Machine learning based yield prediction

Patent metadata

Field	Value
Publication number	US-12469124-B2
Application number	US-202318113030-A
Country	US
Kind code	B2
Filing date	Feb 22, 2023
Priority date	Feb 22, 2023
Publication date	Nov 11, 2025
Grant date	Nov 11, 2025

Abstract

There is provided a system and method of examination of a semiconductor specimen. The method includes obtaining an e-beam image representative of a given layer of a given structure on the specimen in runtime, processing at least the e-beam image using a ML model, and obtaining yield related prediction with respect to the given structure prior to performing an electrical test. The ML model is previously trained using a training set comprising multiple stacks of e-beam images corresponding to multiple sites of the given structure on one or more training specimens, each stack of e-beam images representative of the at least given layer of a respective site; and test data acquired from an electrical test performed at the multiple sites and related to actual yield of the training specimens, the test data respectively correlated with the stacks of e-beam images and used as ground truth thereof.

First claim

The invention claimed is:

1 . A computerized system of examining a semiconductor specimen, the computerized system comprising a processor and memory circuitry (PMC) configured to:
obtain an electron beam (e-beam) image representative of a given layer of a given structure on the semiconductor specimen, the e-beam image acquired in runtime during in-line examination of the semiconductor specimen along a fabrication process thereof; and process at least the e-beam image using a machine learning (ML) model, and obtain, as an output of the ML model, yield related prediction with respect to the given structure prior to performing an electrical test thereon, wherein the ML model is previously trained during a training phase using a training set pertaining to at least the given layer, the training set comprising:
a plurality of stacks of e-beam images corresponding to a plurality of sites of the given structure on one or more training specimens, each stack of e-beam images representative of the at least the given layer of a respective site; and
test data acquired from an electrical test performed at the plurality of sites and related to actual yield of the training specimens, the test data respectively correlated with the plurality of stacks of e-beam images and used as ground truth thereof.

2 . The computerized system according to claim 1 , wherein the training set further comprises one or more stacks of optical data acquired by an optical tool and representative of one or more regions on the training specimens that contain the plurality of sites, the one or more stacks of optical data, together with the plurality of stacks of e-beam images, respectively correlated with the test data.

3 . The computerized system according to claim 2 , wherein the one or more stacks of optical data comprise at least one of optical spectrum data or optical image data captured from the one or more regions, and are informative of at least one of geometrical or material properties of the one or more regions.

4 . The computerized system according to claim 1 , wherein the training set further comprises sensor data acquired at the plurality of sites by one or more process sensors, the sensor data respectively correlated with the test data and being informative of physical properties of the plurality of sites.

5 . The computerized system according to claim 1 , wherein the training set further comprises synthetic e-beam images acquired by performing simulation based on design data of the given structure.

6 . The computerized system according to claim 1 , wherein each stack of e-beam images is associated with metadata informative of fabrication properties of the respective site.

7 . The computerized system according to claim 1 , wherein the yield related prediction with respect to the given structure represents estimated functional defectivity of the given structure following a processing step manufacturing the given layer.

8 . The computerized system according to claim 1 , wherein the one or more training specimens each comprise a plurality of layers, and a first ML model is respectively trained for each individual layer of a sampled set of layers from the plurality of layers using a training set pertaining only to the individual layer.

9 . The computerized system according to claim 1 , wherein the one or more training specimens each comprise a plurality of layers, and a second ML model is respectively trained for each individual layer of a sampled set of layers from the plurality of layers using a training set pertaining to the individual layer and one or more preceding layers thereof.

10 . The computerized system according to claim 1 , wherein the processing comprises obtaining one or more e-beam images acquired for one or more preceding layers of the given layer, and processing the e-beam image and the one or more e-beam images together using the ML model, wherein the ML model is trained using a training set pertaining to the given layer and the one or more preceding layers.

11 . The computerized system according to claim 1 , wherein the PMC is further configured to select the ML model from: a first ML model trained using a training set pertaining to the given layer, and a second ML model trained using a training set pertaining to the given layer and one or more preceding layers thereof.

12 . The computerized system according to claim 11 , wherein the selection of the ML model is based on at least one factor of: processing time for obtaining the yield related prediction, or estimated correlation between the yield related prediction and actual yield of the semiconductor specimen.

13 . The computerized system according to claim 1 , wherein the semiconductor specimen comprises a plurality of layers upon completion of the fabrication process thereof, and the PMC is configured to perform the obtaining and the processing for a sampled set of layers from the plurality of layers during the fabrication process, giving rise to yield related prediction for each sampled layer, and analyze the yield related prediction for each sampled layer to provide a yield alert.

14 . A computerized method, the method comprising:
obtaining a training set pertaining to at least one layer of a given structure, comprising:
a plurality of stacks of electron beam (e-beam) images corresponding to a plurality of sites of the given structure on one or more training specimens, each stack of e-beam images representative of the at least one layer of a respective site; and
test data acquired from an electrical test performed at the plurality of sites and related to actual yield of the training specimens, the test data respectively correlated with the plurality of stacks of e-beam images and used as ground truth thereof; and
training a machine learning (ML) model using the training set, wherein the trained ML model is usable to provide yield related prediction with respect to the given structure in runtime based on at least one e-beam image representative of the at least one layer of the given structure on a semiconductor specimen to be examined, the at least one e-beam image acquired during in-line examination of the semiconductor specimen along a fabrication process thereof.

15 . The computerized method according to claim 14 , wherein the training set further comprises one or more stacks of optical data acquired by an optical tool and representative of one or more regions on the training specimens that contain the plurality of sites, the one or more stacks of optical data, together with the plurality of stacks of e-beam images, respectively correlated with the test data.

16 . The computerized method according to claim 14 , wherein the one or more training specimens each comprise a plurality of layers, and a first ML model is respectively trained for each individual layer of a sampled set of layers from the plurality of layers using a training set pertaining only to the individual layer.

17 . The computerized method according to claim 14 , wherein the one or more training specimens each comprise a plurality of layers, and a second ML model is respectively trained for each individual layer of a sampled set of layers from the plurality of layers using a training set pertaining to the individual layer and one or more preceding layers thereof.

18 . The computerized method according to claim 14 , wherein the yield related prediction with respect to the given structure represents estimated functional defectivity of the given structure following a processing step manufacturing the given layer.

Assignees

• Applied Materials Israel Ltd

Inventors

• Levant Boris
• Tal Noam
• Yacoby Ran
• Choona Lilach
• Pres Shaul
• Linshiz Jasmin Sonia
• Yogev Shay
• Ariel Assaf

Classifications

• H10P72/0616 Primary

  Monitoring of warpages, curvatures, damages, defects or the like · CPC title

• H10P74/23

  characterised by multiple measurements, corrections, marking or sorting processes · CPC title

• H10P74/203

  Structural properties, e.g. testing or measuring thicknesses, line widths, warpage, bond strengths or physical defects · CPC title

• H10P72/0604

  Process monitoring, e.g. flow or thickness monitoring · CPC title

• H10P74/20 Primary

  characterised by the properties tested or measured, e.g. structural or electrical properties · CPC title