Key pointSpatial morphology guides the selection of the next verification check; it cannot automatically translate rings, repeating spots, or scratches into a single definitive root cause.
Wafer Map
Graphical representation of inspection or metrology measurements mapped to wafer coordinates
Shot
Single lithography step-and-scan optical exposure area containing one or multiple dies
Review SEM
High-resolution electron microscopy used for defect morphology and composition review
Care Area
Designated pattern region on the wafer subjected to focused inspection or metrology

Three Scales of Observation Answer Three Engineering Questions

  • Full Wafer Scale:Where do anomalies concentrate, and how do they vary with radius or azimuthal orientation?
  • Exposure Field / Die Scale:Do defects recur at identical relative coordinates across repeated exposure fields?
  • Microscopic Scale:What physical morphology, material contrast, and structural characteristics appear at representative sites?

Repeated anomalies within exposure fields may relate to masks, layout hotspots, process sensitivity, or inspection artifacts. Top-view SEM, FIB-prepared cross-sections, and EDS elemental analysis are distinct techniques, not universal evidence already available at every inspection point.

Process IllustrationProgressing from Macro Clues to Local Defect Review
  1. 01
    Macro ViewIsolate the Anomaly Region

    Annular rings, edge clusters, or linear scratches indicate where to focus investigation efforts.

  2. 02
    Field ViewEvaluate Die and Exposure Field Periodicity

    Repeated positions may involve masks, layout, process conditions, or inspection algorithms; comparison is needed.

  3. 03
    Micro ViewInspect Representative Defect Morphology

    Local SEM or other review helps classify morphology; cross-sections and elemental analysis may require additional preparation.

  4. 04
    SynthesisSynthesize Observations Across All Three Scales

    Preserve coordinates, inspection coverage limits, and sampling bias to formulate discriminative verification checks.

Synthetic multiscale illustration; enlarged views are not real nanoscale images recovered directly from full-wafer pixels.

Comparing Defect Density: Define the Denominator First

Defect density is defined as the count of selected defects divided by the effective inspected area. Before comparing runs, align care areas, inspection recipes, classification bins, and optical sensitivity thresholds. When inferring full-wafer counts from sparse sampling, state estimation uncertainty rather than multiplying by a crude sampling ratio.

Advanced: Inspection Density Does Not Equal Killer Defect Density

Raw inspection hits frequently include nuisance detections, repeating non-critical patterns, or benign surface dust. The killer defect density used in yield models cannot be directly replaced by raw defect counts. Nonetheless, raw counts remain highly informative when optical coverage and inspection recipes are held strictly identical.

Edge Exclusion and Clustering: Do Not Remove Real Anomalies

Edge exclusion zones, care areas, and false alarm filtering must be configured according to process and metrology objectives; wafer edges may represent the exact locus of current process excursions. Defect clustering distance thresholds must reflect coordinate accuracy and physical mechanisms, rather than applying a single rigid radius to scratches and particle clouds alike.

From Visual Clues to Executable Checks

For edge enrichment, compare reference wafers over the same region, check inspection-sensitivity changes, and use actual histories to choose process steps to examine. If new SEM sampling is needed, propose locations and objectives; automatic sampling dispatch depends on interfaces, authorization, and validation.

WAFERPILOT / APPLICATION TO THIS TOPIC

Preserving the Evidence Basis for Every Visual Observation

WaferPilot's ImageLens module analyzes ready inspection images and defect maps to generate source-attributed findings. The goal is to synthesize full-wafer patterns, local morphology, and process conditions into verifiable recommendations. Inline review dispatch and deeper data integration build upon fab interfaces and on-site qualification.

Deliverables for Engineers:Inspectable findings linked back to original wafer maps and defect logs, with highlighted representative regions, review objectives, and prioritized verification steps.

View the Current Demo and Capability Scope →

Related Reading & References

Below are verifiable industry references; specific processes, permissions, and acceptance criteria are governed by fab site specifications.