Wafer Fabrication Parameter Optimization Using Surrogate Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The process of identifying optimal process parameter values for fabrication processes to achieve target specifications is time-consuming and resource-intensive, often requiring numerous iterations of trial and error, especially when multiple specifications are involved.
Innovation Solution
Utilizing machine learning algorithms and statistical inference techniques to guide process development through a surrogate model, an acquisition function, and an inference engine to iteratively select process parameter values, balancing exploration and exploitation to efficiently converge on optimal parameter settings.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If manual trial and error methods are used to identify process parameter values, then process engineers can achieve target specifications, but the process becomes time-consuming and resource-consuming with many iterations needed
Solution Approach 1:
The system performs preliminary actions by building a surrogate model from historical process data before actual optimization begins. This pre-computed model captures the relationship between process parameters and wafer characteristics, enabling rapid predictions without repeated physical experiments. The acquisition function is also pre-configured with exploration-exploitation balance parameters, allowing the system to immediately begin efficient optimization rather than starting from scratch with manual trial-and-error.
Solution Approach 2:
The surrogate model acts as an intermediary between the complex fabrication process and the optimization algorithm. Instead of directly experimenting with the physical fabrication process, the system uses the surrogate model to predict outcomes, reducing the need for time-consuming physical iterations. The acquisition function serves as another intermediary, translating model predictions into informed decisions about which parameter values to test next, thereby reducing overall development time while maintaining precision.
2Manufacturing precision
If numerous iterations of trial and error are performed to achieve target specifications, then optimal process parameters can be identified, but resource consumption increases significantly
Solution Approach 1:
The system implements feedback through the iterative optimization loop where the acquisition function uses predictions from the surrogate model to select the next set of parameter values to test. After each experiment, new data is fed back into the surrogate model to refine its predictions. This feedback mechanism ensures that each iteration is informed by previous results, avoiding redundant experiments and reducing overall resource consumption while maintaining optimization effectiveness.
Solution Approach 2:
The system changes parameters strategically rather than exhaustively. The acquisition function dynamically adjusts which parameter values to explore based on the current state of knowledge from the surrogate model. This intelligent parameter selection focuses computational and experimental resources on the most promising regions of the parameter space, significantly reducing the number of iterations needed compared to exhaustive or random search methods.
3Manufacturing precision
If manual process development with multiple iterations is used, then target specifications can be achieved, but the complexity of managing multiple experiments increases
Solution Approach 1:
The system performs self-service by automatically managing the entire optimization process without requiring manual intervention for each iteration. The acquisition function autonomously selects the next parameter values to test based on surrogate model predictions, and the system automatically tracks results and updates the model. This automation eliminates the complexity of manually designing, executing, and tracking numerous experiments, while still achieving the same specification targets.
Solution Approach 2:
The surrogate model serves multiple functions simultaneously: it predicts wafer characteristics for new parameter values, identifies regions of high uncertainty for exploration, and guides the acquisition function in selecting optimal experiments. This multi-functionality reduces the need for separate tools and processes for each task, simplifying the overall experiment management while maintaining the ability to achieve target specifications through systematic optimization.
Data Source
AI summary
Methods, systems, and media for optimization of fabrication processes are provided. In some implementations, a method of automatically optimizing fabrication processes comprises: (a) providing a first set of process parameter values associated with a first experiment to a model representing a fabrication process; (b) characterizing a statistical uncertainty of predictions made by the model; (c) using an acquisition function to select a second set of process parameter values, wherein the acquisition function identifies the second set of process parameters based on both: (i) a difference between predicted wafer characteristics and a target specification; and (ii) the statistical uncertainty; (d) receiving results of the fabrication process performed using the second set of process parameter values; and (e) determining whether the performance of the fabrication process generates a post-processed wafer having wafer characteristics that meet the target specification.