Monolithic, uniformly-wavelength-graded, vertical-cavity surface- emitting laser (VCSEL) and wavelength-selective resonance- enhanced photodetector (REPD) arrays are key enabling technologies for many wavelength-multiplexed optical network and interconnect architectures. These arrays can be produced in a repeatable and uniform manner by controlling the local MOCVD growth rate of the epilayers on a topographically patterned substrate, which resulted in wavelength-differentiated devices with more uniform optical characteristics. Multi-wavelength VCSEL and REPD arrays have been monolithically integrated on the same substrate to improve the wavelength matching between the source and detector elements. The performance characteristics of these arrays are discussed. We also describe a wavelength-division- multiplexing and demultiplexing experiment in which several channels of optical data are multiplexed together using a multi- wavelength VCSEL array and transmitted on a single optical fiber, and are demultiplexed at the other end using a wavelength- selective REPD array with closely-matching wavelengths.

We present an array of GaAs-based light-emitting diodes emitting at 850 nm, which is designed to be flip-chip joined on carriers like a silicon CMOS circuit. The light-emitting diodes are grown by MBE. After processing of the array and flip-chip joining using indium-bumps, the substrate of the LED array is removed completely. Individual light-emitting diodes reach an external quantum efficiency of 3.3% after the complete process.

Arrays of interdigitated metal-semiconductor-metal photodetectors carried on thin semiconductor membranes have been transferred and grafted by epitaxial liftoff techniques to non-conducting AlN host substrates for advanced multichip module development purposes. Progress includes the introduction of full surface passivation by thin polyimide films resulting in dramatically reduced dark currents for GaAs photodetectors. Device modeling by means of a computer simulation has suggested new interdigitated electrode configurations which can only be applied to these types of devices carried on thin film membranes. Integration of InGaAs photodetector devices on the same host substrate with transimpedance amplifiers provides an intermediate development prototype of the first stage of a high speed multichip module receiver product for fiber optic telecommunications applications. The use of integrated circuit-sized membranes to carry small discrete devices, broad-area inverted photodetectors, OEICs, focal plane arrays, and multi-spectral sensors continues to provide a manageable path to automated packaging using commercial off-the-shelf hardware.

We present a novel smart pixel composed of an optoelectronic threshold switch with gain and a vertical cavity surface-emitting laser (VCSEL). In this smart pixel two surface-normal input optical beams control an output optical beam emitted by the VCSEL. In its present hybrid version the VCSEL-based smart pixel is capable of opto-optical switching with an output contrast ratio in excess of 30 dB at an optical power of about 1.5 mW. For quasi-stationary operation we achieve an optical gain of up to 3 X 10⁵. We also report drastic improvements on the switching dynamics. Operating the receiver with an optical input power of 130 (mu) W we achieve bitrates of up to 160 Mbit/s and an optical gain of 11, while optical inputs of 410 (mu) W result in a maximum bitrate of 400 Mbit/s and an optical gain of 3.6. The minimum input optical energy required for switching is 765 fJ, the AC output contrast ratio is 9 dB. Optically performed NAND and NOR logic operations are demonstrated at 40 Mbit/s with a fan-out of 7.6. We further show that the functionality of this smart pixel can easily be extended to electric read-out of input optical data and to direct electric control of the VCSEL within the smart pixel configuration. In particular, we demonstrate conversion of electric input to optical output data at 1 Gbit/s.

The performance imitations have been studied for various optical bistable symmetric self-electrooptic effect devices (S-SEEDs) using extremely shallow quantum wells (ESQWs). We consider the four ESQW SEEDs; anti-reflection (AR)-coated ESQW S-SEED, back- to-back AR-coated ESQW S-SEED, asymmetric Fabry-Perot (AFP)-ESQW S-SEED, and back-to-back AFP-ESQW S-SEED. As the optical input power increases, the ohmic heating dominantly reduces the on- state reflectivity, while the exciton saturation dominantly increases the off-state reflectivity, both of which seriously degrade the device performances limiting the maximum input power for fast switching. For the quantitative estimation of the limitation, we simulate and analyze the bit-rates of the optically cascaded set-reset latch systems employing the various ESQW S-SEED. The optimized number of ESQWs for each structure is also obtained.

We have experimentally identified and characterized two dynamic components of a VCSEL/multimode optical fiber link: (1) a high temporal frequency fluctuation of the VCSEL's numerical aperture (NA) and, (2) source power spectrum shift in time (chirp) both upon ac modulation. We use these variables of the excitation irradiance distribution at a fiber input to simulate the resulting modal noise; without the typical assumptions that all fiber modes are equally filled and all relative fiber mode delays are equally probable. Results indicate that fluctuations in the NA can generate significant modal noise at the fiber output. On the other hand, source chip, in the presence of mode selective loss, is only moderately effective in generating modal noise. Moreover, since chirping increases with the VCSEL's series resistance, it's significance will diminish as design efforts to decrease the potential drop across the VCSEL are successful.

Planar monolithically integrated GaAs-on-silicon photoreceivers consisting of a metal-semiconductor-metal (MSM) photodiode and GaAs metal-semiconductor field-effect transistor have been fabricated. The performance of GaAs on Si MSM photodetectors has been characterized and the dark current was measured and compared to the conventional p-i-n photodiodes. The investigated GaAs on Si photoreceivers were subjected to thermal accelerated stress testing up to 200 degree(s)C and duration of 1000 hours, in order to study their thermal stability. A failure mode has been determined to consist of an increase of dark current and a decrease of photocurrent as a function of aging time. The degradation mechanism is shown to be the interdiffusion in the Au/GaAs interfaces and the degradation of the photosensitive surfaces after temperature stress testing.

The technical approach and progress achieved under the Polymer Optical Interconnect Technology (POINT) program are described in this paper. The POINT program is a collaborative effort among GE, Honeywell, AMP, AlliedSignal, Columbia University, and University of California at San Diego (UCSD), sponsored by DARPA/ETO, to develop affordable optoelectronic packaging and interconnect technologies for board and backplane applications. Specifically, progress is reported on (a) development of a plastic VCSEL array packaging technology using batch and planar fabrication, (b) demonstration of high-density optical interconnects for board and backplane applications using polymer waveguides to a length of 50 cm at an I/O density of 250 channels per inch, (c) development of low-loss optical polymer waveguides with loss less than 0.1 dB/cm at 850 nm, and (d) development of passively alignment processes for efficient coupling between a VCSEL array and polymer waveguides. Significant progress has also been made under the POINT program at Columbia University, in applying CAD tools to simulate multi-mode-guided wave systems and, at UCSD, to assist mechanical and thermal design in optoelectronic packaging. Because of space limitations, these results will be described elsewhere in future publications.

We report for the first time, 3D tapered polymeric waveguides fabricated by the compression-molding technique. The waveguides presented herein provide a feasible solution to bridge discrete optoelectronic devices having the apertures of a few microns to hundreds of microns. One-cm long tapered channel waveguides with the cross-sections of 5 micrometers X 5 micrometers at one end and 100 micrometers X 100 micrometers at the other end were fabricated. These waveguides have a propagation loss of 0.5 dB/cm when the 632.8 nm He-Ne laser light is coupled from the small end and of 1.1 dB/cm when coupled from the large end. By confining the energy to the fundamental mode, when coupling from large end to the small end, a low-loss packaging can be achieved bidirectionally.

This paper addresses the design, fabrication, and packaging issues of optoelectronic clock distribution networks on multichip modules (MCM) by combining guided-wave optoelectronic interconnect, MCM packaging, and microelectromechanical system fabrication techniques. The proposed prototype employs the silica glass waveguide networks on silicon substrates and innovative I/O coupling method which utilizes the micromachined through-holes across MCM substrates and silicon mirror arrays. Microstructures for I/O coupling have been fabricated using KOH etchants at various processing conditions and the silica glass optical waveguide networks are fabricated using FHD and RIE processes. The four fanout optoelectronic clock distribution networks are prototyped and characterized by measuring the overall insertion losses and the power equalities at the fanout nodes.

Embedding of end-tapered, thin-cladding fiber bundles for board- level large bandwidth optical clock distribution is proposed and implemented. Fan-outs of 1 to 64- and 128-nodes on a printed circuit board of area 13 X 9 cm² are experimentally demonstrated. Dispersion measurement shows a 30 picosecond pulse broadening over 30 cm length of the fiber, thereby indicating multi Gb/s clock delivery capability. Power coupling efficiency of 3 dB with coupling nonuniformity of 4.7 - 5.5 dB is observed.

We demonstrate a four channel integrated wavelength division multiplexer (WDM) and demultiplexer (WDDM) based on volume holographic gratings and substrate-guided waves at near IR wavelengths. The four operating wavelengths are centered at 750, 780, 810 and 840 nm respectively. The WDM and WDDM are demonstrated using 50/125 multimode fibers. The channel-to- channel crosstalk level is measured to be less than -40 dB. The system insertion losses are -23 dB, -21 dB, -20 dB, -22 dB respectively for 750 nm, 780 nm, 810 nm and 840 nm.

An electro-optic polymer with photolime gel and chlorophenol red dye is prepared using a newly developed mixed solvent of water and ethyline-glycol. This new solvent not only prevents the EO chromophore from re-crystallization, but also help the guest and host materials to mix well in the molecular level. As a result, high EO coefficient of 40 pm/v has been obtained with improved stability of EO effect. A new pulsed poling technique is developed, which increase the poling efficiency up to 35% compared with the traditional parallel-plate poling method.

In this paper, we discuss multihop optical bus interconnection networks. For massive parallel computing systems, such interconnection can interconnect a large number of processing elements and reduce system latency at the same time. We give the qualitative results of De Bruijn bus network and Kautz bus network as an engineering options based on combinatorial hypergraph theory and a uniform alphabetic constructive method. We also give an implementation method via polymer based waveguide holograms.

As fiber optics becomes a standard means of transmitting long haul telecommunication signals. costs of adopting fiber solutions for various other communication applications steadily decrease. Optics starts to challenge copper-based solutions in many short-distance data communication domains. Recent progress in developing vertical -cavity surlaceemitting lasers (VCSEL) has lead to a belief that a proper combination of the VCSEL and fibers can provide a cost-effective solution to many bandwidth-demanding local data communication bottleneck problems. A trend to develop VCSEL-hased fiber optic parallel data links has been witnessed [1-2]. Among the most notable approaches are the parallel fiber link prolect by the Opto-electronic Technology Consortium (OETC) [1]. the Optobus project by the Motorola Corp [21. These, along with many other similar, approaches feature 1D array of bit-parallel connections of optical data between a ID VCSEL array and a ID photodetector array. A lD fiber ribbon containing various fiber channels is used as flexible parallel data channels. As demand for bandwidth rapidly increases, wider and wider 1D ribbons containing more and more parallel channels are being proposed. To cope with future bandwidth demand, a direct extension of the 1D approach to a 2D form has also been studied. 2D VCSEL arrays are under development. Methods of fabricating small 2D arrays of individual fibers are being explored [341. It was noted that unless active alignment methods are successfully developed, it is difficult to control core pitch fluctuations of such a fiber array. A somewhat different 2D approach was recent proposed [5]. It uses a fiber image guide (FIG) which contains a large number of closely packed fibers to transmit the image of the VCSEL to the other side. A FIG simply images between a 2D input and output arrays. Among mentioned advantages of this approach are its capability to transmit a scalpel number of parallel channels and its relaxed alignment condition. In this talk, we quantitatively study various interface problems this FIG-based approach may experience and try to understand its capabilities and limitations.

Surface relief gratings with tilted grating profiles have been studied for efficient optical signal coupling for polymeric waveguide based optoelectronic interconnects. In this article the theoretical formulation used to calculate the diffraction efficiency in a pre-defined direction is discussed. The designed gratings provide an effective unidirectional surface-normal optical coupling for polymer-based optoelectronic interconnects. The gratings with a tilt-angle of 32 degrees and periods ranging from 0.5 to 3 micrometers were fabricated. Both surface-normal input and output grating couplers have been demonstrated in polyimide waveguides on silicon substrate.

We have been developing a special purpose parallel processor system with a compact size and a high performance specific for the scientific computation. A multichip module (MCM) with a high speed bus is indispensable to achieve such parallel processor system. We have developed a high speed bus for such MCM using a polyimide optical waveguide with micro-mirrors. We can easily obtain many output signals from one input signal in this optical waveguide. Such polyimide optical waveguide with micro-mirrors is formed on a MCM substrate which is called an optical plate. An optical coupling between the polyimide optical waveguide and the laser diode or the photodetector on LSI chips is performed through the micro-mirror. We observed twenty output signals for the photodetectors from one input signal in the polyimide waveguide of fabricated test module. We have also evaluated the performance of parallel processor system using such MCM with the polyimide optical waveguide by computer simulation.

Performance of polarization dependence of photopolymer-based substrate-guided-wave optical interconnects with surface-normal configuration is characterized. Energy optimization, and trade- off between energy equalization and polarization insensitivity of 1-to-many cascaded interconnects are discussed. DuPont photopolymer film HRF 600X001-20 is employed in our experiment. A 1-to-9 fanout device with +/- 5% energy fluctuation for a randomly-polarized input optical wave at 632.8 nm is fabricated for evaluating the trade-off under p- and s-waves. As the Myler substrate of the photopolymer films has strong effects on changing incident linearly polarized light into elliptically polarized light, the energy uniformity of a practical device can be improved significantly. The developed theory is applicable for any volume hologram.

A new device concept for fabricating polymer-based reconfigurable electro-optic interconnects based on a guide-wave electro-optic beam deflector in conjunction with multiplexed waveguide holograms is presented. The presented device takes fully advantage of a thin-film electro-optic prism array and multiplexed waveguide holograms, capable of reconfiguring more than 50 optical interconnection channels using a single electrode and a single electrical driving source. No any moving component and DC bias are required. The presented device is advantageous whenever there is a need for high-speed, reliable, compact, los- loss and low-power reconfigurable optical interconnection.

In order to utilize the high speed of optical interconnects and overcome the latencyproblem of a large bus structure, we proposed an architecture of optoelectronic hierarchical bus system. Waveguide hologram implementations, and its associated cache coherence problem are addressed. In our configuration, bus hierarchy is controlled with electronic programmable plates. Optical signals can be transmitted on all-optical paths without intermediate conversions.

The decomposition of 2D shuffle multistage interconnection networks without loss of functionality is presented and analyzed. The major costs for the decomposition arise by the interconnection of the subdivided switches. The attempts considered differ with regard to the implementation of the dimension-dependent switches (hybrid and all-optical) and with regard to the organization of the interconnection scheme (hierarchical and direct). These attempts are described, evaluated and compared. The application of the decomposition to distributed switching/processing is briefly illustrated.

A Code V simulation of a free-space optoelectronic data distributor, commonly known as the Kaleidoscope, is presented. The device is designed to interconnect processing elements in massively parallel processors, and the simulation results indicate that the device is capable of interconnecting 64 processing elements with a 64-bit word in a fully connected topology. The results further suggest that the number of processing elements, in the future improved version of the device, may be increased to 1024 and beyond.

This paper reports progress toward the experimental demonstration of a smart pixel based optical interconnection prototype currently being developed under the Free-space Accelerator for Switching Terabit Networks (FAST-Net) project. The prototype system incorporates 2D arrays of monolithically integrated high- bandwidth vertical cavity surface emitting lasers (VCSELs) and photodetectors (PDs). A key aspect of the FAST-Net concept is that all smart pixels are distributed across a single multi-chip plane. This plane is connected to itself via an optical system that consists of an array of matched lenses (one for each smart pixel chip position) and a mirror. The optical interconnect system implements a global point-to-point shuffle pattern. The interleaved 2D arrays of VCSELs and PDs in the prototype are arranged on a clustered self-similar grid pattern with a closest element pitch of 100 micrometers . The circular VCSEL elements have a diameter of 10 micrometers and the square PDs have an active region that is 50 micrometers wide. These arrays are packaged and mounted on circuit boards along with the CMOS driver, receiver, and FPGA controller chips. Micro-positioning mounts are used to effect alignment that is consistent with current MCM chip placement accuracy. Shuffled optical data links between the multiple ICs have been demonstrated in preliminary evaluation of this system. These results suggest that a multi-Terabit optically interconnected MCM module is feasible.

A linearized external electro-optic modulator with a number of parallel Mach-Zehnder interferometers cascaded via directional couplers is proposed. Fast-Fourier-Transform method is used to analyze both the third harmonic spurious signal and the third intermodulation distortion. By adjusting the number of Mach- Zehnder interferometers contained in the modulator, the linearity of the transfer curve of the modulator can be modified while obtaining nearly 100% optical modulating depth. The modulator is based on an innovative domain-interval electro-optic polymer film, where the two arms of each Mach-Zehnder interferometer are inversely poled. As a result, only one driving voltage is required upon a pair of uniform traveling-wave electrodes, for high-speed operation.

Multi-layer switching architectures, where the layers and the parallel waveguides form N-gon prisms (N is the number of waveguides), represent a novel class of all-optical switches which (1) approach shuffle results through their 2 X 2- switches are interconnected in a nearest-neighbor manner and moreover (2) improve shuffle results by increasing the connectivity (equalsVnumber of 2 X 2-switches) between the prisms. However, as multi-layer switches may be difficult to implement by the current technology of photonics, the paper provides a first step towards the planar realization of this class of multi-layer switches.

Optoelectronic integrated circuits (OEICs) containing 32 X 32 arrays of smart pixels on 512 micrometers centers are under development for use in high density input/output (I/O) applications. Each smart pixel will comprise a monolithically integrated detector, a CMOS circuit, and a flip-chip bonded, oxide confined GaAs/AlGaAs vertical cavity surface emitting laser (VCSEL). The CMOS circuit will include a receiver, a digital section, and a laser driver. The digital sections of adjacent pixels will be connected in series to provide high speed 1D and 2D optical I/O into the smart pixel array. This low-power architecture will combine functionality, testability, and electrical/optical access in a single compact circuit. The OEIC will serve as a robust optoelectronic device and also as a test-bed upon which to develop techniques for flip-chip bonding the GaAs/AlGaAs VCSELs onto CMOS circuitry. The smart pixel device will operate at a speed of up to 1 GHz. The 32 X 32 array will consume < 10 W.

A packaging technology for global (N⁴) free space optical interconnects has been developed and implemented by means of hybrid integration using solder engineering. Solder arrays, defined by photolithography and electrodeposition, are used to physically attach, optically align and electrically connect the microoptical components. A processing sequence has been fully developed for the fabrication of the emitter sub-assembly consist of 8 X 8 and 16 X 16 arrays of vertical cavity surface emitting lasers and diffractive optical interconnect elements. The first level of integration consisting of the metal signal lines, ground plane and solder arrays has been completed with excellent uniformity and yield of the deposited solder.

In this paper, in order to properly promote usage's of optics for future interconnection applications, we revisit some fundamental physics governing light propagation in linear media. In particular, we wish to address how the point-topoint data communication capacity is related to the degree of freedom (DOF) of optics, and how packaging and other constraints limit the DOF of optics for exercising its power of interconnections. We hope that a correct understanding of the fundamentals of optics for interconnections can help maximize its advantages while bypassing its drawbacks.

The design, assembly and packaging of a multichip module on silicon with a holographic-optical 1 to 4 interconnect is presented. The potential of the system containing commercially available components is demonstrated. Holographic optical elements in dichromated gelatine as well as computer-generated holograms are used to distribute and focus a clock signal generated by a vertical-cavity surface-emitting laser diode onto silicon photodiodes. The two different kinds of holograms are compared with respect to their efficiency, fabrication effort and compatibility with the demands of a microelectronic system. Alignment issues, thermal behavior and the total optical link efficiency are discussed.

A robust polymeric waveguide technology is proposed for affordable optoelectronic interconnects in massively parallel processing applications. We have developed high-performance organic polymeric materials that can be readily made into both multimode and single-mode optical waveguide structures of controlled numerical aperture and geometry. These materials are formed from highly-crosslinked acrylate monomers with specific linkages that determine properties such as flexibility, toughness, loss, and stability against yellowing. These monomers are intermiscible, providing for precise adjustment of the refractive index from 1.30 to 1.60. Waveguides are formed lithographically, with the liquid monomer mixture polymerizing upon illumination in the UV via either mask exposure or laser direct-writing. A wide range of rigid and flexible substrates can be used, including glass, quartz, oxidized silicon, glass-filled epoxy printed circuit board substrate, and flexible polyimide film. Our waveguides are low loss (0.02 dB/cm at 840 nm) as well as temperature resistant (over 65 years at 100 degree(s)C) and humidity resistant (no effect on unpackaged guides after 600 hours at 85 degree(s)C 85% RH), enabling use in a variety of demanding applications. We discuss the use of these materials on multi-chip modules, boards, and backplanes. Waveguiding structures measuring tens of inches in length can be produced on backplanes, and guides that are meters long can be laser-written on rolls of plastic. We also discuss the fabrication of symmetrically-clad flexible strips of waveguide arrays that are compatible with MT- type connectors.