Bridging the gaps in semiconductor digital twins: A structured review

Semiconductor manufacturing is a highly constrained industrial domain, where extreme precision and tightly coupled processes make even minor disruptions critical. Within this domain, digital twins are increasingly recognized as valuable tools for improving system understanding, prediction, and decision support. Despite this potential, current research remains fragmented, with heterogeneous implementations, inconsistent terminology, and uneven technological maturity. Based on a structured screening of 134 records retrieved from Web of Science and Scopus, this review analyzes 28 peer-reviewed core contributions reporting implemented digital twin or digital shadow solutions in semiconductor manufacturing. We examine the literature by mapping these contributions in terms of operational maturity, architectural intent, technological foundations, and alignment with FAIR principles. The analysis highlights a predominance of system-level implementations, a progressive shift from observability-oriented digital shadows toward more decision-capable digital twins, and a growing adoption of hybrid physics-based/data-driven models, optimization strategies, and AI-enabled control solutions. At the same time, persistent limitations emerge in vertical integration, interoperability, architectural standardization, reproducibility, and data openness. The result is not only a taxonomy of tools and architectures, but also an evidence-based outlook on the future of digital twins in semiconductor manufacturing. This outlook points toward interoperable and reusable infrastructures that bridge simulation and reality, combine human and machine intelligence, and support faster, more efficient, transparent, and collaborative manufacturing.