PhD. of Yufei Xing

Author: Yufei Xing Promotors: Wim Bogaerts and Tom Dhaene Dutch Title: Gedragsmodellen, parameterextractie en voorspelling van opbrengst voor circuits in siliciumfotonica English Title: Behavioural Models, Parameter Extraction and Yield Prediction for Silicon Photonic Circuits Date: 5/12/2019 Main Language: English ISBN number: 978-94-6355-311-7 Thesis: (39.9MB) Presentation: (11.5MB)

Abstract

Silicon photonics is maturing quickly to become more commercially viable. Its high material contrast facilitates tight light confinement and high-density integration, which enables photonic circuits with ever-increasing size, complexity, and functionality. Meanwhile, high material contrast also makes silicon photonic circuits very susceptible to fabrication induced performance variations. These variations accumulate with the increasing circuit size, which increases the cost and limits the capacity of large-scale photonic integration. In this thesis, we focus on how to make a realistic yield prediction for silicon photonics. To predict the yield of photonics circuits, we can map variations of low-level behavior parameters (effective index, coupling coefficient) or fabricated geometry parameters (linewidth, thickness) to high-level circuit performance variations. By assigning location-aware variations that mimic the actual fabrication variation, we can derive the high-level circuit variations using Monte-Carlo simulations. This thesis addresses three key problems in a realistic yield prediction: How to collect data accurately from fabricated circuits on wafer-scale? How to construct a model to separate systematic and random variations on different spatial levels? How to generate virtual wafers maps based on extracted fabricated wafer maps to enable Monte-Carlo simulations for yield prediction?

Related Research Topics

• Variability Aware Design
• Compact Circuit Models and Extraction