材料科学
蚀刻(微加工)
开槽
停留时间
光电子学
硅
等离子体刻蚀
反应离子刻蚀
碳化硅
咬边
干法蚀刻
电介质
接口(物质)
晶体硅
支柱
电子工程
深反应离子刻蚀
占空比
共发射极
纳米技术
准静态过程
过程(计算)
滤波器(信号处理)
绝缘体上的硅
电容
光学
频道(广播)
图层(电子)
薄脆饼
微电子机械系统
作者
J. D. Guo,Zhengming Liu,Mingqiang Geng,Kunrun Song,Huiyun Jiang,Dawei Gao,Dong Ni
出处
期刊:Journal of vacuum science & technology
[American Institute of Physics]
日期:2026-07-22
卷期号:44 (5)
摘要
As dielectric metasurfaces move toward CMOS-compatible wafer-scale manufacturing, the plasma-etch transfer of high-aspect-ratio (HAR) silicon pillars on oxide-containing stacks has become a key challenge for process integration. Here, we investigate ∼10:1 Si pillar etching on a SiO2 underlayer in a 12-in. CMOS production-line environment under continuous-wave (CW) and pulsed-bias conditions. Under CW operation, aspect-ratio-dependent etching (also termed RIE)-lag-induced asynchronous etch-front arrival causes isolated pillars to reach the Si/SiO2 interface earlier than dense regions. The resulting local interface-exposure dwell time promotes foot loss, charge-assisted ion deflection, and mask-related profile degradation, leading to notching and collapse. Pulsed bias modifies this late-stage evolution nonmonotonically: 50% duty preserves vertical propagation and suppresses CW-type collapse, whereas lower duty cycles reduce continuous interface exposure but can induce taper or height-deficient profile loss. To rationalize these observations, we develop a mechanism-oriented two-dimensional feature-scale simulator that tracks etch-front timing, local interface-exposure dwell time, bottom-critical dimension evolution, and charge-assisted foot degradation. Guided by these results, a stage-dependent pulsed recipe using 50% duty for the main-etch step and 30% duty for the over-etch step yields substantially improved vertical pillar profiles. This work provides a process-level framework for interpreting HAR silicon pillar etching on oxide-containing stacks, with implications for silicon photonics, metasurface manufacturing, and other CMOS-compatible nanophotonic platforms.
科研通智能强力驱动
Strongly Powered by AbleSci AI