Semiconductor Process Performance Prediction With Hybrid ML Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current semiconductor manufacturing processes face challenges in predicting the performance of deposition, etch, and clean processes due to the high computational cost and time required by physics-based simulations, and the limitations of data-driven models in capturing complex plasma environments, leading to lengthy design and testing cycles.
Innovation Solution
Hybrid models combining physics-based and data-driven approaches, utilizing machine-learning techniques, are employed to integrate disparate datasets and predict process behaviors, enabling faster and more accurate design iterations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If physics-based simulations are used to predict process performance, then prediction accuracy is improved, but computational cost and time increase
Solution Approach 1:
The patent creates a simplified copy or surrogate model that replicates the behavior of complex physics-based simulations. This surrogate model captures essential process characteristics while requiring minimal computational resources, enabling rapid predictions without full-scale physics simulations.
Solution Approach 2:
The patent transforms complex physics-based simulation parameters into simplified empirical parameters that can be quickly evaluated. By changing the parameter representation from detailed physical quantities to aggregated empirical indicators, the system achieves fast predictions while maintaining acceptable accuracy.
2Productivity
If data-driven models are used to predict process performance, then computational speed is improved, but ability to capture complex plasma environments deteriorates
Solution Approach 1:
The patent combines data-driven models with physics-based insights to create a composite predictive system. This hybrid approach integrates the speed of data-driven methods with the physical accuracy needed for plasma environments, merging empirical correlations with fundamental process understanding.
Solution Approach 2:
The patent introduces physics-based principles as an intermediary layer that guides and constrains the data-driven model. This intermediary ensures that the fast data-driven predictions remain physically plausible and accurate for complex plasma environments by incorporating fundamental process knowledge.
3Loss of information
If high-fidelity simulations are used to understand physical and chemical mechanisms, then understanding of process mechanisms is improved, but design and testing cycles lengthen
Solution Approach 1:
The patent performs preliminary analysis using simplified models to identify key mechanisms and trends before committing to full high-fidelity simulations. This preliminary action filters out less important scenarios, allowing focused use of computationally intensive simulations only where needed, thus shortening overall design cycles.
Solution Approach 2:
The patent divides the design and analysis process into segments: initial screening with fast models, detailed analysis with physics-based simulations for critical cases, and validation with experiments. This segmentation allows teams to maintain deep mechanistic understanding while avoiding the time cost of running high-fidelity simulations for every design iteration.
Data Source
AI summary
Methods, systems, and computer programs are presented for predicting the performance of semiconductor manufacturing equipment operations. One method includes an operation for obtaining machine-learning (ML) models, each model related to predicting a performance metric for an operation of a semiconductor manufacturing tool. Further, each ML model utilizes features defining inputs for the ML model. The method further includes an operation for receiving a process definition for manufacturing a product with the semiconductor manufacturing tool. One or more ML models are utilized to estimate a performance of the process definition used in the semiconductor manufacturing tool. Additionally, the method includes presenting, on a display, results showing the estimate of the performance of the manufacturing of the product. In some aspects, the use of hybrid models improves the predictive accuracy of the system by augmenting the capabilities of data-driven models with the reinforcement provided by the physics-based models.