Department of Material Science and Engineering, Massachusetts Institute of Technology, 77 Massachusetts Avenue,. Cambridge, Massachusetts 02139. John P.
September 1, 2005 / Vol. 30, No. 17 / OPTICS LETTERS
2251
Trimming of microring resonators by photooxidation of a plasma-polymerized organosilane cladding material Daniel K. Sparacin,* Ching-yin Hong, Lionel C. Kimerling, and Jurgen Michel Department of Material Science and Engineering, Massachusetts Institute of Technology, 77 Massachusetts Avenue, Cambridge, Massachusetts 02139
John P. Lock* and Karen K. Gleason Department of Chemical Engineering and Institute for Soldier Nanotechnologies, Massachusetts Institute of Technology, 77 Massachusetts Avenue, Cambridge, Massachusetts 02139 Received February 24, 2005; revised manuscript received April 7, 2005; accepted April 18, 2005 As the complexity of microphotonic devices grows, the ability to precisely trim microring resonators becomes increasingly important. Photo-oxidation trimming uses UV irradiation to oxidize a cladding layer composed of polymerized hexamethyldisilane (6M2S) deposited with plasma-enhanced chemical vapor deposition (PECVD). PECVD 6M2S has optical properties that are compatible with microring devices, and its high cross linking renders it insoluble. Photo-oxidation decreases the refractive index of PECVD 6M2S by nearly 4%, permitting large resonance shifts that are not feasible with thermal trimming techniques. Resonance shifts from single-mode, 100 m diameter Si3N4 共n = 2.2兲 rings were as large as 12.8 nm for the TE mode and 23.5 nm for the TM mode. © 2005 Optical Society of America OCIS codes: 230.7380, 160.4670.
Microring resonators (Fig. 1) are a basic building block of photonic circuits, permitting complex functionality for optical systems. Ring resonators can serve as filters for multiplexing and demultiplexing broadband optical signals,1 dispersion compensators for accurately controlling phase,2 lasers,3 and ultrafast all-optical switches.4 At resonance, light coupled to the input port propagates through the bus waveguide, evanescently couples into the ring, and exits the drop port. At other wavelengths, light continues in the bus waveguide and exits at the through port. Ring resonators can be theoretically designed to have ideal channel dropping characteristics: a broad, steeply sloped, flat-topped spectral response with 100% efficiency.1 The resonance condition [Eq. (1)] relates D, the diameter of the ring; 0, the free space wavelength of resonant light; m, an integer indicating the resonator mode number; and n, the effective index of the ring: 0 = n共D/m兲.
共1兲
Precise control over 0 in each ring is critical for microphotonics integration. As ring diameters shrink to less than 10 m, nondeterministic pattern transfer errors limit dimensional precision and preclude the fabrication of identical devices. Thus, a postproduction trimming process to modify n and control 0 is essential. Conventional trimming methods utilize resistive microheaters to induce a thermo-optic response in the core material,5,6 where dn/dT is typically 10−5 – 10−4 K−1 for dielectrics and negative 1 – 4 ⫻ 10−4 K−1 for polymers. The small magnitude of these thermo-optic coefficients corresponds to a feasible thermal trimming range of a couple of nanometers. This architecture adds several steps to the fab0146-9592/05/172251-3/$15.00
rication of a photonic circuit, limits the density of devices to maintain thermal isolation, and requires significant power consumption for the rings to be kept “trimmed.” An alternative trimming method uses an organosilicon polymer film as the cladding material and adjusts its refractive index via photo-oxidation when irradiated with UV light. This effect has been demonstrated using a dip-coating technique to coat ring resonators with a polysilane material.7 Having a refractive index similar to SiO2, which is the predominant material of choice for cladding layers, organosilicon polymers can be integrated with Si, SiON, and Si3N4 high-index-contrast microring resonators. The material is transmissive over a broad range of visible and near-IR light,8,9 which is useful for microphotonic applications.10 However, polymers deposited from solution by dip coating or spin coating can often be redissolved, making the film incompatible with subsequent rinse steps in the microfabrication process. The low degree of cross linking in soluble-based
Fig. 1. Schematic of a ring resonator device. © 2005 Optical Society of America
2252
OPTICS LETTERS / Vol. 30, No. 17 / September 1, 2005
⌬ = 0,2 − 0,1 = n2共D/m2兲 − n1共D/m1兲.
共2兲
Since the resonant wavelength shift is continuous, the mode number remains fixed 共m1 = m2兲 throughout the shift, yielding ⌬ = 0,1
Fig. 2. Refractive index of PECVD 6M2S versus UV flux.
polymers can also make them prone to swelling when they are irradiated or are in contact with chemical solvents. We demonstrate a vapor-phase technique for depositing an amorphous and highly cross-linked top cladding layer directly onto ring resonator devices. PECVD 6M2S is insoluble, does not swell in organic solvents, and demonstrates good stability in ambient light, atmosphere, and temperature. In this Letter, photo-oxidation trimming was tested on single-mode Si3N4 ring resonators with PECVD 6M2S top cladding, and resulting resonance shifts were compared to a theoretical model. Si3N4 microring resonators, designed for singlemode operation at = 1550 nm, were fabricated on top of a 3 m oxide undercladding layer on (100) Si. The microrings have a diameter of 100 m and have cross-sectional dimensions of 400 nm ⫻ 750 nm. A 1 m 6M2S top cladding layer was deposited directly onto the ring resonators by use of a PECVD process that has been described elsewhere.11 Spectral characterization of the microrings was done at the through port in both TE and TM polarizations by a C + L band, JDS Uniphase swept wavelength system (tunable laser and broadband photodetector) used in conjunction with a Newport AutoAlign System. A Mineralight hand-held lamp (Model UVGL-25) emitting = 254 nm UV light was used to irradiate samples with a flux of 1.7 W / cm2. The same UV irradiation process was conducted separately on a sample from the same wafer, and the PECVD 6M2S film was measured after each exposure with a Woolam M-2000 variable-angle spectroscopic ellipsometer. Ellipsometry data were fitted to the Cauchy–Urbach model,12 yielding the film thickness and the Cauchy constants for calculating the refractive index at 1550 nm. The Woolam ellipsometer operates from 450 to 720 nm, but extrapolated refractive index values at 1550 nm were validated by repeating a few measurements with a Sopra GES5 ellipsometer operating from 800 to 1750 nm. Assuming single-mode operation throughout, the predicted shift of the resonance after UV trimming (subscript 2) from the initial state (subscript 1), ⌬, can be derived from Eq. (1):
冉 冊 n2
n1
−1 .
共3兲
A controllable decrease in the refractive index of the PECVD 6M2S cladding layer was achieved using photo-oxidation induced by UV irradiation (Fig. 2). The overall decrease is ⬃4%, from 1.52 to 1.46 at = 1550 nm. UV light having a wavelength less than ~300 nm causes chain scission in organosilicon polymeric materials, and subsequent oxidation converts Si–Si bonds into Si–O–Si bonds13 (Fig. 2 inset). Incorporation of oxygen reduces the molecular density of the material and causes the refractive index to decrease. The overall index change is about 50% greater than the response observed with dip-coated polysilane material. Output power from the rings was monitored in situ during UV trimming. Figure 3 displays the results for two specific exposure times where minima occur at the resonant wavelengths. The resonant wavelengths were observed to shift continuously with irradiation of the PECVD 6M2S cladding. The overall resonance shifts, after a UV flux of 1000 J / cm2, were 12.8 nm for the TE mode and 23.5 nm for the TM mode (Fig. 4), exceeding the free spectral range (FSR) for both the TE 共3.9 nm兲 and the TM 共4.5 nm兲 polarizations. Resonance shifts were not observed for UV-irradiated Si3N4 microrings without a PECVD 6M2S top cladding. To compare experiment with theory, a model was devised using the relation found in Eq. (3) and the exposure curve data in Fig. 2. Apollo Mode Solver software was used to calculate the effective index of the ring resonator for a given refractive index of the PECVD 6M2S cladding material for each polarization. The resulting theoretical resonance shifts (Fig. 4) agree fairly well with the experimental data in
Fig. 3. TM mode spectral measurements of a 100 m Si3N4 ring resonator 共0,1 = 1564.5 nm兲 after 300 and 420 s of UV irradiation at 1.7 W / cm2.
September 1, 2005 / Vol. 30, No. 17 / OPTICS LETTERS
2253
variances in 0 could be counteracted with a global heater, assuming uniform aging of the locally trimmed film. Reducing the refractive index of PECVD 6M2S is an irreversible process because of the nature of the photo-oxidation reaction. However, the process allows for some error. For ring resonators that have small FSRs compared with the maximum desired resonant wavelength shift, as was the case with the rings used in this experiment, overexposing a ring can be remedied by simply using more UV light and shifting the resonance one more FSR length. In conclusion, photo-oxidation trimming of PECVD 6M2S cladding provides a precise, localized, and controllable technique for specifying the resonance condition of ring resonators. Fig. 4. Experimental and theoretical resonance shifts for TE and TM modes.
terms of functional form and magnitude. Although swelling of PECVD 6M2S with oxidation is very minor due to cross linking, the thickness of the top cladding layer varies by about 1% from the beginning to the end of the trimming process. Additionally, deposition uniformity of the PECVD 6M2S thickness can vary by another 1 or 2% from sample to sample. These thickness variations are the probable cause of deviations between theoretical results and experimental data for the TM mode compared with the TE mode. In this waveguide geometry, the effective index of the TE mode is primarily sensitive to in-plane variations of the core/cladding thickness and or index, such as sidewall roughness, which is a major source of loss for waveguide devices. Analogously, the effective index of the TM mode is sensitive to variations out of plane (cladding thickness), resulting in uncertainty in the modeling. Use of a lower-power UV source can reduce the shift rate and presumably increase the precision if needed. Additionally, localization of the UV exposure, can be used to preserve the spectral response of higher-order filters that require multiple rings. Differences between ring-to-ring and ring-to-bus coupling require localized index trimming on separate areas of the filter so that all rings are kept in resonance with one another. Trimmed ring resonators were measured over a range of temperatures from 25 to 70 ° C, and the thermal-tuning coefficient of the resonance shift was found to be −0.10 nm/ K, corresponding to a thermooptic coefficient (dn/dT) for the system of −1.3 ⫻ 10−4 K−1. This value is on par with other polymerbased ring resonator devices.6 Many polymers used as waveguide cladding materials undergo densification over a span of weeks or months, causing drifts in the refractive index of the order of 10−4(Ref. 14), corresponding to ⌬ of ⬃0.1 nm. Although this effect has not yet been quantified for PECVD 6M2S, long-term
We thank Gilles Benoit for his help in using the Sopra GES 5 spectroscopic ellipsometer and Yoel Fink for access to the equipment. This work was supported in part by the MRSEC Program of the National Science Foundation under contract DMR 02-13282 and by the U.S. Army through the Institute for Soldier Nanotechnologies under contract DAAD-19-02-D0002 with the U.S. Army Research Office. The content does not necessarily reflect the position of the Government, and no official endorsement should be inferred. *Co-first-authors. References 1. B. E. Little, S. T. Chu, H. A. Haus, J. Foresi, and J.-P. Laine, J. Lightwave Technol. 15, 998 (1997). 2. C. K. Madsen, G. Lenz, A. J. Bruce, M. A. Capuzzo, L. T. Gomez, T. N. Nielsen, and I. Brener, Opt. Lett. 24, 22, 1555 (1999). 3. S. J. Choi, Z. Peng, Q. Yang, S. J. Choi, and P. D. Dapkus, IEEE Photon. Technol. Lett. 16, 356 (2004). 4. V. R. Almeida, C. A. Barrios, R. R. Panepucci, and M. Lipson, Nature 43, 1081 (2004). 5. P. Heimala, P. Katila, J. Aarnio, and A. Heinamaki, J. Lightwave Technol. 14, 2260 (1996). 6. P. Rabiei and W. H. Steier, IEEE Photon. Technol. Lett. 15, 1255 (2003). 7. S. T. Chu, W. Pan, S. Sato, T. Kaneko, B. E. Little, and Y. Kokubun, IEEE Photon. Technol. Lett. 11, 688 (1999). 8. L. A. Hornak, T. W. Wedman, and E. W. Kwock, J. Appl. Phys. 67, 2235 (1990). 9. P. K. Tien, G. Smolinsky, and R. J. Martin, Appl. Opt. 11, 637 (1972). 10. L. Eldada, Proc. SPIE 5225, 49 (2003). 11. J. P. Lock and K. K. Gleason, Appl. Opt. 44, 1691 (2005). 12. H. G. Tomkins and W. A. McGahan, Spectroscopic Ellipsometry and Reflectometry (Wiley-Interscience, 1999). 13. R. D. Miller and J. Michl, Chem. Rev. (Washington, D.C.) 89, 1359 (1989). 14. C. G. Robertson and G. L. Wilkes, Polymer 39, 2129 (1988).