3.1. Analysis of Three-Dimensional Birefringence
The birefringence of the injection-molded COC plates was examined with the polarimeter. First, as shown in
Figure 6, the birefringence parallel to the specimen’s xy-plane was evaluated.
The experiment reveals that, on the x-axis and thus parallel to the melt flow direction, the COC plate exhibits maximum birefringence at the injection gate. With increasing distance from the gate, the birefringence decreases linearly and approaches zero at the COC plate’s opposite end. Parallel to the y-axis, and thus orthogonal to the melt flow direction, the birefringence is constant, whereas the absolute value is again determined by the distance between the measurement region and gate. The local surface birefringence at the positions A, B and C (see
Figure 6a) is also determined by means of the prism coupler. While the surface birefringence at positions A and B is below the device’s theoretical measurement resolution (≤2 × 10
−4), a surface birefringence of 5 × 10
−4 is observed at position C. This value differs significantly from the birefringence observed via polarimetry, which amounts to approximately 1.2 × 10
−4 at this position. This discrepancy can be explained by an inhomogeneous birefringence distribution throughout the COC plate’s z-axis. To examine this, slab-shaped samples with a length of 10 mm and a width of 0.2 mm were cut from the 1.5 mm thick COC plate. Again, the samples were taken from regions A, B and C, whereas their smallest extension (width) was oriented parallelly to the measurement plane. Thus, depending on the specimen orientation, the birefringence in the xz- as well as in the yz-plane was evaluated. A summary of the obtained results is provided in
Figure 7.
The polarimetry results reveal an inhomogeneous birefringence distribution within injection-molded COC plates. In all cases, the specimens exhibit maximum birefringence at and just underneath both of the COC plate surfaces. Thereby, the maximum birefringence value depends on the specimen position as well as the measurement direction. Parallel to the x-axis and thus in the melt flow direction, the maximum birefringence (yz-plane) is determined as 5 × 10
−4 at the sample surface. The birefringence then decreases exponentially and saturates at values around 1.8 × 10
−4, in a depth of 0.4 mm. In this measurement direction, the results are independent of the sample position. In contrast, when measuring the birefringence along the COC plate’s y-axis, i.e., orthogonal to the melt flow direction (xz-plane), the maximum birefringence can increase significantly and is also position-dependent. At position A, the maximum birefringence as well as its depth profile are comparable to the situation parallel to the yz-plane. However, at positions B and C, maximum birefringence values of up to 8.4 × 10
−4 were determined. Again, the values were compared to the surface birefringence obtained by means of the prism coupler. The results are summarized in
Table 2.
In general, the surface birefringence values determined with both experimental setups are comparable. While the surface birefringence is constant in the yz-plane, the determined values in the xz-plane are position-dependent, whereas position A exhibits the lowest and C the largest birefringence value. It is worthwhile to note, however, that the surface birefringence determined with the prism coupler is approximately 10% larger than the values determined with the polarimeter.
This is attributed to the fact that the imaging-based evaluation technique suffers from inaccuracies near the edges of samples with finite expansion in the measurement direction. This is especially true at large magnifications since they come with a limited depth of focus. It is consequently assumed that the polarimeter underestimates the surface birefringence of the examined COC specimens.
Overall, the determined results correlate well with the known formation of near-surface birefringent regions when amorphous thermoplastics are injection-molded [
24]. As birefringence directly correlates with the local degree of molecular orientation [
23,
25], the results allow interpretation of the respective preferred polymer chain orientation direction within the injection-molded COC plates. Overall, a high degree of molecular orientation is located in the near-surface regions close to the injection gate, while a linear decrease in molecular orientation with increasing distances from the gate is determined. At the center of the COC plate, the birefringence and thus the molecular orientation is significantly reduced and constant throughout the whole specimen. The fact that the maximum optical birefringence observed in the xz-plane is significantly higher than in the yz-plane leads to the conclusion that the preferred polymer chain orientation direction is parallel to the melt flow direction in these regions.
3.2. Impact on Mechanical Properties
Multiple strain gauges from different regions of the birefringent COC plate were manufactured and their stress–strain behavior was examined. Overall, 18 specimens, i.e., 6 samples per position, were evaluated. The results as well as the respective sample positioning is shown in
Figure 8.
Generally, all examined COC samples exhibited brittle behavior. This means that rupture occurs suddenly, without prior indication of plastic deformation. The average values for the respective tensile modulus, tensile strength as well as elongation at break were determined from the stress–strain analysis data and are summarized in
Table 3.
While the tensile strength is comparable, the determined tensile modulus as well as the elongation at break values differ from previous data and supplier specifications [
34,
35]. This is attributed to the fact that the employed sample geometry is significantly different. However, while the impact of the sample positioning and orientation of the tensile properties is not pronounced, a comparison of the obtained data still indicates a non-negligible impact of the local molecular orientation on the COC plates’ behavior. Especially the sample from position C, which is oriented orthogonally to the preferred polymer chain orientation direction, exhibits the largest tensile modulus and the smallest elongation at break. Furthermore, this specimen group exhibits increased standard deviation values for all parameters. This infers that a high degree of molecular orientation has a negative impact on the mechanical reliability of injection-molded COC. However, the results first and foremost demonstrate that a quantitative analysis and assessment of the birefringence in injection-molded COC plates by means of classical stress–strain analysis is difficult.
The flexural properties of the COC plates as well as the glass transition temperature were determined via DMTA analysis. Therefore, appropriate samples were cut from the same regions of the plate, as outlined in
Figure 9a. The respective results, i.e., flexural modulus as well as normalized tan(δ), in a temperature range from 18 °C to 180 °C, are shown in
Figure 9b.
The overall behavior of the flexural modulus is equal for all three sample locations. At room temperature (20 °C) and up to 30 °C, the flexural modulus is at a maximum, with values of 3461 MPa, 3351 MPa and 3307 MPa for the regions A, B and C, respectively. With increasing temperatures of up to 150 °C, the flexural modus decreases linearly. Afterwards, based on the sudden and drastic reduction of the flexural modulus, the glass transition range is reached. It is worthwhile to note that no rubbery plateau was observed, which means that the injection-molded COC does not exhibit a rubbery state. Instead, all specimens directly translate from the glassy to the viscous state. The glass transition temperature is provided by the location of the tan(δ) maximum, which is found at 163 °C for the samples originating from region A. For specimens B and C, however, the determined glass transition temperature reduces to 159 °C.
While the observed flexural modulus values are in good agreement with the respective literature and the specifications of the wrought material [
7,
36], the glass transition temperature values are noticeably reduced in comparison with earlier studies [
37]. A possible explanation for this is the impact of deviating injection-molding parameters. Furthermore, the results indicate that increased molecular orientation has a negative impact on the glass transition temperature of injection-molded COCs. However, as with stress–strain analysis, the impact of the COC plate’s local birefringence on the mechanical parameters determined via DMTA are not well pronounced. DMTA as well as static stress–strain analysis are thus suboptimal methods to assess the impact of a COC substrate’s birefringence on integrated photonic structures.
3.3. Impact on the Properties of Near-Surface Integrated Waveguides and Bragg Gratings
Multiple near-surface waveguides were fabricated in different regions of the injection-molded COC. Since they exhibited the most different behavior with respect to their inherent birefringence, the upcoming experiments focus on regions A (opposite to the injection gate) and C (close to the injection gate). In both regions, waveguides oriented parallel to the plate’s x- and y-axis were generated, and the polarization-dependent transmitted power, at wavelengths around 1550 nm, was determined as a function of the waveguide length via the cut-back method. The results are summarized in
Figure 10.
It was found that the transmission signals of all waveguides fabricated in region C are significantly larger than those of the waveguides fabricated in region A. Furthermore, the optical attenuation achieved in region C is significantly smaller. Here, the TE mode of the waveguide oriented parallelly to the x-axis exhibits an attenuation of −1.1 dB × cm−1. In region A, however, the minimum attenuation amounts to −6.7 dB × cm−1 for the TE mode of the waveguide oriented parallel to the x-axis. Furthermore, in this region, no optical radiation is discernable if the waveguide is oriented parallel to the y-axis and its TM mode is excited. The results also show that waveguides oriented parallel to the x-axis, and thus parallel to the preferred orientation of the polymer chains, exhibit a reduced attenuation in comparison to waveguides oriented parallel to the y-axis. In terms of optical attenuation, it is thus preferable to fabricate near-surface waveguides in the birefringent zone of the COC plate and parallel to the melt flow direction. On the other hand, waveguides generated in region A, and parallel to the COC plate’s y-axis, have the potential to serve as modal filters, since they do not transmit the TM mode while the TE mode is guided with reasonable attenuation.
The local birefringence of the injection-molded COC plate also influences the BG’s reflection signal, as demonstrated in
Figure 11. Again, waveguides and BGs located in regions A and C, oriented either parallel to the x- or the y-axis, are examined.
The results show that the obtained reflection from BGs located in region C is larger compared to the reflected optical power received by photonic structures fabricated in region A. This observation correlates well with the respective transmission signals and optical attenuation values determined for both regions. Furthermore, the difference in reflection peak amplitude of the polarization modes can also be unambiguously attributed to the determined, polarization-dependent loss properties of the respective waveguide orientations. In contrast to the transmission signal configuration, a residual TM mode reflection is determined from the photonic structures oriented parallelly to the y-axis, located in region A. However, the TM mode’s reflection peak amplitude is reduced by 10 dB in comparison to the TE mode’s reflection signal. It is assumed that the significantly shorter propagation distance in the reflective configuration enables detection of the residual TM mode signal. Qualitatively, the observed Bragg wavelength difference of the polarization modes follows the local near-surface birefringence behavior of the COC plate, i.e., increased birefringence values lead to a more pronounced peak splitting. However, the quantitative values do not correlate with the determined local birefringence. According to Equation (10), the polarization-dependent peak splitting yields values of half the near-surface birefringence in most cases. This behavior, as well as the observed deviations in optical waveguide attenuation (see
Figure 10), are both correlated with the fact that the chemical material modification processes triggered by UV-light irradiation and thermal treatment, which lead to the formation of photonic structures, are based on reorientation and reconfiguration of the macromolecules, i.e., polymer chain scission [
38,
39]. Consequently, the birefringence in the irradiated and thus modified area can differ from the pristine COC plate. It is worthwhile to note that the TE and TM peak Bragg wavelength difference of the BG oriented parallel to the y-axis in region A does not follow this behavior. However, since the TM mode is not well supported in this configuration, the peak splitting is not caused by the near-surface birefringence alone. The weak waveguiding of the TM mode possibly leads to penetration into the pristine and unmodified COC material, which exhibits a significantly reduced refractive index. It is worthwhile to note that, in practice, most Bragg grating devices are evaluated with unpolarized radiation. In this case, the reflection spectrum comprises the superposition of the TE and TM reflection peaks. In cases where the spectral separation of the polarization mode reflection peaks is sufficiently pronounced, this leads to visible peak splitting in the unpolarized reflection signal, an unwanted property in most cases. With this in mind, fabrication of BGs in region A of the COC plate is advantageous, since these structures show sufficient reflection peak amplitude but negligible peak splitting effects.
The polarization-dependent modal intensity distribution was examined by means of the near-field setup. First, the waveguides located in region A of the COC plate are evaluated, as shown in
Figure 12.
Again, like in previous experiments, the TM mode is not sufficiently supported by the waveguide oriented parallel to the y-axis. Thus, there is no evaluable near-field signal. In the horizontal plane, the modal shape is comparable in all cases, independent of the waveguide’s orientation and the polarization mode. In the vertical plane, however, the waveguide orientation and the excited polarization influence the near-field intensity distribution significantly. If the waveguide is oriented parallelly to the y-axis, the TE mode reaches about 20 µm into the substrate (e−2-value). If the waveguide is oriented parallelly to the x-axis, the TE mode penetrates 26 µm deep into the substrate, while the TM mode’s edge is located more than 55 µm underneath the substrate surface.
The near-field evaluation results of similarly oriented waveguides from region C of the COC plate are depicted in
Figure 13. In this region, the waveguide parallel to the y-axis supports both polarization modes. Overall, the radiation is spatially more tightly confined compared to waveguides fabricated in region A of the plate. While, on the horizontal cross-section, the intensity distribution is comparable to the waveguides generated in region A, the radiation propagates closer to the substrate surface on the vertical cross-section. For waveguides oriented parallelly to the y-axis, the radiation penetrates up to 13 µm into the substrate, while it reaches up to 20 µm into the COC in waveguides oriented parallelly to the x-axis. Overall, the results of the near-field evaluation correlate well with the determined transmission signal and waveguide attenuation properties. First, the reduced spatial extension of the waveguide modes leads to an increased modal overlap between the butt-coupled fiber-optic pigtail and the integrated waveguide.
This, in turn, yields significantly reduced coupling losses and thus an increase in transmission signal. This thesis is furthermore supported by the observed Bragg reflection peak amplitudes, which are also larger for BGs located in region C of the COC plate. Second, the increased modal confinement of waveguides fabricated in region C indicates stronger waveguiding, which is in good agreement with the determined optical attenuation values. As all waveguides are fabricated with the same parameters, this underlines that irradiation of the COC plate regions with pronounced birefringence leads to favorable waveguiding conditions.
3.4. Buried Waveguides and Bragg Gratings
Multiple waveguides, oriented parallelly to a COC plate’s y-axis, were fabricated in region C of an injection-molded plate by means of the femtosecond laser-based direct writing approach. The waveguides were generated at different depths underneath the substrate surface, whereas the uppermost photonic structure was located at a depth of 55 µm. The other waveguides were fabricated at increasing depths, with a step width of 15 µm. All waveguides were laterally offset by 100 µm in x-direction to ensure that they did not interfere with each other and the structures were not irradiated repeatedly during fabrication. A schematic of the layout is given in
Figure 14a.
All waveguides were equipped with Bragg gratings.
Figure 14b depicts the reflection signal of two photonic structures fabricated at depths of 55 µm and 145 µm, respectively. The waveguides are excited with unpolarized light. Thus, the reflection spectrum comprises the superimposed signal of the TE and the TM reflection peaks. While the BG close to the surface exhibits significant peak splitting, the BG buried deeper in the substrate comprises a single reflection peak, as it is located in the less birefringent volume of the substrate. The birefringence, calculated according to Equation (10) from the determined peak splitting, is visualized as a function of the BG’s z-axis position in
Figure 14c. The graph furthermore comprises the substrate’s birefringence depth profile acquired via the prism coupler. Therefore, the substrate thickness is reduced layer by layer with the micro mill and then polished back to optical quality in between measurements. The birefringence depth profile agrees well with the birefringence data acquired via polarimetry (see
Figure 7).
Furthermore, the data shows an unambiguous correlation between the peak splitting and the birefringence depth profile. Finally, the experiment demonstrates that unwanted peak splitting effects, exhibited by near-surface BGs fabricated in injection-molded COC plates, can be completely omitted by positioning the photonic structures deeper in the substrate volume.
3.5. Guidelines for the Fabrication of Photonic Structures in Injection-Molded Cyclic Olefin Copolymer Substrates
Based on the observations discussed in this work, it is possible to outline some general guidelines for the fabrication of photonic structures in injection-molded COCs. While it is theoretically possible to influence the degree of molecular orientation in the workpieces by means of adapting the injection-molding parameters, e.g., process and mold temperature, injection and back pressure as well as mold geometry [
40], it is challenging to avoid these effects completely. Nevertheless, injection molding is the preferred choice to manufacture low-cost and large-quantity polymer-based products, a prerequisite for commercial success in attractive markets, such as medical platforms and sensing applications [
21,
22]. Therefore, as the behavior of near-surface integrated waveguides and BGs strongly depends on their location and orientation, the inherent birefringence of the injection-molded plates needs to be adequately quantified.
According to this study, mechanical analysis is insufficient to appropriately detect and analyze the anisotropic properties of the workpieces. Instead, the usage of optical measurement techniques in the form of a prism coupler or a polarimeter is mandatory. Analysis of the birefringence in the plane parallel to the melt flow (xy-plane) is sufficient to derive some basic design decisions. For example, if the intended application requires low-loss waveguides, fabrication of the photonic structures in the more birefringent area (region C, near the injection gate) is recommended. If, on the other hand, peak splitting or double peaks in the reflection spectrum of a BG-based device are to be avoided, the photonic structures need to be located in the less birefringent region of the injection-molded plate (region A). However, in order to accurately predict the behavior of a near-surface integrated photonic structure, three-dimensional analysis of the injection-molded plate’s birefringence is necessary. This is the only way to quantify the inhomogeneous molecular orientation distribution in the COC substrates. Alternatively, femtosecond laser direct writing processes can omit birefringence effects completely, as they allow the fabrication of photonic structures deeper within the volume of the substrate. Consequently, this fabrication method is preferred if the target application does not necessitate near-surface structures, as in the case of devices that interact with sensitization coatings [
12,
13].