ETRI-Knowledge Sharing Plaform

ENGLISH

성과물

논문 검색
구분 SCI
연도 ~ 키워드

상세정보

학술대회 Feasibility Study of Predicting Semiconductor Thin Film Thickness Based on 1D Convolutional Neural Network
Cited 0 time in scopus Download 2 time Share share facebook twitter linkedin kakaostory
저자
박효빈, 임길택, 김광주
발행일
202011
출처
International Conference on Consumer Electronics (ICCE) 2020 : Asia, pp.266-269
DOI
https://dx.doi.org/10.1109/ICCE-Asia49877.2020.9276883
협약과제
20ZD1100, 대경권 지역산업 기반 ICT 융합기술 고도화 지원사업, 문기영
초록
Recently, the increased requirement of data processing and reduced data latency has driven the demand for 3D stacks for high-performance semiconductors. The process of stacking hundreds of layers of semiconductor thin film is essential. However, there is a problem that the uniformity of thickness is degraded due to the defect of the semiconductor thin film. This causes deformation of the device structure and is a significant factor in the performance degradation of the semiconductor. To prevent this, it is important to measure the thickness of the thin film quickly and accurately. Optical spectrum analysis is widely used for measuring the thickness of semiconductor thin films. In this paper, we proposed a method to measure the thickness of semiconductor thin films by learning the reflectance spectrum according to the thickness and wavelength of four-layer thin films using a modified model of the well-known deep learning structure (VGG-16 Network). Our semiconductor thin film thickness prediction method performance of about 0.589 MAE was shown. Experimental results show that the proposed model can be a baseline method for predicting the thickness of semiconductor thin films.
키워드
1D CNN, Convolutional Neural Network (CNN), Ensemble, Semiconductor thin film thickness
KSP 제안 키워드
Convolution neural network(CNN), Data Latency, Data processing, Device structure, Feasibility study, Four-layer, High performance, Method performance, Optical spectrum analysis, Prediction methods, Proposed model