Artificial Intelligence-Driven Approaches in Semiconductor Research

Abstract

To address the persistent challenges of scaling and power consumption in integrated circuits and chips, recent research has focused on exploring novel semiconductor materials beyond silicon and designing new device architectures. The vastness of the material and parameter space poses significant challenges in terms of cost and efficiency for traditional experimental and computational methods. The rise of artificial intelligence (AI) offers a highly promising avenue for accelerating semiconductor technology development. AI-driven methods demonstrate significant advantages in analyzing and interpreting large datasets, potentially freeing researchers to focus on more creative endeavors. This review provides a detailed and timely overview of how AI-driven approaches are assisting researchers across the entire semiconductor research pipeline, encompassing materials discovery, semiconductor screening, synthesis, characterization, and device performance optimization, highlighting how their integration facilitates a holistic understanding of the entire processing-structure-property-performance (PSPP) relationship. Remain challenges related to dataset quality, model generalizability, and autonomous experimentation, as well as the under-application of AI to critical needs are discussed in the semiconductor field, such as wafer-scale growth of high-quality, single-crystal semiconductor thin films beyond silicon. Addressing these challenges requires collaborative efforts from researchers across various organizations and disciplines, and represents a key focus for future research.