Semiconductor Device Electrical Parameter Prediction During Fabrication
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
It is challenging to predict the grade of semiconductor devices prior to electrical tests, leading to potential misprocessing and wastage of wafers, as variations in fabrication steps affect device performance, and current methods do not allow for timely detection of issues until late in the fabrication process.
Innovation Solution
A method and system that collect fabrication data during the process, update initial characteristic values with this data to generate modified sets, and predict electrical characteristics of semiconductor devices using a prediction unit within a manufacturing system, incorporating metrology and process data to guide process settings and scheduling.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If electrical tests are performed late in the fabrication process to determine device grade, then device performance can be accurately measured, but wafer wastage increases and throughput decreases due to undetected misprocessing
Solution Approach 1:
The system performs preliminary prediction of device electrical characteristics during fabrication using process data and models, before completing the full fabrication process and performing actual electrical tests. This allows early identification of defective wafers, preventing wastage of subsequent processing steps and maintaining high throughput while still achieving accurate performance assessment.
2Adaptability or versatility
If fabrication process variations are allowed to occur naturally, then manufacturing flexibility is maintained, but device performance consistency deteriorates
Solution Approach 1:
The system continuously monitors fabrication process data and uses prediction models to provide feedback on expected device electrical characteristics. This feedback loop allows the system to maintain manufacturing flexibility while identifying deviations that would lead to poor performance consistency, enabling corrective actions to be taken before defects are finalized.
Solution Approach 2:
The system replaces physical electrical testing with a virtual prediction model that uses process data and algorithms to estimate device electrical characteristics. This substitution allows performance assessment without actual electrical measurement, maintaining flexibility while ensuring consistency through predictive control.
3Ease of operation
If static predetermined priority rules are used for lot processing, then scheduling simplicity is maintained, but responsiveness to actual device performance needs deteriorates
Solution Approach 1:
The system transitions from static predetermined priority rules to dynamic scheduling based on predicted device electrical characteristics. The prediction model provides real-time insights into device performance, allowing the scheduling system to adapt priorities dynamically while maintaining operational simplicity through automated decision-making based on predictive data.
Data Source
AI summary
A method includes providing a set of initial characteristic values associated with the semiconductor device. A first fabrication process is performed on the semiconductor device. Fabrication data associated with the first fabrication process is collected. At least one of the initial characteristic values is replaced with the fabrication data collected for the first fabrication process to generate a first modified set of characteristic values. A first value for at least one electrical characteristic of the semiconductor device is predicted based on the modified set of characteristic values. A system includes a first process tool, a first data collection unit, and a prediction unit. The first process tool is configured to perform a first fabrication process on the semiconductor device. The first data collection unit is configured to collect fabrication data associated with the first fabrication process. The prediction unit is configured to provide a set of initial characteristic values associated with the semiconductor device, replace at least one of the initial characteristic values with the fabrication data collected for the first fabrication process to generate a first modified set of characteristic values, and predict a first value for at least one electrical characteristic of the semiconductor device based on the modified set of characteristic values.