Citation

Tin, Tze Chiang (2022) Modelling of overlay virtual metrology system in photolithography process. PhD thesis, Multimedia University.

Full text not available from this repository.

Abstract

Prior works on overlay VM system applied wafer-level modelling approach by utilizing fault detection and classification (FDC) data from the photolithography equipment as the source data to construct process representatives for each wafer. However, owing to active research for continuous improvement, FDC system may be taken offline for enhancement works. When FDC system undergoes these activities, FDC-based VM system would be rendered inefficacious as FDC data would not be available. Hence, during such events, a non-FDC based VM system with sufficient competency is required to sustain the production line until the FDC-based VM system is able to resume service. Motivated by a real-world production environment of a 200mm semiconductor manufacturing plant (fab), this work aims to realise a non-FDC based overlay VM system to sustain the production line when FDC system undergoes offline activities. Realizing non-FDC based system is a non-trivial task. Without the availability of the FDC system, process data can only be sampled at low frequency, and only an averaged reading per sensor is available for each wafer’s fabrication process from the fabrication equipment. This limitation resulted in process data that have low process characteristic depiction capability and thus, decreases the prediction capability of the data mining algorithms for a competent VM system. Hence, a different VM modelling approach is required to realise a competent VM system. By using a different modelling paradigm, this work proposes a lot-level modelling approach to realise the non-FDC based overlay VM. Next, this work proposes a joint prediction modelling approach to create the prediction system of the VM. The joint prediction model first performs a classification task to identify wafer lots with faulty wafers. Using the proposed modelling, a smart sampling system termed C2O is realised in this work. The experimental results showed that using the proposed approaches, C2O is capable to achieve a true positive rate (TPR) of 71.34% for the classification task and mean absolute scaled error (MASE) of 8.48 for the regression task.

Item Type:	Thesis (PhD)
Additional Information:	Call No.: QC88 .T56 2022
Uncontrolled Keywords:	Metrology—Technological innovations
Subjects:	Q Science > QC Physics > QC81-114 Weights and measures
Divisions:	Faculty of Computing and Informatics (FCI)
Depositing User:	Ms Nurul Iqtiani Ahmad
Date Deposited:	11 Jan 2024 00:57
Last Modified:	11 Jan 2024 00:57
URII:	http://shdl.mmu.edu.my/id/eprint/12036

Downloads

Edit (login required)