Semiconductor Process Correction Using Sparse-to-Dense CD Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current semiconductor manufacturing processes face challenges in accurately modeling and controlling Critical Dimension (CD) and overlay performance due to sparse data representation, leading to sub-optimal corrections and increased throughput issues, as sparse-to-dense data mismatches result in poor fingerprint descriptions and erratic behavior across different wafers and lots.
Innovation Solution
A method is introduced to address these challenges by obtaining both sparse and dense data sets, applying a model to determine sparse-to-dense mismatches, adapting the model based on these mismatches, and using the adapted model to calculate corrections, thereby improving process control and reducing the need for frequent dense data sampling.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If sparse data sampling is used for process control, then throughput is improved and measurement time is reduced, but measurement precision and manufacturing precision deteriorate due to sparse-to-dense mismatches
Solution Approach 1:
The system performs preliminary dense data measurements and model training in advance to establish a robust predictive model. This pre-established model enables subsequent sparse data measurements to achieve high precision without requiring frequent dense sampling, thus improving throughput while maintaining accuracy
Solution Approach 2:
A predictive model acts as an intermediary between sparse measurements and dense data requirements. The model translates sparse measurements into accurate process corrections by learning the relationship between sparse and dense data patterns, eliminating the need for direct dense sampling in routine control
2Manufacturing precision
If dense data sampling is used for accurate process modeling, then measurement precision and manufacturing precision are improved, but productivity decreases and measurement time increases
Solution Approach 1:
Dense data measurements are performed preliminarily to train the predictive model, but not continuously during routine operation. This preliminary dense sampling establishes the model once or occasionally, while sparse sampling handles ongoing control, thus achieving high precision without continuous productivity loss
Solution Approach 2:
Instead of continuously performing full dense sampling, the system uses partial dense sampling only when needed for model updates or validation. The majority of operations use optimized sparse sampling with the predictive model, reducing measurement time while maintaining sufficient accuracy through the model's compensation
3Productivity
If sparse data is used for process corrections, then productivity is improved, but reliability deteriorates due to erratic behavior and poor fingerprint descriptions across different wafers and lots
Solution Approach 1:
The predictive model serves as an intermediary that stabilizes sparse measurements by learning and compensating for systematic variations across different wafers and lots. The model translates variable sparse measurements into consistent corrections, eliminating erratic behavior while maintaining high throughput
Solution Approach 2:
The system dynamically adjusts model parameters based on learned patterns from dense data to account for variations across different wafers and lots. This parameter adaptation enables reliable process control despite using sparse sampling, as the model compensates for systematic changes in measurement conditions
Data Source
AI summary
A method for configuring a semiconductor manufacturing process, the method including: obtaining a first value of a first parameter based on measurements associated with a first operation of a process step in the semiconductor manufacturing process and a first sampling scheme; using a recurrent neural network to determine a predicted value of the first parameter based on the first value; and using the predicted value of the first parameter in configuring a subsequent operation of the process step in the semiconductor manufacturing process.