Large Language Model-Based Hierarchical Framework for Emerging Trends in Extreme Ultraviolet Research

Abstract

Translate Article

As artificial intelligence advances rapidly, demand for node scaling and improved device performance is increasing in semiconductors. This necessitates a systematic analysis of extreme ultraviolet (EUV) technology developments. This study introduces a hierarchical framework based on large language models (LLM) and applies it to 6,859 EUV-related papers published from 2015 to 2024. The framework combines clustering, summarization, evaluation, selection, and a decision stage guided by an LLM-based AI agent to build a hierarchical tree, and it then analyzes the hierarchy step by step to track emerging trends in the field. Emerging trends in EUV research include algorithmic metrology and overlay modeling, illumination engineering with pupil mapping, and advances in multi-beam mask writing along with the supplier ecosystem. This hierarchical framework is expected to not only determine the future direction of EUV technology but also broadly apply LLM-based methods to identify various emerging trends.