GaAs/AlGaAs based multiquantum well device for learning and decision making in optical neuro-computers

Theoretical and experimental examinations are made of two classes of devices which have the requisite properties for learning and neuronlike decision making. These properties involve responding to a time sequence of optical pulses (for learning) and integrating the thresholding and are realized in GaAs/AlGaAs multiquantum wells (MQWs) using the quantum confined Stark effect. The device is compatible with heterojunction bipolar transistor (HBT) technology since the controller is simply an HBT structure with a built-in MQW structure. Therefore, it should be possible to make use of the advances in HBT technology and develop large neuron arrays. Also, due to the current gain in the controller-modulator (C-M) device, the optical power requirements (~10 μW per device) are consistent with semiconductor laser structures, and one does not require high-power lasers. Based on the estimation of optical power dissipation (⩽10 W/cm²) with current technology it should be possible to achieve a 1000-b array with each C-M device being about 20 μm in diameter