Bayesian Model Decomposition for Semiconductor Tool Mismatch

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing methods for identifying and correcting mismatched performances between semiconductor processing devices rely on human intuition and are prone to cognitive biases, leading to longer times to identify root causes of engineering failures.

Innovation Solution

The use of Bayesian modeling and decomposition to identify intrinsic and extrinsic variations, allowing for the adjustment of device operating parameters to optimize results, by accessing a network of causal relationships between processes and hardware in semiconductor processing devices.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If human intuition and expert ranking systems are used to identify root causes of engineering failures, then domain knowledge can be leveraged, but the process is prone to cognitive biases and takes longer to identify root causes

Engineering Contradiction:

Improveaccuracy of root cause identificationVSAvoidtime to identify root causes

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The patent replaces the mechanical human expert ranking system with an automated computational system that uses statistical models and machine learning algorithms to objectively analyze performance data, eliminate cognitive biases, and rapidly identify root causes of engineering failures

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The patent introduces an intermediary computational layer between raw performance data and root cause identification, using statistical models and data analysis algorithms to process and interpret data objectively, thereby improving both accuracy and speed of failure analysis

Inventive Principle:

Principle #24Intermediary (Mediator)

2Reliability

If human expert ranking systems are used to assess failure risks, then domain expertise can be applied, but cognitive biases affect the rankings and reduce objectivity

Engineering Contradiction:

Improveobjectivity of failure risk assessmentVSAvoidcomplexity of assessment system

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The patent substitutes the human expert assessment system with an automated computational system that applies consistent statistical methods and algorithms, eliminating cognitive biases and ensuring objective, reproducible failure risk assessments across all evaluations

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The patent transforms subjective expert judgments into objective quantitative parameters through statistical modeling, converting qualitative risk assessments into measurable data points that can be consistently analyzed and compared

Inventive Principle:

Principle #35Parameter changes

3Measurement precision

If comprehensive analysis of all system functions is performed to identify failure causes, then thoroughness is improved, but the complexity of analysis increases and time required grows

Engineering Contradiction:

Improvethoroughness of failure analysisVSAvoidcomplexity of analysis system

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The patent segments the complex system into modular functional components, allowing individual analysis of each component's contribution to failures while maintaining the ability to synthesize comprehensive results, thereby reducing analytical complexity without sacrificing thoroughness

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent employs automated computational algorithms and machine learning models to perform comprehensive system-wide analysis, replacing manual analytical methods with computational systems that can efficiently process complex multi-functional relationships

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Data Source

PatentUS12314043B2Bayesian decomposition for mismatched performances in semiconductor equipment

Publication Date: 2025.05.27 APPLIED MATERIALS INC

AI summary

Bayesian inference and modeling techniques, along with model decomposition may be used to improve mismatch performances in semiconductor processing devices by identifying sources of intrinsic and extrinsic variations in performance. A network of causal relationships between processes and hardware in a semiconductor processing device may be accessed to generate a first Bayesian model for a first semiconductor processing device using the causal relationships in the network and performance data. A second Bayesian model may also be generated for a second semiconductor processing device using the causal relationships in the network and associated performance data. Response distributions generated by the first Bayesian model and the second Bayesian model may be compared to determine whether a performance of the of the first semiconductor processing device matches a performance of the second semiconductor processing device by decomposing the model transfer functions to identify the effects of intrinsic/extrinsic variables.