Photonic bandgap crystal resonator enhanced, laser controlled modulations of optical interconnects for photonic integrated circuits

Transcription

1 Photonic bandgap crystal resonator enhanced, laser controlled modulations of optical interconnects for photonic integrated circuits Selin H. G. Teo 1*, A. Q. Liu 2, J. B. Zhang 3, M. H. Hong 3, J. Singh 1, M. B. Yu 1, N. Singh 1, G. Q. Lo 1 1 Institute of Microelectronics, 11 Science Park Road, Singapore teohg@ime.a-star.edu.sg 2 School of Electrical & Electronic Engineering, Nanyang Technological University, Singapore Data Storage Institute, 5 Engineering Drive 1, Singapore Abstract: Ultrafast high-density photonic integrated circuit devices (PICDs) are not easily obtained using traditional index-guiding mechanisms. In this paper, photonic bandgap crystal resonator enhanced, laser-controlled modulations of optical interconnect PICDs were achieved in slab-type mix-guiding configuration - through developed CMOS-compatible processing technologies. The devices, with smallest critical dimensions of 90nm have footprints of less than 5 5μm 2. Quality-factors an order larger than previously realized was achieved. Through use of effective coupling structures; simultaneous alignment for probing and pumping laser beams, optical measurements of both instantaneous free carriers induced device modulations were obtained together with thermo-optical effects characterizations Optical Society of America OCIS codes: ( ) Nanophotonics and photonic crystals; ( ) Optical interconnects; ( ) All-optical devices; References and links 1. E. Yablonovitch, Inhibited spontaneous emission in solid-state physics and electronics, Phys. Rev. Lett. 58, (1987). 2. S. John, Strong localization of photons in certain disordered dielectric superlattices, Phys. Rev. Lett. 58, (1987). 3. S. Noda, and T. Baba, Roadmap on photonic crystal, Kluwer Academic Publisher, London, S. G. Johnson, C. Manolatou, S. Fan, P. R. Villeneuve, J. D. Joannopoulos, and H. A. Haus, Elimination of cross talk in waveguide intersections, Opt. Lett. 23, (1998). 5. H. G. Teo, A. Q. Liu, J. Singh, M. B. Yu, and T. Bourouina, Design and simulation of MEMS optical switch using photonic bandgap crystal, Microsys. Tech. 10, (2004). 6. M. F. Yanik, S. Fan, M. Soljacic, and J. D. Joannopoulos, All-optical transistor action with bistable switching in a photonic crystal cross-waveguide geometry, Opt. Lett. 28, (2003). 7. H. Nemec, P. Kuzel, L. Duvillaret, A. Pashkin, M. Dressel, and M. T. Sebastian, Highly tunable photonic crystal filter for the terahertz range, Opt. Lett. 30, (2005). 8. M. J. Escuti, J. Qi, and G. P. Crawford, Tunable face-centered-cubic photonic crystal formed in holographic polymer dispersed liquid crystals, Opt. Lett. 28, (2003). 9. K. Umemori, Y. Kanamori, and K. Hane, Photonic crystal waveguide switch with a microelectromechanical actuator, Appl. Phys. Lett. 89, (2006). 10. S. Leonard, W. Van Driel, M. Schilling, and J. Wehrspohn, Ultrafast band-edge tuning of a twodimensional silicon photonic crystal via free-carrier injection, Phys. Rev. B 66, (2002). 11. Selin H. G. Teo, A. Q. Liu, J. B. Zhang, and M. H. Hong, Induced free carriers modulation of photonic crystal optical intersection via localized optical absorption effect, Appl. Phys. Lett. 89, (2006). 12. Selin H. G. Teo, A. Q. Liu, J. Singh, and M. B. Yu, High resolution and aspect ratio two-dimensional photonic band-gap crystal, J. Vac. Sci. Technol. B 22, (2004). 13. Selin H. G. Teo, A. Q. Liu, M. B. Yu, and J. Singh, Synthesized Processing Techniques for Monolithic Integration of Nanometers-Scale Hole Type Photonic Bandgap Crystal with Micrometers-Scale Microelectromechanical Structures, J. Vac. Sci. Technol. B 24, (2006).

2 1. Introduction Two decades after the first proposal of the photonic crystals (PhCs) bandgap effect [1, 2], many important integrated photonic circuit devices have been designed and proposed [3] based on the unique advantages PhCs possess for precise photonic dispersion control and compact device sizes - which scales directly with the wavelength of application. One of which, particularly important for photonic integrated circuit devices (PICDs), is the application of orthogonally intersecting optical waveguides to make PhCs optical interconnects [4, 5]. Such optical interconnect devices may be used to make the smallest size right-angle circuit routing connectors, which with the addition of active control functionality [6], can be further applied to applications requiring signal processing and optical routings at such junctions. For PICDs with the above-mentioned active control functionality, much effort had been expended internationally in exploration of several key methodologies to affect the material systems of the PhC devices. Some of the most typical examples of which are, through thermo actuation effects [7], liquid crystal infiltration tuning [8], and micro-mechanical actuation circuits [9] etc. Of which, the method of material excitation via light injection mechanisms [10] is one of the most promising due to its potential for fastest control, since its speed is inprinciple, limited only by the duration of carriers relaxation. However, in silicon (the established integrated circuit material of choice), such desired dynamic control of device property is not straightforward since silicon does not possess strong nonlinearities necessary for most common active control methods. At the same time, the distinctive non-interfering property of light, coupled with its typically high group velocities also hinders effective lightmaterial interactions [2] (critical for efficient properties manipulations). Such limitations are minimized with implementation of devices designed based on the PhCs effect; since group velocity of light may be effectively reduced in PhC structures to enhance light-material interactions. In this letter, PhC optical interconnects comprising of orthogonally intersecting single-line defect waveguides of widths 0.8 μm, incorporated with partially reflecting PhC micro-mirrors, together with a small volume optical resonator, were designed with tailored photonic dispersion properties and realized in both high aspect ratio and also in slab-type mix guiding configuration for optical waveguiding at 1.55 μm centered at the canonical optical fiber communications wavelengths. With smallest critical dimensions of 90 nm and device footprints of less than 5 by 5 μm 2, integrated optical coupling structures were designed and incorporated to enable self-aligned optical measurements. Base on design of higher-order resonator confinement factors, PhC resonators with enhanced laser-controlled device modulations were achieved with cavity sensitive optical intersection devices (Fig. 1). The paper is organized as follows, first, the mechanisms for quality-factor enhancement in the PhC optical interconnects that are responsible for cavity sensitive optical interconnect control operations are presented. Following which, the fabrication techniques required to obtain the sub-100 nm critical dimension structures are described. This is then followed by the results of the optical experiments, which are then analyzed, with respect to expected simulation results, and discussed. This paper is then concluded with an overview of the key findings together their potential applications. 2. Design of resonator enhanced PhC optical interconnects The design of the PhC optical intersection device with center resonant cavity rod sensitive properties is as shown in Fig. 1(a). Here, the use of a pair of degenerate modes at the resonator (right inset) to cancel out unwanted crosstalk in orthogonal waveguide branches is similar to that previously reported [11]; except that in the present case, higher order resonators - as shown in the schematic of Fig. 1(b) - were created through addition of partially reflecting PhC micro-mirrors at each branch of the intersection [3] for progressively higher order resonators in configurations of 1 1; 3 3; and for nil-, one-, and two- partially

3 reflecting PhC rods at each waveguide branch respectively. The frequency responses of these various orders of resonators are as calculated by finite difference time domain (FDTD) method and shown on the right inset of Fig. 1(b). From the spectrums, it can be seen that higher order of resonator configurations indeed gave higher quality-factors in Lorentzian peaks - as expected based on analogous Fabry-Perot cavity principles. At the same time, in place of using only PhC waveguiding in very high aspect-ratio quasi-2d PhC rod lattices [11], index-guiding mechanism is also introduced in the present case for out-of-plane waveguiding - through introduction of reduced-radii defect rods along previously empty defect waveguides (where rods along waveguide path in the PhC lattice were completely removed). Such reduced-radii PhC rods waveguide can be shown, to give single-mode guiding within the span of the photonic bandgap frequencies (generated by the bulk PhC lattice) through planewave-expansion mode calculations. Also, instead of straight-through optical paths, each input waveguiding path is now isolated from the output measurement ports, so as to avoid inclusion of leakage measurements achieved through addition of optimized tapers and waveguide bends. Output coupling optical fiber z y x Input coupling optical fiber 1.0 Buried oxide cladding layer x z y Handle (substrate) silicon PhC (device) silicon Deeply etched fiber grooves -1.0 (a) (b)

4 Fig. 1. (a) Design of PhC optical intersection device, with degenerate mode profiles of resonator structure calculated as given in the inset on the right. (b) Configurations of high order resonators with their FDTD calculated frequency responses. 3. Experimental The process flow of the PhC optical intersection device follows the basis as that previously reported with respect to the high aspect ratio (>50) PhC rods with radii 115 nm [12] and silicon-on-insulator (SOI) PhC with macro-guiding coupling structures [13]. Starting with a p-type (~10Ω-cm) SOI wafer with top silicon 3 μm and buried oxide (BOX) 1 μm thick, PhC lattice of 517 nm period and bulk rods radii 140 nm with defect rods radii 69 nm were patterned using KrF lithography. The smallest features the defect waveguides made up of reduced-radii PhC rods were reduced in CD all around to 45 nm by resist trimming process. This is followed by dry etching of trimmed CDs into hardmask, down onto the device silicon surface. After the resist is stripped, the cleaned wafers were again patterned for macro coupling structures of waveguide bends and tapers - Fig. 2(b). This is then followed by hardmask-, silicon-, oxide-, silicon-etch, down to a total etch depth of 62.3 μm. The PhC lattices were then uncovered by O 2 plasma and transferred into Si using a fine tuned timemultiplexed reactive ion etch process with passivation for smooth vertical PhC rods sidewalls. In this way, the notching issues of device silicon etch on SOI wafer was circumvented through the process flow sequence design. Figure 2(a) shows the SEM image of the PhC optical interconnect device with the various embedded features described. Partially reflecting rods 90 nm reduced radii defect rods waveguides High density PhC bulk lattice Resonant cavity rod Tapered waveguides y z x PhC optical interconnect PICD Second tapered waveguide First tapered waveguide Curved path isolator waveguide Deeply etched fiber grooves (a) (b) Fig. 2. Plane view SEM image of Si PhC optical interconnect device (a) with zoomed in embedded features; (b) with macro coupling structures and integrated waveguides. 4. Results & discussions The fabricated PhC optical interconnect PICDs with photonic bandgap wavelengths spanning 1.2 to 1.8 μm, has single photonic waveguide mode centered at 1.54 µm wavelength; resonator orders ranging form 1 1 to 5 5, and center resonator rod radii CDs that varies between 0 to 325 nm. In the experiments, these devices were characterized for their passive optical properties and also their active modulation operations using Bragg with index-assisted waveguiding together with resonator-enhanced optical carriers injection mechanisms. Figure 3 presents the typical transmission characteristics of the PhC optical interconnect PICD, for a 3 3- and a 5 5- order resonator structure respectively. As expected, Lorentzian spectrum - Fig. 3(a) - responses were obtained in contrast to the case of the 1 1 order devices (which typically gave much broader band spectrum responses [11]), due to the improvement in quality-factors of the higher-order resonator structures. At the same time, the optical

5 intersection devices remained sensitive to the dispersion properties of the resonator, as can be seen from the plot of transmittance against variations in center resonator rod radii - Fig. 3(b); where it can be seen that there is reasonable agreement between the filled symbols - which represent the optical measurements - with the unfilled symbols representing the 3D-FDTD simulation results. Here, although modulation contrast is reduced as opposed to the quasi-2d case due to finite cladding and asymmetrical optical waveguide mode losses, cavity effected relative modulations are still ostensible. Similarly, these factors also contribute to the higher propagation loss coefficients for the slab-type as opposed to the quasi-2d type [11] PhC rods lattice defect waveguides; as measured through the use of the cut-back method (which makes use of variable lengths singleline defect waveguides with similar optical terminations) to give 0.46 and 0.72 db/µm respectively for waveguides with and without reduced-radii rods defects waveguiding. The higher propagation loss in the devices without reduced-radii rods defect waveguides may be attributed to the lack of vertical confinement, which is otherwise present in the index-assisted reduced-radii PhC rods defect waveguide case. Also, the relatively larger propagation losses of these waveguides as compared to the quasi-2d case may be attributed to the fact that the latter throughput was measured without isolation structures, such as bends etc - so that the straight-through paths of the Fabry-Perot measurements [11] may have resulted in greater transmittance due to the presence of top/bottom leakage readings. Another reason would be that albeit the slab-type structure has index-guiding in the out-of-plane direction; coupling with TE-like modes from the symmetry-breaking substrate induces such guided modes to become leaky too. Finally, while scattering losses at the optical interconnects are frequencydependent due to the strongly dispersive band structure near the partially reflecting mirror layers around the resonator; it was found that the insertion loss of such slab-type devices were much higher than that measured for the quasi-2d case, due to the mismatch in optical mode size and shape (between the input fiber and the slab type waveguide), which resulted in a large amount of input radiation being reflected at the input interface. Such high insertion loss would be reduced in future experiments through the design and use of 3D mode converters. Transmitted power Normalized transmission Wavelength (nm) Resonator rod radius (nm) (a) (b) Fig. 3. Optical measurement of (a) Lorentzian spectrum response from a 3 3, and (b) 5 5 order resonator cavity response. Next, we investigate the inducement of light based modulation through optical injection, for substantial increase in population of free carriers density within the cavity rod material (in excess of those lost to on-going diffusion and relaxation processes); to obtain proportional changes in the cavity s optical density [10]. Here, application of spatially localized pump beam with photon energies ~ 2.33 ev (much greater than the silicon bandgap of 1.1 ev), was focused onto the cavity resonator rod (which, with enhanced quality-factor acts to improve

6 sensitivity). At the same time, low intensity signal beam with wavelength centered in the optical telecommunication range was used for device probing though lateral injections. By the Drude model, the change in real dielectric constant may be expressed as Ne Δε = 2 2 m* ε 0( ω + τ d ) (1) where N is the free electron concentration, e is the electron charge, m * is the optical reduced mass, ε 0 is the optical frequency dielectric constant, ω is the frequency, and τ d is the Drude damping time. Thus, the induced change in refractive index via change in dielectric constant is directly determined by the optically induced free-charge carriers density. From Fig. 4, it can be seen that for applied pump intensity of MW/cm 2, instantaneous modulations of transmittance was obtained - resulting in corresponding increase in probing transmittance of ~ 57 % while FWHM was decreased from 40 to 36 nm. When pump intensity was further increased to ~ 1.06 MW/cm 2, the transmittance spectrum flattens and device can be observed for melt-down. As such, it was verified that lower pump intensity may be applied for device signal modulations in the current slab-configuration as compared to the high aspect ratio case [11] - since there is higher Q-factors in such cavities as compared to the HAR 1 1 order resonator case (without surrounding, partially reflecting PhC micromirrors for retention of radiation). Therefore, as an accompanying consequence, much lower power for device meltdowns was also observed. At the same time, for the time-scale of operation, the free-carriers plasma effect was limited mainly by the relaxation time of the excited carriers which is ~ 80 ps for silicon. To resolve such time resolutions, it is necessary to synchronize both applied pump and probe beams, and at the same time introduce a variable delay-line setup with appropriate detectors for measurements. However, as the current experiment setup resources allow for only unsynchronized pump and probe beam injections, with disparate wavelengths and energies without optical choppers, it remains for future works that a set-up with synchronized pump and probe sources be incorporated with a variable delay-line system, so as to enable accurate time-domain switching-speed measurements. 2 With pump power Power level 36 nm Without pump power 40 nm PhC device melt down Wavelength Fig. 4. Instantaneous photonic crystal output modulation by the on/ off toggling of the high intensity applied continuous wave pump laser. If unlike the case of Fig. 4, the duration of pump-beam application allows for fast energy exchange between the optically excited electronic sub-system and the optically excited material s atomic lattice, then thermo-optic effect modulations may then be obtained as shown in Fig. 5. In the thermo-optics induced modulations, the effect of temperature increase in the PhC optical intersection device can be shown to yield both expected red-shifts in the resonant

7 spectrums and also corresponding bandwidth widening in the much slower thermomodulation effect. In Fig. 5, red-shifts in the spectrum response for progressively increased optical pump power applied are plotted. It can be seen that on average, 0.01 mw increase in pump intensity gives ~ 0.1 nm shifts in center wavelength. This can be extended, up to applied pump-power of nearly 0.11 mw, beyond which device melt-downs are observed to occur. Near melt-down, cumulative center wavelength shift of about 2.2 nm can be obtained. By FDTD simulations, center wavelength red-shifts of 2 nm are obtained for refractive index increase of 0.2. Similarly, results obtained in the bandwidth measurement front also showed that bandwidth widening of up to 3 nm was measured prior to device melt-down indicating refractive index increase slightly more than 0.2. Hence, these results are corroborated by the thermo-optics coefficient of silicon, which indicates that an increase in refractive index of 0.2 occurs at over a thousand degrees elevation in temperature (which gives nearly the melting point of silicon). Center Center wavelength shi shift nm/div nm/div Incident power (mw) Fig. 5. Center wavelength shifts of the 3 3 PhC optical intersection resonator as it was thermo-optically modulated by progressively higher pump-power. In these experiments, repeatability of measurements was good apart from after device melt-down - whereby no further modulation effects may be obtained. In future experiments, further coupling efforts will be incorporated to reduce the coupling losses 5. Conclusions Photonic integrated PhC optical intersection devices of the slab-type configuration was demonstrated by fully CMOS compatible process technologies, with enhanced resonator quality-factors and improved waveguiding, coupling schemes. The simulation results revealed that optical modulation limited in speed only by the relaxation times of excited carriers in the order of picoseconds can be achieved on such device due to the small control volume of the center resonator cavity rod size, for which the PhC PICDs are sensitive to. These active demonstrations with both ultrafast free-carriers injection and also thermooptics induced modulations are promising for future high-density integrated active photonic circuits applications. Acknowledgments The authors would like to thank all research colleagues who have helped with their insightful suggestions. This work was supported through the Agency for Science, Technology and Research Singapore, A*STAR Graduate Fellowship program.