Creation and application of very large format high-fill-factor GaAs-on-CMOS binary and gray-scale modulator and emitter arrays

Abstract

JohnAlfredTrezzaJeffry S. Powell, Charles Garvin, Keith Kang, Richard StackSanders, A Lockheed Martin Company, 65 Spit Brook Road, NHQ6-1551, Nashua, NH 03060Vertical cavity optoelectronic modulators are the key component for future real-time electro-opticalsignal processing, computing, optical memory access, free-space optical interconnect, informationprocessing, laser printing, and data conversion systems. Here we will discuss the operation and operatingprinciples of our optical devices, the integration of these devices into large arrays, the experimentalcharacteristics of large arrays of devices, and the use of the arrays.The quantum well device arrays that we produce are comprised of pixels. These are surfacenormal devices constructed (by molecular beam epitaxy) out ofthe 111-V semiconductor material system.MED pixels are so called because they:1) Modulate light: The devices can be operated in a mode in which they alter either the amplitude (onor off), the phase (0° to 360°), or the direction (either unimpeded, or altered by simple reflection).2) Emit light: The devices can be operated as resonant cavity light emitting diodes, providing outputlight that, while not coherent, is spatially collimated and spectrally confined.3) lletect light: The devices can be operated as detectors for incident light.A given pixel can perform all three of these types of operations (though one typically chooses eitherphase or amplitude modulation as one of the three properties); the type of behavior that a given MEDpixel exhibits at any one time is determined by the control signal which is applied at any given moment.When such arrays consist of close-packed pure modulating or light emitting pixels, they are referred toas 'spatial light modulator' (SLM) or 'spatial light emitter' (SLE) arrays respectively. We typicallycombine optical chips containing large (up to 1x2048 and 256 x 256) arrays of GaAs MED pixels withcustom SRAM CMOS chips. The GaAs optical and silicon driver chips are combined by placing solderbumps on the chips and then using these bumps as electrical contacts between individual optical pixelsand silicon transistor circuits. The resultant hybrid electro optical chip is then sealed, wirebonded intoan IC package, and hermetically contained. We have used this 'flip-chip' bonding technology to connectpixelated GaAs devices to silicon drivers on chips over 2 square centimeters in size, containing as manyas 307,000 pixels with pixel yields of 99.98% uncorrected and 99.998% with electrical correction. TheSRAM chips (currently 2 x 65k) can store data bits and output data optically through the optical chip.These arrays, then, are ultra-fast electrical to optical (and optical to electrical) transducers where hugeparallelism exists.Electrical data input into our 256x256 array is through a series of 256 input lines, each of which isclocked in parallel into the array. Thus, the array takes 256 clocks to load an image (our 128x128 arrayhandles 128 input lines and takes 128 clocks to load). The CMOS SRAM driver chip allows two fullframes of binary mode data can be stored at any one time. Electronic data is loaded into the CMOS's'back buffer page' while another page of memory is displayed through the optical chip. When the back