表面粗糙度
表面光洁度
光子学
退火(玻璃)
波长
电子工程
振幅
光电子学
材料科学
等离子体增强化学气相沉积
硅光子学
传输损耗
光学
吸收(声学)
材料性能
化学气相沉积
作者
Pedro B. Veiga,Vinicius G. Antunes,C. Petit-Etienne,J. Moeyaert,Karen Ribaud,Quentin Wilmart,Erwine Pargon
标识
DOI:10.1109/jlt.2026.3651821
摘要
This article aims to identify the key parameters controlling optical propagation losses in PECVD SiN waveguides (WG) at telecom wavelengths and to provide guidance for achieving ultra-low-loss photonic platforms, both with and without thermal-budget constraints. To this end, the contribution of bulk material loss to the total propagation loss is evaluated by systematically comparing experimental loss measurements with surface-scattering losses predicted by the Payne and Lacey model, which incorporates robust experimental measurements of the roughness parameters (unbiased roughness amplitude and correlation length). Several parameters are investigated: the physico-chemical properties of the PECVD SiN (more or less Si-rich) and of the SiO $_{2}$ capping, the sidewalls roughness parameters of the patterned SiN WGs and the impact of an annealing process on the materials properties and consequently on the optical losses. The SiN patterning uses state-of-the art 300mm tools and results in low roughness parameters (LER=2.1nm and correlation length of 78nm) enabling the lowest reported losses at 1.31 $\mathrm{\mu }$ m for PECVD SiN (0.34dB/cm). However, the Payne and Lacey model predicts that reducing the correlation length down to 30nm while keeping low roughness amplitude would offer the possibility to reach sub-0.2 -0.1dB/cm loss if all material loss are suppressed. At 1.55 $\mathrm{\mu }$m, the PECVD SiN used in this work contain too many absorbing centers (NH, SiH) to achieve sub 1dB/cm losses. Annealing PECVD SiN at 1100°C allows the removal of SiH and NH absorption sites, thus enabling ultra-low losses of 0.18–0.24 dB/cm at 1.31 $\mathrm{\mu }$m. However, the annealing introduces also some new defects (Si-Si nanoclusters for Si-rich SiN and N dangling bonds for N-rich SiN) that introduces some unwanted loss at 1.55 $\mathrm{\mu }$m.
科研通智能强力驱动
Strongly Powered by AbleSci AI