| Title |
AI-Assisted Screening of Athermal Candidate Geometries for TiO2 Clad Silicon Ring Resonators |
| Authors |
문형진(Hyeong Jin Mun) ; 이용복(Yong Bok Lee) |
| DOI |
https://doi.org/10.5573/ieie.2026.63.10.9 |
| Keywords |
Athermal ring resonator; Fast screening; Silicon photonics; Surrogate model; TiO2 cladding |
| Abstract |
This study proposes an AI-based screening framework for rapidly identifying athermal candidate geometries for TiO2-clad SOI ring resonators under high-temperature operation. The ring-core width and height, bus-waveguide width, ring radius, and coupling gap were defined as design variables, and two-dimensional MPB simulation results were integrated with an RBF surrogate model and an analytical all-pass ring-resonator model. An ANN emulator was trained using 637,245 analytically generated labels to prioritize near-athermal candidates based on thermal sensitivity, Q-factor, and free spectral range (FSR). FDTD evaluation of the top-ranked candidates followed by a local parameter sweep yielded a refined geometry exhibiting a thermal sensitivity of 0.0134 pm/°C, a Q-factor of 1,901, and an FSR of 48.9 nm over 20?120 °C. The ANN prediction time for a single design point was 5.598 s, corresponding to an approximately 98.41% reduction compared with three-dimensional FDTD simulation. Under dimensional deviations of ±1 nm, the changes in thermal sensitivity caused by variations in the Si core height and ring width were approximately 0.32 and 0.20 pm/°C, respectively. These results demonstrate that the proposed framework can efficiently screen promising athermal candidates over a broad design space. Future work should address robust optimization that accounts for fabrication and material variations, together with experimental validation using fabricated devices. |