense concentration of the ionization events occurred randomly [
29]. The existence of glow spots in the negative-going cycle meant that the electrons transferred from the exposed electrode to the dielectric surface, combined with the fact that more discharge power was consumed by 1# device after 20 h. It can be inferred that the discharge energy deposited on each electron transport channel also became higher. Therefore,
Te increased greatly because it was determined by the optical emission from the strong and bright discharge channels. At the same time, no noteworthy temperature perturbations can be seen in
Figure 9f,h.
The values of
Trot are further determined by comparing the experimentally measured data to the theoretically calculated data with a least-square procedure.
Trot is obtained by making the squared difference between measured and calculated normalized intensity minimum. In addition, gas temperature (
Tgas) can be regarded as being almost equal to
Trot because the equilibrium between rotational and translational motion is easily obtained due to frequent and fierce collisions among the heavy particles [
39]. The calculating results of three cases are shown in
Figure 10, along with fits based on Function (3). The fitting values of
A were 1.04, 0.60 and 0.50, while the values of
B were 370.40, 370.65 and 370.17, for the cases of 1#, 2# and 3# devices, respectively. As presented in the embedded figure, the simulated spectrum at
Trot = 392 K fits the experimental one with good agreement. It can be seen that plasma discharge aging had more effects on 1# device than the other two devices. After 20 h, the gas temperature of 1# device increased from about 370 to 390 K; this was due to the increase of electron density and electron mean energy, more frequent collisions between electrons and heavy particles occurred, and therefore more energy transferred from electrons to heavy particles. At the same time, the increase ranges of 2# and 3# devices were lower than that of 1# device, suggesting that more stable plasma discharge can be obtained after dielectric surface fluorination.
3.4. Surface Morphology and Chemical Structure
Figure 11 shows surface morphologies of the PI dielectrics before plasma discharge aging. As can be seen, there was no significant difference between surfaces of the un-fluorinated PI dielectric and the fluorinated PI dielectric; all of them had smooth surfaces without potholes, except some occasionally distributed micro size dust particles. Cross-section image of the 60 min fluorinated PI dielectric is given in
Figure 11d. It can be seen that a fluorinated layer was formed after fluorination, but the matrix of the dielectric is not affected.
Figure 12 shows surface morphologies of the PI dielectrics after 10 h of plasma processing. All the images were taken in the plasma discharge region around the edge of the exposed electrode. It can be clearly observed that the plasma processing caused distinct changes of surface morphologies. As shown in
Figure 12a, the surface of the un-fluorinated PI was severely etched and damaged, many irregular small gibbosities remained on the surface, around which there were lots of bubbles. These gibbosities and bubbles were the PI material and the underlying Si-based adhesive, respectively. The isolated gibbosity-like morphology can be ascribed to the space among the randomly distributed micro-discharge, where the active species are generated along the filamentary channel and spatially uneven distributed in the discharge regime [
40]. By contrast,
Figure 12b,c show that surface morphologies of 30 and 60 min fluorinated PI dielectrics were more integrated, and the surface roughness was increased after discharge aging. However, no obvious differences in the morphologies can be found between the two surfaces.
XPS analysis was further conducted to investigate changes in the surface chemical composition of un-fluorinated PI and fluorinated PI, before and after plasma processing. As shown in
Table 1, for the un-fluorinated PI after 10 h of plasma processing, the concentrations of both carbon and nitrogen elements decreased sharply, while the concentration of oxygen element increased remarkably from 17.81% to 38.72%. Besides, a new element (Si) was detected and accounted for 28.57% of chemical composition; this was due to the fact that long plasma processing caused the exposure of the underlying Si-based adhesive. Compared to un-fluorinated PI, F element was detected on the surfaces of fluorinated PI. When the fluorination time increased from 30 to 60 min, there was no distinct difference between surface chemical composition of 30 min (2#-0 h) and 60 min (3#-0 h) fluorinated PI dielectrics. However, after 10 h of plasma processing, it can be seen that F element concentration of 60 min fluorinated PI was about 2.3 times as the 30 min fluorinated one.
In order to examine the changes of functional carbon groups on the surfaces, deconvolution analysis of the C1s peaks were executed; the results are shown in
Figure 13 and
Table 2. It is noted that C1s spectra of un-fluorinated PI had four peaks, namely, C–C/C–H, C–N, C–O, and C–N, which respectively corresponded to the binding energies of 284.7, 285.6, 286.3, and 288.6 eV [
24,
25]. In addition to the above four peaks, the C1s spectra of surface fluorinated PI incorporated two new peaks, at the binding energies of 287.1 and 289.5 eV, which corresponded to the C–F
n and C–F peaks [
26], respectively.
For un-fluorinated PI, the concentrations of C–C/C–H, C–N, C–O, and C=O groups were 69.65%, 11.87%, 8.3% and 10.17%, respectively. After 10 h of plasma processing, the C–C/C–H group increased sharply to 85.78% while the concentration of C–N group decreased significantly to 2.64%. Meanwhile, the concentration of C=O also showed a sharp decreasing tendency. These changes can be explained by the reactive and heat effect of DBD micro-discharge region on the PI surface. For the fluorinated PI before plasma processing, when fluorination time was increased from 0 to 30 and 60 min, the concentration of C–C/C–H group decreased sharply from 69.65% to 61.28% and 56.44%, while concentration of C–F group increased to 13.06% and 13.49%, and concentration of C–Fn group increased to 13.05% and 13.01%. This result revealed that the C–C/C–H bonds can be easily broken during plasma processing. Besides, there was no significant concentration variation of fluorine-containing groups when fluorination time was increased from 30 to 60 min, which agreed well with the surface elements composition. After 10 h of plasma processing, the fluorine-containing group concentration of the 60 min fluorinated PI (~22.15%) was much higher than that of the 30 min fluorinated PI (~12.84%).