GAA Transistor Scatterometry for Dimple Recess Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional scatterometry and machine learning-based methods struggle to accurately measure individual dimple recesses in gate-all-around (GAA) transistors due to low signal sensitivity and model complexity, particularly during upstream process changes, leading to insufficient measurement of critical parameters like dimple etch in RibbonFET technology.
Innovation Solution
A machine learning-based feed forward technique using a feed forward neural network (FFN) decouples dimple recess measurements from interfering signals by training with transmission electron microscope (TEM) data, allowing for accurate prediction of individual dimple recesses and average recess depths in GAA transistors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional scatterometry is used to measure transistor dimensions, then the measurement is non-destructive and indirect, but the measurement precision is insufficient for accurately measuring certain dimensions in GAA transistors
Solution Approach 1:
The patent introduces a feed forward neural network as an intermediary between the scatterometry measurement system and the dimension extraction process. The FFN processes the optical measurement data and separates the dimple recess signal from interfering signals, enabling accurate measurement without increasing the complexity of the physical measurement apparatus.
Solution Approach 2:
The patent transforms the measurement approach by changing from direct optical measurement to machine learning-based parameter extraction. The FFN learns the relationship between optical signals and physical dimensions, allowing accurate measurement of dimple recesses by identifying patterns in the optical data that correlate with actual dimensions.
2Productivity
If machine learning-based scatterometry is used, then measurement speed increases, but signal sensitivity remains low for individual dimple recesses
Solution Approach 1:
The feed forward neural network acts as a signal processing intermediary that enhances the weak dimple recess signals. By training the FFN with TEM data, it learns to amplify and isolate the subtle optical signatures of individual dimple recesses from the noisy background, maintaining high sensitivity while enabling fast measurement throughput.
Solution Approach 2:
The patent performs preliminary training of the neural network using transmission electron microscope data before actual measurements. This preliminary action prepares the FFN to recognize and sensitively detect dimple recess signals during production measurements, ensuring both high sensitivity and fast throughput without requiring real-time signal enhancement.
3Adaptability or versatility
If conventional measurement techniques are used, then the process is simpler, but adaptability to upstream process changes is insufficient
Solution Approach 1:
The patent implements a dynamic measurement system where the feed forward neural network can adapt to process changes. The FFN is designed to be retrained or adjusted when upstream process variations occur, allowing the measurement system to maintain accuracy despite changes in transistor fabrication processes, unlike static conventional methods.
Solution Approach 2:
The measurement system uses parameter-based adaptation where the neural network learns to adjust its interpretation of optical signals based on process variations. When process parameters change, the FFN can be retrained with new reference data, enabling the system to adapt to different process conditions without changing the physical measurement apparatus.
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
The FFN approach enables fast, high-volume, and early inline detectability of critical parameters in GAA transistors, improving within-wafer uniformity and mitigating short-channel effects by providing precise dimple recess measurements despite upstream process variations.
Implementation Method 1
Scatterometry/Optical Critical Dimension (OCD) ellipsometry is an indirect, non-destructive, non-imaging measurement technique that may be used to measure transistor dimensions
Data Source
AI summary
This disclosure describes systems, methods, and devices for estimating dimple etch recess depth in a gate-all-around transistor. A method may include receiving, by a device, first measurements of the gate-all-around transistor, the first measurements based on first optical data from a spacer etch stage of fabricating the gate-all-around transistor; inputting, by the at least one processor, using a feed forward network, the first measurements to a machine learning model trained to estimate dimple etch recess in the gate-all-around transistor; inputting, by the at least one processor, to the machine learning model, second optical data from a dimple etch stage of fabricating the gate-all-around transistor; and generating, by the at least one processor, using the machine learning model, based on the second optical data and the first measurements, second measurements comprising the first measurements and dimple etch recess estimates for the gate-all-around transistor.