Silicon-on-insulator microring add-drop filters with free spectral ranges over 30 nm

Purdue University Purdue e-pubs Birck and NCN Publications Birck Nanotechnology Center January 2008 Silicon-on-insulator microring add-drop filters with free spectral ranges over 30 nm Shijun Xiao Purdue University Maroof H. Khan Birck Nanotechnology Center, Purdue University, mhkhan@purdue.edu Hao Shen Birck Nanotechnology Center, Purdue University, shen17@purdue.edu Minghao Qi Birck Nanotechnology Center, Purdue University, mqi@purdue.edu Follow this and additional works at: http://docs.lib.purdue.edu/nanopub Xiao, Shijun; Khan, Maroof H.; Shen, Hao; and Qi, Minghao, "Silicon-on-insulator microring add-drop filters with free spectral ranges over 30 nm" (2008). Birck and NCN Publications. Paper 180. http://docs.lib.purdue.edu/nanopub/180 This document has been made available through Purdue e-pubs, a service of the Purdue University Libraries. Please contact epubs@purdue.edu for additional information.

228 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 26, NO. 2, JANUARY 15, 2008 Silicon-on-Insulator Microring Add-Drop Filters With Free Spectral Ranges Over 30 nm Shijun Xiao, Member, IEEE, Member, OSA, Maroof H. Khan, Student Member, IEEE, Hao Shen, Student Member, IEEE, and Minghao Qi, Member, IEEE Abstract We demonstrate highly compact optical add drop filters based on silicon-on-insulator microring resonators. The microring resonators have a small radius of 2.5 m and a very large free spectral range 32 nm at the 1.55 m communication band. The propagation loss in such small micoring resonators was experimentally determined and shown to be extremely important in designing microring add drop filters with low add drop crosstalk, low drop loss, and maximally flat drop passband. For box-like channel dropping responses, second-order optical add drop filters with two coupled microring resonators are designed and demonstrated, and the simulation matches well with the experiment. Devices were patterned with electron-beam lithography. Two fabrication procedures utilizing different polarity of resists were introduced and compared, and the process with negative resist resulted in much smaller sidewall roughness of waveguides, thus reducing the propagation loss in microring resonators. Index Terms Electron-beam lithography, integrated optical devices, microring add drop filters, microring resonators, microstructure fabrication, photonic integrated circuits, silicon-on-insulator. I. INTRODUCTION SILICON-ON-INSULATOR (SOI) has become an attractive technology to enable the miniaturization of photonic integrated circuits down to micrometer length scale, with submicron cross-sections for individual waveguides. In SOI configuration, high-index-contrast between the core (silicon) and the bottom cladding material (silicon dioxide) enables sub-wavelength light confinement in the core with very little optical leakage into the silicon substrate, thus making SOI suitable for highly integrated photonic devices with very sharp bends, e.g., microring add-drop filters [1] [3], where the bending radius can be on the order of several micrometers. SOI microring add-drop filters are promising for WDM signal processing in a chip. The miniaturization of filtering devices may reduce the cost and the complexity of WDM systems, and SOI devices can benefit from low-cost fabrication using well-developed CMOS technologies Manuscript received January 31, 2007; revised August 4, 2007. This work was supported in part by a grant from the Defense Threat Reduction Agency (DTRA) under Contract HDTRA1-07-C-0042 and in part by the National Science Foundation under Contract ECCS-0701448. S. Xiao was with the School of Electrical and Computer Engineering and the Birck Nanotechnology Center, Purdue University, West Lafayette, IN 47907 USA. He is now with the National Institute of Standards and Technology, Boulder, CO 80305 USA (e-mail: sxiao@boulder.nist.gov). M. H. Khan, H. Shen, and M. Qi are with the Birck Nanotechnology Center, Purdue University, West Lafayette, IN 47907 USA (e-mail: mhkhan@purdue. edu; shen17@purdue.edu; mqi@purdue.edu). Digital Object Identifier 10.1109/JLT.2007.911098 [4]. Unfortunately, large free spectral range (FSR), box-like response with maximally flat passband and fast filtering roll-off, and high out-of-band signal rejection (filtering contrast) are still major challenges for SOI microring optical add-drop filters. Recently, a group reported microring add-drop filters both in silicon nitride (SiN) [5], [6] and in silicon [7]. However, their reported micoring resonators have a FSR of 20 nm (microring s radius m) and 16 nm (microring s radius m) for SiN devices and Si devices, respectively. In order to achieve a large FSR over 30 nm for a sufficiently large spectral window covering the entire C band 1565 nm) supported by matured erbium-doped fiber amplification technology, the microring s radius should be less than 3 m [1]. In this paper, we demonstrate a very large nm at 1.55 m in SOI microring add-drop filters, and the microring s radius is 2.5 m. Nevertheless, the propagation loss in small microring resonators can be large, which is attributed to the bending loss (the major one) as well as sidewall roughness scattering [8]. This nontrivial propagation loss has increased significantly the complexity to design microring add-drop filters with maximally flat channel dropping passband, low drop loss and low add-drop crosstalk, and it also sets a lower limit on the bandwidth achievable in these filters. Low-loss microring resonator is therefore the key to implement high-performance add-drop filters. In this work, the propagation loss in microring resonators is minimized significantly by increasing the waveguide width by only 100 150 nm. The propagation loss in microring resonators is analyzed with a relatively new method reported by us recently [9]. For low add-drop crosstalk and low channel dropping loss, a complete design of SOI microring add-drop filters is verified with detailed experimental results. For box-like responses with flat passband, a second-order add-drop filter with two mutually coupled microring resonators was designed, fabricated and characterized. Although the loss in microring resonators is optimized, it is still nontrivial, and the condition for flat passband in add-drop filters is then modified compared to that with the loss neglected [10]. With simulation, we illustrate the significant effect of the nontrivial propagation loss as well as microrings mutual coupling in the design of second-order add-drop filters. The second-order filter s design was verified and demonstrated with our detailed experimental results. All simulations agree well with experiments. Compared to the majority of previously reported microring add-drop filters, our demonstrated microring add-drop filters have the smallest footprint for integration and the largest FSR covering almost the whole spectral window of C band. By incorporating this work with our recently reported work on multiple-channel filters [11], we believe that 0733-8724/$25.00 2008 IEEE

XIAO et al.: SILICON-ON-INSULATOR MICRORING ADD-DROP FILTERS WITH FREE SPECTRAL RANGES OVER 30 NM 229 these demonstrated results here are very useful to implement high-performance multiple-channel filters that are truly compatible with WDM systems for multiple users. All microring add-drop filters were patterned with the Vistec (formerly Leica) VB6 electron-beam lithography (EBL) system installed in the Birck Nanotechnology Center at Purdue University. A significant advantage of this state-of-the-art EBL tool is the high acceleration voltage (100 kv) that mitigates electron-beam proximity effects [5]. With a beam spot diameter of around 6 nm according to specifications, we used a beam deflection step of 2 nm to reduce digitization errors discussed in previous reports [5]. Two different fabrication strategies were investigated and compared. In contrast to the conventional patterning method, which uses a positive-tone electron-beam resist [5], a relatively new procedure, based on a negative-tone resist, is superior in providing much smoother device sidewalls, which reduces the process complexity and provides much faster device turn-around time. The smoother waveguide sidewall reduces significantly the sidewall roughness scattering loss. The remainder of this paper is structured as the following. Section II discusses and compares two different fabrication processes for SOI devices. Section III presents basic theories to design microring add-drop filters and extract the propagation loss in microring resonators. Section IV presents our experimental work on the optimization of the propagation loss in microring resonators. We present both simulation and experiment for the synthesis of second-order microring add-drop filters in Section V. Section VI is the conclusion. Fig. 1. Schematic of the PMMA-based fabrication procedure. (a) structural configuration of the sample, (b) electron-beam exposure and development, (c) chromium deposition via electron beam evaporation, (d) lift-off by immersing the wafer in a solvent that dissolves undeveloped PMMA and a complementary pattern is transferred into the chromium layer, (e) reactive-ion etching of silicon, (f) removal of chromium mask layer via wet etch. II. DEVICE FABRICATION The basic structure of SOI waveguides consists of three layers from bottom to top: the silicon substrate, the buried oxide layer and the silicon core (Fig. 1(a)). The oxide layer is 3 m thick in our case, which is enough to suppress the optical power leakage from the silicon core to the substrate. The height of the waveguide core is fixed at 250 nm in our case, which is also the thickness of the top silicon layer of our SOI wafers. The waveguide width can be adjusted via lithography. The waveguide is approximately of single mode (the lowest TE) at 1.55 m for a width up to 600 nm in microring resonators. The lowest TM and higher order modes have a much larger loss and resonate at different wavelengths in resonators. Two strategies were applied for the device fabrication. One was with the positive resist poly-methyl-methacrylate (PMMA) and the other one was with the negative resist hydrogen silsesquioxane (HSQ). Fig. 1(a) (f) shows the flow for the PMMA process. Our devices were fabricated on silicon-on-insulator (SOI) wafers (from SOITEC). The device patterns were exposed in a 200 nm-thick PMMA with the 100 kv EBL system. The high accelerating voltage of 100 kv significantly reduces waveguide width dependence on whether another feature is present at close proximity. Such width variations are especially prominent at the coupling regions between resonators and bus waveguides, leading to smaller gaps and wider waveguides. In EBL, the primary electron beam excites the substrate and generates secondary electrons. Unfortunately, such electrons are generated in an area that is much larger than Fig. 2. Schematic of the HSQ-based fabrication procedure. (a) structural configuration of the sample; (b) electron-beam exposure followed by resist development, (c) reactive-ion etching of Si in a Chlorine based plasma. (d) Remove HSQ mask layer in HF solution (Optional). the spot of the primary beam, and contribute to the exposure of resist. Thus the total dose the resist receives at a specific location is also dependent on the exposed pattern in close proximity to itself. This effect is called proximity effect [5]. Thus, a negative bias is typically applied to the pattern layout in order to pre-compensate the proximity effect. At high accelerating voltages, the range of the secondary electrons are large, up to 50 m, therefore providing a uniform background dose that are less sensitive to the local patterns. This advantage helped to eliminate the requirement of pre-compensation of waveguide width in our exposure. After the exposed resist was developed, we deposited a 40 nm-thick chromium layer with electron-beam evaporation. This was followed by a lift-off procedure to remove the unexposed PMMA and to leave the chromium pattern as a mask for the etching of silicon. We used gas at a pressure of 35 mtorr in a PlasmaLab reactive-ion

230 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 26, NO. 2, JANUARY 15, 2008 Fig. 3. Scanning electron micrograph of ring waveguides fabricated with the PMMA (top) and the HSQ (bottom) respectively. For the HSQ sample (bottom), the HSQ mask layer was removed with wet etch for micrograph purpose only. Clearly, the HSQ process yields much smoother waveguides. etch (RIE) tool to etch through the 250 nm silicon layer. Finally, the chromium mask was removed via wet chemical etch. Fig. 2(a) (d) shows the flow of fabrication with the HSQ on the same type of SOI wafers. The same device patterns were exposed in a 200 nm-thick HSQ with the 100 kv EBL system. Still, no pre-compensation of waveguide width was applied. In contrast to the PMMA-based process, the developed HSQ pattern is a good etch mask without further need of pattern transfer through the lift-off procedure. This simplified process reduces the waveguide roughness caused by the lift-off. Then, we used inductively-coupled-plasma (ICP) reactive-ion-etch (RIE) to etch through the 250 nm silicon layer. The chamber pressure was 5 mtorr and the gases were and Ar with flow rates of 15 sccm and 2.5 sccm, respectively. The exposed HSQ pattern can be removed in a diluted HF solution for a very short time period. However, the HF solution also etches slightly the bottom cladding oxide layer. The HSQ pattern can also be annealed at high temperature (1100 1200 ) to convert to an oxide layer. As the HSQ has a refractive index and a very low absorption loss at 1.55 m bands [12], it doesn t affect the optical performance appreciably in high-index-contrast silicon waveguides. Therefore, it was kept intact as a top cladding layer in our case for device characterization. Fig. 3 shows two scanning-electron micrographs of fabricated microring waveguides with the PMMA (the top picture) and HSQ (the bottom picture) respectively. The sidewall roughness is noticeably reduced on the waveguide fabricated with the HSQ process. Thus, the HSQ process was adopted to fabricate the microring add-drop filters for optical characterization. III. DESIGN OF MICRORING ADD-DROP FILTERS Fig. 4 shows the basic configuration (top) of an add-drop filter based on a microring resonator with its schematic responses (bottom). In a symmetrically coupled microring resonator, two bus waveguides have the same width and the same coupling distance to the microring resonator. is the resonator waveguide s width, which may be different from. is the average radius of the microring. is the power coupling coefficient between the bus waveguide and the resonator, which depends exponentially on the coupling gap, and is the propagation power loss coefficient per round-trip in the Fig. 4. Theoretical model (top) and schematic power responses (bottom) of a symmetrically coupled microring add-drop filter (first order). microring resonator. According to the theory developed in [9], [10], close to each center resonance wavelength, power transmission responses of the through port and the drop port are written respectively as (1.a) (1.b) where FSR is the free spectral range of the resonator s response, and is the center resonance wavelength. Equation (1.b) indicates a first-order Butterworth filter. The add-drop crosstalk is defined as at, which is expressed by in db. The drop loss is defined as at, which is expressed by in db. The through extinction is defined as at, which is expressed by in db. For low add-drop crosstalk and low drop loss, the waveguide power coupling coefficient must be much larger than the propagation power loss coefficient, i.e.,. As there is always a nonzero propagation loss in practical devices, for sufficiently large waveguide coupling to achieve low add-drop crosstalk and low drop loss, the gap between the bus waveguide and the microring resonator can t be too large due to the fact of [11], where is a decaying coefficient. According to (1.b), the channel dropping db bandwidth is db (2) With constraints for low add-drop crosstalk and low drop loss, the channel dropping db bandwidth is mainly limited by the waveguide coupling coefficient. To minimize the drop db bandwidth for high spectral resolution, it is very important to minimize the propagation loss in the resonator so that weaker waveguide coupling can be used. With FSR,

XIAO et al.: SILICON-ON-INSULATOR MICRORING ADD-DROP FILTERS WITH FREE SPECTRAL RANGES OVER 30 NM 231 Fig. 5. Scanning-electron micrographs of one fabricated microring add-drop filter., and db, and can be calculated by and [9]. For microring resonators, we have or, where at 1.55 mis the group index in SOI waveguides with core cross section of nm nm and a top HSQ mask layer of 200 nm. For a large FSR over 30 nm or larger to cover a sufficiently large optical spectrum window supported by erbium-doped fiber amplifiers, the radius of SOI microring should be mor smaller. To our best knowledge, there was few reported work on microring add-drop filters with a large FSR over 30 nm as the relatively large propagation loss (we will show in the next section) in such small microring resonators might have been the bottleneck. In this paper, we optimize the filtering performance of SOI microrings with a radius of 2.5 m. The propagatin loss in such small microring resonators can be significantly reduced by increasing the waveguide width over a range of 100 150 nm while the height remains constant. IV. PROPAGATION LOSSES IN MICRORING RESONATORS Fig. 5 is a scanning-electron micrograph of a fabricated microring resonator in add-drop configuration. The microring radius is 2.5 m, and the waveguide s width is nm. The bus waveguide s width is nm, and the gap between the bus waveguide and the microring waveguide is nm. We used a tunable laser source to characterize responses of microring resonators. The laser was coupled into the bus waveguide through a lensed single mode fiber tip mounted on an XYZ nanopositioning stage. A fiber-based polarization controller was used at the input to adjust the polarization to excite the fundamental TE mode with the lowest propagation loss. The output was collected by another lensed single mode fiber tip mounted on a second XYZ nanopositioning stage and then fed into an optical power meter. To obtain a power transmission spectrum, the laser wavelength was swept at a step continuously while the output power of each wavelength was recorded sequentially. Fig. 6 plots the measured responses of the microring adddrop filter shown in Fig. 5, and Fig. 6(b) is a zoom-in view of Fig. 6(a) at one resonance wavelength. For comparison, theoretical curves (solid lines) are also plotted in Fig. 6(b) according to the discussion below. The maximum filtering contrast (out of band signal rejection) is db. The measured channel dropping loss is db, and the measured add-drop crosstalk is db. For the resonance m, the minimum transmission of the through-port is db Fig. 6. (a) Measured add-drop responses of the fabricated first-order microring resonator. The waveguide coupling gap is 200 nm. (b) A zoom-in view with simulated responses plotted in solid line. The simulation is based on (1.a) and (1.b). The drop-port response is normalized to the through-port response by setting the maximum through-port transmission at 0 db when not at the resonance. (through-port extinction db). The channel dropping db bandwidth is nm, and the FSR is nm. The extracted is, and the extracted is. The theoretical drop loss is db, and the theoretical add-drop cross-talk is db. The small discrepancy between the theoretical result and the experimental result could be due to slightly different fiber-to-waveguide couplings or waveguide propagation losses between through-port and drop-port during measurement, and this explains why there is a shift of 1 2 db in the vertical axis between theoretical curve and experimental curve for the drop-port s response in Fig. 6(b). The propagation loss is or db/round-trip, and the corresponding intrinsic quality-factor is (the maximum quality-factor that can be achieved). We also fabricated and tested microring add-drop filters with different ring width of 550, 500 and 450 nm, and we found that the propagation loss increases dramatically as the ring width is reduced. For resonance wavelengths m, the extracted propagation losses are, and for of 550, 500 and 450 nm respectively. The waveguide power coupling coefficient doesn t vary much compared to that for nm, e.g., for nm. Fig. 7 plots the propagation loss in microring resonators m for different of 450, 500, 550 and 600 nm. The four data points are connected with lines to guide the

232 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 26, NO. 2, JANUARY 15, 2008 Fig. 7. Propagation losses in microring resonators for different waveguide width (height 250 nm). The ring radius is 2.5 m. Fig. 9. Schematic modeling of a high-order microring add-drop filter with series coupled resonators in parallel. Fig. 8. Measured add-drop responses of one fabricated first-order microring resonator where the waveguide coupling gap is 100 nm. The drop-port response is normalized to the through-port response. eye, and the propagation loss decays slowly and appears to become approximately constant for nm. The insets in parenthesis show the loss per round-trip and the corresponding intrinsic quality-factor for each data point. For a low add-drop crosstalk db, Fig. 8 plots the add-drop response of another fabricated microring resonator, where the coupling gap is nm for a large waveguide coupling (extracted value is for wavelengths m). The microring width is nm. The drop db bandwidth is nm, and the measured add-drop crosstalk is db. Equation (1) predicts that the through-port extinction will increase to 24 db, which is verified by the measurement shown in Fig. 8. V. HIGH-ORDER MICRORING ADD-DROP FILTERS For WDM signal processing, it is desirable to have box-like responses with flat passband and fast rolling-off from passband to stopband. For this purpose, a high-order add-drop filter consisting of multiple coupled microring resonator should be used. Fig. 9 shows a schematic configuration of an th order filter. All microring resonators are identical and have center resonance wavelengths, and respectively. Waveguide coupling gaps and corresponding power coupling coefficients are represented by parameter sets. For symmetrically coupled add-drop filters, we have,, and In case of second-order filters with maximally flat drop passband (Butterworth filter), if the propagation loss in resonators is negligible [10], and must meet the condition of.however, the nontrivial propagation loss in resonators can affect this condition, and to our best knowledge, this has not been reported or discussed before. We have derived that the required condition is modified to.if holds in a second-order filter, the power transmission of the through-port and the drop-port, in the vicinity of center resonance wavelength, are derived to be (3.a) (3.b) at the bottom of the page, where we assume. According to (3.b), the drop db bandwidth is db (4) The add-drop crosstalk at is expressed by in db, and the drop loss at is expressed by in db. Theoretically, compared to the single microring add-drop filter, assuming the same, and, the second-order filter has 6 db increase in add-drop crosstalk but smaller drop db bandwidth. The channel dropping loss is approximately (3.a) (3.b)

XIAO et al.: SILICON-ON-INSULATOR MICRORING ADD-DROP FILTERS WITH FREE SPECTRAL RANGES OVER 30 NM 233 Fig. 10. The filtering effect of the mutual coupling coefficient between two resonators in second-order add-drop filter. The inset shows the coupling coefficients,, and. The propagation loss is neglected =0. The FSR is 30 nm in simulation. Fig. 11. The filtering effect of the propagation loss of resonators in second-order add-drop filter. The inset shows the coupling coefficients, and.the FSR is 30 nm in simulation. negligible if is satisfied in both single microring filters and second-order microring filters. Additionally, it is worthwhile to understand the effect of different mutual power coupling coefficients in second-order filters. Fig. 10(a) (b) plot simulated power transmission responses of through-port and drop-port respectively. In this simulation, the free spectral range is set at 30 nm. The waveguide power coupling is fixed, and the propagation loss is neglected in order to see the effect of mutual power coupling. We simulate three values of, which are 0.01, (maximally flat passband condition) and 0.001. According to our simulation, in addition to maximally flat passband, the case with also yields the lowest add-drop crosstalk at the center passband. With the propagation loss included, the filter design becomes more complex. Fig. 11(a) (b) plot simulated power transmission responses of through-port and drop-port respectively. The waveguide power coupling is kept the same. For reference, the ideal case with and is also plotted. Then, we use to be comparable to the waveguide coupling coefficient. We simulated two cases, which are for (ideal maximally flat passband condition) and (modified maximally flat passband condition) respectively. The simulation verifies that yields a flat passband but higher add-drop crosstalk at the center of the passband. The nontrivial propagation loss causes an obvious channel dropping loss. In order to achieve maximally flat passband, low channel dropping loss and low add-drop crosstalk simultaneously, the waveguide power coupling coefficient should be much larger than the propagation loss coefficient, i.e.,, thus,. On the other hand, there may also be channel dropping bandwidth constraints for DWDM applications, and and can t be too large. Therefore, low-loss microring resonators are very important in implementing high-performance add-drop filters for high spectral resolution WDM signal processing. Although third or higher order filters can yield channel dropping responses with faster rolling-off, all coupled resonators don t have the same center resonance wavelength due to the coupling induced frequency shift (CIFS) [13]. For example, in a third-order filter considered for maximally flat passband, e.g., for [11], the middle resonator sees different power couplings compared to the other two resonators close to bus waveguides, and the coupling variation results in different center resonance wavelengths between the middle resonator and the other two resonators. Compared to third-order filters, second-order filters have only two coupled resonators that have the same center resonance wavelength in design, thus are immune to CIFS. Here, we illustrate a detailed design for a second-order microring add-drop filter. According to our discussions above, for low add-drop crosstalk db at center passband, should

234 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 26, NO. 2, JANUARY 15, 2008 Fig. 12. Scanning-electron micrographs of fabricated second-order microring add-drop filters. Fig. 13. Measured responses in the C band of the fabricated second-order microring add-drop filter. The drop-port response is normalized to the through-port response. hold. We have obtained in microring resonator with nm, so we need. For this purpose, the coupling gap between the bus waveguide and the microring resonator is set to be 100 nm. In our design, the bus waveguide width is fixed at 500 nm. For maximally flat channel dropping passband, the resonator s mutual power coupling coefficient should be.for coupled waveguides with fixed width, the waveguide power coupling coefficient is described by the following relation [11] where and are bending radii of two coupled waveguides. The radius of a straight waveguide is taken as infinity. If and are both fixed, the power coupling coefficient is an equal ratio series for a uniform variation step of coupling gaps. For nm and m nm, we have obtained at nm and at nm. We design nm (the gap between two coupled resonators) for to approach the desired value 0.0106. Fig. 12 shows scanning-electron micrographs of one fabricated second-order microring add-drop filter and its waveguide coupling region. The coupling gaps are calibrated to be nm and nm. The waveguide widths are calibrated to be nm and nm. Although our symmetrical design eliminates mismatched center resonance wavelength, it is inevitable that the fabrication has imperfections, e.g., waveguide dimension variation (several nm) due to EBL beam deflection error, etc. These small fabrication defects can result in different (5) Fig. 14. Experimental (dotted and dashed lines) and simulated responses (solid lines) of the fabricated second-order microring add-drop filter. The experimental response is compared with the simulation for two resonance matched resonators (a) and for two resonators with a wavelength mismatch of 1 =0:1nm (b). center resonance wavelength for two resonators as resonance is phase sensitive. This point will be illustrated here with both experiment and simulation. Fig. 13 plots measured add-drop response over a broad spectrum covering C band. The filter shows a large nm and a high out-of-band signal rejection of 40 db. Fig. 14 plots zoom-in view of measured response at m compared with simulation, which is repeated in both Fig. 14(a) and (b) in order to be compared with two different cases of simulation. We demonstrate an add-drop crosstalk of db at center passband and an extremely low drop loss db. The drop db bandwidth is nm, the drop db bandwidth is nm, and the drop db bandwidth is nm. The channel dropping response has a flat passband and fast rolling-off. For the simulated response plotted in Fig. 14(a), we used,, and as they were designed. We achieved a very good match between the design and the experiment. For the simulated result plotted in Fig. 14(b), to match with measured responses (especially the through-port response) as close as possible, we used,, (slightly different from 0.0106) and center resonance wavelength mismatch nm. In both cases, the channel dropping response matches well with our simulation,

XIAO et al.: SILICON-ON-INSULATOR MICRORING ADD-DROP FILTERS WITH FREE SPECTRAL RANGES OVER 30 NM 235 (A1) (A2) and we have achieved approximately maximally flat passband with carefully designed waveguide couplings, i.e., holds in the fabricated filter. The through-port response is very sensitive to mismatched center resonance wavelength, and our simulation indicates 0.1 nm shift of center resonance wavelength between two microring resonators in the fabricated second-order filter. VI. CONCLUSION We have demonstrated ultracompact SOI microring add-drop filters with an average radius of 2.5 m and a large free spectral range around 32 nm at around 1.55 m. The propagation losses in such small microring resonators were extracted and found to be very large for ring waveguide width of around 450-500 nm, but reduced significantly for waveguide width of around 550-600 nm. A propagation loss of was demonstrated in the microring resonator with ring width of 600 nm (core height 250 nm), and is believed to be the lowest and first reported number in microring resonators with a small bending radius of 2.5 m. Ultracompact second-order microring add-drop filters were designed for maximally flat passband in channel dropping. In contrast to previous reports that ignored the propagation loss in microring resonators, we present the theoretical add-drop response with consideration of a nontrivial propagation loss in small microring resonators and modified condition for maximally flat passband. The fabricated filter was characterized and demonstrated with a very flat passband as designed and a fast rolling-off in the drop-port response, and, at the center passband, we show low add-drop crosstalk db, very low channel dropping loss db and high out-of-band signal rejection of around 30-40 db. We also point out that fabrication imperfections can cause small mismatched resonance between two resonators, which results in an asymmetric response curve of the through-port but no obvious effects on the drop-port (as long as the mismatch is small in our case). The drop-port response is predicted well with our theories even if this mismatch is ignored in our simulation. By including this resonance wavelength shift in our simulation, we have demonstrated well matched responses between the simulation and the experiment for the through-port response. APPENDIX In general cases, the add-drop response of a second-order filter are written as see equation (A1) (A2) at the top of the page. REFERENCES [1] B. E. Little, J. S. Foresi, G. Steinmeyer, E. R. Thoen, S. T. Chu, H. A. Haus, E. P. Ippen, L. C. Kimerling, and W. Greene, Ultra-compact Si-SiO microring resonator optical channel dropping filters, IEEE Photon. Technol. Lett., vol. 10, no. 4, pp. 549 551, Apr. 1998. [2] A. Vörckel, M. Mönster, W. Henschel, P. H. Bolivar, and H. Kurz, Asymmetrically coupled silicon-on-insulator microring resonators for compact add-drop mutiplexers, IEEE Photon. Technol. Lett., vol. 15, no. 7, pp. 921 923, July 2003. [3] P. Dumon, W. Bogaerts, V. Wiaux, J. Wouters, S. Beckx, J. V. Campenhout, D. Taillaert, B. Luyssaert, P. Bienstman, and D. V. T. R. Baets, Low loss SOI photonic wires and ring resonators fabricated with deep UV lithography, IEEE Photon. Technol. Lett., vol. 16, no. 5, pp. 1328 1330, May 2004. [4] Y. A. Vlasov and S. J. McNab, Losses in single-mode silicon-on-insulator strip waveguides and bends, Opt. Express, vol. 12, pp. 1622 1631, 2004. [5] T. Barwicz, M. A. Popovic, M. R. Watts, P. T. Rakich, E. P. Ippen, and H. I. Smith, Fabrication of add-drop filters based on frequencymatched microring resonators, J. Lightw. Technol., vol. 24, no. 5, pp. 2207 2218, May 2006. [6] M. A. Popovíc, T. Barwicz, M. R. Watts, P. T. Rakich, L. Socci, E. P. Ippen, F. X. Kärtner, and H. I. Smith, Multistage high-order microring-resonator add-drop filters, Opt. Lett., vol. 31, pp. 2571 2573, 2006. [7] M. A. Popovíc, T. Barwicz, F. Gan, M. S. Dahlem, C. W. Holzwarth, P. T. Rakich, H. I. Smith, E. P. Ippen, and F. X. Kärtner, Transparent wavelength switching of resonators, in Proc. CPDA2, CLEO 2007, Baltimore, MD. [8] K. K. Lee, D. R. Lim, and L. C. Kimerling, Fabrication of ultralowloss Si/SiO2 waveguides by roughness reduction, Opt. Lett., vol. 26, pp. 1888 1890, 2001. [9] S. Xiao, M. H. Khan, H. Shen, and M. Qi, Modeling and measurements of losses in silicon-on-insulator resonators and bends, Opt. Exp., vol. 15, no. 12, pp. 10553 10561, 2007. [10] B. E. Little, S. T. Chu, H. A. Haus, J. Foresi, and J. P. Laine, Microring resonator channel dropping filters, J. Lightw. Technol., vol. 15, no. 6, pp. 998 1005, Jun. 1997. [11] S. Xiao, M. H. Khan, H. Shen, and M. Qi, Multiple-channel silicon micro-resonator based filters for WDM applications, Optics Express, vol. 15, no. 12, pp. 7489 7498, 2007. [12] C. W. Holzwarth, T. Barwicz, and H. I. Smith, Optimization of HSQ films for photonic applications, in Proc. EIPBN 2007, Denver, CO. [13] M. A. Popovíc et al., Coupling-induced resonance frequency shifts in coupled dielectric multi-cavity filters, Opt. Exp., vol. 14, pp. 1208 1222, 2006. Shijun Xiao (S 03 M 05) was born in Chengdu, China, in 1979. He received the B.S. degree in electronics from Beijing University, Beijing, China, in 2001 and both the M.S. degree and the Ph.D. degree in electrical and computer engineering from Purdue University, West Lafayette, IN, in 2003 and in 2005, respectively. From 2005 to 2007, Dr. Xiao was a Postdoctoral Research Associate in electrical and computer engineering at Birck Nanotechnology Center of Purdue University, West Lafayette, IN He is currently a research scientist at National Institute of Standards and Technology (NIST), Boulder, CO, where he is doing research on optical frequency comb technology for telecom applications. He has been the author or co-author of nearly 50 journal articles and conference papers.

236 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 26, NO. 2, JANUARY 15, 2008 His research interests include optical frequency comb, silicon photonics, optical signal processing, optical pulse shaping, optical fiber communications, microwave photonics. His Ph.D. research was focused on virtually imaged phased array and its novel application in optical signal processing and optical fiber communication with a sub-gigahertz channel resolution. His postdoctoral research was focused on microring resonators and filters on silicon-on-insulator platform, and he was experienced in microring device modeling, fabrication and characterization Dr. Xiao is a member of the IEEE, the IEEE Lasers and Electro-Optics Society (LEOS) and the Optical Society of America (OSA). He has served as a reviewer of IEEE/OSA Journal of Lightwave Technology, IEEE Photonics Technology Letters, OSA Optics Letters, OSA Optics Express and Optics Communications. He was also a reviewer for International Association of Science and Technology for Development Antennas, Radar, and Wave Propagation, 2007. He received the Andrews Fellowship from Purdue University for two years in 2001 2003. He received the IEEE LEOS Graduate Student Fellowship in 2004. He was a Finalist for the Dimitris N. Chorafas Foundation Award of Purdue University in recognition of top Ph.D. research in 2005. Maroof H. Khan (S XX) photograph and biography not available at the time of publication. Hao Shen (S XX) photograph and biography not available at the time of publication. Minghao Qi (M XX) photograph and biography not available at the time of publication.