1. Introduction

Optical integrated circuits, also known as planar lightwave circuits (PLCs) consist of optical waveguides which may generate, transport, filter, modulate, and detect light, among other functions. In order to support continued increase in data-rates that would allow video-on-demand and other high volume applications, costs need to be further reduced. In addition, new applications for optical communications at the system and board-level also require significant reduction in costs. Unlike electronic circuit testing, where current and voltages are applied to the circuit by simple electrical contacts, PLCs require coupling light from/to waveguides and fibers, a complex process that requires the matching of modes and indices of refraction to be efficient. The ability to carry out wafer-level testing of PLCs would greatly aid in the reduction of testing costs by eliminating the requirements in dicing and polishing. Several methods to couple light to waveguides on chip have been proposed, approaches such as using prism coupling have been used to study slab waveguides, and can be used to study channel waveguides with angle adjusted coupling versus wavelength [1] but they are time consuming and expensive. Johannessen proposed an optical probe for wafer testing based on a rigid prism and possibly an embedded waveguide, which contacts the surface of the wafer. However, this may require removal of cladding to couple to the desired waveguide [2]. Fiber tapers have been used to test optical cavities [3]. We originally proposed an approach that overcomes these difficulties with the use of a single-mode flexible probe that enables the coupling length to be tuned for different top cladding thicknesses [4].

In this paper we demonstrate optical coupling into waveguides at the wafer-level with a multi-mode flexible probe based on directional coupling. We describe the theoretical performance of single-mode and multimode probes, the challenges with single-mode and multimode probe manufacturing. Finally, the experimental observation of flexible probe directional coupling into waveguides is demonstrated.

2. Device design

The design of the optical probe was based on the directional coupling concept. The optical probe consists of a flexible waveguide which enables variable contact length to the waveguide under test, hence allowing one to optimize the coupling length in the directional coupler. A schematic of the proposed system is shown in Fig 1. The flexible waveguide probe comes into contact with the waveguide at the surface of a wafer. The flexible nature of the probe enables it to deform as downward pressure is applied. This deformation increases the contact length between the waveguide probe and the waveguide under test. The region where contact is obtained will be referred to as the overlap length between the two waveguides. When two waveguides are brought into close proximity, directional coupling can be achieved, and light couples from one waveguide to the next. The ability to control the overlap length by applying an external pressure to the probe is a key feature of the proposed device. Multiple flexible waveguides could be used to couple to multi-port devices.

 

Fig. 1. Schematic of device in operation: A single-mode flexible optical waveguide probe on a support layer coupling light directionally into a wafer-level optical waveguide on-chip. Inset shows the simulation geometry considered in this work.

Download Full Size | PDF

Previously, we proposed a single mode waveguide probe designed to couple light into a waveguide of same cross-section and materials [4]. This results in matched propagation constants, which yield the best directional coupling results. The coupling coefficient κ will decrease exponentially as the gap s between waveguide and probe increases. This gap could be due to roughness, residual cladding, or particulates. In our simulations, and experiments, the gap medium is comprised of air. Use of another material will lead to a different coupling coefficient κ. Coupling of 100% of the power can be achieved for matched structures if the overlap length equals the coupling length or the critical length Lc=π/2κ. For any two waveguides, some coupling can always be achieved, but the maximum value increases as modes, propagation constants and indices of refraction are matched. Since typically one can not change the waveguide index of refraction, the waveguide probe properties (width, index, thickness, cladding, etc…) need to be chosen carefully to closely match the device under test [5, 6]. In this work we focus on the parallel coupling section of the waveguide to verify the feasibility of the concept, as shown in the inset of Fig. 1. Simulations of the structures were carried out using software from a commercial vendor [7]. The single-mode probe was simulated using the beam-propagation method (BPM) which was validated using the 2D FDTD method. This was important to show the effect of reflections at the boundaries of the structures used and how they limit the coupling efficiency for the geometry simulated here. The core of the waveguide had a thickness of 1.2 µm an index of refraction 1.6 and the cladding had an index of 1.46, since our ultimate target application is a 1.2 µm polymethylmethacrilate waveguide [8, 9]. The multimode probe was studied using 3D BPM simulations, as the results are not trivial. The core of the waveguide had a thickness of 15 µm and a width of 30 µm and an index of refraction 1.6, while the cladding had an index of 1.46. The launch field wavelength was 1550 nm for all simulations.

 

Fig. 2. Graph showing the simulation of coupling power versus length of probe contact for SU8 probe into an identical core waveguide. (a) single mode; and (b) multimode probes. Different gaps between probe and waveguide in the range 0–400 nm are shown. Note the longer length scales in the multimode case (b) when compared with single mode case (a).

Download Full Size | PDF

Coupled power was monitored as a function of the overlap length between probe and waveguide under test, with probe and waveguides with identical core for single-mode as well as multimode cases. Figure 2(a) shows the coupled power versus overlap length for gap values of 0, 100, 200, 300 and 400 nm for the single-mode case. The graph shows clearly the oscillatory behavior of the coupled power. The coupling cycle for the 0 nm gap is much shorter than the one for the 100 nm and it increases super linearly. The longer coupling lengths offer more precise tuning by virtue of the broader peaks, which can be adjusted by residual top-cladding thickness. Interestingly, the maximum output power is larger for increasing gaps in the range investigated, as mentioned earlier; this is due to the reduction in the reflection at the waveguide transition. Figure 2(b) shows a significant coupling for the 0 gap condition, but significantly lower values for the other gap values. The power coupled to a given mode of the chip-waveguide is found by tracking the overlap integral of the coupled field with the mode of interest. This was investigated and we verified that nearly 100% of the coupled power for zero gap is in the fundamental mode. Also, the coupling lengths in the multimode case are also significantly large. Finally, the simulation results for the coupling are based on a fundamental mode being launched inside the probe.

3. Fabrication of flexible probe

The choice of materials for the flexible probe construction was based on optical and mechanical properties, and ease of fabrication. SU8 is a negative photoresist that has been proven to act as a functional material in suspended MEMS devices [10, 11]; it can also serve as flexible optical waveguide core with a 1.6 index of refraction [12]. Originally PDMS was considered as the flexible substrate, with a 1.46 index of refraction, however, bending resulted in strain to the top layer film of SU8 which exceeded the yield strain of SU8 of 0.015 [13], leading to failure and formation of cracks. However, it has been shown that a more rigid, and significantly thinner, flexible substrate of NOA61 polymer or Topas® can replace the PDMS substrate leading to minimal strain [14, 15].

SU8 layers thicker than 10 µm were shown to be stiff enough to allow the probe structure to be self supported. Figure 3(b) shows a SEM micrograph of the probe at this stage, prior to release. In order to release the probe and complete the fabrication process, the SU8 patterned substrate was immersed in buffer oxide etch, allowing the SU8 layer to detach from the substrate. S-shaped test waveguides were fabricated as described above and set to have 15 µm thickness with a 50 µm width. The curves had radii of 500 µm leading to a total nominal waveguide length of 10 mm. A Heidelberg uPG101 laser writer was used to produce the masks. Figure 3(c) shows a scanning electron micrograph of the probe bent over a 2 mm capillary to demonstrate probe flexibility. Figure 3(d) shows a close-up SEM of the probe sidewall and the roughness resulting from the process, which we attribute primarily to mask roughness. The losses of the probe and the waveguides are normalized in our experiments, so they are of no consequence here.

 

Fig. 3. (a) Cracked single-mode SU8 Probe with 1.2 um thickness over PDMS flexible substrate; (b) multimode self-supported SU8 probe prior to release; (c) released SU8 probe bent over a 2 mm capillary; (d) Flexible probe sidewall roughness estimate of 200nm.

Download Full Size | PDF

We will focus on the multimode flexible probe in this section since the single-mode probe fabrication method is not optimized, and a functional probe was not obtained. The assembled probe was mounted on a micromanipulator for positioning over the target chip. Probe assembly was described elsewhere, and relies on a support structure and guiding elements [4], the assembly of the probe is still not optimal, and as a result, the power exiting the probe saw a loss of 10 dB, which we believe can be substantially improved to less than 1 dB. The mechanical behavior of the probe was verified to lead to variable contact of the probe to the surface of the chips.

4. Optical characterization

For the optical characterization, independent xyz micrometer stages were used to mount the flexible probe and the cleaved chip containing the waveguide under test. The test waveguides had the exposed output facet at the edge of the chip to facilitate imaging. Imaging was achieved by two independent objectives mounted on micrometer stages atop and in front of the device. One objective imaged the light emitted from the waveguide facet, while the second monitored the waveguide from the top of the stage. A visible video camera as well as an infrared camera (SU320 from Sensors Unlimited) were mounted next to the stage and were used for imaging the alignment and the output of the device. Initial results were obtained with illumination by 532 nm laser light from a Nd:YAG laser coupled to the optical fiber, using visible light had the advantage of facilitating the alignment. In addition, an AQ4321D tunable laser source with fiber optic cable output was used to investigate the performance of the device in the 1550 nm wavelength range. Initial testing of the flexible probe was carried out using a simple approach that achieved wafer-level butt coupling, which was observed in the visible wavelength regime. We then pursued coupling light from the probe to the waveguide using the directional coupling method.

Figure 4(a) shows an optical micrograph of the flexible probe during the alignment process. This procedure achieved overlap of a section approximately 2 mm long over the S-shaped waveguide on the substrate. Note that this is a bird’s eye view hence the short depthof- focus. Figure 4(b) shows the result of green light directionally coupling through the aligned flexible probe into the S-shaped SU8 waveguide. Coupling is demonstrated by excess scattered light at the curves and at the tip of the S-shaped waveguide. Figure 4(c) shows the capillary guide that enables the manipulation of the fiber with attached flexible probe.

 

Fig. 4. (a) Optical micrograph showing the SU8 flexible probe in contact with waveguide on the surface of the chip. (b) 532 nm laser light (Green) coupling from flexible waveguide probe to S-shaped waveguide of identical cross section. (c) Broad area view of setup. Note reflection of capillary guide tube and probe off of the chip surface.

Download Full Size | PDF

The successful coupling of visible light into wafer-level waveguides was followed by an identical process with infrared wavelengths in the telecommunication range of interest around 1550 nm. In order to verify the coupling into the test SU8 waveguide structures, direct IR imaging of the output facet of followed by optical power measurements were conducted. The laser was set to 1550 nm wavelength with 5 mW butt coupled to the waveguide input facet. Three methods of coupling light into the waveguide to enable a comparison were performed. First, the ideal case of direct coupling from a fiber optic cable into a waveguide under test was conducted. Second, facet to facet (butt coupling), where the probe facet was aligned with the facet of the waveguide under test on an uncleaved chip. Third, directional coupling where the probe was aligned on top of the waveguide under test was carried out. Figure 5 shows the infrared output power from the waveguide facets for the three coupling approaches described above: (a) direct fiber butt-coupling; (b) wafer-level butt coupling of probe-facet to waveguide-facet; and (c) flexible probe directional coupling. The results show that the power is constrained in modes inside the core of the SU8 waveguides. Different modes are excited and imaged at the output as would be expected from such a large structure.

For direct fiber coupling using 5 mW power, a measured power of 21.5 µW was observed at the output facet. The theoretical coupling loss from fiber to waveguide is 3.4 dB, resulting in 19.6 dB loss due to propagation through the waveguide under test. For the second part of the test the flexible probe attached to fiber was used, which had an output power of 0.5 mW, a 10 dB loss as mentioned earlier. Also, a chip with an uncleaved input to enable the probe-facet to be butt coupled to the waveguide-facet was used. A measured output power of 1.5 µW was observed. Finally, directional coupling of the probe to the top of the waveguide was achieved after a difficult alignment step of the long probe to the thin waveguide. The alignment of the probe consisted of approaching the test waveguide from the top and pushing to create an overlapping section of approximately 2 mm in length while maintaining the alignment. A measured power of 0.24 µW was measured at the waveguide output. It is important to note that the coupling length was not optimized due to the difficulty in keeping the probe aligned on the waveguide under test while increasing the contact length. Clearly distinct modes were observed at the output facets for the three cases, an indication of the multi-mode nature of the waveguide under test.

 

Fig. 5. Infrared pictures of the output facet of S-shaped waveguide chips for 50 µm waveguides (a) Direct fiber optic to waveguide coupling; (b) Direct flexible probe-to-waveguide butt coupling; (c) Flexible probe directional coupling into waveguide.

Download Full Size | PDF

Taking into account the waveguide losses, the flexible probe directional coupling corresponds to 4% coupling, or an insertion loss of 13.6dB. This result is significant considering that the probe contact length and the alignment could not be optimized, and the gap due to roughness of the waveguide over the 2 mm length is non-zero. Furthermore, the probe carries optical power in several modes after the sharp 90o bend, which couple power less efficiently. Also very interesting is the high efficiency of 70% (-1.5dB) of the probe-to-waveguide butt coupling approach when compared to direct fiber to waveguide coupling. Unfortunately the probe-to-waveguide butt coupling cannot easily implement a wafer-level probing scheme.

4. Conclusion

An optical flexible waveguide probe suitable for wafer-level testing was proposed and experimentally demonstrated. Wafer-level testing of planar lightwave circuits of high densities can be performed without requiring die separation and polishing. This flexible waveguide probe approach enables tuning of the contact length in the directional coupler formed between the probe and a waveguide under test. We have simulated the device performance for single-mode and multimode probes, and optimized conditions in 2D and 3D models using the beam propagation method. The approach can compensate for variation in thickness of waveguides, a thin upper cladding and other parameters. Simulation results indicate that for single-mode case, tuning the length of the overlap region always leads to an optimal coupling higher than 80%. We successfully fabricated suspended SU8 probes with 15µm×30 µm cross section and lengths in excess of 6 mm. A coupling efficiency of 4% from the multimode flexible probe into the waveguide under test was achieved, compared to direct butt-coupling of the fiber to the waveguide. These are exciting results which pave the way to the development of higher efficiency broadband single-mode wafer-level testing of PLCs that could lead to shorter development cycles, yield improvement, and savings in die separation and polishing. Finally, a novel flexible optical wire bonding strategy can be implemented.