Date of Award

8-2026

Document Type

Thesis

Degree Name

Master of Science (MS)

Department

Electrical Engineering and Computer Science

First Advisor

Md Selim Habib

Second Advisor

Yakov Berchenko-Kogan

Third Advisor

Naveed Mahmud

Fourth Advisor

Brian A. Lail

Abstract

Anti-resonant hollow-core fibers (AR-HCFs) offer promising capabilities as a trans formative technology for next-generation photonics. By guiding light primarily in air rather than solid silica, they enable light to travel faster, presenting a fiber with sig nificantly reduced latency, high power handling, and broadband transmission. These advantages make them a promising product for future telecom systems. Accurate nu merical modeling of such fabricated structures is essential to capture the potential of a given design, particularly for predicting and minimizing surface scattering loss. Surface scattering in AR-HCFs is governed by the Rayleigh Criterion, which depends on nanoscale roughness and geometric irregularities at the air-glass interfaces. There fore, converting Scanning Electron Microscope (SEM) figures into precise vector-based simulation geometries is critical. However, raw SEM images often suffer from edge frag mentation, noise (e.g dust particles), and imperfections in fabrication which translate into deformations that challenge conventional edge detection methods and complicate tube identification by popular methods like the Hough Transform (H.T). The aim of this project is to design a fully automated workflow which identifies the tubes present in an AR-HCF design given a SEM, approximates the design with a geometry that minimizes the surface scattering loss while still resembling the original geometric de sign, and simulates such design in COMSOL to evaluate the potential of the AR-HCF. The framework compares various approaches to detect, group, and reconstruct edge segments into coherent structural components. Among the different approaches the paper compares, deep learning techniques are included. This automated characterization tool bridges the gap between fabrication and high-fidelity modeling, accelerating the development of next-generation AR-HCFs.

Available for download on Sunday, August 01, 2027

Share

COinS