The design and analysis of novel integrated phase-change photonic memory and computing devices

Abstract

The current massive growth in data generation and communication challenges traditional computing and storage paradigms. The integrated silicon photonic platform may alleviate the physical limitations resulting from the use of electrical interconnects and the conventional von Neuman computing architecture, due to its intrinsic energy and bandwidth advantages. This work focuses on the development of the phase-change all-photonic memory (PPCM), a device potentially enabling the transition from the electrical to the optical domain by providing the (previously unavailable) non-volatile all-photonic storage functionality. PPCM devices allow for all-optical encoding of the information on the crystal fraction of a waveguide-implemented phase-change material layer, here Ge2Sb2Te5, which in turn modulates the transmitted signal amplitude. This thesis reports novel developments of the numerical methods necessary to emulate the physics of PPCM device operation and performance characteristics, illustrating solutions enabling the realization of a simulation framework modelling the inherently three-dimensional and self-influencing optical, thermal and phase-switching behaviour of PPCM devices. This thesis also depicts an innovative, fast and cost-effective method to characterise the key optical properties of phase-change materials (upon which the performance of PPCM devices depend), exploiting the reflection pattern of a purposely built layer stack, combined with a smart fit algorithm adapting potential solutions drawn from the scientific literature. The simulation framework developed in the thesis is used to analyse reported PPCM experimental results. Numerous sources of uncertainty are underlined, whose systematic analysis reduced to the peculiar non-linear optical properties of Ge2Sb2Te5. Yet, the data fit process validates both the simulation tool and the remaining physical assumptions, as the model captures the key aspects of the PPCM at high optical intensity, and reliably and accurately predicts its behaviour at low intensity, enabling to investigate its underpinning physical mechanisms. Finally, a novel PPCM memory architecture, exploiting the interaction of a much-reduced Ge2Sb2Te5 volume with a plasmonic resonant nanoantenna, is proposed and numerically investigated. The architecture concept is described and the memory functionality is demonstrated, underlining its potential energy and speed improvement on the conventional device by up to two orders of magnitude.