Information Encoding and Retrieval in Photonic Systems

Abstract

Photonics is an enabling technology for information communication and is of profound importance in the evolution of information processing. The expeditious proliferation of high volume data applications in recent years further advances photonics from traditional fiber-optic communications to be either the de facto interconnect technology among computing systems, or an emerging solution to active research fields such as neuromorphic photonics. Photonics aims to address new challenges arising in these fields with innovative information encoding and retrieval mechanisms.

This thesis starts off discussing novel information encoding and retrieval in traditional fiber-optic communications against security concerns over eavesdropping. Conventional digital cryptography techniques encode data into unreadable codes, but still expose the existence of communication through metadata information. Instead, we propose a steganographic communication scheme to hide the existence of optical communications itself, i.e., optical steganography. We present the first practical steganographic communication system between real computers over long-distance fiber-optic links, and rigorous security evaluations against stealth signal detection.

This dissertation then moves on to investigate spike coding, a coding mechanism which benefits from both the bandwidth efficiency of analog pulse encoding and from the digital on/off nature of the spikes. We realize the demand for photonic platform to interface with commonly used digital systems, and present an all-optical digital-to-spike conversion scheme at a single graphene excitable laser. Moreover, we identify the lack of inhibition capability in current brain-inspired photonic hardware, and enable simultaneous excitatory and inhibitory dynamics within a graphene excitable laser. These efforts further lead to the construction of spike-based logic gates and their application to optical encryption of data-carrying spike sequence.

This dissertation also puts an unequivocal emphasis on studying an analog photonic approach to address blind source separation problem. State-of-the-art digital approaches digitize redundant multi-dimensional inputs and lacks tunability across multiple frequency bands. We design and fabricate a photonic chip that is capable of processing multi-channel inputs and performing source separation as an analog front-end. Furthermore, we develop novel moment observation algorithms for a variety of multivariate photonic applications. Eventually, we demonstrate a practical blind source separation pipeline for multi-antenna systems receiving microwave signals transmitted over the air.