# US 12,007,335 B2 · Automatic optimization of an examination recipe > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 12,007,335 B2 - Document ID: US12007335B2 - Status: Granted - Kind: B2 - Application number: 18/196,655 - Filed: 2023-05-12 - Published: 2023-09-14 - Granted: 2024-06-11 - Earliest claimed: 2020-08-17 - Assignees: Applied Materials Israel Ltd. - Applicants: Applied Materials Israel Ltd. - Inventors: Amir Bar - Web page: https://patentgrep.com/patent/US12007335B2 ## CPC - G01N 21/9501: {Semiconductor wafers (manufacturing processes per se of semiconductor devices implementing a measuring step H10P74/20)} - G06T 7/001: {using an image reference approach} - G03F 7/7065: {Defects, e.g. optical inspection of patterned layer for defects} - G03F 7/706841: {Machine learning} - G06T 7/0004: {Industrial image inspection} - G06T 2207/10061: from scanning electron microscope - G06T 2207/20081: Training; Learning - G06T 2207/20084: Artificial neural networks [ANN] - G06T 2207/20224: Image subtraction - G06T 2207/30148: Semiconductor; IC; Wafer - G01N: INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES (measuring or testing apparatus or processes other than immunoassay, involving enzymes or microorganisms C12M, C12Q) - G06T: IMAGE DATA PROCESSING OR GENERATION, IN GENERAL - G03F: PHOTOMECHANICAL PRODUCTION OF TEXTURED OR PATTERNED SURFACES, e.g. FOR PRINTING, FOR PROCESSING OF SEMICONDUCTOR DEVICES; MATERIALS THEREFOR; ORIGINALS THEREFOR; APPARATUS SPECIALLY ADAPTED THEREFOR; (phototypographic composing devices B41B; photosensitive materials or processes for photographic purposes G03C; electrophotography, sensitive layers or processes therefor G03G) ## Abstract A method of automatic optimization of an examination recipe includes obtaining inspection data of a given layer of a semiconductor specimen acquired by an inspection tool during runtime examination, the inspection data including inspection images representative of defect candidates from a defect map of the given layer, extracting inspection features characterizing the inspection images, and using a classifier to classify the defect candidates based on the inspection features, giving rise to a list of defect candidates having a higher probability of being defects of interest (DOIs). The semiconductor specimen includes multiple layers, and the classifier is a general-purpose classifier (GPC) usable for runtime classification of inspection data from any layer of the multiple layers of the semiconductor specimen, the GPC being previously trained using training data including inspection features characterizing training inspection images of various types of DOIs and nuisances collected from the multiple layers and label data associated therewith. ## Claims ### Claim 1 A computerized system of examining a semiconductor specimen, the computerized system comprising a processing and memory circuitry (PMC) configured to: obtain inspection data of a given layer of a semiconductor specimen acquired by an inspection tool during runtime examination, the inspection data comprising a plurality of inspection images representative of a plurality of defect candidates from a defect map of the given layer; extract a first set of inspection features characterizing the plurality of inspection images; and use a classifier to classify the plurality of defect candidates based on the first set of inspection features, giving rise to a list of defect candidates having a higher probability of being defects of interest (DOIs), wherein the semiconductor specimen comprises multiple layers, and the classifier is a general-purpose classifier (GPC) usable for runtime classification of inspection data from any layer of the multiple layers of the semiconductor specimen, the GPC being previously trained using training data comprising a second set of inspection features characterizing training inspection images of various types of DOIs and nuisances collected from the multiple layers and label data associated therewith. ### Claim 2 (depends on 1) The computerized system according to claim 1, wherein the GPC is usable for performing classification for inspection data from one or more new layers without being re-trained using training data pertaining to the one or more new layers. ### Claim 3 (depends on 1) The computerized system according to claim 1, wherein the PMC is configured to, for a new layer, use the GPC to perform classification for inspection data of the new layer, and determine, based on performance of the classification, whether to train a layer-specific classifier so as to improve the performance of the classification. ### Claim 4 (depends on 3) The computerized system according to claim 3, wherein the performance can be measured based on at least one performance measure parameter of: false alarm rate (FAR), capture rate (CR), filter rate, accuracy, purity, and rejection rate. ### Claim 5 (depends on 1) The computerized system according to claim 1, wherein the second set of inspection features comprises a set of features extracted from the multiple layers by an unsupervised model, the unsupervised model being previously trained using a set of training inspection images for extracting representative features thereof. ### Claim 6 (depends on 1) The computerized system according to claim 1, wherein the second set of inspection features comprises a set of features extracted from the multiple layers by a supervised model, the supervised model being previously trained using one or more training inspection images with respectively associated label data for determining types of defect candidates thereof. ### Claim 7 (depends on 1) The computerized system according to claim 1, wherein the training data further comprises a set of features comprising at least one of: one or more tool features, one or more defect features and one or more specimen features, in addition to the inspection features. ### Claim 8 (depends on 1) The computerized system according to claim 1, wherein the PMC is configured to extract the first set of inspection features using at least one of an unsupervised model or a supervised model. ### Claim 9 (depends on 1) The computerized system according to claim 1, wherein the GPC is further configured to classify the list of defect candidates into different types of DOIs. ### Claim 10 (depends on 1) The computerized system according to claim 1, wherein the PMC is further configured to provide the list of defect candidates to a review tool, and wherein the review tool is configured to capture review images at locations of the list of defect candidates, and review the review images for ascertaining whether a defect candidate in the list is a DOI or a nuisance. ### Claim 11 A computerized method of examining a semiconductor specimen, the method comprising: obtaining inspection data of a given layer of a semiconductor specimen acquired by an inspection tool during runtime examination, the inspection data comprising a plurality of inspection images representative of a plurality of defect candidates from a defect map of the given layer; extracting a first set of inspection features characterizing the plurality of inspection images; and using a classifier to classify the plurality of defect candidates based on the first set of inspection features, giving rise to a list of defect candidates having a higher probability of being defects of interest (DOIs), wherein the semiconductor specimen comprises multiple layers, and the classifier is a general-purpose classifier (GPC) usable for runtime classification of inspection data from any layer of the multiple layers of the semiconductor specimen, the GPC being previously trained using training data comprising a second set of inspection features characterizing training inspection images of various types of DOIs and nuisances collected from the multiple layers and label data associated therewith. ### Claim 12 (depends on 11) The computerized method according to claim 11, wherein the GPC is usable for performing classification for inspection data from one or more new layers without being re-trained using training data pertaining to the one or more new layers. ### Claim 13 (depends on 11) The computerized method according to claim 11, further comprising, for a new layer, using the GPC to perform classification for inspection data of the new layer, and determining, based on performance of the classification, whether to train a layer-specific classifier so as to improve the performance of the classification. ### Claim 14 (depends on 13) The computerized method according to claim 13, wherein the performance can be measured based on at least one performance measure parameter of: false alarm rate (FAR), capture rate (CR), filter rate, accuracy, purity, and rejection rate. ### Claim 15 (depends on 11) The computerized method according to claim 11, wherein the second set of inspection features comprises a set of features extracted from the multiple layers by an unsupervised model, the unsupervised model being previously trained using a set of training inspection images for extracting representative features thereof. ### Claim 16 (depends on 11) The computerized method according to claim 11, wherein the second set of inspection features comprises a set of features extracted from the multiple layers by a supervised model, the supervised model being previously trained using one or more training inspection images with respectively associated label data for determining types of defect candidates thereof. ### Claim 17 (depends on 11) The computerized method according to claim 11, wherein the training data further comprises a set of features comprising at least one of: one or more tool features, one or more defect features and one or more specimen features, in addition to the inspection features. ### Claim 18 (depends on 11) The computerized method according to claim 11, wherein the inspection features characterizing the plurality of inspection images are extracted using at least one of an unsupervised model or a supervised model. ### Claim 19 (depends on 11) The computerized method according to claim 11, wherein the GPC is further configured to classify the list of defect candidates into different types of DOIs. ### Claim 20 A non-transitory computer readable storage medium tangibly embodying a program of instructions that, when executed by a computer, cause the computer to perform a method of examining a semiconductor specimen, the method comprising: obtaining inspection data of a given layer of a semiconductor specimen acquired by an inspection tool during runtime examination, the inspection data comprising a plurality of inspection images representative of a plurality of defect candidates from a defect map of the given layer; extracting a first set of inspection features characterizing the plurality of inspection images; and using a classifier to classify the plurality of defect candidates based on the first set of inspection features, giving rise to a list of defect candidates having a higher probability of being defects of interest (DOIs), wherein the semiconductor specimen comprises multiple layers, and the classifier is a general-purpose classifier (GPC) usable for runtime classification of inspection data from any layer of the multiple layers of the semiconductor specimen, the GPC being previously trained using training data comprising a second set of inspection features characterizing training inspection images of various types of DOIs and nuisances collected from the multiple layers and label data associated therewith.