The purpose of this paper is to set the scene for what promises to be an outstanding conference. To this end the paper will survey the early work in silicon photonics from the late 1980s to the mid 1990s. This was when the more fundamental studies of basic building blocks were carried out, such as study of the silicon optical waveguide itself, the contributions to loss and improvement of waveguiding devices. Issues such as how to achieve modulation, and how to implement a modulator, the criteria for single mode propagation will also be covered, as well as work on the beginnings of optical circuits in silicon and SOI. The focus will be upon pure silicon, usually, but not exclusively in the form of Silicon on Insulator (SOI), as opposed to work on compounds such as SiGe or SiC. Much of this work still resonates with work being carried out today, because the move to smaller and more efficient devices means that some of these issues must be revisited in order to achieve optimal device performance. Hence the paper will provide a summary of the early work on silicon photonics, and attempt to relate it to some of the issues being studied today.

Due to recent investments by government and industry, silicon-based photonics has a chance of becoming “the” mainstream photonics technology. This paper presents a survey of recent results found in journal articles and conference proceedings. Emerging trends in silicon-based photonic components (waveguides, ultrafast modulators, switches, light sources, detectors, direct bandgap SiGeSn/GeSn devices, photonic-crystal and plasmonic devices) are identified and discussed. In principle, Si PICs and OEICs can operate anywhere within the 0.3 to 100 μm wavelength range -- enabling transceivers, on-chip processing, and interfacing with fibers or free-space light beams. Thus, in addition to the very important 1.55 μm telecomm applications, there are significant Si photonic opportunities throughout the infrared-and-visible spectrum. The push towards smarter, ever-denser on-chip photonic networks, drives a “convergence” of micro-, nano- and plasmo-photonic techniques for progressively smaller devices (Moore’s law for photonics) and for improved functionality of modulators, switches, emitters, detectors, waveguides, resonators, tapers and filters. This convergence includes composite components: monolithic integration of microstrip waveguides, 2D and 3D photonic-crystal elements and metal/Si plasmon-optics that ultilize buried or surface-mounted 2D arrays of metal stripes or nanodots.

We report broad-spectrum electroluminescence (EL) in metal-insulator-silicon (MIS) tunnel diodes. In addition to Si-band-edge EL near 1.1 eV, hot-electron EL in Si can span a detector-limited range from 0.7 eV to 2.6 eV (1780 nm to 480 nm). The maximum EL photon energy increases with forward-bias voltage. In one implementation, sub-micron-size sites for light emission appear during forward-bias stress. The number of sites is linear in the applied current, consistent with formation of an anti-fuse at each site following breakdown of the insulator. We compare the post-stress current-voltage data to the quantum-point-contact model. Results are presented for various p-type Si(100) MIS devices having thin (8 nm or less) insulating layers of SiO₂, Al₂O₃, and HfO_xN_y. We also describe novel MIS devices in which electron-beam lithography of an 18-nm-thick SiO₂ insulator is used to define the EL sites.

Scaling properties of two photon absorption, free carrier scattering, Raman scattering and Kerr effect in silicon waveguides is reported. It is shown that the dependence of minority carrier lifetime on waveguide dimensions has a profound impact on the performance of nonlinear optical devices built using silicon waveguides.

Two Ge on Si photodetectors are reported: MSM photodiodes with an amorphous Ge layer to increase the Schottky barrier height and Ge/Ge_xSi_1-x/Si heterojunction photodiodes. Monolithic fabrication of Ge PIN photodiodes with Si MOSFETs has also been successfully demonstrated.

In this paper we report on the dependency of quantum efficiency of an avalanching light emitting junction on current density and on the injection current from an adjacent lying forward biased junction. In particular, we report on the interpretation of results and modelling of the physical processes responsible for the light emission. The phenomenon was observed in a three terminal silicon bipolar junction CMOS light emitting device (Si BJ CMOS LED). Our observations show that the overall quantum efficiency and light emission from these type of devices can be improved to the 10^-3 regime. The optical emissions is about four orders higher than the low frequency detectivity for silicon CMOS detectors of comparable dimension. The three terminal device also enable modulation of the light emission by a third terminal contact. The device has the potential of being fully integratable with standard CMOS integrated circuitry with no adaptation to the CMOS design and processing procedures.

Recently a number of successful free-space chip-to-chip and board-to-board optical interconnects have been demonstrated. Here we present some of the results that can be derived as a result of this work.

Due to the mature silicon fabrication technology and vast existing infrastructures, silicon photonics has a chance to offer low cost solutions to telecommunications and data communications. It could also enable a chip-scale platform for monolithic integration of optics and microelectronics circuits for applications of optical interconnects for which high data streams are required in a very small footprint. Two key building blocks needed for any silicon based optoelectronics are silicon based light source and high-speed optical modulator. This paper gives an overview of recent results for a fast (>1GHz) silicon modulator and a silicon Raman laser. We present optical characterization of a high speed metal-oxide-semiconductor (MOS) capacitor-based silicon optical modulator. We show that a Mach-Zehnder interferometer (MZI) structure with a custom-designed driver circuit results in the realization of a silicon modulator transmitting data at 2.5 Gb/s with an extinction ratio of up to 2.8 dB. In addition we show that by reducing the waveguide dimensions one can improve the phase efficiency. In addition, as single crystal silicon possesses higher (four orders of magnitude) Raman gain coefficient as compared to silica, it is possible to achieve sizeable gain in chip-scale silicon waveguide for optical amplification and lasing. With a 4.8 cm long waveguide containing a reverse biased p-i-n diode, we demonstrate lasing operation using a pulsed pump laser. We achieve ~10% slope efficiency. We in addition model a continuous-wave silicon Raman laser and show that higher conversion efficiency and lower threshold power can be realized with optimised cavity device design.

We fabricated and characterized an electro-optic Si-based amplitude light modulator working at 1.5 μm. It is a Bipolar Mode Field-Effect transistor (BMFET) integrated within a Si rib waveguide. The devices, 100 μm long, were fabricated using epitaxial Si wafers and standard clean room processing. The light is absorbed during its travel in the device optical channel when a plasma of free carriers, electrically driven, is generated and placed inside the channel. We experimentally monitored the plasma formation and localization in the device using standard Emission Microscopy analysis. The optical characterization in static conditions provides a modulation depth of ~ 90%, well above the 25% minimum required to consider a device a modulator. Furthermore, dynamical measurements show a modulation depth of 65% at 100KHz of operation frequency. Finally, an experimental evidence of a frequency threshold, at about 500 KHz, is observed in the plasma behavior. Theoretical considerations and experimental data suggest that at frequencies below threshold the dominant phenomenon is the plasma generation/recombination, while above threshold the carrier drift leads the plasma motion and redistribution in the device channel.

Si photonics could enable a chip-scale platform for monolithic integration of optics and microelectronics for applications of optical interconnects in which high data streams are required in a small footprint. This paper discusses mechanisms in silicon photonics for modulating and switching light. These mechanisms together with recent advances of fabrication techniques have enabled the demonstration of ultra-compact active silicon photonic components with very low loss.

Silicon is the most diffused material for microelectronic industry and, in recent times, it is becoming more and more widespread in integrated optic and optoelectronic fields. Nowadays it is possible to realize in silicon, using the traditional microelectronic techniques, a huge variety of passive and active devices, like waveguide, switch, modulators, whose operation range is in the second and third optical telecommunication windows. In this poster we propose a simulation of an integrated waveguide-vanishing-based modulator realized by ion implantation in SOI wafer. The active region is 3x3 micron wide and the lateral confinement is guaranteed by two highly-doped As (8×10⁺19 cm^-3) and B (2×10¹⁹ cm^-3) implanted regions with a depth of one micron. This type of structure allows to obtain a planar device; thus an easier integration with electronic devices is possible to obtain. The implantation process has been carefully tuned in order to get high doping uniformity and sharp profile. The resulting channel waveguide shows single mode operation and propagation loss of about 1.8 dB/mm. The modulation is based on a lateral p-i-n diode, that allows to inject free carriers into the rib volume between the doped regions. The resulting optical behavior is the vanishing of the confinement in the rib region and consequently the cut-off of the supported mode. This phenomenon occurs at driving voltage of about 1.0 V, with electrical power consumption of 2 mW, and implies a rapid propagating characteristics variation, with following optical beam lateral redistribution into the structure. Both fabrication process and operation (electrical and optical) simulations have been carried out. Results show that an optical modulation depth close to 100% can be reached with a rise time of about 10 ns. Fall time can be even faster, if a proper reverse polarization is applied to the device.

Systematic development of electro-optic (EO) polymers is leading to optical and material properties such that they present an increasingly viable alternative to crystalline-based technologies for integrated optics. EO polymers demonstrate an inherent velocity match between radio-frequency and optical waves, making them excellent candidates for applications in high-speed telecommunication switching and optical interconnects for VLSI circuitry. In addition, EO polymer devices are relatively simple to fabricate at conditions compatible with microelectronics industry processes, making same-substrate integration of optical and electronic circuitry possible. In this paper, we describe two vertical integration schemes whereby a polymer-based electro-optic modulator may be controlled by metal-oxide semiconductor field effect transistor (MOSFET) circuitry. One scheme described is an insitu integration on the same silicon (Si) substrate. The second scheme is the integration of a modulator built on a flexible substrate with a MOSFET circuit on a second Si substrate. Both schemes have potential applications for integrated electro-optics.

As integrated circuit interconnect dimensions continue to shrink and signaling frequencies increase, interconnect performance degrades. The performance degradation is due to several factors such as power consumption, cross-talk, and signal attenuation. On-chip optical interconnects are a potential solution to these scaling issues because they offer the promise of providing higher bandwidth. In this paper, progress on the major on-chip optical building blocks will be reviewed. It will be shown that significant advances have been made in the design and fabrication of waveguides, detectors, and couplers. However, major challenges in high speed electrical to optical conversion and signaling remain.

The use of optical waveguides as the data link for off-chip and on-chip interconnects is attracting considerable interest. This paper examines the requirements and some possible solutions suitable for FPGA applications.

Reduction of the stresses produced in hybrid integrated structures due to thermal expansion coefficient difference requires removal of the substrate as one of the key elements. Commonly used epitaxial lift-off technique can hardly be employed for the fabrication of VCSELs with all-epitaxial DBRs due to the low etching selectivity between AlAs sacrificial layer and DBR layers with high Al contents. Novel method of substrate removal named oxidation lift-off was proposed and demonstrated. This process shows higher selectivity against Al-content than epitaxial lift-off method, that allows for the release of a VCSEL structure with epitaxial DBRs and separate individual components on Si, reduces the number of process steps and eventually reduces the cost of the fabricated/integrated devices. Au-Ge alloy was used for the metal bonding of the test oxidation lift-off structures grown by MBE. 1 μm thick AlAs imbedded sacrificial layer was laterally oxidized to release the partially processed devices from the GaAs substrate. 2D array of separated VCSELs was fabricated on top of the Si substrate. Contact annealing, substrate removal, device separation, bonding and formation of the oxide apertures were completed within a single processing step. Electroluminescent spectra, I-V and P-I characteristics of fabricated devices were measured. Series resistance of fabricated devices was found to be about 100 Ohms. Lasing with threshold current of 8 mA was demonstrated for the device with 25 μm aperture.

We review the use of the oxide cladding stress induced photoelastic effect to eliminate the modal birefringence in silicon-on-insulator (SOI) ridge waveguide components, and highlight characteristics particular to high index contrast (HIC) systems. The birefringence in planar waveguides has its origin in the electromagnetic boundary conditions at the waveguide boundaries, and can be further modified by the presence of stress in the materials. It is shown that geometrical constraints imposed by different design and fabrication considerations become increasingly difficult to satisfy with decreasing core sizes. On the other hand, with typical stress levels of -100 MPa to -400 MPa (compressive) in SiO₂ used as the upper cladding, the effective indices are altered up to the order of 10^-3 for ridges with heights ranging from 1 μm to 5 μm. We demonstrate that the stress can be effectively used to balance the geometrical birefringence. Birefringence-free operation is achieved for waveguides with otherwise large birefringence by using properly chosen thickness and stress of the upper cladding layer. This allows the waveguide cross-section profiles to be optimized for design criteria other than zero-birefringence. Since the index changes induced by the stress are orders of magnitude smaller than the waveguide core/cladding index contrast, changes in the mode profiles are insignificant and the associated mode mismatch loss is negligible. We study the stress-induced effects in two parallel waveguides of varying spacing, to emulate the condition in directional couplers and ring-resonators. In the arrayed waveguide grating (AWG) demultiplexers fabricated in the SOI platform, we demonstrated the reduction of the birefringence from 1.3x10^-3 (without the upper cladding) to below 1x10^-4 across the spectral band by using a 0.6 μm oxide upper cladding with a stress of -320 MPa (compressive). Design options for relaxed dimensional tolerance and improved coupler performance made available by using stress engineering are discussed.

The recent interest in silicon based photonics, and the trend to reduced device dimensions in photonic circuits generally, has led to the need for mode converters to couple from optical fibres to such small devices. A range of structures have been proposed and in some cases demonstrated, including three dimensional tapers, inverted tapers and micromachined prisms. We have previously reported theoretical analyses of a Dual Grating Assisted Directional Coupler (DGADC), which promises high efficiency coupling over modest spectral linewidths. In this paper we report preliminary experimental results on the fabrication of such devices, together with an evaluation of the coupling efficiency. The approach has been to fabricate a demonstrator device for a particular arrangement of waveguide coupling parameters, i.e. we have fabricated a device that couples easily from fibre, because the input waveguide is approximately 5μm in cross sectional dimensions. The mode converter then couples to a 0.25μm silicon waveguide, primarily because comparisons exist in the literature. These results are compared with the predicted efficiency, and the results are discussed both in terms of the constituent parts of the DGADC, as well as the fabrication limitations. Whilst our device is not optimised we demonstrate that it has promise for very high efficiency coupling.

A new on-chip silicon-based Bragg cladding waveguide with full CMOS compatibility is developed. This novel optical waveguide has a low refractive index core (SiO₂) surrounded by a 1D photonic crystal cladding. The cladding consists of several dielectric bilayers, where each bilayer consists of a high index-contrast pair of layers of Si and Si₃N₄. This new waveguide guides light based on omnidirectional reflection, reflecting light at any angle or polarization back into the core. Its fabrication is fully compatible with current microelectronics processes. In principle, a core of any low-index material can be realized with our novel structure, including air. Potential applications include tight turning radii, high power transmission, nonlinear properties engineering and biomaterials sensors on silicon chip.

In this study, we design and fabricate a hollow optical waveguide with omni-directional reflectors in silicon-based materials. A groove is etched by inductive coupled plasma (ICP) with photolithographic process on (100) silicon wafer. The width of the groove is varied from 3.5 to 5.5 micrometer for different waveguide designs. The depth of the groove is 1.2 micrometers. Plasma enhanced chemical vapor deposition is used to deposit six pairs of Si/SiO₂(0.111/0.258micrometers) on the samples. Finally, the top of the sample is covered by another silicon substrate on which the identical omni-directional reflector has been also deposited. By wafer bonding technology, the top omni-directional reflector can be combined with the groove to form a hollow optical waveguide. Light with the wavelength at 1.55 micrometers can be confined by the omni-directional reflectors at single mode operation. Polarization independent hollow optical waveguides can be achieved with this fabrication process.

Recently, we have realised a polarisation independent optical racetrack resonator whose resonant dips for TE and TM align to better than 1pm. The devices had a Free Spectral Range (FSR) of only several hundred picometres. This in large part was to the relatively large bend radius (~ 400μm) designed and fabricated with initial focus on producing low bend loss devices. Modelling of the bend loss of the same dimension devices shows that the bend radius can be reduced significantly (down to ~25μm) to produce race track ring resonator with an FSR that is approximately 400% larger than that of those previously fabricated, whilst retaining polarisation independence. This paper will focus on the proposed enhancement of these devices as well as the impetus for their investigation.

MEMS devices can be successfully commercialized in favour of competing technologies only if they offer an advantage to the customer in terms of lower cost or increased functionality. There are limited markets where MEMS can be manufactured cheaper than similar technologies due to large volumes: automotive, printing technology, wireless communications, etc. However, success in the marketplace can also be realized by adding significant value to a system at minimal cost or leverging MEMS technology when other solutions simply will not work. This paper describes a thermally actuated, MEMS based, variable optical attenuator that is co-packaged with existing opto-electronic devices to develop an integrated 10Gb/s SONET/SDH receiver. The configuration of the receiver opto-electronics and relatively low voltage availability (12V max) in optical systems bar the use of LCD, EO, and electro-chromic style attenuators. The device was designed and fabricated using a silicon-on-insulator (SOI) starting material. The design and performance of the device (displacement, power consumption, reliability, physical geometry) was defined by the receiver parameters geometry. This paper will describe how these design parameters (hence final device geometry) were determined in light of both the MEMS device fabrication process and the receiver performance. Reference will be made to the design tools used and the design flow which was a joint effort between the MEMS vendor and the end customer. The SOI technology offered a robust, manufacturable solution that gave the required performance in a cost-effective process. However, the singulation of the devices required the development of a new singulation technique that allowed large volumes of silicon to be removed during fabrication yet still offer high singulation yields.

The integration of optical transducers is generally considered a key issue in the further development of lab-on-a-chip microsystems. We present a technology for the integration of miniaturized, polymer based lasers, with planar waveguides, microfluidic networks and substrates such as structured silicon. The flexibility of the polymer patterning process, enables fabrication of laser light sources and other optical components such as waveguides, lenses and prisms, in the same lithographic process step on a polymer. The optically functionalised polymer layer can be overlaid on any reasonably flat substrate, such as electrically functionalised Silicon containing photodiodes. This optical and microfluidic overlay, interfaces optically with the substrate through the polymer-substrate contact plane. Two types of integrable laser source devices are demonstrated: microfluidic- and solid polymer dye lasers. Both are based on laser resonators defined solely in the polymer layer. The polymer laser sources are optically pumped with an external laser, and emits light in the chip plane, suitable for coupling into chip waveguides. Integration of the light sources with polymer waveguides, micro-fluidic networks and photodiodes embedded in a Silicon substrate is shown in a device designed for measuring the time resolved absorption of two fluids mixed on-chip. The feasibility of three types of polymers is demonstrated: SU-8, PMMA and a cyclo-olefin co-polymer (COC) -- Topas. SU-8 is a negative tone photoresist, allowing patterning with conventional UV lithography. PMMA and Topas are thermoplasts, which are patterned by nanoimprint lithography (NIL).

Here we present an analysis of a fully planar optical sensor based on ARROW waveguides. We consider a device geometry that consists of two intersecting ARROW waveguides, one with solid and one with liquid core. The pump wavelength is input from the solid core waveguide and penetrates through the wall into the liquid core of the other waveguide where the fluorescence is excited in the sample material. Then it is captured by the ARROW waveguide and guided to a detector at the end. Pump and signal wavelength can be guided with low loss through the solid and hollow core respectively. At the same time, high loss discrimination inside the core can be obtained by tailoring the thickness of ARROW layers, leading to efficient filtering. The pump wave can be transmitted efficiently in and out of the core, allowing for multiple intersections. In the following experimental part, we investigate the waveguide loss of the liquid core as a function of core width, we measure values as low as 1.7cm^-1 that can be further reduced by improving the thickness control of ARROW layers in the fabrication process. Finally, we report fluorescence experiments on Alexa dye molecules. We demonstrate fluorescence detection at concentrations as low as 10^-8 mol/l from a detection volume of 21pl. Photo-bleaching is observed and discussed as a function of input power intensity.

Optical sensing, and the integration of sensors and electronics into Sensor on a Chip and Sensor on a Package systems are an approach to the creation of miniaturized, portable, customizable, low cost sensor systems for rapid health diagnostics, medical research, environmental monitoring, and security monitoring. To integrate optical sensing systems that are autonomous, it is essential to integrate the sensor, light source, and light detection into a single substrate or chip. The integration of this optical system with signal control and processing electronics enable discrimination with individually customized sensors in sensor arrays, and high sensitivity levels. Thin film optoelectronic active device integration with planar optical passive devices is a heterogeneous integration method for fabricating planar lightwave integrated circuits at the chip level and planar lightwave integrated systems at the substrate and package level.

Fiber Bragg Gratings (FBG) sensors are a very promising solution for strain and/or temperature monitoring in hostile or hazardous environments. In particular, their typical immunity to EMI and the absence of electrical signals and cables, encourage the use of FBG sensors in aerospace structure. Moreover, FBG sensors can be embedded in composite materials, allowing the fabrication of the so-called smart-materials. In this paper we experimentally demonstrate that a Fabry-Perot cavity, integrated in a low-loss all-silicon rib waveguide, and realized by standard dry etching technique, is suitable for FBG monitoring. The reflected signal for the sensor passes through the cavity which is tuned by means of thermo-optic effect. The optical circuit ends with a photodetector that, for each tuning step, produces a photocurrent proportional to the convolution integral between the FBG and the FP spectral response. Because the finesse of a silicon FP cavity in air is not so high (about 2.5), it is advantageous an extended tuning over a wavelength range longer than the cavity free spectral range, that is convolving the FBG response with more than one FP transmission peak. The photodetector output signal, once acquired, is elaborated using standard FFT algorithm and pass-band filtered, in order to extract the main harmonic. After a final I-FFT step, a fitting procedure returns the FBG reflection peak position. The experimental accuracy, using as reference the peak wavelength measure made with a commercial high-performance Optical Spectrun Analizer, is in the order of few tenths of picometers.

Infrared absorption photoinduced by visible light in a-Si_xC_1-x:H is characterized by in guide pump and probe measurements in order to test its applicability to a low-cost micromodulator, fully integrable as a post-processing on-top of a standard microelectronic chip. The Photoinduced Absorption phenomenon in amorphous silicon arises from an alteration of the defect state population by decay of carriers photogenerated by visible light. These levels, deep in the gap, are strongly involved in interactions with IR radiation, and then the VIS illumination modifies their optical properties by increasing the IR absorption coefficient value. Test waveguiding devices are fabricated by Plasma Enhanced Chemical Vapour Deposition on silicon wafers, at temperatures lower than 180°C, and consist of a a-SiC:H/oxide stack. In particular, devices having a-Si_xC_1-x:H cores with different doping and different carbon concentration are characterized. The 1.55 μm probe radiation generated by a DFB laser diode is efficiently transmitted through the a-Si_xC_1-x core thanks to the step index waveguide structure. The pump system consists of low cost AlInGaP LEDs pulsed by a function generator, for an illumination intensity ranging from 0.15 up to 0.85 mW/mm². Results show that the modulation effect increase for longer pump penetration depth and for higher doping concentration. The phenomenon strongly depends on the carbon introduction in a-Si:H. Digital transmissions tests at 300 kbit/s were performed.

In this work, slab and strip optical waveguides were fabricated onto silicon substrates using silicon oxynitride (SiO_xN_y) films, with different chemical compositions, as core and cladding layers. In order to obtain high optical quality and low attenuation levels the nitrogen composition in the core and cladding films were varied from 0% up to ~31% and the index contrast from 1% up to 6%. The constituent materials were deposited by plasma-enhanced chemical vapor deposition (PECVD) technique and characterized by a prism coupling system in order to obtain the refractive index and the thickness values. On the other hand, the slab and strip optical waveguides were annealed at 550°C in vacuum during 2 hours and characterized optically by the moving fiber method and by the end-fire coupling technique, respectively. The results of the optical characterizations in the waveguide structures showed a decrease in their optical losses of up to 50% after the annealing treatment, which can be related with an improvement in the local structure and in the quality of the interface of the constituent films.

This work describes a dual-rate optical transceiver designed for power-efficient connections within and between modern high-speed digital systems. The transceiver can dynamically adjust its data rate according to the performance requirements, allowing for power-on-demand operation. To implement dual rate functionality, the transmitter and receiver circuits include separate high-speed and low-power datapath modules. The high-speed module is designed for gigabit operation and optimized to achieve the maximum bandwidth. A simpler low-power module is designed for megabit data transmission and optimized for low power consumption. The transceiver was fabricated with a 0.5μm Silicon-on-Sapphire (SOS) CMOS technology. The vertical cavity surface-emitting lasers (VCSELs) and photodetector devices were attached to the transceiver IC using flip-chip bonding. A free-space optical link system was set up to demonstrate power-on-demand capability. Experimental results show reliable link operations at 2Gb/s and 100Mb/s data transfer rates with about 104mW and 9mW power consumption, respectively. The transceiver’s switching time between these two data rates was demonstrated at 10μs which was limited by on-chip register reconfiguration time. Improvement of this switching time can be obtained by using dedicated IO pads for dual-rate control signals. At the circuit level, the incorporation of dual rate functionality into a typical gigabit optical transceiver would require 255 additional MOS transistors.