AI Digital Twin Models for Semiconductor Fabrication Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current semiconductor and photonics integrated circuit fabrication processes require extensive trial and error, leading to costly and time-consuming iterations to optimize process flows, making it difficult to achieve desired yields and performance, especially when transferring processes between fabrication facilities or equipment.
Innovation Solution
A physics and chemistry-based artificial intelligence-driven modeling tool and method that uses machine learning to create digital twin models of target devices, optimizing fabrication processes by reducing the number of input features, employing advanced Design of Experiments algorithms, and integrating data visualization, regression, and optimization modules to minimize time and cost.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If traditional trial and error methods with physical wafers are used to optimize process flow, then process optimization can be achieved, but it leads to multiple costly iterations and extended time periods of weeks to months
Solution Approach 1:
The patent creates virtual copies (digital twins) of physical wafers and fabrication processes through AI-driven modeling. These digital replicas allow engineers to test and optimize process parameters virtually without consuming physical materials, thereby eliminating the need for multiple costly physical iterations while maintaining optimization accuracy.
Solution Approach 2:
The system performs preliminary virtual experimentation and process optimization through AI simulations before committing to physical wafer fabrication. By pre-optimizing process flows in the digital domain, the system prevents costly trial-and-error iterations on physical wafers, significantly reducing the overall optimization cycle time.
2Reliability
If extensive Design-of-Experiments testing is performed on physical wafers to optimize process parameters, then desired yield and performance goals can be achieved, but material and labor costs increase significantly
Solution Approach 1:
The patent uses AI-driven digital twin models to create virtual representations of wafers and fabrication processes. These digital copies enable extensive experimentation and process optimization without consuming physical wafer materials, thereby maintaining high device yield while eliminating material waste associated with traditional trial-and-error testing.
Solution Approach 2:
The system replaces physical mechanical experimentation with AI-based computational modeling and simulation. Instead of physically fabricating and testing numerous wafer batches to optimize process parameters, the system uses algorithms to predict optimal settings, eliminating material consumption while achieving the same reliability goals.
3Adaptability or versatility
If process flows are transferred to different fabrication facilities or equipment, then manufacturing flexibility is improved, but process control difficulty increases due to variations in equipment and conditions
Solution Approach 1:
The patent develops AI-driven digital twin models that can universally represent different fabrication equipment and process conditions. These models are designed to adapt to various manufacturing environments, allowing process flows to be transferred between facilities while maintaining consistent control through the standardized digital representation framework.
Solution Approach 2:
The system uses AI algorithms to automatically adjust and optimize process parameters when transferring processes between different fabrication facilities or equipment. By dynamically changing parameters based on digital twin simulations, the system maintains process control despite variations in physical equipment, reducing the complexity of cross-facility process transfer.
Data Source
AI summary
A physics/chemistry-based artificial intelligence driven modeling tool and method of using the same are provided for optimizing fabrication processes for microelectronic devices. Generally the method begins with assembling a training dataset including features of a target device and performance parameters of the process flow from test machine and/or simulation results. The features of the device are narrowed to a number of input features, and the device digitally modeled using a neural network and AI algorithm, based on the input features and performance parameters. The model is analyzed to find an optimal condition for at least one of the input features. A second number of input features for the device are selected based on the optimal condition(s) are found and the modeling and optimizing are repeated using a design of experiment (DoE) advanced algorithm until a digital twin model of the target device is generated.