AI Mask Image Generation for Semiconductor Mask Rule Compliance

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing optical proximity correction (OPC) methods for generating semiconductor masks rely heavily on user expertise, limiting performance and speed in determining optimal rules for mask fabrication.

Innovation Solution

Employing artificial intelligence (AI) models to generate mask images by training with target patterns, applying loss functions to correct mask images based on diffraction considerations, and updating the AI models to meet mask rules, including operations like opening and closing to adjust pattern size and spacing.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Manufacturing precision

If traditional OPC methods are used to generate mask images, then mask fabrication rules can be met, but the performance and speed are limited by user expertise in choosing optimal rules

Engineering Contradiction:
Improvemask fabrication rule complianceVSAvoidmask generation speed
Core Design Contradiction:
Manufacturing precisionVSProductivity

Solution Approach 1:

The AI model automatically learns and applies optimal mask correction rules through training on target patterns and corrected mask images, eliminating the need for manual rule selection by users. The system self-optimizes by minimizing loss functions that measure the difference between generated and corrected masks, thereby achieving both rule compliance and high-speed generation without human intervention.

Inventive Principle:
Principle #25Self-service

2Productivity

If AI models are used to generate mask images automatically, then generation speed improves, but the initial model performance may not meet mask fabrication rules

Engineering Contradiction:
Improvemask generation speedVSAvoidmask fabrication rule compliance
Core Design Contradiction:
ProductivityVSManufacturing precision

Solution Approach 1:

The system implements a feedback mechanism where the generated mask images are compared against corrected reference masks using loss functions. The AI model receives feedback through gradient-based optimization, adjusting its internal parameters to minimize the difference between generated and corrected masks. This iterative feedback loop ensures the model learns to produce rule-compliant masks while maintaining high generation speed.

Inventive Principle:
Principle #23Feedback

Solution Approach 2:

The system performs preliminary training actions by pre-processing target patterns and preparing corrected mask images as training data before actual mask generation. The AI model is pre-trained on diverse pattern data to learn fundamental correction rules, enabling it to quickly adapt to specific fabrication requirements while ensuring initial rule compliance.

Inventive Principle:
Principle #10Preliminary action

3Manufacturing precision

If multiple corrections are applied to mask images to meet fabrication rules, then manufacturing precision improves, but the complexity of the generation process increases

Engineering Contradiction:
Improvemask pattern accuracyVSAvoidcorrection process complexity
Core Design Contradiction:
Manufacturing precisionVSDevice complexity

Solution Approach 1:

The system merges multiple correction operations into a single integrated AI model. Instead of applying separate correction steps for different fabrication rules, the model learns to simultaneously satisfy multiple constraints (minimum pattern size, minimum spacing, etc.) through unified training. This consolidation reduces process complexity while maintaining high manufacturing precision by learning the interrelationships between different correction requirements.

Inventive Principle:
Principle #5Merging (Combining)

Data Source

PatentEP4600735A1Method and apparatus with ai model for mask image generation
Publication Date: 2025.08.13 SAMSUNG ELECTRONICS CO LTD
  • EP4600735A1 patent drawingFigure 1
  • EP4600735A1 patent drawingFigure 2A
  • EP4600735A1 patent drawingFigure 2B

AI summary

A method for generating a mask image for fabricating a mask includes: generating the mask image from a target pattern using a first artificial intelligence (AI) model; calculating a first loss function based on a difference between the mask image and a mask image corrected according to a mask rule; and updating the first AI model based on the first loss function.