Full wafer measurement based on a trained full wafer measurement model

JP2026517537APending Publication Date: 2026-06-02KLA CORP
View PDF 1 Cites 0 Cited by

Patent Information

Authority / Receiving Office
JP · JP
Patent Type
Applications
Current Assignee / Owner
KLA CORP
Filing Date
2024-03-26
Publication Date
2026-06-02

AI Technical Summary

Technical Problem

Current measurement techniques for semiconductor structures face challenges due to high resolution demands, multiple parameter correlations, increasing geometric complexity, and the use of opaque materials, leading to limited measurement accuracy and increased costs, with existing methods failing to utilize wafer-level process information effectively.

Method used

A whole-wafer measurement model is trained using Design of Experiments (DOE) data across an entire wafer, incorporating process behavior information to reduce parameter correlations and improve measurement robustness and accuracy by estimating parameter values at each measurement site.

Benefits of technology

The whole-wafer measurement model enhances measurement performance and robustness by reducing parameter correlations and dimensionality, enabling accurate estimation of parameters across the entire wafer surface.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 2026517537000001_ABST
    Figure 2026517537000001_ABST
Patent Text Reader

Abstract

This specification describes a method and system for measuring semiconductor structures based on a trained whole-wafer measurement model that is valid for any measurement location on a wafer. The whole-wafer measurement model is trained based on Design of Experiments (DOE) measurement data collected across an entire wafer or a set of wafers to which the same series of process steps are applied. By using DOE measurement data from one or a set of wafers, information on the process behavior across the wafer is implicitly incorporated into the trained model at every location across the wafer under measurement. This model training process advances the physical process behavior, thereby reducing the degrees of freedom of the underlying model, eliminating correlations between parameters, and reducing the dimensionality of the solution space. As a result, measurement performance and robustness are improved.
Need to check novelty before this filing date? Find Prior Art