GAA Transistor Scatterometry for Dimple Recess Measurement
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional scatterometry methods struggle to accurately measure individual dimple etch recesses in gate-all-around (GAA) transistors due to low signal sensitivity and model complexity, especially with upstream process changes, leading to insufficient measurements and throughput issues.
Innovation Solution
Implementing a machine learning-based feed forward technique using a feed forward neural network (FFNN) to de-correlate recess signals from interfering reflections, trained with transmission electron microscope (TEM) data to provide accurate, individual dimple etch recess measurements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional scatterometry methods are used to measure dimple etch recesses in GAA transistors, then the measurement process is simple, but the signal sensitivity is low and measurement precision is insufficient
Solution Approach 1:
The patent applies preliminary action by training the machine learning model with TEM data beforehand to establish accurate correlations between optical signals and recess dimensions. This pre-training enables the model to de-correlate recess signals from interfering reflections during actual measurements, improving measurement precision without increasing real-time measurement complexity
Solution Approach 2:
The patent introduces machine learning algorithms as an intermediary between traditional scatterometry and final measurements. The ML model acts as a mediator that processes optical signals, separates recess information from interfering reflections, and outputs accurate recess depth measurements, thereby improving signal sensitivity and measurement precision
2Productivity
If traditional scatterometry methods are used, then the measurement approach is straightforward, but throughput is insufficient for high-volume manufacturing
Solution Approach 1:
The patent replaces traditional mechanical/optical scatterometry analysis with machine learning-based signal processing. The ML model rapidly analyzes optical signals to extract individual recess depth information, achieving both high throughput for high-volume manufacturing and accurate individual recess measurements simultaneously
3Adaptability or versatility
If traditional scatterometry is used, then the measurement system is simple, but it cannot effectively handle upstream process changes
Solution Approach 1:
The patent applies parameter changes by training the machine learning model with TEM data that captures variations in upstream process parameters. This enables the model to adapt to process changes by learning the relationships between process variations and resulting recess characteristics, improving adaptability without requiring system redesign
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
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 etch recess measurements.
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.