Monolithically integrated optical computing component for photonic matrix multiplication based on electro-optically modulated vertical-cavity surface-emitting lasers

Leading-edge machine learning algorithms require large amounts of matrix multiplications, which absorb significant computational resources in modern digital electronic systems. Analogue optical computing systems consisting of light sources, modulators and receivers may perform these operations with much higher efficiency. We introduce an integrated device based on electro-optically modulated (EOM) vertical-cavity surface-emitting lasers (VCSELs) that can perform analog multiplication at >28 GHz and at <20 mW power consumption. Due to its monolithic integration, the EOM VCSEL may be used as a building block for integrated optical computing devices. Development of such devices can help create 3-dimensionally integrated computing and communication systems.