Short distance optical interconnects, which has long promised to solve a number of problems assumed to exist in electronic systems, seems to be at the threshold of providing real value to future processing systems. In many cases, it has been the cost of optical interconnects which has prevented their widespread adoption. However, while data rates required in future computing systems are increasing, and fiber optic component costs have decreased, the distances over which optical interconnects become attractive have been shrinking. In addition the increasing demand for smaller footprints for the I/O at the card edge, plays into one of the strong advantages unique to optics. An effort called Optical Micro-Networks is underway, funded by DARPA, whose objective is to demonstrate cost effective system level benefits in parallel optics for intra- and inter-rack interconnects by leveraging recent advances in VCSEL arrays, high sped CMOS, and low cost glass fiber cabling, connectors and on-board fiber routing. The approach will integrate optical interconnect functionality directly into an ASIC package thus reducing size, cost, and power.

We report the formation of polyimide-based H-tree waveguides for a multi-GBit/sec optical clock signal distribution in a Si CMOS process compatible environment. Such a clock distribution system is to replace the existing electronic counterpart associated with high-speed supercomputers such as Cray T-90 machine. A waveguide propagation loss of 0.21 dB/cm at 850 nm was experimentally confirmed for the 1-to-48 waveguide fanout device. The planarization requirement of the optical interconnection layer among many electrical interconnection layers makes the employment of tilted grating a choice of desire. Theoretical calculation predicts the 1-to-1 free-space to waveguide coupling with an efficiency as high as 95 percent. Currently, a coupling efficiency of 35 percent was experimentally confirmed due to the limited index difference between guiding and cladding layers. Further experiments aimed at structuring a larger guiding/cladding layer index differences are under investigation. To effectively couple an optical signal into the waveguide through the tilted grating coupler, the accuracy of the wavelength employed is pivotal. This makes the usage of the vertical cavity surface-emitting lasers (VCSELs) and VCSEL arrays the best choice when compared with edge-emitting lasers. Modulation bandwidth as high as 6 GHz was demonstrated at 850 nm. Such a wavelength is compatible with Si-based photodetectors.

Conventional electrical bus system technologies are encountered their physical borders in terms of transmission length and parasitic capacity of lines, and thus, many optical interconnection technologies are proposed to solve the problems in the electrical circuits. In this paper, we propose a new optical bus with sheet like shape technology using diffusion controlled light transmission in the multi- planar waveguides that makes capability of broadcasting, simultaneous bidirectional transmission, easy alignment and plural multiplex transmission. We have fabricated a basic model using optical sheet bus technology for the first time and evaluate them at 500MHz clock operation as follows: (1) synchronous broadcasting transmission to control transmission delay; (2) simultaneous bidirectional transmission in the same waveguide; (3) multiplex transmission such as time division multiplexing, wavelength division multiplexing and so on. Through this fabrication and evaluation, we can pick up some approach to put the optical sheet bus technology into practice with feature as follows: (a) simple design rule for multi-accessible bus systems like as shared memory bus systems, which needs faster speed; (b) added-value of multiplex transmission ways that reduce circuit frequency, minimize power consumption, maximize transmission bandwidth, and bring about no-wait communications for the bus systems.

A novel optical interconnect architecture based on temporally accessed spectral multiplexing (TASM) technology is presented. The use of TASM optical interconnects will be demonstrated in multiple access telecommunication applications. TASM technology builds on optical code- division multiple access concepts with promise of low cost implementation of Terabit/second optical networks. The use of TASM technology to enhance wavelength division multiplexing in optical communication systems will be discussed. The working principles and design of TASM system architectures will be presented with emphasis on the key functional component - the TASM encoder.

Limited bandwidth because of too few and too slow external pins is one of the major problems in current VLSI systems. Increasing clock rates and the growing transistor density in future microprocessor will enlarge the imbalance between satisfying computing power and insufficient communication performance. Optoelectronic VLSI (OE-VLSI) circuits using highly dense 3D optical interconnections offer the potential to overcome these problems. To lead OE-VLSI processing to success it is necessary to point out a diversity of architectures that profit extremely from a 2D optical input/output interface. Such architectures have to be developed especially for an optoelectronic solution. We demonstrate this for various architectures like binary neural associative memories and fine-grain 3D processor cores for integer and digital signal processing. We specify the electronic circuits and the optical interconnection schemes. We found out that an optoelectronic approach for the associative memory offers two orders of magnitude more performance than all-electronic solutions. The stacked 3D integer processor offers a performance increase of about 10 to 50 over current RISC processors. For the realization of the OE-VLSI circuits we developed a CMOS-SEED chip and a smart detector test chip consisting of CMOS circuitry monolithically integrated with a silicon based array of photo diodes.

We review recent progresses in various POF related optical interconnect projects at NEC Research Institute. These projects were initiated to study cost-effective sub-local optical interconnect solutions that can potentially compete with copper-based solutions in future. The projects include both coupling schemes for networking various POF data links, and point-to-point polymer circuits which can route data in both serial and parallel formats. Many of the mentioned projects are still on-going and thus the report here has the nature of progress update.

It is demonstrated that the bandwidth limitation of multimode fiber links arising from modal dispersion can be overcome by using passband modulation in the frequency range beyond the 3dB bandwidth. Transmission of 1.25 GBit/s over 500m of near-worst case 62.5micrometers multimode fiber is achieved using two subcarrier multiplexed channels at the 850nm wavelength. In contrast the conventional baseband transmission distance is limited to 300m at this data rate.

The concept of a bi-directional optical backplane with multi-bus lines for a high performance system containing multi-chip module boards, operating at 850 nm, is introduced. The backplane reported here employs VCSEL arrays and photodetector arrays as transceivers. Optical beams from a transmitter are fanned out by deflectors which are arrays of multiplexed polymer-based waveguide holograms, and then undergo total internal reflection inside a waveguiding plate, and finally reach the target receiver arrays on the edge of boards. We have designed a bi-directional optical backplane for the data communications application in a nine- board system. Packaging related issues such as misalignment, crosstalk and signal to noise ratio have been studied for the reliable system based on the Gaussian beam approximation. By integrating 140 micrometers -pitch or 250 micrometers - pitch VCSEL arrays, microlens arrays, and photodetector arrays into our design of backplane, we have demonstrated optical backplane with multi-bus lines and experimentally realized this architecture with 1D and 2D bus lines which greatly increase the bus bandwidth of the backplane. Frequency responses of our devices shows a bandwidth of 2.5 THz, which is higher bandwidth. Eye diagrams up to 1.5 GHz have been demonstrated with clear eyes.

It is well known that a wavelength can be regarded as an address code of a packet in the optical interconnection network. In our scheme, we use wavelength as a part of an address to build a multiple wavelengths optical interconnection ring network. The network consists of double layer rings, the routing inside a ring is electronic, and wavelength routing technology is applied to switch between rings. Two wavelength are used in the network, one is for transmitting data inside local ring, another is for switching to another ring. The detail of the network is described in this paper. Besides, a low-cost virtual parallel optical link and optical interconnection interface, which is bound up with the multiple-wavelengths network, is presented.

Clustering supercomputer nodes requires a large amount of bandwidth in as small an area as possible. The distance between nodes can be quite varied, and interconnects covering 30 meters between nodes are not uncommon. Current advances in parallel optic interconnects allow these devices to be given consideration to replace the more traditional electrical and copper cable interconnect. However, supercomputer communication boards often need to support and route 1,000 to 1,200 communications lines. We build and demonstrate a parallel optic interconnect module with a bandwidth density of 15Gb/s/inch² and an optical connector housing supporting 16 MT ferrules to satisfy the needs of a system communication module.

The next generation giga-Hertz microprocessors require a high-speed clocking scheme with a robust and low-skew clock distribution network. The guided-wave optoelectronic clock distribution networks on multichip modules (MCMs) is believed to satisfy the high-speed clocking requirements by providing superior network bandwidth, low power consumption,and large fanout compared to the electrical interconnect counterpart. In this paper, we report a sixteen-fanout H-tree clock distribution network on MCMs, which utilizes silica glass waveguides and micromachined silicon microstructures. The proposed optoelectronic multichip modules (OE-MCMs) can be fabricated in a CMOS compatible batch process without modifying the conventional IC fabrication facilities and the OE-MCM assembly/packaging processes are simple and economical due to the arrays are characterized at wavelengths of 1310 nm and 1550 nm. The issues of system-level modeling and prototyping are also addressed by discussing VHDL/FD-BPM-based simulation tool: Optoelectronic System Simulator, such that OE-MCMs can be rapidly-designed and mass-produced through the design automation tools.

Guided-wave optics is a promising way to deliver high-speed clock-signal in supercomputer with minimized clock-skew. Si- CMOS compatible polymer-based waveguides for optoelectronic interconnects and packaging have been fabricated and characterized. A 1-to-48 fanout optoelectronic interconnection layer (OIL) structure based on Ultradel 9120/9020 for the high-speed massive clock signal distribution for a Cray T-90 supercomputer board has been constructed. The OIL employs multimode polymeric channel waveguides in conjunction with surface-normal waveguide output coupler and 1-to-2 splitters. Surface-normal couplers can couple the optical clock signals into and out from the H-tree polyimide waveguides surface-normally, which facilitates the integration of photodetectors to convert optical-signal to electrical-signal. A 45-degree surface- normal couplers has been integrated at each output end. The measured output coupling efficiency is nearly 100 percent. The output profile from 45-degree surface-normal coupler were calculated using Fresnel approximation. the theoretical result is in good agreement with experimental result. A total insertion loss of 7.98 dB at 850 nm was measured experimentally.

We report the demonstration of a compact laser-beam deflector based on electro-optic prisms formed within a thin-film polymer waveguide. We fabricate planar waveguides using a polymer that can be readily poled and cured through the simultaneous application of a poling voltage and heat. The index of refraction of each prism in the cascade, but not of the surrounding polymer, is modulated by the electro- optic effect through the application of a drive voltage. A laser beam, to be deflected, is coupled into and out of the planar waveguide by cylindrical lenses. The application of a drive voltage creates a sequence of prisms in the planar waveguide, which change the path of propagation of beam through the planar waveguide with a variable angle of refraction depending upon the voltage. The deflection efficiency is observed to be nearly 100 percent and the laser beam maintains its Gaussian intensity profile after propagating through the device.

Electrooptic polymer modulators demonstrate ultra-high frequency response to 1 1 0 GHz at chip level. To interconnect these integrated polymer waveguide devices to fiber-optic networks and to fully use their ultra-wide bandwidth capacity, a packaging technology that includes optical interfaces to standard fibers, electrode transitions to wideband coaxial cable connectors, and a stable mechanical enclosure is required. This paper summarizes TACAN's efforts in high-speed electrooptic polymer modulator chip interfacing package. Fiber-to-waveguide pigtailing, mode mismatch loss reduction, and the optimization of extinction ratio are important considerations for optical interfaces. Due to materials limitations and the low driving voltage requirement, the waveguide mode size is often much smaller than that of a standard fiber which results in large mode mismatching loss. A fiber mode size conversion coupler has been developed to reduce the mode mismatching loss. For microwave interfaces, the emphasis is on the electrode to coaxial cable connector transitions that include the use of commercial components and the test of connectorized circuits. The packaged electrooptic modulators were tested for their performance characteristics. Connector-to-connector frequency response of the electrode circuits, analog and digital signal transmissions, long-term halfwave voltage stability, and DC bias stability have been tested using packaged electrooptic modulators.

Interferometric optical switches have enabled high-speed gates suitable for demultiplexing applications while still allowing the use of slow recovering optical materials. We present a new form of switch that lends itself to a compact format using reflective switching elements. This is an important distinction form previous designs in that planar reflective elements can now be incorporated into this class of switches which were previous designs in that planar reflective elements can now be incorporated into this class of switches which were previously considered unsuitable. In addition, the form of this switch is compatible with parallel processing architectures.

Telecommunications transport networks have to face the fast development of Internet-based data communications. Wavelength Division Multiplexing (WDM) is a key enabling technology to satisfy this huge increase of traffic. In order to improve the flexibility and the survivability ofthese WDM transport, the introduction of optical routing techniques is foreseen during the next decade. In this context, the availability of reliable and cost-effective optical cross-connects (OXC) is a critical issue. Two main functions may be performed in these nodes, namely the spatial routing of the wavelength demultiplexed channels and the wavelength conversion of these channels. Many architectural options can be envisioned for OXC'3. In all cases, the performance and the cost of the space-switching technology used in OXC are key factors for the realization of these nodes. Among the different technologies under development for space routing in OXC the actually most advanced are optomechanical and thermo-optic switches4. But, in spite they are commercially available, the capacity of these devices is strongly limited. In this context, free space routing is an attractive solution because it could enable to benefit from the large theoretical capacity provided by two-dimensional beam deflection techniques. Moreover, the holographic approach which has yet proven its efficiency5 can be performed by using nematic liquid crystals (NLCs) gratings which may operate with infra-red wavelengths. These components take advantage ofthe rather good maturity ofNLC technology thanks to the display industry, including recent progresses in high resolution microdisplays. We present here our results in the perspective of developing large capacity switching matrices based on high efficiency twodimensional liquid crystal gratings. After a short review of the present state-of-the-art of electrically addressed liquid crystal gratings, we describe the approach followed in order to achieve accurate and continuous deflection angles, which is a strong requirement when interconnecting single mode fibers. We demonstrate that parallel aligned NLC devices can procure the blazed-profile phase needed for high efficiency routing. This approach is supported by experimental results obtained both with transmissive or reflective cells. The problem of the switch dimensioning with either 1 -D or 2-D deflectors, is then addressed. We emphasize the requirements in terms of electrical interconnection technology arising from the large number of electrodes (at least 1,000 I dimension), this analyze leads us to suggest the LC/VLSI technology. Then, we present our work for the development of a 8x8 capacity system based on 1-D deflection, including the adaptation of the LC process on Silicon and test-circuit evaluation results.

2D networks of directional couplers (DCs) and Mach-Zehnder interferometers for of all optical switching in the 3D physical space are presented and arising problems and their solutions are made transparent. The result also applies to optical signal processing. The all-optical 2D switching networks, based on 4-gon switches, are compared of 2 by 2, 3 by 3 and 4 by 4 couplers and nearest-neighbor interconnections where the crucial problem is the implementation of >= 3 by 3 couplers. However, various advantages are expected (1) combination of the massive parallelism of optics and the parallelism in the frequency domain (2) considerable reduction of the number of stages (3) novel properties which are caused by the increase of freedom contributed by the additional space coordinate and (4) an increase of the range of possible applications.

We have recently described a wavelength-recognizing switch (WRS) which we showed to be capable of truly all-optical routing. Although other authors had previously reported "all-optical" networks [1, 2, 3], the term has generally referred to all—optical data paths only. In such implementations only the data remains in optical format as it propagates along the network paths. Optical-to-electronic conversions are still allowed for what is termed "control signals," namely address bits and additional signaling entities, which are assumed to have lower speed than the data. In contrast, the WRS we presented has the capability to route data by interpreting the control signals in the optical domain, thus avoiding the overhead and the latency ofthe optic-to-electronic conversion. In a series of previous publications [4, 5] we demonstrated the experimental viability of the WRS and measured some of the relevant system parameters of the device. More recently, we published details on how the device could be used to build multistage all-optical self-routing networks. We also developed a simulation model and estimated the maximum number of stages that can be cascaded in such networks. The work presented in this paper carries the simulation results one step further, and investigates some of the possible topologies that can be used for WRS networks, as well as the system implications of these topologies. To facilitate a better understanding of the concepts presented, we briefly review the device functionality and the main experimental results that affect the system performance. We then show how to implement some of the building blocks we use in the self-routing topologies, and explain the equalization mechanism necessary for using WRS in multistage networks. We then compare the practical advantages of the topologies of interest and decide which topology is the most probable implementation. Finally we present details on a new generation of WRS, which is waveguide based, thus more easily fabricated and integrated with optical fiber systems

Problems arising throughout the design of 3D lightwave circuits are briefly addressed in the following. (1) The implementation of M X N-gon switches. (2) The routing of M X N-gon prism switches and (3) the routing of all- optical 3D girds and (5) dual switches. (6) The improvement of the switches towards the crossbar and (7) the generalization of all-optical grids to any N >= 3.

Electrooptic polymer-based modulators have been investigated intensively due to their potential applications in optical communication systems. In this paper, we report a polymeric modulator with a domain-inverted Y-coupler configuration. Both of the modulation depth and linearity were improved due to the novel device structure. The Y-coupler modulator was automatically set at 3dB point with no need of DC bias, which eliminate the DC drift phenomena in Mach-Zehnder or co-direcitonal modulators. At the same time, a domain- inversion poling technique was developed, which can be used to fabricate other type of active EO devices in the future.

High optical loss due to mode mismatch at the interfaces of different components in a hybrid photonic integrated circuit (PIC) poses a major challenge in the implementation of such devices. Increased coupling efficiency can be achieved by incorporating an optical mode converter at the interface. This converter basically consists of a tapered waveguide section adapting different modal spot-sizes. Optimized coupling requires total control of the transverse optical field. This can be achieved by the ability to shape both the vertical as well as the lateral waveguide dimension. We present design, simulations, and fabrication considerations for a 3D tapered waveguide structure for low loss mode conversion. Our mode converter concept is based on polymeric optical waveguides on silicon substrate. A gradually deeper trench is formed in the silicon substrate, using diffusion limited wet etching with a laterally tapered mask pattern. The structure is then planarized with a polymer and patterned laterally. Our method thus allows control of both the lateral and vertical waveguide dimensions. The concept is consistent with low-loss coupling to singlemode fibers as well as between laser and amplifier arrays and single mode waveguides in a low-cost hybrid PIC solution.

This paper presents the current status of a new class of liquid crystal material being developed for latching electrooptic applications. This new material has the unique property of being electrooptic and fully latching. That is, in one state, the material has the properties of a conventional liquid crystal, capable of being aligned with either an electric or magnetic field; in its other state, it is an optical quality solid that maintains the molecular alignment set while in the fluid state. Experiments have shown that current materials can be switched on the order of milliseconds, as is the case with conventional nematic liquid crystals. In the solid state, the electric field can be removed with no change to the previously set optical properties because the molecular alignment is frozen in place, which should last for an extended period of time. In addition, the material exhibits broad temperature stability in the solid state, enabling devices to be developed that operate from cryogenic temperatures to 80 degrees C without the use of a temperature controller. This new material is ideally suited for applications where the size and mechanical robustness of an electrooptic device is desired, along with the latching capability of optomechanical devices. This materials technology alone will currently not meet high-speed switch requirements. However, this technology can be integrated with other state-of-the-art high-speed materials to provide a high-speed latching device. Devices currently under investigation using this materials include optical switches, optical attenuators and tunable filters.

Photonic phased array antennas represent one of the most critical technologies for both national defense and civilian wireless communications. In this paper, we present a novel compact detector-switched polymeric waveguide true-time- delay module, which is a crucial building block for advanced wideband photonic phased array antennas. The photolithographically defined ultra-low-loss polymeric waveguides provide us an attractive solution for achieving ultra long delay time over tens of nsec with ultra fine resolution of less than 1 ps. The 2D distributed waveguide grating couplers tap the optically encoded microwave signal, propagating along the polymeric channel waveguide, to high- speed photodetectors. These photodetectors can be electrically switched on and off independently for selecting different delay times. The appropriate delay time is equal to the time of flight along the waveguide. We have demonstrated that such optical true-time-delays can be implemented in such a device scheme with the RF spectrum of 11 GHz to 40 GHz. The optically encoded microwave signals are obtained by using semiconductor-laser-based optical heterodyne technique. Such a monolithic integrated module not only reduces the cost associated with optoelectronic packaging, but also reduces the system payload with an improved reliability.

A fully packaged opto-electronic cross-connect interconnect fiber circuit for data communication applications is demonstrated. Using bit-parallel optical link technology, the circuit offers EO and OE conversion capability, as well as 100 X 100 cross-connectivity. Using this technology together with an array of 10 X 10 digital electronic switches, fast cross-connect switches of the scale of 100 X 100 can be implemented. The insertion loss of the compact 100 X 100 cross-connect interconnect circuit ranges from 0.4 dB to 2.9 dB among all possible connections. Optical transmissions with bit-error-rate of < 10^-12 can be maintained at near 1 Gb/s channel bandwidth when the circuit is powered by the OPTOBUS chips. The system is expected to have an aggregated interconnect bandwidth near 100 Gb/s when being fully connected with the OPTOBUS chips.

With the rapid expansion of short haul data communication networks and the conversion of long haul telecommunication networks from voice to IP based data protocols, an optical technique to combine these two networks as well as other mode mismatched networks and components must be developed. Since the interface of these networks will be with-in centralized routers/hub/switching stations which involves multiple channels/optical fibers, a low loss cost effective solution is required for a multi-channel physical fiber connection between the networks. Currently one standard multi-channel fiber connector, the (MT) mechanically transferable, is widely used on both types of networks and fortunately shares with one exception, the same physical fiber ferrule from factors that could facilitate the physical interface between these networks. To compensate for the one exception a tapered waveguide that is physically matched to each network fiber core type has been package to facilitate low loss coupling with MT connectors. A technical description follows for packaging tapered waveguides supplied from Radiation Research Inc. into a form suitable for connection between a single and multi-mode MT connector. This form will be referred as a MT stand-off.

Using a beveled-edge to coupled optical signals into a waveguiding plate, a path-reversed substrate-guided wave holographic interconnection is employed for a WDDM with a channel wavelength spacing as small as 2 nm. A waveguide grating with a 45 degree incident angle and a 45 degree diffraction angle is fabricated using 20 micrometers thick DuPont photopolymer film HRF 600 X 001. The dispersion and the 3 dB bandwidth of the device are measured to be 0.18 degrees/nm and 20 nm, respectively. A four-channel wavelength demultiplexing is demonstrated at 796 nm, 798 nm, 800 nm and 802 nm with no crosstalk observed. A one-to-five cascaded 4-channel WDDM with +/- 5 percent energy uniformity under an s-polarization is also demonstrated to increase the user-sharing capacity. Twenty fanout channels are experimentally achieved.

Polymer thermooptic waveguide taps have a potential application as light routers for guided wave optical interconnects involving cascaded fanouts. The taps can guide light form an optical bus bar and direct it into other devices in a switching/modulation network. Thermooptic waveguide taps are designed and fabricated on silicon wafers using standard VLSI fabrication techniques. Coupling of light into an adjacent waveguide tap is observed to increase by 12.3 percent from 38.7 percent to 51.0 percent with the application of 34 mW of power.

The issue of interfacing holographic memory with an electronic processor is discussed. The high speed and parallel access of 2D, page formatted optical data from holographic memory can be utilized to reconfigure an electronic processor at a rate much faster than traditionally available. This new technique could be the stepping stone to a new class of high performance device for a variety of image/signal processing tasks. We will first give a review of the holographic memory activity at Holoplex, in particular, our research on holographic optical disk as a read-only memory device. We will then discuss the optical architecture for interfacing an optical ROM with a programmable gate array processor.

Novel optical memory systems offer ultra large storage capacity and with fast access time. The current commercial system can produce in access of 300 Mb/s aggregate data rate and near future system will yield aggregate data rates on the order of 1-10 Gb/s. However, full exploitation of this feature is possible only if memory to processor interface is fast enough to handle such a data rate. In this presentation, a unique optoelectronic interconnect architecture based on WDM and WDDM are described.

In this paper we introduce a micro-optical architecture that uses meso-scopic diffractive optical elements (DOEs) as 3D interconnects in a memory system for high level instruction level parallelism (ILP) processors. By using meso-scopic DOEs we can rescue the scale of integration to the VLSI scale, ie.e., the micron scale, achieve submicron alignment tolerance, improve reliability due to monolithic integration, facilitate integration into the current manufacturing infrastructure, and offer the ability for higher bandwidth and high interconnect densities. To this end we are developing the component technologies needed to realize this system. In this paper we present our work in the development of a theoretical and experimental framework for the design and characterization of meso-scopic DOEs, preliminary experimental results of meso-scopic beam splitters, and a large scale demonstration of the ILP memory system.

A calibration-free 64-Gbps/board parallel optical interconnection subsystem was mounted directly on the 4-CPU processor board of the optically connected parallel- processing machine RWC-1/ON. The subsystem was compared of eight pairs of an array optical transceiver model and a one- chip link LSI. Even though it used multimode fiber parallel optical modules, parallel data could be transmitted over 1 km because of the de-skewing and synchronization action of the link LSI.

An image fiber based optical space code division multiple access system is an important candidate for future 2D parallel optical interconnects. We clarify the advantages of the image fiber in terms of parallelism, channel density, alignment, skew control and so on, by comparing it with conventional parallel optical transmission media such as a fiber ribbon. We also compare the image fiber with the other candidates for 2D parallel optical interconnects, such as 2D fiber arrays and lens systems. We then show that 2D parallel optical interconnects using an image fiber are applicable not only to 1-to-1 optical interconnection, but also to more complicated network topologies connecting many processing devices. We describe key technical subjects for the image fiber network including an image fiber coupler, an alignment method, and a multiplexing scheme. An experiment is reported for each subject. Space-CDMA for the multiplexing scheme is explained in particular detail, and the advantages of this scheme over other schemes such as time division multiplexing and wavelength division multiplexing are clarified.

We report here micro-transmission measurements of a highly selective and a widely tunable optical filter at 1.55 micrometers . This filter is formed by a (lambda) /2 air cavity sandwiched between two 3.5 alternances InP-air Bragg mirrors. Reflectors and air cavity are fabricated by selective micro- machining of InGaAs sacrificial layers. Measurements have shown a very large wavelength tuning of 108nm for only 4.7V and a filter linewidth of 0.6 nm. Such performances have been possible because of the high reflectivity, large bandwidth and good optical confinement of the InP-air Bragg mirrors.

Circular grating surface emitting distributed Bragg reflector lasers exhibit extremely small divergence angles and moderately high output power levels which makes them an attractive candidate for free-space interconnects. Computational simulations are used to model the observed results and assist in the design of new functionality such as focusing and beam shaping. Numerical analysis of the coupled mode equations for DBR lasers and free-space propagation of the emitted field have been performed. Surface emission is realized through grating outcoupling provided by a second order grating while feedback is accomplished with an inner annulus consisting of either a first order or second order grating. These gratings are written using electron beam lithography (EBL) and subsequently etched using ECR-RIE. The flexibility of the EBL process enables a variety of different grating designs to be created on a single sample and subsequently compared during device testing.

A detailed design and fabrication procedure of high-speed traveling-wave electrodes for EO polymer-based modulator has been developed. Design consideration, thick photoresist deposition and electroplating are specially focused on. A lot of practical experiences are introduced as well. This kind of modulator can be used in satellite receiver systems, remote connection of cellular radio systems, and LANs.

Oxide-confined vertical cavity surface-emitting laser diodes (VCSELs) are fabricated for applications in chip-level optical interconnects. 980 nm wavelength devices in arrays with 4 by 8 elements are investigated. Threshold voltages of 1.5 V and operation voltages below 2V of submilliamp threshold current lasers are fully comparable to 3.3 V CMOS technology. Modulation bandwidths of 9.5 GHz at 1.8 mA laser current with a modulation current efficiency factor of 10 GHz/(root)mA is demonstrated for 3 micrometers diameter VCSELs. No error floors are observed down to bit error rates of 10^-11 at 12.5 Gb/s data transmission. VCSEL based top illuminated resonant cavity enhanced photodetectors show peak efficiencies of 50 percent combined with full spectral half-widths of 5 nm.

We report a surface-micromachined electrostatic deflector as a controlling element for a novel micro-opto-electro- mechanical (MOEM) bandpass tunable filter for a wavelength- division-multiplexed optical-fiber sensor system. Such a tunable MOEM filter involves multiplexed volume holograms, a surface micromachined electrostatic deflector, and an array of Si photodetectors. The micromachined electrostatic deflector, providing fast and repeatable adjustments, greatly enhances the dynamic tuning range of the filter. To micromachine the electrostatic deflector, we use thick photoresist as a sacrificial layer that is patterned using conventional microlithography, followed by electroplating. Ten deflectors with different lengths have been fabricated, and the electrostatic actuation of each device has been demonstrated. DuPont photopolymer film is employed for forming multiplexed volume holograms, which in conjunction with a photodetector array will allow the filter to operate at many wavelength windows. The incorporation of the multiplexed volume hologram with the electrostatic deflector will allow us to tune dynamically the filter.

A characteristic feature of a conventional von Neumann computer is that computing power is delivered by a single processing unit. Although increasing the clock frequency improves the performance of the computer, the switching speed of the semiconductor devices and the finite speed at which electrical signals propagate along the bus set the boundaries. Architectures containing large numbers of nodes can solve this performance dilemma, with the comment that main obstacles in designing such systems are caused by difficulties to come up with solutions that guarantee efficient communications among the nodes. Exchanging data becomes really a bottleneck should al nodes be connected by a shared resource. Only optics, due to its inherent parallelism, could solve that bottleneck. Here, we explore a multi-faceted free space image distributor to be used in optical interconnects in massively parallel processing. In this paper, physical and optical models of the image distributor are focused on from diffraction theory of light wave to optical simulations. the general features and the performance of the image distributor are also described. The new structure of an image distributor and the simulations for it are discussed. From the digital simulation and experiment, it is found that the multi-faceted free space image distributing technique is quite suitable for free space optical interconnection in massively parallel processing and new structure of the multifaceted free space image distributor would perform better.

Optical interconnects are being used increasingly in military systems. Sensor platforms in particular have seen great benefit from the use of fiberoptics as a replacement for coaxial cables. Optical interconnects is continuing to find new application at shorter and shorter distances throughout the sensor platform. High density free space optical components will extend the use of optical interconnects down to the inter-chip level. However, optical interconnect technologies can do more than just transmit data at high speeds. The ability to simultaneously process and transmit optical data, as could be done in optical signal processing, gives optical interconnects the ability to inject new capabilities into sensor system on-board processing. In order to see actual field deployment, these new technologies will need to be rigorously tested.