AI Wafer Defect Detection System Using Dynamic Inspection Switching

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

The complexity of semiconductor photolithography processes leads to various types of wafer defects, necessitating a quick and accurate detection method to prevent yield decreases in semiconductor manufacturing.

Innovation Solution

An AI system integrating a machine vision CIS device and deep learning for real-time monitoring, which includes a server configuration to learn and distinguish defects, control equipment operations, and perform total inspections, thereby preventing continuous defect occurrence by immediately stopping affected equipment.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If periodic sampling inspection is used, then productivity is maintained, but measurement precision and defect detection accuracy deteriorate

Engineering Contradiction:
Improvedefect detection accuracyVSAvoidinspection throughput
Core Design Contradiction:
Measurement precisionVSProductivity

Solution Approach 1:

The inspection system dynamically adjusts the inspection method based on defect occurrence. When defects are detected, the system transitions from periodic sampling inspection to total inspection, and can further escalate to stopping equipment operation. This dynamic adaptation allows the system to maintain high productivity during normal operation while ensuring high detection accuracy when defects occur.

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The AI system continuously monitors inspection results and provides feedback to adjust the inspection strategy. When defects are detected in periodic sampling, the system triggers a switch to total inspection method, and if defects persist, it triggers equipment stoppage. This feedback mechanism ensures that measurement precision is maintained without permanently sacrificing productivity.

Inventive Principle:
Principle #23Feedback

2Measurement precision

If total inspection method is implemented, then measurement precision improves, but productivity decreases

Engineering Contradiction:
Improvedefect detection accuracyVSAvoidinspection throughput
Core Design Contradiction:
Measurement precisionVSProductivity

Solution Approach 1:

The system implements total inspection dynamically and temporarily, only when defects are detected in periodic sampling. Once the defect cause is identified and corrected, the system can return to periodic sampling inspection, thereby maintaining high productivity while achieving high detection accuracy when needed.

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The system alternates between periodic sampling inspection and total inspection based on defect occurrence. During normal operation, periodic sampling maintains productivity. When defects occur, total inspection is activated temporarily to ensure complete detection, then the system returns to periodic sampling, creating a periodic cycle that balances both productivity and detection accuracy.

Inventive Principle:
Principle #19Periodic action

3Reliability

If equipment operation is not stopped upon defect detection, then productivity is maintained, but reliability deteriorates due to continuous defect occurrence

Engineering Contradiction:
Improvedefect prevention capabilityVSAvoidmanufacturing output
Core Design Contradiction:
ReliabilityVSProductivity

Solution Approach 1:

The system dynamically determines whether to stop equipment based on defect patterns. For isolated defects, periodic sampling continues without stoppage, maintaining productivity. For recurring or systematic defects, the system triggers equipment stoppage to prevent continuous defect occurrence, thereby improving reliability while minimizing impact on productivity.

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The AI system automatically analyzes defect patterns and makes decisions about equipment stoppage without external intervention. When defects are detected, the system self-determines whether to maintain operation or stop equipment based on the severity and pattern of defects, enabling autonomous reliability management while preserving productivity when safe to do so.

Inventive Principle:
Principle #25Self-service

Data Source

PatentUS20240419158A1Ai system for wafer defect detection
Publication Date: 2024.12.19 APPLE T CO LTD
  • US20240419158A1 patent drawing
  • US20240419158A1 patent drawing
  • US20240419158A1 patent drawing

AI summary

An artificial intelligence (AI) system for wafer defect detection may include a CIS device configured to mount to track equipment and to transmit a scanned image of a wafer to a database; and a server configured to learn the scanned image, to distinguish a defect of the wafer for each defect category, to learn defects that occur in exposure equipment and the track equipment for each type, and to perform real-time monitoring. The server is configured to change a periodic sampling inspection to a total inspection method through the CIS device in response to occurrence of the defect in the wafer.