A graph-driven placement framework for Si photonic circuits

作者：Yu Li, Jiyao Yu, Bowen Li, G. Cao, Junbo Feng · 发表于：Applied Optics and Photonics China · 年份：2023 · DOI：10.1117/12.2650839 · 研究领域：Engineering

This article presents a graph-driven placement framework for Si photonic circuits. In this framework, a netlist exported from the schematic diagram is transferred into an adjacency matrix, and further parameterized to an undirected graph. By this method, optical devices and waveguides are quantified as nodes and edges, respectively. Non-Euclidean data structures between nodes can be extracted which includes parallel relations, sequential relations and connecting patterns, by matching those patterns with pre-defined database, certain layout strategies formulated by human experts can be properly applied. By extracting the geometric information and the preset spacing requirements of each device in the Process Design Kit library, the layout strategy requirements of each component can be assigned, so as to determine the geometric position. This work designed the graph-driven placement framework, tested the identification accuracy for connection pattern, and applied the framework in practical chip designs including artificial intelligent and Wavelength Division Multiplexing circuits.