1. Introduction

Optical polymer planar Bragg gratings (PPBG) have developed to an increasingly interesting sensor concept throughout recent years. Similar to fiber Bragg gratings, a PPBG comprises a waveguide and a periodic refractive index modulation, also denoted as Bragg grating, which acts as a narrowband reflector for light waves propagating within the waveguide. In comparison to classical silica substrates, polymer-based photonic sensing devices offer several general advantages, such as inherent photosensitivity, a reduced Young’s modulus and an increased thermo-optic coefficient [1–3]. Moreover, polymer planar photonic platforms are easy to handle while they maintain reasonable manufacturing costs and effort [4]. Up to date, a variety of PPBGs based on poly(methyl methacrylate), Ormocer hybrid polymers and epoxy resins were demonstrated, facilitating multiple sensing applications like temperature sensing, multidimensional stress or strain sensing, vital sign monitoring and accelerometry [5–11]. Additionally, in combination with microfluidic channels and / or appropriate functionalization, PPBGs can serve as a basis for medical lab-on-a-chip devices and other biochemical applications [12–14]. In practice, however, PPBGs based on these materials suffer from limited heat deflection properties and / or unwanted cross-sensitivities towards relative humidity changes [15]. This can be overcome by employing cyclic olefin copolymers (COC), a novel high-grade thermoplastic which offers glass transition temperatures up to 250 °C and negligible water absorption [16–18]. Thus, COC-PPBGs are able to outlast harsh environments and perform temperature measurements up to 160 °C [19,20].

Usually, fabrication of COC-based PPBGs is achieved by employing a sophisticated single-writing-step (SWS) method, where a stack comprised of a phase mask and an amplitude mask is brought into close contact with the bulk polymer substrate. Afterwards, the sample is exposed to UV radiation generated by a KrF excimer laser. This leads to the simultaneous generation of a waveguide and an integrated Bragg grating structure, buried underneath the COC substrate’s surface. Using this technique, in dependence of the employed UV dosage, PPBGs with a reflection peak exhibiting a reflectivity of 98% and a 400 pm spectral full width at half maximum (FWHM) were demonstrated. A sophisticated discussion of the SWS method as well as the optical and photosensitive properties of COCs is given by the authors elsewhere [21,22]. However, the flexibility of the SWS procedure is limited by the available amplitude and phase masks, which are responsible for defining the waveguide’s shape as well as its width and the resulting wavelength of main reflection, respectively. Furthermore, in order to enable the usage of COC-PPBGs in the outline of microfluidic applications, additional processing technologies, such as femtosecond laser ablation or micromilling, need to be employed to equip the device with microfluidic channels.

Alternatively, it is possible to make use of femtosecond laser direct writing in order to manufacture PPBGs with buried waveguides in bulk COCs. Waveguides as well as grating structures are generated by employing the point-by-point method, whereas a spatial light modulator is used to form a disk-shaped focal volume pixel (voxel) to modify the substrate’s refractive index in a controlled manner. COC-PPBGs fabricated with this method exhibit a reflectivity up to 95% at a FWHM of 288 pm [23]. This technique offers increased flexibility, since the photonic structures can be generated in various depths within the substrate and the Bragg grating period can be adapted by varying the point-by-point inscription parameters. Moreover, femtosecond-laser based processes enable hybrid micromachining, which features the fabrication of photonic structures and microfluidic channels in one processing step [24]. However, this comes with the increased cost of a sophisticated femtosecond laser system.

In contrast, micromilling, already identified as a cost efficient and versatile tool for rapid prototyping of microfluidic devices [25], can also be used for the generation of photonic structures. Bundgaard et al. reported on micromilling-assisted rapid prototyping of COCs, including the fabrication of microfluidic channels and optical waveguides, as a novel alternative manufacturing method for planar polymer-based photonic devices. Nevertheless, the demonstrated platforms still are heterostructures consisting of multiple COC-based layers. Furthermore, the generated 200 × 200 µm² waveguides exhibit highly multi-modal behavior and are thus suboptimal for Bragg grating sensor concepts [26,27].

Against this background, this article demonstrates and discusses the fabrication of ridge-type waveguides from bulk COC polymers by employing a combination of micromilling and UV-radiation exposure. Lateral cross-sections with dimensions down to 30 × 30 µm² are reached and, by means of using an additional phase mask during irradiation, Bragg grating structures are integrated within the waveguide. Furthermore, a Fabry-Pérot (FP) cavity, also generated during the micromilling process, enables refractive index measurements.

2. Fabrication process and micromilling parameters

All photonic structures generated in the outline of this study are based on a high-temperature stable COC grade (TOPAS6017-S04, TOPAS Advanced Polymers, Raunheim, Germany). Its glass transition temperature (defined as 178 °C by the manufacturer) is significantly higher than that of previously demonstrated COCs used for milling-based fabrication of photonic devices [26,27]. A desktop micro mill (CNC Mini-Mill/4, Minitech Machinery, Norcross, GA, USA) is used to prepare COC samples out of an injection-molded polymer plate. Except for an air stream directed at the tip of the end mill, no additional cooling is employed throughout all experiments discussed in this contribution. The bulk substrate exhibits a length of 20 mm, a width of 10 mm and a thickness of 1.5 mm. An end mill with a diameter of 1 mm (ES-PS-0100-3-040-40, vhf camfacture, Ammerbuch, Germany) is employed to cut the sample to appropriate size while successive generation of the ridge structure is achieved by removing excess material with a 0.2 mm diameter single-tooth end mill (ES-SC-0020-3-006-40, vhf camfacture). At this point, an optional Fabry-Pérot cavity can be integrated into the waveguide structure using the same process parameters and milling tool. Subsequently, the structure is covered with a phase mask and the sample is irradiated with 248 nm UV radiation generated by a KrF excimer laser (BraggStar Industrial, Coherent Europe B.V., Utrecht, Netherlands). Finally, the PPBG is thermally cured at a temperature of 130 °C for 2 hours. The process results in a monolithic ridge-type waveguide with an asymmetric refractive index profile. Tangentially to the substrate surface (x-direction, see Fig. 1(b)), the light is confined by the physical extension of the micromilled ridge structure. Perpendicular to the substrate surface however (z-direction), the waveguide exhibits an exponential refractive index depth profile which prevents the guided mode from leaking into the untreated polymer substrate. Thus, the optical properties of the generated waveguide are defined by its cross-section, controlled via the micromilling step, and the amount of UV radiation absorbed by the bulk polymer [22]. Please note that there is no additional masking of the laser beam during the process. Thus, a significant amount of UV light hits the milled COC substrate surface next to the ridge. However, it is found that this does not diminish the light-guiding properties of the fabricated waveguide, as long as the induced refractive index increase is large enough to confine the guided light within the elevated structure. A comprehensive process overview as well as an image of multiple ridge-type waveguides on a single substrate and a schematic of their asymmetric refractive index profile is shown in Fig. 1.

 

Fig. 1. (a) Fabrication process schematic for micromilling-assisted generation of monolithic ridge-type waveguides on a bulk cyclic olefin copolymer (COC) substrate. (b) Image of a finalized specimen comprising three parallel waveguides with integrated Bragg gratings (grating structures are indicated). The zoom in depicts a sketch of the waveguide and schematics of its asymmetric refractive index (RI) profile.

Download Full Size | PDF

There are two main challenges for the generation of photonic microstructures by employing a cutting technology with successive contact irradiation. First, the milled ridge needs to exhibit a sufficiently reduced surface roughness to prevent excessive scattering losses. Second, burr and debris forming on the waveguide’s upper edge needs to be omitted in order to guarantee reproducible results when the phase mask is brought into direct contact with the ridge structure during UV irradiation [28,29]. Overall, there is an enormous variety of adjustable milling parameters such as milling strategy, feed rate, rotational speed and end mill composition including style, material, diameter as well as number of flutes or teeth. In consequence and due to the general machinability variance of polymers, especially in the outline of micromilling processes, an application-related parameter study is necessary to gain an elaborate insight into the fabrication method and the quality of the resulting photonic structures [30,31]. Thus, multiple COC specimens are machined with different feed rates v_f and rotational speeds N, which results in varying feed per tooth values f_t according to

Here, Z represents the number of end mill flutes, which is 1 for all tools employed in this study, as they exhibited superior performance in preliminary experiments on polymer-based waveguides [32]. Thus, this parameter is equal to the feed per revolution, which is also commonly denoted as chip load. Moreover, for micro-scale end mills, the study compares the impact of up-cut milling, also denoted as conventional milling, and down-cut milling. While in the former case, cutting edge travel and workpiece feed are oriented in opposite directions, down-cut milling describes a process where both directions are equal [33,34]. Figure 2(a) depicts a schematic of the side milling procedure while Fig. 2(b) shows the surface roughness resulting from the process, acquired by means of a white-light interferometer (WLI) (ContourGT-I, Bruker, Billerica, MA, USA). The surface roughness is quantified by evaluating the unfiltered arithmetical mean roughness of the surface S_a [35]. Based on the used 50x objective, lateral and height resolution of the WLI are specified as 0.5 µm and less than 0.1 µm, respectively.

 

Fig. 2. (a) Schematic of the milling process and its parameters. The inset depicts a contour plot of the resulting areal surface profile, acquired via white-light interferometry. (b) Arithmetic mean deviation of the assessed surface profile, or surface roughness, as a function of varying chip loads when using end mills with a diameter of 1 mm and 0.2 mm.

Download Full Size | PDF

According to Fig. 2(b), an unambiguous correlation of chip load and the resulting surface roughness of the waveguide’s side walls is observed. Thus, smoother surfaces can be generated by employing a combination of low feed rates and high milling tool rotational speeds. In this case, the number of interactions between tool cutting edge and workpiece is significantly higher over a finite length, as it is the case with a polishing process. Furthermore, increased rotational speeds lead to elevated temperature generation, which possibly has a positive effect on the residual surface characteristics as well. It is also found that, in comparison to down-cut milling, up-cut milling results in reduced S_a values. This behavior is attributed to the fact that down-cut milling implicates increased engagement forces. Furthermore, the force impact is mainly oriented orthogonally with respect to the machining direction when the cutting edge touches the unmachined section of the workpiece once per revolution. Especially when machining ductile materials such as polymers, this may lead to excessive workpiece deformation. In the case of up-cut milling, this effect is mitigated since this strategy features reduced engagement forces which are also oriented in parallel to the machining direction [36]. While the minimum S_a parameter is limited to 300 nm when using an end mill with a diameter of 0.2 mm, employing a tool with a diameter of 1 mm yields surface roughness values down to 200 nm. Furthermore, larger tools generate a comparable surface roughness at increased feed rates. Thus, faster overall fabrication can be achieved by employing an end mill with a larger diameter. However, this means that tool changes are required for the fabrication of microscale structures, such as a microfluidic channel or an FP cavity. Additionally, it is worthwhile to note that only the final tool passage of a side-milling process defines the residual roughness of the resulting surface. At earlier stages, faster milling parameters can be chosen to reduce fabrication time without sacrificing any surface quality. Cross feed Δx and depth feed Δz are set to 10% of the employed tool’s diameter. No noticeable impact of these parameters on the achievable surface quality is observed in this study. In summary, while there is a deviation in the observed absolute surface roughness values, the general relation of surface roughness and chip load compares well to previous micromilling studies conducted with different milling tools and polymers [37,38].

Burr as well as debris appearance on ridge structures, milled with a comparable parameter set, are examined by means of a light microscope (Eclipse LV-N, Nikon, Tokyo, Japan). Figure 3(a) provides a qualitative overview about the employed feed rate / rotational speed combinations, the residual chip load and the resulting quality of the milled waveguide structures. It is observed that reduced chip loads again lead to preferable results, as the occurrence of burr and debris is drastically reduced. Furthermore, an additional finishing step, wherein the cross-feed parameter is reduced from 10% to 1% during the last passage, can further mitigate burr while the abundance of debris is completely omitted. Figure 3(b) depicts the microscope image of a ridge structure fabricated with optimized parameters, namely a feed rate of 50 mm·min^-1 and a rotational speed of 50,000 rpm, which results in a chip load of 1 µm. According to the cross-sectional waveguide profile, which is derived from a WLI measurement, width and height of the top burr are determined as 8 µm and 5 µm, respectively. Up-cut milling is used as fabrication process strategy in all cases, since down-cut milling results in excessive top burr formation, as shown in Fig. 3(c). In the case of ridge structures with widths below 100 µm, the whole ridge surface is completely covered with burr. Thus, they are not suitable for the subsequent illumination step. Again, this behavior is attributed to the chip cutting process which is different for both milling strategies. In the case of up-cut milling, burr is generated in travel direction of the tool. Consequently, every subsequent tool revolution removes burr generated by the preceding revolution. In contrast, down-cut milling leads to burr formation at the side of the tool path and thus no automatic burr removal.

 

Fig. 3. (a) Qualitative analysis of residual top burr and debris in dependence of feed rate and rotational speed of the up-cut milling process. Moreover, the resulting chip load is stated. (b) Microscope image of a waveguide milled with optimized up-cut milling parameters. Furthermore, a cross-sectional height profile of the fabricated ridge structure is depicted. (c) Exemplary comparison of burr formation resulting from up-cut or down-cut milling processes.

Download Full Size | PDF

In order to prevent manual tool changes, a single-tooth end mill with a diameter of 0.2 mm is used for the fabrication of all photonic structures demonstrated in this study. The parameter set of the final tool passage always comprises a rotational speed of 50,000 rpm in combination with a feed rate of 50 mm·min^-1, which yields a chip load of 1 µm. While a cross feed of 10% tool diameter is employed for most of the process, a final finishing step is realized by reducing this parameter to 1% during the final tool passage. This way, the formation of top burr and debris, as well as the resulting surface roughness parameter of the ridge’s side walls, is minimized.

3. Bragg grating temperature sensor

Multiple ridge structures are milled and subsequently irradiated to generate monolithic ridge-type waveguides (see Section 2). During UV exposure, a +1/-1 order phase mask (Ibsen Photonics, Farum, Denmark), with a grating period of 1032 nm over a length of 10 mm, is aligned and brought into direct contact with the 20 mm long ridges. After thermal curing at a temperature of 130 °C for two hours, the photonic structures are aligned with a pigtailed single-mode fiber to enable quantification of their respective Bragg reflection signals by means of an interrogation unit (HYPERION si155, Micron Optics, Atlanta, GA, USA). A permanent connection between fiber and waveguide is realized by employing a UV-curable adhesive (NOA78, Norland, Cranbury, NJ, USA), which additionally acts as a refractive index matching medium in the fiber-to-waveguide intersection.

It is found that smaller waveguide cross-sections require increased UV dosages during irradiation of the ridge structure (see Fig. 1(a)), in order to obtain comparable and evaluable reflection signals from the integrated Bragg gratings. Depending on the waveguide cross-section, which is quadratic with an edge length of 30, 40 or 50 µm, the specimen is exposed to a fluence of 400, 350 or 300 J cm^-2, respectively. At a repetition rate of 200 Hz, the overall exposure is defined by varying the pulse quantity from 30,000 to 40,000 pulses. Thus, the overall irradiation time also varies from 150 s to 200 s. Figure 4(a) to (c) depict an overview on the obtained Bragg peaks as well as the respective waveguide’s mode profile, acquired via beam propagation method simulations (RSoft, Synopsys, Mountain View, CA, USA) [39]. The simulation is based on refractive index depth profiles acquired via phase shifting Mach-Zehnder interferometry [22]. While the demonstrated structures are exposed to various quantities of UV radiation, which implies a different amount of refractive index modification, no straightforward relation between employed fluence F and residual Bragg wavelength λ_B is present. However, the obtained Bragg signals show that the residual top burr of the ridge structure does not prevent successful contact irradiation through a phase mask. Furthermore, the wavelength of main reflection correlates well to the simulated effective refractive index n_eff values of the respective waveguide, as dictated by the Bragg relation

 

Fig. 4. Bragg peaks and simulated mode profiles of micromilled ridge-type waveguides with cross-sections of (a) 30 × 30 µm², (b) 40 × 40 µm² and (c) 50 × 50 µm². Additionally, the respective fluence F, absorbed by the ridge structure during UV exposure, is given. (d) Simulated mode intensity I profile cross-sections for various waveguide geometries irradiated with a fluence of 300 J cm⁻². The inset shows a magnification of the mode cross section near the ridge-structure to polymer-substrate transition region.

Download Full Size | PDF

The grating period Λ therein is defined by the employed phase mask, which means this value is constant for all fabricated structures [40].

The inverse proportionality of waveguide cross-section and UV radiation necessary to enable suitable reflection peaks is examined by means of Fig. 4(d). Therein, the simulated vertical intensity distribution of several waveguide geometries, fabricated with a fluence of 300 J cm^-2, is shown. It reveals that the mode is propagating closer to the waveguide surface in larger structures. Additionally, smaller ridge cross-sections lead to mode penetration into the bulk, unmachined COC material, where there is no more physical light confinement in the horizontal plane (x-direction). This behavior is attributed to the fact that the fabricated ridge-type waveguides exhibit a large refractive contrast at the upper as well as the lateral COC-to-air interfaces. However, towards the unirradiated COC substrate, the guided light experiences relatively weak modal confinement by the waveguide’s exponential refractive index depth profile. Consequently, when the physical extension of a ridge-type waveguide is reduced, the guided mode will begin to leak through the interface with the smallest refractive index contrast, hence, towards the lower regions of the polymer. This needs to be countered with larger UV radiation doses during fabrication, which increases the waveguide’s residual refractive index modification. Additionally, a steeper refractive index depth profile is generated which, in consequence, leads to increased modal confinement [22]. Therefore, overall irradiation in combination with waveguide cross-section needs to be considered when designing a monolithic ridge-type waveguide since they both influence the resulting effective refractive index of the propagating mode as well as the general functionality of the device. However, the results demonstrate that a variety of monolithic ridge-type waveguides, with cross-sections down to 30 x30 µm², can be manufactured with the proposed micromilling-assisted fabrication method.

Since both, effective refractive index n_eff and grating period Λ of a Bragg grating are functions of environmental physical quantities, such as strain and temperature, the fabricated specimens can be used as a sensing device. Thus, an exemplary study is conducted by positioning a sample with a 40 × 40 µm² waveguide on a hotplate with active temperature control. During the experiment, the temperature is incrementally increased and subsequently held for at least 10 minutes in order to guarantee stable conditions. Figure 5 depicts the Bragg wavelength shift induced by environmental temperatures up to 70 °C.

 

Fig. 5. Temperature response of a COC-PPBG with monolithic ridge-type waveguide.

Download Full Size | PDF

It is found that increasing the temperature leads to a negative Bragg wavelength shift correlated to the negative thermo-optic coefficient of COCs, or polymers in general [41]. Linear regression of the measurement data yields a sensitivity of -5.1 pm K^-1, a value which is characteristic for PPBGs based on bulk COCs. Additionally, the observed quality of the Bragg reflection peak is comparable to the signal of COC-PPBGs with buried photonic structures [18,20,42,43]. Hence, the residual top burr has no or negligible influence on the irradiation process and the resulting photonic platform. Conclusively, the device is equally suitable for temperature sensing or, alternatively, for temperature referencing in the outline of other applications.

4. Fabry-Pérot refractive index sensor

Fabry-Pérot interferometers, often denoted as etalons, principally consist of two plane parallel mirrors separated by a cavity. If broadband light traverses the etalon structure, multiple internal reflections lead to characteristic interference fringes in the transmission as well as in the reflection spectrum. Since spectral location and distance of these interference fringes are strongly influenced by perturbations of the cavity’s optical path length, FP-based devices have been successfully developed for various purposes, such as temperature, pressure and refractive index sensing. Within an optical waveguide, etalons can be straightforwardly generated by interrupting the photonic structure in a controlled manner. In this case, the waveguide-to-air interfaces will act as the mirrors of the FP etalon [44,45].

A COC-PPBG with monolithic ridge-type waveguide is fabricated by micromilling a ridge structure with a cross-section of 40 × 40 µm². During the micromilling process, prior to UV irradiation, the waveguide structure is additionally equipped with a FP cavity (see Fig. 1(a)). After connecting the photonic structure to a pigtailed single-mode fiber, the cavity’s interference-based reflection signal is interrogated (SYS:Lab 2, Stratophase, Hampshire, UK), as shown in Fig. 6(a). In order to optimize signal evaluation, the raw signal is filtered with a digital Butterworth filter via Matlab. Both, raw and filtered reflection signal are depicted in Fig. 6(b). Therein, a microscopic image of the FP cavity is shown as well. According to

 

Fig. 6. (a) Schematic of the employed experimental setup. (b) Reflection spectrum of the milled Fabry-Pérot cavity. The raw signal is filtered with a digital Butterworth filter to optimize tracking of the reflection minima. A microscopic image of the cavity is shown in the inset. (c) Wavelength response of the FP reflection minima when filling the cavity with various refractive index liquids.

Download Full Size | PDF

As a result, the reflection signal of the milled device can be utilized to determine the refractive index of a medium within its cavity [46].

Multiple refractive index liquids are prepared by mixing distilled water with various weight percentages of sugar. Each solution’s refractive index is determined by employing an Abbe refractometer (DR-M2/1550, Atago, Tokyo, Japan) which is specified to offer a maximum refractive index deviation of ±0.001 at a wavelength of 589 nm. Before every measurement, the solution is stirred to ensure optimum mixing conditions. Subsequently, the sensor is dipped into the liquid, so the cavity is fully immersed. This leads to an immediate shift of the reflection minima, whereas the observed response time cannot be resolved by the 2 Hz sampling rate of the employed interrogation unit. In-between measurements, the FP cavity is purged with nitrogen gas to remove possible residues. Figure 6(c) shows the resulting wavelength shift of the first seven reflection minima of the spectrum’s short wavelength side, when the cavity is flooded with different refractive index liquids. All tracked minima exhibit a positive wavelength shift when the cavity is filled with increasing refractive indices. Linear regression yields a sensitivity between 1141 nm RIU^-1, for the shortest wavelength minimum (M1), and 1154 nm RIU^-1 for the longest wavelength minimum (M7). The determined shifts of all minima, as well as the wavelength-related sensitivity variation, is in good agreement with the theoretical values obtained by Eq. (4). Consequently, the demonstrated device can be employed in the outline of high-precision refractive index measurement applications.

5. Conclusion

In conclusion, this article demonstrates a novel waveguide generated from a bulk injection-molded cyclic olefin copolymer platform. A combination of micromilling with subsequent UV irradiation enables the fabrication of monolithic ridge-type waveguides. While the milling procedure defines the horizontal extension of the waveguide, light is prevented from leaking into the substrate by means of a graded refractive index depth profile. Thus, in contrast to all other state-of-the-art ridge waveguides, the whole platform consists of only one single bulk material. Additionally, photonic sensing structures, such as Bragg gratings or a Fabry-Pérot cavities, can be integrated.

The study reveals that quadratic ridge structures with edge lengths down to 30 µm can be manufactured via micromilling. By optimizing the parameters of the final tool passages, a mean surface roughness of 300 nm is achieved whereas residual top burr can be reduced to a height of 5 µm and a width of 8 µm. It is additionally demonstrated that it is possible to equip the ridge structure with a Bragg grating by irradiating the waveguide through a phase mask. It is found that the amount of absorbed radiation necessary to achieve comparable Bragg reflection signals reduces with increasing waveguide cross-sections. By means of numerical simulations, this behavior is attributed to modal leakage, which predominantly occurs in smaller waveguides and thus needs to be countered with increased UV radiation. Temperature measurements based on evaluating the Bragg wavelength shift of the integrated structure yield a temperature sensitivity of −5.1 pm K^-1, which correlates well to preliminary data based on COC-PPBGs with buried sensing structures. Finally, the study presents that it is possible to employ micromilling in order to equip the ridge with a Fabry-Pérot cavity. By tracking the wavelength shift of the reflection minima, this enables refractive index measurements with sensitivities up to 1154 nm RIU^-1.

Therefore, the proposed fabrication method constitutes a promising approach for novel and robust polymer-based photonic platforms. In comparison to state-of-the-art heterostructure ridge waveguides, the demonstrated devices consist of only one single material. Furthermore, in comparison to conventional lithography-based fabrication techniques, employing a micro mill to define the cross-section of the resulting waveguide omits the necessity of inflexible amplitude masks. Thus, the proposed method enables rapid prototyping of novel waveguide designs since dimension and shape can be straightforwardly adapted and directly implemented onto the photonic platform. Furthermore, the fabricated ridge structures can be equipped with a milled Fabry-Pérot cavity for refractive index sensing purposes and a Bragg grating structure for temperature referencing. Consequently, and due to the fact that micro milling is already an established process in the field of microfluidics, this contribution paves the way towards rapid prototyping of polymer-based lab-on-a-chip devices for biomedical sensing applications.