Back to search

University of Patras

Smart industrial networks and digital twins for optimizing the overlay modeling in advanced photolithography processes

Abstract

dc:description

Photolithography is a microfabrication process used to transfer intricate patterns onto semiconductor wafers through the use of light-sensitive materials. It serves as a foundational technology in semiconductor manufacturing, enabling the creation of integrated circuits (IC) with ever-decreasing feature sizes. In photolithography, yield refers to the percentage of semiconductor devices that meet quality and performance standards after the manufacturing process, serving as a critical indicator of process reliability and product quality. Throughput, on the other hand, measures the efficiency of the manufacturing process by quantifying the number of wafers processed within a given time period. Manufacturers constantly grapple with a fundamental dilemma: increasing throughput to meet high production demands often risks reducing yield if process parameters are compromised, while an exclusive focus on yield may slow production rates. This tension between quality and efficiency drives ongoing research aimed at optimizing process parameters, modeling techniques, and equipment performance to achieve a balance where improvements in yield do not come at the expense of throughput. This study contends that targeted modifications to the overlay (OL) modeling process in advanced photolithography can significantly enhance yield without compromising throughput. By systematically analyzing process parameters and modeling techniques, the research seeks to identify optimal strategies that balance efficiency and output, thereby offering valuable insights for improving semiconductor manufacturing. This research highlights the essential challenge of balancing throughput and quality, particularly due to the time-intensive nature of precise OL sensor measurements. To address this, the study proposes a Sensor Markers Selection (SMS) algorithm utilizing Genetic Algorithms (GA), optimizing measurement selection to achieve an improved balance between uniformity and information gain without reducing throughput. Additionally, a novel Surface Reconstruction (SR) technique is introduced to enhance OL modeling accuracy. By leveraging spatial interpolation, this approach enables more precise parameter tuning while minimizing the number of necessary sensor measurements. The research further investigates Ridge Regression as an effective alternative to traditional Least Squares (LSQ) methods to address multicollinearity in high-dimensional OL models. Lastly, the thesis proposes a Digital Twin (DT) architecture for self-adaptive OL control. This architecture integrates Advanced Process Control (APC) and computational techniques, dynamically estimating and adjusting OL models to reduce sensor dependency while maintaining throughput. This approach demonstrates significant potential for advancing Moore’s Law through data-driven methodologies in photolithography. The proposed methods were validated using an actual industrial process involving 300 mm wafers. The results indicate substantial improvement in OL accuracy in both the x and y directions, underscoring the practical effectiveness of the presented techniques.

Degree

thesis:*
Grantor dc:publisher
University of Patras
Year dc:date
2025

Author and committee

dc:creator, dc:contributor.*
Authors dc:creator
  • Magklaras, Aristeidis
  • Μαγκλάρας, Αριστείδης

Subjects

dc:subject × 15

Rights

Language dc:language
eng

Identifiers

dc:identifier.*
Identifier
10.12681/eadd/58982

Chain of custody

source
Harvested from
Greek National Archive of PhD Theses
Base URL
phdtheses.ekt.gr/eadd_oai/request
Last updated
2026-11-18
Source record
OAI-PMH GetRecord
citation

Magklaras, Aristeidis; Μαγκλάρας, Αριστείδης. Smart industrial networks and digital twins for optimizing the overlay modeling in advanced photolithography processes. University of Patras, 2025. http://hdl.handle.net/10442/hedi/58982