Etching Model Grouping for Accurate Cross-Tool Recipe Tuning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Establishing a universal prediction model for etching processing across various etching processing apparatuses and film compositions is challenging due to differences in apparatus specifications, mechanical variations, and film characteristics, making it difficult to adjust setting data effectively for accurate etching processing.
Innovation Solution
An etching processing system that classifies processing conditions into groups based on effects, performs learning processing for each group, updates the prediction model using a test wafer with similar elemental composition and film structure, and searches for optimal setting data to achieve the desired effect, allowing for adjusted setting data application on individual apparatuses.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a universal prediction model is established for multiple etching processing apparatuses, then the model can be applied broadly across different apparatuses and film types, but the model accuracy deteriorates due to apparatus-specific variations and film composition differences
Solution Approach 1:
The patent segments the universal prediction model into apparatus-specific sub-models by performing classification processing that groups processing conditions according to apparatus characteristics. This allows the system to maintain a broad applicable framework while achieving accurate predictions for each specific apparatus by selecting and applying the appropriate segmented model.
Solution Approach 2:
The patent changes parameters by introducing apparatus-specific classification parameters and film composition parameters that dynamically adjust which prediction model is applied. This enables the system to adapt the model parameters based on the specific apparatus and film being processed, resolving the contradiction between universality and accuracy.
2Manufacturing precision
If setting data is adjusted for each specific apparatus and film type, then etching precision is improved, but the complexity of the processing system increases due to multiple prediction models and classification requirements
Solution Approach 1:
The patent performs classification processing and model selection in advance before the actual etching process. By pre-grouping processing conditions and selecting the appropriate prediction model beforehand, the system reduces the complexity during actual operation while maintaining high precision through apparatus-specific adjustments.
Solution Approach 2:
The system performs self-classification and self-selection of prediction models based on the input processing conditions and apparatus characteristics. This automated self-service approach reduces the need for manual intervention and complex external control systems, thereby managing system complexity while achieving precise etching results.
3Measurement precision
If classification processing is performed to group processing conditions, then the prediction model accuracy is improved for each group, but the processing time increases due to additional classification and model selection steps
Solution Approach 1:
The classification processing and model selection are performed as preliminary steps before the actual prediction and etching processes. By completing the classification beforehand and caching the selected model information, the system minimizes the time impact during critical processing stages while maintaining high prediction accuracy through group-specific models.
Data Source
AI summary
An etching processing apparatus includes a storage that stores a learned model of each group generated by each learning processing in a case where each processing condition acquired during execution of a specific step of an etching processing, is classified into a plurality of groups according to a difference in effects when executing the specific step, and the learning processing is performed for each group; an updating unit that updates the learned model of a specific group when an effect of executing the specific step on a test wafer using setting data included in a processing condition associated with the specific group is not equivalent to an effect associated with the specific group; and a searching unit that searches for, using the updated learned model, setting data capable of obtaining the effect associated with the specific group when the specific step is executed on the test wafer.