semiconductor defect inspection
**Semiconductor Defect Inspection** is the **systematic detection and classification of pattern defects, particles, and process anomalies on wafers during fabrication** — enabling rapid identification of yield-killing defects so that process problems can be corrected before hundreds of wafers pass through the defective step, where a single undetected systematic defect can destroy millions of dollars of product.
**Inspection vs. Metrology**
- **Inspection**: Find defects (something wrong present on the wafer).
- **Metrology**: Measure dimensions (whether features are the right size).
- Both critical for process control, but inspection is about finding the unexpected.
**Defect Types**
| Category | Examples | Cause |
|----------|---------|-------|
| Particles | Dust, slurry residue | Contamination, CMP |
| Pattern defects | Bridging, opens, shorts | Litho, etch, deposition |
| Scratch/damage | Mechanical scratches | Handling, CMP |
| Film defects | Pinholes, voids, delamination | CVD, PVD, plating |
| Stacking faults | Crystal defects | Epitaxy, oxidation |
**Inspection Technologies**
| Technology | Resolution | Throughput | Use |
|-----------|-----------|-----------|-----|
| Brightfield optical | ~30 nm | 10-50 WPH | Patterned wafer inspection |
| Darkfield optical | ~20 nm | 50-150 WPH | Particles, macro defects |
| E-beam inspection | ~5 nm | 0.1-1 WPH | Voltage contrast, buried defects |
| E-beam review (SEM) | ~1 nm | Review only | Defect classification |
**KLA Dominance**
- KLA Corporation holds ~80% market share in semiconductor inspection.
- Key tools: 29xx/39xx series (broadband plasma optical inspection).
- Defect sensitivity: Can detect defects smaller than the illumination wavelength (using scattering, interference).
- Each tool: $30-80 million.
**Inspection Flow**
1. **After each critical process step**: Run wafers through inspection tool.
2. **Defect detection**: Tool scans wafer surface, identifies anomalies.
3. **Defect review**: SEM (scanning electron microscope) images each defect at high resolution.
4. **Classification**: Automatic defect classification (ADC) sorts defects by type.
5. **Yield correlation**: Statistical analysis links defect types to yield loss.
6. **Root cause**: Engineering team identifies process problem and corrects it.
**Inline vs. Offline Inspection**
- **Inline**: Inspection tool integrated in fab — automated, wafer never leaves FOUP.
- **Offline**: Manual load for engineering analysis — slower but more flexible.
- Modern fabs: 95%+ of inspections are inline — speed is critical for process control.
**Defect Budget**
- At advanced nodes: < 0.01 defects per cm² (killer defect density).
- A 100 mm² chip can tolerate roughly 1 defect per 100 chips → 99% yield for that defect type.
- Total killer defect density budget: Sum of all process steps must stay below yield target.
Defect inspection is **the guardian of semiconductor yield** — the ability to detect nanometer-scale defects across 300mm wafers at production speed directly determines how quickly yield problems are identified and resolved, making it one of the highest-ROI investments in fab operations.