special cause variation

**Special cause variation** (also called **assignable cause variation**) is process variability that arises from a **specific, identifiable source** — a discrete event or change that pushes the process outside its normal operating behavior. It is the opposite of common cause variation and indicates the process is **out of control**. **Characteristics of Special Cause Variation** - **Identifiable**: A specific root cause can be found and addressed. - **Not Always Present**: Special causes come and go — they represent abnormal conditions, not the system's baseline behavior. - **Detectable by SPC**: Control charts are designed specifically to distinguish special cause variation from common cause variation. - **Correctable**: Once identified, the cause can be fixed to return the process to its in-control state. **How SPC Detects Special Causes** - **Point Beyond 3σ**: A sudden large shift caused by a dramatic event (wrong recipe, hardware failure). - **Trends**: 6+ consecutive points trending upward or downward — gradual degradation of a component. - **Runs**: 8+ consecutive points on one side of the center line — a sustained shift in process mean. - **Clustering**: Points oscillating between the center and one control limit — possible alternating between two states. **Examples in Semiconductor Manufacturing** - **Sudden Shift**: A gas bottle change introduces slightly different gas composition → etch rate shifts by 2%. - **Gradual Drift**: Electrode erosion slowly reduces plasma uniformity over weeks → trending EWMA alarm. - **Intermittent**: A sticking valve occasionally delivers incorrect gas flow → random OOC points. - **Step Change**: A PM restores chamber performance but at a slightly different operating point → sustained offset after PM. **Responding to Special Causes** - **Immediate**: Stop production on the affected tool (for critical steps). - **Investigate**: Use 5-Why analysis, fishbone diagrams, or systematic troubleshooting to find the root cause. - **Correct**: Fix the root cause — not just the symptom. - **Prevent**: Implement controls to prevent recurrence (improved PM procedures, better monitoring, alarm limits). - **Verify**: Confirm the process is back in control through requalification monitoring. **The Statistical Foundation** - In a process with only common cause variation, approximately **99.73%** of points fall within ±3σ of the mean. - A point beyond 3σ has only a **0.27%** chance of occurring naturally — so it very likely indicates a special cause. - Run rules further reduce the probability of false alarms by looking for patterns that are extremely unlikely under common cause alone. Special cause variation is what SPC is designed to detect — identifying and eliminating special causes is the **primary mechanism** by which manufacturing processes are stabilized and improved.