Wafer Defect Prediction Using Non-Correctable Error Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The semiconductor manufacturing process faces challenges in efficiently detecting and identifying defects on wafers, particularly in high-throughput production, as current inspection methods are time-consuming and require extensive operator intervention, leading to reduced yield and increased costs.
Innovation Solution
A method is developed to predict defect locations on wafers using non-correctable error information from photolithography systems, employing computational models and defect detection algorithms to identify areas requiring inspection, thereby reducing the number of locations that need to be inspected and increasing throughput.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If every location on the wafer is inspected with a microscope, then defect detection completeness is improved, but inspection time increases and throughput decreases
Solution Approach 1:
The system performs preliminary defect prediction using a computational model that analyzes process data from photolithography and etching steps before actual inspection. This preliminary action identifies high-probability defect locations, allowing the inspection system to focus only on these areas rather than scanning the entire wafer, thus maintaining high defect detection completeness while significantly reducing inspection time and improving throughput
Solution Approach 2:
Instead of uniform inspection across the entire wafer surface, the system applies local quality by concentrating inspection resources on specific regions predicted to have defects. The computational model generates a defect probability map that highlights areas requiring detailed inspection, allowing the microscope to target only those locations with high defect probability, thereby optimizing the balance between detection completeness and inspection efficiency
2Measurement precision
If comprehensive wafer inspection is performed, then defect identification accuracy is improved, but inspection time and costs increase
Solution Approach 1:
The computational model performs preliminary analysis of process parameters and historical defect data to predict defect locations before inspection begins. This preliminary action provides the inspection system with a targeted list of high-probability defect areas, enabling accurate defect identification to be achieved by focusing inspection efforts on these predetermined locations rather than conducting a time-consuming comprehensive scan of the entire wafer
Solution Approach 2:
The computational prediction model acts as an intermediary between the photolithography/etching processes and the physical inspection system. It processes process data and generates defect probability predictions that guide the inspection system, serving as a mediator that translates process information into actionable inspection targets, thereby improving defect identification accuracy while reducing the time required for actual microscopic inspection
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach significantly reduces inspection time and increases wafer throughput by focusing on predicted defect locations, enhancing overall yield and maintaining high production efficiency while minimizing operator intervention.
Implementation Method 1
predicting optical imaging effects of heating on optical elements of the projection optical system
Implementation Method 2
predicting optical imaging effects of heating on optical elements of the projection optical system
Data Source
AI summary
A computer-implemented defect prediction method for a device manufacturing process involving processing a pattern onto a substrate. Non-correctable error is used to help predict locations where defects are likely to be present, allowing improvements in metrology throughput. In an embodiment, non-correctable error information relates to imaging error due to limitations on, for example, the lens hardware, imaging slit size, and/or other physical characteristics of the lithography system. In an embodiment, non-correctable error information relates to imaging error induced by lens heating effects.