The design of an optical image guide network for distributed multiprocessing is described. The network supports multiple high bandwidth rings between workstations over distances of 10s of meters. Traditionally, error and flow control functions for multiprocessor networks are implemented in custom high speed electronic Application Specific Integrated Circuits which are physically removed from the interconnect's physical layer. In this paper, we consider migrating these functions directly into the optoelectronic physical layer, yielding an 'Intelligent Optical Network'. Conventional error control protocols are infeasible with dense bit parallel optical systems based on image guides since they require excessive amounts of hardware. The designs of efficient error and flow control protocols for such networks are proposed and analyzed. The key blocks of the protocols have been designed, fabricated and demonstrated in 0.8 micron and 0.5 micron CMOS/SEED devices. The protocols require significantly less hardware then alternative schemes, and CMOS/VCSEL devices supporting these protocols are scalable to very high bandwidths, i.e., 10s of Terabits per second.

The series coupler access encoder (SCAE) applied to optical code division multiple access (CDMA) system generates the optical signals that cross-correlation of each user can be minimized according to the increase of the length of optical delay line connected among N couplers. A SCAE can act as either an encoding or a decoding structure depending on which terminals are used for the input/output of the data. This concept had already been published. However, in previous studies, the system performance was analyzed only considering the first order signals. The SCAE and SCAD have the high order signals of various patterns as the number of coupler increases, resulting in the change of auto- and cross-correlation intensities. Also, the number of users connected in an optical CDMA network could be limited since the side-lobe peaks are increased by the third order signals. In this paper, we analyzed the properties of optical matched filter considering the third order signals in optical CDMA network using SCAE and SCAD. We also estimate the variance of peak to side-lobe ratio in accordance with the increase of the number of couplers.

A novel approach for laser beam deflection using the thermal optic prism array in a polymeric planar waveguide is developed. This approach is based on the different thermo- optic properties between polymer and silica, the two optical materials employed for the guided wave beam deflector. A waveguide structure with the core layer composed of inversely positioned polymer and silica triangles forming a polymer/silica prism array has been fabricated. Through electrical heating, a temperature change results in an index difference between the two optical materials and creates an optical prism structure in the polymer/silica planar waveguide. A beam deflection of 5.4 degree was observed under a temperature change of 60 degrees C in the fabricated prism array. The sensitivity of the device is 0.09 degrees/degrees C. A maximal number of resolvable spots of 8 was achieved at a low driving power from the thermo-optic prism array structure. The accuracy of beam deflection approaches 16 micro-radian. The device fabricated has a thickness of 5 microns, a prism aperture width of 600 microns, and a device length of 7 mm. Optimal design to maximize the deflection angle and the number of resolvable spots has been evaluated.

This paper details several patented optical backplane interconnect system design for harsh environments. These optical backplanes provide an extremely high bandwidth interconnect structure capable of supporting today's and tomorrow's optical-electrical transducer technology, permitting the insertion of the next generation processing technology without upgrading/replacing the backplane interconnect structure itself. Additionally, the design of several robust, high optical density, cylindrical connectors is discussed, which extend the high-bandwidth optical connectivity beyond the backplane, permitting the integration of multiple physically federated 'boxes' into one computationally integrated system.

Optical interconnects are expected to overcome the limitations imposed by electrical interconnects. For board- to-board and board-to-multiboard communication we have developed an optical backplane for applications in mobile systems. Compared to existing realizations it is compact, rugged and has the potential to be fabricated at low cost. The main features of the optical backplane in planar technology are free space expanded beam transmission between boards and backplane and guided wave transmission within the backplane. No optical connectors are required. Due to the expanded beams and highly multimode waveguides large coupling tolerances of several 100 micrometers are achieved. Low loss polymer backplane waveguides allow transmission length of more than 19 inches. Demonstrators for board-to-board interconnections and for ring and star networks have been realized. Transmission experiments at 1GBit/s have been successfully performed. First environmental test with respect to dust, moisture, temperature and vibrations showed the feasibility of the concept.

Skew of an image fiber, which has more than ten thousands of cores in a common cladding, was measured by a novel measurement method for the first time. This method can measure the time-of-flight difference between individual cores over the whole area of an image circle. The measurement results reveals that a test 100-m-long image fiber has skew of 5 ps/m, and the time-of-flight distributes randomly in the whole area of the image circle due to nonuniformity of the core dimension. It is also experimentally shown that the skew of an image fiber increases by bending. The theoretical analysis reveals that the bending-induced skew depends neither on the radius of curvature nor the shape of the curve but it depends only on the number of turns it is wound. The numerical calculation of skew by using typical parameters of image fibers shows that the winding have to be restricted to less than 12.5 turns to achieve a transmission speed of over 1 Gbps/ch. Lastly we propose twisted image fiber and a 8-shaped bobbin to suppress the skew due to bending.

Guided-wave optoelectronic clock distribution networks on multichip modules (MCMs) are designed and fabricated to meet the high-speed clocking requirements of next-generation digital systems through a realization of superior network bandwidth, low power consumption, and large fanout capabilities compared to electrical interconnect counterparts. The sixteen-fanout H-tree clock distribution networks on MCMs is demonstrated by combining silica glass waveguides and micromachined silicon microstructures. The proposed optoelectronic multichip modules (OE-MCMs) with silica glass networks can be fabricated in a CMOS compatible batch process without modifying the conventional IC fabrication facilities. The proposed OE-MCM assembly/packaging processes are simple and cost-effective by sing self-alignment silicon microstructures. The design issues of optoelectronic H-tree networks and micromachined alignment structures for assembly/packaging processes are presented and discussed to improve the overall system performance while minimizing the system cost. The silica glass clock distribution networks and mosaic silicon mirror arrays are characterized at wavelengths of 1310 nm and 1550 nm by measuring overall optical power losses and signal uniformity.

This paper describes a 622-Mbit/s/ch, 12-channel optical transmitter and receiver designed to transmit any type of coding data with large tolerance to optical fiber. It incorporates a low power consumption and high speed 12- channel transmitter and receiver LSIs with automatic decision threshold control, a low threshold and highly uniform 12-channel 1.3 micrometers laser diodes array and a 12- channel InGaAs PIN photo diodes array, and a passive optical alignment technology. The applicability of newly developed 622-Mbit/s/ch 12-channel parallel optical transmitter and receiver has been demonstrated through a transmission experiment over 100m length of multi-mode optical fibers.

Parallel optical interconnects may become the communication method of choice to achieve future high bandwidth data transfer between chips or MCMs. For this purpose, an integrated CMOS detector approach is favorable at the light- reception side, so Flip-chip of detectors is no longer required. In this paper we present an integrated differential CMOS detector layout which gives a flat frequency response of 0.1A/W with a -3dB bit rate over 450 Mbit/s/ch in standard 0.6(mu) technology. The detector works following the SML-detector principle. Based on this SML-detector we fabricated a dense detector/receiver array consisting of 100 channels on one square mm Si area in 0.6(mu) standard CMOS. The detector area is 50 X 50(mu) ². The detector signal is amplified and latched by a self-regenerative sense-amplifier, which is self- calibrating for increased array homogeneity and receiver yield. The power consumption per receiver channel is an low as 1.1mW and the received light power at 200Mbit/s is 25.1(mu) W. Measured standard deviation on the output jitter is 96ps. Future CMOS technology will improve the maximum detector/receiver bit-rate as well as the attainable sensitivity.

We demonstrated the integration of 32 by 32 p-i-n photodiode arrays with Si-dummy chips for potential use in massively parallel short-distance optical interconnects. Individual devices in 32 by 32 InGaAs/InP photodetector arrays were successfully tested and demonstrated a small signal modulation speed above 10 GHz under dark condition, corresponding to an aggregate data transmission capacity in excess of Tera-bit/s.

The switching characteristic of wide-band MSM and PIN photodetectors has been studied in theory and experiments. MSM detector has threshold bias voltage to activate its response and so has a better performance than PIN photodetectors when working as a photo-electronic switch. However, our study tells that through a suitable designed bias circuit, the PIN photodetector also can provide a switching operation with considerable performance. Especially for RF photonic signal, the extinction ratio can reach around 30dB. At different bias condition, the gain of PIN can be continually tuned and it has very important application in photonic phased-array antenna system.

In this paper, we report on some transmitter and optical fiber technologies for 10 Gb/s speed datacom systems. Primarily to enable direct flip-chip integration with silicon CMOS circuits for highly parallel interchip interconnects, we have fabricated 2D vertical-cavity surface-emitting laser arrays featuring 4*8 elements on a 250 micrometers pitch. Arrays emitting in the 980 nm spectral region exhibit excellent homogeneity and operation data like 0.8 mA threshold current, below 1.5 V threshold voltage, 1 mW output power at 2 mA current, and more than 35 percent conversion efficiency in the 2 to 5 mW power range. Bottom emission of 850 nm arrays is enabled by etching holes into the opaque GaAs substrate. Due to less efficient heat removal, devices are currently limited to about 2 mW light output at 6 mA current. Top-emitting 850 nm array elements have been employed for high-speed data transmission experiments. First we describe the characteristics of a new 50 micrometers core diameter silica multimode fiber that allows for 10 Gb/s data rate operation over a record distance of 1.6 km. Finally, a 9 Gb/s transmission experiment over 100 m length of a rather promising perfluorinated plastic optical fiber for low-cost data links is presented.

In this paper we present the design, fabrication and characterization of a module which directly connectorizes a 1 by 8 red VCSEL array to a small diameter polymer optical fiber array, using a standard MT ferrule. The facets of the POF are prepared by a hot knife cutting, followed by a short polishing step. First coupling results show total losses in the range of 1.1 dB/channel for a 30 cm POF link. Optical crosstalk between adjacent channels is below -45 dB. Plastic micromachined parts surrounding the VCSEL chip ensure the correct alignment of the connector, using the connector, using the connector guiding points. The parts themselves are aligned to the chip with a n index-alignment technique, using an excimer laser ablated mastertool. In a deconnectorizable version of the module, a thin, flat glue layer on the chip acts as a window between the VCSEL chip and the MT terminated POF array. Integrated in a standard ceramic package, clear eye diagrams have been measured at 150 MHz for a 10m POF link, coupled to the VCSEL array. Further efforts on higher speed measurements using dedicated drivers, will also be presented.

In many VCSEL applications, it is essential to know the transverse beam characteristics. This paper reports an experiment of transverse modal characterization based on intensity measurement. The beam form a VCSEL is imaged by a microscope objective and intensity profiles are recorded by scanning an apertured detector. The second moment of the intensity profile is found to vary quadratically with the distance along the direction of beam propagation. An effective Rayleigh range is extracted by means of quadratic data fitting. Once this parameter is obtained, Fourier analysis of one intensity profile yields the relative weights of the Hermite-Gaussian (HG) modes, provided that the beam is indeed a superposition of independent HG modes. It is found that a VCSEL driven at low current generates a beam that is approximately HG or a superposition of independent HG modes. At high drive current, however, the transverse modal structure becomes more complicated. The experiment demonstrates that intensity-based Fourier analysis is a convenient method to assess the closeness of approximating the outputs form semiconductor lasers by superposition of independent HG modes without using sophisticated spatial modal filters. The experiment also measures the M² modes without using sophisticated spatial modal filters. The experiment also measures the M² parameter of beam quality versus the drive current.

This paper presents novel surface mount type fiber optic transmitters and receivers suitable for mass-production and high-density installation to the PCBs of the communication equipment. The transmitters and receivers are applicable for SONET/SDH equipment and fulfill the specifications of ITU-T G.957¹ and G.958² recommendations. Conventional fiber optic transmitters and receivers have a pigtail cord fasten to their housing. Such a pigtail cord is a constraint factor for reflow soldering and machine mounting of the transmitters and receivers to the PCBs in the communication equipment manufacturing process. Therefore, simplified optical connectors, the pigtail fiber cords can be easily attached to them, has been developed. A conventional feedback APC circuit needs classical variable resistors to adjust LDs bias current (Ib) and pulse current (Ip). This method requires manual adjustment and sites for the variable resistors. The developed fiber optic transmitters are equipped with a digital feed-forward APC circuit to compensate Idc and Iac versus the ambient temperature change. In this method a ROM built into LSI stores data for Idc and Iac. APC operation is carried out by loading the data for the ROM and giving the data to digital-analog converters. The fiber optic receivers have clock regeneration function utilizing the PLL technology instead of conventional band-pass filter. This scheme aligns the phase delay between received data and extracted clock at the optimum position, thus no manual circuit adjustment is needed.

InP/InGaAsP weakly confined wave guides, are of interest for the fabrication of a variety of optoelectronic devices, including optical cross point switches. Previous studies have shown that stress from metallization alters the behavior of these waveguides and the strain-induced changes in refractive index have been investigated by studying the degree of polarization (DOP) from the facets of ridge waveguide (RWG) lasers. However, in devices intended for waveguide routing or switching, e-beam deposited TiPtAu contacts, caused broadening and splitting of the optical near field mode profile after thermal annealing and increased waveguide loss. In this paper, we report the results of a study of process-induced changes on the optical near field mode profile, for different dielectric depositions, metallizations and anneal cycles. It was found that the addition of a layer of the TiPtAu. This not only eliminated the mode splitting, but for sufficiently high values of compressive stress, was found to actually sharpen the mode and enhance the guiding. On the basis of these results stable waveguides, suitable for routing or switching were achieved for a range of device parameters, operating conditions and process parameter variations.

This paper describes the two-layer scalable wavelength routing optical interconnection network being developed in Tianjin University. The top layer of the network is multi- wavelength bi-directional optical bus, which has high bandwidth and low latency. The optical bus is made up of passive components, no wavelength-tunable devices have been sued. As a result, the optical bus has low communication latency that is mainly decided by the optical fiber length. The sub-layer of the network is single-wavelength ring, which has low communication latency and high-scalability. In each ring, a wavelength routing node is used for data transmission between the ring and the optical bus. Each node computer is connected to the ring using an optical network interface card, which is based on peripheral component interconnect bus. The communication latency inside the ring is decreased using synchronous pipelining transmission technique. The scale of the ring is mainly limited by the efficient bandwidth required by each node computer. The number of rings is mainly decided by the optical power of the laser diodes and the sensitivity of the optical detectors. If Erbium doped fiber amplifier is used in the optical bus, the scale of the network can be further developed.

Electrical interconnects are reaching their fundamental limits and are becoming the speed bottleneck as processor speeds are increasing. A polymer-based interconnect technology was developed for affordable integrated optical circuits that address the optical signal processing needs in the telecom, datacom, and performance computing industries. We engineered organic polymers that can be readily made into single-mode, multimode, and micro-optical waveguide structures of controlled numerical apertures and geometries. These materials are formed from highly-crosslinked acrylate monomers with specific linkages that determine properties such as flexibility, robustness, optical loss, thermal stability, and humidity resistance. These monomers are intermiscible, providing for precise continuous adjustment of the refractive index over a wide range. In polymer form, they exhibit state-of-the-art loss values and exceptional environmental stability, enabling use in a variety of demanding applications. A wide range of rigid and flexible substrates can be used, including glass, quartz, silicon, glass-filled epoxy printed circuit board substrates, and flexible plastic films. The devices we describe include a variety of routing elements that can be sued as part of a massively parallel photonic integrated circuit on the MCM, board, or backplane level.

We propose a new concept of optoelectronic (OE) interconnect hardware 'OE Scalable Substrate (OE-SS)' and 'Film Optical Link Module (FOLM)', which have potentiality to remove optics excess. The structure is as follows: OE-films, in which waveguides, thin-film OE devices, LSIs, capacitor chips etc. are integrated with via/pad/electrode, are stacked by electrical joints (Z-connections). This gives rise to standardized-interface capability and scalability. Using one basic technology 'film/Z-connection', all levels of interconnection will be achieved, including massive parallel optical link, inter-board optical connect, and 3D- stack-OE-MCM. We prose a new process 'Device Integration with Self-Organizing Transfer', which is essential for low- cost OE-SS and FOLM, especially for WDM applications.

This paper reports our efforts to develop an optical True- Time-Delay line module for Phased Array Antenna applications using optical polymeric waveguides. We first give a brief description of a targeted phased array antenna, having chosen a 16-element sub-array as our demonstration system. Then we address the design considerations of the True-Time- Delay lines pattern based on the sub-array antenna's parameters, including simulations we have done to optimize the building blocks of the pattern: splitters, arcs' curvature, and crossings. Finally, we describe the steps of a modified fabrication process and present the primary result. Our experiment shows that the polyimide-based waveguide has a promising future because it has high fabrication precision and packaging density.

Conventional optical packaging technologies cannot provide low-loss coupling among various mode-mismatched optoelectronic devices in a cost-effective manner. Our approach based on polymeric waveguide technology provides a unique solution that can alleviate the existing coupling losses among various optoelectronics components, including optical fibers, laser diodes, and photodetectors. In this work, the 3D tapered waveguide coupler was simulated to match MT connectors. According to this MT stand-off design, we fabricated and tested waveguide couplers in molded MT-RJ fiber connect ferrules and integrated optical cross-connects with fiber MT-like connectors. Three fabrication technologies for low-loss polymeric waveguides are also discussed here.

Free-space smart-pixel optical interconnect architecture promises to relive the interconnect bottleneck in high-speed parallel interconnection and switching systems. Vertical cavity Fabry-Perot multiple quantum well modulators, which offers high on-off contrast, low insertion loss, low operating voltage swing, low chirp and easy to be integrated with silicon electronics ins one of the most promising devices. A low chirp vertical cavity Fabry-Perot modulator where the top and bottom mirrors are made of distributed Bragg reflector , the undoped regions between top mirrors and intracavity contact region contain 26 GaAs/Al_0.35Ga_0.65As quantum wells. By applying a voltage across the quantum well regions through intracavity contacts, the absorptive characteristics of the active region at the cavity's resonance wavelength can be modified through the quantum-confined Stark effect. The device under test is integrated with microwave probe for efficient wafer level GHZ probing without further high frequency packaging. To minimize the parasitic capacitance of the probe pads, the conducting region underneath probe pads is etched away and planarized with low dielectric material-electronics resins BCB. In this paper, we report the low switching voltage with only 3.5V, high contrast ratio of 10:1 and high bandwidth of 16GHz. The effects of incident laser power on the contrast ratio and modulation bandwidth were also discussed.

CMOS compatible optical polyimide based thermo-optic switches have the potential use as low-power switches. These switches would have many advantages over other switches based on inorganic crystals. For one, they can be integrated into module-to-module systems using currently available VLSI fabrication techniques. Polyimide based, 1 by 2 thermo-optic switches are fabricated onto silicon wafers and tested. We report the properties and characteristics of digital thermo- optic switches designed to operate at 1.3 micrometers . Also, the switching characteristics at different heating electrode voltages are tested and compared.

We present on the experimental results of the NpnP optoelectronic switching device with a significant optical sensitivity. This structure consists of a novel fully depleted optical thyristor (DOT) with a bottom mirror layers. We measure the emission efficiency for various sizes and injected currents. This device shows 20 percent and 45 percent enhancement in switching voltage change and spontaneous emission efficiency, respectively, which is very important for the sensitivity and the low power consumption of DOT.

A growable multistage ATM switching architecture based on optical interconnect is presented in this paper. The interconnect stage, core of the 3 stages architecture, is composed of 16 by 16 CMOS-SEED optoelectronical hybrid ATM switching modules. Since the interconnect stage is memory- less, electronic buffers are provided in the output concentrator stage, and the buffers are partial shared to be used effectively. Optical interconnects between the pair- input expansion stage and the interconnect stage, also the interconnect stage and the output concentrator stage provide high-speed data paths, for example 622Mb/s or 2.4 Gb/s. Both the with lower speed control signal and the complicated logical processing are carried out in the electronic devices. With 64 ports of OC-12 interface, the maximum throughput of the prototype system is about 40 gigabits per second, an the packet loss ratio of this ATM switching system is less than 10e-9. Taking advantages of high speed of the optical interconnect and the high density, flexible logical processing of the electronic devices, the ATM switching of the optical interconnect and the high density, flexible logical processing of the electronic devices, the ATM switching system has favorable potential to scale easily to very large network size, for example 256 ports of OC-48 interface.

In this paper, wavelength routing technology is applied to computer interconnection network. By analyzing the relationship between wavelength and networks routing, we describe a concept of wavelength being used as network address, and propose a wavelength routing topology to extend the scale of networks and realize scalability of networks. A twin-wavelength ring network to implement and test the function of wavelength routing is presented, and the main units of the twin-wavelength ring network are presented as well. In addition, we put forward tow methods to implement it. The design method of physical link driver software for Linux, which uses optical interconnection interface as network interface for parallel computing, is also introduced.

Steered Agile Beam (STAB) technology is a revolutionary concept for the development and employed of laser devices that perform in a manner similar to phased array radar. In effect, a STAB device, like the phased array radar, will be able to transmit multiple beams simultaneously. This will permit STAB to acquire and track multiple objects simultaneously; a STAB device will be able to direct its laser beams in different directions without having to move the device. Consequently, STAB will be able to acquire, track, and communicate with multiple objects within its field-of-regard. There are several battlefield applications that would benefit significantly from STAB technology. Some of the tactical areas that may use STAB technology include communications, target designation, optical countermeasures, target acquisition and tracking, identification friend or foe, and battle damage assessment. The DARPA is sponsoring the STAB development program, and during the next few years will be evaluating the work of several contractors in bringing the technology to the user.

Free-space optical transmission provides large bandwidth, small size, lightweight, low cost and good security. Diffuse IR link configuration is also rather robust against shadowing. Its disadvantages are, however, bandwidth degradation due to multipath dispersion, sensitivity to ambient light and limited transmission distance due to the limitations of optical power budget. To specify the bandwidth and power budget requirements of the diffuse link, we performed ray-trace simulations for different room geometries and dimensions, and different transmitter and receiver locations. We considered both diffuse and specular reflections as well as shadowing and reflection effects due to blocking objects, such as furniture. The simulations were verified by analytically calculating the impulse response in simple diffuse reflection geometry. We also analyzed stray light induced shot noise effects. Furthermore, we simulated some properties of a quasi-diffuse link comprising of multi- beam transmitters with restricted beam divergences as well as detectors with narrow fields of view. Based on the study, novel Monte Carlo ray-tracing software packages, such as ASAP, can be used for diffuse link multipath dispersion and optical power path loss analysis. Ray tracing can also be used for parallel channel crosstalk and stray light analysis. Potential applications for these system are high- bit-rate wireless LANs and free-space optical interconnects.

The design and experimental characterization of the substrate-mode guided wave optical backplane using 0.5mm, 0.75mm and 1mm pitch 2D optical beam array are described. The system sues arrays of multiplexed holograms to implement board to board interconnects and employs 250micrometers pitch 2D vertical-cavity surface-emitting lasers and microlens array as transmitter to provide 0.5mm to 1mm pitch 2D beam array, operating at 850nm. By comparing the optical beam properties at the detector plane including the spot size and power uniformity of optical beam array, as well as signal to noise ratio, the acceptable design parameters are justified. Furthermore, the improvement of throughput that can be achieved by 1D and 2D crosstalk analysis within the same design concept is presented. The results of crosstalk analysis obtained here are used for application to the standard five-board holographic optical backplane system.

A new free-space multistage optical interconnection network which is called the Comega interconnection network is presented. It has the same topological construction for the cascade stages of the Comega interconnection. The concept of the left Comega and the right Comega interconnection networks are given to describe the whole Comega interconnection network. The matrix theory for the Comega interconnection network is presented. The route controlling of the Comega interconnection network is decided based on the matrix analysis. The node switching states in cascade stages of the 8 by 8 Comega interconnection network for the route selection are given. The data communications between arbitrary input channel with arbitrary output channel can be performed easily.

A low power 1 X 12 array of VCSEL driver that each runs at 1.32 Gbps has been designed and fabricated in 0.25 micrometers CMOS. The transmitter array operates with minimal crosstalk, and with edge times as low as 59 ps, and power consumption less than 5mW per transmitter. Zero-bias vertical-cavity surface-emitting lasers are considered, and the trade-off between speed and power consumption with be discussed. The edge time for this transmitter can be adjusted with a novel speed-up circuit. The application of this speed-up circuit will be discussed for short-haul optical links that uses multimode fibers. The drivers have been successfully integrate on a single chip with temperature compensation circuits, clock generation phase-lock loop (PLL), serializing and 8B10B encoding, and deserializer and 8B10B decoding circuits, and a 1 X 12 receiver post-amplifier array.

We report the fabrication and processing of thin film MSM silicon photo-detector and thin film VCSEL for optoelectronic interconnects. These two components, together with polyimide wave-guide can be used in constructing the high sped, low power, low cost optical interconnection system. Such a system will provide the fast board level data transmission. The DC and AC characteristics of thin film silicon MSM photo-detector, and the I-V and L-I characteristics of thin film VCSEL are measured.

As processor speeds rapidly approach the Gigahertz regime, the disparity between process time and memory access time plays an increasing roll in the overall limitation of processor performance. In addition, limitations in interconnect density and bandwidth serve to exacerbate current bottlenecks, particularly as computer architectures continue to reduce in size. To address these issues, we propose a 3D architecture based on through-wafer vertical optical interconnects. To facilitate integration into the current manufacturing infrastructure, our system is monolithically fabricate din the silicon substrate and preserves scale of integration by using meso-scopic diffractive optical elements (DOEs) for beam fan-out. We believe that this architecture can alleviate the disparity between processor speeds and memory access times while increasing interconnect density by at least an order of magnitude. We are current working to demonstrate a prototype system that consists of vertical cavity surface emitting lasers, diffractive optical elements, photodetectors, and memory units integrate don a single silicon substrate. To this end, we are currently refining our fabrication and analysis methods for the realization of meso-scopic DOEs. In this paper, we present our progress to date and demonstrate through-silicon optical data transmission using DOEs that were designed, fabricated, and characterized at the University of Delaware. We present the validation of our theoretical models for the design of such DOEs with experimental data and discuss applications for our proposed architecture including instruction level parallel processors and field programmable gate arrays.

We have designed and assembled two generations of integrated micro-optical systems that deliver pump light and detect broadband laser-induced fluorescence in micro-fluidic separation systems employing electrochromatography. The goal is to maintain the sensitivity attainable with larger, tabletop machines while decreasing package size and increasing throughput. One type of micro-optical system uses vertical-cavity surface-emitting lasers (VCSELs) as the excitation source. Light from the VCSELs is relayed with four-level surface relief diffractive optical elements (DOEs) and delivered to the chemical volume through substrate-mode propagation. Indirect fluorescence from dye- quenched chemical species is collected and collimated with a high numerical aperture DOE. A filter blocks the excitation wavelength, and the resulting signal is detected as the chemical separation proceeds. Variations of this original design include changing the combination of reflective and transmissive DOEs and optimizing the high numerical aperture DOE with a rotationally symmetric iterative discrete on-axis algorithm. We will discuss the result of these implemented optimizations.

The significantly increasing transmission capacity in the subscriber system are accelerating demands for higher data transmission speed, low cost, and system miniaturization in the millennium approaches. The key solution for the demands, smaller, lower cost, and high performance package, is surface technology. OptoBGA package has been developed to meet this challenge, a surface mountable type optical package that is designed for high frequency transmission speeds. OptoBGA package has a BGA structure to bring forth effectively size and cost reduction for the packaging. OptoBGA consists of ceramic material, a robust material, which has many advantages such as design flexibility with fine design rule, ease of process technology, high performance, and high reliability. The design of OptoBGA was made possible by improvement in the physical structure. The results of this research produced an OptoBGA with a return loss of about -20dB at 10GHz. Furthermore, reliability test evaluations have been conducted to demonstrate that the OptoBGA can withstand harsh environments.

A novel concept has been developed for a fiber optic switch. It is based on miniaturized prisms deflecting the nearly collimated beam from the input fiber. These prisms are moved by the use of piezoelectric bending actuators. The deflected beam is directed to the output fiber, there is one prism for each output fiber. Coupling of the deflected light beam into the output fibers is achieved by microlenses. In order to simplify the integration process regular arrays of both the output fibers and the coupling microlenses have been used. A special technique and setup has been developed for fast and accurate adjustment of the lens array with respect to the fiber array. Both arrays are subsequently fixe to each other by gluing. The second critical integration process is the mounting procedure of the prisms to the bending actuators. For this purpose a special vacuum gripper has been built. With the help of this gripper all microprism can be mounted in one step. Using the developed integration processes, fiber switch prototypes have been built up with excellent optical parameters.

A low cost optical transmitter/receiver module is essential for optical subscriber systems. Hybrid integration of optical components and alignment free assembly are effective techniques for such low cost module. We newly developed a bi-directional optical transmitter/receiver module using passive alignment techniques for a laser diode (LD) mounting on a Planar Lightwave Circuit (PLC). Compact module size of 26mm X 10mm X 3mm is achieved. A spot-size converted laser diode is passively mounted on a PLC chip by detecting each alignment marks on the LD and the PLC chip. Low optical coupling loss of less than 4.5 dB is attained. The PLC chip has the function of Y branch. Alignment marks for LD mounting and a V-groove for fiber alignment are also fabricated on the PLC chip. A fiber is self-aligned on the V-groove. The PLC chip and the other components such as a sub assembled dual photo diode and a receiver-amplifier are also built in a ceramic package. The output power of the transmitter/receiver module is +2dBm at the LD drive current at 45mA. Responsivity of over 0.35A/W is realized. Minimum optical received power of -38dBm at the 50Mb/s burst signal was obtained.

Polymer microlens fabrication techniques that enable easy integration with VCSELs are presented. We have developed a high-tolerance coupling structure with microlenses formed on both sides of the optical components for inter-chip optical interconnections, and have developed two types of drop-on- demand techniques for producing microlenses: an ink-jetting method and a dispensing method. Both methods use the surface tension of liquid UV-curable epoxy polymer. We have fabricated various microlenses that have a geometrical diameter from 20 micrometers to over 1 mm with F/1 to F/12 by controlling the volume and viscosity of the droplets and their wettability to the substrate. The measured uniformity in arrayed lenses was within +/- 1 percent in lens diameter and +/- 3 micrometers in pitch. Examples of how we have integrated microlenses with VCSELs are also presented. An ink-jetted microlens ona VCSEL coupled to a single-mode fiber enabled highly efficient coupling: 4 dB greater than without the microlens. A dispensed microlens on a VCSEL coupled to a multimode fiber increased the coupling efficiency by 20 dB compared to without a microlens. In the multimode case, large tolerances of +/- 2 mm in axial misalignment and +/- 10 micrometers in lateral misalignment were obtained for a coupling loss increase of 1 dB.

Silica-based photonic integrated circuits (PICs) have been making major advances and are finding increasing applications in optical communications, networking, and signal processing. For the next generation of photonic integrated circuits, it is desirable to add more functionality as well as to increase the integration level. This would involve introducing a variety of heterogenous materials and devices on the same substrates, using a monolithic and/or hybrid integration method. In this paper we describe the results of our efforts of developing/incorporating new functions to the silica-based integrated circuits. 1) Optical amplifiers, suitable for monolithic integration with other guided-optic devices, are promising as loss-compensating devices for photonic integrated circuits. 2) Silicon is the most commonly used substrate for silica-based PICs. A novel method has been developed for forming 2D waveguides on silicon substrates, utilizing the photoelastic effect in Si induced by thin-film stress. This method does not require any separate guiding layer nor etching of silicon, and therefore is expected to increase the flexibility in designing/implementing advanced PICs on Si. 3) Ferroelectric materials possess various functional properties and are expected to play an important role in advanced PICs. The major challenges and progress are discussed in achieving monolithic integration of functional films, such as PZT, on silica and Si substrates.

Graded index fiber has a limited bandwidth due to defects introduced in the fiber manufacturing process. In this paper, an alternative launch technique is presented using a diffractive element to excite specific modes in a fiber to maximize the bandwidth of graded index fiber.

The efficiency of laserdiode-monomode fiber coupling is one of the essential assignments for interactive optical telecommunications. In this paper, we present our investigations and results of the development of such couplers, especially of a soldering process for hybridly joining a laser on a silicon microoptical bench by self adjustment. To achieve the necessary adjustment tolerances of better than 0.5 micrometers , the influence of the solder height for the self-alignment process is simulated with respect to the lowest allowable solder thickness, which still generates the necessary alignment forces. Because of the excellent properties at their coupling facet U-grooves were fabricated to mount the fiber by deep and high precision anisotropic reactive ion etching of silicon using SF₆/CBrF₃ gas- mixtures. The solder was deposited in buried structures by electroplating gold and tin at the eutectic ratio to allow optimum adjustment in vertical and lateral dimensions. Furthermore, the results of a fluxless soldering process in a UV-light activated forming gas atmosphere is presented and compared to the simulations. The measurement of the coupling efficiency to fibers will also be illustrated. The design of the microoptical bench allows to extend the system by incorporating e.g. a faraday rotator or other optoelectronic integrated circuits, like transceivers.

The assembly of fast-axis collimating lenses (FAC-) with high power laser diode (HPLD) bars and stacks is discussed. FAC-lenses are cylindrical lenses of high numerical aperture and short focal length for the collimation of the strongly divergent radiation emitted perpendicular to the active region of a high power laser diode bar. These FAC-lenses are necessary for all kinds of beam transformation systems using HPLD bars and stacks. The lenses have to be fixed at the heat sink of the laser diode bar with an accuracy of < 0.3 micrometers and for special transformation system with a definite tilt of 1.2 degrees and an accuracy of better than mrad. In most cases an active assembly of the whole beam transformation system is necessary. Here we propose another approach.

The microjet printing method is being used to fabricate microlens arrays for use in massively parallel, VCSEL-based datacom switches and to deposit lenslets of various configurations onto the tips of single-mode telecom fibers. Applications in the latter case include collimation of the output beams for free space optical interconnection and increasing the fiber numerical aperture for collection of light from edge-emitting diode lasers. Additional applications of this technology include point of arrays of active sensor elements onto the tips of imaging fiber bundles and fabrication of microlenses with axial index of refraction gradients to reduce focal spot size, utilizing multiple print heads with differing fluids. This low-cost, data-driven process, based on 'drop-on-demand' inkjet technology, involves the dispensing the placing of precisely sized microdroplets of optical material onto optical substrates. The micro-optical elements are printed with 100 percent solid, UV-curing optical epoxies, utilizing printing devices that can dispense picoliter-volume droplets at temperatures up to 300 degrees C.