Optical Spectrum Data Prediction for Semiconductor Electrical Reliability

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

The increasing complexity of semiconductor processes makes it difficult to predict electrical reliability properties of semiconductor devices before shipment, hindering the optimization of process conditions and productivity.

Innovation Solution

A method and device that utilize optical spectrum data to predict electrical reliability properties by generating and processing optical and electrical data through a neural network-based prediction model, allowing for early defect detection and improved process optimization.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Manufacturing precision

If semiconductor process integration degree increases and line width becomes finer, then manufacturing complexity increases, but productivity and defect detection capability remain insufficient

Engineering Contradiction:

Improveline width precisionVSAvoidprocess complexity

Core Design Contradiction:

Manufacturing precisionVSDevice complexity

Solution Approach 1:

The patent applies preliminary action by performing optical inspections and reliability predictions at early process stages (after lithography and etching) rather than waiting for final device completion. The system captures optical data from intermediate process stages and uses machine learning models to predict final device reliability, enabling early defect detection and process optimization before shipment.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent introduces optical data and machine learning prediction models as intermediaries between the complex manufacturing process and the final reliability assessment. Instead of directly measuring electrical properties at each process stage, the system uses optical characteristics as intermediate indicators that correlate with final device performance, simplifying the complexity of direct electrical measurement throughout the process.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Reliability

If electrical reliability properties are verified before shipment, then product quality improves, but development period and time consumption increase

Engineering Contradiction:

Improveelectrical reliabilityVSAvoiddevelopment period

Core Design Contradiction:

ReliabilityVSLoss of time

Solution Approach 1:

The system performs preliminary reliability assessment by analyzing optical data captured during intermediate process stages using trained machine learning models. This allows reliability prediction to occur before final device completion and shipment, eliminating the need for lengthy post-manufacturing testing while maintaining high reliability verification accuracy.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent replaces traditional electrical measurement and testing mechanisms with optical measurement and machine learning prediction. Instead of performing time-consuming electrical reliability tests on finished devices, the system uses optical microscopy data combined with trained prediction models to assess reliability, significantly reducing testing time while maintaining accuracy.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

3Difficulty of detecting and measuring

If optical spectrum data is used to predict electrical reliability, then defect detection capability improves, but data processing complexity increases

Engineering Contradiction:

Improvedefect detection capabilityVSAvoiddata processing complexity

Core Design Contradiction:

Difficulty of detecting and measuringVSDevice complexity

Solution Approach 1:

The system performs preliminary data processing by training machine learning models in advance using paired optical data and electrical measurement results. Once trained, the models can rapidly predict reliability from new optical data without requiring complex real-time processing, separating the complex training phase from the simpler prediction phase.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent transforms complex optical spectrum data into simplified feature representations that the machine learning models can process efficiently. By converting raw optical data into extracted features and using trained models to map these features to reliability metrics, the system reduces the computational complexity of real-time data processing while maintaining defect detection capability.

Inventive Principle:

Principle #35Parameter changes

Data Source

PatentUS20240319259A1Electrical device reliability properties prediction device and electrical device reliability properties prediction method

Publication Date: 2024.09.26 SAMSUNG ELECTRONICS CO LTD

AI summary

Provided is an electrical reliability properties prediction method including generating a plurality of pieces of optical spectrum data of a substrate, performing a wafer level reliability (WLR) process on the substrate, measuring electrical reliability property data based on the WLR process, matching an inspection region to the plurality of pieces of optical spectrum data and the electrical reliability property data, generating a data set, performing data pre-processing, training an electrical reliability properties prediction model, acquiring a plurality of pieces of target optical data from a database, and extracting, with respect to the plurality of pieces of target optical data, a feature vector from the plurality of pieces of target optical data, and detecting predicted electrical reliability property data of the plurality of pieces of target optical data based on the feature vector.