One of the challenges of the high refractive index contrast of silicon photonics platform is the high sensitivity of the resonance wavelength of resonators to dimensional variations caused by fabrication process variations. In this work, we have experimentally demonstrated an accurate post-fabrication trimming technique for optical devices that is robust to process variations. Using this technique, we have reduced the random variation of the resonance wavelength of 4 µm diameter resonators by a factor of 6 to below 50 pm. The level of accuracy achieved in this work is adequate for most of the RF-photonic, interconnect, and optical signal processing applications. We also discuss the throughput of this technique and its viability for wafer-scale post-fabrication trimming of silicon photonic chips.
©2013 Optical Society of America
One of the advantages of silicon (Si) material platform for photonics is the high refractive index contrast of Si with respect to the silicon dioxide (SiO2) cladding [1, 2]. This property enables high field confinement in compact waveguides and very low-loss sharp bends, which combined with the mature and high-quality fabrication processes for Si devices has resulted in achieving ultra-compact traveling-wave resonators (TWRs) with high quality factors (Qs) . These compact and high-Q resonators are valuable for many applications including optical signal processing and on-chip interconnection, as they enable compact, low loss, and ultra-low power active and passive devices [4–6]. However, one of the challenges of using these resonators is the high sensitivity of their resonance wavelength to random dimensional variations that is caused by the variations in the fabrication processes. This sensitivity is a direct result of the sensitivity of the optical path-length of the optical field to size variations in tightly-confined guided-wave devices .
The proposed techniques to address the random variation of the resonance wavelength fall into two categories: active tuning mainly through the thermooptic effect [4, 5], and permanent post-fabrication trimming (PFT) [8–15]. One of the main challenges of active tuning techniques is the complexity of signaling and control electronics of the tunable elements when the number of these components becomes very large. At the same time, the power consumption of the tunable elements can be prohibitive for applications such as on-chip interconnects that have tight power budgets . The alternative approach is PFT in which the refractive index of the cladding or core materials is permanently changed for correction of the optical path-length. The main PFT techniques demonstrated so far are based on either of these two mechanisms: 1) changing the level of compaction or stress/strain of the cladding or core materials through high-energy electron or laser beams [8, 9]; 2) changing the refractive index of a photo-sensitive cladding material (e.g., chalcogenide glass, or polymer) by applying high-energy UV or visible light [10–12]. Other alternative PFT approaches such as polymerization of a liquid crystal cladding , electron-beam bleaching of chromophore-doped polymer cladding , and local oxidation of the surface of Si  have also been demonstrated. One of the concerns with most of the proposed PFT techniques is the long-term stability of the trimming mechanism. In the case of cladding compaction it is known that the level of compaction relaxes over time ; and in the case of photo-sensitive polymers and chalcogenide glasses it is known that refractive index drifts over time due to densification  and structural relaxation , respectively. These effects cause a great deal of reliability concern for the proposed PFT techniques.
In this paper, we demonstrate a PFT technique that is based on patterning a thin silicon nitride (SiN) film on top of the Si device and does not have any long-term stability issue. We were able to reduce the resonance variation of 144 4 µm-diameter resonators by a factor of 6 to below 50 pm in a single fabrication step. This indicates that this technique is robust to process variations and combined with its relatively high throughput, it can be a viable solution for wafer-scale trimming in many applications.
2. Proposed PFT technique
The PFT approach discussed in this paper is based on partial etching of a thin film of SiN that is deposited on top of the device. By controlling the amount of SiN removed from the device, we can adjust the level of perturbation on the optical field and consequently the optical path-length. Figures 1(a) and 1(b) schematically show the cross-section of a ridge waveguide on silicon-on-insulator (SOI) platform before and after PFT using this approach. Here, we have intentionally placed a thin buffer layer of SiO2 between the Si and SiN layers to protect the top surface of the Si layer in the etching process of the SiN layer. At the same time, since the device is covered with a SiO2 cladding after the trimming step, the small amount of the buffer SiO2 layer that might be etched during the etching step of the SiN layer is filled back with SiO2. As a result, this technique is not sensitive to variation of the etch-rate or slight over-etching of the SiN layer [Fig. 1(b)]. Also, the thickness of the SiO2 buffer layer is an additional parameter to control the total amount of resonance wavelength trimming. The advantage of trimming SiN layer instead of the Si directly is that it enables us to considerably reduce the sensitivity of the trimming step to process-dependent variations by appropriately adjusting the total amount of perturbation of the SiN film on the optical mode.
In this work, we focus on the PFT of the ultra-compact (4 µm diameter) microring resonators because of their practical significance in many applications [5, 6, 18]. Figure 1(c) shows the cross-section of the Hz field component of the first radial order TE (i.e., electric field parallel to the plane of the ring) mode of a 4 µm diameter microring with a width of 0.75 µm at λ = 1.55 µm. These are the dimensions used in the experimental demonstration of the proposed PFT approach (Section 3). It is observed from Fig. 1(c) that most of the optical field is concentrated towards the outer edge of the microring. In practice, the sidewalls of the microring are not perfectly normal (i.e., perpendicular to the interface of Si and SiO2) resulting in slight etching of the Si layer at the outer and inner edges of the microring during the etching step of the SiN layer for trimming. Because of the strong concentration of the field near the outer edge of the device, we avoid etching the SiN in this area. Thus, to control the level of trimming, we etch a disk with radius rt, concentric with the microring, inside the SiN layer. Therefore, after the PFT process, the SiN layer is in the form of a ring with inner radius rt sitting on top of the Si microring as schematically shown in Fig. 1(d).
3. Experimental results
We implemented the PFT technique by fabricating a series of 4 µm diameter microrings on an SOI substrate with Si and buried oxide (BOX) layer thicknesses of 230 nm and 3 µm, respectively. The SOI substrate is first coated with SiO2 (40 nm) and SiN (30 nm) films (for trimming) using plasma-enhanced chemical vapor deposition (PECVD). The film thicknesses used here provide a resonance wavelength trimming range of approximately 2.5 nm, which is sufficient to correct for the fabrication-induced resonance variations of the compact microrings studied in this work. The microrings are patterned using electron-beam lithography (EBL) and dry-etched of the SiN/SiO2/Si stack by inductively coupled plasma (ICP) etching. The sample is then coated with a polymer (aliphatic polyester-based urethane diacrylate from Sartomer) with an index very close to that of SiO2 so that it can be easily removed after the characterization for applying PFT. The device is characterized using a swept-wavelength transmission measurement setup with TE-polarized input light (see  for the details of the characterization). After the characterization of the device, the polymer cladding layer is removed and the PFT pattern is written in ZEP 520A (Zeon Chemicals) resist using EBL and then dry-etched in SiN. The trimmed device is finally covered with 1 µm of PECVD SiO2 as the top cladding layer. Figure 2(a) shows the scanning electron micrograph (SEM) of one of the trimmed microrings with rt = 1.75 µm.
The above process is first applied to achieve a set of calibration data in which the amount of resonance wavelength shift (Δλr) is measured as a function of the inner radius of the SiN trimming ring (rt). The amount of resonance wavelength shift is measured by varying rt from 1.3 µm to 1.95 µm and the results are shown in Fig. 2(b) (blue curve). It is observed that approximately 2.5 nm of wavelength shift can be achieved over the whole trimming range. Δλr is also theoretically modeled using a mode-solver code in COMSOL and the simulation results are shown in Fig. 2(b) (red curve). It is observed that the experimental curve has an offset of approximately −0.26 nm with respect to the theoretical curve. We believe that this offset is caused by the slight etching of the inner edge of the Si microring during SiN etching. By excluding this offset, we observe good agreement between the modeling and experimental results, which indicates that the demonstrated PFT technique is robust to the random variations in the fabrication process. Since in practice some resonators are shifted towards higher wavelengths (red-sifted) and some towards lower wavelengths (blue-shifted) with respect to their designed resonance, we need to be able to trim the resonance in both positive and negative wavelength directions. For this purpose, we position the trimming bias point at rt = 1.75 µm, which enables us to achieve positive and negative wavelength shift by varying rt below and above this value, respectively.
The calibration data is then used to correct for the resonance wavelength variations of 8 resonator arrays each consisting of 19 resonators. 4 of these arrays are in an add-drop configuration [Fig. 3(a)], and the other 4 are in a single-bus configuration with critically coupling condition. The first resonator in each array has an outer diameter of 4 µm and the diameters of the rest of the resonators are incremented by 8 nm with respect to their adjacent resonator to avoid the overlap of their resonance features. The first resonator in each array serves as a reference to correct for the temperature-induced resonance shift observed during the characterization of the device. We applied the proposed PFT technique on all of the resonators except for the reference resonator (i.e., a total of 18 resonators) in each array. The blue curves in Fig. 3(b) show the overlay of the normalized transmission spectra of the through port of the 18 resonators in one of the tested add-drop devices before PFT. Here, the horizontal axis is the wavelength detuning of each resonator in the array from its designed resonance wavelength. Because of the fabrication-induced variations in the dimensions of the resonators it is observed that the resonances are spread over ± 0.5 nm around their otherwise perfect (theoretically designed) positions. The transmission spectra of the same resonators after PFT are shown in red in Fig. 3(b) showing considerable improvement in the variation of the resonance wavelengths. In this particular device, the standard deviation (SD) of the resonance wavelength variation of the 18 resonators in the array is improved by a factor of 7 through PFT to approximately 30 pm. The blue and red curves in Fig. 3(c) show similar plots as in Fig. 3(b) for one of the critically coupled single-bus resonator arrays. It is again observed that our PFT technique can considerably reduce the variations of resonance wavelength of ultra-compact resonators. In all of the tested devices, the intrinsic Q of the resonators was in the range of 25-50k and variation of the intrinsic Q before and after PFT was less than 5%.
Figure 4(a) shows the SD of the resonance wavelength variation in all of the 8 resonator array devices tested in this work before (blue triangles) and after (red triangles) PFT. The histograms of the resonance wavelength variation before and after PFT for the 144 tested resonators are shown in Fig. 4(b). The SD of resonance wavelength variation is improved from 253 pm (before PFT) to 45 pm (after PFT) by a factor of 5.6 over all of the 144 tested resonators. To the best of our knowledge, this is the only technique demonstrated to this date that has achieved this level of improvement in resonance wavelength variations for an ensemble of devices in a single processing step.
4.1. Practical significance of the achieved resonance wavelength accuracy
In Section 3, we demonstrated better than 50 pm (~6 GHz) accuracy in the resonance wavelength (frequency) of ultra-compact resonators. To demonstrate the practical significance of this level of accuracy, we used the measurement results in Section 3 to evaluate the performance of an 18th order coupled-resonator filter [Fig. 5(a)] when there is random variation in the resonance wavelengths of the resonators in this device. The filter is designed to have a flat-top response with a 3 dB bandwidth of 25 GHz matching the wavelength spacing of an ITU grid . The blue curve in Fig. 5(b) shows the response of the drop port of this filter in the ideal case when there are no resonance wavelength variations. To observe the effect of random resonance variations on the response of the coupled-resonator devices, we used the resonance data of the 18 resonators in each of the resonator arrays that we measured in the Section 3 and assigned their wavelength variations to the 18 resonators of the studied coupled-resonator devices. The drop-port transmission spectra of the eight coupled-resonator devices modeled using the resonance variation data before PFT are shown by the curves in the gray shaded area in Fig. 5(b). It is observed that the transmission response of the filter is highly distorted with an insertion loss of more than 50 dB. By applying the resonance variation data after PFT, the response of the drop port is considerably improved and approaches the ideal response as shown by the curves in the red shaded area in Fig. 5(b). In these simulations, we assumed an intrinsic Q of 100k for all of the resonators in the filter. This particular example shows that the demonstrated PFT technique can be very useful for narrow-band applications and especially in the devices where the number of resonators is large.
4.2. Throughput of the PFT technique
The average electron beam exposure time for the trimming of a single resonator in this work is 7.6 ms. In other words, the exposure time of a single die with 1000 resonators (with the same size as those in this work) is 7.6 s; and the exposure time of 5000 of these dies is 10.5 hours. It is possible to further reduce the exposure time by a factor of 3-4 with the optimization of the design (trimming film thicknesses and trimming pattern) and electron beam exposure parameters. The other factor that affects the overall throughput of the PFT step is the device characterization step that is needed for every single chip on the wafer. This step is indispensable for any PFT approach to extract the exact parameters of the critical components on the chip (e.g., resonators). However, with the automated wafer-scale optoelectronic characterization equipment available today, this step is not a bottleneck of the PFT approach. Therefore, with such fast PFT mask exposure times and the achieved level of accuracy, the demonstrated technique can enable wafer-scale PFT of Si photonic devices for low to medium-volume manufacturing.
Here, we demonstrated a PFT technique that is robust to the variations in the fabrication process. Using this technique, we demonstrated a factor of 6 improvement in the resonance wavelength variations of 144 resonators and achieved better than 50 pm accuracy in the trimming of the resonance wavelengths of ultra-compact resonators. Considering that the demonstrated PFT is achieved in a single fabrication step and the fact that the electron-beam lithography exposure time for the trimming pattern of a single device is in the order of a few milliseconds, this technique has the potential for wafer-scale trimming of Si photonic devices for low to medium-volume manufacturing.
This work was funded by AFOSR STTR contract FA9550-11-C-0070 under Dr. Gernot Pomrenke.
References and links
1. D. A. B. Miller, “Device requirements for optical interconnects to silicon chips,” Proc. IEEE 97(7), 1166–1185 (2009). [CrossRef]
2. R. Soref, “The past, present, and future of silicon photonics,” IEEE J. Sel. Top. Quantum Electron. 12(6), 1678–1687 (2006). [CrossRef]
4. M. R. Watts, W. A. Zortman, D. C. Trotter, G. N. Nielson, D. L. Luck, and R. W. Young, “Adiabatic Resonant Microrings (ARMs) with Directly Integrated Thermal Microphotonics,” in Conference on Lasers and Electro-Optics/International Quantum Electronics Conference, OSA Technical Digest (CD) (Optical Society of America, 2009), paper CPDB10. [CrossRef]
5. A. H. Atabaki, A. A. Eftekhar, S. Yegnanarayanan, and A. Adibi, “Novel Micro-Heater Structure for Low-Power and Fast Photonic Reconfiguration,” in Conference on Lasers and Electro-Optics/International Quantum Electronics Conference, OSA Technical Digest (CD) (Optical Society of America, 2010), paper JThE44. [CrossRef]
6. M. R. Watts, W. A. Zortman, D. C. Trotter, R. W. Young, and A. L. Lentine, “Vertical junction silicon microdisk modulators and switches,” Opt. Express 19(22), 21989–22003 (2011). [CrossRef] [PubMed]
9. H. Haeiwa, T. Naganawa, and Y. Kokubun, “Wide range center wavelength trimming of vertically coupled microring resonator filter by direct UV irradiation to SiN ring core,” IEEE Photon. Technol. Lett. 16(1), 135–137 (2004). [CrossRef]
10. D. K. Sparacin, C. Y. Hong, L. C. Kimerling, J. Michel, J. P. Lock, and K. K. Gleason, “Trimming of microring resonators by photo-oxidation of a plasma-polymerized organosilane cladding material,” Opt. Lett. 30(17), 2251–2253 (2005). [CrossRef] [PubMed]
11. A. Canciamilla, F. Morichetti, S. Grillanda, P. Velha, M. Sorel, V. Singh, A. Agarwal, L. C. Kimerling, and A. Melloni, “Photo-induced trimming of chalcogenide-assisted silicon waveguides,” Opt. Express 20(14), 15807–15817 (2012). [CrossRef] [PubMed]
12. L. Zhou, K. Okamoto, and S. J. B. Yoo, “Athermalizing and Trimming of Slotted Silicon Microring Resonators With UV-Sensitive PMMA Upper-Cladding,” IEEE Photon. Technol. Lett. 21(17), 1175–1177 (2009). [CrossRef]
13. S. Lambert, W. De Cort, J. Beeckman, K. Neyts, and R. Baets, “Trimming of silicon-on-insulator ring resonators with a polymerizable liquid crystal cladding,” Opt. Lett. 37(9), 1475–1477 (2012). [CrossRef] [PubMed]
15. Y. Shen, I. Divliansky, D. Basov, and S. Mookherjea, “Perfect set-and-forget alignment of silicon photonic resonators and interferometers,” in Optical Fiber Communication Conference, OSA Technical Digest (CD) (Optical Society of America, 2011), paper PDPC3.
16. C. B. Norris and E. P. Eernisse, “Ionization dilatation effects in fused silica from 2 to 18-Kev electron-irradiation,” J. Appl. Phys. 45(9), 3876–3882 (1974). [CrossRef]
17. C. G. Robertson and G. L. Wilkes, “Refractive index: A probe for monitoring volume relaxation during physical aging of glassy polymers,” Polymer (Guildf.) 39(11), 2129–2133 (1998). [CrossRef]
18. Z. Xia, A. A. Eftekhar, M. Soltani, B. Momeni, Q. Li, M. Chamanzar, S. Yegnanarayanan, and A. Adibi, “High resolution on-chip spectroscopy based on miniaturized microdonut resonators,” Opt. Express 19(13), 12356–12364 (2011). [CrossRef] [PubMed]
19. Q. Li, M. Soltani, S. Yegnanarayanan, and A. Adibi, “Design and demonstration of compact, wide bandwidth coupled-resonator filters on a silicon-on- insulator platform,” Opt. Express 17(4), 2247–2254 (2009). [CrossRef] [PubMed]