split-lot experiments
**Split-lot experiments** is the **controlled manufacturing trials that divide a lot into baseline and experimental subsets to isolate process effects** - they are the operational backbone of fab experimentation because they enable causal comparison under near-identical context.
**What Is Split-lot experiments?**
- **Definition**: Lot-level experiment where matched wafer groups receive different recipe settings or process conditions.
- **Control Principle**: Keep all non-target variables constant so observed differences map to intentional change.
- **Measured Outcomes**: Yield shift, parametric movement, defect signatures, and reliability impact.
- **Experimental Types**: Single-factor splits, paired tool comparisons, and staged split verification runs.
**Why Split-lot experiments Matters**
- **Causal Clarity**: Split design provides stronger evidence than uncontrolled historical comparisons.
- **Change Qualification**: New process settings can be validated with limited production risk.
- **Yield Optimization**: Direct A/B data reveals whether proposed changes improve key metrics.
- **Escalation Control**: Helps distinguish true process fixes from random run-to-run variation.
- **Learning Traceability**: Results become reusable knowledge for future process tuning decisions.
**How It Is Used in Practice**
- **Split Planning**: Define objective metric, baseline condition, experimental condition, and success criteria.
- **Execution Discipline**: Maintain strict run-order and metrology consistency across split branches.
- **Statistical Review**: Use significance testing and effect-size analysis before adopting recipe change.
Split-lot experiments are **the most practical controlled method for fab process decision-making** - disciplined split execution converts hypotheses into reliable production actions.