Automated Anomaly Detection and Root Cause Analysis in Semiconductor Wafer Fabrication via Multi-Modal Graph Neural Networks

**Abstract:** This paper introduces a novel framework for automated anomaly detection and root cause analysis (RCA) within semiconductor wafer fabrication processes. Leveraging recent advances in multi-modal data fusion and graph neural networks (GNNs), our system, termed "WaferGraph Insights" (WGI), integrates process parameter data, inspection images, and equipment logs to identify anomalous wafer batches and pinpoint their underlying root causes with unprecedented accuracy. WGI demonstrates a potential 25-40% reduction in defect rates and a 15-20% improvement in RCA resolution time compared to current statistical process control (SPC) methods, significantly impacting semiconductor manufacturing efficiency and cost reduction. This system is immediately deployable in existing fabs with minimal hardware modifications.