Semiconductor Wafer Polishing Control With Adaptive Causal Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing techniques for determining control settings in semiconductor wafer polishing are inefficient and fail to accurately model the causal relationships between input settings and environment responses, particularly in dynamic environments where uncontrollable factors affect the quality of the wafers.
Innovation Solution
A control system that employs a causal model to repeatedly select and measure the impact of different control settings on semiconductor wafer quality, updating the model to quickly and accurately determine these relationships, while accounting for uncontrollable environmental characteristics.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If modeling-based techniques are used to determine control settings, then the system can passively observe historical data and learn patterns, but the accuracy of capturing causal relationships is insufficient
Solution Approach 1:
The patent transforms the modeling approach by changing the fundamental parameter from pattern recognition to causal inference. The causal model uses do-calculus and structural causal models to fundamentally alter how relationships are represented, moving from probabilistic correlations to causal mechanisms. This allows the system to accurately determine which control settings actually cause quality improvements rather than merely correlating with them.
2Reliability
If active control techniques are used for knowledge generation, then the system can actively control the environment for experimentation, but the complexity and time consumption increase
Solution Approach 1:
The patent applies preliminary action by pre-specifying the causal structure and relationships before experimentation begins. The structural causal model is built beforehand based on domain knowledge, and the do-calculus framework is established in advance. This allows the system to efficiently conduct experiments by simply collecting data that tests the pre-defined causal hypotheses, rather than exploring all possible relationships through extensive experimentation.
Solution Approach 2:
The system implements feedback by continuously updating the causal model with new observations and using the causal inference results to guide subsequent control settings. The causal model learns from the outcomes of active control experiments and refines its understanding of causal relationships, creating a closed-loop system that improves over time while reducing the need for extensive repeated experimentation.
3Adaptability or versatility
If traditional control models are used, then the system can operate in dynamic environments, but vulnerability to uncontrollable environmental factors increases
Solution Approach 1:
The patent extracts and isolates the causal relationships from the complex environmental noise by using do-calculus to condition on confounding variables. The structural causal model explicitly separates the effects of controllable input settings from uncontrollable environmental factors by identifying and conditioning on backdoor paths. This extraction allows the system to focus only on the relevant causal mechanisms while filtering out environmental interference.
Solution Approach 2:
The causal model acts as an intermediary layer between the input settings and the quality outcomes. Rather than directly observing the relationship between settings and outcomes in a noisy environment, the system uses the causal model to mediate this relationship by incorporating domain knowledge and causal structure. This intermediary filters out environmental noise and provides a clearer understanding of the true causal effects.
Data Source
AI summary
Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for optimizing a process of polishing semiconductor wafers. In one aspect, the method comprises repeatedly performing the following: i) selecting a configuration of input settings for polishing a semiconductor wafer, based on a causal model that measures current causal relationships between input settings and a quality of semiconductor wafers; ii) receiving a measure of the quality of the semiconductor wafer polished with the configuration of input settings; and iii) adjusting, based on the measure of the quality of the semiconductor wafer polished with the configuration of input settings, the causal model.