1. Introduction

Polymer-based integrated photonics offers a number of advantages over silicon or III–V materials in a number of applications. The electrical, thermal and mechanical properties of polymers can be tailored to great extents to meet particular requirements in a case-by-case basis, as the past decades of material science evolution has demonstrated. For integrated optics, polymers have been extensively employed, for instance, to improve electro-optical [1,2] and optical nonlinear effects [3,4]. In general, their notable disadvantage is a lower tolerance to the high temperatures that are often employed in conventional micro- and nanofabrication processes. On the other hand, if high temperatures can be avoided, polymers usually represent a lower-cost fabrication platform, which is always desirable, but specifically so for disposable devices, as is the case for optical biosensors, as an example.

By replacing the standard silicon or III–V technology with polymers, additional advantages can be conferred to biosensors, such as fabrication flexibility, reliability, low power consumption and potential for mass production accompanied by a reduction of production costs [5–9].

Polymeric photonic biosensors are at the core of the concept of photonic lab-on-a-chip [10], aimed towards solving existing limitations in achieving economically viable devices for point-of-care diagnosis [11,12].

There are a number of choices for optically transparent polymers for fabrication of photonic structures. One of the most common and well-developed of those polymers is SU-8, an epoxy usually employed due to its mechanical stability, high aspect ratio, low cost, and well established fabrication process. Micro-electro-mechanical devices have been developed in SU-8 for decades [13–15], and more recently, optical devices followed [16–18]. I-line (365 nm) is the recommended exposure wavelength, although great results have also been demonstrated with electron beam and X-rays.

Recently, studies have shown that by mixing the photoinitiator H-nu 470 to the conventional SU-8 epoxy, it is possible to improve the chemical properties and thermal stability of fabricated structures with this photoresist, increasing its degradation temperature up to 400 °C [19]. This addition aims to change the maximum absorption peak of the photoresist from 365 nm to 470 nm wavelength, allowing a material conventionally used in near UV photolithography to be polymerized with a laser in the visible spectrum. The advantage of allowing exposure at longer wavelengths is that less expensive hardware can be used to allow the full benefits from technologies such as the photonic lab-on-a-chip.

The photoinitiator presence does not change the optical properties after the bake process, thus representing an excellent alternative for the fabrication of integrated devices for applications in the visible, near infrared and infrared domains, i.e., not only in biosensor applications, but also in telecommunications [5,20–23].

In this paper we present the fabrication and characterization of low-loss optical waveguides using the modified SU-8 with H-line lithography (405 nm wavelength). We show that the use of the photoinitiator allows direct laser writing (DLW) of sub-micron structures with low sidewall roughness, low fabrication tolerance, and transparency comparable to those of the original SU-8. Characterization of thin film was performed by X-ray photoelectron spectroscopy (XPS) and atomic force microscopy (AFM), and single- and multimode waveguides propagation losses were measured at 633 nm wavelength.

2. Viscosity and surface roughness characterization

Optical waveguides in the visible range require sub-micron dimensions for single mode operation, thus we first reduced the viscosity of the SU-8 to obtain a 600 nm layer upon spin coating a SiO2 substrate [20,24]. We diluted SU-8 2100 (75% solid) in cyclopentanone to obtain solutions with 14% and 20% solid concentrations. The mixtures were stirred for 24 hours to ensure uniform coverage of the substrate later on. Spin calibration curves for both dilutions are presented in Fig. 1(a), which shows that we are able to reach the desired thickness range for optical waveguide fabrication, in our case around 600 nm.

 

Fig. 1 Bulk modified SU-8 characterization. (a) Spin coating calibration curves for 2 different solid concentrations after dilution of the epoxy in cyclopentanone. (b) Surface roughness of the original SU-8 layer (14% solid) obtained by AFM. (c) Surface roughness of the SU-8 layer with the addition of the photoinitiator.

Download Full Size | PDF

The 14% solid SU-8 received the photoinitiator H-nu 470, the cationic cure accelerator AN-910E, and the OPPI photoacid generator to form concentrations of 0.1%, 0.1%, and 2.5% in weight respectively, and was then stirred for 48 hours. The exact amounts used of each component in the mixture are detailed in Table 1. After spin coating, the original (diluted) and modified SU-8 layers showed great uniformity under visual inspection. Surface roughness characterization by AFM showed RMS values of 0.44 nm ± 0.05 nm for the original resist and 0.20 nm ± 0.05 nm for the modified version. The scans were performed over a 4 μm² area, as presented in Fig. 1(b) and 1(c). The chemical contents of each sample were measured by XPS and are displayed in Fig. 2(a) and 2(b). The spectroscopy results indicate only a small concentration of antimony (2.45%) in the modified mixture that is not present in the original SU-8, which is known to increase the mechanical hardness and improve the thermal properties of the material [25].

 

Fig. 2 SU-8 surface composition. (a) Original SU-8 layer obtained by XPS. (b) SU-8 layer with the addition of the photoinitiator.

Download Full Size | PDF

Table 1. Components of the modified SU-8 mixture for H-line lithography.

View Table

The other 3 prominent energy peaks are similar in both curves, with a slight increase in oxygen and fluorine counts (accompanied by a decrease in carbon counts) in the modified mixture when compared to the original one.

3. H-line Laser lithography

The SU-8 structures were exposed using DLW, which allows rapid prototyping and development cycle of nanometer-scale structures [26, 27] without compromising the possibilities of mass-production at lower costs and scalability to larger substrates.

The preparation of the samples, illustrated in Fig. 3, followed the usual SU-8 processing: after cleaning in piranha bath, they were dehydrated on a 200 °C hot plate for 20 minutes, spin coated with the modified photoresist, soft-baked at 65 °C for 1 minute, and then ramped up to 95 °C for a 3-minute pre-bake. We used a Heidelberg DWL 66FS system with a 2.0 mm head lens and 720 mJ/cm² maximal dose on substrate at 405 nm for exposure. After exposition process the photoresist was subjected to a post-exposure bake starting at 65 °C for 1 minute, ramped up to 95 °C and held for another minute. The fabricated structures were then developed in SU-8 Developer by submersion and stirring for 60 s and, finally, hard-baked for 5 minutes on a 150 °C hot plate.

 

Fig. 3 Scheme of the fabrication process. Fabrication steps for the polymerization of the modified SU-8 with 405 nm DLW.

Download Full Size | PDF

A dose test was performed with 3 different filters (5%, 15%, and 30%) and linewidths (600 nm, 10 μm, and 100 μm) to characterize the polymerization of the modified SU-8. The results presented in Figs. 4(a)–4(c) show that, as expected, the required dose for polymerization of the full thickness of the photoresist layer is quite high. Nevertheless, with both the 5% and 15% transmission filter, complete exposure is attainable for all linewidths with laser energy above 70%. By inspecting Figs. 4(a)–4(c), we can also see that the exposure is not impacted by the linewidth of the structure, specially with the 30% filter, which indicates that the process is robust enough to allow the fabrication of structures with dimensions of different orders of magnitude in a single step. This benefit can be seen in the scanning electron microscope images of fully exposed structures in Fig. 5. These structures were obtained with the 30% filter and the geometrical tolerance measured for the 600 nm linewidth was 8%.

 

Fig. 4 Characterization of polymerization dose. (a) DLW dose calibration for 600 nm, 10 μm and 100 μm linewidth structures with a 5% transmission filter. (b) Same dose calibration as (a) but with a 15% transmission filter, i.e., overall 10% higher dose. (c) Same as (a) but with a 30% transmission filter (25% higher dose).

Download Full Size | PDF

 

Fig. 5 Scanning electron microscope images of the fabricated structures at different scales..

Download Full Size | PDF

The sidewalls of the structures were also characterized using AFM scanning, which resulted in an RMS roughness of 5.60 nm ± 0.02 nm. Despite the addition of the photoinitiator, the sidewall quality of the device is not be affected, which represents an improvement compared to reported results [28], because the implementation of fabrication with direct writing reduces the roughness on the device sidewalls when compared to conventional photolithography.

One last interesting feature of this process is the possibility of controlling the thickness on the final structure by modulating the laser energy almost linearly, as Fig. 4(a) show for the 5% filter. The use of varying core thickness for optical confinement has been show to provide an excellent platform for the fabrication of gray scale photonic devices [29–32], i.e., devices that present multiple levels of core thickness—or even a continuum of thickness variation across the device—fabricated in a single exposure step. This technique is usually employed in the fabrication of large diffraction lenses, but with the length scales achieved in our experiments, it could be used in the future to enable multimode polymeric devices for optical interconnects and biosensors [33,34].

4. Optical characterization

We characterized the absorbance and transmittance of the SU-8 samples with and without the addition of the photoinitiator over a wide range of wavelengths.

The results are displayed in Fig. 6(a) and 6(b). The addition of the photoinitiator results in a small increase in absorbance (loss in transmission), still providing a more than adequate platform for optical applications beyond 400 nm. Below that wavelength, the modified photoresist presents an accentuated absorbance due to the influence of H-nu 470, indicating that the material still absorbs at lower wavelengths even after hard-baking. The refractive indices measured at 633 nm were 1.59 for both samples, which allows the fabrication of well-confined waveguides on SiO2.

 

Fig. 6 Optical characterization of the SU-8 with and without addition of the photoinitiator. (a) Absorbance and (b) transmittance curves show that the addition does not significantly impact the optical properties of the material for wavelengths beyond 400 nm.

Download Full Size | PDF

Waveguides were then fabricated over a 2.0 μm thick SiO2 layer using DLW and left air-clad. Single-mode square waveguides with 600 nm sides were patterned with lengths of 5 mm, 10 mm, and 20 mm for transmission loss measurements Fig. 7(a) shows an scanning electron microscope image of a few devices and Fig. 7(b) presents their optical characterization setup. We used two objective with numerical apertures 0.65 and 0.40 for coupling in and out of the waveguides, respectively. The source was a 633 nm wavelength laser with electric field polarized parallel to the substrate (preferential direction for the quasi-TE mode in conventional rectangular waveguides). The collected data was linearly fitted and resulted in propagation losses of only 4.38 ± 0.55 dB/cm, as shown in Fig. 7(c), an excellent value for applications where fabrications costs are more critical than ultra-low loss devices.

 

Fig. 7 Characterization of optical waveguides fabricated on SU-8 with the H-nu 470 photoinitiator. (a) Scanning electron microscope images of a few air-clad waveguides on a SiO2 layer. (b) Insertion loss measurement setup with objectives for coupling light in and out of the waveguides at 633 nm wavelength. (c) Propagation loss calculation based on a linear fit of the measured loss for waveguides of different lengths.

Download Full Size | PDF

Similarly, multimode waveguides with core width of 1.70 μm were also fabricated on the same 600 nm thick SU-8 layer and with the same length variations. Their optical measurements are also presented in Fig. 7(c), where the linear fit to the data resulted in a propagation loss of 6.40 ± 0.3 dB/cm. Single- and multi-mode waveguide insertion losses were measured, achieving lower values than previously reported [35], demonstrating an excellent performance by fabricated devices with this novel technique.

5. Conclusions

This work describes the preparation and characterization of SU-8 as an optical material suitable for H-line photolithography. We show that the addition of the H-nu 470 photoinitiator that allows absorption at 405 nm does not significantly impact other characteristics of the photoresist. The fabrication process is the same as for the original SU-8 processing and the resulting layer has excellent surface roughness and similar optical absorbance at longer wavelengths.

Patterning via DLW allowed us to reproducibly obtain sub-micron structures including single-mode waveguides for the 633 nm visible wavelength. Propagation losses on those were measured at 4.38 dB/cm, which enables the use of this material in optical applications where low-cost is critical, such as disposable lab-on-a-chip for point-of-care diagnosis.