A new technique for automatic alignment of small optical components is presented. It involves vibrating one of the components, and analyzing the resulting fluctuations in the coupled light to steer the components into alignment using a negative feedback network. A feedback approach enables the use of compact, inexpensive transducers and electronics. Automatic aligners fabricated in the lab have demonstrated simultaneous dual axis alignment, have achieved optimum alignment in times as short as 2 milliseconds and have been able to operate on as little as -80dBm of peak coupled power.

We have designed and built a robot system for the automated alignment of optical fibers. A generalized alignment problem consists of a photodetector and electro-optical circuitry contained in a package. This is a difficult task that requires very high precision (1gm) translational movement with little backlash, and high speed manipulation (> 5 mm/sec). The robot system has closed-loop computer control and intelligent feedback to allow active alignment of the system, i.e., the optical components need to be manipulated while monitoring the output of the optical system. High-precision adjustments are needed in three dimensions, with sub-micron step increments. The robot system consists of an xy stage and an x-y-z translation stage, with closed-loop feedback control. As the optical fiber is inserted into the package, the optical fiber comes in close proximity to the photodetector. The alignment of the optical fiber to the photodetector occurs by a device which grasps the optical fiber. The fiber is translated in three dimensions within the package. The optical power coupled into the fiber serves as the control signal. Optimum alignment time can be achieved in < 10 s despite the complications caused by localized minima, with -40 dB of peak coupling power. The purpose of this paper is to discuss the general problem of precision alignment.

A reliable package for edge emitting LEDs has been developed, and the assembly and characteristics are described. This package is easily fabricated, hermetic, and provides excellent coupling stability between the chip and single mode fiber. A minimum of 7,447 of coupled power into single mode fiber is typically achieved, and long-term aging indicates less than 1dB change in light output after 2000 hrs at 90°C.

A description, from a systems perspective, of the needs and approaches to interface fiber-optics to electronic equipment is given in this paper. This systems approach addresses the fiber interface to the different levels of interconnection in the electronic equipment such as cabinets, backplanes, and circuit packs. Various existing and new AT&T components or subsystems hardware designed to interface fibers to these levels of interconnection in the electronic equipment are described.

This paper discusses various aspects of fiber-waveguide interconnection technology. Theoretical alignment tolerances for various degrees of freedom are considered. Methods for attaching individual fiber and fiber arrays to guided wave devices are described. Recent results on self-aligned fiber array-waveyuide attachment and on self-oriented arrays of polarization maintaining fibers are presented.

A packaging technique that enables a low-loss optical couple between a single-mode laser diode and a single-mode optical fiber has been developed. The technique uses a miniature thin-film substrate with an onboard resistor to mount and subsequently align a lensed and metallized fiber to an emitting laser diode. The resistor serves as a local heat source to first effect the attachment of the fiber to the substrate and second to effect the attachment of the fiber-substrate assembly into aligned position with the laser. Thermal stability of optical couples made with this technique is demonstrated.

A comprehensive review of the coupling of laser diodes to single mode fiber is given. The characteristics of coupling are clarified for the following specific coupling schemes: PC-GRIN lens PC-GRIN lens in confocal system

The use of edge-emitting LEDs in short-distance single-mode transmission applications has recently generated interest due to cost and reliability advantages compared with conventional laser sources. However, a primary reason for the low cost of conventional packaged LEDs has been the relative' insensitivity of LED-to-multimode-fiber coupling. In this paper we will consider the impact of switching from a multimode to single-mode fiber on coupling efficiency and sensitivity. Experimental results for several different coupling schemes will be presented, including both direct coupling to single-mode fiber and indirect coupling to single-mode fiber through a multimode pigtail. We show that coupling sensitivity in the direction perpendicular to the junction plane is comparable to that for a laser, regardless of the coupling scheme used. However, in spite of increased sensitivity, some relaxation of alignment tolerances can be obtained by trading off coupling efficiency for decreased sensitivity. Simple expressions are derived which predict the coupling efficiency/sensitivity trade-offs. A discussion of these options and their impact on LED packaging will be included.

Since the development of fiber optics, epoxy resins have been used exclusively for bonding and potting optical fibers into a myriad of connectors and cast geometries. Along with describing some of these fiber optic epoxy applications, this paper will illustrate one of the outstanding characteristics of this family of bonding agents, the glass transition temperature. This thermodynamic and thermomechanical characteristic of epoxy resins is defined as the midpoint of the temperature region wherein the properties of epoxy resins change from those of a glassy state to those of a rubbery state. Furthermore, this paper will eliminate some of the mystique surrounding Tg, as the glass transition temperature is usually abbreviated, as well as highlight how the Tg is a useful engineering tool for selecting epoxy resin systems for optical fiber bonding.

At speeds near 1 GHz, power requirements for driving coax and microstrip inter-connections become significant. The difficulty of properly terminating microstrips, particularly fanouts or those with complex conductor paths (such as clock distribution) leads to pulse distortion and reduced noise immunity.

We report on the status of a monolithically integrated transceiver chip being developed in our lab. The functional components of this chip consist of a TJS laser, a photodetector, an amplifier, a 4:1 multiplexer, a 1:4 demultiplexer, and a gate-array. The total gate count is -500, and the chip is designed for operation at 1.0 Gbps. The receiver section of the chip has been tested with optical inputs at a 1.0-GHz clock frequency. Fabrication of the entire transceiver chip is in progress.

Fiber optic VLSI networks which use optical processing of the medium access protocol are investigated. Two protocols are demonstrated in an experimental all-optical network with a passive star architecture: asynchronous code-division multiple access (spread spectrum), at 100 Mbps, and fixed assignment time-division multiple access, at 500 Mbps. An all-optical active star architecture is also implemented with an optically-controlled lithium niobate integrated optic switch. Optical clock distribution to four fiber optic receivers, integrated on three Si VLSI chips, is demonstrated using a coupling technique which reduces the footprint of the fiber-optic interconnect to its theoretical minimum, i.e., the diameter of a single-mode waveguide.

The combination of a low capacitance pin photodetector-low noise GaAs FET (pin/FET) pre-amplifier has been shown to provide excellent receiver sensitivity for high bit rate long-wavelength lightwave systems. For some time it has been recognized that monolithic integration of these optical and electronic components on a single semiconductor chip offers the potential of improved performance at lower cost. Reduction in the cost of such components is expected to have significant impact on the introduction of lightwave systems into the local loop and of optical interconnections in high speed switching systems. Efforts at monolithic integration of receiver components to achieve competitive sensitivity have met with only modest success to date. In this paper we review the status of this work.

Concepts for optical interconnects between electronic circuits and systems based on multichip integration, a new form of hybrid integration, are presented. In this integration scheme, chips of different substrate types, such as Si and GaAs, are imbedded in a potting material and connected with photolithographically defined metallizations to form a multichip. Within a multichip, the interconnection scheme would retain the high packing density and minimal numbers of packaging operations achievable with monolithic integration and still permit the individual optimization of material types and device designs to perform electronic and optical functions. Also, the elimination of large bonding pads on the component chips within a multichip will alleviate the pinout problem of complex integrated circuits at the chip level by allowing the pitch between input/output lines to be decreased. Multiplex-ing many slow lines with fast GaAs multiplexers can alleviate the same problem at the multichip level. Off-chip connections between multichips can be made either electrically or optically. Optical interconnections would have low crosstalk. Preliminary results with an epoxy potting material show that continuous lines can be photolithographically patterned from the chip to the epoxy. Tests with a discrete breadboard demonstrate that fast edge transitions can be retained with a diode laser driven directly by a commercial GaAs digital circuit at rates of 1.4 GHz.

Photoneural systems are a class of hybrid, optoelectronic processors designed to mimic rather closely the neural mechanisms and architectures involved in visual perception and other cognitive processes. In these systems optical linkages provide complex and adaptable neural-like interconnections between different synaptic centers which may be realized as autonomous elements of monolithically integrated optoelectronic arrays. The perception of sensory stimuli is mediated by a vastly complex hierarchy of "permeatively" linked processes. However, each step in perception seems to involve relatively simple "antagonistic" responses and the structure of perceptual responses remain topographic coherent when "mapped" to various subsystems of the brain. The topographic invariance of cognitive data transmission is suggestive of optical imaging operations and is the essential rationale for photoneural design. The basic synaptic elements are autonomous, optoelectronic processors which are, essentially, optical transceivers activated by con-trasting input stimuli. Recent developments in technology make feasible a class of promising photoneural devices which are simple integrated configurations of light emitting diodes, heterojunction phototransistors and liquid crystal light switches.

A new technology for opto-electronics has been developed, semiconductor MQWs. These MQWs have an electroabsorption effect 30-60 times larger than conventional semiconductors. They are compatible with existing source and detector material systems and produce devices that are compact and high speed, which makes them useful for monolithic integrated optoelectronic devices.

Recent work on grating-surface-emitting diode lasers has resulted in the production of lasers which have very narrow farfields (far field angles 0.25°x6°) and dynamically-stable-single-wavelength operation. These lasers use a distributed-Bagg-reflecting (DBR) grating to act as at least one of the reflectors required to complete the laser cavity. A second order grating with period of aproximately 2400 Å is used. The grating is connected to the laser-gain section through a thin-film planar or rectangular cross-section optical waveguide which is incorporated in the gain section as a large-optical-cavity and continues through a taper-transition section to the DBR section. Thus, light is launched into an optical waveguide as part of the intrinsic structure of the grating-surface-emitting laser (GSEL).

Requirements of advanced generations of computers and signal processors dictate not only high computational speed, but highly parallel architectures. These highly parallel systems require significant communication at gigabit data rates with high fanout capabilities. These systems also require multiple switching of high rate data paths to cilitate efficient functioning of parallel architectures. To realize these requirements, MSI level GaAs optoelectronic integrated circuits have been developed to overcome the limitations associated with conventional electrical interconnections. Digital GaAs circuits have been integrated with optoelectronic laser transmitters and detectors to achieve high speed, high bandwidth, optoelectronic ICs for chip level and board level optical data communications. More recently these chips are being interfaced high speed VLSI silicon ICs to form optical fiber crossbar switches, under the sponsorship of DARPA/NOSC. These circuits are expected to operate with data rates of 250 MHz and reconfiguration times of 1 nanosecond. Optical fiber crossbar switches up to 32 X 32 arrays are planned.

Integration of Si MOSFETs with GaAs MESFETs and with GaAs/AlGaAs double-heterostructure LEDs on mono-lithic GaAs/Si substrates is reviewed. Both Si MOSFETs and GaAs MESFETs show characteristics comparable to those for devices fabricated on separate Si and GaAs substrates. In LED/MOSFET integration, the cathode of each LED is connected with the drain of a MOSFET. LED modulation rates up to 27 Mbps have been achieved by applying a stream of voltage pulses to the MOSFET gate.

Silicon provides a natural substrate base for the development of channel waveguides and their integration with optoelectronic components. Using epitaxial growth, selective doping, and plasma etching, channel waveguides can be fabricated using single crystal silicon alone. Oxide layers of low optical index are readily formed by thermal means on silicon to provide a base upon which low-loss film waveguides can be formed by ion exchange and implantation, chemical vapor deposition, and physical vapor deposition. Thermally oxidized and nitrided layers provide a simple means for developing waveguides. The channel shape for ridge waveguides can be delineated by chemical etching and ion milling techniques. The anisotropic etch characteristics of silicon provide a natural channel for imbedding waveguides using organic and inorganic materials. This paper will review common semiconductor processing techniques used for the formation of channel waveguides on silicon and the performance results obtained to date. The use of channel waveguides for specific device developments will be described and the most promising areas for future development will be addressed.

A black, patternable polyimide coating has been developed for optoelectronics applications. Dielectric properties, adhesion, thermal stability, spectral characteristics, and techniques for processing the material are discussed.