Machine Learning Analysis Reveals Pathways to Overcome Oxide Semiconductor
A data-driven study using machine-learning analysis has mapped why oxide semiconductor thin-film transistors struggle to scale to high-speed, stable switching. Reviewing more than 1,000 paired experimental records drawn from 2,005 relevant papers, researchers quantified an enduring trade-off: attempts to raise electron mobility typically worsen threshold-voltage stability under electrical and thermal stress. The team crawled 64,852 documents from 30 seed papers and 13 key investigators, then validated 1,030 data points linking field-effect mobility with stability metrics. Extraction accuracy reached 95.2% for atomic layer deposition and 94.9% for physical vapor deposition. The work attributes the “frontier” to band-structure physics and subgap defect states.






