Semiconductor Process Correction Using Sparse-to-Dense Mapping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for determining corrections in semiconductor manufacturing processes face challenges with sparse data models not accurately representing densely measured parameter values, leading to sub-optimal corrections and poor fingerprint descriptions due to crosstalk and over-dimensioning, as well as limitations in throughput and alignment control.
Innovation Solution
A method that involves 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 the accuracy of process adjustments.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If sparse sampling layout is used for measurements, then throughput is improved, but measurement precision deteriorates
Solution Approach 1:
The patent introduces an intermediary mapping model that transforms sparse measurement data into dense measurement equivalents. The model learns the relationship between sparse and dense measurements from training data, then uses this mapping to predict what dense measurements would look like based on sparse inputs, thereby achieving both high throughput and high measurement precision
Solution Approach 2:
The patent creates a virtual copy of dense measurement data by training a mapping model on paired sparse-dense datasets. Once trained, the model generates synthetic dense measurement predictions from sparse inputs, effectively copying the information content of dense measurements without requiring actual dense sampling during production
2Measurement precision
If dense sampling layout is used for measurements, then measurement precision is improved, but productivity deteriorates
Solution Approach 1:
The patent performs preliminary training of the mapping model using dense measurement data during an offline training phase. This preliminary action captures the relationship between sparse and dense measurements, so that during actual production, only sparse measurements are needed and the model handles the rest, avoiding the need for time-consuming dense measurements during throughput-critical operations
3Productivity
If model parameters are based on sparse ADI data, then productivity is improved, but measurement precision deteriorates
Solution Approach 1:
The patent introduces a mapping model as an intermediary layer between sparse measurements and model parameters. Instead of directly using sparse measurements to determine model parameters, the mapping model first transforms sparse measurements into equivalent dense measurements, which then feed into the parameter determination process, ensuring both productivity and precision
4Measurement precision
If model is over-dimensioned for sparse data set, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The patent implements a dynamic two-stage modeling approach where the system adapts its complexity based on data availability. The mapping model dynamically transforms sparse inputs into dense-equivalent representations, allowing the downstream model to operate at optimal complexity without being over-dimensioned for the actual sparse input size
Solution Approach 2:
The patent adds a new dimensional layer to the processing pipeline by introducing the mapping model transformation step. This additional dimension converts the problem from directly modeling sparse data to modeling transformed dense-equivalent data, allowing the use of more sophisticated models without the penalties of over-dimensioning
Data Source
Figure 1~2
Figure 3
Figure 4
AI summary
A method for determining a correction to a semiconductor manufacturing process, the method comprising: - obtaining first sparse data (504) representing measured values of a parameter across one or more substrate subject to the process, measured (502) using a sparse sampling layout; - obtaining dense data (512) representing measured values of the parameter across one or more substrate subject to the process, measured (510) using a dense sampling layout that is more spatially dense than the sparse sampling layout; - applying (506) a model (508) to the sparse data and dense data to determine a sparse-to-dense mismatch; - obtaining second sparse data (522) representing measured values of the parameter across the substrate subject to the process, measured (516) using a sparse sampling layout; - adapting (524) the model (508) based on the sparse-to-dense mismatch; - applying the adapted model to the second sparse data to determine a sparse model result (526); and - determining (542) a correction to the process based on the sparse model result (526).