We report the design, fabrication, and measurement of high efficiency, compact 45° single air interface bends in low refractive index contrast waveguides in a low refractive index material system. Using standard microfabrication techniques, the bends are fabricated on silicon substrates using perfluorcyclobutyl (PFCB) copolymers, which feature a high glass transition temperature and low absorption loss. The measured 45° bends have a loss of 0.30±0.03dB/bend for TM polarization and 0.33±0.03dB/bend for TE polarization.
©2004 Optical Society of America
It is difficult to realize very small planar lightwave circuits (PLCs) in low loss waveguide material systems in which the refractive index contrast is typically in the range of 0.3% to 1.5%. This is primarily because the minimum bend radius typically must be at least several millimeters to achieve high optical throughput. In the literature, several different methods for reducing the area of a waveguide bend have been reported. For example, the minimum bend radius can be significantly reduced by creating a region of large refractive index contrast at the bend by etching deep trench structures . These structures define not only the bends, but also tapered transitions into and out of the bends . The total area required for a bend therefore consists of both the bend itself and the taper regions. Another method consists in using corner mirrors based on hybrid integration of photonic crystals with conventional waveguides  or single air interface bends (SAIBs) , both of which can change the mode’s direction in a very small area. SAIBs have been proposed for a variety of waveguide material systems and operating wavelengths [4–10]. However, high efficiency single air interface bends in low refractive index contrast waveguides have yet to be demonstrated.
As proposed in Ref. 4, SAIBs with bend angles smaller than 90° can be used to achieve high optical efficiency in low refractive index material systems. This strategy insures that total internal reflection occurs for all angular spectrum components of the waveguide mode, thus resulting in a high efficiency bend. In this paper we report the design, fabrication, and measurement of high efficiency single air interface 45° bends in waveguides with a refractive index contrast of 1.3% using perfluorocyclobutane (PFCB) copolymer materials [11,12]. These polymers offer low propagation loss and a high glass transition temperature of over 200°C .
In the following sections, we first discuss the design of SAIBs and determine their placement tolerance. This is followed by a description of our fabrication process. In the last section, we report our measurement results.
2. Single air interface bend analysis
Based on core and cladding PFCB copolymer formulations supplied by Tetramer Technologies, Inc, we designed single mode channel waveguides to operate at a wavelength of 1.55µm. The core dimensions are 3.6µm×3.6µm. The core refractive index is 1.4836 for TE polarization (electric field in the plane of the substrate) and 1.4816 for TM polarization (electric field out of the plane). The TE and TM cladding refractive indices are 1.4644 and 1.4625, respectively. To form a small area bend, a SAIB structure is placed at the intersection of two straight waveguides as shown in Fig. 1. The red region represents an air trench etched vertically into the waveguide cladding/core/cladding stack. It provides a PFCB/air interface to reflect the waveguide mode through total internal reflection (TIR). In the following discussion, the origin (labeled O in Fig. 1) is taken as the point at which the centers of the waveguides intersect, and the distance between the PFCB/air interface and the origin along the z axis is zp.
To numerically calculate the optical performance of SAIB structures, we use a rigorous two-dimensional finite difference time domain (2-D-FDTD) method. The geometry and magnitude squared time averaged fields at λ=1.55µm for both TE and TM polarizations for our 45° bend are shown in Figs. 2(a) and 2(b). The edge of the air interface is placed at zp=-0.2µm. At this position, the Goos-Hanchen shift is such that the bend efficiency is nearly equal for both TE and TM polarized light. The bend efficiencies are calculated from the FDTD simulations with a mode overlap integral (MOI), which gives the ratio of the power in the guided mode in the output waveguide to the power in the incident guided mode. At a wavelength of 1.55µm, the calculated efficiencies using 2-D FDTD are 99.0% and 98.7% for TE and TM polarizations, respectively. To verify these results, we determined the efficiencies using a 3-D FDTD method implemented on a Linux cluster, which gave 98.5% for TE polarization and 98.0% for TM polarization. Since the 2-D and 3-D FDTD results are within 1% of each other and 3-D FDTD analysis is so computationally intensive, we use 2-D FDTD calculations throughout the remainder of this paper.
For device fabrication it is important to understand the effect of misalignment and etch undercut on the optical performance of SAIBs. In both cases, the result is misplacement of the SAIB interface with respect to the designed position. The bend efficiency as a function of this misplacement is shown in Fig. 3. The alignment accuracy must be better than ±0.3µm around the designed placement at zp=-0.2µm to have a bend efficiency of 95% or greater for both TE and TM polarizations. The same applies to undercut. Thus, the misplacement of the air interface caused by the both misalignment and undercut must be less than ±0.3µm to achieve the aforementioned bend efficiency.
We fabricated SAIBs in PFCB waveguides on a silicon substrate using typical microfabrication techniques. First, a 12 µm thick undercladding layer was spun and cured at 190°C followed by a 3.7 µm thick core layer. Next, we deposited SiO2 as a hard mask for the waveguide etch via RF sputtering and defined the waveguides using contact lithography with AZ 701 MiR i-line photoresist. Then, the SiO2 hard mask was etched in a CHF3 chemistry with a reactive ion etcher (RIE) followed by etching of the PFCB core layer in an inductively coupled plasma reactive ion etcher (ICP-RIE) using a He and O2 etch chemistry. After the waveguide core etch, the SiO2 layer was stripped in buffered hydrofluoric acid. The overcladding was then spun and cured at 190°C. Another SiO2 hard mask was deposited followed by photoresist pattern definition with contact lithography to define the air trench patterns. The SiO2 was then etched in the RIE. Finally, the air trenches were etched in the ICP-RIE using a He and O2 chemistry.
As shown in the 2-D FDTD simulation results in Fig. 3, the placement of the air trench interface with respect to the bend is critical. Moreover, contact photolithography doesn’t allow us to align the SAIBs with the necessary tolerance. Thus, we introduced a range of predefined offsets into the photomask to compensate for actual misalignment during the fabrication process. These offsets compensate for a mask misalignment of up to ±1µm in 0.1µm increments in both the x and y directions. Hence, if our alignment is less than or equal to one micron in x and y, at least one set of bends will be aligned to within 0.1 µm and the expected bend efficiency as per the 2-D FDTD simulations in Fig. 3 will stay within 1% of the maximum. Another fabrication issue that must be considered is that during the SAIB deep etch process the PFCB material beneath the edge of the etch mask is slowly undercut. After a 14µm to 16µm deep etch this undercut causes considerable misplacement of the SAIB interface. Thus, we introduced another set of offsets into the mask layout for undercut compensation. The range of this compensation goes from 0.3µm to 1.3µm of undercut in 0.2µm increments. This means that if the air interface misplacement is due only to undercut then the interface for at least one set of SAIBs will be located to within 0.1µm of its designed position for any undercut value in the 0.2µm to 1.4µm range. Figure 4 shows a cross-sectional SEM image of the undercut typically observed for our fabrication process, which is 1.1µm.
For each combination of misalignment and undercut compensation, a set of waveguides was designed containing 2, 4, 8, and 16 bends. Each of these is termed a group. Nine of these groups are included in one 1.5cm×1.5cm die, and two die are required to account for all of the misalignment offsets for each undercut compensation. Twelve die containing all possible misalignment and undercut combinations are fabricated on a single 3” wafer.
Even though we did not quantify the amount of mask misalignment on fabricated die we evaluated it qualitatively by looking at the symmetry of the air trench placement with respect to the waveguides from microscope images as illustrated in Fig. 5. This image of a pair of bends was taken with a DIC filter with the focus at the waveguide plane. It shows that the bends’ symmetry with respect to the waveguides is quite good. This symmetry implies good alignment for this bend. We used this criterion to determine which waveguide groups to initially measure since better alignment results in higher bend efficiency.
Figure 6 shows an SEM image of a finished air trench bend. It consists of a 25µm wide trench that is 87µm long. The waveguide isn’t visible because it is buried between the under and overcladding layers. As illustrated in Fig. 7, the verticality of the etched sidewall is excellent (typically better than ±0.5°). However, as seen in Fig. 6, there is some evidence of sidewall roughness in the form of vertical striations. A clearer view of the sidewall roughness is shown in Fig. 8. Note that there are two different types of roughness that can be observed. At the top of the etched face, we can see some fine scale roughness that is related to how long the sidewall is exposed to the etch. Further down, vertical lines that start at the top of the sidewall are visible. These seem to be due to roughness of the edge of the etch mask that is transferred to the polymer sidewall. We are currently working on reducing this roughness.
4. Experimental measurement and discussion
Our measurement setup consists of a laser operating at 1550nm coupled to a polarization maintaining (PM) fiber, which is butt coupled to an input waveguide on the device die. For detection, we use a single mode (SM) fiber butt coupled to the corresponding output waveguide. The fiber is connected to a photodetector. Both input and output fibers are on computer controlled three axis translation stages with piezo actuators. Index matching oil is used with the input and output fibers to improve the coupling efficiency and reduce scattering due to end face roughness on the waveguides. For TM measurements, the PM fiber is rotated to orient the electric field perpendicular to the substrate, while for TE it is oriented parallel to the substrate.
We automated our test bench to perform a simultaneous conical scan with the input and output fibers to find the maximum throughput for a waveguide. This yields measurement standard deviations typically less than 1%. However, the separation between each fiber and the waveguide is set manually using a reticle while looking through the eyepiece of a microscope set at a total magnification of 225x. At this magnification, the resolution of the reticle is 10 µm/division. To determine the effect of the inherent variability of a manual process on our measurement accuracy, we first examined the relationship between the measured throughput power and the separation between an output waveguide and output fiber. For simplicity we varied only one fiber’s position. Figure 9(a) shows the results for such a scan. Note that there is only a 13% variation in the measured optical power for a 25µm displacement of the fiber. We then tested the repeatability of the measured power when a user sets the fiber separation for both input and output fibers using the method described above. The results are shown in Fig. 9(b). The standard deviation of the measurements is only 3% of the average value, which results in an acceptably small measurement error for the bend efficiency.
To measure the optical loss of fabricated SAIBs we measured the power transmitted through waveguides containing 2, 4, 8, and 16 bends. Each of these waveguides in a group has the same misalignment and undercut compensation. Moreover, each waveguide has the same length, so that bend loss can be measured independent of the propagation loss. The measured power for a group is plotted in dB as a function of the number of waveguide bends and the data is fitted to a straight line. The line’s slope is the loss per bend.
We measured several groups among the ones that had the greatest symmetry when observed with a microscope as in Fig. 5. The measured data for TM polarization of one of these is shown in Fig. 10. The optical loss per bend is 0.30 ± 0.03dB/bend, which is a bend efficiency of 93.4%. For the same waveguide group, the measured bend efficiency for TE polarization is 0.33±0.03dB/bend.
By measuring the rest of the groups with 1.1µm undercut compensation, but with different misalignment compensations, we are able to compare the predicted variation in bend efficiency as a function of misalignment with the measured data of the fabricated samples. Measurement results are shown in Fig. 11. The experimental data is similar to the 2-D FDTD simulation data of Fig. 3, except for the 0.25 to 0.3 dB excess loss of the measured data. We believe that this is primarily due to scattering at the bend interface, although a small amount is likely owing to residual undercut and mask alignment errors as well as a slight sidewall bowing or tilt. Also, note that the loss per bend is similar for both TE and TM polarizations for bends with the lowest loss. The polarization dependant loss for these bends is therefore quite low.
We have presented results for compact high efficiency 45° bends in PFCB waveguides. Total internal reflection is used to reflect the waveguide mode from a single air interface. Bend design and tolerance analysis for fabrication was carried out using 2-D and 3-D FDTD simulations. The SAIB fabrication process and fabrication results have been presented. Finally, experimental measurements show an achieved bend loss of 0.30±0.03dB/bend for TM polarization and 0.33±0.03dB/bend for TE polarization.
Future research will focus on reducing the etched interface roughness to obtain higher bend efficiencies. Moreover, we plan to significantly reduce the air trench size to enable higher bend packing densities to be achieved. We estimate that 40µm by 10µm air trenches are feasible.
This work was supported in part by DARPA Grant N66001-01-1-8938 and National Science Foundation Grant EPS-0091853.
References and Links
1. L.H. Spiekman, Y. S. O., E.G. Metaal, F.H. Groen, P. Demeester, and M.K. Smit, “Ultrasmall waveguide bends: the corner mirrors of the future?” IEEE Proc. Optoelectron. 142, 61–65 (1995). [CrossRef]
2. Milos Popovic, Shoji K. W., Hermann A. Akiyama, Jurgen Haus, and Michel, “Air Trenches for Sharp Silica Waveguide Bends,” J. Lightwave Technol. 20, 1762–1772 (2002). [CrossRef]
3. Seunghyun Kim, Gregory P. Nordin, Jianhua Jiang, and Jingbo Cai, “High Efficiency 90° Silica Waveguide Bend Using an Air Hole Photonic Crystal Region,” IEEE Phot. Technol. Lett. 16, 1846–1848 (2004). [CrossRef]
5. Kazuhiko Ogusu, “Transmission Characteristics of Optical Waveguide Corners,” Opt. Commum. 55, 149–153 (1985). [CrossRef]
6. Regis Orobtchouk, Suzanne Laval, Daniel Pascal, and Alain Koster, “Analysis of Integrated Optical Waveguide Mirrors,” J. Lightwave Technol. 15, 815–820 (1997). [CrossRef]
7. L. Faustini, C. Coriasso, A. Stano, C. Cacciatore, and D. Campi, “Loss Analysis and Interference Effect in Semiconductor Integrated Waveguide Turning Mirrors,” Photon. Technol. Lett. 8, 1355–1357 (1996). [CrossRef]
8. John E. Johnson and C. L. Tang, “Precise Determination of Turning Mirror Loss Using GaAs/AlGaAs Lasers with up to Ten 90° Intracavity Turning Mirrors,” Photon. Techn. Lett. 4, 24–26 (1992). [CrossRef]
9. P. D. Swanson, D. B. Shire, C. L. Tang, M. A. Parker, J. S. Kimmet, and R. J. Michalak, “Electron-Cyclotron Resonance Etching of Mirrors for Ridge-Guided Lasers,” Photon. Techn. Lett. 7, 605–607 (1995). [CrossRef]
10. Y. Z. Tang, W. H. Wang, T. Li, and Y. L. Wang, “Integrated Waveguide Turning Mirror in Silicon-on-Insulator,” Phot. Techn. Lett. 14, 68–70 (2002). [CrossRef]
11. Smith Jr., D.W. Chen, S. Kumar, S. Ballato, J. Shah, H. Topping, and C. Foulger, “Perfluorocyclobutyl Copolymers for Microphotonics,” S. Adv. Mater. 14, 1585 (2002). [CrossRef]
12. Smith Jr., D.W. Hoeglund, A.B. Shah, H.V. Radler, M.J. Langhoff, and C.A., “Perfluorocyclobutane Polymers for Optical Fibers and Dielectric Waveguides,” in Optical Polymers, J. Harmon; Ed., ACS Symp. Ser.795, Ch. 4, pp. 49–62 (2001). [CrossRef]
13. John Ballato, Dennis W. Smith Jr, and S. H. Foulger, “Optical Properties of Perfluorocyclobutyl (PFCB) Polymers,” J. Opt. Soc. Am. B 201838 (2003). [CrossRef]