Scatterometric APC Modeling for Semiconductor Knob Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional process control methods in semiconductor manufacturing face challenges with increasing accuracy demands due to high dimensional knob spaces and variability across multiple production lines, especially when applying machine learning techniques, and require expensive reference parameters for training.
Innovation Solution
A machine learning model is trained using scatterometric data before and after processing steps, correlating process control knob settings to reduce variations in pattern parameters without relying on expensive reference parameters, employing neural networks with encoder-decoder structures and dual loss functions to optimize knob settings.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If traditional process control methods are used, then process control is simple, but manufacturing precision deteriorates due to high dimensional knob spaces and variability across multiple production lines
Solution Approach 1:
The patent introduces scatterometric data as an intermediary measurement that captures optical properties of patterns at wafer sites. This intermediary data serves as a bridge between process control knobs and manufacturing outcomes, enabling precise control without directly measuring all dimensional parameters. The scatterometric measurements act as a mediator that translates complex process variations into actionable insights for control adjustments.
Solution Approach 2:
The patent transforms the control approach by changing from direct dimensional measurement to optical property measurement. By measuring scatterometric data (optical properties) instead of directly measuring critical dimensions, the system achieves higher precision control. The parameter transformation allows the system to handle high dimensional knob spaces more effectively by working in an optimized measurement space.
2Manufacturing precision
If machine learning techniques are applied to improve APC, then manufacturing precision improves, but loss of information increases due to difficulty in training with high dimensional knob spaces
Solution Approach 1:
The patent extracts the essential information needed for process control by focusing scatterometric measurements on specific optical properties that are most relevant to pattern formation. Rather than attempting to process all possible process parameters, the system extracts and measures only the critical optical characteristics, thereby preserving the most important information while filtering out noise and redundancy in high dimensional spaces.
Solution Approach 2:
The patent performs preliminary scatterometric measurements before final process adjustments are made. These preliminary measurements provide advance information about pattern properties, allowing the machine learning model to predict optimal control knob settings before actual manufacturing deviations occur. This preliminary action enables proactive rather than reactive process control.
3Manufacturing precision
If reference parameters are used for training machine learning models, then manufacturing precision improves, but loss of substance increases due to expensive reference measurements
Solution Approach 1:
The patent creates optical copies (scatterometric measurements) of the physical patterns at wafer sites. Instead of using expensive reference measurements that consume material or require destructive sampling, the system uses non-destructive optical copying techniques to capture pattern properties. These optical copies provide sufficient training data for machine learning models without the cost and material loss associated with traditional reference measurement methods.
Solution Approach 2:
The patent replaces expensive, resource-intensive reference measurements with inexpensive, rapidly acquirable scatterometric data. The optical measurements are cheap to obtain, require no consumable materials, and can be performed quickly and repeatedly. This substitution enables abundant training data collection without the substantial costs associated with traditional reference parameter acquisition.
4Productivity
If multiple production lines with multiple manufacturing routes are used, then productivity increases, but manufacturing precision deteriorates due to time scales drifts and variability
Solution Approach 1:
The patent develops a universal scatterometric measurement approach that works across multiple production lines and manufacturing routes. The optical measurement technique is line-independent and route-agnostic, providing consistent data format and measurement principles regardless of which specific manufacturing equipment or process sequence is used. This universality enables the machine learning model to generalize across diverse production environments, maintaining precision while supporting high productivity through multiple concurrent manufacturing routes.
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 model effectively recommends process control knob settings to minimize variations in pattern parameters, enhancing process uniformity and accuracy in semiconductor manufacturing without the need for costly reference measurements.
Implementation Method 1
Scatterometric data (also referred to herein as 'spectra data') is typically acquired as reflected light radiation that is indicative of optical properties of patterns at wafer sites
Data Source
AI summary
A system and methods for Advance Process Control (APC) in semiconductor manufacturing include: for each of a plurality of waiter sites, receiving a pre-process set of scatterometric training data, measured before implementation of a processing step, receiving a corresponding post-process set of scatterometric training data measured after implementation of the process step, and receiving a set of process control knob training data indicative of process control knob settings applied during implementation of the process step; and generating a machine learning model correlating variations in the pre-process sets of scatterometric training data and the corresponding process control knob training data with the corresponding post-process sets of scatterometric training data, to train the machine learning model to recommend changes to process control knob settings to compensate for variations in the pre-process scatterometric data.