Multi-Stage Etching Parameter Control for Hard Mask Stack Variations
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The multi-stage etching technology faces challenges in accurately setting numerous variation factors due to material changes and varying stacking relationships of hard mask layers, making it difficult to achieve the desired etching target.
Innovation Solution
A machine learning model is employed to generate parameter setting recipes for hard mask layers based on stack information and etching target conditions, enabling precise control of etching processes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If traditional monitoring of one or two hard mask layers is used, then the monitoring process is simple, but it cannot achieve accurate control when material changes and stacking relationships vary
Solution Approach 1:
The machine learning model is designed to universally handle multiple hard mask layer configurations, material compositions, and stacking relationships. Instead of creating separate monitoring systems for each scenario, a single multi-functional model processes diverse inputs (stack information, material types, layer configurations) to generate appropriate parameter settings, thereby achieving accurate etching control across varying conditions without proportionally increasing system complexity
Solution Approach 2:
The system dynamically adjusts etching parameters based on input stack information and material characteristics. The machine learning model processes variations in hard mask layer parameters (material composition, thickness, stacking order) and automatically optimizes etching conditions, transforming the approach from static monitoring to adaptive parameter control that maintains precision across different configurations
2Manufacturing precision
If numerous variation factors are manually set by engineers, then the etching process can be controlled, but it creates a technical development bottleneck due to difficulty in setting
Solution Approach 1:
The machine learning model performs self-service by automatically analyzing stack information, material compositions, and process history to generate optimized parameter settings without requiring manual engineer intervention. The system serves itself by learning from past data and autonomously determining the optimal etching parameters for new configurations, eliminating the bottleneck of manual parameter setting while maintaining high precision
Solution Approach 2:
The system performs preliminary analysis of stack information and material characteristics before the actual etching process. By pre-processing the input data and predicting optimal parameters in advance through machine learning, the system prepares the necessary configuration settings beforehand, making the actual etching operation straightforward and eliminating the complexity of real-time manual parameter adjustment
3Manufacturing precision
If multi-stage etching is used to reduce etching width, then the etching target precision is improved, but the process complexity increases due to multiple hard mask layers
Solution Approach 1:
The machine learning model serves as a universal controller that handles the complexity of multiple hard mask layer configurations. It processes diverse stack information (number of layers, material types, thicknesses, stacking orders) and generates coordinated parameter settings for all stages, thereby managing the inherent complexity of multi-stage etching while maintaining the precision benefits of the approach
Data Source
AI summary
A control method of a multi-stage etching process and a processing device using the same are provided. The control method of the multi-stage etching process includes the following step S. A stack information of a plurality of hard mask layers is set. An etching target condition is set. Through a machine learning model, a parameter setting recipe of the hard mask layers is generated under the etching target condition. The machine learning model is trained based on the stack information of the hard mask layers, a plurality of process parameters and a process result.