Resumen de Multipurpose programmable integrated photonics: principles and applications
In recent years, programmable integrated photonics (PIP) has evolved from a promising, new paradigm to deploy photonics to a larger scale to a solid, revolutionary reality, bringing up the attention of numerous research and industry players. Based on the same theoretical foundations than field-programmable gate arrays (FPGAs), this technology relies on common, two-dimensional integrated optical hardware configurations based on the interconnection of programmable unit cells (PUCs), which -by suitable programming of their phase actuators- can implement a variety of functionalities that can be elaborated for basic or more complex operation in many application fields, such as artificial intelligence, deep learning, quantum information systems, 5/6-G telecommunications, switching, data center interconnections, hardware acceleration and sensing, amongst others.
In this work, we will dedicate ourselves to explore several software capabilities of these processors under different chip designs. We explore different cutting-edge approaches based on computational optimization and graph theory to precisely control and configure these devices. One of these, self-configuration, deals with the automated synthesis of optical circuit configurations -even in presence of parasitic effects such as nonuniform losses, optical and electrical crosstalk- without any need for prior knowledge about hardware state. There are occasions, though, in which accessing to this information may be of use. Self-calibration and self-characterization tools allow us to perform a quick check to our photonic processor's status, allowing us to retrieve useful pieces of information such as the electrical current needed to supply to each phase actuator to change its corresponding PUC state arbitrarily or the insertion loss of every unit cell and optical interconnection surrounding the structure. These mechanisms not only allow us to quickly identify any malfunctioning PUCs or chip areas in our design, but also reveal another alternative to program photonic circuits in our design from current pre-sets.
These strategies constitute a gigantic step to unleash all the potential of these devices. They provide solutions to handle with hundreds of variables and simultaneously manage multiple configuration actions, one of the main limitations that prevent this technology to scale up and become disruptive in the years to come.
