This paper presents an overview of our study on the subject that we categorically termed VLSI (very large scale integration) microphotonics. We examine the scientific and technological issues and challenges concerning three essential steps in this technology: miniaturization, interconnection, and integration of microphotonic devices, circuits and systems in micron or submicron scale. In miniaturization, the issues on the size effect, proximity effect, energy confinement effect, microcavity effect, single photon effect, optical interference effect, high field effect, nonlinear effect, noise effect, quantum optical effect, and chaotic noise effect should be addressed. In interconnection, the issues of connecting identical devices (homogeneous interconnection) or nonidentical devices (heterogeneous interconnection) have to be examined. Optical alignment between micron-scale devices, minimizing interconnection losses, and maintaining optical modes between devices, are to be considered. In integration, the issues of interfacing same kind of devices, two different kinds of devices, and several or many different kinds of devices have to be addressed. Other issues include the design and packaging of the integrated devices and circuits as a system for reliable function and operation. In the course of this study, we closely follow the experiences of VLSI microelectronics so that they can provide lessons, learnings, and insights that microphotonics can benefit from. Similarities, dissimilarities, advantages, and disadvantages of the two technologies are explored in such a way that they can be more effectively utilized by mutual support and complement. Directions for future studies are also discussed.

New generations of network application continue to demand component solutions of higher speed (10 Gbps and beyond) with good reliability and affordability. The potential of VCSEL arrays technology in meeting those demands has long been recognized and is being actively explored by many in the field. In this paper we present a few examples of VCSEL array based technology developments including monolithic integration of VCSELs and photodetector (PD) in both 1D and 2D arrays, and their hybrid integration with micro-lens array and electronic integrated circuits for optical interconnects. As we explore to achieve more functionality through integration, we also emphasize on the merits of usability of the electrical and optical interfaces of the new components, and producibility and manufacturability aspects of the new technologies.

Lithium Niobate (LiNbO₃) films produced by the crystal ion slicing (CIS) method are introduced in various components in use in the optical telecommunications market. The CIS technique employs high-energy ion implantation to create a narrow (~0.2 micrometers ) planar layer of localized damage, buried ~10 micrometers beneath the surface of the implanted LiNbO₃ wafers. This sacrificial layer allows for slicing of microns-thick LiNbO₃ films, either by selective wet chemical etching or by thermal shock. The obtained films have bulk material properties and morphology suitable for integrated optics applications. Slices of X-cut LiNbO₃ were used to produce zero-order wave retarders that can be inserted in slots cut into planar lightwave circuits, resulting in TE-TM polarization mode conversion with high extinction ratio (30 dB) and low excess loss (<0.1 dB). Conventional LiNbO₃ waveguide fabrication techniques were combined with the CIS process to produce CIS films of Z-cut LiNbO₃ with optical circuits patterned prior to lift-off, having propagation losses typical of bulk LiNbO₃ waveguides. Using thin sheets of LiNbO₃, velocity- and impedance-matched modulators can be fabricated with low V_(pi )L(~7.6 V.cm) and low microwave losses (0.3 db/cm.GHz^1/2). The CIS film optical circuits can be integrated into hybrid systems with otherwise incompatible, yet technologically important materials platforms.

This paper reviews some of the key enabling technologies for present and future WDM communications systems. This review concentrates mainly on advances in material growth, lasers, modulators, and fast photodetectors as the basis modern communication systems. Emphasis is placed on integration of components, the need for reconfigurable switching elements, and mass production of a generic intelligent transceiver.

The results of a detailed experimental characterization of a three-port, all-active SOA-MZI wavelength converter are presented. It is shown that for counter-propagating input signals, the ASE noise is intensity modulated due to gain saturation in the output SOA. Consequently, the waveform for the wavelength-converted signal is determined by both cross- phase modulation and self-gain modulation. Time- and frequency-domain measurements are used to characterize the properties of the modulated ASE noise, wavelength-converted signal and total signal. For both co- and counter- propagating signals, the small-signal chirp properties of the wavelength converted along the conversion curve are considered. The measured results for the small-signal chirp and optical conversion are incorporated into a device model that can be used to obtain the large-signal pulse response. Based on this model, good agreement is demonstrated between calculated and measured results for the time dependence of the intensity and chirp of the wavelength-converted signal.

Optical hybrid integration technology has been developed using a planar lightwave circuit (PLC) platform. The platform combines a silica optical waveguide and optical active devices to realize various optical functions without restricting the type of device. We employ two additional techniques to fabricate these optical modules, namely, a multichip mounting technique for optical active devices and a PLC-PLC attachment technique. The use of a PLC platform and these techniques will make it possible to apply our technology to various optical modules from an optical transceiver module for subscriber systems to a multichannel large-scale optical module such as that employed for WDM systems.

An optical transceiver is presented consisting of a polarization insensitive directional coupler, a hybrid integrated self-aligned laser diode and a hybrid integrated photodiode. In order to increase waveguide-fiber coupling, the fiber is connected to a tapered waveguide. The polarization insensitive directional coupler is designed for separation of the transmitting ((lambda) equals1.3 micrometers ) and the receiving ((lambda) equals1.55 micrometers ) wavelengths. The coupler is based on Al₂O₃ waveguides realized by MOPECVD processes. These waveguides show attenuation below 0.3 dB/cm at the wavelengths of interest after annealing at ~700 degree(s)C. The hybrid integration of the laser diode is realized by a self-aligned soldering process. Electroplated tin/gold is used as the solder, while the pre-positioning of the laser diode is achieved by a fine-placer. After self-alignment, the misalignment of the laser has to be smaller than +/- 0.5 micrometers vertical and +/- 1.5 micrometers lateral to achieve coupling losses below 3 dB. Vertical mirrors are used for guiding the signal on the chip to reduce optical losses and chip size. The waveguide-fiber coupling is optimized by a tapered waveguide during the deposition of the Al₂O₃ layer by a KOH-etched silicon mask. The lateral positioning of the fiber is guaranteed by the vertically etched walls of the waveguide improving the properties of the coupling facet. The depth of the fiber groove is machined by an isotropic silicon plasma etch process.

Magneto-optics is an area that is uniquely enabling for the production of nonreciprocal components such as optical isolators and circulators. The concepts behind the nonreciprocity include nonreciprocal polarization rotation (Faraday rotation) and nonreciprocal phase shift. A magneto- optic material that is magnetized in the direction of propagation of light acts as a Faraday rotator. An asymmetric magneto-optic waveguide that is magnetized perpendicular to the propagation direction acts as a nonreciprocal phase shifter. Both effects can be utilized to realize nonreciprocal devices. Today, commercial isolators and circulators are strictly bulk components, and as such they constitute the only type of optical component that is not available in integrated form. However, the technology for integrated nonreciprocal devices has been maturing and is expected to have a considerable impact in the communication industry by enabling the integration of complete optical subsystems. We report on the development of integrated optical isolators and circulators that consist of polymer-based planar interferometers with inserted thin films of cerium-substituted Yttrium Iron Garnet (Ce-YIG) for efficient Faraday rotation, and thin films of LiNbO₃ for wave-retarders that enable polarization-independent operation.

Growing attention is paid to intraboard optical interconnects among multichip modules because of its potential to solve so-called pin bottleneck problem in constructing ultrahigh speed information processing unit. Vertical integration of waveguides and their coupling as well as input/output coupling of guided waves from/to free- space waves are key issues. Three types of high performance coupling via bridging modes by integrated gratings are reviewed. Use of radiation mode as bridging mode is discussed with advantage of selective coupling between two waveguides among multistory structure. Three waveguides were integrated with grating couplers, and wavelength-division demultiplexing with 5 nm separation and 0.5 nm selectivity from one waveguide to two other waveguides was demonstrated. Utilization of supermode as bridging mode is shown for optical add/drop multiplexing (OADM) function between two waveguides. Device was designed to have high power transfer efficiency of almost 100% with high wavelength selectivity of 2 nm, while preliminary experimental results were 40% efficiency with 1 nm selectivity. OADM between guided wave and free-space wave is also discussed with utilization of supermode. Power transfer efficiency of 25% with wavelength selectivity of 4 nm was obtained in coupling of guided wave to free-space wave, while theoretically predicted efficiency was 70% with 2 nm selectivity.

We describe design of 622 Mb/s 16-channel CMOS optical transceiver array using the 0.35 micrometers CMOS technology. The transceiver array consists of Laser Diode (LD) driver and limiting amplifier with trans-impedance amplifier. CMOS LD driver offers the capability of independent dc and modulation current adjustments. The dc circuit used to pre- bias LD is adjustable for the dc current at a range of 0~30 mA. Because each amplifier block is dc-coupled, there are several sources of nonlinearity in the amplifier chains. These problems deteriorate the magnitude and timing performance. In order to solve these problems, we employ a compensate circuit which consists of positive and negative peak detectors and decision comparator. This scheme forces the data to be sliced in the middle eliminating timing errors. With this design technique, we have succeeded in developing a CMOS optical transceiver array with a high performance.

We have investigated various conductive and nonconductive polymer materials suitable for use as cladding layers in nonlinear optic (NLO) polymer based opto-electronic devices. Our goal was to maximize the nonlinearity of the NLO core materials, while minimizing the total poling voltage and minimizing the absorption loss. Using a cladding material that is more conductive than the NLO core material, the majority of the applied poling voltage is dropped across the core, realizing a maximum EO coefficient with minimum applied poling voltage. We found, however, that there are tradeoffs between absorption loss, conductivity, refractive index, materials processability and materials compatibility when using off-the-shelf materials. Results are presented for a 3-layer device structure using a conductive polymer material for both the top and bottom cladding layers.

Many photonics component manufacturers are struggling to lower the costs of pump- and source-laser module assembly through automation. Even contract manufacturing companies with decades of electronics manufacturing experience are learning the new rules of laser diode assembly. A few pages from the rulebook - specific to automated vision and precision placement - are discussed here.

Through wafer via hole connection has found applications in photonic and microelectronic devices. Such vias provide higher packing densities, improved gain and simplified device layout. In this paper, reactive ion etching of GaAs via hold has been systematically studied using CCl₂F₂ as the reactive gas. The effects of process pressure and r.f. power on the etch rate and the resultant etch profiles have been investigated. It was found that the etch rate increased linearly with the increase of process pressure for values below 50 mTorr. The process pressure had a significant influence on the etch profiles. At low process pressure, anisotropic profiles were observed. The etch rate increased as the r.f. power was increased due to the increased excitation of reactive species as well as higher ion energies and improved sputter desorption of the etch products. Reproducible, and good etch profiles with etch rate as high as 1.55 micrometers /min. could be obtained at a process pressure and r.f. power of 50 mTorr and 150 W, respectively. Devices have been successfully fabricated which employed the via hole process developed in this study. The via hole connections for the grounding were found to be working very well as confirmed by DC measurements.

In pigtailing of a single mode fiber to a semiconductor laser for optical communication applications, the tolerance for displacement of the fiber relative to the laser is extremely tight, a submicron movement can often lead to a significant misalignment and thus the reduction in the power coupled into the fiber. Among various fiber pigtailing assembly technologies, pulsed laser welding is the method with submicron accuracy and is most conducive to automation. However, the melting-solidification process during laser welding can often distort the pre-achieved fiber-optic alignment. This Welding-Induced-Alignment-Distortion (WIAD) is a serious concern and significantly affects the yield for single mode fiber pigtailing to a semiconductor laser. This work presents a method for predicting WIAD as a function of various processing, laser, tooling and materials parameters. More specifically, the degree of WIAD produced by the laser welding in a dual-in-line laser diode package is predicted for the first time. An optimal welding sequence is obtained for minimizing WIAD.

A significant advance in technical capability has recently been achieved in the fabrication of refractive microlens arrays by microjet printing. This advance enables control of lens diameter and center-to-center distances to accuracies on the order of +/- 1 micrometers , and of focal length variations within an array to less than the +/- 1%. Such accuracies are especially important for microlens arrays used for MOEMs device interconnects to optical fibers because of the relatively long free space optical path lengths required for such applications. The new process also enables the printing of microlenses of a given diameter with aspect ratios and focal lengths varying over a wide range (e.g., f/1-f/5). The profile of plano-convex microlenses printed by this method have exhibited less than a quarter wavelength of deviation from spherical surface. The thermal durability of the optical epoxies used for microlens printing enables both cycling temperature up to 200 degree(s)C and continuous exposure to thousands of hours at 85 degree(s)C, without affecting microlens performance. The microjet printing of lensed fibers provides another solution for fiber beam shaping and collimation with low production cost.

In order to couple the light emitted from a semiconductor laser into the core of an optical fiber, where the core and laser emitting are not matched, it is necessary to use a lens system. This lens system then focuses the divergent light into the fiber core; often this lens is fabricated on the end of the fiber. When the lens is on the end of the fiber, it must be fixed in place in front of the emitting laser using a method that will hold the fiber in position relative to the laser during the lifetime of the laser. This reliability requirement is particularly critical in the case of lasers with highly elliptical output beams. The tolerances for the displacement of the lensed fiber relative to the laser are extremely tight; a movement of less than 0.5 microns in the vertical direction can cause a drop in coupled power of greater than 10%. In this paper existing techniques of fixing the fiber relative to the laser are outlined and contrasted with an improved method using glass solder. The improved method is described with salient features such as the fixing of the fiber to a rigid support member and the laser welding process used to fix the rigid support member in front of the laser explained in detail. Other aspects of the laser design including thermal dissipation are also discussed in detail. Finite element analysis (FEA) was used extensively to demonstrate the capacity of the design. Device performance and reliability information is also presented, demonstrating the capability of this technique in the fixing of an optical fiber relative to a laser device.

A number of optical components being developed exhibit, due to the material used in their fabrication, temperature sensitive properties. These properties necessitate that the component is accurately controlled in temperature, in certain cases to +/- 0.1 degree(s)C (ambient temperature 0 degree(s)C-70 degree(s)C). Using standard techniques this temperature control can be made especially difficult when the material being controlled has a very low thermal conductivity, for example, devices made using Lithium Niobate or Polymers. In this paper the thermal aspects of the packaging of these materials are analyzed, and a heat shielding technique is proposed which is capable of controlling the temperature of the component extremely accurately. Extensive Finite Element Analysis is used to demonstrate the improved performance of the technique. Details of prototype construction and results for these prototypes with Lithium Niobate and Polymer devices are reported and discussed.

We propose a silicon microlens that is mountable on a silicon v-groove platform for low-cost telecommunication optical modules. The proposed silicon microlens has a rod shape which has the diameter of 125 micrometers , identical to a single-mode fiber. Therefore, it can be passively embedded in the v-groove in conjunction with the optical fiber with high precision. We employed diffractive microlens that can be fabricated on the top surface of the rod by conventional LSI technology. Better coupling efficiency between a laser diode and an optical fiber is expected with silicon diffractive lenses as compared with conventional silica ones. The coupling efficiency of -0.7 dB between two single-mode fibers was obtained with silicon lenses significantly better than -1.4 dB of insertion loss obtained by a pair of silica diffractive lenses. With good coupling efficiency and the ease of packaging by surface mount technology (SMT), silicon diffractive microlens is promising for low-cost and high- performance optical module applications.

In the collimator array development effort for MEMS applications, the ability to align devices automatically that meets industry target specifications of power throughput for multiple channels has been demonstrated. Results show eight-channel devices have been successfully aligned and attached within 2-dB power loss. Three key factors of success include: (1) Automated Pitch/Yaw using virtual pivot, (2) initial mirror alignment to verify parallelism between fiber array and lens array, (3) channel balance algorithm to compensate for lens array pitch and focal length variations.

Extremely high is nowadays researchers' interest towards amplification techniques alternative to Erbium Doped Fiber Amplifiers (EDFA); driving force is the exploitation of new and larger fiber optic communication bandwidths. While EDFA gain relies on population inversion in an active material and thus amplification bandwidth is directly related to the doping ion energy levels, well known effects such as Raman and third order nonlinear susceptibility in optical fiber allow amplification not subject to this constraint. In optical parametric amplification, in particular, bandwidth shape and spectral collocation are determined by the choice of pump wavelength and power. Thus almost every spectral region can be reached provided that proper pump is available. We present a detailed analysis of optical parametric amplification in standard communication fiber, discussing different operation regime according to pump wavelength placement with respect to fiber zero dispersion wavelength. In particular attention is paid to the optical parametric amplifier gain bandwidth and saturation. We evidence that a key OPA project parameter is the relative phase between interacting waves, i.e., pump, signal and idler. This parameter not only allows comprehension of gain dynamics in the amplifier, but also can be controlled by engineering the nonlinear medium to improve OPA performances. Experimental verifications are also presented to validate the proposed analysis.

Interconnect bottleneck in emerging integrated circuitry (IC) has generated a need for alternative signal transmission solutions, such as optical technologies, in chip-level applications. The present paper discusses target parameters for chip-level optical interconnects (CLOIs) that yield superior performance starting with the 70 nm IC node, and possibly extending down to the 25-15 nm node. The benefits and disadvantages of various CLOI system and component solutions are reviewed. In particular, this paper discusses critical fundamental and technological challenges that need resolution to enable massively parallel CLOI links with a total throughput of 10-25 Tb/s, reduced power consumption in comparison with electrical wires, and enhanced density. Recent results from the presenting authors are summarized with an emphasis on CLOI specific solutions. These results include the development of InAs quantum dot gain medium to increase the operating temperature of laser arrays above that of Si ICs. Controllable routing of VCSEL- emitted beams is carried out through a microsystem-based reconfigurable free-space interconnect system which employs optical diffractive or reflective structures. This work also explores a novel hybrid integration protocol that allows self-aligned bonding of massive arrays of III-V components to Si electronics, and ensures low thermal budget and reduced stress.

Two-dimensional (2D) scanners can be used for displays, printers, optical data storage devices, optical scanning microscopes, and free-space optical interconnects. In this paper, we will describe the modeling and simulation of a novel cantilever microscanner. The scanner is actuated using electrostatic force. The cantilever beam connects to the top electrode. The bottom four electrodes on the substrate provide extra feedback for the control of the cantilever beam. A thorough mechanical analysis (both static and dynamic) using Finite Element Analysis has been performed. Key design parameters such as driving voltage, tilt angle and resonant frequencies have been investigated. The model has not been verified by experimental data but a fabrication process flow has been designed. The fabrication of this novel cantilever microscanner is in progress.

In this paper, we present a 40 Gb/s VCSEL driver (4 x 10 channels, 1 Gbps/ch) designed and fabricated in HYNIX 0.35 micrometers 2-poly 4-matel CMOS technology. The CMOS driver designed for a free space optical interconnect system consists of two NMOS for driving a VCSEL and protection circuit rejecting influence of electrostatic discharge (ESD) or unexpected input signal with several tens voltage amplitude. Two NMOS with CMOS channel length of 0.4 micrometers and width of 100 micrometers are used for adjusting dc bias current from 0 to 27 mA and ac modulation current from 0 to 13.8 mA. Protection circuit is made of two diodes. The purpose of the protection circuit is to permit the input modulation voltage range only from -5 to 5 V.

A continuously variable optical true time delay module, based on substrate-guided-wave and holographic optical elements, is designed and fabricated. The dispersion effect of volume holographic gratings is combined with a novel structure to obtain continuously variable time delay intervals from tens of picosecond down to sub-picosecond, which can be employed for K-band (18 GHz - 26.5 GHz) and Ka band (26.5 GHz - 40 GHz) phased-array antennas. The insertion loss of the packaged module is 10 dB, including fan-out loss and propagation loss, while the non-uniformity of insertion loss is within 0.6 dB among all channels. The crosstalk among channels is less than -40 dB. An optical true time delay module is designed to provide delay signals for -60 degrees to +60 degrees continuously steering of an eight-element K-band phased-array antenna system. Antenna field patterns are simulated with the change of optical wavelengths. The demonstrated architecture can be used in both the transmitting and the receiving modes and can be easily scaled up for large arrays.

We report both system design and experimental evaluation of SMOP (Simple and small Multi/demultiplexer consisting of OPtical elements) system used for intra-cabinet-level optical interconnections. To increase the interconnection bandwidth, wavelength division multiplexing (WDM) is an important technology. However, dense WDM is not suitable for the intra-cabinet-level interconnections due to its system complexity and expensive production cost. Instead, we chose wide-channel-spacing wavelength division multiplexing (WWDM) since it is expected to reduce fabrication cost by using uncooled lasers. The SMOP consists of three dielectric interference filters and a reflecting mirror, and aspheric lenses. The lens array pitch is set to be 0.25 mm to fit the optical fiber pitch at the end facet of MT connectors. Three dielectric interference filters are stacked to perform wavelength multiplexing and demultiplexing functionalities. An incident signal light from an MT connector with four different wavelengths (1280 nm, 1300 nm, 1320 nm, 1340 nm) is first collimated and deflected by an off-axis collimator lens. Three layers of dielectric filters differentiate different wavelengths by reflecting the incident beam at different locations. The SMOP is packaged as small as 6.4 mm by 2.5 mm with 8.1 mm thickness small enough for the intra- cabinet-level interconnections. The proto-type experiments confirmed low insertion loss of below 5 dB. The proposed wafer-level fabrication of SMOP may be effective in packaging such applications.

Widely tunable lasers are an enabling technology that eliminate many of the limitations of fixed wavelength lasers, and enable new system architectures with increased functionality. Beyond solving the logistical and inventory problems of growing channel counts, tunable lasers enable system architects to implement truly dynamic networks that can adjust to changing network demands in real time. This paper will discuss the general application of tunable lasers and enable functionalities such as real-time channel provisioning, dynamic add/drop and network channel reconfiguration.

Course Wavelength Division Multiplexing (CWDM) provides access to next generation optical interconnect data rates by utilizing conventional electro-optical components that are widely available in the market today. This is achieved through the use of CWDM multiplexers and demultiplexers that integrate commodity type active components, lasers and photodiodes, into small optical subassemblies. In contrast to dense wavelength division multiplexing (DWDM), in which multiple serial data streams are combined to create aggregate data pipes perhaps 100s of gigabits wide, CWDM uses multiple laser sources contained in one module to create a serial equivalent data stream. For example, four 2.5 Gb/s lasers are multiplexed to create a 10 Gb/s data pipe. The advantages of CWDM over traditional serial optical interconnects include lower module power consumption, smaller packaging, and a superior electrical interface. This discussion will detail the concept of CWDM and design parameters that are considered when productizing a CWDM module into an industry standard optical interconnect. Additionally, a scalable parallel CWDM hybrid architecture will be described that allows the transport of large amounts of data from rack to rack in an economical fashion. This particular solution is targeted at solving optical backplane bottleneck problems predicted for the next generation terabit and petabit routers.

The regeneration of optical data signals is an important requirement of WDM and OTDM networking schemes. All-optical regenerators based on cross phase modulation schemes currently require complicated systems of several optical components and modules together with complex control. Here we demonstrate an optical 3R regeneration in a single DFB laser. The key to this regeneration process is to gain switch the DFB laser with a clock signal extracted from the data signal in order to retime the converted signal. A data signal at 10 Gbits/s sent to 80 Km in a SMF with poor Q is reamplified, retimed and reshaped by this laser. However, the uncorrelated jitter (amplitude and phase noise) primarily present due to the spontaneous emission in the directly modulated transmitting laser as well as in the gain switched DFB laser (Regenerator) can very easily distort the whole regeneration process. In this paper we propose a systematic approach to tackle this problem to effectively enhance the performance of optical regenerator. An improvement of orders of magnitude in BER measurement can be achieved. The proposed approach results in a regenerated signal whose optical power, Q factor and extinction ratio are considerably improved by the regeneration process.

We fabricated and replicated in semiconductor compatible plastics a multichannel free-space optical interconnection module designed to establish intra-chip interconnections on an Opto-Electronic Field Programmable Gate Array (OE-FPGA). The micro-optical component is an assembly of a refractive lenslet-array and a high-quality microprism. Both components were prototyped using deep lithography with protons and were monolithically integrated using a vacuum casting replication technique. The resulting 16-channel module shows optical transfer efficiencies of 50% and interchannel crosstalks as low as -22 dB. These characteristics are sufficient to establish multichannel intra-chip interconnects with OE- FPGAs. The OE-FPGA we used was designed within a European co- founded MEL-ARI consortium, working towards a manufacturable solution for optical interconnects between CMOS ICs. The optoelectronic chip combines fully functional FPGA digital logic with the drivers, receivers and flip-chipped optoelectronic components. It features 2 optical inputs and 2 optical outputs per FPGA cell, amounting to 256 photonic I/O links based on multimode 980-nm VCSELs and InGaAs detectors.