公开(公告)号：US12218015B2

公开(公告)日：2025-02-04

申请号：US17397989

申请日：2021-08-10

Applicant: Taiwan Semiconductor Manufacturing Company, Ltd.

Inventor： Chansyun Yang ,  Chan-Lon Yang ,  Keh-Jeng Chang ,  Perng-Fei Yuh

IPC: H01L21/67 ,  G03F7/00 ,  H01L21/02 ,  H01L21/66 ,  H01L21/027

Abstract: An apparatus includes a beam conditioning assembly configured to output one or more wavelengths to a substrate being processed and receive one or more reflected wavelengths from the substrate, and a machine learning device configured to process the one or more reflected wavelengths to predict a process variable and compare the predicted process variable with a measured process variable to obtain a comparison result.

公开(公告)号：US20220336294A1

公开(公告)日：2022-10-20

申请号：US17397989

申请日：2021-08-10

Applicant: Taiwan Semiconductor Manufacturing Company, Ltd.

Inventor： Chansyun Yang ,  Chan-Lon Yang ,  Keh-Jeng Chang ,  Perng-Fei Yuh

IPC: H01L21/66 ,  G03F7/20 ,  H01L21/67 ,  H01L21/027

Abstract: An apparatus includes a beam conditioning assembly configured to output one or more wavelengths to a substrate being processed and receive one or more reflected wavelengths from the substrate, and a machine learning device configured to process the one or more reflected wavelengths to predict a process variable and compare the predicted process variable with a measured process variable to obtain a comparison result.

公开(公告)号：US20240395637A1

公开(公告)日：2024-11-28

申请号：US18789686

申请日：2024-07-31

Applicant: Taiwan Semiconductor Manufacturing Company, Ltd.

Inventor： Chansyun Yang ,  Chan-Lon Yang ,  Keh-Jeng Chang ,  Perng-Fei Yuh

IPC: H01L21/66 ,  G03F7/00 ,  H01L21/027 ,  H01L21/67

Abstract: A method for real-time compensation control of an etch process includes: providing a substrate having a layer in a process chamber; performing the etch process on the layer; directing one or more wavelengths to a region of the layer by a beam conditioning assembly; receiving one or more reflected wavelengths from the region of the layer; predicting a process variable by processing the one or more reflected wavelengths using a machine learning model; and comparing the predicted process variable with a predetermined process variable to obtain a comparison result.