spatial correlation in yield
**Spatial correlation in yield** is the **statistical relationship where neighboring dies on a wafer show similar pass-fail behavior because they share local process conditions** - when one region is weak, nearby dies often fail together, so yield cannot be modeled as fully independent Bernoulli events.
**What Is Spatial Correlation in Yield?**
- **Definition**: Dependence between die outcomes as a function of physical distance on the wafer map.
- **Physical Drivers**: Local film non-uniformity, equipment zones, contamination streaks, and thermal gradients.
- **Modeling Impact**: Independent defect assumptions understate risk when clustering exists.
- **Key Metric**: Correlation length, which estimates how far local process effects persist.
**Why Spatial Correlation Matters**
- **Yield Forecast Accuracy**: Clustered failures require non-Poisson models for realistic yield prediction.
- **Root Cause Isolation**: Correlated failure regions point to tool or module-specific issues.
- **Screening Strategy**: Spatial outlier rules can catch latent weak dies that still meet absolute limits.
- **Cost Control**: Better map interpretation reduces unnecessary rework and scrap.
- **Process Monitoring**: Correlation trend shifts are early warning indicators for process drift.
**How It Is Used in Practice**
- **Map Statistics**: Compute spatial autocorrelation metrics such as Moran I or variograms.
- **Cluster Detection**: Identify contiguous fail regions and compare against known tool signatures.
- **Adaptive Action**: Escalate diagnostics when local fail density exceeds control thresholds.
Spatial correlation in yield is **a core manufacturing reality that turns wafer maps from simple pass-fail grids into actionable process diagnostics** - understanding neighborhood dependence is essential for accurate yield management.