Photomask Pattern Generation Using ML Hotspot Detection

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Conventional semiconductor manufacturing processes face challenges with hotspots in photomask designs that can damage semiconductor devices, as some hotspots are difficult to detect during the manufacturing process.

Innovation Solution

A system utilizing a machine learning-based hotspot detection model generates a hotspot image from a design layout image, identifying adjacent hotspot objects and separating them into distinct photomask patterns to prevent defects in the semiconductor substrate.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If conventional photomask design methods are used, then manufacturing process is simple, but hotspots occur and may damage semiconductor devices

Engineering Contradiction:

Improvedefect preventionVSAvoidphotomask design complexity

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The system performs preliminary hotspot detection and identification during the photomask design phase using machine learning models. By detecting potential hotspots before photomask manufacturing, the system prevents defects proactively rather than reacting to them during semiconductor fabrication, thus improving reliability while managing design complexity through automated analysis.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent replaces conventional manual or rule-based hotspot detection methods with machine learning-based automated detection systems. This substitution enables more accurate and comprehensive hotspot identification, improving defect prevention capability while the automation helps manage the complexity of analyzing complex photomask designs.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Reliability

If comprehensive hotspot detection is performed, then defect prevention improves, but detection and measurement difficulty increases

Engineering Contradiction:

Improvehotspot detection accuracyVSAvoidhotspot identification complexity

Core Design Contradiction:

ReliabilityVSDifficulty of detecting and measuring

Solution Approach 1:

The system replaces difficult manual hotspot detection with machine learning-based automated detection. The machine learning models automatically analyze photomask designs, identify hotspot patterns, and predict potential defects, thereby improving detection accuracy while reducing the complexity and difficulty of the detection process through automation.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The system creates hotspot detection models trained on representative photomask design data. These models serve as virtual copies that can rapidly analyze new designs without requiring manual expertise, enabling comprehensive hotspot detection while simplifying the measurement and identification process through pattern recognition.

Inventive Principle:

Principle #26Copying

Data Source

PatentUS11320742B2Method and system for generating photomask patterns

Publication Date: 2022.05.03 TAIWAN SEMICONDUCTOR MANUFACTURING CO LTD

AI summary

The present disclosure provides a method and a system for generating photomask patterns. The system obtains a design layout image, and generates a hotspot image corresponding to the design layout image based on a hotspot detection model. The system generates two photomask patterns based on the hotspot image. The at least two photomask patterns are transferred onto a semiconductor substrate.