Pattern Prediction Using 3D Shape Data and Residual Error Functions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional Process Proximity Correction (PPC) methods struggle to accurately predict and correct pattern shapes due to limitations in reflecting three-dimensional variations and etching process biases, leading to inaccuracies in semiconductor fabrication.
Innovation Solution
A pattern prediction method using conversion functions and residual error amount functions to connect contour shapes of patterns across different process steps, incorporating etching process bias models and lithography conversion difference models to correct mask and resist patterns, ensuring high accuracy in pattern formation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If planar (two-dimensional) shape data are used for pattern correction, then the processing time is short, but the prediction accuracy of pattern shapes is low
Solution Approach 1:
The patent transitions from using only two-dimensional planar shape data to incorporating three-dimensional shape data that includes height information. This dimensional enhancement allows the system to capture variations in cross-sectional shapes of resist patterns, thereby improving prediction accuracy without requiring excessively long processing times, as the 3D data structure enables more efficient modeling of process variations.
Solution Approach 2:
The patent introduces additional parameters beyond contour shapes, specifically incorporating height information and other three-dimensional characteristics into the pattern correction model. By changing the parameter set from 2D contour data to 3D shape data including height, the system achieves higher prediction accuracy while managing processing time through optimized calculation methods.
2Manufacturing precision
If three-dimensional shape data are used for pattern correction, then the prediction accuracy of pattern shapes is high, but the processing time is lengthened
Solution Approach 1:
The patent segments the pattern correction process into multiple stages: first obtaining 3D shape data including height information, then using this segmented data in a multi-step correction process that applies different models (etching process bias model, lithography conversion difference model) to specific aspects of the pattern formation. This segmentation allows high accuracy to be achieved while managing computational complexity and processing time.
Solution Approach 2:
The patent performs preliminary acquisition of three-dimensional shape data and preliminary modeling of etching process biases and lithography conversion differences before the main pattern correction calculation. By preparing 3D shape information and relevant models in advance, the system reduces the computational burden during the actual correction process, thereby maintaining high prediction accuracy while improving overall processing efficiency.
3Manufacturing precision
If conventional PPC methods are used, then the process is simple, but the accuracy in reflecting etching process biases and three-dimensional variations is low
Solution Approach 1:
The patent introduces intermediary models that mediate between the complex 3D shape data and the final pattern correction. Specifically, an etching process bias model and a lithography conversion difference model serve as intermediaries that process 3D shape information and translate it into corrective adjustments. This intermediary approach enables high accuracy in reflecting etching biases and 3D variations while managing the complexity of the overall correction system through modular modeling.
Solution Approach 2:
The patent changes the parameter set used in the correction model to include not only contour shapes but also height information and other three-dimensional characteristics. By expanding the parameters to include etching process biases and 3D variations, the model achieves higher accuracy while organizing the increased complexity through systematic parameter management and dedicated modeling approaches for each type of variation.
Data Source
AI summary
A pattern prediction method according to an embodiment includes: predicting a second pattern shape from a first pattern shape by using a conversion function and a conversion difference residual error amount function, wherein; the conversion function makes the connection between the first pattern formed by a first step and the second pattern formed by a second step following the first step based on contour shapes of the first pattern and the second pattern, and the conversion difference residual error amount function makes the connection between a residual error amount between a predicted shape of the second pattern obtained from the conversion function and the second pattern shape obtained by actually using the second step, and factors other than the contour shapes of the first pattern and the second pattern.