defect review
**Defect Review** is the **high-resolution imaging step that follows optical wafer inspection**, in which a scanning electron microscope (SEM) navigates to the coordinates of each flagged defect to capture a detailed image — converting the inspection tool's abstract "something is anomalous at (X,Y)" into a classified, identifiable defect image that enables root cause analysis, process debugging, and yield learning.
**Why Review Is Necessary**
Optical inspection tools operate at high throughput (100+ wafers/hour) using visible or UV light, achieving ~30–100 nm detection sensitivity. However, the resulting images have insufficient resolution to distinguish a metallic particle from a dielectric void, or a bridging short from a pattern roughness artifact. Without review, engineers see defect counts but cannot determine what the defects are — making corrective action impossible.
**Defect Review SEM (DR-SEM) Workflow**
**Coordinate Transfer**: The optical inspection tool outputs a KLARF file containing defect (X,Y) coordinates in wafer reference frame. The DR-SEM (KLA eDR7380, Hitachi RS-3000) imports this file, converting coordinates to stage positions using calibrated wafer alignment.
**Auto Navigation**: The SEM stage drives autonomously to each defect coordinate, centers the beam on the flagged location, and captures a high-resolution SEM image (5–50 nm pixel size, 3–20 kV beam energy). A typical DR run images 50–200 defects per wafer at throughput of ~30–60 defects/hour.
**Image Capture**: Each defect is imaged at two magnifications — a low-mag context image (showing surrounding pattern) and a high-mag detail image (showing defect morphology). The SEM's spatial resolution (< 2 nm) and materials contrast (Z-contrast in backscatter mode) reveal particle composition, shape, dimensions, and relationship to the underlying pattern.
**Defect Classification Output**
From the SEM images, engineers classify each defect into categories: Particle (in-contact or nearby), Bridge/Short, Missing Feature, Void, Scratch, Crystal Defect, Etch Residue, Deposition Blob — each pointing to different process modules and failure mechanisms.
**Integration with ADC**: Modern DR-SEMs feed images directly to Automated Defect Classification (ADC) engines that apply machine learning classifiers to categorize defects without human review of each image — enabling real-time feedback at production throughput.
**Defect Review** is **the forensic microscopy step** — zooming from the "license plate number" provided by optical inspection to the "mugshot" resolution of SEM that reveals exactly what each defect is and provides the visual evidence needed to trace it back to its process source.